Tobacco Use and
The Risk of Cardiovascular Diseases
In Developed and Developing Countries
Myriam Alexander
Queen’s College
University of Cambridge, UK
2008-2012
This dissertation is submitted for the Degree of Doctor in Philosophy
NETSIM Bloodomics Department of Public Health
and Primary Care

IThis thesis is dedicated
to my parents
who have taught me to value intellectual knowledge
to my husband
who has supported me daily
and to my two wonderful children
who were born during the course of this PhD
and have taught me that science is like children
and requires patience, endless learning, unconditional love, formidable energy and a
willingness to approach the future positively and with confidence.
II
SUMMARY
Background & objective: The association between cigarette smoking and the risk of
cardiovascular diseases (CVD) is well established. However, the effect of other, less common,
types of smoking on CVD risk, such as pipes and cigars in developed countries, remains
uncertain. By contrast, in developing countries, a large panel of smokeless tobacco products
are consumed alongside smoking products, with unknown effects on the risk of CVD. The aim
of this thesis is to investigate the association between various forms of tobacco use with the
risk of CVD in the setting of developed countries and of a developing country with a large
population, Pakistan.
Data Sources: Firstly, for the investigation of cigarette, pipe and cigar smoking, the analysis
was based on the Emerging Risk Factor Collaboration database. It included, in April 2011, up to
929,335 individuals with baseline information on smoking status from 135 prospective cohort
studies in developed countries, who experienced 40,218 incident coronary artery disease
events during an average of 14.2 years. Secondly, for the investigation of the association
between chewing and dipping forms of tobacco and the risk of CVD, the analysis was based on
the Pakistani Risk of Myocardial Infarction Study, which had recruited, by May 2012, a total of
7,905 first ever myocardial infarction (MI) cases and 7,458 age and sex frequency matched
controls.
Results: All forms of tobacco use were significantly associated with excess risk of CHD and
CVD. Current cigarette smoking was most strongly associated and produced a hazard ratio of
1.99 (95 % confidence interval: 1.86; 2.13) of MI and 1.64 (1.54; 1.75) of cerebrovascular
events, compared to never-smokers, in developed countries. By contrast, compared to never-
smokers, the risk of MI was 1.35 (1.20; 1.51) for current cigar use and 1.84 (1.69; 2.01) for
current pipe use in developed countries. For smokeless tobacco, which was investigated in the
South Asian context, odds ratios of non-fatal MI compared to never tobacco users were 1.71
(1.46; 2.00) when currently chewing paan, supari or gutka products, and 1.46 (1.21; 1.78) when
currently dipping South Asian snuff called naswar. The risks associated with current tobacco
use, and in particular, current smoking, were significant even at low intensities (<5 products a
day) in both developed and developing countries. Quitting smoking or stopping the use of
smokeless products was associated with a significantly lower excess risk of CVD worldwide.
Few individuals in developing countries are ex-tobacco users. The risk of past smokers was
investigated in a developed population and was shown to become not significantly different
from that of never-smokers 20 years after stopping.
Conclusion: All forms of tobacco use increased CVD risk, and the highest risk was found
amongst cigarette users. Neither pipe or cigar smoking nor smokeless tobacco constituted safe
alternatives to cigarette smoking and current users carried a higher CHD risk than ex- and
never-users. Rather than advocating alternatives to cigarette smoking on the basis that they are
less harmful for CVD risk, this thesis emphasizes the need in both developed and developing
countries to adopt consistent policies restricting the use of all types of tobacco use, to prevent
the very high burden of tobacco related CVD worldwide.
III
CONTENTS
Page
Acknowledgments 1
List of abbreviations 3
Chapter 1: Introduction 4
Section A: Smoking and the risk of CVD in developed countries
Chapter 2 Description of the Emerging Risk Factors Collaboration 54
Chapter 3 Distribution and cross-sectional correlates of smoking in
developed countries
78
Chapter 4 Cigarette smoking and the risk of CVD, lung cancer deaths
and all causes mortality in developed countries
105
Chapter 5 Pipe or cigar smoking and the risk of CVD and lung cancer
deaths in developed countries
146
Section B: Tobacco use and the risk of CVD in a developing country
Chapter 6 Description of the Pakistan Risk of Myocardial Infarction Study 175
Chapter 7 Correlates of tobacco use in a developing country 188
Chapter 8 Smoking and smokeless tobacco use and the risk of MI in
Pakistan
218
Chapter 9: Discussion 244
APPENDICES
Appendix 1 List of publications authored during PhD 260
Appendix 2 Socio-economic status and the risk of MI in Pakistan 261
Appendix 3 Diet and the risk of MI in Pakistan 273
Appendix 4 The 9p21.3 locus and risk of MI in Pakistan 290
Note: Relevant tables, figures and references are provided at the end of each chapter. See p.3
for abbreviations.
IV
PREFACE
The aim of this thesis was to investigate, in more detail than has been done before, the
association between different types of tobacco use and the risk of cardiovascular
diseases in developed and developing countries. The work is presented, following an
Introduction, in two Sections looking at the effect of different types of tobacco use:
firstly in developed countries and secondly in a developing country, taking the example
of Pakistan. Section A contains four Chapters and Section B contains three
Chapters. At the end, a Discussion reviews the main findings and considers future
work.
During my doctoral studies, I have also conducted research on other topics relevant to
cardiovascular diseases, including blood pressure, the chromosome 9p21.3, diet and
socio-economic status, using the Pakistani Risk of Myocardial Infarction Study dataset.
In addition, I have been involved in side projects regarding tobacco use, including a
project researching the relationship between access to tobacco outlet and smoking
abstinence in the UK. Brief description of these projects and list of publications that
arose from these works are presented in the Appendices.
This dissertation is the result of my own work except where collaborations are
specifically acknowledged in the Acknowledgments section.
1ACKNOWLEDGEMENTS
Although this thesis is the result of my own work, it would not have been possible without the
precious help of several people. I would like to acknowledge my funders, NETSIM Bloodomics,
for offering me an EU funded Marie Curie scholarship, and Dr Willem Ouwehand for
supervising and providing and inspirational leadership to all Bloodomics students. I am
indebted toward Professor John Danesh for welcoming me into the Cardiovascular
Epidemiology Unit initially as an intern, then as an MPhil student, and presently as a PhD
student. On a day to day basis, I am grateful to my two co-supervisors, Dr Emanuele Di
Angelantonio and Dr Danish Saleheen, who have guided me and encouraged me during my
time as a PhD student and especially Dr Emanuele Di Angelantonio for commenting on this
thesis. I was privileged to work on datasets which had already been collected, and am grateful
to all the people involved in setting up the Emerging Risk Factors Collaboration (ERFC) and the
Pakistan Risk of Myocardial Infarction Study (RPOMIS). I would like to thank all the members of
the Cardiovascular Epidemiology Unit for offering their support when needed and providing a
friendly environment. My gratitude goes in particular to fellow statisticians Dr Stephen Kaptoge
and Dr Angela Wood, Dr David Wormser, Dr Rao Kondapally Seshasai, Dr Emma Heydon, Dr
Sebhat Erqou, to my fellow PhD students, to the data management team and to the
administrative team. I detail below the contribution of colleagues to the analyses presented in
this thesis.
Introduction
I performed reviews of the published epidemiological literature using search engines such as
PubMed Google Scholar and Web of Science. I organized the Introduction into several
paragraphs, and synthesized the evidence regarding each aspect of the relationship between
tobacco use and CVD risk. Dr E. Heydon and Dr D. Wormser commented helpfully.
Section A:
ERFC was a project which was started before my arrival at the Cardiovascular Epidemiology
Unit and was initiated by Professor J. Danesh with the help of others (Dr N. Sarwar, Dr Di
Angelantonio, Dr A. Thompson, Dr S. Thompson, Dr I. White, Dr M. Walker and S. Watson). A
list of collaborators who contributed individual data is available at www.phpc.cam.ac.uk. I am a
member of the Coordinating Centre of ERFC. Actual collation and cleaning of data was done by
the data management team including Dr M. Walker and S. Watson. Statistical methods were
developed by a team of statisticians who were members of the Coordinating Centre of ERFC to
whom I am part, and were published in the Am J Epidemiol in 2007. I conducted all the
statistical analyses presented in Chapters 2-5, except for the estimation of years of life lost as
a result of smoking in the last paragraph of Results of Chapter 4 which was done by Dr S
Kaptoge. I wrote appropriate programming code in STATA language, calling programs
developed by Dr S Kaptoge and Dr P Perry specifically for the analysis of ERFC dataset. I
produced relevant tables and figures and drafted the text, being inspired by previous reports
2published by ERFC (Lancet 2011 March, N Engl J Med. 2011 March, BMJ 2011 Feb, Lancet
2010, JAMA 2009). Dr S. Thompson, Dr A. Wood, Dr S. Kaptoge, P. Gao and Dr D. Wormser
provided statistical advice. Dr E. Di Angelantonio was my daily supervisor for the analyses
presented in these Chapters. Dr D. Wormser, Dr E. Heydon and S. Warnakula helpfully
commented on these Chapters.
Section B:
The second dataset I worked on is PROMIS. PROMIS’ principal Investigator is Dr. D. Saleheen
and he is supported in Cambridge by Professor J. Danesh, and in Pakistan by a large group
dedicated to the enrolment of cases and controls (http://www.cncdpk.com/projects/the-pakistan-
risk-of-myocardial-infarction-study-promis.html). Data entry was performed in 6 hospitals
located in urban centres across Pakistan. Checks and cleaning of this dataset were done by
data manager Dr M. Walker and by myself. I conducted all the statistical analyses, wrote
appropriate programming code in STATA language, produced relevant tables and figures and
drafted the text. Developing programming code for these analyses was a time consuming task
and I shared my codes with other members of the Cardiovascular Epidemiology Unit who were
involved in the analysis of PROMIS and other datasets informally and at seminars. Dr. D.
Saleheen was my main supervisor for these analyses, seconded by Dr E. Di Angelantonio. Dr
D. Wormser, Dr E. Heydon and W Kee Ho commented helpfully on these Chapters.
Appendices 2-3:
Analyses were conducted at the same time as for Section B. Dr L. Johnson and Dr N. Naswar
provided helpful comment on the dietary analyses.
Appendix 4:
The questionnaire data was provided by Dr. D. Saleheen, after cleaning done by Dr M. Walker
and myself. The genetic data was generated at the Wellcome Trust Sanger Institute
Cambridge, England. Analysis plan was drafted by Dr. D. Saleheen and by me, with helpful
comments from Dr E. Di Angelantonio. I produced all tables and figures. The manuscript was
written by me and Dr D. Saleheen, with corrections from Dr E. Di Angelantonio and Professor J.
Danesh. Other co-authors and their contributions are detailed in the body of the manuscript.
3ABBREVIATIONS
BMI: Body mass index
CAD: Coronary artery disease
CHD: Coronary heart disease
CI: Confidence interval
CO: Carbon monoxide
CPD: Cigarettes per day
CPS: Cancer Prevention Study
CVD: Cardiovascular diseases
DBP: Diastolic blood pressure
DALYs: Disability adjusted life years
ERFC: Emerging Risk Factors Collaboration
EU: European Union
FAV: Floating absolute variance
FCTC: Framework Convention on Tobacco Control
GWA: Genome wide association study
HDL-C: High density lipoprotein cholesterol
HR: Hazard ratios
LDL-C: Low-density lipoprotein cholesterol
MI: Myocardial infarction
OR: Odds ratio
PROMIS: Pakistan Risk of Myocardial Infarction Study
RR: Relative risks
RYO: “Roll your own” cigarettes
SBP: Systolic blood pressure
SNP: Single nucleotide polymorphism
UK: United Kingdom
UN: United Nations
USA: United States of America
WHO: World Health Organization
WHR: Waist to hip ratio
4Chapter 1: INTRODUCTION
Summary
Cardiovascular disease (CVD) is the leading cause of death worldwide, responsible
for over 17 million deaths globally in 2008. Approximately eight out of ten of these
deaths occur in developing countries, which are currently experiencing an epidemic
of non-communicable diseases. A rise in tobacco use concomitant with a reduction in
infectious disease is thought to contribute to this epidemic, and tobacco has been
estimated to directly cause 10% of all CVD worldwide. There are currently more than
1.1 billion smokers worldwide, and while developed countries remain at the top in
terms of prevalence, with around one fifth of the total adult population smoking, nine
out of ten smokers now reside in developing countries. This is due to especially high
and rising prevalence in men in developing countries, with men in lower middle
income countries averaging 40% smoking prevalence, whilst prevalence in
developed countries has been approximately halved in men and women over the
past 50 years. Alternative forms of tobacco use are increasingly being used: cigar
smoking in developed countries and smokeless tobacco in developing countries. The
epidemiological evidence on cigarette smoking in relation to CVD is compelling, but
the relationships between CVD and other smoking types including pipe, cigars, and
smokeless types of tobacco, remain uncertain. Biological mechanisms by which
tobacco causes CVD are partially understood and are thought to involve nicotine,
oxidant chemical, and carbon monoxide. The objective of this thesis is to address
remaining epidemiological uncertainties on tobacco in relation to CVD risk using the
Emerging Risk Factors Collaboration and the Pakistan Risk of Myocardial Infarction
Study.
51.1 Introduction
The aim of this Chapter is to define several types of tobacco used globally and
present the current prevalence of tobacco use. In addition, CVD is defined and the
potential biological mechanisms of the association between CVD and tobacco use
are also described. Finally, a review of the literature on the epidemiological evidence
of causal associations between tobacco use and CVD is presented, which motivates
the overall objectives of this thesis.
1.2 Definition of tobacco uses
The WHO Framework Convention on Tobacco Control defines tobacco products as
“products entirely or partly made of the leaf tobacco as raw material which are
manufactured to be used for smoking, sucking, chewing or snuffing” 1. The main form
of tobacco use in developed countries is smoking, while sucking, chewing and
snuffing are common in developing countries. There is a wide variety of tobacco
products and here I give examples of products found in Pakistan and more generally
South Asia.
1.2.1 Smoking tobacco
1.2.1.1 Cigarettes: Cigarettes are the most popular form of tobacco use worldwide.
The modern cigarette evolved from a 16th century variant of the cigar 2 and consists
of a roll of tobacco wrapped in paper or a substance not containing tobacco 3.
However, tobacco companies have been inclined to blur the difference between
cigars and cigarettes to avoid the heavy taxes imposed on cigarettes, and as a result
the US code of federal regulation has extended the definition of cigarettes to “any roll
of tobacco wrapped in any substance containing tobacco which, because of its
appearance, the type of tobacco used in the filler or its packaging and labelling, is
likely to be offered to or purchased by consumers as cigarettes [rather than a cigar]”
4. The tobacco contained in cigarettes is mainly heat- or air-cured, and in a small
percentage sun-cured. Air-curing involves hanging the whole tobacco plant or primed
tobacco leaves in barns for 30 to 40 days, while heat-curing involves hanging the
leaves of tobacco on tiers in barns where the air is gradually warmed to a
temperature of 70°C to 75°C over a period of 5 to 7 days. These curing processes
make the smoke acidic and its nicotinic content easier to inhale, promoting nicotine
addiction 5. Conventionally, cigarette smoke is divided into two phases: a tar phase
and a gas phase 6. The tar or particulate phase is defined as the material that is
6trapped when the smoke stream is passed through the Cambridge glass-fibre filter
that retains 99.9% of all particulate material with a size >0.1 μm, while the gas phase
is the material that passes through the filter. Cigarette smoke comprises
approximately 90% of gaseous components and 10% of tar.
1.2.1.2 Roll-your-own (RYO): Roll-your-own cigarettes are handmade cigarettes
made by wrapping paper around tobacco. RYO are very popular in developing
countries for cultural reasons and increasingly used in developed countries where
they are sometimes used as a cheaper replacement to heavily taxed industrial
cigarettes2. Bidies (also written beedies or bidis) are a form of RYO cigarette popular
in Pakistan and other South-East Asian countries (Figure 1.1). They are small hand-
rolled cigarettes which require more frequent puffing and pulmonary effort than
cigarettes due to their non-porous wrapper of temburini leaf. Bidies usually contain a
smaller amount of tobacco than cigarettes (0.15g to 0.25g versus 1g) and produce a
smaller volume of smoke; but more frequent puffing means they produce up to 3
times more carbon monoxide and nicotine, and 2 to 3 times more tar than regular
cigarettes 7.
1.2.1.3 Cigars: Cigars are made of a roll of tobacco wrapped in a leaf of tobacco or in
any substance which contains tobacco 4. The main difference between cigar and
cigarette resides in the processing of tobacco, which is air cured and fermented in
cigars, while it is not fermented in cigarettes 3. Fermentation entails packing the
tobacco leaves and placing them in fermentation rooms for 3 to 5 weeks. They are
subsequently removed, repacked, and returned to the fermentation rooms several
times to achieve the desired flavour and aroma. The fermentation process is
responsible for making cigar tobacco richer than cigarettes in nitrate, carcinogens
formed from nornicotine and nicotine, and nitrogen oxides. Cigars also have a higher
pH than cigarettes, which increases the amount of free nicotine in the gas and solid
phases of the smoke. As a result, the smoke of cigars dissolves more easily in the
saliva than the smoke of cigarettes. The desired dose of nicotine is achieved without
the need to inhale the smoke into the lung and high levels of dependency can be
created even if the smoke is not inhaled.
1.2.1.4 Pipes: Pipes are a device consisting of a tube of wood, clay or other material
with a small bowl at one end which is filled with tobacco. Tobacco for smoking in
pipes is often carefully treated and blended to achieve flavour nuances not available
7in other tobacco products. Similar to cigars, the smoke produced by pipes tends to
have a higher pH than cigarette smoke and thus does not need to be directly inhaled,
or even lit up, to sustain high levels of nicotine addiction. Owing to the relatively large
quantities of tobacco put into pipes, pipe smokers are generally exposed to smoke
equivalent to that from several cigarettes 2. Chilum is a straight conical pipe, usually
made of clay, in use in South and Southeast Asia for smoking tobacco, cannabis and
opiates, and requiring deep pulmonary effort. Chilum is entirely filled with tobacco
and is held vertically for smoking. To prevent the tobacco from entering the mouth, a
pebble or stopper is interested into its top, usually made of a wet piece of cloth which
protects the mouth from the heat and serves as a filter 8.
1.2.1.5 Water pipes: called shisha or narghile in Middle Eastern countries and
Hookah in India and South Asia. Water pipes consist of a receptacle for water, with
an opening on the top, to which a long wooden stem is fixed. At the top of this stem,
a small bowl is attached for tobacco 9. The smoke of tobacco is made to pass
through the water before being inhaled, which cools and filters it. In hookahs, around
20g of tobacco is burned over smouldering charcoal at each sitting, 20 times the
amount contained in a cigarette. The water filter is not effective in removing tar,
nicotine or carbon monoxide. Whereas a cigarette is typically smoked over
approximately five minutes with 300-500 ml of smoke inhaled, hookah smoking
sessions last from 20 to 60 minutes with volumes of 10 litres or more of smoke
inhaled 2.
1.2.2 Smokeless tobacco
Smokeless tobacco is used to describe tobacco that is consumed without burning it 8.
Oral use of smokeless tobacco has existed for thousands of years in South America
and has gained popularity in other areas of the world after the discovery of the new
continent by Christopher Columbus at the end of the 15th century. Smokeless
tobacco is used in two ways. One way is to place the preparations in various parts of
the mouth and suck it, which is called “dipping tobacco”. Another way is to chew the
preparations placed in the mouth which is called “chewing tobacco”.
1.2.2.1 Chewing tobacco: Tobacco is shredded like short cut grass, generally mildly
acidic and intended to be chewed throughout the day as desired. In Pakistan and
other parts of South Asia the most popular chewing product is betel quid. There are
several ways of preparing these products but the main ingredients are betel, areca
8and tobacco. Paan means “leaves of the betel vine” and is commonly used to refer to
a chewing mixture wrapped in betel leaves. Most paan fillings are made of areca
nuts, slaked lime paste and flavouring agents such as menthol, camphor, sugar,
rosewater, aniseed, mint etc (paan supari, paan masala, sada paan); but tobacco
can also be added to the filling (tambako paan). In tambako paan, tobacco is sun-
dried, roasted, powdered and flavoured; or alternatively boiled, then made into a
paste and scented with rosewater or perfume. Paan is placed in the mouth and
gently sucked and chewed as a palate cleanser, a breath freshener and for digestive
purposes, usually at the end of a meal. The speed of nicotine absorption, and hence
the strength of the nicotine effect, increases with the pH of the mixture which is
raised by the slaked lime 2. Supari means “nut” in Indic language and indicates all
mixtures from areca nuts, including paan. The areca nut can be either broken into
pieces and chewed by itself; mixed with slaked lime, flavourings and optionally
tobacco as paan; or mixed with sugar, spices, flavourings such as menthol and
optionally tobacco 9. Gutka (or gutkha) is in use in India, Pakistan, South-East Asia,
and in the UK. It is a dry mixture of tobacco, betel nut and catechu, which are mixed
together with slaked lime, flavouring (menthol, saffron) and sweetening spices (e.g.:
cardamom, clove), and is held in the mouth to be chewed. Saliva is generally spit
out, but sometimes swallowed. Gutka was introduced half a century ago as a
manufactured and cheaper substitute of hand-prepared products such as paan and
supari. Compared to these products, gutkha is industrially prepared and has a longer
shelf life. Its growth has been so rapid that it has overtaken the smoking of tobacco in
India, being especially popular amongst younger generations and women9.
1.2.2.2 Snuff dipping refers to tobacco chopped into particles like large coffee
grounds, moistened and used by holding between gum and cheek, rather than
chewed. Naswar is also called nass or niswar. It is used in Pakistan, Central Asia,
Iran, Afghanistan, Baluchistan and India 8. Naswar is made with fresh tobacco leaves
dried in the sun or hot room, slaked lime, ash from a tree bark, flavouring
(cardamom, oil, menthol) and colouring agents (indigo) mixed together. Water is
added and the mixture is rolled into balls. It is held in the mouth for 10 to 15 minutes
and is sometimes chewed slowly 9.
91.3 Definition of cardiovascular diseases
The term cardiovascular disease encompasses all disorders of the heart, blood
vessels and blood circulation 10. Development of the disease might be clinically silent
for years until stenosis impairs the function of the heart or another organ and CVD
becomes symptomatic. Coronary heart disease (CHD), also called ischaemic heart
diseases and coronary artery disease, is the most common form of CVD. CHD
manifests as myocardial infarction (MI) when atherosclerosis leads to narrowing and
obstruction of the coronary arteries by formation of a thrombus or embolus, resulting
in interruption of the blood supply to the heart and damage of the heart muscle 10.
Cerebrovascular diseases are another common form of CVD caused by either
atherosclerosis developing in an artery directly supplying blood to the brain or emboli
from a distant artery causing obstruction to the blood flow directed to the brain 10.
Cerebrovascular outcomes can be further classified as ischaemic stroke,
haemorrhagic stroke and subarachnoid haemorrhage. Stroke refers to the damage to
part of the brain caused by an interruption of its blood supply which can be due to a
blockage of cerebral artery by a blood clot (ischaemic stroke) and rupture of a blood
vessel in or near the brain (haemorrhagic stroke). Subarachnoid haemorrhage is a
type of brain haemorrhage in which a blood vessel ruptures into the cerebrospinal
fluid that surrounds the brain and spinal cords caused by burst aneurysm.
Other types of CVD include coronary aortic aneurysm, pulmonary embolism, heart
failure, cardiac dysrhythmia and peripheral vascular diseases 10. Aortic aneurysm
refers to an abnormal dilatation of an artery located in the aorta caused by the
pressure of the blood flowing through a weakened area as a result of atherosclerosis
or MI. Pulmonary embolism occurs when a pulmonary artery or one of its branches
becomes obstructed by an embolus, usually after a deep vein thrombosis. Heart
failure means an inability of the heart to cope with its workload or pumping blood to
the lungs and to the rest of the body. Cardiac dysrhythmia refers to a disturbance of
heart rhythm caused by a problem with the electrical impulses in the heart. Peripheral
vascular diseases refer to a narrowing of blood vessels in the legs or arms, causing
blood flow and pain and, in severe cases, development of gangrene.
CVD are complex diseases mediated through multiple risk factors and pathways
including tobacco use, dyslipidaemia, hypertension, inflammation, and insulin
10
resistance. Atherosclerosis is used to mean the development of fatty plaques on the
artery linings, which begins in childhood, and develops usually by middle age into
plaques several centimetres across the arterial intima, narrowing the arteries and
impairing blood flow 11. Lipoprotein particles bound to arterial intima have increased
susceptibility to oxidative and other chemical modifications 12. Early after initiation of
hypercholesterolemia, leucocytes adhere to the endothelium and enter the intima,
where they begin to accumulate lipids and become foam cells. Once macrophages?
have taken up residence in the intima and become foam cells, they replicate and
serve as a reservoir for excess lipid. Macrophages also provide a source of pro-
inflammatory mediators. This is turn promotes the progression of lesions. Whereas
the early events in atheroma initiation involve primarily altered endothelial function
and recruitment of leukocytes, the subsequent evolution of atheroma involves
smooth muscle cells as well. These smooth muscle cells multiply by cell division and
also sometimes die, both mechanisms contributing to the complication of the
atherosclerotic process. Another mechanism of plaque progression is thought to
involve angiogenesis in plaques. Micro-vessels within plaques form in response to
angiogenic peptides over-expressed in atheroma. The micro-vessels are known to be
friable and prone to rupture 13. A third mechanism is through development of areas of
calcification in plaques and mineralization. In most cases, MI is caused by a fracture
of the fibrous cap of the plaque 12. Mechanisms which lead to the rupture of the
fibrous cap include reduced collagen synthesis by smooth muscle cells, increased
catabolism of the extra-cellular matrix macromolecules, deaths of smooth muscle
cells and the presence of a large lipid pool.
1.4 Biological mechanisms by which tobacco use leads to CVD
Cigarette smoking has been shown to promote inflammation of the vessel wall,
reduce oxygen availability and activate the sympathetic nervous system, all factors
contributing to an imbalance between supply and demand of myocardial blood,
oxygen and nutrients, and ultimately leading to clinical outcomes 14.
Cigarette smoking is associated with a more atherogenic lipid profile; including lower
HDL-C and higher LDL-C, apolipoprotein A1, VLDL-C and triglycerides levels
compared to non-smokers 15. Smokers also have higher levels of oxidized LDL,which
are taken up by macrophages to become foam cells that are an integral part of the
atherosclerotic plaque 16 (Figure 1.2).
11
Inflammation is believed to contribute to atherogenesis; and white blood cell counts,
C-reactive protein, and fibrinogen are predictors of future cardiovascular events 17.
Cigarette smoking results in a chronic inflammatory state, and smokers have been
shown to have higher levels of reactive oxygen species, circulating leukocytes, C-
reactive protein, and acute phase reactants such as fibrinogen 18 (Figure 1.2).
Cigarette smoking enhances the recruitment and adhesion of leukocytes to blood
vessel walls 19 and activates monocytes 20, which cause damage to the vessel walls.
Smoking also induces a hyper-coagulable state in individuals with relatively low
levels of atherosclerosis, such as young smokers 21. In men who experienced a
sudden death, pathological findings of acute thrombosis are more likely to be present
in smokers than in non-smokers 22. Thrombosis can result from platelet activation,
itself a consequence of endothelial dysfunction. Smoking may also affect the
thrombogenecity of atherosclerotic plaques, through higher levels in tissue factor
which is known to contribute to thrombosis after plaque disruption, higher levels of
vascular cell adhesion molecule 1 and greater numbers of macrophages in
atherosclerotic plaques 23.
Smoking not only acts as a major risk factor for CVD, but may also serve to induce or
aggravate type 2 diabetes, therefore enhancing another major risk factor 24. Smokers
have been shown to have higher levels of free fatty acids and triglycerides after
meals, which has been associated with insulin resistance.
Tobacco contains nicotine and 4,000 other compounds which are partially transferred
into the smoke when a cigarette, cigar or pipe is lit up 25. In pipes, cigars and bidies,
tobacco is easily absorbed across the oral mucosa while the paper wrapping up
tobacco in cigarettes reduces its absorption, necessitating inhalation of the smoke
into larger surfaces of the lungs in order to satisfy the smoker’s addiction 14. The tar
phase of the smoke, referring to particles of the smoke which can be measured by
machine tests methods, contains a large number of toxic constituents and
carcinogenic compounds. The vapour phase of the smoke mainly contains carbon
monoxide produced by the combustion. Out of all the known components of cigarette
smoke, only a few have been examined in isolation 6 (Figure 1.3).
12
1.4.1 Effects of nicotine
Nicotine has been shown to increase heart rate, and to produce endothelial
dysfunction, lipid abnormalities and insulin resistance in smokers 26. The amount of
nicotine recovered from the smoke taken into the mouth varies from less than 0.3 mg
to 3 mg per cigarette, depending on the degree of inhalation. Nicotine is then
absorbed rapidly, with peak arterial blood levels of 10 to 100 ng/ml after a cigarette,
and eliminated within 6 to 8 hours 27.
Intravenous nicotine, nicotine nasal spray, and nicotine chewing gum have been
shown to increase heart rate and systolic blood pressure 28-30. Nicotine stimulates the
sympathetic nervous system by increasing plasma levels of norepinephrine and
epinephrine, which translates into an acute rise in heart rate (up to 20 beats per
minute) after smoking a cigarette 31 (Figure 1.3). In contrast, nicotine from chewing
tobacco is absorbed slowly and peak arterial levels are much lower than those seen
in cigarette smokers 14 (Figure 1.4).
Nicotine may also cause endothelial cells dysfunction 25, 32, 33. Nicotine in
concentrations similar to those found in the blood of cigarette smokers alters the
structural and functional characteristics of cultured vascular smooth muscle and
endothelial cells 34. In studies of cultured endothelial cells, nicotine enhances the
release of basic fibroblast growth factor and inhibits the production of transforming
growth factor 1, increases DNA synthesis, mitogenic activity, and endothelial
proliferation 35. Nicotine has also been reported to act on human monocyte-derived
dendritic cells involved in adaptive immunity, which have been detected in the wall of
arteries and in atherosclerotic lesions and are known to stimulate an inflammatory
response 36.
Finally, nicotine has been demonstrated to increase insulin resistance in smokers 37
and long term nicotine users 38 by activating the sympathetic nervous system,
increasing release of corticosteroids and growth hormone.
1.4.2 Effects of carbon monoxide (CO)
CO is a major constituent of cigarette smoke, while users of smokeless tobacco
remain unexposed. CO has been directly associated with CVD risk in a dose-
response manner 22. Non-smokers have average levels of carboxyhemoglobin
13
ranging from 0.5% to 2%, compared to levels of 5% to 10% in smokers 39. CO binds
avidly to haemoglobin, reducing the amount of haemoglobin available to carry
oxygen, and impedes oxygen release by haemoglobin. CO inhalation in people with
CAD has been shown to provoke reduced exercise tolerance and exercise-induced
ventricular dysfunction, including ventricular arrhythmias 40. Long term CO exposure
also results in an elevated red cell mass in smokers, to compensate for hypoxemia,
which contributes to increased blood viscosity and promotes a hyper-coagulable
state in smokers.
1.4.3 Effects of oxidants
Tobacco smoke also delivers a high concentration of oxidizing chemicals to the
smoker 41. These chemicals include oxides of nitrogen and a number of different free-
radicals, found both in the gas and tar phases of the smoke. Oxidant stress
translates into elevated levels of peroxides and decreased levels of traditional
plasma antioxidants such as vitamins A and C in smokers 42. Oxidant stress
provokes a reduction in nitric oxide release amongst smokers 43 and is believed to
contribute to a number of the potential mechanisms of CVD, including inflammation,
endothelial dysfunction, oxidation of LDL-C and platelet activation. Nitric oxide and
prostacyclin are vasodilators and have anti-platelet aggregation effects. Oxidant
chemicals also produce hyper-aggregability of platelets through peroxidation of free
fatty acids.
1.4.4 Effects of carcinogenic chemicals
Some pollutants found in the tar fraction of cigarette smoke called polycyclic
aromatic hydrocarbons, such as benzo(a)pyrene, benzo(a)anthracene 44 and
butadiene 45, have been reported to accelerate atherosclerosis in experimental
animals, at doses below those that produce tumours. A mechanism of atherogenesis
is speculated to be a mutation of smooth muscle or other cells that become the
source of an atherosclerotic plaque.
1.5 Prevalence of tobacco use
Amongst all forms of tobacco use, smoking is the most prevalent globally. In 1995,
more than 1.1 billion people smoked worldwide, with about 82% of smokers residing
in low- and middle-income countries 46. Worldwide, male smoking far exceeds female
smoking, with a smaller gender difference in high-income countries. Smokeless
14
tobacco use is more common in developing countries, but migration and tourism
mean its use is being exported to developed countries, where they are most
fashionable in younger age groups.
1.5.1 Past and current prevalence of tobacco use in developed countries
The use and culture of tobacco is old, dating back to around 1,000 BCE, by natives
of the Americas who smoked tobacco in pipes for medicinal and ceremonial
purposes 47. Christopher Columbus was the first Westerner to discover tobacco in
1492, bringing back leaves and seeds with him to Europe. However, the use of
tobacco did not spread until the mid-16th century, popularized by adventurers, sailors
and travellers such as the French Jean Nicot, who gave his name to the world
nicotine. Smoking was introduced to France, Portugal, Spain and England in the mid-
16th century and in North America at the beginning of the 17th century. At that time,
the main forms of tobacco use were pipe smoking, chewing and snuff. Cigars
became popular at the beginning of the 19th century while cigarettes did not start to
be mass-produced in the USA until the end of the 19th century with the invention of a
machine to replace hand-production 48. With the mass production of cigarettes, the
proportion of smokers increased enormously in developed countries during the first
part of the 20th century. In the meantime, pipes and cigars became less prevalent
and snuffing and chewing became on the verge of extinction. Between 1900 and
1960, the annual number of cigarettes sold in the USA was multiplied by a factor of
80, going from an average of 50 cigarettes per person per year to an average of
3,900 cigarettes per person per year (Figure 1.5). During the same period, the sale
of pipe tobacco was divided by 3 (from 1.6 to 0.6 pounds per person per year), the
number of cigars per person per year was halved, and the consumption of chewing
tobacco and snuff, already low, kept decreasing (from 110 to 60 usages per person
per year) 2 (Figure 1.5).
Cigarette smoking prevalence began to decline in the USA in 1960s, following the
publication of reports on the dangers of tobacco by the UK Royal College of
Physicians and US Surgeon General 7, 47. The public became aware of the dangers
of cigarette smoking and governments started to regulate cigarette sale and
advertisement. In the USA for example, health warnings were imposed in 1965 on
cigarettes packs, broadcast advertising was forbidden in 1971, and smoking was
banned in buses and domestic flights in 1990; the Food and Drug Administration was
created in 1995 with the aim of regulating the sale of tobacco, and its role was
15
reinforced in 2009 7. As a result, the proportion of smokers has been approximately
halved over the past 50 years in developed countries. Between 1960 and 2005, the
proportion of adult male smokers dropped from 52% to 27% in the USA, from 81% to
43% in Japan, and from 61% to 25% in the UK. In women, the drop in prevalence
was less dramatic: from 13% to 12% in Japan, from 42% to 24% in the UK, and from
34% to 19% in the USA 46. At baseline of the British Doctors Study (BDS) initiated
1951 by Doll and Hill, 87% of UK male doctors reported smoking cigarettes and/or
pipes 49 while, in 2000, the proportion of male physicians smoking in the UK was 8%
46. Concomitant with a reduction in prevalence, smoking intensity also declined. In
the USA, in 1965, 56% of smokers were smoking at least 20 cigarettes per day
(CPD), while smokers of more than 20 cigarettes a day represented 23% of smokers
in California and 40% of smokers in the remaining United States in 2007 50.
However, the addictive nature of smoking and the peer pressure existing amongst
teenagers to initiate smoking have meant that roughly 1 in 5 adults still smoke in
developed countries (Figure 1.6). Current smokers comprise 1 in 5 adults in the
USA, 1 in 4 in the UK and 1 in 3 in Japan. In the European Union, the proportion of
people currently smoking varies between 25% and 35% for men and between 15 and
25% for women 51. With a lesser decline in prevalence amongst women compared to
men, the gender gap is narrowing and women are catching up with men. The
European Union has the highest worldwide prevalence of women smoking 46.
Teenage girls are at least as likely and sometimes more likely than teenage boys to
start smoking. In the USA, more than 20% of both boys and girls aged 13-15 years
old were tobacco users in 2000 46. A youth survey in Portugal in 2001-2002 found
that 26% of girls versus 18% of boys smoked at least once a week 52.
While cigarette smoking has experienced a major decrease, other forms of smoking
tobacco remain stable or are on the rise. Roll-your-own (RYO) cigarettes represent a
cheaper option than manufactured cigarettes, and are often used to combine tobacco
with illegal products such as marijuana. Their use is increasing in the USA and
Europe, especially amongst young people 2. Studies have found that RYO users are
heavier smokers, more addicted to nicotine and less likely to consider quitting 53.
Another popular alternative to cigarettes has been cigars, particularly small cigars
called “cigarillos” which are made to look like cigarettes in shapes and format. The
proportion of cigar users has been increasing in the USA since 1990. Since 1995,
initiation of cigar smoking has outnumbered initiation of cigarette smoking in the
16
USA, being particularly appealing to teenagers and women 54. In 2003, current cigar
smokers made up 5.5% of the population in USA and the highest rates were reported
amongst young adults aged 18 to 25, where 11% reported having smoked a cigar in
the past month. In 2007, in industrialized countries from the former Commonwealth
(Australia, Canada, UK and USA), the prevalence of cigar use varied between 3%
and 13%, pipe use between 0.3% and 2.1%, smokeless tobacco was used by less
than 2.3% and RYO by 12% to 21% of the population 55.
1.5.2 Past and current prevalence of tobacco use in developing countries
Tobacco use originated in South America, and spread to other parts of the world,
including Africa and Asia, by European travellers and merchants, during the 16th and
17th century 2. It is during this period that mixing tobacco with various chewable
mixtures of herbs, spices, areca nut, betel leaf and other substances became popular
in the South Asian subcontinent. In the 18th and 19th centuries, dry powdered tobacco
to be snuffed into the nose became popular in parts of East-Asia.
The production and consumption of tobacco has expanded rapidly in developing
countries during the second half of the 20th century with the introduction of
industrially and mass produced tobacco products. Developing countries represented
in 1970 around 40% of the world production and consumption of cigarettes, and they
now represent 70% (Figure 1.7). The proportion of smokers is now highest
worldwide in lower-middle income countries, where it attains 40% (Figure 1.6). There
are regional differences: prevalence is lowest in the African Region as defined by the
World Bank (<15% men smoking), and highest in the Western Pacific Region (>45%
men smoking). In the South-East Asian region, prevalence of daily smoking is above
30% in men and it is close to 30% in the Eastern Mediterranean region. In particular,
in China and the Philippines, above 65% of men currently smoke 56. One shared
aspect of tobacco uses across developing countries is the presence of a gender gap
due to cultural and social reasons. On average, fewer than 50% women smoke in
developing countries (Figure 1.6). However, women are nowadays aggressively
targeted by the tobacco industry, which seeks to associate tobacco use with
feminism, sophistication, weight control, and Western-style independence. Recent
increases in female smoking prevalence have been reported in Cambodia, Malaysia
and Bangladesh 7.
17
Until recently, governments have been showing a weak response to the rise in
tobacco use. In terms of per capita public spending, for every $1 dedicated by
developed countries to tobacco control, middle income countries dedicate $0.005,
and low income countries $0.001 51. Philip Morris, the world’s biggest cigarette
company, was also in 1996 the world’s ninth largest advertiser, spending more than
$3 billion on promoting its products, mainly in developing countries. Following the
adoption in 2003 of the WHO Framework Convention on Tobacco Control, which
binds states to ban tobacco promotion, several developing countries have passed
anti-smoking laws. However, their impact remains limited. In Pakistan, anti-smoking
laws which ban consumption in public areas and storage near educational institutions
were passed in 2009, but are not being enforced 57. In India, a ban on the sale of
gutka issued in December 2010 has had no practical effects due to a lack of support
from the general population, insufficiently educated on the dangers of tobacco use 58.
In China, where the prevalence of smoking among health-care professionals reaches
40%, many (health professionals) smoke in front of their patients and as many as
29% of non-smoking physicians accept cigarettes as gifts, making them ineffective in
discouraging smoking initiation and promoting smoking cessation to the wider
population 59. Whilst prevalence of smokers who have quitted their habit is generally
above 35% in developed countries, it is less than 20% in developing countries 60. In
Bangladesh, India and China, the prevalence of former smokers is below 10% before
age 45 years old.
1.6 Burden and economic impact of tobacco use and cardiovascular
diseases
1.6.1 Tobacco overall morbidity and mortality
During the 20th century, 100 million people worldwide died from tobacco-related
diseases; and the tobacco burden is predicted to reach 1 billion individuals during the
21st century 7, 46. While the tobacco epidemic has reached its peak in developed
countries, it is only at its early phases in developing countries, with a gap between
men and women and an increasing prevalence of smoking over time (Figure 1.9) 61.
In 2000, nearly 5 million premature deaths in the world were attributable to smoking,
half of them in developing countries and half of them in developed countries 51.
Nowadays, 6 million people die from tobacco use each year, both from direct tobacco
and second-hand smoke 62. Based on current smoking patterns, annual smoking
18
related deaths will rise to 8.3 million by 2030, with four out of five deaths occurring in
developing countries. Smoking represented 3% of all deaths in 2000 for medium and
low income countries, and this proportion is expected to rise to 8% by 2030 (Figure
1.10) 63.
1.6.2 Cardiovascular burden from any cause
CVD is the main cause of death worldwide and over 80% of CVD deaths now occur
in developing countries 51. CVD were responsible for the largest proportion of non-
communicable deaths under 70 years old in 2008, representing 17 million deaths of
48% of deaths from non-communicable disease. Non-communicable diseases were
collectively responsible for 36 million deaths, representing 63% of all deaths (57
million individuals in 2008), exceeding in all regions except Africa the burden of
communicable, maternal, perinatal and nutritional conditions combined (Figure 1.11).
Population ageing is a significant trend in most parts of the world and translates into
increasing burden of CVD. Whereas annual infectious disease deaths are projected
to decline, annual CVD deaths are projected to increase by 6 million worldwide over
the next 20 years. The burden of non-communicable disease is expected to increase
by 15% globally between 2010 and 2020, to 44 million deaths a year. Greatest
increases are predicted to happen in Africa, South-East Asia and the Eastern
Mediterranean region, where the burden will increase by over 20%, and lowest
increases in the European Region where the annual burden is expected to remain
stable. By 2020, the regions that are projected to have the greatest total number of
deaths from non-communicable diseases are South-East Asia (10.4 million deaths)
and the Western Pacific (12.3 million deaths).
1.6.3 Cardiovascular burden as a result of tobacco use
Tobacco is estimated to cause 10% of CVD worldwide 63. In 2000, amongst
individuals aged ≥30 years old, smoking accounted for 22% of CVD mortality in North 
America, 13% in Western Europe, 10% in South Asia including India and 4% in
South-East Asia including China. Because female smokers are attaining similar
levels to male smokers in developed countries, smoking related CVD mortality is now
roughly similar across genders, while a gap remains in developing countries. In North
America, in 2000, smoking accounted for 23% of CVD mortality in men versus 21%
in women. Inversely, CVD represents a leading cause of the smoking related burden.
19
It represents 51% of smoking related deaths in Japan 64; compared to 13% in China
65.
CVD represented 1.7 million tobacco related deaths worldwide in 2000 and is
projected to reach 1.9 million deaths in 2015 (0.92 million CHD deaths, 0.52 million
stroke deaths and 0.24 million other CVD deaths) 66. In the USA, cigarette smoking is
the cause of nearly 500,000 premature deaths a year, about 1 in every 5 deaths, and
nearly half of them are due to cardiovascular diseases 67
1.6.4 Economic impact of cardiovascular diseases
Once thought of as a diseases of the rich, CVD is now the leading cause of death in
low- and middle-income countries and adds up to a heavy socio-economic toll
worldwide. Nearly 30% of all deaths from non-communicable diseases in developing
countries occur before 60 years old, whereas in high-income countries the proportion
is only 13% 51. Each year, 100 million people are pushed into poverty because they
have to pay directly for health services and for a large proportion this is as a result of
CVD. Developing economies with large population will bear the highest cost of the
CVD epidemic. From 2005 to 2012, China and India are projected to lose 1% of the
Gross Domestic Product as a result of heart diseases and 1.5% of Gross Domestic
Product as a result of strokes and diabetes.
1.6.5 Economic impact of tobacco use
There are multiple economic impacts of tobacco, ranging from health costs and
impoverishment of families to damage done to the environment (for example, forest
fires caused by lit cigarettes carelessly thrown to the ground). The personal
economic cost of buying tobacco impoverishes the most deprived sections of the
population, and slows economic development, especially in developing economies 7.
Indeed, tobacco use is more common amongst people in low socio-economic
categories, who dedicate an important part of their resources to buying tobacco
products and are often unable to cover the medical costs incurred as a result of
tobacco related diseases. The poorest households in Bangladesh spend almost 10
times as much on tobacco as on education; and tobacco use has been demonstrated
to exacerbate child malnutrition diverting household funds away from food and other
necessities 68. In Pakistan, a pack of imported manufactured cigarettes costs more
than half the average daily income 46. In Indonesia, the lowest income group spends
20
15% of its total expenditure on tobacco. In China, poor households spend between
7% in cities and 11% in the countryside of their income on cigarettes 69. Tobacco is
also a main cause of disability and death among middle-aged men, often the main
breadwinners of their families, causing considerable economic losses and
accelerating entry into poverty 63.
At a population level, tobacco health burden represents a sizeable part of total health
expenditure in both industrialized countries and developing countries. Developed
regions account for 12% of the worldwide burden from all causes of death and
disability, but for 90% of health expenditure 7. Total annual health expenditure related
to smoking, including diseases and deaths from passive smoking, runs to $81 billion
in the USA, $7 billion in Germany and $1 billion in Australia 46. In China, the annual
cost of health expenditure related to smoking reaches $3.5 billion 51.
1.7 Epidemiological evidence on the cardiovascular risks associated with
tobacco use
This section combines synthetizes literature reviews of smoking in relation to CVD
with regard to 1) dose-response relationship, 2) the effect of duration, age at risk and
starting age, 3) the effect of cessation, 4) the importance of confounding and 5) the
potential for effect modification. In addition, the relationship between smoking and
other risk factors or markers of CVD is reviewed. Finally, a review of the evidence
regarding other types of tobacco use is conducted and epidemiological evidence is
compared to cigarette smoking. Studies reviewed in this Chapter are listed at the end
of the Chapter in Table 1.1.
1.7.1 Landmark epidemiological studies and evaluation of the causal effect
of smoking on CVD risk
Until the landmark studies published in 1954 in the UK by Doll and Hill, and in 1958
in the US by Hammond and Horn 70 (Table 1.1), epidemiological studies on the
effects of tobacco had been small, focused on the risk of lung or mouth cancer, and
retrospective. These studies were unable to prove beyond doubt a causal association
between smoking and CVD. Therefore, Doll and Hill recognized the need for “some
entirely new approach. That approach I considered should be prospective. It should
determine the frequency with which the disease appeared, in the future, among
21
groups of persons whose smoking habits were already known”. Doll and Hill enrolled
40,000 British doctors in November 1951 who filled in a short questionnaire on their
smoking habits. After 3 years of follow-up, the authors found: “…There is a rise in the
mortality from deaths attributed to coronary thrombosis as the amount smoked
increases, but the gradient is much less steep than that revealed by cancer of the
lung”. After a follow-up of 12 years, the authors were able to conclude more
assertively “In short, that cigarette smoking is a cause of coronary thrombosis is not, I
think, proved; but it is the most reasonable interpretation of the available facts” 71 and
after a 20 years follow-up, they classified the burden of coronary heart diseases
(CHD) as “probably partly or wholly attributable to smoking” 72.
Complementary to the study by Doll and Hill, Hammond and Horn initiated in 1952 a
study on a larger scale in the USA, enrolling 188,000 white men age 50 to 69 years
old, and published their first results in 1958. During four years of follow-up, 11,870
deaths occurred, including 5297 due to CHD 70. With a larger data size and longer
follow-up than Doll and Hill, the authors had enough statistical power to conclude
already in 1958 that “coronary heart disease and other circulatory diseases showed a
high degree of association with cigarette smoking”, and to estimate a doubling in risk
of CHD death in individuals smoking at least 1 pack of cigarettes per day (CPD)
compared to never-smokers. Since these two landmark studies, cigarette smoking
has been demonstrated to cause myocardial infarction, stroke, aortic aneurysm,
sudden cardiac death and peripheral vascular disease 73.
1.7.2 Dose-response relationship
Two prospective studies including each around 7,000 people, the Goteborg Study
(GOTO) in Sweden and the British Regional Heart Study (BRHS) in the UK, found a
similar effect on the risk of fatal and non-fatal MI independently of the level of
exposure to cigarette smoking 77,78 (Table 1.1). However, most studies have
observed a dose-response relationship between CPD and the risk of death from CVD
79-81. In 1976, Doll & Peto published the first significant tests for trends with increasing
CPD for the risk of deaths from ischaemic heart disease, myocardial degeneration,
arteriosclerosis, aortic aneurysm and cerebral thrombosis 72. In the Nurses’ Health
Study (NHS), a prospective cohort study of 100,000 women (Table 1.1), the
multivariate adjusted HR for CHD deaths was 2.8, 4.8, 7.0, 7.8 in women smoking
respectively 1-14, 15-24, 25-34 and ≥35 CPD versus never-smokers 74. For stroke,
the test of linear trend was non-significant but the increased risk was 2.1 (95% CI:
22
1.6-2.8) in women smoking 1-4 CPD versus 3.3 (2.1-5.4) in women consuming ≥35 
CPD. INTERHEART, an international case-control study of the risk of MI conducted
in 52 countries (Table 1.1) estimated an increase of 6% in risk per additional
cigarette 75.
Even low levels of smoking have been associated with a significantly higher risk of
CVD. The risk of first ever non-fatal MI was significant in INTERHEART at 3-4 CPD.
In the NHS, smoking 1 to 4 CPD was associated with a twofold increase in the risk of
MI. In the Copenhagen City Heart Study (CCHS, see Table 1.1), individuals
consuming 3-5 grams of tobacco per day (corresponding approximately to 3-5 CPD)
were shown to carry a significantly increased risk of developing MI with a RR of 2.14
(1.11; 4.13) 76. In a prospective study conducted in Norway, started in 1970 and
followed up until 2002, smoking 1-4 CPD was positively associated with a
significantly higher risk of dying from MI 77,78. A rapid rise in CVD risk even in light
smokers is consistent with observations made in passive smokers, who are exposed
to relatively low levels of smoke compared to even light current smokers, but who are
nevertheless at significant increased risk of CVD 79. Cotinine is a metabolite of
nicotine with a relatively long half-life of 16 hours commonly used as a proxy to
exposure to passive smoking and daily cigarette consumption. A prospective study
measuring cotinine found that never-smokers exposed to high levels of passive
smoking were also at higher risk of MI: the HR for top versus bottom 4th of the
distribution was 1.57 (1.08-2.28) after adjustment for conventional risk factors 80. Law
and Wald proposed a shape of association with CHD with an exponential increase up
to 5 CPD, and then a linear increase up to 30 CPD. (Figure 1.12).
The existence of a plateau effect at high levels of smoking status remains
controversial. INTERHEART observed a linear increase in MI risk up to 21 cigarettes
per day with no evidence of a threshold 81. In the Framingham Study (Table 1.1), the
risk of stroke increased linearly with CPD and the RR was 2 for smoking greater than
40 CPD compared to fewer than 10 CPD 82. In a prospective study of 325,384 white
US males followed for 5 years, age-adjusted mortality rates were linearly increasing
with no plateau at consumption >35 CPD 83. In the Stroke Prevention and Young
Women Study (SPYW, see Table 1.1), a strong dose-response relationship was
observed for the risk of stroke, with an OR of 9.1 (95% CI: 3.2-26) for women
smoking 40 or more cigarettes per day compared to never-smokers 84. However, in
the NHS, risk ratios for all CHD were the same in women consuming either 25-34 or
23
≥35 CPD compared to never-smokers: 3.7 (3.1; 4.4) and 3.7 (1.9; 4.7) respectively 
after 24 years of follow-up 74, 85. Similarly for cerebrovascular disease, risk ratios were
2.9 (2.0; 4.2) and 2.9 (1.8; 4.8). The convergence effect found in this study has been
attributed to a steeper increase in exposure to smoking at low levels of smoking, as
heavier smokers may take lighter and shorter puffs, therefore inhaling less smoke. A
study which measured cotinine found that cotinine levels increased more steeply
from 0 to 10 CPD than from 20 to 30 and were reaching a plateau above 30 CPD 86.
1.7.3 Duration, age at risk and starting age
The effect of duration of smoking in relation to disease is difficult to disentangle from
the effect of intensity, due to the addictive nature of smoking, as amount is likely to
increase with increasing duration. In addition, most smokers start during adolescence
so age and duration are nearly collinear variables, and the estimation of their
separate effects requires large sample sizes 73. In the Multiple Risk Factors
Intervention Trial (MRFIT, see Table 1.1), a study conducted among 361,662 men
followed up for 10 years, the association of duration with risk of CHD death was not
significant, after adjustment for age, cholesterol levels, blood pressure and CPD 87.
However, rate ratios compared to never-smokers were increasing with duration at
any level of cigarette consumption and in any age strata, in individuals younger than
70 years old in CPS-1, and in everyone in CPS-2 85, and this finding has been
replicated in other studies 88. The effect of smoking on disease is often represented
by the number of pack-years increase, measured as the number of packs of CPD
times the number of years smoking, each pack containing 20 cigarettes, and makes
the assumption that an increase in 1 year of duration is equivalent to increasing
smoking consumption by 1 pack per day. In 3 City Study (3CS, see Table 1.1),
women smoking more than 30 pack-years had a hazard ratio of 3.2 (2.4-4.5)
compared to never-smokers 89.
Regarding age at risk, studies have consistently shown a decrease in risk ratios with
increasing age 71, 90-92. In the British Doctors Study (BDS, see Table 1.1), death rates
from CHD were 5.7 times higher among cigarette smokers than among non-smokers
at ages 35 to 44 but were approximately equal to those of non-smokers at ages 75 to
84. In a large case-control study of 14,000 cases and 32,000 controls, the rates of MI
in smokers were 5 times those in non-smokers at age 30-49 years old, 3 times at age
50-59, and twice at age 60-79 75. A similar effect has been observed for most risk
factors of CVD, with stronger associations observed at younger ages 85, 93, 94 . In
24
individuals below 45 years old, CHD was the dominant cause of increased mortality
attributable to cigarette smoking in CPS-2 85.
The age of starting smoking has also been associated with CVD risk, independently
of duration, amount, age, and known intermediate risk factors; with young starters
experiencing higher risks than late starters. In CPS-2 and in the US Veteran Study,
individuals starting earlier were generally at higher risk independently of their number
of CPD and their age 102. In the NHS, at 12 years of follow-up, the adjusted risk ratio
of CVD compared to never-smokers was 9.2 (95% CI: 5.3-16.2) in women who had
started before age 15, while it was 3.2 (2.1-4.8) for women who had started aged ≥26 
years 71.
1.7.4 Smoking cessation
While reducing smoking amount does not reduce CVD risk, smoking cessation is
beneficial, although studies differ in their estimation of time necessary to achieve a
risk comparable to that of never-smokers 85, 95. In the BDS, after correcting for
reverse causality by excluding the first 5 years of follow-up, CHD mortality decreased
relatively slowly in ex-smokers: the rate was 60% that of non-smokers after 5-9 years
and 29% the rate of non-smokers after ≥20 years 74. However, ex-smokers were
relatively rare at the time and these results were based on small numbers. In the US
Veteran Study, CHD deaths rates depended on the amount formerly smoked, were
not reduced until 20 years after stopping smoking, and became comparable to that of
never-smokers only 30 years after cessation 75. In the BRHS, the incidence of major
CHD event was significantly raised even 20 years after giving up smoking and
increased with the number of years smoked 96. In INTERHEART, while ORs for MI
risk were halved within the first 1-3 years of abstinence, they remained significant
after 20 years of cessation (1.31; 95%CI: 1.13-1.51) for ex- versus never-smokers
unexposed to passive smoking, independently of the amount previously smoked. By
contrast, in the NHS, the multivariate adjusted risk of CHD death was halved within 5
years and reached the level of non-smokers in 10-15 years, with a similar decrease
observed for cerebrovascular death 90. In the Third National Health and Examination
Survey (NHANES-3, see Table 1.1), inflammatory markers of atherosclerotic disease
such as C-reactive protein, white blood cell count, albumin and fibrinogen returned to
population levels 5 years after smoking cessation 97.
1.7.5 Confounding
25
1.7.5.1 Lifestyle risk factors: There has been a suggestion from the tobacco
industry that the association of smoking with CHD risk reflects an inadequate control
for confounding by lifestyle risk factors. In CPS-2, smokers tended to be less
educated, drank more alcohol and ate fewer vegetables. Male smokers in particular
were less likely to be employed, more likely to be in a unskilled job, consume a fatty
diet and were less physically active 14. However, adjustment of the risk ratios for race,
education level, marital status, unskilled job, weekly consumption of vegetables and
citrus fruit, aspirin use, alcohol consumption, body mass index, physical activity and
weekly consumption of fatty food decreased the estimate for CHD death for current
versus never-smokers by only 9% in men and 5% in women. For the risk of stroke
death, the biggest decrease upon adjustment was in men: the age adjusted HR for
current versus never was 2.1 (1.9-2.4) while the full adjusted HR was 1.7 (1.5-2.0) (in
women it was 2.3; 95%CI: 2.0-2.6 compared to 2.2; 95%CI: 2.0-2.5). In a British
study of 7,142 men followed up for 15 years, adding BMI and physical activity as
covariates to the Cox model provoked a modest increase in the HR of MI and stroke
98.
1.7.5.2 Intermediate CVD risk factors: Cigarette smoking has been shown to
influence several intermediate risk factors of CVD. It has been found to acutely
increase blood pressure and heart-rate, and with regular use throughout the day, the
increase remains persistent 99. In the Edinburgh Artery Study (EAS), smoking was
associated with reduced dietary antioxidant vitamin intake, HDL-C and diastolic blood
pressure (DBP), and with increased alcohol intake, serum triglycerides, blood
viscosity, plasma fibrinogen and markers of endothelial disturbance 100. In the BRHS,
smokers had lower BMI, DBP and HDL-C; while they had higher systolic blood
pressure (SBP) and serum triglycerides than non-smokers. Ex- versus never-
smokers had higher SBP, BMI, total cholesterol and triglycerides. In the Munster
Heart Study (MHS-PROCAM), the same observations were made, as well as higher
levels of fibrinogen in smokers versus non-smokers 101. Changes in lipid levels were
of a greater magnitude in women compared to men 74. However, adjusting for
intermediate risk factors of CVD only marginally affected RR in these two studies. In
the Seven Countries Study (SCS), after adjustment for cohort effect, age, BMI, serum
cholesterol levels, SBP and the presence of clinical CVD, the HR for smokers ≥10 
CPD remained highly significant at 1.8 (1.6-2.1) for CHD death compared to never-
smokers 102.
26
Overall, both lifestyle and intermediate risk factors fail to explain the association
between smoking and CVD, which indicate that smoking independently increases
risk and causes CVD by independent biological pathways. It also means that even a
tight control of other risk factors such as blood pressure and cholesterol level would
not counterbalance the effect of smoking. In NHS, adjustment for hypertension,
diabetes, high cholesterol levels, BMI, change in weight between age 18 and
baseline, alcohol intake, physical activity, contraceptive and hormone use,
menopause status, parental history of MI, diet, daily number of CPD and age of
starting smoking slightly raised the association with vascular deaths for current
versus never-smokers: 3.0 (2.4;3.2) to 3.3 (3.0; 3.8) 75.
1.7.6 Effect modifications
1.7.6.1 Sex: Early epidemiological studies observed significantly lower risk of all
causes of death and CVD for women than for men when they smoked, even after a
20 years follow-up in the case of the BDS 85, 103. However, women smokers included
in these studies were mostly born before the First World War; and these women
differed substantially from men of the same era who smoked, and from younger
women smokers. In particular, these women were less likely to inhale and had
started smoking later than their male counterparts. By contrast, in more recent
studies, the risk for women smoking has been shown to be equal 90 or even higher
than the risk for men 76, 104, in particular for the risk of early CVD 12, 89. In the Finmark
study, the incidence of MI was increased six-fold in women versus threefold in men
who smoked ≥20 CPD compared to never-smokers. In all age groups, the HR for 
women smoking was 3.3 (2.1-5.1) compared to 1.9 (1.6-2.3) in men 105. Reasons for
this difference are unknown and may include hormonal effects of smoking (female
sex hormones may affect CVD risk), synergy with oral contraceptive agents or even
arteriolar differences.
1.7.6.2 Intermediate factors and medical history: Whether the effect of smoking is
exacerbated by the presence of other known CVD risk factors remains uncertain
because of sparse evidence. In ARIC, the association of smoking with fatal and non-
fatal CHD was stronger in individuals with higher LDL-C levels 106. In individuals with
LDL-C ≥130 mg/dl and smoking ≥15 CPD, the HR was 2.81 and higher than 
expected if these factors were independently affecting CHD risk (1.15X1.71=1.97).
Nevertheless, confidence intervals were wide and this difference was not significant.
In the Asia Pacific Study Collaboration (ASPC), smoking interacted positively with
27
total cholesterol and HDL-C on CHD risk, but p-values fell between 0.01 and 0.05
and therefore were only modestly significant 107. In diabetic patients, smoking has
been found in a single study to interact with the duration of diabetes to accentuate
atherosclerosis 108. The presence of diabetes or hypertension was also found to
strengthen the association between smoking and subclinical atherosclerosis,
measured by carotid intima-media thickness 109, 109.
1.7.6.3 Lifestyle factors: It has been suggested that smoking worsens the negative
effect of alcohol and offsets the protective effect of low consumption of alcohol on
CVD risk. In two prospective studies, the protective effect of alcohol in light and
moderate alcohol drinkers on CVD mortality was non-significant in smokers, however
there was no statistical evidence of an interaction on a multiplicative scale at high
levels of consumption 110, 111.
1.7.7 Smoking and the progression of CVD
As seen in the section above, the risk of acute CVD events decreases rapidly upon
cessation of smoking, which may suggest that smoking has a greater impact on
plaque rupture and thrombus formation than on the atherosclerotic process of plaque
building 83. In the Atherosclerosis Risk in Community Study (ARIC), the association of
smoking was stronger with advanced rather than with early atherosclerosis measured
by carotid intima media thickness, regardless of levels of smoking exposure; and
smoking was associated with plaque calcification 106. The Multi-ethnic Study of
Atherosclerosis (MSA) found that smoking accelerated plaque progression to thicker,
more fibrous lesions which are more vulnerable to rupture 112. A stronger association
between smoking and CVD has also been reported at high compared to low levels of
LDL-C, which is consistent with the view that smoking accelerates the progression of
cholesterol-filled regions 113.
1.7.8 Other smoking and smokeless types of tobacco use and CVD risk
1.7.8.1 “Filtered”, “low tar” and “roll-your own” cigarettes: The tobacco
companies have been quick to develop alternatives to the standard cigarette with the
claim that these new products are less harmful. However, by changing the way they
smoke, for example, inhaling more deeply, or increasing the number of CPD, it is
possible to obtain as much nicotine from these new products as from regular
cigarettes, in order to satisfy one’s addiction, counterbalancing any “protective” effect
of these new products 114.
28
“Filtered” cigarettes, which aim to reduce the amount of toxicants that go into the
smoke inhaled by the smoker, have been shown to carry the same CVD risk as
“unfiltered” cigarettes 115. Similarly, “Low-tar” or “light cigarettes”, designed to burn
more quickly than standard cigarettes, and to produce lower machine-measured
yields of tar and nicotine, have not been shown to reduce the risk of non-fatal MI 116
or subclinical atherosclerosis 117.
“Roll your own” and bidies have not been demonstrated to carry either more or less
risk than cigarettes. In a case-control study in India, the risk of vascular death
associated with cigarette smoking was 1.8 (1.7-1.9) in an urban setting, where
manufactured cigarettes are predominant (26% smoked only cigarettes and 69%
both cigarettes and bidies); and was nearly identical to the OR in rural areas, where
bidies are most popular (38% smoked only bidies and 57% both bidies and
cigarettes): 1.7 (1.6-1.9) 118. In a smaller case-control study of Indians, individuals
smoking ≥10 cigarettes or bidies per day had an OR of 6.7 (p<0.001) 49. A systematic
review on bidi smoking found that bidi smokers inhale on average 2–3 times more
nicotine and tar than smokers of conventional cigarettes, due to the poor
combustibility of the bidi wrapper and greater puff frequency needed to keep the bidi
alight; with resulting health hazards at least as great as for cigarettes. A small scale
case-control study in Bangalore found higher OR for bidi versus cigarette smoking 3.
1.7.8.2 Pipe and cigars: The BDS was the first study to investigate the effect of pipe
and cigar on the risk of CVD and found a non-significant association, leading to the
early belief that these products were safer alternatives to cigarettes 119. Smoke from
pipes and cigars contains the same toxic substances as cigarette smoke, but those
who use a pipe or cigar tend to smoke at lower intensity and tend not to inhale the
smoke, thus reducing their exposure to its toxic substances. For example, 2/3 of
those who smoke both cigars and cigarettes (>40% cigar smokers) inhale cigar
smoke, compared with less than 15% of cigar smokers who have never smoked
cigarettes 120.
More recent studies have established that pipe and cigar smokers are not protected
even if their risk is smaller than cigarette smokers 121. In the Zutphen Study, current
cigarette smokers experienced a reduced life-expectancy by 6.8 years (disease-free
life years decreased by 5.8 years) compared to never cigarette smokers, whereas
29
current pipe or cigar smokers had on average 4.7 years of life lost (5.2 disease-free
years) compared to never pipe or cigar smokers122. In the Kaiser Permanente Study
(KPS), current cigar smokers were at higher risk of CHD after multiple adjustment
(RR compared to never-smokers: 1.27; 95%CI: 1.12 to 1.45) with evidence of a
dose-response relationship 123. In CPS-2, the association between current versus
never cigar smoking and CHD death was stronger among younger men and was
non-significant in ex-smokers 124. The association with current versus never pipe
smoking in this cohort was similar to that of cigar smoking, with a RR of 1.30 (1.18-
1.43) for CHD.
1.7.8.3 Smokeless tobacco: Smokeless tobacco is varied in its forms, and
epidemiological evidence on each type is limited and generally derived from small-
scale studies. Snuff users have been shown to have levels comparable to never-
smokers in terms of inflammation (fibrinogen 125 and C-reactive protein 126),
endothelial dysfunction (levels of carotid intima media thickness measured in the
carotid and femoral arteries), and oxidant stress 127. A case-control study conducted
in Bangladesh found that, compared to never users of tobacco, ever cigarette
smokers had an OR for CHD of 3.6 (1.5-8.5), ever bidi smokers had an OR of 2.9
(1.3-6.3), while ever use of betel nut with quid carried an excess risk of 3.8 (1.9-7.7)
and ever use of dried tobacco leaf (corresponding to Pakistani naswar) carried an
OR of 2.8 (1.0-4.5) 128. In a meta-analysis of smokeless tobacco in industrialized
countries, with data mainly coming from CPS conducted in the USA, the current use
of snuff and other smokeless tobacco products was significantly associated with fatal
MI and stroke risk 129. Pooled RR were respectively 1.13 (1.06; 1.21) and 1.40 (1.28;
1.54) taking never users as the reference group. The authors estimated that
smokeless tobacco contributed to 5.6% of all fatal MI, and to 5.4% of all fatal strokes
in Sweden which occurred in 2001. Combining information from both developing and
developed countries, INTERHEART obtained an OR of acute MI of 2.23 for only
users of smokeless tobacco compared to non-tobacco users; comparable to the
excess risk experienced by current cigarette smokers in that study (OR: 2.95 for
current cigarette smokers versus never users of tobacco) 81.
1.8 Aims of the thesis and outline
1.8.1 Strengths and weaknesses of the available epidemiological evidence
30
The evidence of a causal relationship between smoking and CVD risk is compelling
and has been gathered over more than 50 years of epidemiological research since
seminal papers in the UK and the USA 13. However, some aspects of this relationship
still remain unclear or subject to controversy. A stronger effect of smoking in women
compared to men has been observed by some studies but not others, and the
interplay between smoking and obesity, raised cholesterol and elevated blood
pressure has been rarely studied due to requirements of large sample sizes to detect
any significant interaction. The effect of smoking has been mostly investigated in
relation to most common CVD including MI and cerebrovascular events, but
uncertainty remains regarding the strength of an association between smoking and
pulmonary embolism, aortic aneurysm and other rare types of CVD. Review of the
literature on CVD and tobacco use showed that most studies investigated the effect
of cigarette smoking, which is the most common form of tobacco use in developed
countries. Therefore, the effect of pipes and cigar remain relatively unknown, and
there have not been any large scale studies conducted in Western European
populations on the effect of these other types of smoking. Regarding smokeless
forms of tobacco which are relatively uncommon in developed countries but common
in developing countries, there have been few studies investigating their effect on
cardiovascular health. These studies have been either set up in the USA or Northern
Europe and results may not be applicable to developing settings, or characterized by
small sample size.
1.8.2 Aims of the thesis
In this context, the aim of this thesis is to investigate the association between several
forms of tobacco and the risk of major cardiovascular events using epidemiological
data from developed and developing countries. Smoking is the most common form of
tobacco use in developed countries and the effect of cigarette smoking on the risk of
CVD has already been investigated by several studies, so the objectives are to:
(1) Summarize the evidence on cigarette smoking with the risk of CVD in
developed countries using meta-analyses an in a South Asian developing
country, Pakistan.
(2) Investigate the effect of pipe and cigar smoking with the risk of CVD, two
alternative forms of smoking popular in developed countries.
Regarding smokeless tobacco, its use remains mainly confined to the developing
world and here the objectives are to:
(1) Look at the effect of chewing tobacco in relation to MI risk in South Asia.
31
(2) Look at the effect of dipping tobacco in relation to MI risk in South Asia.
1.8.3 Outline of the thesis
This thesis is organised in two sections. Section A focuses on developed countries,
and Section B focuses on a developing country with a high prevalence of smoking
and smokeless tobacco use as well as a substantial population, Pakistan. In Section
A, the dataset used for my analysis is presented in Chapter 2, and is a large
collaboration of prospective cohorts with detailed information on smoking. Chapter 3
presents correlates of smoking with a range of other lifestyle and biochemical risk
factors of CVD. Chapter 4 looks at the associations between cigarette smoking and
CVD, which is compared in Chapter 5 to that of cigar and pipe smoking with CVD. In
Section B of this thesis, the relationship between smoking and smokeless tobacco
use with the risk of myocardial infarction is investigated in the context of Pakistan.
The Pakistan Risk of Myocardial Infarction Study is presented in Chapter 6.
Correlates of tobacco use with other conventional and locally relevant risk factors of
MI are presented in Chapter 7. In Chapter 8, I investigate strength of the association
between smoking and smokeless tobacco use with MI, with several levels of
adjustment for confounders and tests of effect modifications by other risk factors. The
Discussion exposes the limits of my results, and discusses their public health
relevance. It also envisions future work in the field of tobacco use and CVD risk.
Finally, in the Appendices, (1) a list of contributions to other analyses and resulting
publications is given, (2) the association between socio-economic status and the risk
of MI is presented, (3) the association between diet and the risk of MI using principal
component analyses is presented, and (4) the association between the chromosome
9p21 and the risk of MI in Pakistanis is shown.
32
Figures and Tables
Figure 1.1: Main types of tobacco use in Pakistan
Cigarettes ChilumHookahBeedies
SupariPaanGutka
Smoking tobacco
Smokeless tobacco
Naswar
Sources: 3rd International Conference on Smokeless Tobacco 2002 “Fact sheet”.
33
Figure 1.2: Potential sites of effect of smoking on thrombosis through oxidative
stress and other mechanisms
1. Increased number and activation of polymorphonuclear leukocytes; increased production of
superoxide radicals; and increased expression of integrins and adhesion molecules on
leukocytes and endothelial cells. 2. Increased oxidation of LDL-C; oxidized LDL-C taken up
more easily into macrophages to produce foam cells; and increased adhesiveness of
monocytes to endothelial cells. 3. Increased levels of fibrinogen; increased nitration of
tyrosine residues on fibrinogen, rendering it more thrombogenic; impaired activity of plasmin;
and decreased thrombolysis. Source: US Surgeon General Report, 2010 130.
34
Figure 1.3: Overview of mechanisms by which cigarette smoking causes an acute
cardiovascular event
Source: Benowitz, 2003 131.
35
Figure 1.4: Rapidity of absorption of nicotine according to types of tobacco use
Legend: Mean (standard error) blood concentrations of nicotine in 10 subjects who smoked
cigarettes for 9 minutes (1.3 cigarettes), used oral snuff (2.5g), used chewing tobacco (7.9 g)
and chewed nicotine gum (4 mg). Shaded bars above the time axis indicate the period of
exposure to tobacco or nicotine gum. Source: Benowitz, 1988 85.
36
Figure 1.5: Per capita consumption of different forms of tobacco in the United States
1880-1995
Source: U.S. Department of Agriculture 1996 51
37
Figure 1.6: Age-standardized prevalence of daily tobacco smoking in adults aged
15+ years, by WHO Region and World Bank income group, comparable estimates,
2008
Note: AFR: African Region, AMR: Region of the Americas, EMR: Eastern Mediterranean
Region, EUR: European Region, SEAR: South-East Asia Region, WPR: Western Pacific
Region. World Bank income groups are created dividing all Member States into 4 income
groups based on 2004 Gross National Income per capita: low, lower middle, upper middle,
and high. Source: WHO Global status report on non-communicable diseases 2010 7.
38
Figure 1.7: Share of cigarette production and consumption in developing countries.
Source: Based on data from Food and Agriculture Organization FAOSTAT, United Nations
Commodity Trade Statistics Database, United Nations Common Database, United States
Department of Agriculture Economic Research Service, World Health Organization Statistical
Information System, and ERC Group Plc.’s world Cigarettes Report 2005, extracted from
WHO report on tobacco epidemic 2008 132.
39
Figure 1.8: Average per capita cigarette consumption in persons aged ≥15 years by 
WHO region
Source: United Nations Statistics Division. 2003. Commodity Trade Statistics Database 51
40
Figure 1.9: Four stages of the tobacco epidemic
Note: x-axis indicates the number of years since smoking began. Source: Thun, 2012 133.
41
Figure 1.10: Projected number of tobacco-related deaths for high and middle plus
low income countries, 2002-2030
Source: Mathers 2006 Plos Medicine 62.
42
Figure 1.11: Total deaths by broad cause group, by WHO Region, World Bank
income group and by sex, 2008
Note: AFR: African Region, AMR: Region of the Americas, EMR: Eastern Mediterranean
Region, EUR: European Region, SEAR: South-East Asia Region, WPR: Western Pacific
Region. World Bank income groups are created dividing all Member States into 4 income
groups based on 2004 Gross National Income per capita: low, lower middle, upper middle,
and high. Source: WHO Global status report on non-communicable diseases 2010 104.
43
Figure 1.12: Model of the dose-response relationship between smoking and the risk
of ischemic heart disease
Note: Summary of evidence from a meta-analysis of 5 large cohort studies of active smoking
combined with the summary estimate from the studies of environmental tobacco smoke
exposure (taken to be equivalent to actively smoking 0.2 cigarettes per day). Source: Law &
Wald 2003 134
.
44
Table 1.1: Literature review of epidemiological studies on the association between tobacco use and CVD risk
Ref.
no
Study
acronym
Study
full name
Study design No. people recruited Country Baseline
survey
Follow-
up
period
Number of CVD events
107 3CS 3 Copenhagen study (CCHS + Glostrup
Population Study
PC 13,897 individuals Denmark 1976-78 (CCHS)
1964 (GPS)
7-16
years
-
49, 71,
72, 103
ARIC Atherosclerosis Risk in Community Study PC 15,792 individuals aged 45-64 years USA 1987-89 13 years 932 CHD events
96, 99 ASPC Asia Pacific Studies Collaboration Meta-analysis
of 40 PCs
500,000 Asians and 100,000
Australians
Asia 1961-99 7 years 4183 fatal & non fatal MI; 5930
fatal & non fatal strokes
76 BDS British Doctors Study PC 35,000 Male UK 1951 50 years in 2001, 7628 CHD deaths,
3307 stroke deaths
129 BRHS British Regional Heart Study PC 7,735
individuals
UK 1978 20 years 1,766 fatal or non fatal CVD
events at 15 years follow-up
135 CCHS Copenhagen City Heart Study PC 6,505 women and 5,644 men Denmark 1976-78 21 years 1348 fatal and non fatal MI
136 CPS-1 Cancer Prevention Study 1 PC ~1 million individuals USA 1960 6 years -
105 CPS-2 Cancer Prevention Study 2 PC ~1 million individuals USA 1982 6 years 14,585 CHD & 3,539 strokes
deaths
82,
114,
137
EAS Edinburgh Artery Study PC 1,592 individuals aged 55-74 years UK 1988 5 years 141 fatal & non fatal CHD
81 Finmark Finmark Study PC 11,843 individuals aged 35-52 years Finland 1977 12 years 498 fatal & non fatal MI
111 Framingham Framingham Heart Study PC 4,255 individuals age 36-68 years USA 1960 26 years 459 strokes
GOTO Goteborg study PC 6879 men aged 47-55 years Sweden 1970-80s 12 years 2277 CHD event
122 INTERHEART INTERHEART CC 12,461 MI cases and 14,637 controls World 2000-2 - 12460 MI cases
101 Iowa Study Iowa Study PC 41,836 women USA 1986 13 years 757 CHD deaths
87 KPS Kaiser Permanente Medical Care Program PC 60,838 individuals aged ≥35 years USA 1979 8 years -
112 MHS Munster Heart Study (also called PROCAM) PC 20,696 men and 10,212 women aged
40-65 years
Germany 1978-95 8 years 39 CHD events
138 MRFIT Multiple Risk Factors Intervention Trial Clinical trial 12,866 men USA 1980s 10 years -
74 MSA Multiethnic Study of Atherosclerosis CS 6,384 individuals aged 45-84 years USA 2000-2 - -
102 NHANES-III Third National Health and Examination
Survey
CS 33,994 individuals USA 1988-1994 - -
85 NHS Nurses' Health Study PC 100,000 female nurses USA 1980 24 years 1385 CHD & 734 strokes
deaths
121 SCS Seven Countries Study PC 12,763 men aged 40-89 years Europe, USA,
Japan
1957-
1964
25 years 1827 CHD & 797 stroke deaths
70 USVS US Veteran Study PC 300,000 men USA 1954 26 years 16,586 CVD deaths
78 Zutphen Zutphen study PC 1373 men Netherlands 1960 40 years -
- Hammon & Horn study PC 188,000 white men age 50-69 years old USA 1952 6 years 5,297 CVD deaths
Norway Oslo city and 3 counties in Norway PC 23,521 men and 19,201 women aged
35-49 years
Norway 1970 32 years 2253 CHD deaths
116 ISIS International Study of Infarct Survival CC 14,000 MI cases and 32,000 controls UK 1990s - -
Ref. no: Reference number; CC: case-control study. PC: prospective cohort study; CS: cross-sectional study; CT: clinical trial.
45
References
(1) www.who.int/fctc/en. WHO Framework Convention on Tobacco Control. 24-8-2012.
(2) World Health Organization. Tobacco: deadly in any form or disguise. 2006.
(3) Baker F, Ainsworth SR, Dye JT et al. Health risks associated with cigar smoking. Stat
Med 2002 June 15;21(11):1539-58.
(4) US code of federal regulation. 2011 April 24;Available at: URL:
http://cfr.vlex.com/vid/40-11-meaning-terms-19672629.
(5) Advocacy forum for tobacco control. 2011 May 1;Available at: URL:
http://www.aftcindia.org/tob_pro_india.htm .
(6) Ambrose JA, Barua RS. The pathophysiology of cigarette smoking and
cardiovascular disease: an update. J Am Coll Cardiol 2004 May;%19;43(10):1731-7.
(7) World Health Organization. WHO report on the global tobacco epidemic; the
MPOWER package. 2008.
(8) Gupta PC, Hamner JE. Control of tobacco-related cancers and other diseases.
International Symposium. 1992.
(9) Third International Conference on Smokeless Tobacco; Fact sheet. 2002.
(10) The British Medical Association. Illustrated Medical Dictionary. 2nd edition ed. 2007.
(11) Aaronson PI, Ward JPT. The cardiovascular system at a glance. 3rd edition ed. 2007.
(12) Libby, Bonow, Mann, Zipes. Braunwald's heart disease: a textbook of cardiovascular
disease. 8th edition Saunders Elsevier ed. 2009.
(13) Licastro F, Candore G, Lio D et al. Innate immunity and inflammation in ageing: a key
for understanding age-related diseases. Immun Ageing 2005 May 18;2:8.:8.
(14) Benowitz NL. Cigarette smoking and cardiovascular disease: pathophysiology and
implications for treatment. Prog Cardiovasc Dis 2003 July;46(1):91-111.
(15) Craig WY, Palomaki GE, Haddow JE. Cigarette smoking and serum lipid and
lipoprotein concentrations: an analysis of published data. BMJ 1989 March
25;298(6676):784-8.
(16) Harats D, Ben-Naim M, Dabach Y, Hollander G, Stein O, Stein Y. Cigarette smoking
renders LDL susceptible to peroxidative modification and enhanced metabolism by
macrophages. Atherosclerosis 1989 October;79(2-3):245-52.
(17) Libby P, Ridker PM, Maseri A. Inflammation and atherosclerosis. Circulation 2002
March 5;105(9):1135-43.
(18) Wannamethee SG, Lowe GD, Shaper AG, Rumley A, Lennon L, Whincup PH.
Associations between cigarette smoking, pipe/cigar smoking, and smoking cessation,
and haemostatic and inflammatory markers for cardiovascular disease. Eur Heart J
2005 September;26(17):1765-73.
46
(19) Lehr HA, Frei B, Arfors KE. Vitamin C prevents cigarette smoke-induced leukocyte
aggregation and adhesion to endothelium in vivo. Proc Natl Acad Sci U S A 1994
August 2;91(16):7688-92.
(20) Adams MR, Jessup W, Celermajer DS. Cigarette smoking is associated with
increased human monocyte adhesion to endothelial cells: reversibility with oral L-
arginine but not vitamin C. J Am Coll Cardiol 1997 March 1;29(3):491-7.
(21) Metz L, Waters DD. Implications of cigarette smoking for the management of patients
with acute coronary syndromes. Prog Cardiovasc Dis 2003 July;46(1):1-9.
(22) Burke A, Fitzgerald GA. Oxidative stress and smoking-induced vascular injury. Prog
Cardiovasc Dis 2003 July;46(1):79-90.
(23) Matetzky S, Tani S, Kangavari S et al. Smoking increases tissue factor expression in
atherosclerotic plaques: implications for plaque thrombogenicity. Circulation 2000
August 8;102(6):602-4.
(24) Xie XT, Liu Q, Wu J, Wakui M. Impact of cigarette smoking in type 2 diabetes
development. Acta Pharmacol Sin 2009 June;30(6):784-7.
(25) Rahman MM, Laher I. Structural and functional alteration of blood vessels caused by
cigarette smoking: an overview of molecular mechanisms. Curr Vasc Pharmacol 2007
October;5(4):276-92.
(26) Heeschen C, Jang JJ, Weis M et al. Nicotine stimulates angiogenesis and promotes
tumor growth and atherosclerosis. Nat Med 2001 July;7(7):833-9.
(27) Henningfield JE, Stapleton JM, Benowitz NL, Grayson RF, London ED. Higher levels
of nicotine in arterial than in venous blood after cigarette smoking. Drug Alcohol
Depend 1993 June;33(1):23-9.
(28) Arcavi L, Jacob P, III, Hellerstein M, Benowitz NL. Divergent tolerance to metabolic
and cardiovascular effects of nicotine in smokers with low and high levels of cigarette
consumption. Clin Pharmacol Ther 1994 July;56(1):55-64.
(29) Jolma CD, Samson RA, Klewer SE, Donnerstein RL, Goldberg SJ. Acute cardiac
effects of nicotine in healthy young adults. Echocardiography 2002 August;19(6):443-
8.
(30) Teter CJ, Asfaw B, Ni L, Lutz M, Domino EF, Guthrie SK. Comparative effects of
tobacco smoking and nasal nicotine. Eur J Clin Pharmacol 2002 August;58(5):309-
14.
(31) Symons JD, Stebbins CL. Hemodynamic and regional blood flow responses to
nicotine at rest and during exercise. Med Sci Sports Exerc 1996 April;28(4):457-67.
(32) Sabha M, Tanus-Santos JE, Toledo JC, Cittadino M, Rocha JC, Moreno H, Jr.
Transdermal nicotine mimics the smoking-induced endothelial dysfunction. Clin
Pharmacol Ther 2000 August;68(2):167-74.
(33) Csonka E, Somogyi A, Augustin J, Haberbosch W, Schettler G, Jellinek H. The effect
of nicotine on cultured cells of vascular origin. Virchows Arch A Pathol Anat
Histopathol 1985;407(4):441-7.
(34) Krupski WC, Olive GC, Weber CA, Rapp JH. Comparative effects of hypertension
and nicotine on injury-induced myointimal thickening. Surgery 1987
August;102(2):409-15.
47
(35) Aicher A, Heeschen C, Mohaupt M, Cooke JP, Zeiher AM, Dimmeler S. Nicotine
strongly activates dendritic cell-mediated adaptive immunity: potential role for
progression of atherosclerotic lesions. Circulation 2003 February 4;107(4):604-11.
(36) Axelsson T, Jansson PA, Smith U, Eliasson B. Nicotine infusion acutely impairs
insulin sensitivity in type 2 diabetic patients but not in healthy subjects. J Intern Med
2001 June;249(6):539-44.
(37) Eliasson B, Taskinen MR, Smith U. Long-term use of nicotine gum is associated with
hyperinsulinemia and insulin resistance. Circulation 1996 September 1;94(5):878-81.
(38) Louie D. The effects of cigarette smoking on cardiopulmonary function and exercise
tolerance in teenagers. Can Respir J 2001 July;8(4):289-91.
(39) Sheps DS, Herbst MC, Hinderliter AL et al. Production of arrhythmias by elevated
carboxyhemoglobin in patients with coronary artery disease. Ann Intern Med 1990
September 1;113(5):343-51.
(40) DeBias DA, Banerjee CM, Birkhead NC, Greene CH, Scott SD, Harrer WV. Effects of
carbon monoxide inhalation on ventricular fibrillation. Arch Environ Health 1976
January;31(1):42-6.
(41) Yanbaeva DG, Dentener MA, Creutzberg EC, Wesseling G, Wouters EF. Systemic
effects of smoking. Chest 2007 May;131(5):1557-66.
(42) Newby DE, McLeod AL, Uren NG et al. Impaired coronary tissue plasminogen
activator release is associated with coronary atherosclerosis and cigarette smoking:
direct link between endothelial dysfunction and atherothrombosis. Circulation 2001
April 17;103(15):1936-41.
(43) Bauer V, Sotnikova R. Nitric oxide--the endothelium-derived relaxing factor and its
role in endothelial functions. Gen Physiol Biophys 2010 December;29(4):319-40.
(44) Penn A, Snyder C. Arteriosclerotic plaque development is 'promoted' by polynuclear
aromatic hydrocarbons. Carcinogenesis 1988 December;9(12):2185-9.
(45) Penn A, Snyder CA. Butadiene inhalation accelerates arteriosclerotic plaque
development in cockerels. Toxicology 1996 October 28;113(1-3):351-4.
(46) World Health Organization. The Tobacco Atlas. 2002.
(47) CNN's brief history of tobacco. http://edition cnn
com/US/9705/tobacco/history/index.html Accessed on 2011 April 28.
(48) White C. Research on smoking and lung cancer: a landmark in the history of chronic
disease epidemiology. The Yale Journal of Biology and Medicine 1990;(63):29-46.
(49) Doll R, HILL AB. The mortality of doctors in relation to their smoking habits: a
preliminary report. BMJ 1954;328(7455):1529-33.
(50) Pierce JP, Messer K, White MM, Cowling DW, Thomas DP. Prevalence of heavy
smoking in California and the United States, 1965-2007. JAMA 2011 March
16;305(11):1106-12.
(51) World Health Organization. Global status report on noncommunicable diseases 2010.
2011 Apr.
48
(52) Allender S, Scarborough P, Peto V et al. European cardiovascular disease statistics:
2008 edition. 2008.
(53) O'Connor RJ, McNeill A, Borland R et al. Smokers' beliefs about the relative safety of
other tobacco products: findings from the ITC collaboration. Nicotine Tob Res 2007
October;9(10):1033-42.
(54) Delnevo CD. Smokers' choice: what explains the steady growth of cigar use in the
U.S.? Public Health Rep 2006 March;121(2):116-9.
(55) Tomar SL. The prevalence of smokeless tobacco use remained low and unchanged
among US men in 2000-2005 while smoking declined. J Evid Based Dent Pract 2010
September;10(3):183-5.
(56) Nishtar S. Prevention of coronary heart disease in south Asia. Lancet 2002
September 28;360(9338):1015-8.
(57) IRIN. Pakistan: anti-smoking law have little impact. IRIN Humanitarian news and
analysis 2004 July 14.
(58) Simpson D. India: gutka plastic pack ban. Tobacco Control 2011;20:89-92.
(59) Liu Y, Chen L. New medical data and leadership on tobacco control in China. Lancet
2011 March 23.
(60) Giovino GA, Mirza SA, Samet JM et al. Tobacco use in 3 billion individuals from 16
countries: an analysis of nationally representative cross-sectional household surveys.
Lancet 2012 August 18;380(9842):668-79.
(61) Ezzati M, Lopez AD. Estimates of global mortality attributable to smoking in 2000.
Lancet 2003 September 13;362(9387):847-52.
(62) Mathers CD, Loncar D. Projections of global mortality and burden of disease from
2002 to 2030. PLoS Med 2006 November;3(11):e442.
(63) Murray CJ, Lopez AD. Global mortality, disability, and the contribution of risk factors:
Global Burden of Disease Study. Lancet 1997 May 17;349(9063):1436-42.
(64) Katanoda K, Marugame T, Saika K et al. Population attributable fraction of mortality
associated with tobacco smoking in Japan: a pooled analysis of three large-scale
cohort studies. J Epidemiol 2008;18(6):251-64.
(65) Liu BQ, Peto R, Chen ZM et al. Emerging tobacco hazards in China: 1. Retrospective
proportional mortality study of one million deaths. BMJ 1998 November
21;317(7170):1411-22.
(66) Ezzati M, Henley SJ, Thun MJ, Lopez AD. Role of smoking in global and regional
cardiovascular mortality. Circulation 2005 July 26;112(4):489-97.
(67) National Institute on drug abuse. NIDA InfoFacts: Cigarettes and other tobacco
products. [http://www.nida.nih.gov/infofacts/tobacco.html#References], Accessed on
01/06/2010.
(68) Best CM, Sun K, de PS, Bloem MW, Stallkamp G, Semba RD. Parental tobacco use
is associated with increased risk of child malnutrition in Bangladesh. Nutrition 2007
October;23(10):731-8.
49
(69) Hu TW, Mao Z, Liu Y, de BJ, Ong M. Smoking, standard of living, and poverty in
China. Tob Control 2005 August;14(4):247-50.
(70) Hammond EC, Horn D. Smoking and death rates: report on forty-four months of
follow-up of 187,783 men. 2. Death rates by cause. J Am Med Assoc 1958 March
15;166(11):1294-308.
(71) Doll R, Hill AB. Mortality in relation to smoking: ten years' observations of British
Doctors. Br Med J 1964 May 30;1(5395):1399-410.
(72) Doll R, Peto R. Mortality in relation to smoking: 20 years' observations on male British
doctors. Br Med J 1976 December 25;2(6051):1525-36.
(73) U.S.Department of Health and Human Services. How tobacco smoke causes
disease: the biology and behavioral basis of smoking-attributable disease: a report of
the surgeon general. 2010.
(74) Kenfield SA, Stampfer MJ, Rosner BA, Colditz GA. Smoking and smoking cessation
in relation to mortality in women. JAMA 2008 May 7;299(17):2037-47.
(75) Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in
52 countries in the INTERHEART study: a case-control study. Lancet 2006
August;368(9536):647-58.
(76) Prescott E, Scharling H, Osler M, Schnohr P. Importance of light smoking and
inhalation habits on risk of myocardial infarction and all cause mortality. A 22 year
follow up of 12 149 men and women in The Copenhagen City Heart Study. J
Epidemiol Community Health 2002 September;56(9):702-6.
(77) Willett WC, Green A, Stampfer MJ et al. Relative and absolute excess risks of
coronary heart disease among women who smoke cigarettes. N Engl J Med 1987
November;317(21):1303-9.
(78) Bjartveit K, Tverdal A. Health consequences of smoking 1-4 cigarettes per day. Tob
Control 2005 October;14(5):315-20.
(79) Gallo V, Neasham D, Airoldi L et al. Second-hand smoke, cotinine levels, and risk of
circulatory mortality in a large cohort study of never-smokers. Epidemiology 2010
March;21(2):207-14.
(80) Whincup PH, Gilg JA, Emberson JR et al. Passive smoking and risk of coronary heart
disease and stroke: prospective study with cotinine measurement. BMJ 2004 July
24;329(7459):200-5.
(81) Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in
52 countries in the INTERHEART study: a case-control study. Lancet 2006
August;368(9536):647-58.
(82) Wolf PA, D'Agostino RB, Kannel WB, Bonita R, Belanger AJ. Cigarette smoking as a
risk factor for stroke. The Framingham Study. JAMA 1988 February;259(7):1025-9.
(83) Neaton JD, Kuller LH, Wentworth D, Borhani NO. Total and cardiovascular mortality
in relation to cigarette smoking, serum cholesterol concentration, and diastolic blood
pressure among black and white males followed up for five years. Am Heart J 1984
September;108(3 Pt 2):759-69.
50
(84) Bhat VM, Cole JW, Sorkin JD et al. Dose-response relationship between cigarette
smoking and risk of ischemic stroke in young women. Stroke 2008
September;39(9):2439-43.
(85) U.S. National Cancer Institutes of Health. Smoking and tobacco control: eighth
monograph. Feb. 1997. Available online at wwww.cancer.gov.
(86) Caraballo RS, Giovino GA, Pechacek TF et al. Racial and ethnic differences in serum
cotinine levels of cigarette smokers: Third National Health and Nutrition Examination
Survey, 1988-1991. JAMA 1998 July 8;280(2):135-9.
(87) Kuller LH, Ockene JK, Meilahn E, Wentworth DN, Svendsen KH, Neaton JD.
Cigarette smoking and mortality. MRFIT Research Group. Prev Med 1991
September;20(5):638-54.
(88) Tverdal A. Calculation of risk for the development of acute myocardial infarction in the
normal population based on long-term follow-up studies: smokers compared with
non-smokers. J Cardiovasc Risk 1999 October;6(5):287-91.
(89) Prescott E, Bjerg AM, Andersen PK, Lange P, Vestbo J. Gender difference in
smoking effects on lung function and risk of hospitalization for COPD: results from a
Danish longitudinal population study. Eur Respir J 1997 April;10(4):822-7.
(90) Thun MJ, Apicella LF, Henley SJ. Smoking vs other risk factors as the cause of
smoking-attributable deaths: confounding in the courtroom. JAMA 2000 August
9;284(6):706-12.
(91) Dept.of Health and Human Services. The health consequences of smoking: a report
of the Surgeon General. 2004.
(92) Woodward M, Lam TH, Barzi F et al. Smoking, quitting, and the risk of cardiovascular
disease among women and men in the Asia-Pacific region. Int J Epidemiol 2005
October;34(5):1036-45.
(93) Anand SS, Islam S, Rosengren A et al. Risk factors for myocardial infarction in
women and men: insights from the INTERHEART study. Eur Heart J 2008
April;29(7):932-40.
(94) Thun MJ, Day-Lally C, Myers D et al. Trends in tobacco smoking and mortality from
cigarette use in Cancer Prevention Studies I (1959 through 1965) and II (1982
through 1988). Changes in Cigarette-Related Disease Risks and Their Implication for
Prevention and Control.Smoking and Tobacco Control Monograph No.8.Rockville
(MD) [a], 305-382. 1997. Shopland DR, Burns DM, Garfinkel L, Samet JM.
(95) Godtfredsen NS, Osler M, Vestbo J, Andersen I, Prescott E. Smoking reduction,
smoking cessation, and incidence of fatal and non-fatal myocardial infarction in
Denmark 1976-1998: a pooled cohort study. J Epidemiol Community Health 2003
June;57(6):412-6.
(96) Cook DG, Shaper AG, Pocock SJ, Kussick SJ. Giving up smoking and the risk of
heart attacks. A report from The British Regional Heart Study. Lancet 1986 December
13;2(8520):1376-80.
(97) Bakhru A, Erlinger TP. Smoking cessation and cardiovascular disease risk factors:
results from the Third National Health and Nutrition Examination Survey. PLoS Med
2005 June;2(6):e160.
51
(98) Wannamethee SG, Shaper AG, Walker M, Ebrahim S. Lifestyle and 15-year survival
free of heart attack, stroke, and diabetes in middle-aged British men. Arch Intern Med
1998 December 7;158(22):2433-40.
(99) Shaper AG, Pocock SJ, Walker M, Phillips AN, Whitehead TP, Macfarlane PW. Risk
factors for ischaemic heart disease: the prospective phase of the British Regional
Heart Study. J Epidemiol Community Health 1985 September;39(3):197-209.
(100) Price JF, Mowbray PI, Lee AJ, Rumley A, Lowe GD, Fowkes FG. Relationship
between smoking and cardiovascular risk factors in the development of peripheral
arterial disease and coronary artery disease: Edinburgh Artery Study. Eur Heart J
1999 March;20(5):344-53.
(101) Cullen P, Schulte H, Assmann G. Smoking, lipoproteins and coronary heart disease
risk. Data from the Munster Heart Study (PROCAM). Eur Heart J 1998
November;19(11):1632-41.
(102) Jacobs DR, Jr., Adachi H, Mulder I et al. Cigarette smoking and mortality risk: twenty-
five-year follow-up of the Seven Countries Study. Arch Intern Med 1999 April
12;159(7):733-40.
(103) Doll R, Gray R, Hafner B, Peto R. Mortality in relation to smoking: 22 years'
observations on female British doctors. Br Med J 1980 April 5;280(6219):967-71.
(104) Prescott E, Hippe M, Schnohr P, Hein HO, Vestbo J. Smoking and risk of myocardial
infarction in women and men: longitudinal population study. BMJ 1998 April
4;316(7137):1043-7.
(105) Njolstad I, Arnesen E, Lund-Larsen PG. Smoking, serum lipids, blood pressure, and
sex differences in myocardial infarction. A 12-year follow-up of the Finnmark Study.
Circulation 1996 February 1;93(3):450-6.
(106) Hozawa A, Folsom AR, Sharrett AR, Payne TJ, Chambless LE. Does the impact of
smoking on coronary heart disease differ by low-density lipoprotein cholesterol level?:
the Atherosclerosis Risk in Communities (ARIC) Study. Circ J 2006
September;70(9):1105-10.
(107) Nakamura K, Barzi F, Huxley R et al. Does cigarette smoking exacerbate the effect of
total cholesterol and high-density lipoprotein cholesterol on the risk of cardiovascular
diseases? Heart 2009 June;95(11):909-16.
(108) Karim R, Buchanan TA, Hodis HN, Li Y, Mack WJ. The association of smoking and
subclinical atherosclerosis in Type 2 diabetes: modification by duration of diabetes.
Diabet Med 2005 January;22(1):81-7.
(109) Richey SA, Coady SA, Folsom AR, Couper DJ, Heiss G. Smoking and diabetes differ
in their associations with subclinical atherosclerosis and coronary heart disease-the
ARIC Study. Atherosclerosis 2004 January;172(1):143-9.
(110) Xu WH, Zhang XL, Gao YT et al. Joint effect of cigarette smoking and alcohol
consumption on mortality. Prev Med 2007 October;45(4):313-9.
(111) Ebbert JO, Janney CA, Sellers TA, Folsom AR, Cerhan JR. The association of
alcohol consumption with coronary heart disease mortality and cancer incidence
varies by smoking history. J Gen Intern Med 2005 January;20(1):14-20.
52
(112) Sharrett AR, Ding J, Criqui MH et al. Smoking, diabetes, and blood cholesterol differ
in their associations with subclinical atherosclerosis: the Multiethnic Study of
Atherosclerosis (MESA). Atherosclerosis 2006 June;186(2):441-7.
(113) Stratton K, Shetty P, Wallace R, Bondurant S. Clearing the smoke: the science base
for tobacco harm reduction--executive summary. Tob Control 2001 June;10(2):189-
95.
(114) Castelli WP, Garrison RJ, Dawber TR, McNamara PM, Feinleib M, Kannel WB. The
filter cigarette and coronary heart disease: the Framingham study. Lancet 1981 July
18;2(8238):109-13.
(115) Baldassarre D, Castelnuovo S, Frigerio B et al. Effects of timing and extent of
smoking, type of cigarettes, and concomitant risk factors on the association between
smoking and subclinical atherosclerosis. Stroke 2009 June;40(6):1991-8.
(116) Gajalakshmi V, Peto R, Kanaka TS, Jha P. Smoking and mortality from tuberculosis
and other diseases in India: retrospective study of 43000 adult male deaths and
35000 controls. Lancet 2003 August 16;362(9383):507-15.
(117) Gallus S, Randi G, Negri E, Tavani A, La VC. Tar yield and risk of acute myocardial
infarction: pooled analysis from three case-control studies. Eur J Cardiovasc Prev
Rehabil 2007 April;14(2):299-303.
(118) Pais P, Pogue J, Gerstein H et al. Risk factors for acute myocardial infarction in
Indians: a case-control study. Lancet 1996 August 10;348(9024):358-63.
(119) Pais P, Fay MP, Yusuf S. Increased risk of acute myocardial infarction associated
with beedi and cigarette smoking in Indians: final report on tobacco risks from a case-
control study. Indian Heart J 2001 November;53(6):731-5.
(120) Shaper AG, Wannamethee SG, Walker M. Pipe and cigar smoking and major
cardiovascular events, cancer incidence and all-cause mortality in middle-aged British
men. Int J Epidemiol 2003 October;32(5):802-8.
(121) Streppel MT, Boshuizen HC, Ocke MC, Kok FJ, Kromhout D. Mortality and life
expectancy in relation to long-term cigarette, cigar and pipe smoking: the Zutphen
Study. Tob Control 2007 April;16(2):107-13.
(122) Iribarren C, Tekawa IS, Sidney S, Friedman GD. Effect of cigar smoking on the risk of
cardiovascular disease, chronic obstructive pulmonary disease, and cancer in men. N
Engl J Med 1999 June 10;340(23):1773-80.
(123) Jacobs EJ, Thun MJ, Apicella LF. Cigar smoking and death from coronary heart
disease in a prospective study of US men. Arch Intern Med 1999 November
8;159(20):2413-8.
(124) Henley SJ, Thun MJ, Chao A, Calle EE. Association between exclusive pipe smoking
and mortality from cancer and other diseases. J Natl Cancer Inst 2004 June
2;96(11):853-61.
(125) Eliasson M, Asplund K, Evrin PE, Lundblad D. Relationship of cigarette smoking and
snuff dipping to plasma fibrinogen, fibrinolytic variables and serum insulin. The
Northern Sweden MONICA Study. Atherosclerosis 1995 February;113(1):41-53.
(126) Wallenfeldt K, Hulthe J, Bokemark L, Wikstrand J, Fagerberg B. Carotid and femoral
atherosclerosis, cardiovascular risk factors and C-reactive protein in relation to
53
smokeless tobacco use or smoking in 58-year-old men. J Intern Med 2001
December;250(6):492-501.
(127) Stegmayr B, Johansson I, Huhtasaari F, Moser U, Asplund K. Use of smokeless
tobacco and cigarettes--effects on plasma levels of antioxidant vitamins. Int J Vitam
Nutr Res 1993;63(3):195-200.
(128) Rahman MA, Zaman MM. Smoking and smokeless tobacco consumption: possible
risk factors for coronary heart disease among young patients attending a tertiary care
cardiac hospital in Bangladesh. Public Health 2008 December;122(12):1331-8.
(129) Boffetta P, Straif K. Use of smokeless tobacco and risk of myocardial infarction and
stroke: systematic review with meta-analysis. BMJ 2009 August 18;339.
(130) Boersma E, Mercado N, Poldermans D, Gardien M, Vos J, Simoons ML. Acute
myocardial infarction. Lancet 2003 March 8;361(9360):847-58.
(131) Benowitz NL, Porchet H, Sheiner L, Jacob P, III. Nicotine absorption and
cardiovascular effects with smokeless tobacco use: comparison with cigarettes and
nicotine gum. Clin Pharmacol Ther 1988 July;44(1):23-8.
(132) Shafey O, Dolwick S, Guindon EG, World Health Organization, American Cancer
Society, International Union Against Cancer. Tobacco control country profile. 2nd ed.
2003.
(133) Thun M, Peto R, Boreham J, Lopez AD. Stages of the cigarette epidemic on entering
its second century. Tob Control 2012 March;21(2):96-101.
(134) Law MR, Wald NJ. Environmental tobacco smoke and ischemic heart disease. Prog
Cardiovasc Dis 2003 July;46(1):31-8.
(135) Steenland K, Thun M, Lally C, Heath C, Jr. Environmental tobacco smoke and
coronary heart disease in the American Cancer Society CPS-II cohort. Circulation
1996 August 15;94(4):622-8.
(136) Price JF, Mowbray PI, Lee AJ, Rumley A, Lowe GD, Fowkes FG. Relationship
between smoking and cardiovascular risk factors in the development of peripheral
arterial disease and coronary artery disease: Edinburgh Artery Study. Eur Heart J
1999 March;20(5):344-53.
(137) Dawber TR, Kannel WB, REVOTSKIE N, STOKES J, III, KAGAN A, GORDON T.
Some factors associated with the development of coronary heart disease: six years'
follow-up experience in the Framingham study. Am J Public Health Nations Health
1959 October;49:1349-56.:1349-56.
(138) Seeman T, Merkin SS, Crimmins E, Koretz B, Charette S, Karlamangla A. Education,
income and ethnic differences in cumulative biological risk profiles in a national
sample of US adults: NHANES III (1988-1994). Soc Sci Med 2008 January;66(1):72-
87.
54
Section A: Smoking and the risk of cardiovascular diseases in
developed countries
Chapter 2: Description of the Emerging Risk Factors
Collaboration
Summary
The aim of this Chapter is to present the dataset used to investigate the effect of
smoking on CVD in developed populations. The Emerging Risk Factors
Collaboration (ERFC) was set up with the aim of better characterizing associations
between lipids and inflammatory markers with the risk of CVD. Compared to
literature-based meta-analyses, ERFC collected individual participants’ data,
enabling meta-analyses of study specific estimates with consistent adjustment for
confounders and tests for interaction across studies. By April 2011, ERFC had
enrolled 135 prospective epidemiological studies with information on lipids and
inflammatory markers as well as lifestyle, medical and demographic characteristics,
and follow-up to first major cardiovascular event or main cause of death. After
exclusion of studies conducted in developing countries or with insufficient information
on smoking status, a subset of 929,335 individuals from 114 studies was selected for
the analyses of cigarette smoking presented in Chapter 4. A subset of 20 studies
has collected information on cigar use and a subset of 22 studies on pipe use
presented in Chapter 5. During an average follow-up of 14.2 years; participants
experienced 40,218 incident coronary artery disease outcomes, 17,445 strokes,
9,788 lung cancer deaths, and a total of 128,137 deaths from all causes. The ERFC
will allow more detailed analyses of the risk of CVD in developed populations in
relation to smoking cigarettes, pipes or cigars.
55
2.1 Background of ERFC
The aim of the Emerging Risk Factors Collaboration is to “characterize more
precisely and in greater detail than had previously been possible the shape and
strength of associations of several lipids and inflammatory markers with incident
CHD and other cardiovascular outcomes, under a wide range of circumstances”
(http://www.phpc.cam.ac.uk/ceu/research/erfc/) 2. For this purpose, the ERFC
established a central database with ongoing recruitment of suitable studies (List
2.1). By September 2012, the ERFC included more than 2 million participants from
over 135 prospective studies with population-based samples. Subsets of participants
had information on lipid and inflammatory markers, lifestyle and behavioural factors
as well as major CHD and cerebrovascular events, and cause-specific mortality. The
ERFC main objectives have been fulfilled with individual participants’ meta-analyses
on the relationship with CVD of triglycerides 3, major lipids and apolipoproteins 4,
lipoprotein(a) 5, and C-reactive protein 6. To help harvest the information made
available by the collaboration, the ERFC has been enlarging its scope of analysis to
lifestyle factors, such as tobacco, obesity 7, and diabetes 8,9. Data requests were sent
to all the ERFC investigators in December 2010 (List 2.2) requiring more detailed
information on a range of lifestyle and other risk factors, including smoking amount,
duration, pack-years, type, age starting and age stopping smoking. As a result, the
ERFC represents one of the largest collections of information on smoking habits and
other CVD risk factors, with follow-up of participants for major vascular and non-
vascular events, similar in scale to the Cancer Prevention Studies set up in the U.S.A
in 1960s and 1980, which included over 1 million participants 10. Its uniqueness
resides in the length of follow-up of prospective studies included in the Collaboration,
in the depth of information gathered on smoking and other cardiovascular risk
factors, such as lipids and inflammatory markers, and in the way individual
prospective cohort studies enrolled random samples from the general population.
2.2 Methods
2.2.1 Inclusion of studies
As the initial aim of the ERFC was to investigate lipids and inflammatory markers, the
criteria for inviting studies to join the ERFC were (1) data available from baseline
measurements on at least one circulating lipid or inflammatory marker, (2) at least 1
year of follow-up, (3) participants not selected on the basis of having previous
cardiovascular diseases; and (4) information on cause-specific mortality and/or major
cardiovascular morbidity collected during follow-up 2. To enable proper adjustment
56
for confounding factors and investigation of effect modifiers, additional data were
sought on lifestyle, demographic and biochemical risk factors (List 2.1). Studies
were identified through meta-analyses publications, literature search of databases,
scanning of reference lists and correspondence with authors of relevant reports.
Data obtained from each participating study were checked for consistency and
harmonized to a standard format by the coordinating centre. For biochemical factors
expressed in different units of measurement, conversions to a standard unit were
operated 11. Information was stored on the different coding systems and on the assay
methods used by individual studies to measure biochemical factors. For categorical
variables, the number of categories was defined in order to maximize information
whilst taking into account the need for harmonization of differential coding systems
operated by individual studies.
2.2.2 Smoking information
Information relative to smoking was self-reported. Some studies provided smoking
status already coded as “current smoker”, “ex-smoker” and “never smoker” whilst
some studies provided answers to questions such as “when did you last smoke”,
“have you smoked at least once over the past week / month / year preceding the
interview?”. The data managing team of the ERFC, communicating with study
investigators, created a harmonized variable of smoking status categorized as
“current”, “ex-smoker” and “never smoker”.
Smoking amount was recorded for cigarettes, pipes and cigars separately when
available. To enable comparisons across different types of smoking and also to be
able to quantify total exposure to tobacco, self reported amounts of cigars, pipes and
cigarettes were converted into a unique scale using the approximate amount of
tobacco contained in each form of tobacco. One cigar was thought to contain on
average 4 grams of tobacco and correspond to approximately 3 cigarettes. One pipe
was converted into 1.6 cigarettes and 1 cigarillo was converted into 2 cigarettes
(personal correspondence between data management teams of the ERFC and
Prospective Study Collaboration 12). Smoking pack-years was defined as the number
of packs of cigarettes or cigarettes equivalent per day – 1 pack corresponding to 20
cigarettes equivalent per day – multiplied by the number of years smoked.
Regarding type of smoking, individuals reporting the use of small or large cigar, or of
cigarillo, were coded as current cigar smokers. Some studies provided information
57
on only pipe, only cigar or only cigarette smoking. In order to maximize information
available, individuals with missing information on cigar or on pipe were considered as
non pipe or non cigar smokers in the main analyses. This is a reasonable
assumption because currently pipe and cigar smokers represent below 5% of the
population in most populations from developed countries 13,14. Subsidiary analyses in
which only studies providing information on all three types: pipe, cigar and cigarettes
were also conducted to assess robustness of the results.
2.2.3 Follow-up and outcomes
For each individual, data were sought on each of the following outcomes and on their
dates of occurrence: fatal and non-fatal MI events and/or fatal and non-fatal stroke
events; and cause-specific mortality. In registering fatal outcomes, all contributing
studies used coding from the International Classification of Diseases to at least 3
digits or study-specific classification systems, and ascertainment was based on
death certificates (Table 2.1). Attribution of death referred to the primary cause
provided or, in its absence, the underlying cause provided.
The study was approved by the Cambridgeshire Ethics Review Committee and
analyzed independently from its funders.
2.3 Summary of dataset
2.3.1 Characteristics of studies
Out of the 114 studies, 5 were nested case-control studies, 10 were clinical trials and
the rest were cohort studies (Table 2.2). Studies included spanned 4 decades, with
the earliest cohort starting recruitment in 1960 and the most recent having initiated
enrolment in 2001. Most individuals were recruited in the 1970s (32%), 1980s (37%)
or 1990’s (24%) while only 8% were enrolled in 1960s and less than 1% in 2000 and
after. Average follow-up time was 14.2 years (inter-quartile range: 3.3 years to 30.4
years). Length of follow-up varied widely between studies, with the longest follow-up
exceeding 35 years (KARELIA) and the shortest follow-up below 4 years
(MOGERAUG3). Overall, 32% of individuals had less than 10 years of follow-up,
41% between 10 and 20 years, 21% between 20 and 30 years and 6% were
followed up for 30 years or over. Average age at baseline was 49.5 years old
(Standard Deviation: 12.3 years) and 51% of the data were men.
58
2.3.2 Baseline smoking characteristics
Smoking information was available in April 2011 for 117 studies. Out of these, 3
studies were excluded because they were set up in a developing country (China,
Turkey and the Caribbean), leaving 114 studies, which included 929,335 individuals
(List 2.3). In this data, 34% declared being current smokers (316,688 individuals),
23% ex-smokers (214,017 individuals) and 43% never smokers (398,630 individuals)
(Figure 2.1).
Subsets of the 114 studies shared further information on smoking: 43 studies
provided data on number of pack-years (114,662 individuals with non-missing and
non-null values), 20 studies on cigar (131,816 individuals) and 22 studies on pipe
smoking (132,060 individuals), 44 studies on starting age (119,907 individuals), and
54 studies on number of years since quitting smoking (80,318 individuals) (Table
2.3).
The distribution of amount smoked showed evidence of digit preferences for multiple
of 5 which may be due to the fact that cigarettes are usually sold in packs of 20 (and
therefore individuals are more likely to answer that they smoke 1 pack or half a pack
of cigarettes per day, and it could also be that the question was asked in this way in
the original study questionnaire). Indeed, 47% of smokers answered that they were
currently smoking a multiple of 5 (5, 10, 15, 20 or 25) cigarettes per day.
Most individuals had started smoking during teenage-hood or early adulthood: more
than 50% of current and ex-smokers had initiated smoking at age 18 years old or
younger (IQR: 16-22) and 75% had begun before age 22 years old. Quitting age was
more evenly spread, with a median age of 40 years old and an inter-quartile range of
17 years old (31 to 48 years old); and mean 42.9 years old (SD: 11.2). Those who
had stopped were on average 39 years old (SD: 12) and had stopped for 9 years
(IQR: 3 to 18) (Table 2.2). Mean pack-years smoked was 21 (SD: 6.8) in current
smokers and 22 (SD: 19.4) in ex-smokers (Table 2.4).
Cigarette was the most popular form of smoking: 34% of individuals declared being
current cigarette smokers at baseline (Table 2.5). Cigar was the second most
common form of smoking: 8% of individuals who were asked a question on cigar
smoking declared being current cigar smokers at baseline. The least common form
of smoking was pipe: amongst those asked the question, only 5% reported being
current pipe smokers at baseline. Pipe and cigar smokers reported on average a
59
lower number of pack-years, with medians respectively equal to 4 (inter-quartile
range: 2-10) and 5 (2-9), versus 20 (10-32) for cigarette smoking.
2.3.3 Outcomes
A total of 77 out of the 114 studies contributing to this analysis also involved medical
records, autopsy findings and other supplementary sources to help classify deaths
(Table 2.6). In total, 72 studies used standard definitions of myocardial infarction
based on World Health Organization criteria. Out of 114 studies, 111 reported on
ischemic stroke as a stroke subtype and 66 reported diagnosis of strokes on the
basis of typical clinical features and brain imaging. During an average follow-up of
14.2 years; there were 40,218 incident coronary artery disease outcomes (16,390
non-fatal myocardial infarction and 23,828 coronary deaths); 17,445 strokes (5,023
ischaemic stroke and 2,014 haemorrhagic stroke and the rest was unclassified or
information was not provided); 2,108 deaths from heart failures; 1,298 sudden
deaths; 919 deaths from cardiac dysrhytmia; 704 deaths from pulmonary embolism;
1331 deaths from aortic aneurysm and 385 deaths from peripheral vascular disease
(Table 2.7). There were also 9,788 lung cancer deaths, and, in total, 128,137 deaths
from all causes.
2.4 Use of datasets in later Chapters
The dataset described above is used in Chapter 3, 4 and 5. In Chapter 3, the
dataset including all studies with smoking status (114 studies and 929,335
individuals) is further restricted to studies with information on a specific correlate
(blood pressure, lipids, inflammatory markers). To investigate shapes of relationship
between smoking and other factors, subsets with information on pack-years (43
studies, 114,662 individuals) and with information on years since quitting smoking
(54 studies, 80,318 individuals) are restricted to further subsets with information on
specific correlates in Chapter 3. Chapter 4 uses studies which provide information
on smoking status and on the following subsets: amount, duration, starting age,
stopping age and number of years since quitting. Chapter 5 uses studies which
provide information on cigarette, cigar and pipe smoking status. A more detailed
description of the datasets can be found in each Chapter.
60
Figure 2.1: Flow chart of study selection
ERFC – April 2011
135 studies
1,494,051 individuals
Quality control and exclusion of missing data on smoking status
Exclusion of 3 studies based in developing countries (TARFS
based in Turkey, XIAN based in China and PRHHP based in
the Caribbean)
Smoking Status Dataset
114 studies
929,335 individuals
14.2 average years of follow-up
Subsets with information available on:
Number of pack-years (all types): 114,662 individuals;
43 studies
Amount (all types): 300,890 individuals; 84 studies
Duration (all types): 114,978 individuals; 44 studies
Cigar smoking status: 131,816 individuals; 20 studies
Pipe smoking status: 132,060; 22 studies
Starting age (all types): 119,907 individuals; 44 studies
Stopping age (all types): 53,431 individuals; 39 studies
Number of years since quitting: 80,318; 54 studies
ERFC: Emerging Risk Factor Collaboration
61
Table 2.1: Definition of endpoints recorded in The ERFC using the International
Classification of Diseases version 10
Endpoint ICD-10 codes
All cardiovascular* G45, I01, I03-I82, I87, I95-I99, F01, Q20-2Q28, R96
Coronary heart disease (CHD)* I21-I25
Myocardial infarction I21, I22
All cerebrovascular* F01, I60-I69
Ischemic stroke* I63
Hemorrhagic stroke* I61
Subarachnoid stroke* I60
Unclassified stroke†* I64
Other vascular deaths Remainder of cardiovascular disease (fatal)
Cardiac dysrhythmia I47-I49
Hypertensive disease I10-I15
Pulmonary embolism I26
Ill-defined descriptions and complications of the
dearth disease
I51
Sudden death R96
Aortic aneurysm I71
Heart failure I50
Peripheral vascular disease I73-I74, I77-I78
Other Remainder of vascular
All cancer C00-C97, D00-D48
Oral C00-C14
Colorectum C18-C21
Oesophagus C15
Stomach C16
Liver C22
Pancreas C25
Lung C34
Prostate C61
Ovary C56
Bladder C67
Hematological C81-C96
Endocrine & nervous C69-C75
Melanoma C43
Connective tissue C40-C42, C45-C49
Breast (female) C50
Other/unspecified Remainder of cancer/ unspecified to ERFC
All non-cancer, non-vascular A00-A99, B00-B99, D50-D99,E00-E99, F00, F02-F99, G00-G44, G46-G99, H00-H99, I00,
I02, I83-I86, I88-I89, J00-J99, K00-K99, L00-L99, M00-M99, N00-N99, O00-O99, P00-P99,
Q00-Q18, Q30-Q99, S00-S99, T00-T99, U04, V00-V99, W00-W99, X00-X99, Y00-Y99, Z00-
Z99
All external cause S00-S99, T00-T98, U04, V01-V99, W00-W99, X00-X99, Y00-Y98, Z00-Z99
Falls W00-W19
Intentional self-harm X60-X84
Infections A00-A99, B00-B14, B20-B99
Diabetes mellitus E10-E14
Mental disorders F04-F99
Alzheimer's disease and related conditions F00, F02, F03, G30-G32
Liver disease B15-B19, K70-K77
Respiratory system disease J00-J99
Pneumonia J12-J18
COPD and related conditions J40-J47
Digestive system disease (except liver) K00-K69, K78-K93
Renal disease N00-N19
Other/unspecified Remainder of non-cancer, non-vascular/ unspecified to ERFC
Deaths of unknown cause or ill-defined cause R00-R96, R97-R99 and non-vascular deaths defined according to study-specific read-codes
for mortality, and not standard ICD codes.
All-cause mortality A00-Y89
* includes both fatal and non-fatal events. † Unclassified stroke was defined by the ICD
codes stated, or as strokes not specified as ischaemic or hemorrhagic by study-specific
codes. NB: Corresponding ICD-6, 7, 8 or 9 codes were used for studies that recorded
outcomes using earlier ICD versions
62
Table 2.2: Characteristics of individual studies with complete information on at least
smoking status, age and sex
Smoking baseline characteristics
St
ud
y
na
m
e
/s
tu
dy
de
si
gn
C
ou
nt
ry
M
ed
ia
n
fo
llo
w
-u
p
(5
th
&
95
th
pe
rc
en
til
es
)
N
o
w
ith
sm
ok
in
g
st
at
us
A
ge
at
su
rv
ey
(y
rs
)m
ea
n
(s
d)
M
al
e
se
x
N
(%
)
C
ur
re
nt
al
lt
yp
es
sm
ok
in
g
N
(%
)
C
ur
re
nt
pi
pe
or
ci
ga
r
sm
ok
er
s,
%
Sm
ok
in
g
pa
ck
ye
ar
s$
,
M
ed
ia
n
(IQ
R
)
Sm
ok
in
g
ag
e
st
ar
t,
m
ed
ia
n
(IQ
R
)
Sm
ok
in
g
ag
e
st
op
pe
d,
m
ea
n
(s
d)
Sm
ok
in
g
nu
m
be
ry
ea
rs
st
op
pe
d,
m
ed
ia
n
(IQ
R
)
Nested case-control studies
EPICNOR UK 7.5 (3.4 to 9.3) 3334 65 (8) 2110 (63) 352 (11) - - - - -
FIA Sweden 4.0 (0.3 to 9.5) 2684 54 (8) 2161 (81) 633 (24) - - - - -
FLETCHER New Zealand 5.7 (4.9 to 6.4) 706 56 (15) 537 (76) 125 (18) - 18 (10 to 29) 19 (16 to 22) - -
GLOSTRUP Denmark 4.5 (0.5 to 12.5) 390 50 (9) 286 (73) 237 (61) - - - - -
WHIHABPS USA 6.8 (1.2 to 9.3) 1560 68 (6) 0 (0) 91 (6) - - - - -
SUBTOTAL 6.2 (1.0 to 9.3) 8674 61 (11) 5094 (59) 1438 (17) - - - - -
Clinical trials
AFTCAPS USA 5.1 (4.1 to 6.6) 652 57 (7) 540 (83) 652 (100) - - - - -
ALLHAT
USA/Canada/
Puerto Rico/
US Virgin Islands
4.4 (0.4 to 6.7) 28684 66 (8) 14031 (49) 7037 (25) - - - - -
LEADER UK 4.2 (0.6 to 6.1) 471 66 (9) 471 (100) 382 (81) 9 - - - 10 (2 to 21)
MRFIT USA 6.9 (4.4 to 7.8) 12833 47 (6)
12833
(100)
8164 (64) - - - - -
PREVEND Netherlands 7.6 (4.7 to 8.2) 4612 49 (11) 2130 (46) 2381 (52) - - - - -
PROSPER
Scotland/Ireland/
Netherland
3.2 (1.1 to 3.8) 3253 75 (3) 1351 (42) 1083 (33) - - - - -
TPT UK 7.5 (2.8 to 10.5) 7190 55 (7) 7190 (100) 7190 (100) - - - - -
USPHS2 USA 10.9 (4.9 to 11.5) 10707 64 (8)
10707
(100)
507 (5) - - - - -
WHS USA 10.2 (8.4 to 10.8) 27456 55 (7) 0 (0) 3185 (12) - - - - -
WOSCOPS UK 4.8 (2.2 to 5.8) 2717 55 (5) 2717 (100) 2717 (100) - - - - -
SUBTOTAL 7.2 (1.8 to 11.0) 98575 59 (10) 51970 (53) 33298 (34) - - - - -
Cohort studies
ARIC USA 14.0 (5.0 to 15.7) 14584 54 (6) 6294 (43) 3745 (26) - 24 (11 to 38) 19 (16 to 22) 40 (11) 14 (6 to 23)
ATENA Italy 6.7 (5.2 to 8.1) 4740 50 (7) 0 (0) 1900 (40) - 16 (8 to 27) 18 (16 to 24) 41 (10) 8 (3 to 12)
ATS_SAR Italy 8.7 (5.7 to 9.1) 3778 46 (8) 1963 (52) 1201 (32) - 19 (10 to 30) 20 (16 to 24) 43 (8) 7 (3 to 10)
ATTICA Greek 5.0 (0.0 to 5.0) 2113 52 (10) 1076 (51) 897 (42) - 25 (11 to 40) 22 (19 to 29) - -
AUSDIAB Australia 5.0 (4.9 to 8.5) 8260 53 (12) 3619 (44) 1387 (17) 2 - - - 16 (9 to 25)
BHS Australia 26.6 (7.2 to 33.2) 5931 45 (16) 2795 (47) 1857 (31) 5 - - - -
BRHS UK 24.5 (4.7 to 25.4) 6795 50 (6) 6795 (100) 2797 (41) - - - - -
BRUN Italy 20.2 (3.9 to 20.5) 817 58 (11) 398 (49) 198 (24) - - - - -
BUPA UK
23.7 (11.1 to
26.8)
14517 47 (8)
14517
(100)
8232 (57) - - - - 6 (1 to 10)
BWHHS UK 7.3 (3.1 to 8.4) 2789 68 (5) 0 (0) 321 (12) 0 17 (8 to 31) 18 (17 to 22) 46 (13) 22 (13 to 33)
CAPS UK 13.0 (4.0 to 13.0) 2129 52 (5) 2129 (100) 1180 (55) 27 - - - 4 (4 to 5)
CASTEL Italy 11.1 (2.0 to 14.0) 330 72 (4) 245 (74) 330 (100) - 20 (10 to 30) 21 (17 to 31)
CHA USA
32.0 (11.6 to
35.6)
34250 41 (13) 19894 (58) 14243 (42) - - - - -
CHARL USA 24.0 (3.4 to 40.0) 2028 50 (11) 934 (46) 1114 (55) - - - - -
CHS1 USA 12.1 (2.0 to 12.9) 3778 72 (5) 1437 (38) 452 (12) - - 18 (16 to 21) 51 (14) 20 (10 to 31)
CHS2 USA 9.1 (1.9 to 9.5) 462 72 (5) 173 (37) 76 (16) - - - - -
COPEN Denmark 13.2 (2.6 to 14.9) 6166 57 (14) 2525 (41) 4000 (65) 0 - - - -
DISCO Italy 5.5 (5.5 to 9.5) 1825 50 (11) 813 (45) 512 (28) 19 (8 to 34) 18 (15 to 22) 47 (12) 7 (3 to 10)
DRECE Spain
16.4 (15.5 to
16.6)
2748 41 (11) 1338 (49) 1098 (40) 5 (2 to 7) 22 (15 to 30) 37 (11) 7 (3 to 14)
DUBBO Australia 14.1 (2.0 to 15.0) 1523 68 (7) 537 (35) 323 (21) 0 15 (6 to 30)
EAS Scotland 15.2 (2.8 to 15.8) 1036 64 (6) 515 (50) 243 (23) 5 21 (10 to 33) 20 (17 to 25)
EMOFRI Italy 6.8 (6.5 to 7.2) 360 55 (6) 176 (49) 92 (26) 19 (9 to 32) 20 (18 to 23) 41 (10) 15 (6 to 20)
ESTHER Germany 5.0 (0.0 to 6.1) 1286 59 (6) 622 (48) 1286 (100) - - - - -
FINE_FIN Finland 5.3 (0.5 to 10.0) 40 75 (4) 40 (100) 40 (100) - - - - -
FINE_IT Italy 9.9 (1.9 to 21.4) 459 72 (4) 459 (100) 123 (27) - - - 10 (5 to 10)
FINRISK92 Finland 16.9 (7.9 to 16.9) 5737 44 (11) 2647 (46) 1761 (31) 0 - - - 15 (9 to 21)
FINRISK97 Finland 11.8 (6.7 to 11.9) 6306 49 (12) 3059 (49) 1879 (30) 0 - - - 15 (9 to 24)
63
FRAMOFF USA 5.0 (2.0 to 6.9) 335 57 (9) 147 (44) 335 (100) - - - - -
FUNAGATA Japan 7.3 (5.3 to 10.2) 1125 53 (12) 504 (45) 331 (29) - - - - -
GOH Israel
35.0 (12.9 to
36.0)
3874 43 (8) 1957 (51) 1693 (44) - - - - -
GOTO13 Sweden 21.0 (4.0 to 30.5) 404 54 (0) 404 (100) 404 (100) - - - - -
GOTO33 Sweden 12.8 (5.7 to 13.1) 720 51 (0) 720 (100) 264 (37) - - - - -
GOTO43 Sweden 11.0 (7.9 to 11.7) 776 50 (0) 776 (100) 240 (31) - - - - -
GOTOW Sweden
32.2 (10.6 to
32.7)
1425 47 (6) 0 (0) 582 (41) - 10 (5 to 16) 21 (18 to 26) 38 (9) 10 (3 to 17)
GREPCO Italy 7.9 (7.7 to 8.4) 779 44 (8) 0 (0) 319 (41) - 8 (4 to 17)
25 (20 to
30.5)
- -
GRIPS Germany 9.8 (4.9 to 10.0) 5320 48 (5) 5320 (100) 2179 (41) - 26 (19 to 35) 18 (17 to 20 - -
GUBBIO Italy 8.4 (5.8 to 9.4) 3408 55 (13) 1515 (44) 1161 (34) - - - - -
HBS Finland 20.5 (5.0 to 20.5) 1302 60 (4) 1302 (100) 227 (17) - - - - -
HELSINAG Finland 9.3 (2.2 to 11.0) 385 79 (4) 100 (26) 35 (9) - - - - -
HISAYAMA Japan 14.0 (3.2 to 14.0) 2576 59 (12) 1088 (42) 349 (14) - 32 (18 to 47) 20 (20 to 23) 57 (11) 1 (0 to 1)
HONOL USA 6.3 (1.5 to 7.6) 2436 78 (4) 2436 (100) 184 (8) - - - - -
HOORN Netherlands 8.8 (3.2 to 9.9) 2228 61 (7) 983 (44) 706 (32) - - - - -
HPFS1 USA 20.2 (6.2 to 21.9) 45857 54 (10)
45857
(100)
4562 (10) - - - - -
IKNS Japan 11.1 (5.1 to 18.6) 8038 58 (10) 3299 (41) 1911 (24) - 32 (19 to 45) 20 (20 to 21) - -
ISRAEL Israel 23.3 (7.9 to 23.9) 7698 49 (7) 7698 (100) 4111 (53) 5 - - - -
KARELIA Finland 36.7 (6.5 to 36.9) 9735 41 (10) 4638 (48) 3271 (34) 0 - - - 1 (1 to 1)
KIHD Finland 19.0 (1.8 to 23.9) 632 52 (6) 632 (100) 632 (100) - - -
LASA Netherlands 9.9 (1.5 to 10.5) 1236 69 (8) 549 (44) 285 (23) 21 - - - 20 (10 to 31)
MALMO Sweden 18.2 (8.2 to 22.6) 31675 45 (7) 21913 (69) 14413 (46) - - - - -
MATISS83 Italy 18.7 (6.8 to 19.5) 2552 51 (10) 1198 (47) 759 (30) - 22 (11 to 37) 18 (15 to 20) 44 (12) 8 (3 to 15)
MATISS87 Italy 15.6 (7.2 to 16.2) 2019 52 (9) 895 (44) 443 (22) - 23 (12 to 36) 18 (15 to 20) 42 (12) 10 (3 to 18)
MATISS93 Italy 8.3 (7.1 to 9.3) 1214 49 (9) 587 (48) 325 (27) - 22 (10 to 36) 18 (15 to 20) 41 (11) 7 (2 to 13)
MCVDRFP Netherlands
16.7 (10.5 to
18.9)
23522 42 (10) 10946 (47) 10329 (44) 4 14 (6 to 25) 17 (15 to 20) 33 (10) 8 (3 to 15)
MESA USA 4.8 (2.0 to 5.2) 3576 62 (10) 2092 (59) 1017 (28) 3 17 (6 to 34) 18 (15 to 20) 42 (14) 21 (11 to 31)
MICOL Italy 5.9 (4.5 to 7.1) 18858 51 (10) 10650 (56) 6164 (33) - 18 (9 to 30) 21 (18 to 28) - -
MOGERAUG1 Germany 13.0 (3.6 to 13.4) 871 54 (6) 871 (100) 294 (34) - - - - -
MOGERAUG2 Germany 7.9 (2.3 to 8.4) 3962 53 (12) 1949 (49) 969 (24) - - - - -
MOGERAUG3 Germany 3.0 (1.8 to 3.6) 3373 55 (10) 1664 (49) 747 (22) - - - - -
MONFRI86 Italy 16.7 (7.7 to 16.9) 1407 49 (9) 690 (49) 447 (32) - 17 (8 to 26) 19 (16 to 21) 40 (11) 7 (2 to 14)
MONFRI89 Italy 13.6 (7.5 to 13.7) 1341 49 (8) 666 (50) 380 (28) - 17 (8 to 29) 18 (16 to 20) 39 (10) 8 (3 to 13)
MONFRI94 Italy 8.5 (8.0 to 8.8) 1293 49 (8) 629 (49) 349 (27) - 15 (7 to 27) 19 (16 to 20) 38 (10) 10 (5 to 16)
MONICA Italy 6.5 (2.1 to 10.5) 3639 49 (9) 1825 (50) 1242 (34) - 18 (7 to 32) 18 (15 to 22) 47 (10) 7 (3 to 10)
MORGEN Netherlands 10.8 (8.3 to 13.0) 18246 46 (9) 8307 (46) 6826 (37) 6 15 (7 to 26) 17 (15 to 18) 34 (10) 12 (5 to 19)
MOSWEGOT Sweden 13.9 (7.6 to 19.6) 4158 47 (11) 1973 (47) 1222 (29) - - - - -
MRCOLD UK 8.7 (1.2 to 11.7) 10137 80 (4) 3825 (38) 1222 (12) - 12 (5 to 27) - 50 (16) 28 (16 to 43)
NCS1 Norway
16.1 (14.4 to
16.7)
23996 42 (4) 11815 (49) 10467 (44) 20 - - 40 (4) 5 (3 to 5)
NCS2 Norway
17.3 (15.6 to
17.8)
12743 42 (4) 6491 (51) 5070 (40) 8 - - 40 (5) 5 (3 to 5)
NCS3 Norway
18.2 (12.2 to
18.8)
9658 42 (4) 4994 (52) 5485 (57) 5 - - 40 (5) 3 (3 to 5)
NFR Italy 10.2 (6.1 to 11.2) 3075 55 (5) 3075 (100) 1564 (51) - 25 (14 to 38) 20 (17 to 24) 51 (6) 7 (3 to 10)
NHANESI USA 19.0 (3.7 to 21.0) 6629 51 (15) 2456 (37) 1402 (21) 5 22 (11 to 39) 18 (16 to 21) 38 (14) 8 (4 to 15)
NHANESIII USA 14.3 (3.5 to 17.7) 7357 51 (17) 4339 (59) 3979 (54) 3 8 (2 to 19) 20 (17 to 29) 41 (16) 13 (5 to 23)
NHS1 USA
28.6 (12.0 to
30.3)
117887 43 (7) 0 (0) 38884 (33) - - - - -
NPHSI UK 14.6 (4.4 to 18.6) 1392 52 (7) 1392 (100) 704 (51) 54 - - - 9 (4 to 17)
NPHSII UK 8.3 (3.5 to 10.4) 1941 57 (3) 1941 (100) 1101 (57) 30 - - - -
NSHS Canada 9.7 (4.0 to 10.0) 1168 53 (15) 614 (53) 411 (35) - 24 (11 to 37) 17 (15 to 20) 51 (13) 1 (1 to 1)
OB43 Italy 7.5 (5.1 to 9.1) 3531 47 (8) 1707 (48) 1264 (36) - 16 (7 to 30) 21 (17 to 27) 43 (9) 7 (3 to 10)
OSAKA Japan 10.2 (3.9 to 18.8) 12380 52 (10) 8415 (68) 4795 (39) - 29 (17 to 42) 20 (20 to 20) - -
OSLO Norway
29.5 (10.7 to
30.5)
16832 44 (6)
16832
(100)
9600 (57) 40 - - 41 (6) 5 (3 to 5)
OYABE Japan 10.4 (4.2 to 10.6) 973 57 (12) 873 (90) 972 (100) - - - - -
PARIS1 France 22.9 (7.0 to 26.1) 5959 47 (2) 5959 (100) 4771 (80) 12 - - - 6 (2 to 15)
PRIME France / NI 5.2 (5.0 to 7.3) 9486 55 (3) 9486 (100) 2544 (27) 5 - - - 14 (6 to 20)
PROCAM Germany 9.8 (3.8 to 18.9) 12790 44 (10) 10333 (81) 7878 (62) 5 18 (10 to 28) 18 (17 to 21) 35 (9) 10 (4 to 15)
QUEBEC Canada 26.4 (4.2 to 26.9) 1798 46 (7) 1798 (100) 1429 (79) 22 37 (25 to 50) 16 (15 to 19) 38 (10) 6 (2 to 12)
RANCHO USA 14.2 (2.0 to 18.1) 1819 68 (11) 753 (41) 250 (14) - 20 (8 to 40) 22 (18 to 30) 48 (13) 19 (11 to 28)
REYK Iceland 24.7 (6.3 to 37.1) 16760 52 (9) 8028 (48) 7944 (47) 20 20 (12 to 31) 22 (17 to 22) 38 (10) 14 (6 to 22)
64
RF2 Italy
13.7 (11.3 to
14.1)
5392 44 (9) 2536 (47) 1979 (37) - 15 (6 to 28) 20 (16 to 25) 41 (9) 7 (3 to 10)
ROTT Netherlands 12.0 (3.3 to 14.3) 2279 68 (9) 550 (24) 892 (39) - - - - -
SHHEC UK 10.0 (6.3 to 10.0) 13402 49 (8) 6533 (49) 5962 (44) 10 27 (18 to 38) 17 (15 to 19) 38 (11) 11 (4 to 19)
SHS USA 12.4 (2.0 to 14.3) 4132 56 (8) 1613 (39) 1371 (33) - - - - -
SPEED UK 16.7 (3.3 to 18.2) 2114 55 (4) 2114 (100) 1011 (48) 18 - - 50 (5) 10 (7 to 10)
TOYAMA Japan 12.7 (7.8 to 12.8) 4524 46 (7) 2908 (64) 1752 (39) - - - - -
TROMSØ Norway
18.9 (11.4 to
19.3)
13665 39 (10) 6646 (49) 6610 (48) 8 (4 to 15) 18 (16 to 22) - 10 (6 to 15)
ULSAM Sweden 26.5 (6.0 to 37.6) 1629 50 (2) 1629 (100) 1167 (72) 71 13 (7 to 23) 20 (18 to 21) 38 (9) 8 (3 to 10)
VHMPP Austria 13.1 (2.2 to 16.7) 120612 48 (14) 55101 (46) 22370 (19) - - - - -
VITA Italy 3.3 (1.7 to 5.3) 2353 50 (8) 1205 (51) 2353 (100) 0 - - - 25 (14 to 36)
WHITEI UK 8.2 (2.0 to 8.4) 3993 76 (5) 3993 (100) 554 (14) - - - - -
WHITEII UK 12.4 (4.9 to 14.1) 10174 45 (6) 6787 (67) 2624 (26) 16 21 (12 to 31) 18 (16 to 20) - -
ZARAGOZA Spain 5.1 (4.1 to 5.1) 2601 60 (12) 1101 (42) 478 (18) - - - - 2 (1 to 3)
ZUTE Netherlands 7.0 (0.9 to 10.1) 123 75 (4) 123 (100) 99 (80) - 30 (16 to 44) 17 (15 to 19) 73 (4) -
SUBTOTAL 15.7 (4.1 to 31.3) 822086 48 (12)
419745
(51)
281952 (34) - - - - -
TOTAL 14.2 (3.3 to 30.4) 929,335 50 (12)
476809
(51)
316688 (34) 10 18 (8 to 31) 18 (16 to 22) 39 (12) 9 (3 to 18)
No: Number of individuals; -: Not available; $: smoking pack-years in current and ex-smokers
(excluding never smokers); IQR: Inter-quartile range; SD: standard deviation from the mean.
Means and SD were computed as weighted average of study-specific means and SD.
65
Table 2.3: Studies providing information on smoking type
Cohort Type of smoking
abbreviation Cigarettes Pipe or cigars Pipes only Cigars only
AFTCAPS º • • •
ALLHAT • • • •
ARIC º • • •
ATENA º • • •
ATS_SAR º • • •
ATTICA º • • •
AUSDIAB º º º º
BHS º º • •
BRHS º • • •
BRUN • • • •
BUPA º º º º
BWHHS º • • •
CAPS º º º º
CASTEL º • • •
CHA º • • •
CHARL • • • •
CHS1 º • • •
CHS2 º • • •
COPEN • • • •
DISCO º • • •
DRECE º • • •
DUBBO • • • •
EAS º º • •
EMOFRI º • • •
EPICNOR º • • •
ESTHER º • • •
FIA • • • •
FINE_FIN º • • •
FINE_IT • • • •
FINRISK92 • • • •
FINRISK97 • • • •
FLETCHER º • • •
FRAMOFF º • • •
FUNAGATA • • • •
GLOSTRUP • • • •
GOH • • • •
GOTO13 º º • •
GOTO33 • • • •
GOTO43 • • • •
GOTOW º • • •
GREPCO º • • •
GRIPS º • • •
GUBBIO º • • •
HBS • • • •
HELSINAG • • • •
HISAYAMA º • • •
HONOL • • • •
HOORN º • • •
HPFS1 º • • •
IKNS º • • •
ISRAEL º º • •
KARELIA • • • •
KIHD • • • •
LASA • º º º
LEADER º º º º
MALMO º • • •
MATISS83 º • • •
MATISS87 º • • •
MATISS93 º • • •
MCVDRFP º º º º
MESA º º º º
MICOL º • • •
MOGERAUG1 º • • •
MOGERAUG2 º • • •
MOGERAUG3 º • • •
MONFRI86 º • • •
MONFRI89 º • • •
MONFRI94 º • • •
MONICA º • • •
MORGEN º º • •
MOSWEGOT º • • •
MRCOLD º • • •
MRFIT º • • •
NCS1 º º º º
NCS2 º º º º
NCS3 º º º º
NFR º • • •
NHANESI º º º º
NHANESIII º º º º
NHS1 º • • •
NPHSI º º º º
NPHSII • º º •
NSHS º • • •
OB43 º • • •
OSAKA º • • •
OSLO º º º º
OYABE º • • •
PARIS1 º º º º
PREVEND º • • •
PRIME º º º º
PROCAM º º • •
PROSPER • • • •
QUEBEC º º º º
RANCHO º • • •
REYK º º º º
RF2 º • • •
66
ROTT • • • •
SHHEC º º º º
SHS º • • •
SPEED º º • •
TOYAMA • • • •
TPT • • • •
TROMSØ º º º •
ULSAM º º º º
USPHS2 • • • •
VHMPP • • • •
VITA • • • •
WHIHABPS • • • •
WHITEI º º • •
WHITEII º º • •
WHS º • • •
WOSCOPS • • • •
ZARAGOZA • • • •
ZUTE º º • •
•: Information not collected by study investigators or collected by study investigators but not
shared with THE ERFC. º: Information provided
67
Table 2.4: Smoking intensity and demographics at baseline
Smoking status
Characteristics
No of
studie
s
No of
individuals
Mean (SD) in
current
smokers
Mean (SD) in
ex-smokers
All types of smoking combined
Smoking pack-years 43 114,662 21 (6.8) 22 (19.4)
Duration of smoking (yrs) 44 114,978 30.2 (10) 21.8 (10.8)
Amount smoked (cpd) 84 300,890 16.5 (14.2) 17.6 (11.6)
Age at smoking initiation
(yrs) 44 119,907 21 (6.8) 19.7 (5.4)
Age at smoking cessation (yrs) 39 53,431 - 42.9 (11.2)
Time since quitting smoking
(yrs) 54 80,318 - 11.2 (9.5)
No: Number of individuals excluding null values (for example never smokers will have null
value for amount smoked); cpd: cigarettes equivalent per day. No: Number of individuals; -:
Not available; $: smoking pack-years in current and ex-smokers (excluding never smokers);
IQR: Inter-quartile range; SD: standard deviation from the mean. Means and SD were
computed as weighted average of study-specific means and SD.
68
Table 2.5: Smoking status at baseline
Smoking status, %
No
studies
No
subjects
Current
smoker Ex-smoker
Never
smoker Other *
All types of smoking
combined 114 929,335 34% 23% 43% 0%
Cigarette smoking 83 689,104 34% 25% 37% 4%
Cigar or pipe smoking 33 178,566 11% 4% 45% 40%
Cigar smoking only 20 131,816 8% 3% 43% 46%
Pipe smoking only 22 132,060 5% 3% 43% 49%
No: Number. *: Other includes individuals who were classified as “non current” (meaning ex
or never) or “users” (meaning ex or current) without the possibility to classify them further as
either current, ex or never. A total of 19 studies, including 110,456 individuals who were
current or ex smokers, had information on the three smoking types. Some studies only
provided information for “cigar or pipe smoking” without further precision regarding the type of
smoking.
69
Table 2.6: Endpoints characteristics
Study Definition of incident endpoint Classification of incident endpoints
Death Non-fatal MI Non-fatal stroke MI Stroke
C
lin
ic
al
fe
at
ur
e
EC
G
C
ar
di
ac
en
zy
m
es
C
lin
ic
al
fe
at
ur
es
C
T/
M
R
I
im
ag
in
g
D
ef
in
ite
Pr
ob
ab
le
Si
le
nt
Is
ch
em
ic
H
em
or
rh
ag
ic
SA
H
U
nc
la
ss
ifi
ed
AFTCAPS ** √ √ √ √ √ √ - √NC√ √ √ √ 
ALLHAT ** √ √ √ √ √ √ o o √ NC √ NC √ NC √ 
ARIC ** √ √ √ √ √ √ √ √ √ √ √ √ 
ATENA† ** √ √ √ √ √ √ √ √ √ √ √ √ 
ATS_SAR * NA NA NA NA NA √ √ NC o √ √ √ √ 
ATTICA ** NA NA NA NA NA NS NS NS o o o o
AUSDIAB * NS NS NS NS NS NS NS NS √ √ √ √ 
BHS * NA NA NA NA NA √ o o √ √ √ √ 
BRHS * √ √ √ NS NS √ o o √ √ √ √ 
BRUN ** √ √ √ √ √ √ o o √ √ o o
BUPA * NA NA NA NA NA √ o o √ √ √ √ 
BWHHS ** √ √ √ √ √ √ o o √ √ √ √ 
CAPS ** √ √ o √ √ √ √ o √ √ √ √ 
CASTEL ** NA NA NA NA NA √ o o √ NC √ NC √ NC √ 
CHA * NA NA NA NA NA NS NS NS NS NS NS NS
CHARL ** √ √ o √ o √ √ o √ √ √ √ 
CHS 1 ** √ √ √ √ √ √ √ NC √ √ √ o √ 
CHS 2 ** √ √ √ √ √ √ √ NC √ √ √ o √ 
COPEN ** √ √ √ √ √ √ o o √ √ √ √ 
DISCO‡ * NA NA NA NA NA √ √ NC o √ √ √ √ 
DRECE * NA NA NA NA NA √ o o NS NS NS √ 
DUBBO ** √ √ √ √ √ √ NS o √ √ √ √ 
EAS ** √ √ √ √ √ √ √  √ √ √ √ √ 
EMOFRI† ** √ √ √ √ √ √ √ √ √ √ √ √ 
EPIC- * √ √ √ √ √ √ o o √ NC √ NC √ NC √ 
ESTHER * NS NS NS NS NS NS NS NS NS NS NS NS
FIA ** √ √ √ NA NA √ o o √ NC √ NC √ NC √ 
FINE_FIN ** √ √ √ √ NA √ NS √ √ √ √ √ 
FINE_IT ** √ √ √ √ NA √ NS √ √ √ √ √ 
FINRISK9 ** √ √ √ √ √ √ o o √ √ √ √ 
FINRISK9 ** √ √ √ √ √ √ o o √ √ √ √ 
FLETCHE * √ √ √ √ √ √ o o √ NC √ NC √ NC √ 
FRAMOF ** √ √ √ √ √ √ o √ √ √ √ √ 
FUNAGA NS NS NS NS NS NS NS NS NS NS NS NS NS
GLOSTR ** √ √ √ NA NA o o o √ NC √ NC √ NC √ 
GOH ** NA NA NA NA NA √ √ NC o √ √ √ √ 
GOTO13 ** √ √ √ √ √ √ o o √ NC √ NC √ NC √ 
GOTO33 ** √ √ √ √ √ √ o o √ NC √ NC √ NC √ 
GOTO43 ** √ √ √ √ √ √ o o √ √ √ √ 
GOTOW * √ √ √ √ √ √ o o NS NS NS √ 
GREPCO‡* NA NA NA NA NA √ √ NC o √ √ √ √ 
GRIPS ** √ √ √ √ √ √ √ o √ √ √ √ 
GUBBIO‡ * NA NA NA NA NA √ √ NC o √ √ √ √ 
HBS ** NA NA NA NA NA √ o NS √ NC √ NC √ NC √ 
HELSINA ** NA NA NA NA NA √ o o √ √ √ √ 
HISAYAM ** √ √ √ √ √ √ √ √ √ √ √ √ 
HONOL ** √ √ √ √ √ √ √ NC √ √ √ √ √ 
HOORN * √ √ √ √ o √ o o √  √  √  √  
HPFS ** NA NA NA NA NA √ √ NC √ √  √  √  √  
IKNS ** √ √ √ √ √ √ √ o √ √ √ √ 
ISRAEL ** NA NA NA NA NA √ NC o o √ NC √ NC √ NC √ 
KARELIA ** √ √ √ √ √ √ o o √ √ √ √ 
KIHD ** √ √ √ √ √ √ √ NC o √ √ √ √ 
LASA * NS NS NS √ √ √ √ o o o o √ 
LEADER ** √ √ √ √ √ √ o o √ √ √ √ 
MALMO ** √ √ √ √ √ √ o o √ √ √ √ 
MATISS8 ** √ √ √ √ √ √ √ √ √ √ √ √ 
MATISS8 ** √ √ √ √ √ √ √ √ √ √ √ √ 
MATISS9 ** √ √ √ √ √ √ √ √ √ √ √ √ 
MCVDRF NS NA NA NA NA NA NS NS NS √ √ √ √ 
MESA ** √ √ ? √ √ √ √ √ √ √ √ √ 
70
MICOL‡ * NA NA NA NA NA √ √ NC o √ √ √ √ 
MOGERA ** √ √  √  NA NA √ √ NC o √ √ √ √ 
MOGERA ** √ √  √  NA NA √ √ NC o √ √ √ √ 
MOGERA ** √ √  √  NA NA √ √ NC o √ √ √ √ 
MONFRI8 ** √ √ √ √ √ √ √ √ √ √ √ √ 
MONFRI8 ** √ √ √ √ √ √ √ √ √ √ √ √ 
MONFRI9 ** √ √ √ √ √ √ √ √ √ √ √ √ 
MONICA‡ * NA NA NA NA NA √ √ NC o √ √ √ √ 
MORGEN * NA NA NA NA NA NS NS NS √ √ √ √ 
MOSWEG** √ √ √ √ √ √ o o √ √ √ √ 
MRCOLD * NA NA NA NA` NA NA NA NA √ √ √ √ 
MRFIT ** √ √ √ √ √ √ o √ √ √ √ √ 
NCS1 * NA NA NA NA NA √  o o √ √ √ √ 
NCS2 * NA NA NA NA NA √  o o √ √ √ √ 
NCS3 * NA NA NA NA NA √  o o √ √ √ √ 
NFR‡ * NA NA NA NA NA √ √ NC o √ √ √ √ 
NHANES1* √ √ √ √ √ √ o o √ √ √ √ 
NHANES3* NA NA NA NA NA √ NC o o √ NC √ NC √ NC √  
NHS ** NA NA NA NA NA √ √ NC o √  √  √  √  
NPHSI ** √ √ √ √ √ √ √ NC √ √ NC √ NC √ NC √ 
NPHSII ** √ √ √ √ √ √ √ NC √ √ √ √ √ 
NSHS ** √ √ √ √ √ √ o o √ √ √ √ 
OB43‡ * NA NA NA NA NA √ √ NC o √ √ √ √ 
OSAKA ** √ √ √ √ √ √ √ NC o √ √ √ √ 
OSLO * NA NA NA NA NA √ o o √ √ √ √ 
OYABE ** NA NA NA √ √ √ o o √ √ √ √ 
PARIS1 ** NA NA NA NA NA √ NC o o o o o √ 
PREVEN ** √ √ √ √ √ NS NS NS √ √ √ √ 
PRIME ** √ √ √ √ √ √ o o √ √ √ √ 
PROCAM ** √ √ √ √ √ √ √NC √ √ √ o √ 
PROSPE ** √ √ √ √ √ √ √ o o o o √ 
QUEBEC ** √ √ √ √ √ √ o √ o o o √ 
RANCHO * √ √ √ √ √ √ o o √ √ √ √ 
REYK ** √ √ √ √ √ √ √ o √ √ √ √ 
RF2‡ * NA NA NA NA NA √ √ NC o √ √ √ √ 
ROTT ** √ √ √ NA NA √ √ o √ √ √ √ 
SHHEC ** √ √ √ √ √ √ √ o √ √ √ √ 
SHS ** √ √ √ √ √ √ √ o √ √ √ √ 
SPEED ** √ √ √ √ √ √ o √ √ √ √ √ 
TOYAMA ** √ √ √ √ √ NS NS NS √ √ √ √ 
TPT ** √ √ √ √ √ √ √ NC NS √ o o √ 
TROMSO * √ √ √ √ √ √ √ √ √ √ √ √ 
ULSAM ** √ √ √ √ √ √ o o √ √ √ √ 
USPHS2 NS NS NS NS NS NS NS NS NS NS NS NS NS
VHMPP * NA NA NA NA NA √ o o √ √ √ √ 
VITA ** √ √ √ √ √ √ √NC o √ √ √ √ 
WHIHABP** NC NC NC √ √ √ o o √  √ NC √ NC √ 
WHITE I * NA NA NA NA NA NS NS NS √ √ √ √ 
WHITE II * √ √ √ √ √ √ o o √ √ √ √ 
WHS ** √ √ √ √ √ √ o o √ √ √ √ 
WOSCOP ** √ √ √ √ √ √ √ √ o o o √ 
ZARAGO ** √ √ √ NS NS √ √NC o √ o o √ 
ZUTE ** √ √ √ √ √ √ o o √ √ √ √ 
–: Not recorded; +: Self-report only; ++: Self-report supplemented by objective criteria (e.g.,
Electrocardiogram, Physical examination). * Death certificate only; ** Death certificate
supplemented by medical record. SAH: Subarachnoid haemorrhage; NS: Not stated. NC =
reportedly measured but data not contributed to the ERFC; NA = not applicable, where
cohorts contributed data on fatal endpoints only. 0: Feature not included in criteria; √: Feature 
included in criteria.
71
Table 2.7: Summary of endpoints for each study with information on smoking status,
age and sex
Cardiovascular outcomes
Cohort
abbreviation
A
ll
ca
rd
io
va
sc
ul
ar
C
H
D
de
at
h
an
d
no
n
fa
ta
lM
I
M
yo
ca
rd
ia
lI
nf
ar
ct
io
n
(n
on
fa
ta
l)
A
ll
C
H
D
(fa
ta
l)
A
ll
ce
re
br
ov
as
cu
la
r
(fa
ta
la
nd
no
n
fa
ta
l)
Is
ch
ae
m
ic
st
ro
ke
(fa
ta
l
an
d
no
n
fa
ta
l)
H
ae
m
or
rh
ag
ic
st
ro
ke
(fa
ta
la
nd
no
n
fa
ta
l)
Su
ba
ra
ch
no
id
ha
em
or
rh
ag
e
(fa
ta
l)
H
ea
rt
fa
ilu
re
(fa
ta
l)
Su
dd
en
de
at
h
(fa
ta
l)
C
ar
di
ac
dy
sr
hy
tm
ia
(fa
ta
l)
Pu
lm
on
ar
y
em
bo
lis
m
(fa
ta
l)
A
or
tic
an
eu
ry
sm
(fa
ta
l)
Pe
rip
he
ra
lv
as
cu
la
r
di
se
as
e
(fa
ta
l)
Lu
ng
ca
nc
er
(fa
ta
l)
A
ll
ca
us
es
of
de
at
h
AFTCAPS 23 18 17 1 2 2 0 0 0 2 0 0 0 0 0 8
ALLHAT 1666 1124 1119 5 542 0 0 0 0 0 0 0 0 0 0 11
ARIC 1637 872 672 200 562 453 56 33 13 0 21 12 13 7 228 1507
ATENA 30 18 17 1 4 1 2 1 0 0 2 0 1 0 4 40
ATS_SAR 32 19 0 19 8 0 1 1 0 0 0 0 1 0 13 113
ATTICA 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50
AUSDIAB 127 77 41 36 38 12 2 4 2 0 2 1 3 2 21 269
BHS 914 509 0 509 217 22 21 4 37 0 11 9 37 9 84 1932
BRHS 1854 1213 674 539 514 7 13 10 11 0 7 12 50 8 246 1954
BRUN 151 66 28 38 63 43 19 0 9 0 0 7 5 0 0 240
BUPA 1073 723 0 723 176 22 28 8 24 0 2 17 69 2 230 2495
BWHHS 198 91 78 13 91 0 1 0 1 0 0 3 4 2 17 217
CAPS 289 249 138 111 18 3 3 1 1 0 0 7 3 1 46 340
CASTEL 73 16 3 13 15 0 0 0 28 10 0 4 0 0 0 175
CHA 4820 3000 0 3000 786 116 154 39 174 3 119 62 88 32 940 11613
CHARL 967 549 256 293 257 29 34 5 13 0 32 10 11 5 88 1205
CHS1 1063 569 366 203 442 345 62 0 0 0 0 0 0 0 95 1045
CHS2 107 53 31 22 48 39 5 0 0 0 0 0 0 0 15 89
COPEN 1013 382 348 34 430 265 53 14 46 7 8 20 12 0 122 1279
DISCO 10 8 0 8 2 0 1 0 0 0 0 0 0 0 3 26
DRECE 29 15 0 15 6 0 1 0 2 0 0 1 0 0 13 132
DUBBO 386 196 153 43 144 57 14 2 19 0 1 4 3 0 18 338
EAS 169 82 41 41 68 0 3 2 3 0 1 1 4 0 33 284
EMOFRI 8 2 2 0 3 2 1 0 0 0 3 0 0 0 0 9
EPICNOR 503 484 257 227 0 0 0 0 0 0 0 0 0 0 0 353
ESTHER 52 18 16 2 31 1 0 0 0 0 0 0 2 0 7 28
FIA 584 584 448 136 0 0 0 0 0 0 0 0 0 0 0 136
FINE_FIN 10 5 2 3 5 1 0 0 0 0 0 0 0 0 9 24
FINE_IT 208 66 18 48 103 4 5 0 19 2 1 0 1 1 17 328
FINRISK92 323 162 123 39 136 85 39 3 6 2 1 6 1 0 18 268
FINRISK97 219 103 75 28 91 64 16 1 9 0 0 9 1 0 10 190
FLETCHER 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
FRAMOFF 23 18 17 1 5 5 0 0 0 0 0 0 0 0 0 22
FUNAGATA 49 13 6 7 33 17 6 1 0 0 0 0 1 1 4 50
GLOSTRUP 76 76 58 18 0 0 0 0 0 0 0 0 0 0 0 18
GOH 437 201 0 201 99 2 8 2 14 12 50 6 2 3 29 1422
GOTO13 208 131 130 1 59 0 1 0 9 2 0 3 1 0 17 143
GOTO33 44 27 14 13 8 0 0 0 0 0 2 2 0 0 7 81
GOTO43 48 29 28 1 17 12 1 2 0 0 0 0 1 0 3 25
GOTOW 369 146 94 52 178 2 0 0 1 0 2 0 1 1 14 407
GREPCO 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 5
GRIPS 409 272 272 0 95 0 0 0 8 15 0 12 3 0 28 181
GUBBIO 107 69 0 69 29 11 2 1 0 0 0 1 0 1 12 239
HBS 135 91 0 91 27 0 0 0 0 0 0 0 0 0 0 417
HELSINAG 96 41 0 41 32 17 2 0 1 0 2 7 3 2 4 203
HISAYAMA 356 77 67 10 220 148 49 21 2 0 1 2 12 2 30 387
HONOL 298 151 110 41 124 12 39 2 0 0 3 1 7 1 38 493
HOORN 170 73 60 13 53 3 4 0 11 13 11 3 4 0 19 211
HPFS1 4259 2492 0 2492 709 97 127 36 198 341 55 41 141 11 745 12290
IKNS 495 84 37 47 344 158 71 25 57 0 2 1 3 0 18 757
ISRAEL 987 723 0 723 264 0 0 0 0 0 0 0 0 0 0 2531
KARELIA 2970 1854 1303 551 882 64 44 34 80 3 13 35 17 2 142 2357
KIHD 219 154 151 3 48 34 10 2 2 0 0 0 2 1 30 162
LASA 34 22 22 0 12 0 0 0 0 0 0 0 0 0 0 333
LEADER 93 53 20 33 33 24 2 0 2 0 1 1 1 2 16 93
MALMO 2413 2043 1229 814 143 36 49 21 18 1 6 17 46 3 334 3270
MATISS83 335 83 47 36 99 26 10 3 53 0 71 1 0 0 12 410
MATISS87 166 43 21 22 55 9 7 2 26 1 34 0 0 0 7 200
MATISS93 31 14 11 3 7 1 2 1 5 0 4 0 0 0 0 29
MCVDRFP 456 196 0 196 97 15 31 14 27 8 19 8 16 4 247 1775
MESA 105 59 47 12 41 30 9 1 0 0 0 0 0 0 0 102
MICOL 147 102 0 102 33 7 3 0 0 0 0 3 3 2 75 507
MOGERAUG1 108 79 47 32 5 0 2 0 10 2 0 5 2 1 8 126
MOGERAUG2 129 104 63 41 7 1 1 1 8 1 0 4 0 0 16 199
MOGERAUG3 36 18 11 7 5 2 1 0 3 0 0 2 0 0 2 54
MONFRI86 107 28 20 8 26 14 5 2 4 1 43 2 1 0 5 166
MONFRI89 82 28 22 6 20 10 5 0 6 0 23 2 1 0 1 100
MONFRI94 39 10 10 0 17 6 7 1 0 1 9 0 1 0 0 40
MONICA 38 28 0 28 8 0 1 0 0 0 0 0 0 0 10 100
MORGEN 149 77 0 77 24 3 10 7 5 4 4 3 6 0 80 586
MOSWEGOT 307 155 116 39 132 75 19 22 1 0 2 7 2 1 15 234
MRCOLD 2632 1146 0 1146 842 53 61 13 170 0 61 47 93 47 219 6301
72
MRFIT 901 772 588 184 80 5 4 8 0 0 8 6 5 0 62 484
NCS1 545 372 0 372 67 9 17 26 8 43 5 2 12 0 75 1430
NCS2 270 187 0 187 25 2 5 10 1 20 5 1 3 1 43 778
NCS3 446 274 0 274 81 8 21 20 3 38 5 0 5 0 60 952
NFR 124 90 0 90 27 2 9 1 0 0 0 1 4 0 39 330
NHANESI 1221 649 237 412 343 89 27 10 24 1 35 9 11 5 89 1691
NHANESIII 756 423 0 423 131 0 0 0 38 0 0 0 14 0 226 2064
NHS1 5198 2267 0 2267 1327 23 104 648 365 342 1 100 150 155 2194 23241
NPHSI 200 157 89 68 24 0 0 0 0 0 0 0 0 0 0 215
NPHSII 211 140 126 14 46 27 5 3 0 13 0 2 5 0 17 135
NSHS 58 13 0 13 38 1 1 1 3 0 4 0 0 0 0 23
OB43 23 14 0 14 8 1 1 1 0 0 0 0 1 0 5 76
OSAKA 261 42 26 16 144 57 27 16 62 1 1 0 4 0 10 628
OSLO 2575 1585 0 1585 372 55 78 30 60 116 35 14 155 5 503 5816
OYABE 67 13 0 13 45 33 6 5 5 0 0 0 0 0 13 134
PARIS1 436 174 0 174 91 2 29 3 3 24 21 0 7 4 119 1899
PREVEND 127 87 72 15 17 0 9 6 2 0 3 0 8 1 24 183
PRIME 205 144 127 17 42 33 6 0 0 16 0 0 0 0 29 183
PROCAM 594 405 304 101 79 59 14 0 7 76 2 7 8 0 92 753
PROSPER 396 267 202 65 115 0 0 0 0 0 0 0 0 0 0 243
QUEBEC 423 297 260 37 93 0 0 0 4 29 0 0 0 0 0 414
RANCHO 543 246 243 3 197 0 1 0 10 0 9 1 5 5 36 486
REYK 4531 3245 2028 1217 768 183 162 45 82 12 45 78 71 6 532 6672
RF2 90 64 0 64 18 2 7 0 0 0 0 2 1 0 27 320
ROTT 313 100 85 15 79 28 7 1 41 30 1 0 16 2 56 718
SHHEC 674 453 321 132 183 55 21 21 3 2 2 2 6 1 122 757
SHS 782 449 303 146 214 8 10 0 15 4 24 6 2 4 39 1155
SPEED 354 253 98 155 77 66 2 1 0 0 1 5 9 0 69 475
TOYAMA 92 34 33 1 51 24 17 10 4 0 0 0 0 0 6 83
TPT 737 555 367 188 145 82 16 8 6 0 2 5 17 0 146 675
TROMSØ 1001 604 550 54 365 278 40 32 1 9 2 0 7 1 80 505
ULSAM 770 458 336 122 229 134 40 16 15 0 4 6 18 2 65 737
USPHS2 643 310 282 28 259 217 40 0 0 38 0 0 0 0 0 791
VHMPP 3281 1683 0 1683 783 81 122 24 184 1 61 45 57 34 460 6929
VITA 26 18 15 3 5 5 0 0 0 0 2 0 1 0 4 23
WHIHABPS 768 48 42 6 719 719 0 0 0 0 0 0 0 0 0 38
WHITEI 470 218 0 218 138 19 14 4 20 0 12 6 40 4 62 1228
WHITEII 349 317 255 62 10 2 2 2 1 0 0 4 3 1 15 328
WHS 603 236 228 8 286 239 26 19 0 52 0 0 0 0 0 625
WOSCOPS 259 213 174 39 39 0 0 0 0 0 0 0 0 0 0 111
ZARAGOZA 80 43 30 13 37 7 0 0 0 0 0 0 0 0 0 20
ZUTE 41 16 13 3 14 1 1 0 3 0 0 1 7 0 4 65
Total 69174 40218 16390 23828 17445 5023 2014 1313 2108 1298 919 704 1331 385 9788 128137
Note: Some studies did not provide information regarding non-fatal myocardial infarction or
non fatal stroke. Some studies did not provide information on type of cerebrovascular event
(ischaemic versus haemorrhagic stroke).
73
List 2.1: List of core variables sought in the Emerging Risk Factors Collaboration
Source: Extract from the Protocol of ERFC 2.
74
List 2.2: Data request sent to all the ERFC investigators in December 2010
Source: Personal communication from Dr M. Walker, data manager of ERFC.
75
List 2.3: List of study acronyms in the ERFC
AFTCAPS, Air Force/Texas Coronary Atherosclerosis Prevention Study; ALLHAT,
Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial; AMORIS,
Apolipoprotein Related Mortality Risk Study; ARIC, Atherosclerosis Risk in Communities
Study; ATENA, cohort of Progetto CUORE; ATS_SAR, cohort of Risk Factors and Life
Expectancy Pooling Project; ATTICA, ATTICA Study; AUSDIAB, Australian Diabetes,
Obesity and Lifestyle Study; BHS, Busselton Health Study; BRHS, British Regional Heart
Study; BRUN, Bruneck Study; BUPA, BUPA Study; BWHHS, British Women's Heart and
Health Study; CaPS, Caerphilly Prospective Study; CASTEL, Cardiovascular Study in the
Elderly; CHA, Chicago Heart Association Study; CHARL, Charleston Heart Study; CHS-1,
original cohort of the Cardiovascular Health Study; CHS-2, supplemental African-American
cohort of the Cardiovascular Health Study; COPEN, Copenhagen City Heart Study; DISCO,
cohort of Risk Factors and Life Expectancy Pooling Project; CUORE, Progetto CUORE;
DRECE, Diet and Risk of Cardiovascular Disease in Spain; DUBBO, Dubbo Study of the
Elderly; EAS, Edinburgh Artery Study; EMOFRI, part of CUORE; EPESEBOS, The
Established Populations for the Epidemiologic Study of the Elderly Studies, Boston;
EPESEIOW, The Established Populations for the Epidemiologic Study of the Elderly Studies,
Iowa; EPESENCA, The Established Populations for the Epidemiologic Study of the Elderly
Studies, North Carolina; EPESENHA, The Established Populations for the Epidemiologic
Study of the Elderly Studies, New Haven; EPICNOR, European Prospective Investigation of
Cancer Norfolk Study; ESTHER, Epidemiologische Studie zu Chancen der Verhutung und
optimierten Therapie chronischer Erkrankungen in der alteren Bevolkerung; FIA, First
Myocardial Infarction in Northern Sweden; FINE-FIN, Finland, Italy and Netherlands Elderly
Study - Finland cohort; FINE-IT, Finland, Italy and Netherlands Elderly Study – Italian cohort;
FLETCHER, Fletcher Challenge Blood Study; FINRISK-92, Finrisk Cohort 1992; FINRISK-
97, Finrisk Cohort 1997; FRAMOFF, Framingham Offspring Study; FUNAGATA, The
Funagata Study; GLOSTRUP, Research Centre for Prevention and Health; GOH, The
Glucose Intolerance, Obesity and Hypertension Study; GOTO13, Goteborg Study 1913;
GOTO33, Göteborg 1933 Study; GOTO43, Göteborg 1943 Study; GOTOW, Population Study
of Women in Gothenburg, Sweden; GREPCO, cohort of Risk Factors and Life Expectancy
Pooling Project; GRIPS, Göttingen Risk Incidence and Prevalence Study; GUBBIO, cohort of
Risk Factors and Life Expectancy Pooling Project; HBS, Helsinki Businessmen Study;
HELSINAG, Helsinki Aging Study; HISAYAMA, Hisayama Study; HONOL, Honolulu Heart
Program; HOORN, Hoorn 275 Study; HPFS, Health Professionals Follow-up Study; IKNS,
Ikawa, Kyowa, and Noichi Study; ISRAEL, Israeli Ischaemic Heart Disease Study; KARELIA,
North Karelia Project; KIHD, Kuopio Ischaemic Heart Disease Study; LASA, Longitudinal
Aging Study Amsterdam; LEADER, Lower Extremity Arterial Disease Event Reduction Trial;
MALMO, Malmö Study; MATISS-83, cohort of Progetto CUORE; MATISS-87, cohort of
Progetto CUORE; MATISS-93, cohort of Progetto CUORE; MCVDRFP, Monitoring of CVD
Risk Factors Project; MESA, Multi-Ethnic Study of Atherosclerosis; MICOL, cohort of Risk
76
Factors and Life Expectancy Pooling Project; MOGERAUG1, MONICA/KORA Augsburg
Surveys S1; MOGERAUG2, MONICA/KORA Augsburg Surveys S2; MOGERAUG3,
MONICA/KORA Augsburg Surveys S3; MONFRI-86, cohort of Progetto CUORE; MONFRI-
89, cohort of Progetto CUORE; MONFRI-94, cohort of Progetto CUORE; MONICA, cohort of
Risk Factors and Life Expectancy Pooling Project; MORGEN, Monitoring Project on Chronic
Disease Risk Factors; MOSWEGOT, MONICA Göteborg Study; MRCOLD, MRC Study of
Older People; MRFIT, Multiple Risk Factor Intervention Trial 1; NCS 1, 2 and 3, Norwegian
Counties Studies; NFR, cohort of Risk Factors and Life Expectancy Pooling Project;
NHANES I, First National Health and Nutrition Examination Survey; NHANES III, Third
National Health and Nutrition Examination Survey; NHS, Nurses' Health Study; NPHSI,
Northwick Park Heart Study I; NPHSII, Northwick Park Heart Study II; NSHS, Nova Scotia
Health Survey; OB43, cohort of Risk Factors and Life Expectancy Pooling Project; OSAKA,
Osaka Study; OSLO, Oslo Study; OYABE, Oyabe study; PARIS1, Paris Prospective Study I;
PREVEND, Prevention of Renal and Vascular End Stage Disease Study; PRHHP, Puerto
Rico Heart Health Program; PRIME, Prospective Epidemiological Study of Myocardial
Infarction; PROCAM, Prospective Cardiovascular Münster Study; PROSPER, Prospective
Study of Pravastatin in the Elderly at Risk; QUEBEC, Quebec Cardiovascular Study;
RANCHO, Rancho Bernardo Study; REYK, Reykjavik Study; RF2, cohort of Risk Factors and
Life Expectancy Pooling Project; RIFLE, Risk Factors and Life Expectancy Pooling Project;
ROTT, The Rotterdam Study; SHHEC, Scottish Heart Health Extended Cohort; SHS, Strong
Heart Study; SPEED, Speedwell Study; TARFS, Turkish Adult Risk Factor Study; TOYAMA,
Toyama; TROMSØ, Tromsø Study; ULSAM, Uppsala Longitudinal Study of Adult Men;
USPHS, U.S. Physicians Health Study; USPHS2, U.S. Physicians Health Study II; VHMPP,
Vorarlberg Health Monitoring and Promotion Programme; VITA, Vicenza Thrombophilia and
Athrosclerosis Project; WHIHABPS, Women's Health Initiative (Hormones and Biomarkers
Predicting Stroke in Women); WHITE I, Whitehall I Study;WHITE II, Whitehall II Study; WHS,
Womens Health Study; WOSCOPS, West of Scotland Coronary Prevention Study;
ZARAGOZA, Zaragosa study; ZUTE, Zutphen Elderly Study
77
References
1. Strong K, Mathers C, Leeder S, Beaglehole R. Preventing chronic diseases: how many
lives can we save? Lancet. 2005;366:1578-1582.
2. Danesh J, Erqou S, Walker M et al. The Emerging Risk Factors Collaboration: analysis
of individual data on lipid, inflammatory and other markers in over 1.1 million
participants in 104 prospective studies of cardiovascular diseases. Eur J Epidemiol.
2007;22:839-869.
3. Sarwar N, Sandhu MS, Ricketts SL et al. Triglyceride-mediated pathways and coronary
disease: collaborative analysis of 101 studies. Lancet. 2010;375:1634-1639.
4. Di AE, Sarwar N, Perry P et al. Major lipids, apolipoproteins, and risk of vascular
disease. JAMA. 2009;302:1993-2000.
5. Erqou S, Kaptoge S, Perry PL et al. Lipoprotein(a) concentration and the risk of
coronary heart disease, stroke, and nonvascular mortality. JAMA. 2009;302:412-423.
6. Kaptoge S, Di AE, Lowe G et al. C-reactive protein concentration and risk of coronary
heart disease, stroke, and mortality: an individual participant meta-analysis. Lancet.
2010;375:132-140.
7. Wormser D, Kaptoge S, Di AE et al. Separate and combined associations of body-
mass index and abdominal adiposity with cardiovascular disease: collaborative analysis
of 58 prospective studies. Lancet. 2011;377:1085-1095.
8. Seshasai SR, Kaptoge S, Thompson A et al. Diabetes mellitus, fasting glucose, and
risk of cause-specific death. N Engl J Med. 2011;364:829-841.
9. Sarwar N, Gao P, Seshasai SR et al. Diabetes mellitus, fasting blood glucose
concentration, and risk of vascular disease: a collaborative meta-analysis of 102
prospective studies. Lancet. 2010;375:2215-2222.
10. Thun, M. J., Day-Lally, C., Myers, DG, Calle, EE, Flanders, WD, Zhu, B-P, Namboodiri,
MM, and Heath, CW Jr. Trends in tobacco smoking and mortality from cigarette use in
Cancer Prevention Studies I (1959 through 1965) and II (1982 through 1988). Changes
in Cigarette-Related Disease Risks and Their Implication for Prevention and
Control.Smoking and Tobacco Control Monograph No.8.Rockville (MD) [a], 305-382.
1997. Shopland DR, Burns DM, Garfinkel L, Samet JM.
11. Iverson C, Christiansen S, Flanagin A, et al. American Medical Association Manual of
Style: A Guide for Authors and Editors. 10 ed. New York, NY: 2007.
12. Whitlock G, Lewington S, Sherliker P et al. Body-mass index and cause-specific
mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet.
2009;373:1083-1096.
13. O'Connor RJ, McNeill A, Borland R et al. Smokers' beliefs about the relative safety of
other tobacco products: findings from the ITC collaboration. Nicotine Tob Res.
2007;9:1033-1042.
14. Delnevo CD. Smokers' choice: what explains the steady growth of cigar use in the
U.S.? Public Health Rep. 2006;121:116-119.
78
Chapter 3: Distribution and cross-sectional correlates of
smoking in developed countries
Summary
A smoker is exposed to the inhaled smoke as well as to the ingested tobacco when
smoking a cigarette, pipe or cigar. As the tar and gas phases of tobacco consist of
thousands of chemical compounds, it has proven difficult to pinpoint the causal
element which promotes CVD. A better understanding of correlates of tobacco may
shed light on how smoking induces heart diseases. Another reason for investigating
correlates of tobacco is appropriate adjustment of the association of tobacco with
CVD for potential confounders, enhancing quality of the estimates of risk and the
chances that an observed relationship is causal.
In this Chapter, I first summarize the distribution of smoking variables, and then
investigate correlations between smoking and conventional CVD risk factors such as
blood pressure, lipid levels, inflammatory markers and adiposity. Up to 929,335
individuals from 114 studies were included in these analyses. A majority of current
and ex-smokers had initiated smoking at age 18 years old or younger and 75% had
begun before age 22 years old. BMI was slightly lower in current (25.4 kg/m2, SD:
3.8) compared to ex (26.4 kg/m2, SD: 4.0) and never smokers (26.3 kg/m2, SD: 4.2).
Current smokers were more likely to be male and white, to declare being current
alcohol drinkers and to be employed in skilled rather than manual jobs. Current
smokers, especially women, had higher baseline levels of total cholesterol, non-HDL-
C, loge triglycerides, and ApoB, while they had lower levels of HDL-C and ApoA1
compared to never smokers. Smoking was most strongly correlated to markers of
inflammation. There was a positive gradient from never to ex-smokers and to current
smokers for CRP, fibrinogen, leukocyte counts and interleukin-6. In current smokers,
per SD increase in number pack-years, CRP was raised by 0.18 mg/l (95% CI: 0.11
to 0.24) and fibrinogen by 0.22 μmol/l (0.16 to 0.28).  
79
3.1 Background
Vasomotor dysfunction, inflammation and modification of lipids are all responsible for
the development of atherosclerotic plaques in the arteries, later leading to thrombosis
and clinical CVD. Smoking has been shown to accelerate atherosclerosis and
promote thrombosis in the coronary arteries, the aorta, the carotid and cerebral
arteries and the large arteries in the peripheral circulation 1 (see Section 1.4).
Studies have shown a relationship between smoking and hyperlipidaemia 3, blood
pressure 4, inflammation 5,6, oxidative stress and platelet activation 7, and a role for
nicotine, carbone monoxide and oxidant chemicals has been suggested. However,
these studies have been characterized by small sample sizes and inconsistent
adjustment. Studies have, as a result, sometimes found inconsistent results, and the
strength of correlations between smoking and other known cardiovascular risk factors
remains uncertain. Published studies have also not been able to assess how
adjustment for potential correlates modifies the association between smoking and
CVD risk. In this context, ERFC represents a unique opportunity to adjust
associations for a large range of questionnaire-based and blood-based factors
significantly correlated with smoking.
In this context, there is an interest in better quantifying the relationship between
smoking and a range of lifestyle and biochemical risk factors of CHD using
information contained in the ERFC. Smoking status was available in 929,335
individuals from 114 studies characterized by random sampling at baseline from the
general population of the country involved. Subsets had information on
anthropometric measurements, blood pressure, lipid levels, and inflammatory
markers.
3.2 Methods
3.2.1 Dataset
Details of study selection, data collection and harmonization within the ERFC have
been described in Chapter 1. In total, 929,335 individuals included provided
information on age, sex, smoking status and body mass index (BMI). Smaller subsets
had information on other conventional risk factors: 154,674 on waist to hip ratio
(WHR), 754,077 on systolic blood pressure, 738,215 on total cholesterol (on 382,079
80
on HDL-C), 107,879 on C-reactive protein, 10,974 on leukocyte counts and 180,379
on fibrinogen levels.
3.2.2 Statistical methods
Positively skewed variables such as Lp(a), triglycerides and C-reactive protein were
log-transformed, apart from smoking pack-years which was kept untransformed for
ease of interpretation. The proportion of current and ex-smokers at baseline was
drawn by cohort, and ordered by decade of baseline enrollment of the study. Ninety
five percent confidence intervals for the frequencies were derived assuming a
binomial distribution )( p , with p the probability of being a current or ex-smoker in a
specific study. Graphically, this translated into the width of the confidence interval
being inversely proportional to the number of individuals in the study of a specific
sex. For continuous variables, box-plots were represented for each study and
grouped by decade of enrollment. Mean of smoking amount reported at baseline was
drawn by sex and by tenths of smoking duration, using a linear mixed model adjusted
for study and age 50 years old (see below). Unadjusted means of all continuous
variables were calculated within studies and pooled across studies by random effect
meta-analysis; and the pooled variance was obtained as an average of study-specific
variances weighted by study size. Categorical variables were summarized as raw
counts and proportions.
Statistical methods used in the ERFC for the assessment of cross-sectional
correlates using linear mixed models follow the example of methods used in the
Fibrinogen Study Collaboration 8. First, I estimated the effect of 1 standard deviation
increase in smoking pack-years (equal to approximately 17 pack-years) on other
characteristics. Secondly, I graphically assessed the strength and shape of
correlations by plotting predicted means of continuous characteristics based on linear
mixed model regressions (i) over smoking pack-years, in current smokers only,
divided into tenths and (ii) over smoking status categorized as current, ex and never.
In comparison to a standard regression, a so-called “mixed” regression allows the
effect of smoking, and additionally of selected covariates, to truly vary across studies.
Studies included in the ERFC dataset recruited participants from all over the World
and over a time scale of 50 years. Therefore, unexplained heterogeneity between
studies may reasonably be assumed to be real heterogeneity. A mixed model relaxes
the assumption of a fixed true estimate, allowing estimation of true heterogeneous
81
effects of smoking on correlates across studies, even after adjustment for available
confounder and effect modifiers. A mixed model can be written as:
sisisisssi XEuY   )(
Where s=1…S indicates the study. i=1…ns individuals belonging to study s, Ysi the
risk factor levels, Esi the exposure of interest, Xsi other covariates of adjustment.
 represents the constant fixed effect of the cohort and  is a vector of coefficients
for covariates of adjustment.  is the parameter of interest, being the change in risk
factor per unit increase in exposure, adjusted for covariates Xsi. The random noise
around the predicted value of the risk factor level is represented by a random
variable  and the heterogeneous effect of the exposure of interest across study is
represented by another random variable u . I assume a normal distribution of these
two random variables : ),0(~),,0(~ 22 rsius NNu  .
All mixed models included a random effect of smoking (and optionally age, age2 and
sex) at the study level, while the main effect of study was modeled as a separate
fixed effect. In addition to study, models were also adjusted for at least age and sex
(and when stated in the legends for age2, age x sex and age and sex interactions
with smoking). For each fitted mixed model, predicted means and their 95%
confidence intervals were plotted, separately for men and women, and for age fixed
to 50 years old. An inverse-variance weighted polynomial curve was superimposed
on the graph for men and women separately, to investigate whether the association
was consistent with a linear or a quadratic shape. For categorical variables, instead
of the effect of 1 SD increase in pack-years, the difference in the number of pack-
years compared to the reference category was reported. This was obtained by
regressing pack-years over categories of the variable using a linear mixed model,
and adjusting at least for age and sex.
I also reported correlation coefficients between smoking pack-years in current
smokers and continuous characteristics, as they provide useful summaries when the
shape of association is approximately linear. To obtain these correlation coefficients,
I followed the steps listed below:
1. “Partial” Pearson correlation coefficients rs adjusted for baseline age and sex
were computed for each study k=1…s.
82
2. These coefficients were z-transformed to fulfill the assumption of normality 9
using Fisher’s formula:
))1()1((log5.0 sses rrz 
3. Z-transformed coefficients were pooled across studies by random-effects
meta-analysis 10
4. The pooled estimate was back-transformed using the same formula as before
to obtain the overall correlation coefficient:
)1)2(exp())1)2(exp(  pooledpooledpooled zzr
All analyses were performed using STATA statistical software, release 11 (StataCorp
LP, College Station, Texas, USA).
3.3 Results
3.3.1 Baseline characteristics
There was heterogeneity between studies and between sexes regarding the
prevalence of current and ex-smokers, ranging for current users from 5% for
USPHS2 to above 80% for ZUTE (Figures 3.1 & 3.2). Studies which began to enroll
in the 1960s and 1970s had a tendency to report higher prevalence of current
smokers and lower prevalence of ex-smokers than studies started in the 1980s,
1990s and 2000s, for both men and women (Figure 3.3 & 3.4). Substantial
heterogeneity was also observed at a study level for other variables of smoking,
namely number of pack-years in current smokers, and years since quitting smoking
in ex-smokers, which were not explained by the decade in which the study was
started (Figure 3.5 & 3.6). Within each study, women consistently reported lower
prevalence of current smokers and, amongst current users, lower levels of pack-
years than men.
After adjustment for age, baseline amount and duration were positively and
approximately linearly correlated in both men and women, with wider confidence
intervals in women due to their overall lower prevalence of smoking compared to
men (Figure 3.7). For every 10 additional years of smoking, amount was raised by
approximately 2.3 cigarettes per day for both men and women.
83
3.3.2 Smoking association with demographic characteristics
The proportion of current smokers was 41% amongst men and 27% amongst women
(Table 3.1). The proportion of ex-smokers was 28% amongst men whilst it was 18%
amongst women. Mean age was 53.3 years old (SD: 8.5) in current smokers, 55.0
years old in ex-smokers (SD: 8.6) and 54.3 years old in never smokers (SD: 10.2).
One SD increase in age (corresponding to approximately 9 years) translated into an
average of 4.5 additional smoking pack-years in current smokers at baseline. Whites
were 5% more likely to report being current smokers than non-Whites (38% versus
33%)
3.3.3 Smoking association with lifestyle factors
Alcohol use and smoking were positively correlated. Current alcohol drinkers were
slightly more likely than non-current alcohol drinkers (38% versus 32%) to be also
current smokers (Table 3.1). By contrast, being a diabetic had a negative effect on
the likelihood of being a smoker. Non-diabetics were 13% more likely than diabetics
(36% versus 23%) to declare being current smokers and 10% less likely to have
stopped smoking (26% versus 36%). Smoking was also associated with education.
The percentage of current smokers was highest amongst individuals reaching
secondary education (41%), whilst the percentage of never smokers was highest
amongst individuals reporting no schooling. Regarding occupation, the percentage of
smoker was highest amongst office workers (65% current smokers), whilst it was
lowest amongst individuals not working (19%), a heterogeneous category of older
individuals which was comprised of housewives, unemployed individuals and retired
individuals, with a mean age of 57.4 (SD: 11.5). Leaving aside individuals not
working, the percentage of never smokers was highest amongst individuals working
in services (52%). Whites reported an average 9.22 (95% confidence interval of 6.36
to 12.07) additional pack-years than non-Whites at baseline. Number of smoking
pack-years was not significantly different depending on alcohol status, history of
diabetes, level of education and occupation.
3.3.4 Smoking association with adiposity and blood pressure
BMI was slightly lower in current (25.4 kg/m2, SD: 3.8) compared to ex (26.4 kg/m2,
SD: 4.0) and never smokers (26.3 kg/m2, SD: 4.2) (Table 3.2 & Figure 3.8). In
current smokers, correlation with pack-years was low and the Pearson correlation
84
coefficient was equal to 0.02 (Figure 3.9). In contrast, WHR was not different
amongst the smoking groups and not associated with number of pack-years in
current smokers (difference in WHR per 1-SD increase in pack-years was 0.01).
Women current smokers had relatively lower SBP and DBP levels than ex and never
smokers, but there were no significant trends with number of pack-years (partial
correlations equal to 0). Associations of blood pressure with number of years since
stopping to smoke in ex-smokers were also null (Figure 3.10).
3.3.5 Smoking association with lipids
Current smokers, especially women, had higher baseline levels of total cholesterol,
non-HDL-C, loge triglycerides, ApoB; while they had lower levels of HDL-C and
ApoA1 compared to never smokers (Table 3.2). Levels of lipids and lipoproteins
were not significantly different for ex-smokers compared to those of never smokers,
except for total cholesterol, where they were slightly raised (Figure 3.8). When
focusing on current smokers and looking at the relationships with number of pack-
years smoking, there were slight positive relationships for total cholesterol, non-HDL-
C, loge triglycerides, ApoB; and slightly negative relationships for HDL-C and ApoA1.
Correlation coefficients with pack-years smoking were all less than 0.10 with these
variables. In ex-smokers, log triglycerides decreased linearly and HDL-C increased
slightly with number of years since stopping (Figure 3.9).
3.3.6 Smoking association with inflammation
Smoking was most strongly correlated to markers of inflammation. There was a
positive gradient from never to ex-smokers and to current smokers for CRP,
fibrinogen, leukocyte counts and interleukin-6. In current smokers, per SD increased
in pack-years, CRP was raised by 0.18 mg/l (95% CI: 0.11 to 0.24). The
corresponding increase for fibrinogen was 0.22 μmol/l (0.16 to 0.28) (Table 3.2). For
all inflammatory markers (apart from albumin men ex-smokers), levels in ex-smokers
were not significantly different from levels in never smokers. The shape of
association for these markers was curvilinear, with an apparent tailoring of the
association for loge leukocyte counts and loge interleukin-6 above 40 pack-years for
both men and women. Overall, partial correlations were 0.14 for CRP, 0.11 for
fibrinogen, 0.17 for loge leukocyte counts, 0.13 for loge interleukin-6 and -0.04 for
albumin (Figure 3.9). In ex-smokers loge C-reactive protein and loge leukocyte
85
counts were negatively associated with the number of years since stopping smoking,
while associations with other markers were not significant (Figure 3.10).
86
3.4 Discussion
Smoking status, defined as being a current, ex- or never smoker, was available in
929,335 individuals from 114 studies, around a half of participants being women. In
that respect, the ERFC represents an effort similar in scale to Cancer Prevention
Studies conducted in the USA in the 1960’s and 1980’s, which both included more
than 1 million participants but did not have information on biochemical risk factors
relevant to CVD such as blood pressure, lipid levels and inflammatory biomarkers 11.
The present analysis is the most comprehensive and detailed analysis on the cross-
sectional correlates of smoking status and dose-duration (as measured by pack-
years) to date, with adequate power to assess shapes of association separately for
each gender.
3.4.1 Baseline characteristics
The prevalence of active smokers was 34% in the overall data and was higher than
current levels in the developed world which comprise between 20% and 25% 12.
There was evidence of a gender gap, with prevalence of current smoking in men
much higher than in women (41% versus 27%). A majority of studies included in the
ERFC recruited individuals in the 1960’s and 1970’s and these percentages are a
reflection of a time period when smoking was more prevalent, especially amongst
men, in developed countries 13. The drop in smoking prevalence, and the narrowing
in the gender gap observed since the 1960’s, was observable when plotting sex-
specific prevalence within studies by starting decade of recruitment.
A positive and approximately linear association was found between baseline amount
and baseline duration of smoking. For every 10 additional years of smoking, smoking
amount was raised by approximately 2.3 cigarettes per day for both men and women.
This cross-sectional observation shows that continuing smokers are likely to increase
their smoking intake over time. In my dataset, most smokers had started in
teenagehood or early adulthood: median age was 18 years old and inter-quartile
range 16 to 22 years old. Studies have shown that adolescents initiating smoking are
particularly vulnerable to becoming dependent on nicotine and rapidly increase their
intake over time 14. The mechanisms by which nicotine establishes dependence in
smokers are thought to involve adaptative changes of nicotinic acetylcholine
receptors, which are widely distributed through the central nervous system, 15. The
87
activation of these receptors by nicotine increases the release of neurotransmitters
and produces effects on a large number of physiological processes such as
locomotion, anxiety, learning and memory.
3.4.2 Smoking and lifestyle factors
In the ERFC, current smokers were over-represented amongst individuals reaching
secondary education and working in offices, and never smokers were over-
represented amongst individuals without schooling, not working or in manual jobs.
These are surprising findings which contrast with previous reports showing an over-
representation of smokers in deprived socio-economic groups characterized by a
poor diet, low levels of physical activity and high prevalence of alcohol drinking 16,17,
as a result of higher exposure to public smoking 18 and lower education about
smoking hazards 19. One reason for this could be the over-representation of
unqualified women in never smokers as well as older individuals having reached
retirement age. Alcohol drinking and smoking were correlated habits. In that respect,
sociological studies have shown that smoking tends to become socially acceptable
and normative in student parties where alcohol is also consumed, whereas it is
stigmatized the rest of the time 20. Diabetics were less likely to smoke and more likely
to have stopped smoking than non-diabetics. This may reflect a greater awareness of
the health effects of smoking amongst diabetics in the developed world.
3.4.3 Smoking and adiposity
It has been previously reported that smokers have decreased BMI but increased
central adiposity 21-23. In the ERFC, never smokers were leaner than current smokers,
and WHR was not significantly associated with the number of pack-years amongst
current smokers. These paradoxical effects are thought to result from increased
resting metabolic rate, increased sympathetic nervous system activity and
thermogenesis in smokers 23. Never and ex-smokers had similar BMI levels, which
contrasts with available epidemiological evidence showing that smoking cessation is
accompanied by weight gain, a stronger appetite and an increase in adipose tissue
lipoprotein lipase activity 23.
3.4.4 Smoking and blood pressure
88
Smoking is thought to affect heart rate and blood pressure 24 Nicotine uptake has
been shown to immediately activate the sympathetic nervous system 25. In both
animal and human models, studies have shown that active and passive cigarette
smoke exposures are associated with a decrease in vasodilatory function 2. In
humans, cigarette smoke exposure impaired endothelium-dependent vasodilatation
in macrovascular beds such as coronary and brachial arteries and in microvascular
beds. Cigarette smoke has also been shown to alter nitric oxide biosynthesis, a
vasoregulatory molecule which helps regulate inflammation, leucocyte adhesion,
platelet activation and thrombosis. In the ERFC, current, never and ex-smokers had
similar distributions of SBP and DBP, with slightly lower levels in never smokers. This
surprising lack of correlation between smoking and blood pressure could be due to
smoking having short-term rather than chronic effects on blood pressure. One
consequence is that most of the smoking effect on CVD is unlikely to be mediated
through modification of blood pressure levels.
3.4.5 Smoking and lipids
Smoking could promote atherosclerosis, in part, through modifying the lipid profile of
smokers 26. In the ERFC, associations with cholesterol and lipoprotein levels were
significant but modest and correlations were below 0.10. There were positive
correlations between smoking and loge triglycerides, total cholesterol, non-HDL-C
and Apolipoprotein B; and negative correlations with HDL-C and Apolipoprotein A.
Several mechanisms have been proposed to explain these correlations. They include
lipid oxidation, changes in composition of lipoproteins, alteration in plasma- and
lipoprotein- associated lipid transfer enzymes, changes in metabolism of fatty acids,
effects of levels of postprandial lipids and changes in cholesterol fluxes, particularly
reverse cholesterol transport 27. Specifically, higher levels of triglycerides and lower
levels of HDL-C have been suggested to be related to insulin resistance 2. Smoking
has also been shown to increase oxidative modification of LDL and decrease plasma
activity of paraoxanase, an enzyme that protects against LDL oxidation 2.
3.4.6 Smoking and inflammation
In the ERFC, current smoking was positively associated with several markers of
inflammation and most strongly with CRP and fibrinogen; and there was a negative
trend with number of years since stopping. Correlations between smoking and
chronic low grade inflammation, such as leukocyte counts, C-reactive protein,
89
interleukin-6 and fibrinogen, have already been reported in the literature 5-7,28.
Markers of inflammation are known to be associated with an increased CVD risk 29,30,
but whether they are causally related to CVD remains subject to debate. Local
recruitment of leukocytes on the surface of endothelial cells is an early event in
atherosclerosis 2. Elevation of various pro-inflammatory cytokines increases
leukocyte-endothelial cell interaction leading to leukocyte recruitment; and soluble
VCAM-1, ICAM-1 and E-selecting levels have been shown to be elevated in smokers
2. Cigarette smoking has also been found to activate pro-atherogenic molecules
leading to alteration in cell-cell interactions 2. However, the association between CRP
and vascular disease has been shown to attenuate considerably after adjustment for
non-inflammatory risk factors 5,28 and a Mendelian randomization study has reported
a lack of concordance of the associations of CRP genotypes and CRP
concentrations with CVD risk 31. The association between fibrinogen and CVD has
also been shown to be partially altered by progressive adjustment 29. These
observations translate into uncertainties regarding the relevance of inflammation as
mediating the association between smoking and CVD and therefore the need to
consider them as potential confounders.
3.4.7 Limitations
Despite its strengths, this analysis contains several limitations. Prevalence of
smoking was highly variable across studies and only part of this heterogeneity could
be attributed to study characteristics such as decade of baseline survey. Limited
information was available on the criteria used by each study to define categories of
“current”, “ex” and “never” smokers (such as “smoking at least once over the past
day/month/year or the interview”), which may have accounted for part of the
heterogeneity. Efforts to retrieve this information proved unsuccessful because this
information was rarely published by studies and had not been requested by ERFC
investigators. Evidence of digit preference in the reporting of amount and duration
may partly be attributed to misreporting or difficulty in recalling the exact number of
years since starting smoking and precisely estimating the amount smoked each day.
In that respect, distortion in levels of biochemical risk factors and misreporting of
smoking status were minimized by excluding individuals with pre-existing CVD. More
accurate information on smoking amount, for example measurement of biomarkers
such as cotinine levels 32, was not available. Despite efforts in harmonizing coding
across studies, measurement error may have been introduced when recoding level of
education and occupation which were assessed differently across study designs and
90
categorization may not be transferable across countries. The assessment of
correlates was done using cross-sectional information, and investigation of these
correlations using prospective data would be needed to confirm that they are causal.
Despite the large scale of the dataset, information for some biomarkers, especially
markers of inflammation, was only available in relatively small subsets of the data.
91
Figure 3.1: Proportion of current smokers in each cohort, by sex and by decade of
baseline survey
1960 1970 1980 1990 2000
0
20
40
60
80
P
ro
po
rti
on
of
cu
rr
en
ts
m
ok
er
s
(%
)
BH
S
C
H
A
G
O
H
G
O
TO
13
IS
R
A
E
L
R
EY
K
P
A
R
IS
1
C
H
A
R
L
G
O
TO
W
N
H
A
N
E
S
I
BU
PA
BR
H
S
P
R
O
C
A
M
S
P
E
ED
M
A
LM
O
K
A
R
E
LI
A
N
C
S
1
N
P
H
S
I
N
C
S
2
N
FR
U
LS
A
M
R
F2
C
A
PS
O
S
LO
M
R
FI
T
N
C
S
3
G
LO
S
TR
U
P
Q
U
E
B
E
C
N
H
S
1
H
P
FS
1
C
H
S
1
R
A
N
C
H
O
FI
N
E
_F
IN
H
E
LS
IN
A
G
H
B
S
D
U
BB
O FI
A
V
H
M
P
P
E
A
S
C
A
ST
EL
W
H
IT
E
II
FI
N
E
_I
T
A
R
IC
M
O
SW
E
G
O
T
H
IS
A
YA
M
A
M
O
G
E
R
A
U
G
2
K
IH
D
TP
T
M
O
N
FR
I8
9
M
O
G
E
R
A
U
G
1
G
O
TO
33
M
O
N
FR
I8
6
H
O
O
R
N
N
P
H
S
II
G
R
IP
S
S
H
S
M
IC
O
L
M
A
TI
SS
87
W
O
SC
O
PS
TR
O
M
S
Ø
G
U
B
B
IO
M
O
N
IC
A
M
C
V
D
R
FP
AT
S
_S
AR
S
H
H
E
C
D
IS
C
O
O
B
43
IK
N
S
M
A
TI
S
S
83
O
YA
B
E
G
R
E
P
C
O
U
S
PH
S
2
H
O
N
O
L
EP
IC
N
O
R
AF
TC
AP
S
FR
A
M
O
FF
W
H
IT
E
I
M
R
C
O
LD
A
U
SD
IA
B
FL
E
TC
H
E
R
C
H
S
2
E
M
O
FR
I
M
O
G
E
R
A
U
G
3
P
R
IM
E
A
LL
H
A
T
N
S
H
S
V
IT
A
R
O
TT
M
O
N
FR
I9
4
G
O
TO
43
ZA
R
A
G
O
ZA
FI
N
R
IS
K
97
LA
S
A
B
R
U
N
P
R
E
V
EN
D
N
H
A
N
E
S
III
FI
N
R
IS
K
92
M
O
R
G
EN
M
A
TI
S
S
93
LE
A
D
ER
P
R
O
SP
E
R
D
R
E
C
E
O
S
A
K
A
C
O
P
E
N
TO
YA
M
A
FU
N
A
G
AT
A
ZU
TE
W
H
IH
AB
PS
BW
H
H
S
W
H
S
A
TE
N
A
E
ST
H
ER
M
E
S
A
A
TT
IC
A
Male
Female
Cohorts are grouped by decade of baseline survey. Within each decade, cohorts are ordered
by increasing proportion of current smokers. 95% Confidence intervals are derived assuming
a binomial distribution and width is inversely proportional to the number of individuals in the
study of a specific sex. A total of 114 cohorts contributed to this graph.
92
Figure 3.2: Proportion of ex-smokers in each cohort, by sex and by decade of
baseline survey
1960 1970 1980 1990
2000
0
20
40
60
80
P
ro
po
rti
on
of
ex
-s
m
ok
er
s
(%
)
G
O
H
G
O
TO
13
IS
R
AE
L
C
H
A
R
L
P
AR
IS
1
R
EY
K
B
H
S
C
H
A
G
O
TO
W
B
U
PA
Q
U
E
BE
C
M
A
LM
O
N
H
AN
E
SI
R
F2
KA
R
EL
IA
U
LS
AM
N
C
S3
G
LO
S
TR
U
P
M
R
FI
T
N
C
S
1
O
S
LO
N
C
S2
C
A
PS
P
R
O
C
A
M
N
PH
S
I
N
FR
B
R
H
S
S
PE
E
D
N
H
S1
C
A
ST
E
L
FI
N
E_
FI
N
K
IH
D
N
P
H
S
II
TP
T
W
O
SC
O
P
S
O
YA
BE
D
U
BB
O
V
H
M
P
P
TR
O
M
SØ
O
B4
3
AT
S
_S
AR
S
H
H
EC
G
O
TO
33
G
R
IP
S
D
IS
C
O
M
C
VD
R
FP
IK
N
S
FI
A
W
H
IT
E
II
M
AT
IS
S8
3
M
O
SW
E
G
O
T
M
IC
O
L
M
O
N
IC
A
G
U
B
BI
O
M
A
TI
SS
87
M
O
N
FR
I8
9
M
O
N
FR
I8
6
M
O
G
ER
A
U
G
1
SH
S
M
O
G
E
R
A
U
G
2
H
B
S
FI
N
E_
IT
H
P
FS
1
A
R
IC
H
O
O
R
N
EA
S
H
IS
A
YA
M
A
C
H
S1
R
A
N
C
H
O
H
E
LS
IN
A
G
G
R
E
PC
O
AF
TC
AP
S
FR
A
M
O
FF
R
O
TT
V
IT
A
C
O
P
EN
PR
EV
EN
D
LE
A
D
ER
ZA
R
AG
O
ZA
FU
N
AG
A
TA
TO
YA
M
A
D
R
EC
E
ZU
TE
FI
N
R
IS
K9
7
O
SA
KA
FI
N
R
IS
K
92
BR
U
N
M
AT
IS
S9
3
N
H
AN
ES
III
M
O
R
G
E
N
G
O
TO
43
M
O
N
FR
I9
4
A
U
SD
IA
B
U
S
PH
S2
P
R
O
S
PE
R
FL
ET
C
H
ER
M
O
G
ER
A
U
G
3
P
R
IM
E
EM
O
FR
I
N
S
H
S
C
H
S2
AL
LH
AT
H
O
N
O
L
LA
SA
W
H
IT
E
I
E
PI
C
N
O
R
M
R
C
O
LD
A
TE
N
A
BW
H
H
S
W
H
S
W
H
IH
AB
PS
ES
TH
ER
A
TT
IC
A
M
E
SA
Male
Female
Cohorts were grouped by decade of baseline survey. Within each decade, cohorts are
ordered by increasing proportion of ex-smokers. 95% Confidence intervals are derived
assuming a binomial distribution and width is inversely proportional to the number of
individuals in the study of a specific sex. A total of 114 cohorts were included in this graph.
93
Figure 3.3: Proportion of current smokers by baseline year of study and by sex
0
20
40
60
80
P
ro
po
rti
on
of
cu
rr
en
ts
m
ok
er
s
(%
)
19
60
19
63
19
67
19
68
19
69
19
70
19
71
19
72
19
73
19
74
19
75
19
76
19
77
19
78
19
79
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
Male
Female
95% Confidence intervals are derived assuming a binomial distribution and width is inversely
proportional to the number of current smokers recruited in a specific year all studies
combined. A total of 114 cohorts were included in this graph.
94
Figure 3.4: Proportion of ex-smokers by baseline year of study and by sex
0
10
20
30
40
50
P
ro
po
rti
on
of
ex
-s
m
ok
er
s
(%
)
19
60
19
63
19
67
19
68
19
69
19
70
19
71
19
72
19
73
19
74
19
75
19
76
19
77
19
78
19
79
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
19
99
20
00
20
01
Male
Female
95% Confidence intervals are derived assuming a binomial distribution and width is inversely
proportional to the number of ex smokers recruited in a specific year all studies combined. A
total of 114 cohorts were included in this graph.
95
Figure 3.5: Box plots of pack-years in current smokers by studies and decade of
baseline survey
0
50
10
0
15
0
Sm
ok
in
g
am
ou
nt
:c
om
bi
ne
d
(p
ac
k
ye
ar
s)
1960 1970 1980 1990 2000
G
O
TO
W
R
E
YK
U
LS
A
M
R
F2
P
R
O
C
A
M
N
H
A
N
E
SI
N
FR
Q
U
E
B
EC
G
R
E
P
C
O
TR
O
M
S
Ø
M
O
N
FR
I8
6
M
C
V
D
R
FP
M
O
N
FR
I8
9
O
B
43
M
IC
O
L
M
O
N
IC
A
D
IS
C
O
C
A
S
TE
L
W
H
IT
EI
I
AT
S_
S
AR
M
AT
IS
S
83
M
AT
IS
S
87
G
R
IP
S
S
H
H
E
C
E
AS
A
R
IC
IK
N
S
R
A
N
C
H
O
H
IS
AY
A
M
A
D
R
E
C
E
N
H
A
N
E
SI
II
A
TE
N
A
E
M
O
FR
I
M
O
N
FR
I9
4
M
O
R
G
E
N
FL
E
TC
H
E
R
M
A
TI
S
S9
3
N
S
H
S
BW
H
H
S
ZU
TE
O
S
A
K
A
M
E
S
A
A
TT
IC
A
G
O
TO
W
R
E
YK
U
LS
A
M
R
F2
P
R
O
C
A
M
N
H
A
N
E
SI
N
FR
Q
U
E
B
EC
G
R
E
P
C
O
TR
O
M
S
Ø
M
O
N
FR
I8
6
M
C
V
D
R
FP
M
O
N
FR
I8
9
O
B
43
M
IC
O
L
M
O
N
IC
A
D
IS
C
O
C
A
S
TE
L
W
H
IT
EI
I
AT
S_
S
AR
M
AT
IS
S
83
M
AT
IS
S
87
G
R
IP
S
S
H
H
E
C
E
AS
A
R
IC
IK
N
S
R
A
N
C
H
O
H
IS
AY
A
M
A
D
R
E
C
E
N
H
A
N
E
SI
II
A
TE
N
A
E
M
O
FR
I
M
O
N
FR
I9
4
M
O
R
G
E
N
FL
E
TC
H
E
R
M
A
TI
S
S9
3
N
S
H
S
BW
H
H
S
ZU
TE
O
S
A
K
A
M
E
S
A
A
TT
IC
A
In total, 43 studies provided information on the number of pack-years. Cohorts are ordered by
decade of baseline survey and by increasing median of pack-years. A total of 44 studies
including 82.922 individuals were included in this graph.
96
Figure 3.6: Box plots of number of years since stopping smoking by studies and
decade of baseline survey
0
20
40
60
80
N
um
be
ro
fy
ea
rs
si
nc
e
qu
itt
in
g
sm
ok
in
g
1960 1970 1980 1990 2000
PA
R
IS
1
G
O
TO
W
R
EY
K
(to
ta
l)
K
AR
E
LI
A
N
C
S
3
N
C
S1
N
C
S
2
O
SL
O
C
A
PS
B
U
PA R
F2
N
FR
SP
EE
D
Q
U
E
BE
C
U
LS
AM
P
R
O
C
A
M
N
H
A
N
ES
I
N
P
H
SI
(to
ta
l)
H
IS
AY
A
M
A
M
O
N
IC
A
O
B4
3
A
TS
_S
AR
D
IS
C
O
FI
N
E
_I
T
M
O
N
FR
I8
9
M
O
N
FR
I8
6
M
C
V
D
R
FP
M
A
TI
S
S8
3
TR
O
M
SØ
M
AT
IS
S
87
S
H
H
E
C
AR
IC
D
U
BB
O
R
AN
C
H
O
C
H
S
1
(to
ta
l)
ZU
TE
A
TE
N
A
M
AT
IS
S
93
D
R
EC
E
M
O
N
FR
I9
4
M
O
R
G
EN
LE
A
D
ER
E
M
O
FR
I
P
R
IM
E
FI
N
R
IS
K9
2
N
H
A
N
ES
III
FI
N
R
IS
K
97
AU
SD
IA
B
LA
SA
BW
H
H
S
W
H
IT
E
I
M
R
C
O
LD
(to
ta
l)
M
E
SA
(to
ta
l)
In total, 54 studies including 80,318 individuals provided information on number of years since
stopping smoking. Cohort are grouped by decade of baseline survey, and within each
decade, ordered by increasing number of years.
97
Figure 3.7: Smoking amount cigarette by sex and by duration of smoking
5
10
15
20
25
0 10 20 30 40 50
Smoking duration (years)
Male Female
S
m
ok
in
g
am
ou
nt
:,
ci
ga
re
tte
s
eq
ui
va
le
nt
/d
ay
(9
5%
C
I)
Mean baseline smoking amount of cigarette, pipes and cigar combined were plotted by sex
against mean of each tenth of smoking duration amongst current smokers at baseline. Error
bars represent 95% confidence intervals. Means and SD were adjusted for study and age 50
years old using mixed models (see Methods section). 44 studies and 82,922 current smokers
at baseline provided information on both amount and duration at baseline.
98
Figure 3.8: Cross-sectional association between smoking status and continuous
baseline characteristics
25.5
26.0
26.5
27.0
N
ev
er E
x
C
ur
re
nt
BMI, kg/m2
0.8
0.9
0.9
1.0
N
ev
er E
x
C
ur
re
nt
Waist/hip ratio
128.0
130.0
132.0
134.0
136.0
N
ev
er Ex
C
ur
re
nt
SBP, mmHg
78.0
80.0
82.0
84.0
86.0
N
ev
er E
x
C
ur
re
nt
DBP, mmHg
5.8
5.9
5.9
6.0
6.0
N
ev
er
E
x
C
ur
re
nt
Total cholesterol, mmol/l
4.3
4.4
4.5
4.6
4.7
N
ev
er Ex
C
ur
re
nt
Non-HDL-C, mmol/l
1.2
1.3
1.4
1.5
1.6
N
ev
er E
x
C
ur
re
nt
HDL-C, mmol/l
0.1
0.2
0.3
0.4
0.5
N
ev
er E
x
C
ur
re
nt
Log triglycerides (mmol/l)
1.4
1.5
1.5
1.6
1.6
N
ev
er E
x
C
ur
re
nt
Apolipoprotein A1, g/l
1.0
1.1
1.1
1.1
1.1
1.1
N
ev
er E
x
C
ur
re
nt
Apolipoprotein B, g/l
2.2
2.3
2.3
2.4
2.4
N
ev
er E
x
C
ur
re
nt
Log lipoprotein (a)
0.0
0.2
0.4
0.6
0.8
N
ev
er E
x
C
ur
re
nt
Log C-Reactive Protein (mg/l)
8.5
9.0
9.5
N
ev
er Ex
C
ur
re
nt
Male Female
Fibrinogen, µmol/l
Smoking status
43.0
43.5
44.0
44.5
45.0
N
ev
er E
x
C
ur
re
nt
Albumin, g/l
1.8
1.8
1.9
1.9
2.0
2.0
N
ev
er Ex
C
ur
re
nt
Log leucocyte counts (x10^9/l)
0.2
0.3
0.4
0.5
0.6
N
ev
er Ex
C
ur
re
nt
Log Interleukin 6 (ng/l)
M
ea
n
(9
5%
C
I)
Mean risk factor levels were adjusted to age 50 years. Error bars represent 95% confidence
intervals. Mean risk factors were estimated using a linear mixed model. The model was
adjusted for cohort, age, age2, sex, age x sex, age2 x sex, smoking status x sex and smoking
status x age (where x denotes and interaction). Coefficients that were allowed to vary
randomly across studies were age, age2, sex and smoking status. The number of studies and
number of individuals included in each graph can be found in the first two columns of Table 2.
99
Figure 3.9: Cross-sectional association between levels of smoking pack-years in
current smokers and continuous baseline characteristics
21.5
23.5
25.4
27.3
29.2
0 10 20 30 40 50 60
BMI (kg/m2)
0.02 (0.01 to 0.03)
.8
.9
.9
.9
1
0 10 20 30 40 50 60
Waist/hip ratio
0.09 (0.07 to 0.11)
115
124
134
143
152
0 10 20 30 40 50 60
SBP (mmHg)
0.00 (-0.01 to 0.02)
70
75
80
85
90
0 10 20 30 40 50 60
DBP (mmHg)
-0.00 (-0.02 to 0.01)
4.7
5.3
5.8
6.4
7.0
0 10 20 30 40 50 60
Total cholesterol (mmol/l)
0.03 (0.02 to 0.05)
M
ea
n
ris
k
fa
ct
or
le
ve
l(
95
%
C
I)
-1.0
-0.5
0.0
0.5
1.0
0 10 20 30 40 50 60
Non-HDL-C (mmol/l)
0.06 (0.05 to 0.07)
.9
1.1
1.3
1.5
1.7
0 10 20 30 40 50 60
HDL-C (mmol/l)
-0.06 (-0.07 to -0.05)
-1.0
-0.5
0.0
0.5
1.0
0 10 20 30 40 50 60
Loge triglycerides (mmol/l)
0.07 (0.06 to 0.08)
1.2
1.3
1.4
1.6
1.7
0 10 20 30 40 50 60
Apolipoprotein A1 (g/l)
-0.06 (-0.09 to -0.03)
.8
1
1.1
1.3
1.4
0 10 20 30 40 50 60
Apolipoprotein B (g/l)
0.06 (0.03 to 0.09)
1
1.7
2.3
2.9
3.5
0 10 20 30 40 50 60
Loge Lp(a) (mg/l)
-0.01 (-0.07 to 0.05)
-0.3
-0.3
0.8
1.3
1.9
0 10 20 30 40 50 60
Loge CRP (µmol/l)
0.14 (0.10 to 0.17)
7.4
8.5
9.6
10.7
11.8
0 10 20 30 40 50 60
Fibrinogen (µmol/l)
0.11 (0.08 to 0.13)
39.8
41.6
43.3
45.1
46.8
0 10 20 30 40 50 60
Albumin (g/l)
-0.04 (-0.05 to -0.02)
1.7
1.8
2
2.1
2.2
0 10 20 30 40 50 60
Loge leukocyte counts
0.17 (0.13 to 0.21)
0.0
0.3
0.6
0.9
1.2
0 20 40 60 80
Loge interleukin 6
0.13 (0.09 to 0.17)
Male Female
Mean for tenths of smoking pack-years
Mean risk factor levels were adjusted for age 50 years. The values above each figure
correspond to the age- and sex- adjusted partial correlation coefficient (95% CI) between risk
factor and number of pack-years in current smokers, males and females combined. Error bars
represent the 95% CIs. Mean risk factors were estimated using a linear mixed model. The
model was adjusted for cohort, age, age2, sex, age x sex, age2 x sex, risk factor x sex and
risk factor x age (where x denotes and interaction). Coefficients that were allowed to vary
randomly across studies were age, age2, sex and risk factor. The number of studies and
number of individuals included in each graph can be found in Table 2.
100
Figure 3.10: Cross-sectional association between number of years since quitting
smoking and continuous baseline characteristics
22.4
24.4
26.4
28.4
30.5
0 10 20 30 40
BMI (kg/m2)
.8
.9
.9
.9
1
0 10 20 30 40
Waist/hip ratio
119
128
137
146
156
0 10 20 30 40
SBP (mmHg)
72
77
83
88
94
0 10 20 30 40
DBP (mmHg)
4.8
5.3
5.9
6.4
0 10 20 30 40
Total cholesterol (mmol/l)
3.4
3.9
4.5
5
5.6
0 10 20 30 40
Non-HDL-C (mmol/l)
1
1.2
1.3
1.5
1.7
0 10 20 30 40
HDL-C (mmol/l)
-0.5
0.5
-0.2
0.6
0.9
0 10 20 30 40
1.2
1.3
1.5
1.6
1.7
0 10 20 30 40
Apolipoprotein A1 (g/l)
.8
1
1.1
1.3
1.4
0 10 20 30 40
Apolipoprotein B (g/l)
1
1.6
2.3
2.9
3.5
0 10 20 30 40
-0.5
0.5
-0.5
1.2
1.7
0 10 20 30 40
6.9
8
9.1
10.1
11.2
0 10 20 30 40
Fibrinogen (µmol/l)
39.7
41.5
43.3
45.1
46.9
0 10 20 30 40
Albumin (g/l)
1.6
1.7
1.8
1.9
2.1
0 10 20 30 40
-0.5
0.5
-0.2
0.8
1.1
0 10 20 30 40
Male Female
Mean for tenths of years since stopping smoking
Log lipoprotein (a) Log C-Reactive
Protein (mg/l)
Log leucocyte
counts (x10^9/l) Log Interleukin 6 (ng/l)
M
ea
n
(9
5%
C
I)
Log triglycerides
(mmol/l)
Mean risk factor levels were adjusted for age 50 years. The values above each figure
correspond to the age- and sex- adjusted partial correlation coefficient (95% CI) between risk
factor and number of pack-years in current smokers, males and females combined. Error bars
represent the 95% CIs. Mean risk factors were estimated using a linear mixed model. The
model was adjusted for cohort, age, age2, sex, age x sex, age2 x sex, risk factor x sex and
risk factor x age (where x denotes and interaction). Coefficients that were allowed to vary
randomly across studies were age, age2, sex and risk factor.
101
Table 3.1: Summary of demographic and lifestyle covariates by smoking status at baseline
Smoking status Smoking pack-years in current smokers
No of
studies
No of
subjects
Current
smokers
Ex-
smokers
Never
smokers
No of
studies
No of
subjects
Difference (95% CI) in
number of smoking pack-
years compared to the
reference category
Age (years), mean (SD) 114 929,335 53.3 (8.5) 55 (8.6) 54.3(10.2) 44 82922 4.51 (3.67 to 5.35)*
Sex, % 114 929,335 44 82922
Male 476,809 41% 28% 32% 50059 1
Female 452,526 27% 18% 55% 32863 -5.18 (-17.31 to 6.96)
Race, % 90 483,114 37 47099
White 414,085 38% 24% 38% 36637 1
Non white 69,029 33% 24% 43% 10462 -9.22 (-12.07 to -6.36)
Alcohol status, % 86 432,223 37 69297
Not current 153,086 32% 21% 48% 12960 1
Current 279,137 38% 28% 34% 56337 -0.09 (-0.80 to 0.62)
History of diabetes, % 103 749,098 39 76858
No 714,983 36% 26% 38% 74509 1
Yes 34,115 23% 36% 42% 2349 0.89 (-0.07 to 1.84)
Level of education reached, % 59 337,521 26 57483
No schooling 12,863 29% 19% 52% 3146 1
Primary 63,474 36% 23% 41% 12522 0.86 (-0.70 to 2.42)
Secondary 167,910 41% 25% 34% 31365 0.66 (-0.46 to 1.77)
Vocational/University 93,274 30% 31% 39% 10450 -1.05 (-2.31 to 0.21)
Occupation or job, % 51 327,152 25 36228
Not working 81,454 19% 14% 67% 7835 1
Manual 84,173 38% 17% 44% 10034 1.08 (-0.54 to 2.69)
Office 9,401 65% 19% 15% 815 -1.37 (-3.18 to 0.43)
Service 120,953 27% 21% 52% 8776 -0.65 (-3.15 to 1.84)
Student 546 33% 24% 43% 62 -2.74 (-7.24 to 1.77)
Other 30,625 35% 30% 35% 8706 -1.53 (-2.97 to -0.09)
Change in number of smoking pack-years compared to reference category - except for age where it is per 1 standard deviation higher levels of age -,
adjusted for age and sex, pooled across studies using random effects meta-analysis.
102
Table 3.2: Summary of anthropometric covariates, blood pressure, lipids and inflammatory markers, by smoking status at baseline
Correlation with smoking status Correlation with smoking pack-years in currentsmokers
No of
studies
No of
subjects
Mean (SD) or %
in
current
smokers
Mean (SD) or %
in
ex-smokers
Mean (SD) or %
in
never smokers
No of
studies
No of
subjects
Difference (95% CI) in row
variable per 1 SD increase
in number of smoking pack-
years
Anthropometry
Body Mass Index (kg/m2) 114 929,335 25.4 (3.8) 26.4 (4) 26.3 (4.2) 44 82922 0.07 (0.02 to 0.13)
Waist to Hip Ratio 46 154,674 0.888 (0.093) 0.905(0.090)
0.866
(0.090) 21 20203 0.01 (0.00 to 0.01)
Blood pressure
Systolic blood pressure 111 754,077 133.9 (18.4) 137.3 (18.3) 136.5 (19.5) 44 80905 0.10 (-0.18 to 0.39)
Diastolic blood pressure 111 755,576 80.9 (11.1) 82.8 (10.8) 82.2 (10.9) 44 80879 -0.04 (-0.21 to 0.13)
Lipid factors
Total cholesterol (mmol/l) 111 738,215 5.8 (1.1) 5.9 (1.1) 5.8 (1.1) 43 79994 0.04 (0.02 to 0.06)
Non-HDL-C (mmol/l) 93 381,742 4.5 (1.2) 4.5 (1.1) 4.4 (1.1) 39 63904 0.07 (0.05 to 0.09)
HDL-C (mmol/l) 93 382,079 1.3 (0.4) 1.3 (0.4) 1.4 (0.4) 39 63972 -0.02 (-0.03 to -0.02)
Loge triglyceride (mmol/l) 93 574,463 0.4 (0.5) 0.3 (0.5) 0.3 (0.5) 37 59540 0.04 (0.03 to 0.05)
Apo A1 (g/l) 28 111,370 1.4 (0.3) 1.5 (0.3) 1.5 (0.3) 11 11667 -0.02 (-0.03 to -0.01)
Apo B (g/l) 28 113,902 1.1 (0.3) 1.1 (0.3) 1.1 (0.3) 11 11273 0.02 (0.01 to 0.03)
Loge Lp(a) 29 90,111 2.3 (1.3) 2.3 (1.2) 2.3 (1.2) 10 8286 -0.01 (-0.08 to 0.05)
Inflammatory markers
Loge CRP (mg/l) 48 107,879 0.8 (1.1) 0.6 (1.1) 0.4 (1.1) 12 5471 0.18 (0.11 to 0.24)
Fibrinogen (μmol/l)  44 180,379 9.6 (2.2) 9.1 (2.1) 8.9 (2) 14 22768 0.22 (0.16 to 0.28)
Albumin (g/l) 35 122,447 43.3 (3.5) 43.3 (3.6) 43.1 (3) 13 16988 -0.13 (-0.20 to -0.06)
Loge leucocyte
count(x10^9/l) 33 109,974 2 (0.3) 1.8 (0.3) 1.8 (0.3) 14 13229 0.05 (0.04 to 0.07)
Loge Interleukin-6 (ng/l) 10 16,205 0.6 (0.6) 0.5 (0.7) 0.4 (0.6) 6 2283 0.08 (0.05 to 0.10)
Mean and SD for current, ex and never smokers were calculated within studies and pooled across studies using random effect meta-analysis. Change in
row variable per 17 additional pack-years, adjusted for age and sex, and pooled across studies using random effect meta-analysis.
103
References
1. McBride PE. The health consequences of smoking. Cardiovascular diseases. Med Clin
North Am. 1992;76:333-353.
2. Ambrose JA, Barua RS. The pathophysiology of cigarette smoking and cardiovascular
disease: an update. J Am Coll Cardiol. 2004; 43:1731-1737.
3. Cullen P, Schulte H, Assmann G. Smoking, lipoproteins and coronary heart disease
risk. Data from the Munster Heart Study (PROCAM). Eur Heart J. 1998;19:1632-1641.
4. Njolstad I, Arnesen E, Lund-Larsen PG. Smoking, serum lipids, blood pressure, and
sex differences in myocardial infarction. A 12-year follow-up of the Finnmark Study.
Circulation. 1996;93:450-456.
5. Smith CJ, Fischer TH. Particulate and vapor phase constituents of cigarette
mainstream smoke and risk of myocardial infarction. Atherosclerosis. 2001;158:257-
267.
6. Kannel WB, D'Agostino RB, Belanger AJ. Fibrinogen, cigarette smoking, and risk of
cardiovascular disease: insights from the Framingham Study. Am Heart J.
1987;113:1006-1010.
7. Liu J, Liang Q, Frost-Pineda K et al. Relationship between Biomarkers of Cigarette
Smoke Exposure and Biomarkers of Inflammation, Oxidative Stress, and Platelet
Activation in Adult Cigarette Smokers. Cancer Epidemiol Biomarkers Prev.
2011;20:1760-1769.
8. Kaptoge S, White IR, Thompson SG et al. Associations of plasma fibrinogen levels with
established cardiovascular disease risk factors, inflammatory markers, and other
characteristics: individual participant meta-analysis of 154,211 adults in 31 prospective
studies: the fibrinogen studies collaboration. Am J Epidemiol. 2007;166:867-879.
9. Gayen AK. The frequency distribution of the product-moment correlation coefficient in
random samples of any size drawn from non-normal universes. Biometrika.
1951;38:219-247.
10. Riley RD, Higgins JP, Deeks JJ. Interpretation of random effects meta-analyses. BMJ.
2011;342.
11. Thun, M. J., Day-Lally, C., Myers, DG, Calle, EE, Flanders, WD, Zhu, B-P, Namboodiri,
MM, and Heath, CW Jr. Trends in tobacco smoking and mortality from cigarette use in
Cancer Prevention Studies I (1959 through 1965) and II (1982 through 1988). Changes
in Cigarette-Related Disease Risks and Their Implication for Prevention and
Control.Smoking and Tobacco Control Monograph No.8.Rockville (MD) [a], 305-382.
1997. Shopland DR, Burns DM, Garfinkel L, Samet JM.
12. World Health Organization. The Tobacco Atlas. 2002.
13. Pierce JP, Messer K, White MM, Cowling DW, Thomas DP. Prevalence of heavy
smoking in California and the United States, 1965-2007. JAMA. 2011;305:1106-1112.
14. Kandel DB, Chen K. Extent of smoking and nicotine dependence in the United States:
1991-1993. Nicotine Tob Res. 2000;2:263-274.
15. Berrendero F, Robledo P, Trigo JM, Martin-Garcia E, Maldonado R. Neurobiological
mechanisms involved in nicotine dependence and reward: participation of the
endogenous opioid system. Neurosci Biobehav Rev. 2010;35:220-231.
16. Thun MJ, Apicella LF, Henley SJ. Smoking vs other risk factors as the cause of
smoking-attributable deaths: confounding in the courtroom. JAMA. 2000;284:706-712.
104
17. Kenfield SA, Stampfer MJ, Rosner BA, Colditz GA. Smoking and smoking cessation in
relation to mortality in women. JAMA. 2008;299:2037-2047.
18. Nettle D. Large Differences in Publicly Visible Health Behaviours across Two
Neighbourhoods of the Same City. PLoS One. 2011;6:e21051.
19. Niederdeppe J, Kuang X, Crock B, Skelton A. Media campaigns to promote smoking
cessation among socioeconomically disadvantaged populations: what do we know,
what do we need to learn, and what should we do now? Soc Sci Med. 2008;67:1343-
1355.
20. Nichter M, Nichter M, Carkoglu A, Lloyd-Richardson E. Smoking and drinking among
college students: "it's a package deal". Drug Alcohol Depend. 2010;106:16-20.
21. Barrett-Connor E, Khaw KT. Cigarette smoking and increased central adiposity. Ann
Intern Med. 1989;111:783-787.
22. Lissner L, Bengtsson C, Lapidus L, Bjorkelund C. Smoking initiation and cessation in
relation to body fat distribution based on data from a study of Swedish women. Am J
Public Health. 1992;82:273-275.
23. Akbartabartoori M, Lean ME, Hankey CR. Relationships between cigarette smoking,
body size and body shape. Int J Obes (Lond). 2005;29:236-243.
24. Symons JD, Stebbins CL. Hemodynamic and regional blood flow responses to nicotine
at rest and during exercise. Med Sci Sports Exerc. 1996;28:457-467.
25. Benowitz NL, Kuyt F, Jacob P, III. Influence of nicotine on cardiovascular and hormonal
effects of cigarette smoking. Clin Pharmacol Ther. 1984;36:74-81.
26. Craig WY, Palomaki GE, Haddow JE. Cigarette smoking and serum lipid and
lipoprotein concentrations: an analysis of published data. BMJ. 1989;298:784-788.
27. U.S.Department of Health and Human Services. How tobacco smoke causes disease:
the biology and behavioral basis of smoking-attributable disease: a report of the
surgeon general. 2010.
28. Bazzano LA, He J, Muntner P, Vupputuri S, Whelton PK. Relationship between
cigarette smoking and novel risk factors for cardiovascular disease in the United States.
Ann Intern Med. 2003;138:891-897.
29. Danesh J, Lewington S, Thompson SG et al. Plasma fibrinogen level and the risk of
major cardiovascular diseases and nonvascular mortality: an individual participant
meta-analysis. JAMA. 2005;294:1799-1809.
30. Smith FB, Lee AJ, Fowkes FG, Price JF, Rumley A, Lowe GD. Hemostatic factors as
predictors of ischemic heart disease and stroke in the Edinburgh Artery Study.
Arterioscler Thromb Vasc Biol. 1997;17:3321-3325.
31. Kaptoge S, Di AE, Lowe G et al. C-reactive protein concentration and risk of coronary
heart disease, stroke, and mortality: an individual participant meta-analysis. Lancet.
2010;375:132-140.
32. Caraballo RS, Giovino GA, Pechacek TF et al. Racial and ethnic differences in serum
cotinine levels of cigarette smokers: Third National Health and Nutrition Examination
Survey, 1988-1991. JAMA. 1998;280:135-139.
105
Chapter 4: Cigarette smoking and the risk of cardiovascular
diseases, lung cancer deaths and all-cause mortality in
developed countries
Summary
Cigarette smoking is a well-known risk factor for cardiovascular diseases thanks to
numerous prospective cohort studies, mainly conducted in the developed world. The
present analyses used the Emerging Risk Factors Collaboration and aimed to
synthesize the available evidence on the association between smoking and the risk
of CVD in the developed world using appropriate adjustment for cardiovascular risk
factors. In addition, they investigated areas of this association which remain
controversial or based on insufficient evidence, such as the presence of a higher risk
in female versus male current smokers; and the association of smoking with
subtypes of CVD. They also contained novel analyses on potential interplays
between smoking and other predominant risk factors with CVD risk; and an
estimation of risk of life lost to smoking in women. Finally, these analyses looked at
the effect of stopping smoking, in terms of the number of years necessary for an ex-
smoker to reach the level of CVD risk of a never smoker. For comparison,
associations with lung cancer deaths were also presented.
The ERFC had data with smoking status from 114 prospective cohort studies on
929,335 individuals, with an average follow up of 14.2 years; who experienced in
total 46,576 cardiovascular deaths. After adjustment for age and body mass index,
the risk ratios for current versus never smokers were for coronary heart diseases
1.99 (95% CI: 1.86, 2.13), and for cardiovascular deaths 2.01 (1.89, 2.14). All types
of CVD, including pulmonary embolism, heart failure and cardiac dysrhythmia were
associated with smoking. The effect of smoking was strongest on the risk of fatal
aortic aneurysm (OR: 4.34; 3.33 to 5.66) and deaths from peripheral vascular
diseases (OR: 3.81; 2.67 to 5.45). There was no evidence of confounding by other
lifestyle and biochemical risk factors. Associations were stronger in younger age
groups and in women, and for myocardial infarction, in non-diabetics. However, in
absolute terms, more men and diabetics experienced a CVD event as the result of
smoking than women and non-diabetics. Associations with numbers of amount and
duration were curvilinear and converged toward a plateau above approximately 20
pack-years for the risk of CHD and all CVD. Cessation was rapidly beneficial, with an
106
80% decrease in CVD risk within the first 5 years of cessation. However, the risk only
became non-significant after 20 years of cessation. Men who were current smokers
and 50 years old lost an average of 7.5 years, and women of the same age lost 4.6
years due to smoking. These results confirm and enhance previous epidemiological
evidence on the association between cigarette smoking and CVD.
107
4.1 Background
Smoking remains the leading preventable cause of disease, disability and death in
developed countries, and a large majority of smokers are cigarette smokers.
Considerable epidemiological evidence has been accumulated on the association
between cigarette smoking and the risk of cardiovascular diseases and death 1.
Recently, a very large-scale analysis reported on the risk associated with smoking
and the benefits of cessation using a database which included 1.3 million women
recruited between 1996 and 2001 and follow-up for up to 8 years 2. However, some
areas have not been investigated in enough detail and others remain controversial.
Firstly, some types of CVD are rarer than others and studies have often been
underpowered to investigate the effect of smoking on these types, namely
subarachnoid haemorrhage, aortic aneurysm, heart failure, cardiac dysrhythmia,
pulmonary embolism and peripheral vascular disease. Current evidence is mostly
derived from literature based meta-analysis with relatively few cases 3-5, which
sometimes also include retrospective studies subject to recall bias 6, and are
generally unable to consistently adjust for other risk factors of CVD across
substantially heterogeneous studies 7.
A second area of uncertainty relates to the interplay between cigarette smoking and
other well known risk factors of CVD. At a time when the prevalence of glycaemia
and diabetes is rising because of general ageing of the population 8, obesity is
spreading as a result of poor diet and low physical activity 9, and total cholesterol
remains highest worldwide in high income countries 10, it is important to assess the
joint effect with smoking of these well known risk factors of CVD 11-13. The presence
of an effect modification by sex also remains controversial. It has been suggested by
a recent meta-analysis that women experience an increase in risk 25% higher when
smoking than men 14. However, this meta-analysis was literature based, adjustment
was inconsistent across studies included, and when pooling the age-only adjusted
RR ratios, the effect modification by sex was non-significant.
Thirdly, the shape of association between smoking dose, smoking duration and the
risk of CVD is still controversial. A plateau effect at high smoking intensities has been
found in some but not all studies 15. The effect of duration of smoking has been
difficult to assess as it is nearly collinear with age, most smokers starting in their
teenage years 16. The common use of smoking pack-years to quantify total exposure
108
to smoking makes the assumption that an increase of 1 cigarette per day is
equivalent to 1 additional year of smoking and, whilst it may be appropriate as a
crude measure of total exposure to smoking and when assessing the risk of lung
cancer 17, 18, its relevance when assessing CVD risk has yet to be demonstrated.
Fourthly, the rapidity of the benefits obtained by cessation is uncertain. Some studies
have estimated a halving in CHD risk within the first 3 to 5 years 19, 20 whilst others
have observed a slower decline 21. The estimated time for levels of CVD risk to
come back to that of a never smoker have also oscillated between 10 years 20 and
more than 20 years 22. In the ERFC, information was available in 80,318 individuals
from 54 studies on number of years since quitting smoking, enabling more reliable
estimates of risk reduction than previously possible.
For continuing smokers, quantifying the reduction in life expectancy as a result of
smoking relies on precise estimates of its association with disease and death, and in
particular with CVD. Published estimates of an average of 10 years of life lost as a
result of smoking were based on a study of British Male doctors born before 1930,
with lifelong smoking habits likely to differ from more recent generations 23. These
estimates were also based on men only. Nowadays, smoking rates in women are
close to or have reached the same rates as men in some Western countries 24. As
women tend to live longer than men and have been conversely shown to experience
greater increase in risk when smoking 14, the question of years of life lost to smoking
needs to be addressed in women. A recent study estimated a similar number of
years of life lost for women compared to men but this finding has not yet been
replicated 2.
The objective of this Chapter is to address these uncertainties and provide better
estimates of risk of CVD than previously possible using data from up to 114
prospective cohort studies including nearly 1 million participants. The main focus of
this Chapter is on CVD, but some sections present results on lung cancer deaths for
comparison. In addition, the effect of smoking on life expectancy is considered, and
the respective share of CVD and cancer in tobacco related burden is assessed using
ERFC findings.
109
4.2 Methods
4.2.1 The dataset
The subset of the Emerging Risk Factors Collaboration (ERFC) dataset with
information on smoking has been described in Chapter 2. There were 114 studies
with information on smoking status which including 929,335 individuals, 103 studies
with information on smoking pack-years (defined as the number of packs of 20
cigarettes per day times the number of years), 84 studies for amount, 44 studies for
duration, 44 studies for starting age and 54 studies for stopping age, also expressed
as number of years since quitting smoking. Three studies conducted in developing
countries were excluded from the analysis. Harmonization of the coding for smoking
status into 3 categories, current, ex- and never smokers, was done by the data
management team, in correspondence with the individual study investigators. This
analysis was done, combining all type of smoking, because some studies did not
provide information on type of smoking product used, and in order to maximize
power. However, 72% of individuals had information on cigarette smoking status and
information was mostly concordant: over 90% of current smokers of any type were
current cigarette smokers and over 98% of ex-smokers of any type were ex-
cigarette-smokers. When smoking status was provided by the studies with no
mention of the type of smoking, cigarette smoking was assumed.
4.2.2 Statistical methods
Details of the statistical methods have been reported previously 25. Following the
example of previous reports published by the ERFC 26-31, I assessed associations of
smoking status, pack-years, type, cessation and CVD. Participants were censored at
the first non-fatal MI or stroke and at any cause of death. Sensitivity analyses
excluding studies which recorded only fatal CVD events were conducted when
comparing the effect for fatal versus non-fatal MI or stroke. To avoid reverse
causality, sensitivity analyses were done excluding the first 5 years of follow-up.
Relative risks (RRs) were estimated using a 2-stage approach. In the first stage, RRs
were calculated separately for each study. For prospective cohort studies and trials,
a Cox proportional hazards regression model was fitted to estimate the hazard ratio
for each study, stratifying, where appropriate, by sex and trial arm 32. Separately for
110
each study s=1 …S, with strata k=1 …Ks (for most studies Ks=2, just for the two
sexes) and individuals i=1 …ns, with exposure of interest Esi and other covariate Xsi,
the hazard at time t after baseline was modeled as:
The s is the parameter of interest, and corresponds to the log of the hazard ratios
per unit increase in continuous variables for study s, or for one category versus the
reference group in the case of categorical variables. For smoking status, never
smokers were chosen as the reference group. To check whether the assumption of
proportional hazards held, a time-dependent interaction term was added to the
model for each study and its pooled estimate was tested for significance. For nested
case-control studies, odds ratios were calculated fitting a logistic regression rather
than a Cox model. Odds ratios were assumed to approximate hazard ratios in that
case, which is reasonable because cardiovascular disease incidence can be
considered as rare 33. To characterize shapes of associations for continuous
variables such as pack-years, years since quitting and age stopped, RRs were
computed by quintiles or predefined categories of the distribution. To investigate
effect modification, interaction terms were added to the model. For the sex (or race)
interactions, the dataset was reduced to studies recording both Males and Female
(both White and non-White), in order to avoid spurious differences in hazard ratios
due to between-studies heterogeneity rather than within-studies heterogeneity.
The second stage involved a multivariate random-effects meta-analysis of study-
specific hazard ratios or odds ratios, to obtain an overall “risk ratio”. Compared to
fixed effects, a random effects meta-analysis incorporates heterogeneity between
studies 25, 34. A fixed effects model assumes that a single parameter RR is common
to all studies, while a random effects model makes the assumption that the
underlying RR follows a random distribution. The latter is more appropriate for
studies such as the ones involved in the ERFC, which differ considerably in design
and beyond measurable covariates included in the model (for example, decade of
recruitment, country of recruitment, study design, etc.), and are therefore likely to
truly vary in terms of underlying RR. Writing the variance of the estimated s as vs,
the random-effects meta-analysis model is given by:
111
where  represents the average log HR. The true heterogeneity between studies,
beyond variation in the estimates due to chance, is represented by the variance 2 ,
and the percentage of variance in the estimates that is attributable to between-study
variation as opposed to sampling variation is expressed by I2 statistic 35. An I2 close
to 0% suggests that variability in study estimates is entirely due to chance, and an I2
>50% suggests substantial heterogeneity between studies. RR pooled by random
effect meta-analysis provides an estimate of the average or, in other terms, expected
RR of a study (with allowance for random noise around this average), rather than the
common RR across studies which would be provided assuming a fixed effects
model. Specific sources of heterogeneity were explored by fitting interactions with
study-specific risk factors such as decade of enrollment and region of the world
(Western Europe, North America or other).
Another aspect of the regressions and meta-analyses performed was that they were
both multivariate. Compared to univariate models, multivariate models take into
account the correlation between the exposure of interest and covariates. Multivariate
meta-analysis is a meta-analysis of  of main effect and coefficients of covariates in
study-specific models at the same time 36. Multivariate meta-analyses are well
designed for my analysis, because I can assume that the effect of current versus
never smoking and the effect of ex versus never smokers will be correlated. They are
also useful for shape analyses, as coefficients for different quintiles are likely to be
correlated; and when testing for interaction, as interactions and main effects are
likely to be correlated. To reduce excessive computation time, the method of moment
was used to estimate pooled RRs by multivariate meta-analysis, rather than
maximum likelihood methods 36.
To assess shapes, pooled estimates within quintiles of continuous variables were
plotted against the pooled mean of the quintile, obtained by random effect univariate
meta-analysis of study-specific means. For example, for number of years since
quitting smoking in relation to CVD risk, quintiles were defined using individuals who
declared being ex-smokers at baseline and provided information on number of years
since stopping smoking. For the association with number of pack-years smoking,
quintiles were defined using baseline pack-years information available for current
smokers.
112
To enable graphical comparison of RRs between any two groups, and not only with
the reference group, 95% confidence intervals for the RRs were represented using
“floating absolute” variances, and the size of the box representing RRs was chosen
to be proportional to the inverse of the ‘”floating variance” 37. RRs were at least
adjusted for age and body mass index. To explore potential biological pathways’
underlying associations, hazard ratios for smoking status were further adjusted for
systolic blood pressure, history of diabetes, BMI, waist circumference, waist-to-hip
ratio, total and high density lipoprotein cholesterol, triglyceride, C-reactive protein,
fibrinogen, alcohol consumption, or socioeconomic indicators, i.e. educational
attainment and occupational category. The association with number of years since
stopping smoking was further adjusted for age starting smoking and past number of
pack-years. For the purpose of investigating interactions, RRs were adjusted for age,
body mass index, history of diabetes, systolic blood pressure and total cholesterol,
and the dataset was reduced to individuals with information on all these covariates.
To estimate absolute rather than relative risks, event rates were estimated in the
ERFC dataset by bands of 5 years of age at risk (40-45, 45-50, 50-55, 55-60, 60-65,
65-70, 70-75, 75-80, 80-85, 85+) and by sex, for current and never smokers
separately, on the model of the Prospective Studies Collaboration 38. Age at risk is
defined as age of occurrence of an event and requires the data to be split into 5 year
age bands so that individuals contribute to follow-up until they are censored or until
they experience a first CVD event or death. Practically, event rates at 40 years or
older were calculated by applying RR adjusted for age, BMI, history of diabetes, total
cholesterol and systolic blood pressure, and stratified, where appropriate, by sex and
trial arm, to the event rate in the ERFC in current smokers. For Body Mass Index,
RRs were only adjusted for age at baseline. Never smokers were chosen as the
reference group, except for sex, where current smokers were chosen as the
reference group. RRs were based on 21,946 fatal and non-fatal MI (15,759 events
for the graph showing the interaction with sex, selecting only studies including both
Males and Females) and 10,243 cerebrovascular events (7,732 events for the graph
showing the interaction with sex). The age at risk and sex standardized event rate for
fatal and non-fatal MI was 0.00216 in never smokers and 0.004371 for current
smokers in the ERFC dataset. The age at risk and sex standardized event rate for
fatal and non-fatal cerebrovascular events was 0.001204 for never smokers and
0.00201 for current smokers.
113
Age at risk and sex specific rates were combined into an overall rate using
calibrations provided by the European standard population (a notional population of
2 million people commonly used to standardise rates) 39. In a third step, these rates
were applied to estimated RRs.
I followed the example of a previous ERFC report on diabetes to estimate survival
curves for current, ex- and never smokers and corresponding years of life lost 30.
Briefly, estimates of cumulative survival from 35 years of age and older among
current, ex and never smokers at baseline were calculated by applying hazard ratios,
specific to age at risk and sex, for cause-specific mortality associated with smoking
status, to cause-specific rates of death at 35 years of age and older, for residents of
the European Union. Analyses were carried out in Stata release 11 (StataCorp).
4.3 Results
Description of smoking variables and correlations with other baseline characteristics
has been presented in Chapter 3. Information was available in up to 114 prospective
cohort studies including 929,335 individuals; who experienced 40,218 incident CHD
events including 16,390 non-fatal MI and 11,585 coronary deaths; 17,445 strokes
including 8,866 non-fatal and 8,579 fatal strokes; and 11,511 other vascular events,
during an average follow-up time of 14 years.
4.3.1 Current versus never smoking and the risk of CVD
Current smokers experienced roughly a doubling in risk of fatal and non-fatal CVD.
RRs adjusted for age, BMI, sex and study, were 2.01 (1.89, 2.14) for all
cardiovascular deaths and 1.99 (1.86, 2.13) for fatal and non-fatal MI, and 1.64
(1.54; 1.75) for all cerebrovascular events, compared to never smokers (Figure 4.1).
Restricting the data to studies which recorded both fatal and non-fatal MI, RR was
1.89 (1.74; 2.06) for non-fatal MI and 2.10 (1.84; 2.39) for fatal MI (Table 4.1).
Subsidiary analyses excluding the first 5 years of follow-up, and censoring at death
rather than at first cardiovascular event or death only, did not materially impact the
RRs. RRs for ischaemic stroke and haemorrhagic stroke were nearly the same (1.72;
1.54 to 1.93 versus 1.64; 1.45 to 1.85), while RR for subarachnoid haemorrhage was
stronger: 2.69 (2.26; 3.20). Amongst other vascular diseases, the associations for
heart failure, sudden death, cardiac dysrhythmia and pulmonary embolism were all
significant with increases in risk comprising between 60% and 130%. The strongest
associations were observed for aortic aneurysm (RR: 4.34; 3.33-5.66), and
114
peripheral vascular deaths (RR: 3.81; 2.67-5.45). Further adjustment for blood
pressure, history of diabetes, alcohol consumption, levels of education and
occupation, as well as lipids and inflammatory markers did not appreciably alter
these associations (Table 4.2).
There was substantial heterogeneity between studies for the associations with fatal
and with non-fatal MI. Values of I2 were above 50%, indicating that most of the
variability between studies could not be explained by chance. Looking at study
specific determinants, I observed a small variation of RRs across regions of location
of the studies and according to decade of start of enrolment of the study (p-values of
interaction < 0.0001) (Figures 4.2 & 4.3). Pooled RRs were 2.02 (1.90; 2.19) in
Western European countries, 1.75 (1.53; 2.00) in North America, and 1.52 (1.41;
1.65) in “other countries” which grouped studies from Australia, New Zealand, Israel
and Japan. The association was 1.63 (1.44; 1.85) for studies started in the 1960s
while it was above 2.01 (1.78; 2.27) for studies started in the 1970s. It was 1.95
(1.79; 2.13) for studies started in the 1980s and 1.91 (1.66; 2.20) for studies started
in the 1990s.
The proportional hazard assumption was tested for departure from proportionality. P-
values were statistically significant for fatal and non-fatal MI and all CVD. However,
for these outcomes, RRs decreased only very slightly with the number of years of
follow-up. For fatal and non-fatal MI, the difference in the RR for current versus never
smokers was -0.10 (-0.13; -0.08) and for all CVD it was -0.09 (-0.12; -0.06) (Table
4.3).
Because of the relatively low number of events, it was not possible to investigate the
effect of smoking other than that smoking status presented above on the risk of
subtypes of CVD other than CHD and all cerebrovascular events.
4.3.2 Interplay of smoking with other risk factors on CVD risk
Individuals with a lower absolute risk, such as younger age groups, women, non-
diabetics and individuals with low blood pressure levels were most harmed by
smoking, in terms of CVD risk (Figure 4.4 & 4.5). Individuals ≥60 years old 
experienced a 27% (95%CI: 21%; 33%) lower RR of CHD for current versus never
smokers, compared to individuals <60 years old. The difference in pooled RR
between individuals aged ≥60 years versus <60 years old was 17% (95% CI: 9%; 
115
25%) for all cerebrovascular events. Regarding sex, women had a stronger increase
in risk when smoking compared to men by approximately 13%. However, the
interaction was only borderline significant (95% CI for CHD: 4%-21% and for stroke:
2%-21%). For diabetes, an effect modification of smoking status according to history
of diabetes was observed for CHD, with a 30% (21%; 38%) reduction in RR for
diabetics compared to non-diabetics; but not for cerebrovascular events. There was
no effect modification according to BMI or total cholesterol. The interaction with blood
pressure for CHD death and non-fatal MI was statistically significant (P-value<0.001)
but modification was marginal: per SD increase in SBP, smoking relation to CVD risk
decreased by less than 10%. Subsidiary analyses where the models testing for effect
modification with sex, diabetes, BMI, SBP and total cholesterol were adjusted for an
additional interaction term between age and smoking in order to take into account
potential confounding by an interaction with age yielded the same results.
The age at risk and sex standardized rate of fatal and non-fatal MI in the ERFC was
2.16/1,000 for never smokers and 4.37/1,000 for current smokers. For fatal and non-
fatal cerebrovascular events, rates were 1.20/1,000 for never smokers and
2.01/1,000 for current smokers. Using these rates and applying them to RRs for
current versus never smokers, a significantly greater number of CHD events were
caused by smoking amongst individuals with a higher absolute risk such as older age
groups, men, diabetics, and individuals in the top third of BMI, SBP and total
cholesterol (Figure 4.6a). Individuals with higher baseline risk (such as individuals in
the top third of SBP levels and diabetics) also experienced a greater increase in the
number of cerebrovascular events when smoking than individuals with lower
baseline risk (Figure 4.6b).
4.3.3 Dose-response relationship
Relationships between smoking intensity, duration and pack-years and the risk of
CHD were all non-linear (Figure 4.7). Individuals smoking 15-20 CPD had a RR of
2.50 (2.34; 2.66), 20-25 CPD a RR of 2.54 (2.30; 2.81) and >25 CPD a RR of 2.93
(2.63; 3.26), compared to never smokers. Even smoking at low intensities was
significantly associated with an approximate doubling in risk of CHD compared to
never smokers and was significantly higher than the risk of individuals who stopped
smoking: individuals smoking <10 CPD had a RR for CHD of 1.91(1.76; 2.08), and
ex-smokers had a RR of 1.20 (1.09; 1.32) when comparing to never smokers. The
association with duration was as follows: Individuals who had smoked <17 years had
116
a risk of 1.89 (1.68; 2.13), and individuals who had smoked 23-30 years experienced
a risk of 2.60 (2.40; 2.80), with reference to never smokers. The shape of association
for pack-years was approximately linear up to 25 pack-years, with a RR of 2.54
(2.33; 2.77) in individuals reporting 17-25 pack-years and then a tailoring above 25
pack-years, with reference to never smokers.
For cerebrovascular events, there was evidence of a dose-response relationship
between smoking intensity and cerebrovascular events, but somewhat less strong
than for CHD. Individuals smoking less than 10 CPD had a RR of 1.34 (1.09; 1.65)
for ischaemic stroke and 1.61 (1.20; 2.15) for haemorrhagic stroke, compared to
never smokers, and these were significantly greater than the risks for ex-smokers
(0.93; 0.76 to 1.16 for ischaemic stroke and 1.16; 0.83 to 1.62 for haemorrhagic
stroke). For 15-20 CPD, RRs were 2.00 (1.68; 2.36) for ischaemic stroke and 1.56
(1.14; 2.17) for haemorrhagic stroke. For 20-25 CPD, they were 2.12 (1.53; 2.93) for
ischaemic stroke and 2.42 (1.56; 3.78) for haemorrhagic stroke and for >25 CPD
1.65 (1.24; 2.19) and 2.45 (1.60; 3.74) respectively. The association with duration
was approximately flat, with RRs oscillating around 1.8 for ischaemic stroke and 2 for
haemorrhagic stroke. Regarding pack-years, a positive and approximately linear
association was apparent for ischaemic strokes and all cerebrovascular events
combined, apart from the top quintile where RRs were slightly lower compared to the
top 4th quintile.
4.3.4 Smoking cessation and CVD risk
Stopping smoking translated into clear benefits in terms of CVD risk when compared
to current smokers. Ex-smokers experienced an excess risk for CVD which varied
depending on the type of CVD and ranged from 4% to 44% (Figure 4.1). RRs were
1.13 (1.10; 1.17) for CHD, 1.17 (1.03; 1.33) for death from heart failure; 1.44 (1.17;
1.77) for death from aortic aneurysm; 1.04 (099; 1.09) for all cerebrovascular events;
1.04 (0.89; 1.22) for sudden death; 1.10 (0.97; 1.47) for deaths from cardiac
dysrhythmia; and 1.39 (0.91; 2.13) for deaths from peripheral vascular diseases.
Further adjustment for blood pressure, history of diabetes, lifestyle risk factors, lipids
and inflammatory markers did not appreciably alter these associations for CHD
(Table 4.4).
Cessation rapidly reduced CHD risk, from 1.95 (1.78; 2.13) in current smokers to
1.22 (1.14; 1.31) for ex-smokers who had stopped ≤3 years before enrolment in the 
117
study, and became non-significant only 20 years after cessation with a RR of 1.00
(0.93; 1.05) (Figure 4.8, please note that RRs were plotted against the median
number of years since stopping smoking, which for the category of ≥20 years was 
approximately 25 years before entry into the baseline). Individuals who had stopped
4-9 years before enrolment had a pooled RR of 1.21 (1.11; 1.31) and individuals who
had stopped 10-19 years before enrolment a RR of 1.18 (1.10; 1.26). For the risk of
all CVD deaths combined, RRs were 1.87 (1.74; 2.01) for current smokers, 1.23
(1.15; 1.32) for individuals who had stopped ≤3 years before enrolment, 1.15 (1.07; 
1.23) for those who had stopped 4-9 years before enrolment, 1.08 (1.03; 1.14) for
those who had stopped 10-19 years before enrolment and 0.95 (0.91; 1.00) for those
who had stopped 20 years and more before enrolment. The decrease in risk for lung
cancer deaths was more gradual and past smokers remained subjected to an excess
risk even 20 years after cessations. RRs were 12.78 (10.98; 14.86) for current
smokers, and respectively 8.99 (7.23; 11.18), 5.96 (5.03; 7.07), 4.21 (3.53; 5.03);
2.01 (1.64; 2.44) for individuals who had stopped 0-3, 4-9, 10-19 and 20+ years
before entry into the study. Further adjustment for age starting smoking and past
number of pack-years slightly attenuated the associations (Table 4.5).
4.3.5 Number of years of life lost attributable to smoking
Continuous smoking translated into a substantial loss in terms of years of life, for
both men and women. RRs for all causes of deaths were 2.02 (1.96; 2.08) in men
and 1.88 (1.82; 1.93) in women for current versus never smokers, with important
heterogeneity between studies for both sexes (I2 equal to 70% for men and 65% for
women). Amongst individuals who were resurveyed, the proportion of never smokers
who became current smokers during follow-up was below 1% even after 40 years
follow-up, the proportion of individuals who were current smokers and became ex-
smokers was <10% until 15 years of follow-up where it became >10%, and reached
28% in individuals who had been resurveyed 30 to 40 years after baseline entry
(Table 4.6). A small percentage (mostly ≤5%) of individuals had started and stopped 
smoking at the time of the next resurvey, which was indicated by a switch in their
status from never to ex-smokers.
At baseline age of 40, 50, 60 and 70 years old, men who were also current smokers
lost respectively 7.7 years, 7.5 years, 6 years and 4.6 years of life compared to never
smokers. The corresponding years of life lost for women smokers versus never
smokers at these ages were 5.2 years, 4.6 years, 4.2 years and 3.1 years. Median
118
survival was 70 years old for male current smokers whilst it was 77 years old for
male current smokers (Figure 4.9a). For women, median survival was 76 years old
amongst current smokers whilst it was 81 years old amongst never smokers. At
baseline age 50, in men, 28.5% of all deaths were attributable to CVD, 40.9% to
cancer deaths and 28.7% to deaths which were not vascular or cancer related and
the rest to unknown causes (Figure 4.9b). In women aged 50, vascular deaths
represented 30.0%, cancer deaths 35.7% and non-vascular non-cancer deaths
32.2%, the rest being from unknown causes. The proportion of deaths attributable to
CVD decreased with ageing while the proportion attributed to cancer deaths
remained similar and the proportion attributed to non-vascular non-cancer causes
increased, representing 51.7% at age 90 (31.1% for cancer deaths and 16.7% for
CVD deaths).
For ex- versus never smokers, the difference was lower but still represented
respectively 1.6 years, 1.5 years, 1.4 years and 1.1 years at ages 40, 50, 60 and 70
years old in men. In women with age 40, 50, 60 and 70 at baseline, the loss in
number of years of life was respectively 0.7, 0.65, 0.60 and 0.46. Median survival for
ex-smokers was 74 years old for men and 81 years for women (Figure 4.10a). A
majority of the excess deaths of ex- versus never smokers was attributed to causes
other than CVD. At age 50, in men, 23.7% of deaths amongst ex-smokers were due
to CVD, 51.7% to cancer and 23.7% to non CVD non cancer deaths. In women at
the same age, the proportions were 14% for CVD, 6.06% for cancers and 25.3% for
non-CVD non-cancer deaths. These proportions remained similar in different age
groups (Figure 4.10b).
4.4 Discussion
I studied smoking in relation to several major vascular and non-vascular outcomes in
approximately 1 million adults from developed countries. The present analysis
distinguishes itself by the large number of events and relatively long follow-up, with
more than 70,000 cardiovascular events occurring during 15 million person years
follow-up 40. The ERFC censored individuals at their first non-fatal MI and stroke
event as well as CVD mortality 20, 41. My dataset included more than three times as
many women participants as the largest study on the effect of smoking in women 20.
Contrasting with most published studies which have been questionnaire-based 20, 41,
42; information was available on several blood based factors relevant to CVD such as
blood pressure, lipids and inflammatory markers, enabling appropriate adjustment
119
and tests of interaction. The advantage of an individual participant meta-analysis like
the ERFC, compared to literature-based meta-analyses, is that it allows consistent
adjustment across studies and investigation of the sources of heterogeneous effects
across studies. Bias from misreporting of smoking status in sick individuals was
minimized by involving data from prospective cohort studies. The use of a
prospective design and appropriate adjustment for a range of relevant risk factors
increase the likelihood that RRs demonstrate causality between smoking and CVD
43. Over 90% of current smokers (98% of past smokers) were current cigarette
(respectively past cigarette) smokers, and therefore these findings describe the
relationship between cigarette smoking and CVD.
4.4.1 Main findings
Smoking was associated with a wide range of cardiovascular events, both non-fatal
and fatal. The effect of current smoking on vascular diseases and deaths was
relatively homogenous, oscillating between 1.5 and 2.1, with stronger effects on the
risks of subarachnoid haemorrhage, aortic aneurysm and peripheral vascular deaths,
which may point toward a hypertensive effect of smoking 44 in addition to its
atherosclerotic and thrombotic effects 45, 46. The association with death caused by
rupture of an aortic aneurysm was high and was higher than previously reported 47.
Second after aortic aneurysm, smoking had the strongest effect on death from
peripheral vascular diseases. This is in agreement with the most recent meta-
analysis which was literature-based, and included only 4 prospective cohort studies
with 50 times less individuals than the current collaboration 4. Regarding
subarachnoid haemorrhage, I obtained estimates similar to that found by a meta-
analysis of 14 longitudinal studies, which included, in total, 892 cases, compared to
1,313 in the ERFC 5. The association between smoking and pulmonary embolism
death was nearly as strong as the association with MI, contrary to previous belief that
the RR is about half of the RRs for the more common forms of CVD which are MI or
stroke 3, 48. The association between smoking and atrial fibrillation (the most common
type of sustained dysrhythmia) has been conflicting in the literature 7. In the ERFC,
smoking increased the risk of death from cardiac dysrhythmia by ~80% and the
strength of association was similar to that of CHD and stroke. Evidence on smoking
and death from heart failure has so far been limited and in the ERFC the RR for
current versus never smokers was twice as high as previous estimates 49.
120
The relationship between smoking and MI or ischaemic stroke has been reported by
numerous studies. The association was not significantly different for fatal versus non-
fatal events, even if smokers experiencing an MI or stroke have been shown to have
lower levels of atherosclerosis and less filled atherosclerotic plaques than non-
smokers 50. In the Nurses’ Health Study, current smokers had a multivariate HR of
3.26 (2.97; 3.59) compared to never smokers for CHD and 2.81 (2.32; 3.41) for
cerebrovascular diseases, which is higher than my estimates of 1.99 (1.86, 2.13) and
1.64 (1.54; 1.75) respectively 20. Length of follow-up was longer (24 years versus
around 14 years in the ERFC), which could affect RRs for CVD which are diseases
of middle age and generally happen several decades after smoking initiation.
Misclassification due to never smokers starting to smoke or current smokers at
baseline stopping during follow-up may have attenuated associations.
There are several proposed mechanisms by which smoking could cause CVD. As
developed in the Introduction, nicotine, carbon monoxide and oxidant gases are
constituents of cigarette smoke and have been shown to promote atherosclerosis
and thrombosis. Cigarette smoking is thought to contribute to the process of
atherosclerosis by promoting inflammation, insulin sensitivity and lipid abnormalities;
and to produce acute myocardial infarction by adversely affecting the balance of
demand for myocardial oxygen and nutrients with myocardial blood supply 16. One
mechanism that links smoking to ischaemic stroke is structural artery wall damage
and carotid atherosclerosis, leading to thrombosis or embolic phenomena 51. The
mechanism of thrombogenesis is a short-term effect of smoking, which includes
increased levels of inflammatory markers as shown in the previous Chapter, and is
supported by the reduction in risk of cerebrovascular diseases observed after
cessation of smoking. For subarachnoid haemorrhage, there is evidence for an
association of smoking with aneurysm formation, growth and rupture 52. For
haemorrhagic stroke, possible mechanisms are still speculative and may be
mediated through arterial wall damage 53 and blood pressure 54.
The effect of smoking was very slightly stronger in women, younger age groups, non-
diabetics, non-obese and non-hypertensive individuals. These findings suggest that
smoking may act as a trigger of acute thrombotic events even in individuals with low
levels of atherosclerosis. It has been suggested that women smoking are more at
risk than men smoking. Hormonal treatment in post-menopausal women and
different vessel composition would be reasons for this difference. My findings show a
121
modestly increased RR in women compared to men and which was borderline
significant. Therefore, I did not replicate results from a recent literature based meta-
analysis claiming a 25% greater risk of MI in women smoker compared to men
smoker 14. Inconsistent adjustment across studies included in the literature based
meta-analysis and consistent adjustment in the ERFC make ERFC findings more
reliable. For other risk factors, interplays were small except for age where RRs were
about ¼ smaller for individuals ≥60 years old compared to individuals aged 40-59 
years old.
I confirmed previous findings of a nonlinear association between smoking intensity
and CVD risk 20, 55. The first few cigarettes accounted for most of the increased risk
and there was no “safe level” of smoking in terms of CVD risk. This suggests that the
underlying biochemical and cellular processes may become saturated with small
doses of toxic components from cigarette smoking. In that respect, it has been
shown that passive smokers, who experienced lower levels of smoke than smokers
of 1-5 cigarettes per day still have increased levels of platelet activation similar to
that of active smokers56. Another reason may be that heavy smokers take lighter and
shorter puffs than lighter smokers, therefore inhaling less smoke and absorbing less
nicotine and carbon monoxide into the blood stream. In a study measuring cotinine, a
metabolite of nicotine, levels increased most strongly from 0 to 10 CPD, reaching a
plateau above 30 CPD 57. The association between smoking duration and CHD was
relatively weaker than that of intensity and it was approximately flat for
cerebrovascular diseases. This suggests that smoking has mostly a short term
impact on CVD risk rather than a long term damaging effect 51. Because of the
shapes of association between intensity and duration with CVD risk, I would not
advocate summarizing smoking exposure with the use of pack-years when
considering CVD risk, as had been done when investigating lung cancer association,
where both duration and intensity have an approximately linear relationship with
disease 17.
Cessation of smoking was rapidly beneficial. Ex-smokers experienced a risk which
was not significantly higher than that of never smokers for all CVD subtypes, except
for heart failure and aortic aneurysm. Smoking marijuana has been shown to act as a
trigger of MI, but not smoking cigarettes 58. Damage to the arterial wall caused by
smoking takes time to heal, and excess risk still remained significantly higher than
122
that of never smokers for the next 10 years for all CVD and for the next 20 years for
CHD only.
Estimates of number of years of life lost to smoking in current smokers (being at their
maximum at age 40 years old with 7.7 years lost) were lower than previously
published estimates of 10 years in men 23. However, the previous estimate published
by the British Doctor’s Study was based on men born at the end of the 19th century
and first half of the 20th century whose mortality rates and smoking habits differed
from more recent generations. For women, the present study is the first to my
knowledge to give estimates of years of life lost as a result of smoking. Reliability of
these findings is warranted by the large scale of the data and the approximately
equal proportion of men and women enrolled at baseline. For current smokers,
vascular deaths accounted for slightly less than a third of excess deaths in both men
and women, while cancer accounted for over a third of excess deaths. In ex-
smokers, most excess deaths were due to cancer rather than CVD, which was a
consequence of my finding that smoking effects are more easily undone on the
cardiovascular system than on cancer conditions, in particular lung cancer.
4.4.2 Limitations
Despite its several strengths, this analysis contains several limitations. First, I
conducted an independent participant meta-analysis, and heterogeneity between
studies may affect estimates of the risk. However, I investigated the extent of
heterogeneity using the I2 statistic and attempted to explain some of this
heterogeneity by grouping RRs according to geographical and other study
characteristics. The definition of smoking status and assessments of duration and
amount may have differed in studies, but data managers involved prior to the
analyses corresponded with authors and attempted to resolve main discrepancies.
Studies included in the ERFC were conducted over several decades during which
cigarettes have changed in their chemical composition and new forms have
appeared such as “filtered” and “light” cigarettes. However, violation of the
proportional hazard assumption was marginal, showing no clinically significant decay
of the RRs over time. In addition, residual bias could persist due to unmeasured or
imprecisely measured confounding factors, for example alcohol consumption and
socioeconomic factors. Some confounding variables such as menopause and
hormone therapy were not available.
123
Analyses comparing the risk of current, ex and never smokers were not adjusted for
amount and duration of smoking, in order to maximize power of the analyses. This
may have introduced biases in the results asindividuals are likely to differ in terms of
amount and duration of smoking. For example, women in the ERFC dataset are
likely to have been smoking for shorter duration and lower amounts than their male
counterparts, and this may distort the interaction between sex and smoking status on
the risk of CVD. The same applies for individuals with higher body mass index, high
blood pressure or high cholesterol levels.
Apart from for coronary disease and stroke, follow-up for other cardiovascular events
and for lung cancer was only up to a fatal event. It is plausible that individuals
experiencing a non-fatal cardiovascular event or lung cancer would modify their
smoking habit. To limit the influence of potential prior diseases on baseline
measurements, subsidiary analysis was done excluding the first five years of follow-
up and RRs remain unchanged (Table 4.1).
Baseline smoking status was used. Measurement error could be introduced if current
smokers attempt to quit during follow-up and never smokers initiate smoking. In my
dataset, amongst individuals who provided resurvey information, more than 80% had
a stable smoking status <20 years of follow-up, 78% after 20-30 years of follow-up
and 66% after 30-40 years of follow-up (Table 4.6). In particular, the proportion of
never smokers initiating smoking was <1% independent of the time of follow-up. An
increasing proportion of current smokers became ex-smokers: the proportion was
<10% within the first 15 years of follow-up and reached 28% after 30-40 years of
follow-up.
Taking into account measurement error in a categorical variable such as smoking
status remains statistically challenging 59 and was beyond the scope of this thesis.
Sex and age specific mortality rates for the overall population were used instead of
rates in lifelong never smokers for the estimation of number of years lost as a result
of smoking, which may have led to underestimation of the effect of smoking. Finally,
European rates were used as reference rates, which may not be appropriate in the
case of non-European studies included in the ERFC.
4.5 Conclusion
124
Smoking was associated with substantial CVD burden and reduction in life
expectancy in both men and women. All types of CVD were associated with
smoking. Women, younger age groups and non-diabetics experienced the greatest
increase in risk of CVD when smoking, but in absolute terms, smoking caused more
CVD events amongst men, older age groups and diabetics. Rather than reducing
intake, the only safe option in terms of CVD risk was to stop smoking, which was
associated with a rapid drop in CVD risk.
125
Figure 4.1: Risk ratios for non-fatal MI, non-fatal stroke and cardiovascular deaths
by smoking status
Ex versus never
Pulmonary embolism
Ex versus never
Ex versus never
Ex versus never
Aortic aneurysm
Unclassified stroke
Ex versus never
Haemorrhagic stroke
Current versus never
Current versus never
Current versus never
Ex versus never
Current versus never
Current versus never
Current versus never
Ex versus never
Current versus never
Current versus never
CHD death and non-fatal MI
Subarachnoid haemorrhage
Ischaemic stroke
Ex versus never
Current versus never
Heart failure
Ex versus never
Ex versus never
Current versus never
Sudden death
Peripheral vascular deaths
Ex versus never
All cerebrovascular events
Current versus never
Ex versus never
Cardiac dysrhytmia
Current versus never
1964
279
1240
4786
40118
522
1209
3473
17221
1166
796
1854
1.20 (0.97, 1.47)
1.01 (0.93, 1.10)
1.12 (0.87, 1.43)
1.39 (0.91, 2.13)
1.17 (0.96, 1.43)
2.00 (1.77, 2.25)
3.81 (2.67, 5.45)
2.29 (1.92, 2.74)
1.17 (1.03, 1.33)
1.72 (1.54, 1.93)
1.99 (1.86, 2.13)
1.62 (1.30, 2.02)
0.95 (0.86, 1.04)
2.69 (2.26, 3.20)
1.50 (1.37, 1.64)
1.04 (0.99, 1.09)
1.64 (1.54, 1.75)
1.04 (0.89, 1.22)
1.00 (0.88, 1.15)
4.34 (3.33, 5.66)
1.44 (1.17, 1.77)
1.79 (1.47, 2.18)
1.13 (1.10, 1.17)
1.64 (1.45, 1.85)
0 (0, 47)
0 (0, 32)
0 (0, 54)
10 (0, 81)
0 (0, 43)
0 (0, 39)
0 (0, 79)
16 (0, 51)
0 (0, 39)
30 (2, 51)
74 (69, 78)
0 (0, 54)
0 (0, 39)
0 (0, 43)
0 (0, 39)
8 (0, 29)
43 (27, 56)
0 (0, 49)
0 (0, 35)
45 (10, 66)
0 (0, 45)
0 (0, 47)
0 (0, 24)
0 (0, 35)
1 1.5 2 3 4 6 8
Cause of
death\subgroup
No of
events RR (95% CI) I2 (95% CI),%
RR (95% CI)
Ex versus never smokers
Current versus smokers
Study-specific loge risk ratios adjusted for baseline age and body mass, and stratified, where
appropriate, by sex and trial arm, were combined using a multivariate random-effects meta-
analysis. Studies with fewer than 10 cases were excluded from the analysis of the outcome.
Sizes of the data markers are proportional to the inverse of the variance of the log risk ratios.
RRs were calculated with reference to never smokers. Individuals were censored at their first
non-fatal myocardial infarction or stroke and at death.
126
Figure 4.2: Forest plot of risk ratios for fatal and non-fatal MI for current versus
never smokers by region where studies were conducted
North America
CHS2
CHARL
ALLHAT
QUEBEC
RANCHO
NHANESI
MRFIT
SHS
CHS1
HONOL
USPHS2
CHA
NSHS
ARIC
HPFS1
NHS1
WHS
Subtotal (I-squared = 92.0%, p = 0.000)
Western Europe
LEADER
MONICA
LASA
MATISS93
PROSPER
FINE_IT
NFR
EMOFRI
GOTO33
GLOSTRUP
MRCOLD
BRUN
DRECE
NPHSI
MATISS83
MONFRI94
ATS_SAR
FINRISK97
EAS
VHMPP
RF2
GUBBIO
SPEED
KARELIA
MICOL
BUPA
REYK
WHITEI
BRHS
COPEN
WHITEII
HBS
NPHSII
MONFRI86
GOTOW
PRIME
MATISS87
TROMSØ
ROTT
HOORN
FINRISK92
OB43
MONFRI89
PREVEND
CAPS
NCS2
MCVDRFP
HELSINAG
GRIPS
FIA
MOSWEGOT
MALMO
MOGERAUG1
GOTO43
BWHHS
MOGERAUG2
EPICNOR
OSLO
NCS3
NCS1
SHHEC
ATENA
MORGEN
ZARAGOZA
MOGERAUG3
Subtotal (I-squared = 76.2%, p = 0.000)
Other
HISAYAMA
FUNAGATA
IKNS
ISRAEL
OSAKA
DUBBO
GOH
BHS
AUSDIAB
TOYAMA
Subtotal (I-squared = 0.0%, p = 0.555)
0.83 (0.45, 1.55)
1.18 (1.01, 1.37)
1.30 (1.14, 1.49)
1.40 (0.91, 2.16)
1.46 (1.10, 1.93)
1.49 (1.27, 1.75)
1.51 (1.23, 1.86)
1.59 (1.32, 1.92)
1.67 (1.38, 2.02)
1.80 (1.18, 2.74)
1.86 (1.39, 2.48)
1.90 (1.77, 2.03)
2.09 (0.87, 5.01)
2.19 (1.94, 2.47)
2.29 (2.09, 2.52)
2.38 (2.24, 2.54)
3.49 (2.84, 4.29)
1.75 (1.53, 2.00)
0.48 (0.24, 0.94)
0.70 (0.27, 1.79)
1.02 (0.33, 3.17)
1.12 (0.51, 2.46)
1.18 (0.91, 1.53)
1.22 (0.84, 1.78)
1.25 (0.74, 2.11)
1.26 (0.25, 6.40)
1.34 (0.69, 2.61)
1.35 (0.57, 3.16)
1.40 (1.23, 1.58)
1.41 (0.94, 2.11)
1.48 (0.57, 3.82)
1.50 (0.97, 2.30)
1.53 (1.07, 2.18)
1.54 (0.75, 3.18)
1.65 (0.64, 4.24)
1.70 (1.21, 2.39)
1.70 (1.14, 2.54)
1.70 (1.54, 1.89)
1.71 (0.99, 2.97)
1.72 (1.00, 2.95)
1.74 (1.24, 2.45)
1.74 (1.58, 1.93)
1.75 (1.12, 2.73)
1.78 (1.56, 2.04)
1.83 (1.71, 1.97)
1.86 (1.39, 2.48)
1.86 (1.64, 2.11)
1.88 (1.62, 2.18)
1.97 (1.55, 2.50)
1.97 (1.23, 3.14)
1.99 (1.47, 2.69)
1.99 (1.24, 3.20)
2.01 (1.61, 2.50)
2.05 (1.44, 2.92)
2.10 (1.22, 3.62)
2.14 (1.81, 2.54)
2.15 (1.63, 2.83)
2.16 (1.45, 3.23)
2.19 (1.68, 2.85)
2.21 (0.69, 7.07)
2.29 (1.34, 3.93)
2.40 (1.59, 3.62)
2.42 (1.63, 3.61)
2.47 (1.81, 3.38)
2.49 (1.94, 3.21)
2.57 (1.43, 4.63)
2.64 (2.02, 3.45)
2.64 (2.08, 3.34)
2.70 (2.06, 3.55)
2.72 (2.44, 3.03)
2.84 (1.62, 4.96)
2.85 (1.26, 6.44)
2.90 (2.00, 4.21)
2.91 (1.80, 4.68)
2.92 (2.04, 4.17)
2.93 (2.56, 3.35)
2.94 (2.10, 4.10)
2.94 (2.33, 3.70)
3.08 (2.45, 3.86)
3.59 (1.53, 8.43)
3.72 (2.29, 6.05)
4.01 (2.23, 7.21)
5.40 (2.43, 12.02)
2.04 (1.90, 2.19)
1.32 (0.91, 1.93)
1.37 (0.63, 3.01)
1.38 (1.05, 1.82)
1.43 (1.24, 1.65)
1.49 (1.02, 2.16)
1.51 (1.17, 1.95)
1.54 (1.26, 1.88)
1.64 (1.39, 1.93)
2.03 (1.18, 3.48)
2.82 (1.56, 5.12)
1.52 (1.41, 1.65)
57
331
573
22
216
797
109
213
489
103
325
1470
7
538
1570
1727
237
10
12
7
13
133
60
18
3
16
10
1038
91
17
26
138
18
10
85
53
2523
27
46
40
1304
40
305
1194
94
342
264
128
41
60
33
171
49
74
181
173
45
126
5
25
34
28
58
84
63
68
226
87
408
16
8
85
33
142
240
41
93
95
8
21
45
12
180
26
256
274
119
277
212
390
58
29
13
489
414
349
67
239
635
299
143
29
54
2070
19
604
607
2513
151
63
10
8
15
142
52
67
3
21
56
340
40
9
113
129
14
17
67
50
483
49
30
199
1228
60
768
2504
89
912
734
142
31
151
48
175
85
47
600
140
76
129
13
35
85
204
183
300
16
246
204
134
1589
55
21
42
48
96
1933
351
395
427
18
83
25
16
60
20
164
561
93
88
225
317
19
58
1.5 2 4 8 16
Cohort RR (95 CI)
No failure
in never
smokers
No failures
In current
smokers
Region
Boxes sizes are inversely proportional to study weights within each subgroup. Weights are
from a random effect univariate meta-analysis. Study specific estimates were adjusted for
baseline age and body mass index and stratified, when appropriate, by sex and trial arm.
Studies with less than 10 events were excluded.
127
Figure 4.3: Forest plot of risk ratios for fatal and non-fatal MI for current versus
never smokers by baseline decade of survey
1990
LEADER
CHS2
LASA
MATISS93
PROSPER
EMOFRI
ALLHAT
FUNAGATA
MRCOLD
BRUN
DRECE
OSAKA
MONFRI94
FINRISK97
HONOL
WHITEI
USPHS2
COPEN
AUSDIAB
PRIME
NSHS
ROTT
FINRISK92
PREVEND
TOYAMA
GOTO43
BWHHS
EPICNOR
WHS
ATENA
MORGEN
ZARAGOZA
MOGERAUG3
Subtotal (I-squared = 80.4%, p = 0.000)
0.48 (0.24, 0.94)
0.83 (0.45, 1.55)
1.02 (0.33, 3.17)
1.12 (0.51, 2.46)
1.18 (0.91, 1.53)
1.26 (0.25, 6.40)
1.30 (1.14, 1.49)
1.37 (0.63, 3.01)
1.40 (1.23, 1.58)
1.41 (0.94, 2.11)
1.48 (0.57, 3.82)
1.49 (1.02, 2.16)
1.54 (0.75, 3.18)
1.70 (1.21, 2.39)
1.80 (1.18, 2.74)
1.86 (1.39, 2.48)
1.86 (1.39, 2.48)
1.88 (1.62, 2.18)
2.03 (1.18, 3.48)
2.05 (1.44, 2.92)
2.09 (0.87, 5.01)
2.15 (1.63, 2.83)
2.19 (1.68, 2.85)
2.40 (1.59, 3.62)
2.82 (1.56, 5.12)
2.85 (1.26, 6.44)
2.90 (2.00, 4.21)
2.92 (2.04, 4.17)
3.49 (2.84, 4.29)
3.59 (1.53, 8.43)
3.72 (2.29, 6.05)
4.01 (2.23, 7.21)
5.40 (2.43, 12.02)
1.91 (1.66, 2.20)
10
57
7
13
133
3
573
26
1038
91
17
119
18
85
103
94
325
264
58
49
7
173
126
34
29
8
85
142
237
8
21
45
12
63
13
8
15
142
3
414
20
340
40
9
93
14
67
29
89
54
734
19
85
19
140
129
85
58
21
42
96
151
18
83
25
16
.
1980
MONICA
FINE_IT
HISAYAMA
GOTO33
IKNS
RANCHO
DUBBO
MATISS83
SHS
ATS_SAR
CHS1
EAS
VHMPP
GUBBIO
MICOL
WHITEII
HBS
NPHSII
MONFRI86
MATISS87
TROMSØ
HOORN
ARIC
OB43
HPFS1
MONFRI89
MCVDRFP
HELSINAG
GRIPS
FIA
MOSWEGOT
MOGERAUG1
MOGERAUG2
SHHEC
Subtotal (I-squared = 68.0%, p = 0.000)
0.70 (0.27, 1.79)
1.22 (0.84, 1.78)
1.32 (0.91, 1.93)
1.34 (0.69, 2.61)
1.38 (1.05, 1.82)
1.46 (1.10, 1.93)
1.51 (1.17, 1.95)
1.53 (1.07, 2.18)
1.59 (1.32, 1.92)
1.65 (0.64, 4.24)
1.67 (1.38, 2.02)
1.70 (1.14, 2.54)
1.70 (1.54, 1.89)
1.72 (1.00, 2.95)
1.75 (1.12, 2.73)
1.97 (1.55, 2.50)
1.97 (1.23, 3.14)
1.99 (1.47, 2.69)
1.99 (1.24, 3.20)
2.10 (1.22, 3.62)
2.14 (1.81, 2.54)
2.16 (1.45, 3.23)
2.19 (1.94, 2.47)
2.21 (0.69, 7.07)
2.29 (2.09, 2.52)
2.29 (1.34, 3.93)
2.49 (1.94, 3.21)
2.57 (1.43, 4.63)
2.64 (2.02, 3.45)
2.64 (2.08, 3.34)
2.70 (2.06, 3.55)
2.84 (1.62, 4.96)
2.91 (1.80, 4.68)
3.08 (2.45, 3.86)
1.95 (1.79, 2.13)
12
60
180
16
256
216
277
138
213
10
489
53
2523
46
40
128
41
60
33
74
181
45
538
5
1570
25
84
63
68
226
87
16
33
95
10
52
60
21
164
67
88
129
299
17
143
50
483
30
60
142
31
151
48
47
600
76
604
13
607
35
300
16
246
204
134
55
48
427
.
.
1960
CHARL
ISRAEL
GOH
BHS
REYK
CHA
GOTOW
Subtotal (I-squared = 86.3%, p = 0.000)
1970
NFR
GLOSTRUP
QUEBEC
NHANESI
NPHSI
MRFIT
RF2
SPEED
KARELIA
BUPA
BRHS
NHS1
CAPS
NCS2
MALMO
OSLO
NCS3
NCS1
Subtotal (I-squared = 87.4%, p = 0.000)
1.18 (1.01, 1.37)
1.43 (1.24, 1.65)
1.54 (1.26, 1.88)
1.64 (1.39, 1.93)
1.83 (1.71, 1.97)
1.90 (1.77, 2.03)
2.01 (1.61, 2.50)
1.63 (1.44, 1.85)
1.25 (0.74, 2.11)
1.35 (0.57, 3.16)
1.40 (0.91, 2.16)
1.49 (1.27, 1.75)
1.50 (0.97, 2.30)
1.51 (1.23, 1.86)
1.71 (0.99, 2.97)
1.74 (1.24, 2.45)
1.74 (1.58, 1.93)
1.78 (1.56, 2.04)
1.86 (1.64, 2.11)
2.38 (2.24, 2.54)
2.42 (1.63, 3.61)
2.47 (1.81, 3.38)
2.72 (2.44, 3.03)
2.93 (2.56, 3.35)
2.94 (2.10, 4.10)
2.94 (2.33, 3.70)
2.01 (1.78, 2.27)
331
274
212
390
1194
1470
171
18
10
22
797
26
109
27
40
1304
305
342
1727
28
58
408
240
41
93
489
561
225
317
2504
2070
175
67
56
349
239
113
635
49
199
1228
768
912
2513
204
183
1589
1933
351
395
1.5 2 4 8 16
Cohort RR (95 CI)
No failure
in never
smokers
No failures
In current
smokers
Decade
Baseline
survey
Boxes sizes are inversely proportional to study weights within each subgroup. Weights are
from a random effect univariate meta-analysis. Study specific estimates were adjusted for
baseline age and body mass index and stratified, when appropriate, by sex and trial arm.
Studies with less than 10 events were excluded.
128
Figure 4.4: Adjusted relative risk ratios for major vascular morbidity and mortality
and for current smokers versus never by levels of categorical CVD risk factors
CHD death and non-fatal MI
All cerebrovascular events
Ischaemic stroke
Haemorrhagic stroke
CHD death and non-fatal MI
All cerebrovascular events
Ischaemic stroke
Haemorrhagic stroke
CHD death and non-fatal MI
All cerebrovascular events
Ischaemic stroke
Haemorrhagic stroke
20366
9752
3978
1223
15774
7759
3465
1072
21959
10274
4219
1310
0.73 (0.66, 0.79)
0.83 (0.75, 0.91)
0.81 (0.69, 0.96)
1.03 (0.79, 1.35)
0.87 (0.79, 0.96)
0.88 (0.79, 0.98)
0.90 (0.77, 1.05)
0.89 (0.68, 1.18)
0.70 (0.62, 0.79)
0.96 (0.82, 1.12)
0.83 (0.65, 1.07)
1.20 (0.65, 2.22)
<0.001
0.004
0.003
0.240
0.005
0.020
0.198
0.418
<0.001
0.588
0.151
0.550
Outcome
events
No. of
RRR (95% CI) p-value
RRRSubgroup
Comparing Individuals ≥60 years old versus 40-59 years old
Comparing Male versus Female
Comparing diabetics versus non diabetics
Higher RR when
≥60 years old
Higher RR
for Male
Higher RR for
Female
Higher RR
when diabetics
Higher RR when
non diabetics
Higher RR when
40-59 years old
1.2 .4 .6 .8 1.2 1.5 2 3
RRR (95% CI)
RRR: Ratio of risk ratio; 95% CI: 95% confidence interval; p-value: p-value of 1 degree of
freedom test of interaction between baseline characteristics (either categories for a
categorical variable or in their continuous form for numeric variables) and smoking status.
Study-specific loge risk ratios adjusted for baseline age body mass index, history of diabetes,
systolic blood pressure and total cholesterol, and stratified, where appropriate, by sex and
trial arm, were combined using a multivariate random-effects meta-analysis. Studies with
fewer than 5 cases were excluded from the analysis of the outcome. Sizes of the data
markers are proportional to the inverse of the variance of the log risk ratios. RRs were
calculated with reference to never smokers. Individuals were censored at their first non-fatal
myocardial infarction or stroke and at death. For the interaction with sex, the data was
restricted to studies including both Male and Females. All cerebrovascular events include
ischaemic stroke, haemorrhagic stroke, unclassified strokes and subarachnoid haemorrhage.
129
Figure 4.5: Adjusted relative risk ratios for major vascular morbidity and non-
vascular mortality and for current smokers versus never by levels of continuous CVD
risk factors
CHD death and non-fatal MI
All cerebrovascular events
Ischaemic stroke
Haemorrhagic stroke
Lung cancer deaths
CHD death and non-fatal MI
All cerebrovascular events
Ischaemic stroke
Haemorrhagic stroke
CHD death and non-fatal MI
All cerebrovascular events
Ischaemic stroke
Haemorrhagic stroke
21959
10274
4219
1310
4605
21959
10274
4219
1310
21959
10274
4219
1310
0.97 (0.93, 1.02)
0.95 (0.87, 1.02)
1.05 (0.92, 1.19)
0.91 (0.78, 1.06)
0.75 (0.64, 0.88)
0.96 (0.95, 0.98)
1.02 (0.99, 1.04)
1.01 (0.97, 1.05)
1.06 (1.00, 1.13)
1.01 (0.98, 1.04)
1.03 (1.00, 1.07)
1.04 (0.97, 1.11)
1.11 (0.99, 1.24)
0.290
0.168
0.490
0.217
<0.001
<0.001
0.149
0.779
0.061
0.549
0.082
0.258
0.065
1.2 .4 .6 .8 1.2 1.5 2 3
RRR (95% CI)
Per 5 kg/m2 increase in Body Mass Index
Per 10 mmHg increase in systolic blood pressure
Per 1 mmol/l increase in total cholesterol
RRR: Ratio of risk ratio; 95% CI: 95% confidence interval; p-value: p-value of 1 degree of
freedom test of interaction between baseline characteristics (either categories for a
categorical variable or in their continuous form for numeric variables) and smoking status.
Study-specific loge risk ratios adjusted for baseline age body mass index, history of diabetes,
systolic blood pressure and total cholesterol, and stratified, where appropriate, by sex and
trial arm, were combined using a multivariate random-effects meta-analysis. Studies with
fewer than 5 cases were excluded from the analysis of the outcome. Sizes of the data
markers are proportional to the inverse of the variance of the log risk ratios. RRs were
calculated with reference to never smokers. Individuals were censored at their first non-fatal
myocardial infarction or stroke and at death. For the interaction with sex, the data was
restricted to studies including both Male and Females. All cerebrovascular events include
ischaemic stroke, haemorrhagic stroke, unclassified strokes and subarachnoid haemorrhage.
130
Figure 4.6 a-b: Absolute risk of fatal or non-fatal MI for current smokers and never,
baseline characteristics, representing event rates on absolute scale using bar plots
a) Fatal and non-fatal MI
0
2
5
10
15
20
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
40-59
years old
60-69
Years old
70+
Years old
Age at baseline
0
2
5
10
15
20
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Male Female
Sex
0
2
5
10
15
20
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Non diabetics Diabetics
History of diabetes
0
2
5
10
15
20
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Body Mass Index
0
2
5
10
15
20
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Systolic blood pressure
0
2
5
10
15
20
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Bottom
third
Middle
third
Top
third
Total cholesterol
Never smokers Current smokers
Bottom
third
Middle
third
Top
third
Bottom
third
Middle
third
Top
third
Please see the Method section for an explanation of this graphs.
131
b) fatal and non-fatal cerebrovascular events
0
2
5
10
15
20
0
2
5
10
15
20
0
2
5
10
15
20
0
2
5
10
15
20
0
2
5
10
15
20
0
2
5
10
15
20
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
40-59
years old
60-69
Years old
70+
Years old
Age at baseline
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Male Female
Sex
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Non diabetics Diabetics
History of diabetes
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Body Mass Index
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Systolic blood pressure
ev
en
ts
pe
r1
00
0
pe
rs
on
-y
ea
rs
(9
5%
C
I)
Bottom
third
Middle
third
Top
third
Total cholesterol
Never smokers Current smokers
Bottom
third
Middle
third
Top
third
Bottom
third
Middle
third
Top
third
Please see the Method section for an explanation of this graphs.
132
Figure 4.7: Risk ratios for fatal and non-fatal MI, fatal and non-fatal cerebrovascular
events, lung cancer deaths and all causes of death by smoking status and quintiles
of smoking pack-years, duration (years) and amount (cigarettes per day)
1
1.5
2
2.5
3
1
1.5
2
2.5
3
1
1.5
2
2.5
3
1
1.5
2
2.5
3
1
1.5
2
2.5
3
1
1.5
2
2.5
3
1
1.5
2
2.5
3
1
1.5
2
2.5
3
1
1.5
2
2.5
3
0 10 20 30 40 50 0 10 20 30 40 0 10 20 30 40
0 10 20 30 40 50 0 10 20 30 40 0 10 20 30 40
0 10 20 30 40 50 0 10 20 30 40 0 10 20 30 40
1
4
6
10
15
20
30
1
4
6
10
15
20
30
1
4
6
10
15
20
30
0 10 20 30 40 50 0 10 20 30 40 0 10 20 30 40
Mean level
a) Pack-years a) Duration c) Amount
(i) CHD deaths
And non fatal MI
(17,500 events)
(i) Cerebrovascular fatal
And non fatal events
(9,845 events)
(i) Lung cancer
(2,557 deaths)
(i) All causes
mortality
(56,013 deaths)
R
is
k
ra
tio
(9
5%
C
I)
R
is
k
ra
tio
(9
5%
C
I)
R
is
k
ra
tio
(9
5%
C
I)
Current smokers Ex smokers Never smokers
R
is
k
ra
tio
(9
5%
C
I)
Study-specific loge risk ratios adjusted for baseline age and body mass index, and stratified,
where appropriate, by sex and trial arm; were combined using a multivariate random-effects
meta-analysis. Studies with fewer than 5 cases were excluded from the analysis of the
outcome. Sizes of the data markers are proportional to the inverse of the variance of the log
risk ratios. RRs were calculated with reference to never smokers. Confidence intervals are
derived from “floating absolute variances”. “Never” correspond to never smokers and to 0
pack-years/years/cigarettes per day. “Ex” corresponds to “ex-smokers”. All cerebrovascular
events include ischaemic stroke, haemorrhagic stroke, unclassified strokes and subarachnoid
haemorrhage. Data restricted to studies with information on amount, duration and pack-
years. Exclusive pipe or cigar smokers were considered as non-smokers for this analysis.
133
Figure 4.8: Risk ratios for CHD death and non-fatal MI, all fatal CVD and lung
cancer death according to the number of years since quitting smoking
1
1.2
1.5
2
3
4
6
8
10
14
1
1.2
1.5
2
3
4
6
8
10
14
1
1.2
1.5
2
3
4
6
8
10
14
01 5 10 15 20 25 30 01 5 10 15 20 25 30 01 5 10 15 20 25 30
CHD death and non fatal MI
18,213 cases
All CVD events
17,803 cases
Lung cancer deaths
3,931 cases
R
is
k
ra
tio
(9
5%
C
I)
Mean number of years since quitting smoking
Current smoker Ex smoker Never smoker
Study-specific loge risk ratios adjusted for baseline age and body mass index, and stratified,
where appropriate, by sex and trial arm; were combined using a multivariate random-effects
meta-analysis. Studies with fewer than 10 cases were excluded from the analysis of the
outcome. Sizes of the data markers are proportional to the inverse of the variance of the log
risk ratios. RRs were calculated with reference to never smokers. Confidence intervals are
derived from “floating absolute variances”. Note “All CVD events” refers to all cardiovascular
deaths.
134
Figure 4.9: Smoking and survival, according to sex and smoking status in current
versus never smokers
a) Estimated survival of current versus never smokers
a) Men b) Women
Pr
ob
ab
ilit
y
of
su
rv
iv
al
0.0
0.2
0.4
0.6
0.8
1.0
40 50 60 70 80 90 100 110 40 50 60 70 80 90 100 110
Never smoker
Current smoker
Age (years)
0.0
0.2
0.4
0.6
0.8
1.0
Never smoker
Current smoker
Age (years)
b) Estimated future years of life lost owing to smoking in current versus never
smokers
a) Men b) Women
0
1
2
3
4
5
6
7
40 50 60 70 80 90 40 50 60 70 80 90
Ye
ar
s
of
lif
e
lo
st
0
1
2
3
4
5
6
7
All vascular deaths
Cancer deaths
Non-vascular non
cancer deaths
Death from unknown
causes
All vascular deaths
Cancer deaths
Non-vascular non
cancer deaths
Death from unknown
causes
Age (years) Age (years)
Panel a) shows estimated survival curves that were plotted by applying hazard ratios for
death from any cause (specific for sex and age at risk) from the present analyses to mortality
data for the European Union in 2000. Panel B shows the estimated number of years of life
lost owing to smoking. Participants with known pre-existing cardiovascular disease at
baseline were excluded from these analyses.
135
Figure 4.10: Smoking and survival, according to sex and smoking status in ex-
versus never smokers
a) Estimated survival of ex versus never smokers
0.0
0.2
0.4
0.6
0.8
1.0
40 50 60 70 80 90 100 110 40 50 60 70 80 90 100 110
Never smoker
Ex smoker
a) Men b) Women
Pr
ob
ab
ilit
y
of
su
rv
iv
al
Age (years) Age (years)
Never smoker
Ex smoker
0.0
0.2
0.4
0.6
0.8
1.0
b) Estimated future years of life lost owing to smoking in ex- versus never smokers
0
1
2
3
4
5
6
7
40 50 60 70 80 90 40 50 60 70 80 90
a) Men b) Women
Ye
ar
s
of
lif
e
lo
st
All vascular deaths
Cancer deaths
Non-vascular non
cancer deaths
Death from unknown
causes
All vascular deaths
Cancer deaths
Non-vascular non
cancer deaths
Death from unknown
causes
0
1
2
3
4
5
6
7
Age (years) Age (years)
Panel a) shows estimated survival curves that were plotted by applying hazard ratios for
death from any cause (specific for sex and age at risk) from the present analyses to mortality
data for the European Union in 2000 39. Panel b) shows the estimated number of years of life
lost owing to smoking. Participants with known pre-existing cardiovascular disease at
baseline were excluded from these analyses.
136
Table 4.1: Risk ratios for major outcomes for current and ex versus never smokers,
adjusted for age, sex and BMI under different conditions
Current versus never
smoker
Ex versus never
smokers
Analysis Outcome
No of
event
s
RR (95% CI) I2 (95% CI) RR (95% CI) I2 (95% CI)
Including everyone
Coronary heart disease* 40118 1.99(1.86, 2.13) 74 (69, 78)
1.13
(1.10, 1.17) 0 (0, 24)
Ischaemic stroke* 3803 1.74(1.57, 1.93) 7 (0, 37)
1.04
(0.94, 1.15) 6 (0, 36)
CVD deaths 43861 2.01(1.89, 2.14) 72 (66, 77)
1.08
(1.05, 1.12) 7 (0, 28)
Excluding the first 5 years of follow-up
Coronary heart disease* 29820 1.91(1.78, 2.04) 70 (62,76)
1.09
(1.05, 1.13) 2 (0, 22)
Ischaemic stroke* 3262 1.69(1.46, 1.96) 54 (22,73)
1.06
(0.94, 1.19) 0 (0, 50)
CVD deaths 36543 1.98(1.85, 2.11) 70 (63,76)
1.07
(1.03, 1.11) 8 (0, 31)
Excluding studies recording only fatal MI
Non-fatal MI 15997 1.89(1.74, 2.06) 61 (50,70)
1.13
(1.08, 1.19) 2(0, 22)
Fatal MI 6935 2.10(1.84, 2.39) 57 (42,68)
1.11
(1.03, 1.20) 0(0, 31)
Excluding studies recording only fatal Ischaemic strokes
Fatal Ischaemic stroke 2271 1.63(1.43, 1.85) 0 (0, 85)
1.05
(0.86, 1.27) 0(0, 85)
Non-fatal Ischaemic stroke 64 1.59(0.74, 3.43) 0 (0, 52)
1.08
(0.49, 2.36) 33(0, 63)
Censoring individuals at death only rather than at first non-fatal MI/stroke or death
Fatal MI 28532 2.08(1.93, 2.25) 71 (65,77)
1.14
(1.09, 1.19) 9 (0, 30)
Fatal Ischaemic stroke 1334 1.71(1.38, 2.12) 39 (6,61)
0.94
(0.80, 1.11) 0 (0, 40)
All fatal CVD 52698 1.95(1.83, 2.07) 75 (70,80)
1.07
(1.03, 1.12) 25 (4, 41)
Restricting to studies including both Males and Females and fitting and interaction with sex
In Females Coronary heart disease* 8569 2.12(1.87, 2.39) 47 (30, 59)**
1.10
(1.03, 1.18) 0 (0, 28)**
Ischaemic stroke* 1505 1.73(1.51, 1.98) 0 (0, 28)**
1.00
(0.84, 1.18) 0 (0, 28)**
CVD deaths 10557
1.95
(1.77, 2.15)
37 (16, 52)** 0.99(0.92, 1.07) 0 (0, 28)**
In Males Coronary heart disease* 16924 1.99(1.86, 2.14) 49 (33, 61)
1.15
(1.09,1.21) 0 (0, 28)
Ischaemic stroke* 1586 1.59(1.39, 1.82) 0 (0, 38)
0.96
(0.82,1.11) 3 (0, 39)
CVD deaths 15072 2.09(1.95, 2.25) 43 (24, 57)
1.06
(1.00,1.12) 0 (0, 29)
*Includes both fatal and non-fatal events
** I2 for the interaction term (the effect for Female was computed adding the interaction term
on top of the effect for Male). Risk ratios were adjusted for age at baseline and BMI and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of the outcome.
137
Table 4.2: : Risk ratios for coronary heart disease, ischaemic stroke and all cardiovascular mortality for current versus never smokers, with
progressive adjustment for baseline levels of biological, socioeconomic and behavioural risk factors
Coronary heart disease* Ischaemic stroke* All CVD deaths
Progressive adjustment
No of
events RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
Age, sex and BMI 28,095 2.03 (1.88, 2.18) 72 (65, 77) 4,184 1.66 (1.47, 1.87) 31 (0, 53) 24,547 2.04 (1.88, 2.20) 69 (61, 75)
Plus systolic blood pressure 28,095 2.07 (1.92, 2.23) 72 (66, 77) 4,184 1.69 (1.50, 1.90) 31 (0, 53) 24,547 2.08 (1.92, 2.25) 69 (61, 75)
Plus history of diabetes 28,095 2.05 (1.91, 2.21) 71 (64, 76) 4,184 1.70 (1.52, 1.92) 32 (0, 53) 24,547 2.08 (1.92, 2.24) 69 (61, 75)
Plus total cholesterol 28,095 2.01 (1.88, 2.16) 68 (60, 74) 4,184 1.72 (1.53, 1.94) 31 (0, 53) 24,547 2.08 (1.93, 2.24) 67 (58, 73)
Additional adjustment
Lifestyle factors
Age, sex and BMI 19,175 2.02 (1.87, 2.19) 60 (46, 70) 3,126 1.75 (1.49, 2.05) 46 (16, 65) 15,230 2.08 (1.90, 2.28) 60 (46, 71)
Plus education 19,175 1.98 (1.84, 2.14) 57 (42, 69) 3,126 1.73 (1.48, 2.03) 46 (16, 65) 15,230 2.07 (1.89, 2.27) 60 (45, 71)
Age, sex and BMI 14,423 1.90 (1.75, 2.06) 51 (32, 65) 1,733 1.52 (1.33, 1.74) 0 (0, 47) 12,933 1.87 (1.70, 2.06) 58 (40, 70)
Plus occupation/job 14,423 1.87 (1.72, 2.03) 49 (29, 64) 1,733 1.46 (1.10, 1.94) 0 (0, 47) 12,933 1.86 (1.68, 2.06) 57 (39, 69)
Age, sex and BMI 13,159 1.96 (1.80, 2.13) 55 (41, 65) 2,916 1.74 (1.49, 2.02) 28 (0, 53) 10,934 1.87 (1.70, 2.06) 51 (35, 63)
Plus alcohol consumption 13,159 2.01 (1.84, 2.19) 56 (42, 66) 2,916 1.79 (1.54, 2.07) 24 (0, 51) 10,934 1.86 (1.68, 2.06) 52 (37, 64)
Lipids
Basic model$ 11,446 1.95 (1.76, 2.15) 58 (44, 68) 3,203 1.75 (1.58, 1.94) 0 (0, 41) 6,282 1.98 (1.78, 2.22) 43 (22, 58)
Plus non-HDL-C, HDL-C
& loge triglycerides
+ 11,446 1.91 (1.75, 2.08) 50 (33, 63) 3,203 1.76 (1.57, 1.97) 0 (0, 41) 43,861 2.03 (1.81, 2.27) 44 (23, 59)
Inflammatory markers
Basic model$$ 7,240 1.87 (1.63, 2.15) 60 (43, 72) 2,150 1.72 (1.32, 2.23) 63 (37, 78) 4,260 2.11 (1.84, 2.42) 41 (12, 61)
Plus loge CRP 7,240 1.70 (1.48, 1.95) 59 (43, 71) 2,150 1.59 (1.24, 2.04) 59 (30, 76) 4,260 1.87 (1.65, 2.13) 34 (1, 56)
Basic model$$ 6,748 2.08 (1.87, 2.30) 44 (18, 62) 2,445 1.76 (1.46, 2.12) 44 (9, 65) 4,259 2.23 (1.93, 2.59) 55 (35, 69)
Plus fibrinogen 6,748 1.95 (1.76, 2.15) 39 (10, 59) 2,445 1.68 (1.40, 2.01) 40 (4, 63) 4,259 2.04 (1.78, 2.35) 49 (25, 65)
*Includes both fatal and non-fatal events. $: The basic model is adjusted for age, sex, systolic blood pressure, history of diabetes and body-mass index. $$: The
basic model is adjusted for age, sex, systolic blood pressure, history of diabetes, body-mass index and total cholesterol. +: total cholesterol was not included in this
model. RRs were computed within cohorts and combined using random effect meta-analysis. Study-specific RRs were adjusted as shown and stratified where
appropriate by sex and trial arm. Studies recording fewer than 5 events during follow-up were excluded from the analysis of that outcome.
138
Table 4.3: Tests of the proportional hazard assumption
Current versus never smokers Ex versus never smokers
Outcome
Change (95% CI) in
RRs per 5 years
additional follow-up
I2 (95%
CI)
Change (95% CI) in
RRs per 5 years
additional follow-up I2 (95% CI)
Overall
P-value
All cardiovascular -0.09 (-0.12; -0.06) 26 (4, 43) -0.03 (-0.05; -0.01) 2 (0, 21) 0
CHD death and non-fatal MI -0.10 (-0.13; -0.08) 7 (0, 29) -0.04 (-0.07; -0.02) 0 (0, 26) 0
All cerebrovascular events -0.07 (-0.12; -0.01) 33 (11, 49) 0.00 (-0.04; 0.04) 0 (0, 28) 0.065
Ischaemic stroke -0.06 (-0.15; 0.04) 9 (0, 37) 0.06 (-0.04; 0.18) 0 (0, 34) 0.29
Haemorrhagic stroke 0.05 (-0.05; 0.16) 0 (0, 36) 0.09 (-0.04; 0.24) 0 (0, 37) 0.236
Subarachnoid haemorrhage -0.19 (-0.33; -0.03) 14 (0, 47) -0.17 (-0.30; -0.03) 0 (0, 45) 0.005
Unclassified stroke (fatal) -0.04 (-0.13; 0.06) 21 (0, 49) -0.03 (-0.11; 0.07) 0 (0, 42) 0.605
Heart failure (fatal) -0.03 (-0.17; 0.14) 17 (0, 47) -0.01 (-0.23; 0.25) 36 (0, 59) 0.923
Sudden death (fatal) -0.15 (-0.25; -0.04) 0 (0, 51) 0.02 (-0.11; 0.17) 0 (0, 58) 0.028
Cardiac dysrhythmia (fatal) -0.16 (-0.29; 0.01) 0 (0, 48) -0.10 (-0.27; 0.11) 0 (0, 49) 0.108
Pulmonary embolism (fatal) 0.06 (-0.10; 0.25) 0 (0, 54) 0.08 (-0.11; 0.31) 0 (0, 54) 0.565
Aortic aneurysm (fatal) 0.12 (-0.05; 0.33) 10 (0, 45) 0.09 (-0.10; 0.33) 0 (0, 50) 0.285
Peripheral vascular disease
(fatal) -0.20 (-0.48; 0.24) 23 (0, 68) -0.33 (-0.53; -0.04) 0 (0, 79) 0.058
RR: Risk ratio. The proportional hazard assumption was tested by adding a time dependant interaction to study specific models, with adjustment for age and
body mass index, and stratification, when appropriate, by sex and trial arm. The change in RRs per 5 years additional follow-up was obtained by taking the
pooled time dependant interaction estimate multiplied by 5, exponentiating it and subtracting 1. Overall p-values are derived from a 2 degree of freedom test
of nullity of the time dependant interactions, testing jointly the significance of current versus never and ex versus never interaction estimates. All
cerebrovascular events include ischaemic stroke, haemorrhagic stroke, unclassified strokes and subarachnoid haemorrhage.
139
Table 4.4: : Risk ratios for coronary heart disease, ischaemic stroke and all cardiovascular mortality for ex versus never smokers, with
progressive adjustment for baseline levels of biological, socioeconomic and behavioural risk factors
Coronary heart disease* Ischaemic stroke* All CVD deaths
Progressive adjustment
No of
events RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
Age, sex and BMI 28,095 1.14 (1.10, 1.18) 1 (0, 26) 4,184 1.01 (0.92, 1.11) 7 (0, 35) 24,547 1.08 (1.02, 1.14) 12 (0, 33)
Plus systolic blood pressure 28,095 1.15 (1.10, 1.19) 7 (0, 29) 4,184 1.01 (0.92, 1.11) 6 (0, 34) 24,547 1.08 (1.03, 1.14) 11 (0, 33)
Plus history of diabetes 28,095 1.14 (1.09, 1.18) 7 (0, 29) 4,184 1.03 (0.94, 1.14) 5 (0, 33) 24,547 1.08 (1.02, 1.13) 12 (0, 33)
Plus total cholesterol 28,095 1.12 (1.08, 1.17) 11 (0, 32) 4,184 1.04 (0.94, 1.15) 6 (0, 34) 24,547 1.07 (1.02, 1.13) 13 (0, 34)
Additional adjustment
Lifestyle factors
Age, sex and BMI 19,175 1.15 (1.10, 1.20) 4 (0, 30) 3,126 0.94 (0.85, 1.05) 8 (0, 39) 15,230 1.06 (1.00, 1.12) 4 (0, 28)
Plus education 19,175 1.14 (1.07, 1.21) 2 (0, 24) 3,126 0.95 (0.84, 1.08) 7 (0, 38) 15,230 1.08 (1.02, 1.15) 12 (0, 37)
Age, sex and BMI 14,423 1.10 (1.05, 1.16) 0 (0, 34) 1,733 0.97 (0.81, 1.17) 25 (0, 56) 12,933 1.08 (0.99, 1.17) 28 (0, 51)
Plus occupation/job 14,423 1.10 (1.03, 1.18) 0 (0, 34) 1,733 0.98 (0.74, 1.30) 26 (0, 57) 12,933 1.07 (0.99, 1.16) 27 (0, 50)
Age, sex and BMI 13,159 1.10 (1.04, 1.17) 15 (0, 37) 2,916 0.97 (0.86, 1.09) 0 (0, 39) 10,934 1.06 (1.00, 1.13) 7 (0, 31)
Plus alcohol consumption 13,159 1.12 (1.05, 1.19) 17 (0, 39) 2,916 0.99 (0.88, 1.12) 0 (0, 39) 10,934 1.10 (1.02, 1.18) 8 (0, 32)
Lipids
Basic model$ 11,446 1.15 (1.09, 1.22) 0 (0, 30) 3,203 1.09 (0.97, 1.22) 0 (0, 41) 6,282 1.09 (1.01, 1.18) 6 (0, 33)
Plus non-HDL-C, HDL-C & loge triglycerides
+
11,446 1.15 (1.09, 1.22) 0 (0, 30) 3,203 1.11 (0.99, 1.24) 0 (0, 41) 6,282 1.09 (1.00, 1.19) 2 (0, 26)
Inflammatory markers
Basic model$$ 7,240 1.14 (1.06, 1.22) 0 (0, 36) 2,150 1.14 (1.00, 1.31) 3 (0, 52) 4,260 1.16 (1.03, 1.30) 16 (0, 45)
Plus loge CRP 7,240 1.11 (1.04, 1.19) 0 (0, 36) 2,150 1.13 (0.99, 1.28) 0 (0, 51) 4,260 1.13 (1.00, 1.27) 17 (0, 46)
Basic model$$ 6,748 1.15 (1.06, 1.25) 8 (0, 38) 2,445 1.02 (0.90, 1.16) 0 (0, 44) 4,259 1.12 (1.02, 1.23) 0 (0, 37)
Plus fibrinogen 6,748 1.15 (1.07, 1.24) 7 (0, 37) 2,445 1.01 (0.88, 1.14) 0 (0, 44) 4,259 1.11 (1.01, 1.22) 0 (0, 37)
*Includes both fatal and non-fatal events. $: The basic model is adjusted for age, sex, systolic blood pressure, history of diabetes and body-mass index. $$:
The basic model is adjusted for age, sex, systolic blood pressure, history of diabetes, body-mass index and total cholesterol. +: total cholesterol was not
included in this model. RRs were computed within cohorts and combined using random effect meta-analysis. Study-specific RRs were adjusted as shown and
stratified where appropriate by sex and trial arm. Studies recording fewer than 5 events during follow-up were excluded from the analysis of that outcome.
140
Table 4.5: Risk ratios for CVD and fatal and non-fatal MI by number of years since stopping smoking, with progressive adjustment for age started smoking
and past number of pack-years.
No of
events
RRs
(95% CI)
for current versus
never smokers
RRs (95% CI) for ex versus never smokers
Outcome Model adjusted for <5 years 5-10 years 10-20 years ≥20 years
CHD deaths and non-fatal MI Ages & BMI 15578 2.16 (2.03, 2.29) 1.36 (1.20, 1.55) 1.23 (1.06, 1.43) 1.21 (1.07, 1.36) 1.02 (0.92, 1.14)
+ Age started 15578 2.04 (1.92, 2.18) 1.28 (1.12, 1.45) 1.15 (0.98, 1.34) 1.13 (1.00, 1.27) 0.94 (0.84, 1.06)
Ages & BMI 24905 2.01 (1.89, 2.14) 1.40 (1.26, 1.56) 1.18 (1.04, 1.34) 1.03 (0.92, 1.15) 0.85 (0.76, 0.96)
+ Past number of pack-years 24905 1.78 (1.65, 1.92) 1.23 (1.10, 1.38) 1.06 (0.93, 1.21) 0.94 (0.84, 1.06) 0.82 (0.73, 0.93)
All CVD events Ages & BMI 28769 1.92 (1.76, 2.10) 1.26 (1.12, 1.41) 1.12 (0.99, 1.26) 1.09 (0.99, 1.20) 0.93 (0.84, 1.02)
+ Age started 28769 1.87 (1.77, 1.98) 1.20 (1.09, 1.33) 1.05 (0.93, 1.18) 1.02 (0.93, 1.12) 0.86 (0.78, 0.94)
Ages & BMI 44547 2.07 (1.92, 2.24) 1.64 (1.49, 1.80) 1.27 (1.14, 1.43) 1.00 (0.90, 1.11) 0.93 (0.84, 1.04)
+ Past number of pack-years 44547 1.66 (1.55, 1.77) 1.24 (1.12, 1.36) 0.99 (0.88, 1.11) 0.83 (0.75, 0.92) 0.85 (0.76, 0.95)
Lung cancer deaths Ages & BMI 2301 14.81 (11.88, 18.47) 10.49 (7.98, 13.78) 6.40 (4.57, 8.97) 3.84 (2.76, 5.34) 2.09 (1.47, 2.97)
+ Age started 2301 10.80 (8.55, 13.64) 7.38 (5.55, 9.82) 4.41 (3.11, 6.24) 2.64 (1.88, 3.71) 1.38 (0.96, 1.99)
Ages & BMI 15039 2.29 (2.10, 2.50) 1.93 (1.68, 2.22) 1.37 (1.15, 1.63) 1.21 (1.05, 1.41) 1.11 (0.95, 1.31)
+ Past number of pack-years 15039 1.66 (1.50, 1.84) 1.34 (1.15, 1.56) 1.01 (0.84, 1.21) 0.98 (0.84, 1.15) 1.04 (0.89, 1.23)
No: Number; BMI: Body Mass Index; CHD: Coronary heart disease; CVD: Cardiovascular diseases; RR: Risk ratio; Study-specific loge risk ratios stratified,
where appropriate, by sex and trial arm; were combined using a multivariate random-effects meta-analysis. Studies with fewer than 10 individuals
experiencing an outcome were excluded from the analysis of that outcome.
141
Table 4.6: Proportion of individuals changing their smoking status compared to
baseline according to time since baseline
Smoking status
Time between
baseline and resurvey
Stable status
Never smoker
becomes
current smoker
Current
smoker
becomes
ex-smoker
Never
smokers
becomes
Ex-smoker
Total
No
individual
resurveys
<5 years 88.3% 0.1% 7.0% 4.6% 236,639
5-10 years 88.3% 0.3% 7.9% 3.5% 158,646
10-15 years 89.2% 0.6% 8.8% 1.4% 34,846
15-20 years 85.6% 0.4% 12.3% 1.4% 12,835
20-30 years 77.9% 0.3% 20.5% 1.3% 10,715
30-40 years 66.0% 0.8% 28.1% 5.1% 2,093
Total No individual
resurveys
400,794 1,083 36,695 17,202 455,774
Total proportions 88.0% 0.2% 8.1% 3.8% 100%
No: Number. Percentages are row percentages and are computed based on 455,774
individual resurveys. Individuals who were never smokers and changed their status to ex-
smoker at the next resurvey were understood to have started and stopped smoking during
the time interval between the two surveys, or to have answered the question incorrectly.
142
Reference List
(1) Dept.of Health and Human Services. The health consequences of smoking: a report
of the Surgeon General. 2004.
(2) Pirie K, Peto R, Reeves GK, Green J, Beral V. The 21st century hazards of smoking
and benefits of stopping: a prospective study of one million women in the UK. Lancet
2012 October 26.
(3) Ageno W, Becattini C, Brighton T, Selby R, Kamphuisen PW. Cardiovascular risk
factors and venous thromboembolism: a meta-analysis. Circulation 2008 January
1;117(1):93-102.
(4) Willigendael EM, Teijink JA, Bartelink ML et al. Influence of smoking on incidence
and prevalence of peripheral arterial disease. J Vasc Surg 2004
December;40(6):1158-65.
(5) Feigin VL, Rinkel GJE, Lawes CMM et al. Risk Factors for Subarachnoid
Hemorrhage. Stroke 2005 December 1;36(12):2773-80.
(6) Pomp ER, le CS, Rosendaal FR, Doggen CJ. Risk of venous thrombosis: obesity and
its joint effect with oral contraceptive use and prothrombotic mutations. Br J Haematol
2007 October;139(2):289-96.
(7) D'Alessandro A, Boeckelmann I, Hammwhoner M, Goette A. Nicotine, cigarette
smoking and cardiac arrhythmia: an overview. Eur J Cardiovasc Prev Rehabil 2011
May 25.
(8) Danaei G, Finucane MM, Lu Y et al. National, regional, and global trends in fasting
plasma glucose and diabetes prevalence since 1980: systematic analysis of health
examination surveys and epidemiological studies with 370 country-years and 2.7
million participants. Lancet 2011 July 2;378(9785):31-40.
(9) Finucane MM, Stevens GA, Cowan MJ et al. National, regional, and global trends in
body-mass index since 1980: systematic analysis of health examination surveys and
epidemiological studies with 960 country-years and 9.1 million participants. Lancet
2011 February 12;377(9765):557-67.
(10) Farzadfar F, Finucane MM, Danaei G et al. National, regional, and global trends in
serum total cholesterol since 1980: systematic analysis of health examination
surveys and epidemiological studies with 321 country-years and 3.0 million
participants. Lancet 2011 February 12;377(9765):578-86.
(11) Sarwar N, Gao P, Seshasai SR et al. Diabetes mellitus, fasting blood glucose
concentration, and risk of vascular disease: a collaborative meta-analysis of 102
prospective studies. Lancet 2010 June 26;375(9733):2215-22.
(12) Di AE, Sarwar N, Perry P et al. Major lipids, apolipoproteins, and risk of vascular
disease. JAMA 2009 November 11;302(18):1993-2000.
(13) Wormser D, Kaptoge S, Di AE et al. Separate and combined associations of body-
mass index and abdominal adiposity with cardiovascular disease: collaborative
analysis of 58 prospective studies. Lancet 2011 March 26;377(9771):1085-95.
(14) Huxley RR, Woodward M. Cigarette smoking as a risk factor for coronary heart
disease in women compared with men: a systematic review and meta-analysis of
prospective cohort studies. Lancet 2011 August 10.
(15) Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in
52 countries in the INTERHEART study: a case-control study. Lancet 2006
August;%19;368(9536):647-58.
143
(16) U.S.Department of Health and Human Services. How tobacco smoke causes
disease: the biology and behavioral basis of smoking-attributable disease: a report of
the surgeon general. 2010.
(17) Lubin JH, Alavanja MC, Caporaso N et al. Cigarette smoking and cancer risk:
modeling total exposure and intensity. Am J Epidemiol 2007 August 15;166(4):479-
89.
(18) Samet JM, Thun MJ, de Gonzalez AB. Models of smoking and lung cancer risk: a
means to an end. Epidemiology 2007 September;18(5):649-51.
(19) Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in
52 countries in the INTERHEART study: a case-control study. Lancet 2006
August;%19;368(9536):647-58.
(20) Kenfield SA, Stampfer MJ, Rosner BA, Colditz GA. Smoking and smoking cessation
in relation to mortality in women. JAMA 2008 May 7;299(17):2037-47.
(21) Doll R, HILL AB. Mortality in relation to smoking: ten years' observations of British
Doctors. Br Med J 1964 May 30;1(5395):1399-410.
(22) Cook DG, Shaper AG, Pocock SJ, Kussick SJ. Giving up smoking and the risk of
heart attacks. A report from The British Regional Heart Study. Lancet 1986
December 13;2(8520):1376-80.
(23) Doll R, Peto R, Boreham J, Sutherland I. Mortality in relation to smoking: 50 years'
observations on male British doctors. BMJ 2004 June 26;328(7455):1519.
(24) European Union Directorate-General for Health and Consumers. Data and
Information on Women's Health in the European Union. 2009.
(25) Thompson S, Kaptoge S, White I, Wood A, Perry P, Danesh J. Statistical methods for
the time-to-event analysis of individual participant data from multiple epidemiological
studies. Int J Epidemiol 2010 October;39(5):1345-59.
(26) Emerging Risk Factors Collaboration. Lipoprotein(a) concentration and the risk of
coronary heart disease, stroke, and nonvascular mortality. JAMA 2009 July
22;302(4):412-23.
(27) Emerging Risk Factors Collaboration. Major lipids, apolipoproteins, and risk of
vascular disease. JAMA 2009 November 11;302(18):1993-2000.
(28) Emerging Risk Factors Collaboration. C-reactive protein concentration and risk of
coronary heart disease, stroke, and mortality: an individual participant meta-analysis.
Lancet 2010 January 9;375(9709):132-40.
(29) Emerging Risk Factors Collaboration. Diabetes mellitus, fasting blood glucose
concentration, and risk of vascular disease: a collaborative meta-analysis of 102
prospective studies. Lancet 2010 June 26;375(9733):2215-22.
(30) Emerging Risk Factors Collaboration. Diabetes mellitus, fasting glucose, and risk of
cause-specific death. N Engl J Med 2011 March 2;364(9):829-41.
(31) Emerging Risk Factors Collaboration. Separate and combined associations of body-
mass index and abdominal adiposity with cardiovascular disease: collaborative
analysis of 58 prospective studies. Lancet 2011 March 11.
(32) Prentice RL, Kalbfleisch JD. Hazard rate models with covariates. Biometrics 1979
March;35(1):25-39.
144
(33) Kirkwood B, Sterne J. Logistic regression: controlling for confounding and other
extensions. In: Blackwell Science, editor. Medical statistics. 2 ed. Massachusetts,
USA: 2003. p. 212-3.
(34) Riley RD, Higgins JP, Deeks JJ. Interpretation of random effects meta-analyses. BMJ
2011 February 10;342.
(35) Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med
2002 June 15;21(11):1539-58.
(36) Jackson D, White IR, Thompson SG. Extending DerSimonian and Laird's
methodology to perform multivariate random effects meta-analyses. Stat Med 2009
April 30.
(37) Easton DF, Peto J, Babiker AG. Floating absolute risk: an alternative to relative risk
in survival and case-control analysis avoiding an arbitrary reference group. Stat Med
1991 July;10(7):1025-35.
(38) Whitlock G, Lewington S, Sherliker P et al. Body-mass index and cause-specific
mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet
2009 March 28;373(9669):1083-96.
(39) West Midlands Public Health Observatory. MetaData; European Standard
Population. 2011.
(40) Thun MJ, Day-Lally C, Myers D et al. Trends in tobacco smoking and mortality from
cigarette use in Cancer Prevention Studies I (1959 through 1965) and II (1982
through 1988). Changes in Cigarette-Related Disease Risks and Their Implication for
Prevention and Control.Smoking and Tobacco Control Monograph No.8.Rockville
(MD) [a], 305-382. 1997. Shopland DR, Burns DM, Garfinkel L, Samet JM.
(41) Thun MJ, Apicella LF, Henley SJ. Smoking vs other risk factors as the cause of
smoking-attributable deaths: confounding in the courtroom. JAMA 2000 August
9;284(6):706-12.
(42) Doll R, Peto R. Mortality in relation to smoking: 20 years' observations on male
British doctors. Br Med J 1976 December 25;2(6051):1525-36.
(43) Last J. A Dictionary of Epidemiology. 4th edition ed. Oxford University Press; 2001.
(44) Symons JD, Stebbins CL. Hemodynamic and regional blood flow responses to
nicotine at rest and during exercise. Med Sci Sports Exerc 1996 April;28(4):457-67.
(45) Michael PR. Cigarette smoking, endothelial injury and cardiovascular disease. Int J
Exp Pathol 2000 August;81(4):219-30.
(46) Burke A, Fitzgerald GA. Oxidative stress and smoking-induced vascular injury. Prog
Cardiovasc Dis 2003 July;46(1):79-90.
(47) Cornuz J, Sidoti PC, Tevaearai H, Egger M. Risk factors for asymptomatic abdominal
aortic aneurysm: systematic review and meta-analysis of population-based screening
studies. Eur J Public Health 2004 December;14(4):343-9.
(48) Lowe GD. Common risk factors for both arterial and venous thrombosis. Br J
Haematol 2008 March;140(5):488-95.
(49) Hoffman RM, Psaty BM, Kronmal RA. Modifiable risk factors for incident heart failure
in the coronary artery surgery study. Arch Intern Med 1994 February 28;154(4):417-
23.
145
(50) Metz L, Waters DD. Implications of cigarette smoking for the management of patients
with acute coronary syndromes. Prog Cardiovasc Dis 2003 July;46(1):1-9.
(51) Kurth T, Kase CS, Berger K, Schaeffner ES, Buring JE, Gaziano JM. Smoking and
the risk of hemorrhagic stroke in men. Stroke 2003 May;34(5):1151-5.
(52) Juvela S, Poussa K, Porras M. Factors affecting formation and growth of intracranial
aneurysms: a long-term follow-up study. Stroke 2001 February;32(2):485-91.
(53) Howard G, Wagenknecht LE, Burke GL et al. Cigarette smoking and progression of
atherosclerosis: The Atherosclerosis Risk in Communities (ARIC) Study. JAMA 1998
January 14;279(2):119-24.
(54) Omvik P. How smoking affects blood pressure. Blood Press 1996 March;5(2):71-7.
(55) Price JF, Mowbray PI, Lee AJ, Rumley A, Lowe GD, Fowkes FG. Relationship
between smoking and cardiovascular risk factors in the development of peripheral
arterial disease and coronary artery disease: Edinburgh Artery Study. Eur Heart J
1999 March;20(5):344-53.
(56) Celermajer DS, Adams MR, Clarkson P et al. Passive smoking and impaired
endothelium-dependent arterial dilatation in healthy young adults. N Engl J Med 1996
January 18;334(3):150-4.
(57) Caraballo RS, Giovino GA, Pechacek TF et al. Racial and ethnic differences in serum
cotinine levels of cigarette smokers: Third National Health and Nutrition Examination
Survey, 1988-1991. JAMA 1998 July 8;280(2):135-9.
(58) Nawrot TS, Perez L, Kunzli N, Munters E, Nemery B. Public health importance of
triggers of myocardial infarction: a comparative risk assessment. Lancet 2011
February 26;377(9767):732-40.
(59) White IR, Frost C, Tomar SL. Correcting for measurement error in binary and
continuous variables using replicates. Statistics in Medicine 2001;20:3441-57.
146
Chapter 5: Pipe or cigar smoking and the risk of cardiovascular
diseases and lung cancer deaths in developed countries
Summary
The effects of types of smoking other than cigarettes have remained relatively little
studied due to lower prevalence in Western countries. The large scope of the ERFC
provides information on pipe and cigar smoking on a unique scale. In total, 83 studies
provided information on at least cigarette smoking status, 22 studies on pipe status
and 20 studies on cigar status. The prevalence of current smoking of cigarettes only
was 34%, cigar only 8% and pipe only 5% in the subsets of data which provided
information on these types of smoking. Cigar and pipe smoking was mainly a male
habit. Current cigarette-only smokers experienced a total of 13,077 fatal and non-
fatal MI, current pipe-only smokers 712 such events, current cigar-only smokers 356
such events; and current pipe or cigar on top of cigarette smokers, a total of 678
events. All forms of smoking were significantly associated with CVD risk. Compared
to never smokers, RRs for all CVD events were 1.31 (1.19; 1.44) for cigar-only
smokers, 1.68 (1.56; 1.81) for pipe-only smokers; 1.97 (1.85; 2.11) for cigarette-only
users and 1.95 (1.81; 2.10) for individuals who were secondary pipe or cigar smokers
(and also smoked cigarettes). All these RRs were higher than the risk of past
smokers, which was 1.09 (1.05; 1.13) compared to never smokers. I observed a
nonlinear dose response relationship between amount of pipe and cigar smoking with
the risk of MI and all CVD, with strongest increases in risk experienced at low
smoking intensities. The strength of association of pipe or cigar smoking with risk of
CVD was not affected by adjustment for conventional lifestyle and biochemical risk
factors of CVD. These novel findings represent the first demonstration of the effects
of pipe or cigar smoking on the risk of major cardiovascular events, using large scale
prospective cohort studies and after appropriate adjustment for cardiovascular risk
factors. As cardiovascular diseases remain the main cause of death in developed
countries, it is important that public health policies address pipe and cigar smoking in
addition to cigarette smoking.
147
5.1 Background
The prevalence of pipe and cigar smoking has been decreasing over the past 50
years in Western Europe. In 2007, the proportion of British men smoking cigars
attained 2%, compared to 34% in 1974, and the number of women smoking cigars
was very low and scarcely measurable 1. In the same year, only 1% of men in the UK
said they smoked a pipe, and they were almost all aged 50 and over. In contrast,
there is evidence of a rise in cigar smoking in the USA, where they represent an
increasingly popular alternative to cigarettes, especially amongst women and
younger age groups, to the extent that the proportion of teenagers initiating cigar
smoking is presently higher than the proportion initiating cigarettes 2. Whilst the
proportion of the US population declaring themselves as current cigar users
decreased by 66% between 1964 and 1992, it has risen by nearly 50% between
1993 and 1996, and continues to increase 3. In 2008, 5% of American smoked cigars
and 0.7% smoked tobacco in pipes, representing respectively 13 million and nearly 2
million individuals4. This resurgence of cigar smoking has been attributed to
campaigns of glamorization of the image of cigars as a substitute for cigarettes 5. In a
study conducted in the USA, UK, Australia and Canada, O’Connor showed that a
quarter of smokers in these countries believe that pipes, cigars or roll-your-own
cigarettes are safer than manufactured cigarettes6.
However, this belief is based on relatively scarce epidemiological evidence, and the
relationship between cigar and pipe smoking and CVD risk remains uncertain. Early
studies such as the British Doctors Study found a non-significant increased risk when
smoking cigar, while cigarette smoking was a highly significant risk factor 7. In this
study, pipe and cigar smokers were defined as those who had never smoked
cigarettes. These primary smokers tend not to inhale and so are exposed to a
relatively low amount of tar and other harmful constituents of the tobacco smoke
compared with cigarette smokers. Nowadays, as most cigar smokers are former
cigarette smokers, they are likely to have transferred their inhalation techniques,
despite the greater irritancy of the pipe and cigar smoke, and it is plausible that their
risk would have reached significant levels. Two more studies based in the USA and
conducted in 1960’s and 1980’s found that cigar smokers face a CHD risk around 1/3
higher than those who never smoked 8,9. In a study on over 7,000 men in the UK,
pipe or cigar smokers combined experienced a 70% higher risk of major CHD events,
whether or not they had smoked cigarettes previously, but data was too scarce to
disentangle the effect of cigars from that of pipes 10. In a study in Norway, men who
148
switched from smoking cigarettes only to smoking pipes only experienced the same
risk of death as people who remained cigarette only users, and individual who had
only ever smoked pipe had double the risk of dying from all causes compared to
never smokers 11.
In this context, the aim of this Chapter is to investigate the association between pipe
and cigar smoking with CVD risk in Western populations in more detail than has been
previously possible. The association with lung cancer is also briefly presented for
comparison.
5.2 Methods
5.2.1 The dataset
Study design and statistical methods are detailed in Chapter 1. The dataset was
restricted to studies which provided information on at least cigarette smoking status.
Individuals with missing information on pipe (respectively cigar) smoking status
because (1) the study did not ask this information to participants, (2) the study did not
provide it to the ERFC coordinating centre or (3) individuals refused to provide the
information, were coded as non-current pipe (respectively cigar) smokers (Table
5.1). This is a reasonable assumption because cigar and pipe smoking are relatively
rare habits in the general population and studies included in the ERFC were
prospective cohorts sampled from the general population 12. In a subsidiary analysis,
only studies which provided information on all three types of smoking, namely
cigarettes, pipes and cigars were used. All individuals who gave information on
whether they were current, ex or never pipe or cigar smokers also provided
information on cigarette smoking status. Smoking type was defined as “never
smoker”, “ex-smoker”, “current cigarette smoking only”, “current pipe smoking only”,
“current cigarette and cigar/pipe smoking”. “Primary pipe or cigar smokers” were
defined as individuals who currently smoked pipes or cigars but not cigarettes,
“secondary pipe or cigar smokers” as individuals who currently smoked pipes or
cigars as well as cigarettes, and “exclusive pipe or cigar smokers” as individuals who
currently smoked pipes or cigars and had never smoked cigarettes. Outcomes of
interest were fatal and non-fatal MI, fatal and non-fatal cerebrovascular events, all
fatal and non-fatal CVD events and fatal lung cancer. Individuals were censored at
their first non-fatal MI or cerebrovascular event and at any cause of death.
149
5.2.2 Statistical methods
Statistical methods are the same as the methods presented in Chapter 3. Risk
Ratios (RRs) and 95% Confidence Intervals were estimated using Cox proportional
hazard models for prospective cohort studies and clinical trials; and using logistic
models for nested case-control studies. Models were adjusted at least for age and
Body Mass Index and stratified by sex and trial arm where appropriate. For graphical
representation, estimates were represented by boxes with sizes inversely
proportional to their variances, and confidence intervals were represented using
floating absolute variances 13. Robustness of estimates was tested by progressive
adjustment for traditional cardiovascular risk factors, restricting the dataset to studies
providing information on these additional variables. To investigate heterogeneity of
effects across studies, I2 value and its 95% confidence interval was calculated for
each estimate. Subgroup analysis was done by age group, history of diabetes, use of
alcohol, race and geographical region. Interaction between smoking type and other
risk factors on raising the risk for CVD were tested using formal tests of interaction.
All estimates were computed in two steps. In a first step, adjusted estimates were
calculated for each study and in a second step, study-specific estimates were
combined using random effect multivariate meta-analyses. Estimates and tests with
P-value of significance <0.001 were emphasized.
5.3 Results
5.3.1 Description of the dataset
In total, 83 studies provided information on at least cigarette smoking status, 22
studies on pipe status and 20 studies on cigar status (Figure 5.1). Assuming that
individuals with missing information on cigar or pipe status were not cigar or pipe
users, information was available for (in non-overlapping subsets of data) 235,776
cigarette only smokers, 6,255 pipe only smokers, 4,162 cigar only smokers and
5,875 smokers of pipes or cigars on top of cigarettes. During an average follow-up
time of 15.6 years, current cigarettes only smokers experienced a total of 13,077 fatal
and non-fatal MI, current pipe only smokers 712 such events, current cigar only
smokers 356 such events; and current pipe or cigar on top of cigarette smokers, a
total of 678 events. There were 15 additional studies with information on pipe or cigar
smoking, including 2,350 current pipe or cigar users, for which information on
whether either pipe or cigar was smoked was not available. These individuals
150
experienced 202 fatal and non-fatal MI events during an average follow-up of 12.5
years.
5.3.2 Baseline characteristics of pipe and cigar smokers
Half of participants who smoked either pipes only or cigars only were past cigarette
smokers. More than a third (37%) of individuals who smoked pipes or cigars also
declared themselves to be current cigarette smokers. While 44% of current cigarette
only smokers were women, the proportions were <1% of pipe and 7% of cigar only
smokers (Table 5.1a). Pipe and cigar smokers were also overwhelmingly White
(respectively 99% and 97%) whilst non White individuals represented 12% of
cigarette smokers (6% Asian or Oriental and 4% Black). Pipe and cigar smokers
were ~20% more likely to be current alcohol drinkers than cigarette only smokers. A
higher proportion of pipe only users had not reached secondary education (36%
versus 22% for cigarette users), whilst the level of education of cigar smokers was
similar to that of cigarette smokers. Individuals in manual jobs and in services were
over-represented amongst cigar smokers (37% and 41% respectively versus 30%
and 28% for cigarette smokers).
Levels of conventional risk factors were slightly higher in cigar smokers, and to a
lesser extent in pipe smokers (Table 5.1b). Compared to cigarette only users, cigar
smokers had higher levels of anthropometric (Body Mass Index and Waist to Hip
Ratio), hemodynamic (systolic and diastolic blood pressure), lipid (total cholesterol,
non HDL-cholesterol, Apolipoprotein B, loge triglycerides, loge Lp(a)), and some
inflammatory markers (loge CRP). Pipe only smokers had slightly raised WHR and
blood pressure levels, but levels of other risk factors were similar to that of cigarette
only smokers.
5.3.3 Pipe, cigar and cigarette smoking in relation to CVD risk
Pipe, cigar and cigarette smoking were all significantly associated with the risk of
fatal and non-fatal cardiovascular events (Figure 5.2). The highest risk was
experienced by cigarette, then pipe, then cigar smokers. Compared to never
smokers, RRs for fatal and non-fatal myocardial infarction were 1.35 (1.20, 1.51) for
cigar only smokers; 1.84 (1.69, 2.01) for pipe only smokers, 2.11 (1.95, 2.28) for
cigarette only smokers. The excess risk of pipe and cigar smokers who also smoked
cigarettes was not significantly different from that of cigarette only users: RR of 2.05
(1.87, 2.24) compared to never smokers. Compared to never smokers, RRs for all
151
CVD events were 1.31 (1.19; 1.44) for cigar only smokers, 1.68 (1.56; 1.81) for pipe
only smokers; 1.97 (1.85; 2.11) for cigarette only users and 1.95 (1.81; 2.10) for
individuals who were secondary pipe or cigar smokers (and also smoked cigarettes).
All these RRs were higher than the risk of past smokers, which was 1.09 (1.05; 1.13)
for all CVD events compared to never smokers.
In a subsidiary analysis restricting the data to studies which provided information on
all three smoking types to enable a within study only comparison of the association
with cigarettes, cigars and pipes, associations were marginally changed. RRs for
fatal and non-fatal MI became 2.04 (1.62; 2.58) for cigarette only users, 1.28 (1.10;
1.48) for cigar only users, 1.80 (1.57; 2.07) for pipe only users and 1.79 (1.40; 2.29)
for individuals combining cigarettes with pipe or cigar use, compared to never
smokers (Figure 5.3). Results were also similar to those found previously for all CVD
events combined.
The association was similar for current pipe or cigar smokers who were past cigarette
smokers and for smokers who had always only smoked cigars or pipes, and it was
significantly lower than that of current cigarette smokers (Figure 5.4). RRs for fatal
and non-fatal MI were 1.56 (1.38; 1.77) for exclusive pipe/cigar smokers and 1.71
(1.56; 1.88) for individuals who had switched from cigarettes to pipes or cigars,
compared to never smokers. For all cardiovascular events, RRs were respectively
1.49 (1.34; 1.65) and 1.57 (1.45; 1.70), compared to never smokers.
The association between pipe and cigar smoking differed according to type of
cardiovascular event (Figure 5.5). It was slightly higher for fatal than for non-fatal
myocardial infarction: 1.71 versus 1.57 for primary pipe or cigar smoking and 2.30
versus 1.96 for secondary pipe or cigar smoking (on top of cigarette smoking),
compared to never smokers. The risk of MI was also higher than that of
cerebrovascular events; which was 1.28 (1.13; 1.46) for primary and 1.95 (1.61; 2.34)
for secondary pipe or cigar smokers, compared to never smokers. The association
with other types of cardiovascular events, which included aortic aneurysm, heart
failure, pulmonary embolism, sudden death and other complication of the heart, was
similar in strength to that of myocardial infarction: 1.75 (1.52; 2.02) for primary and
2.27 (189; 2.73) for secondary pipe or cigar smoking. Levels of heterogeneity
between studies were low and I2 < 50% for all types of smoking and all outcomes
studied.
152
Estimates remained significant after adjusting for systolic blood pressure, history of
diabetes, non HDL-cholesterol, HDL-cholesterol and log triglycerides levels (Table
5.2a). For the risk of fatal and non-fatal MI, RRs for current pipe or cigar smokers
versus never smokers attenuated from 1.58 to 1.53, for current cigarette only from
2.05 to 1.99; and for current cigarette and pipe or cigar from 2.16 to 2.12. Estimates
of the risk of all CVD combined were also altered only marginally (Table 5.2b).
Adjustment for these risk factors explained some of the heterogeneity between
studies: I2 for the risk of MI for current cigarette smokers versus never smokers was
reduced from 40 to 21 and from 49 to 33 for the risk of all cardiovascular events.
Adjustment for alcohol consumption (categorized as never or past alcohol drinker
versus current alcohol drinker), for occupation (categorized as not working, manual,
services, office, student and other) and for level of education reached (no schooling,
primary, secondary, vocational or university), in subsets which provided information
on these covariates, did not significantly modify estimates for the risk of fatal and
non-fatal MI and for the risk of all CVD events.
5.3.4 Pipe, cigar and cigarette smoking in relation to the risk of lung cancer
death
Significant increases in risk of lung cancer death were also observed for all smoking
forms (Figure 5.6). The risk was strongest for cigarette (RR: 13.13; 11.72 to 14.71),
then pipe (10.05; 8.15 to 12.38), then cigar (6.54; 5.00 to 8.55) smokers, compared
to never smokers. Secondary pipe or cigar smokers experienced a risk similar to that
of exclusive cigarette users (RR of 13.90 versus 13.13). In a subsidiary analysis
where data were restricted to studies which provided information on all three smoking
types to enable a within study only comparison of the association, RRs for lung
cancer deaths were even higher for cigarette only and pipe only smokers (Figure
5.7). Individuals who had switched from cigarettes to pipes or cigars had a RR higher
than exclusive pipe or cigar smokers and lower than current cigarette only smokers:
9.77 versus 5.65 and 12.79; but differences were non-significant due to relatively
small numbers of events in each sub-category resulting in wide confidence intervals
(Figure 5.8).
5.3.5 Effects modifications
RRs for MI and all cardiovascular events did not differ significantly by geographical
region (Western Europe/ North America / Other) (p-value of interaction: 0.28) (Figure
5.9 a-b). The effect of being primary pipe/cigar smoker versus never/ex-smoker was
153
highly significant amongst White with a RR of 1.41 (1.30; 1.53) and non-significant
amongst non-White (0.91; 0.58 to 1.42). There was evidence of a stronger effect of
cigarette smoking in younger age groups (RRs of all CVD events of 2.71, 2.05, 1.80
and 1.10 in individuals aged 20-39, 40-59, 60-69 and 70+ years old respectively) and
there was a similar trend for primary pipe or cigar smoking (RRs of 3.52, 1.19, 1.00
and 2.01 for the same age categories) but it was not significant for the latest.
Smoking cigarettes was associated with the greatest increase in risk amongst non-
diabetics (RRs of 1.93 versus 1.51), but there was no such effect modification for
primary pipe or cigar smokers (RRs of 1.45 versus 1.46).
5.3.6 Dose-response relationship
There was evidence of a non-linear dose response relationship with pipe and cigar
smoking combined for fatal and non-fatal MI, after excluding current cigarette
smokers (Figure 5.10). Compared to never smokers of any type, primary pipe or
cigar smokers experienced RRs of 1.43 (1.29; 1.64) for 1-5 cigarettes equivalent per
day, 1.61 (1.40; 1.86) for 5-10 cigarettes equivalent per day, 1.59 (1.27; 1.99) for 10-
15 cigarettes equivalent per day and 1.86 (1.54; 2.25) for ≥15 cigarettes equivalent 
per day. When considering all cardiovascular events combined, the association was
slightly higher, especially at low intensity of pipe or cigar smoking. Compared to
never smokers, RRs were 1.27 (1.13; 1.43); 1.43 (1.27; 1.61); 1.55 (1.28; 1.87) and
1.73 (1.49; 2.02) for <5, 5-10, 10-15 and ≥15 cigarettes equivalent per day 
respectively.
5.4 Discussion
My analysis contains novel findings on the association between smoking types other
than cigarettes in the Western world. The risk factor profile of cigar only or pipe only
smokers was not more favourable than that of cigarette only smokers when
comparing to never smokers. For some risk factors such as non-HDL-c and loge
triglycerides, unadjusted levels were even higher in cigar only smokers than in
cigarette smokers. In addition, this analysis was well powered to show, for the first
time using large scale prospective cohort studies and appropriate adjustment for
cardiovascular risk factors, that both pipe and cigar smoking are associated with the
risk of major CVD events. As CVD remains the main cause of death in developed
countries, it is important that public health policies address pipe and cigar smoking in
addition to cigarette smoking.
154
5.4.1 Main findings
After adjustment for risk factors, age and body mass index (and stratification for sex
and trial arm), the risk for pipe and cigar smokers was significantly higher than that of
never and of ex-smokers. Compared to never smokers, RRs for all CVD events were
1.31 (1.19; 1.44) for cigar only smokers, 1.68 (1.56; 1.81) for pipe only smokers; 1.97
(1.85; 2.11) for cigarette only users and 1.95 (1.81; 2.10) for individuals who were
secondary pipe or cigar smokers (also smoked cigarettes). My estimate for cigar
smoking is in agreement with previous estimates in the US population 14,15. These
findings were also compatible with results on pipe and cigar smoking combined from
a study of the British population 16. Cigar and pipe smoking exert an independent
effect on cardiovascular risk which is likely to be similar to the way cigarette smoking
acts on the vessel wall. Cigar and pipe smoke have been shown to contain the same
toxic and carcinogenic compounds as cigarette smoke 17. The mainstream smoke
from cigars which is drawn into the mouth from the butt end has been shown to
contain greater concentration of nicotine and carbon monoxide than the mainstream
smoke from cigarettes 18. Persons who smoke ≥4 cigars per day are exposed to an 
increased amount of smoke equivalent to the smoke of 10 cigarettes per day 19, and
in this analysis I have shown that there is a positive dose-response relationship
between pipe and cigar amount and risk of CVD which mimics that with cigarette
smoking discussed in Chapter 4, and makes it more likely that the relationship
between cigar and pipe smoking and CVD risk is causal..
The smoke of pipes or cigars is alkaline and generally not inhaled 20. Early studies
such as the British Doctors Study 21 did not observe an increased risk in cigar and
pipe smokers, and this was attributed to the fact that at the time pipe and cigar
smokers rarely inhaled and smoked smaller quantities of tobacco than cigarette
smokers. In my dataset, 53% of cigar only smokers and 45% of pipe only smokers at
baseline declared being ex-cigarette smokers. Individuals who switch from cigarettes
to pipes or cigars are more likely to inhale smoke than exclusive pipe or cigar
smokers 22. An earlier study found a risk of death similar in exclusive pipe smokers, in
cigarette only smokers and in individuals who switched from cigars to pipes 23. Here,
the CVD risk of exclusive pipe smokers was similar to that of individuals who used to
smoke cigarettes whilst a difference was observable on the risk of lung cancer
155
deaths. This finding warrants further study in the Western population to better
understand the role of smoke inhalation in promoting CVD risk.
Associations were not strongly modified upon adjustment for a range of biochemical
and lifestyle risk factors of CVD. In the Cancer Prevention Study, Henley & al found a
significant increase in risk in exclusive pipe smokers of both coronary heart disease
and cerebrovascular diseases compared to never smokers, even after adjusting for a
range of confounders including alcohol consumption, educational level, body mass
index, occupation and diet 24. In another study of males from the USA, cigar smoking
was associated with a moderate but significant increase in the risk of coronary heart
disease after multiple adjustment, but not with cerebrovascular disease 25. In
Chapter 4 of this thesis, I have shown that the association with smoking was not
importantly confounded by medical and behavioural risk factors. The present results
show that the association of cigarettes, pipes or cigars with CVD risk is independent
of lower socio-economic status, increased alcohol consumption and adverse medical
history of smokers.
5.4.2 Strengths and limitation
This study contains several strengths. Despite the fact that data came from studies
which differ in design, decade of enrolment, geographical region etc., study specific
RRs were comparable and the heterogeneity between studies for RRs with pipes and
cigars was non-significant. In a subsidiary analysis, the dataset was restricted to
studies providing information on all 3 types of smoking analysed and results were
largely unchanged. There was no significant difference in RRs according to region of
the world and therefore these results may apply to all developed countries. Due to
the large size and long follow-up of my data, it was possible to estimate the risk due
to pipes and cigars separately, stratifying by past smoking use of cigarettes and by
current complementary cigarette use. Analyses were adjusted for a range of potential
confounders in a systematic way for all studies, providing more reliable estimates
than those found by pooling literature based estimates in a meta-analysis. In
particular, adjustment for conventional biochemical factors (blood pressure,
cholesterol and triglycerides level, history of diabetes, alcohol status, education and
occupation) did not attenuate the associations.
Despite its strengths, my analysis contains limitations. Individuals with missing
information regarding cigar or pipe status were considered non-current cigar and
156
non-current pipe smokers, introducing measurement error. However, this is unlikely
to have led to substantial under-estimation in my estimate, which was in broad
agreement with previous estimates, especially for cigarette smoking 26-28. Pipe and
cigar smokers were mainly Men and White, and there was a difference in RRs
between White and non-White, even if it was non-significant, indicating that these
results may not be generalizable to non-White populations. Pipe and cigar amounts
were harmonized by the ERFC data management team and expressed in number of
cigarettes equivalent per day using conversion factors (see Chapter 2) which may
not be appropriate for some studies, introducing measurement error and biasing
results. Finally, there were not enough women smoking pipes or cigars to be able to
test for a difference in RRs according to gender.
5.5 Conclusion
Cigar smoking confers a moderately higher risk of CVD than never smokers. Pipe
smoking confers a risk close to that of cigarette smokers, but as amount and duration
of pipe and cigars were not available, it was not possible to find out whether the
lower risk experienced by pipe and cigar smoker in the data was a reflection of lower
intensity of this type of smoking compared to cigarettes. The resurgence of cigar
smoking in the USA over the past decades is a matter of substantial concern. Rather
than switching from cigarette to either pipe or cigar, individuals should be
encouraged to stop all forms of smoking.
157
Figure 5.1: Flow chart of study selection in the Emerging Risk Factors Collaboration Dataset
ERFC – April 2011
Smoking status dataset (See Chapter 1 Figure 1)
114 studies
929,335 individuals
14.2 average years of follow-up
Dataset 1:
-restricted to studies with information on
on cigarette smoking status
-impute missing pipe status and missing
cigar status as non current pipe and non
current cigar smoker
Studies
• 83 studies
• Subsets: 22 studies with information on pipe
only, 20 studies with information on cigar
only, 15 studies with information on pipe or
cigar on top of cigarettes smoking
Individuals
• 235,776 current cigarettes only smokers
• 4,162 cigar only smokers
• 6,255 pipe only smokers
• 5,871 pipe or cigar smokers on top of
cigarettes
Events
• 35,580 MI events during follow-up
• 13,077 MI events in cigarettes only smokers
• 356 MI events in cigar only smokers
• 712 MI events in pipe only smokers
• 678 events in individuals who were pipe or
cigar smokers on top of cigarettes
Average follow-up time of 15.6 years
Used in:
• Fig.2, Fig.6
Dataset 3: Subsidiary analysis
-restricted to studies with
information on cigarettes only, on
pipe only, on cigar only, and on
pipe or cigar smoking
• 12 studies*
• 5249 MI events during follow-up
*1 study had les than 10 MI events
recorded during follow-up and was
excluded from analyses of MI
Used in:
• Fig.3, Fig.7
Dataset 2:
-restricted to Dataset 1
-Adding current/ex pipe or cigar
smokers who were not cigarette
smokers, without information on
whether they were pipe or cigar
smokers
• 15 additional studies with information on
pipe or cigar smoking
• 2,350 current pipe or cigar users
• 202 MI events during follow-up
• Average follow-up time 12.5 years
Used in:
• Fig.4, Fig.5, Fig.8, Fig.9, Fig.10
• Table 2, Table 3
158
Figure 5.2: Risk ratios for fatal and non-fatal MI, and all cardiovascular events
combined, for current and ex cigarette, pipe and cigar smokers versus never
smokers
Study-specific risk ratios were adjusted for age at baseline and body mass index and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of that outcome. Study specific risk ratios
were pooled by random effect meta-analysis and boxes are proportional to the inverse of the
variance of the estimate, except for the reference group – never smokers – where the box is
purely illustrative.
Ex-smoker
Smoking cigarettes & pipes/cigars
Smoking cigars only
Never smokers
Smoking cigarettes only
Smoking pipes only
Smoking cigars only
Never smokers
Smoking pipes only
Smoking cigarettes only
Ex-smoker
Smoking cigarettes & pipes/cigars
16773
678
355
13055
1003
514
711
20907
9752
991
1.09 (1.05, 1.13)
2.05 (1.87, 2.24)
1.35 (1.20, 1.51)
2.11 (1.95, 2.28)
1.68 (1.56, 1.81)
1.31 (1.19, 1.44)
1.84 (1.69, 2.01)
1.97 (1.85, 2.11)
1.13 (1.09, 1.16)
1.95 (1.81, 2.10)
24.63
4.54
2.74
5.87
6.37
3.89
4.61
8.20
32.73
6.41
11,332
1 1.5 2 2.5 3
Fatal and non fatal myocardial infarction
All fatal and non fatal cardiovascular events
21,873
N failures RR (95% CI) I 2 (95% CI), %Outcome
1 (Reference)
1 (Reference)
Risk ratio (95% CI)
159
Figure 5.3: Risk ratios for cardiovascular events for current and ex cigarette, pipe
and cigar smokers versus never smokers, restricting the dataset to 12 studies (11
studies for fatal and non-fatal MI risk) which provided information on cigarettes, on
cigar and on pipe smoking status
Study-specific risk ratios were adjusted for age at baseline and body mass index and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of that outcome, leaving 11 studies for the
risk of fatal and non-fatal MI and 12 studies for the risk of all CVD events. “All CVD events”
includes all non-fatal MI recorded in individual studies as well as all CVD causes of deaths.
Study specific risk ratios were pooled by random effect meta-analysis and boxes are
proportional to the inverse of the variance of the estimate, except for the reference group –
never smokers – where the box is purely illustrative.
Never smoker
Smoking cigars only
Smoking pipes only
Smoking cigarettes only
Smoking cigars only
Ex-smoker
Smoking cigarettes only
Smoking pipes only
Smoking cigarettes & pipes/cigars
Smoking cigarettes & pipes/cigars
Never smoker
Ex-smoker
300
354
2167
222
1472
1471
287
390
530
1008
1.21 (1.06, 1.37)
1.58 (1.40, 1.79)
1.86 (1.44, 2.41)
1.28 (1.10, 1.48)
1.02 (0.88, 1.20)
2.04 (1.62, 2.58)
1.80 (1.57, 2.07)
1.79 (1.40, 2.29)
1.70 (1.32, 2.20)
1.11 (0.97, 1.26)
0 (0, 49)
0 (0, 49)
47 (10, 69)
0 (0, 50)
28 (0, 59)
10 (0, 47)
0 (0, 50)
0 (0, 50)
19 (0, 54)
0 (0, 50)
3057
1871
1 1.5 2 2.5 3
N failures RR (95% CI) I2 (95% CI), %Outcome
All cardiovascular events
Fatal and non fatal myocardial infarction
1 (Reference)
1 (Reference)
Risk ratio (95% CI)
161
Figure 5.4: Risk ratios for cardiovascular events for current and ex cigarettes, pipe
and cigar smokers, according to whether they used to smoke cigarettes
Study-specific risk ratios were adjusted for age at baseline and body mass index and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of that outcome. Study specific risk ratios
were pooled by random effect meta-analysis and boxes are proportional to the inverse of the
variance of the estimate, except for the reference group – never smokers – where the box is
purely illustrative.
Current pipe/cigars & past cigarettes smoker
Exclusive pipes/cigars smoker
Ex smoker
Exclusive pipes/cigars smoker
Current pipe/cigars & past cigarettes smoker
All cardiovascular events
Ex smoker
Current cigarettes only smoker
Fatal and non fatal myocardial infarction
Current cigarettes & pipes/cigars smoker
Current cigarettes & pipes/cigars smoker
Current cigarettes only smoker
784
302
9,752
432
574
16,773
20,871
864
588
13,032
1.57 (1.45, 1.70)
1.56 (1.38, 1.77)
1.13 (1.09, 1.17)
1.49 (1.34, 1.65)
1.71 (1.56, 1.88)
1.09 (1.05, 1.14)
1.98 (1.85, 2.12)
2.07 (1.91, 2.24)
2.19 (2.00, 2.40)
2.12 (1.97, 2.29)
0 (0, 24)
0 (0, 24)
2 (0, 22)
0 (0, 24)
0 (0, 24)
33 (15, 47)
70 (63, 75)
0 (0, 24)
0 (0, 24)
61 (52, 69)
1 1.5 2 2.5 3
N failures RR (95% CI) I2 (95% CI), %Outcome
Never smoker 11,332
Never smoker 21,873
1 (Reference)
1 (Reference)
Risk ratio (95% CI)
162
Figure 5.5: Risk ratios for subtypes of cardiovascular diseases for current and ex
cigarette and pipe or cigar smokers versus never smokers.
Study-specific risk ratios were adjusted for age at baseline and body mass index and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of that outcome. Study specific risk ratios
were pooled by random effect meta-analysis and boxes are proportional to the inverse of the
variance of the estimate, except for the reference group – never smokers – where the box is
purely illustrative. Other fatal cardiovascular events included all cardiovascular deaths not
attributable to either a myocardial or a cerebrovascular infarction.
Smoking cigarettes & pipes/cigars
Never smokers
Smoking pipe or cigar only
Ex-smokers
Smoking pipes or cigars
Never smokers
Smoking cigarettes only
Ex-smokers
Smoking cigarettes & pipes/cigars
Smoking cigarettes only
222
641
3,833
480
4,839
2,532
244
4,177
1.96 (1.69, 2.27)
1.57 (1.42, 1.74)
1.14 (1.07, 1.20)
1.71 (1.53, 1.90)
2.12 (1.98, 2.27)
1.11 (1.04, 1.17)
2.30 (1.99, 2.65)
2.17 (1.99, 2.38)
0 (0, 29)
0 (0, 29)
4 (0, 27)
0 (0, 29)
23 (0, 43)
0 (0, 29)
0 (0, 29)
43 (24, 58)
1 1.5 2 2.5 3
N failures RR (95% CI) I 2 (95% CI), %Outcome
Risk ratio (95% CI)
Fatal myocardial infarction
Non fatal myocardial infarction
4,176
Smoking pipes or cigars
Smoking cigarettes only
Never smokers
Smoking cigarettes only
Ex-smokers
Ex-smokers
Smoking cigarettes & pipes/cigars
Smoking pipes or cigars
Smoking cigarettes & pipes/cigars
Never smokers
316
3,554
4,194
2,737
4,250
142
275
133
1.28 (1.13, 1.46)
2.15 (1.99, 2.33)
1.71 (1.60, 1.84)
1.12 (1.06, 1.18)
1.04 (0.99, 1.09)
2.27 (1.89, 2.73)
1.75 (1.52, 2.02)
1.95 (1.61, 2.34)
0 (0, 26)
23 (0, 43)
19 (0, 39)
0 (0, 28)
10 (0, 31)
0 (0, 28)
0 (0, 28)
0 (0, 26)
Other fatal cardiovascular events
Fatal and non fatal cerebrovascular events
3,894
6,592
3,169 1 (Reference)
1 (Reference)
1 (Reference)
1 (Reference)
163
Figure 5.6: Risk ratio for fatal lung cancer for current and ex cigarette, pipe and cigar
smokers versus never smokers
Never smoker
Smoking pipes only
Smoking cigarettes & cigars/pipes
Ex-smokers
Smoking cigarettes only
Smoking cigars only
163
195
1701
5833
72
10.05 (8.15, 12.38)
13.90 (11.44, 16.89)
3.40 (2.96, 3.90)
13.13 (11.72, 14.71)
6.54 (5.00, 8.55)
0 (0, 31)
0 (0, 31)
23 (0, 45)
6 (0, 31)
0 (0, 31)
1 2 4 6 8 10 12 16
Smoking status N failures RR (95% CI) I 2 (95% CI), %
Risk ratio (95% CI)
778 1 (Reference)
Study-specific risk ratios were adjusted for age at baseline and body mass index and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of that outcome. Study specific risk ratios
were pooled by random effect meta-analysis and boxes are proportional to the inverse of the
variance of the estimate, except for the reference group – never smokers – where the box is
purely illustrative.
164
Figure 5.7: Risk ratio for fatal lung cancer for current and ex cigarette, pipe and cigar
smokers versus never smokers restricting the dataset to studies which provided
information on cigarettes, cigar and pipe smoking status
Smoking cigars only
Smoking cigarettes & pipes/cigars
Ex-smoker
Never smokers
Smoking cigarettes only
Smoking pipes only
27
82
117
608
54
6.57 (3.91, 11.04)
18.84 (12.62, 28.12)
4.83 (3.25, 7.18)
20.61 (14.73, 28.83)
14.52 (9.19, 22.95)
0 (0, 85)
0 (0, 85)
0 (0, 85)
0 (0, 85)
0 (0, 85)
1 2 4 6 8 1012 16
Risk ratio (95% CI)
43
Smoking status N failures RR (95% CI) I2 (95% CI), %
1 (Reference)
Study-specific risk ratios were adjusted for age at baseline and body mass index and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of that outcome, leaving 4 studies for the
analysis of the risk of lung cancer deaths. Study specific risk ratios were pooled by random
effect meta-analysis and boxes are proportional to the inverse of the variance of the estimate,
except for the reference group – never smokers – where the box is purely illustrative.
165
Figure 5.8: Risk ratio for fatal lung cancer for current and ex cigarettes, pipe and
cigar smokers according to whether they used to smoke cigarettes
Ex-smoker
Current pipe/cigars & past cigarettes smokers
Exclusive pipes/cigars smokers
Current cigarettes only smokers
Current cigarettes & pipes/cigars smokers
Never smoker
1701
138
39
5816
194
3.33 (2.89, 3.83)
9.77 (7.79, 12.24)
5.65 (3.99, 8.01)
12.79 (11.35, 14.41)
15.87 (12.99, 19.40)
23 (0, 45)
0 (0, 31)
0 (0, 31)
6 (0, 31)
0 (0, 31)
1 2 4 6 8 1012 16
Smoking status N failures RR (95% CI) I2 (95% CI), %
Risk of fatal lung cancer
778 1 (Reference)
Study-specific risk ratios were adjusted for age at baseline and body mass index and
stratified, where appropriate, by sex and trial arm. Studies with fewer than 10 events for a
specific outcome were excluded from the analysis of that outcome. Study specific risk ratios
were pooled by random effect meta-analysis and boxes are proportional to the inverse of the
variance of the estimate, except for the reference group – never smokers – where the box is
purely illustrative.
166
Figure 5.9 a-b: Subgroup analysis
a) Risk ratios for fatal and non-fatal myocardial infarction for current smokers
compared to never or ex-smokers
By age groups (36,085 events in 102 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and pipe/
cigar) smoker
History of diabetes (32,492 events in 92 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and pipe/
cigar) smoker
Alcohol status (16,708 events in 76 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and pipe
/cigar) smoker
Race (23,728 events in 79 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and pipe
/cigar) smoker
Geographical region (36,085 events in 102 cohorts)
Primary Pipe
/cigar smoker
Cigarette (and pipe
/cigar) smoker
20-39
40-59
60-69
70+
20-39
40-59
60-69
70+
Other
Definite diabetic
Other
Definite diabetic
Other
Current
Other
Current
White
Non-white
White
Non-white
Western Europe
North America
Other
Western Europe
North America
Other
2.15 (0.32, 14.39)
1.03 (0.29, 3.69)
1.00 (0.32, 3.11)
1.25 (0.17, 9.00)
3.28 (2.87, 3.74)
2.15 (2.06, 2.24)
1.77 (1.64, 1.91)
1.12 (0.66, 1.90)
1.57 (1.45, 1.70)
1.76 (1.30, 2.39)
2.06 (1.91, 2.21)
1.61 (1.46, 1.79)
1.45 (1.12, 1.89)
1.41 (1.24, 1.60)
2.25 (2.02, 2.50)
1.97 (1.81, 2.14)
1.49 (1.31, 1.70)
1.42 (0.78, 2.58)
1.93 (1.77, 2.10)
1.83 (1.56, 2.14)
1.54 (1.36, 1.73)
1.18 (0.85, 1.64)
1.77 (1.13, 2.75)
2.00 (1.59, 2.50)
1.89 (1.25, 2.86)
1.66 (0.86, 3.21)
<0.001
<0.001
0.014
0.747
0.283
1 1.5 2 2.5 4
Risk ratio (95% CI)
Smoking status CategoriesSubgroups RR (95% CI) p-value
Interaction
Study-specific adjusted risk ratios were stratified, where appropriate, by sex and trial arm and
then pooled across studies by random effects meta-analysis. Studies with fewer than 5 events
for a specific outcome were excluded from the analysis of that outcome. P-value of interaction
comes from a formal t-test of interaction using age as a continuous variable and a Fisher test
of significance of all interaction categories for the other categorical variables. A nominal p-
value <0.001 was chosen as indicating presence of a significant effect modifier. For example,
in this figure, RRs for current smokers significantly decreased with age and were significantly
different according to diabetes status.
167
b) Risk of all cardiovascular events for current smokers compared to never or ex-
smokers
By age groups (62,357 events in 107 cohort)
Primary Pipe/
cigar smoker
Cigarette (and
pipe/cigar) smoker
By history of diabetes (55,957 events in 97 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and
pipe/cigar) smoker
By alcohol status (29,961 events in 80 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and
pipe/cigar) smoker
By race (39,010 events in 84 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and
pipe/cigar) smoker
By geographical region (62,357 events in 107 cohorts)
Primary Pipe/
cigar smoker
Cigarette (and
pipe/cigar) smoker
Smoking status
20-39
40-59
60-69
70+
20-39
40-59
60-69
70+
Other
Definite diabetic
Other
Definite diabetic
Other
Current
Other
Current
White
Non-white
White
Non-white
Western Europe
North America
Other
Western Europe
North America
Other
CategoriesSubgroups
3.52 (0.72, 17.34)
1.19 (0.44, 3.23)
1.00 (0.43, 2.34)
2.01 (0.39, 10.39)
2.71 (2.44, 3.00)
2.05 (1.98, 2.12)
1.80 (1.69, 1.92)
1.10 (0.74, 1.65)
1.46 (1.37, 1.56)
1.45 (1.12, 1.87)
1.93 (1.81, 2.05)
1.51 (1.39, 1.65)
1.27 (1.03, 1.57)
1.44 (1.30, 1.60)
1.96 (1.81, 2.13)
1.86 (1.71, 2.02)
1.41 (1.30, 1.53)
0.91 (0.58, 1.42)
1.84 (1.71, 1.98)
1.60 (1.43, 1.79)
1.48 (1.34, 1.64)
1.15 (0.88, 1.50)
1.37 (0.94, 1.98)
1.95 (1.62, 2.35)
1.80 (1.28, 2.53)
1.45 (0.85, 2.49)
RR (95% CI)
<0.001
<0.001
0.146
0.007
0.280
p-value
Interaction
1 1.5 2 2.5 4
Risk ratio (95% CI)
Study-specific adjusted risk ratios were stratified, where appropriate, by sex and trial arm and
then pooled across studies by random effects meta-analysis. Studies with fewer than 5 events
for a specific outcome were excluded from the analysis of that outcome. P-value of interaction
comes from a formal t-test of interaction using age as a continuous variable and a Fisher test of
significance of all interaction categories for the other categorical variables. A nominal p-value
<0.001 was chosen as indicating presence of a significant effect modifier.
168
Figure 5.10: Dose-response relationship of pipe and cigar smoking with risk of MI
and CVD
1.0
1.2
1.4
1.6
1.8
2.0
0/Never smokers 10 20 30
b) All fatal and non fatal cardiovascular events
(4,787 events)
R
is
k
ra
tio
(9
5%
C
I)
Mean smoking amount pipes & cigars amount
(in cigarettes equivalent/day)
1.0
1.2
1.4
1.6
1.8
2.0
0 10 20 30
a) Fatal and non fatal MI
(3,052 events)
R
is
k
ra
tio
(9
5%
C
I)
Mean Smoking amount: pipes & cigars (cigarettes/day)
Pipe and cigar smoking amount were converted into number of cigarettes equivalent per day
(see Chapter 1) and categorized as “<5 cigarette equivalent per day”, “5-10 cigarettes
equivalent per day”, “10-15 cigarettes equivalent per day” and “>15 cigarettes equivalent per
day” amongst current pipe or cigar smokers who were not also current cigarette smokers.
Never smokers (of neither cigarettes nor pies no cigars) were chosen as the reference group.
Study-specific risk ratios adjusted for baseline age and Body Mass Index were stratified,
where appropriate, by sex and trial arm and then pooled across studies by random effects
meta-analysis. Studies with fewer than 5 events were not included in this graph.
169
Table 5.1: Cross-sectional correlates of pipe and cigar smoking
a) Demographic and lifestyle variables
Percentage or mean (SD) according to smoking type
No of
studies*
No of
subjects
current
smokers*
Current
cigarette
only
smokers
Current
pipe only
smokers
Current
cigar only
smokers
Age (years) 85 247,002 52.89(8.27)
53.83
(5.95)
51.39
(8.00)
Sex 85 247,002
Male 144,494 56% 99% 97%
Female 102,508 44% 1% 7%
Race 80 228,045
White 133,541 88% 99% 97%
Non white 15,636 12% 1% 3%
Alcohol status 64 126,520
Not current 42,819 35% 18% 11%
Current 83,701 65% 82% 89%
History of diabetes 78 232,102
No 226,526 98% 98% 98%
Yes 5,576 2% 2% 2%
Level of education reached 42 97,354
No schooling 3,570 4% 1% 1%
Primary 17,918 18% 35% 21%
Secondary 51,906 54% 43% 49%
Vocational/University 61,230 25% 21% 29%
Occupation or job 40 612,230
Not working 8,900 15.84 2.36 1.93
Manual 18,620 29.66 33.94 36.92
Office 6,012 9.39 14.93 8.77
Service 17,527 28.2 27.6 40.68
Student 77 0.13 0.1 0.05
Other 10,094 16.78 21.07 11.64
SD: standard deviation. Mean and standard deviation were calculated within studies and
pooled across studies using random effects meta-analyses. *With information on cigarettes,
cigars, pipe or pipe and cigar smoking.
170
b) Medical and biochemical variables
Mean (SD) by smoking type
No of
studies*
No of
subjects
current
smokers*
Current
cigarette only
smokers
Current pipe
only smokers
Current cigar
only smokers
Anthropometry
Body Mass Index (kg/m2) 85 247,002 25.28 (3.77) 25.57 (3.06) 26.15 (3.51)
Waist to Hip Ratio 35 34,490 0.89 (0.08) 0.95 (0.06) 0.95 (0.09)
Blood pressure
Systolic blood pressure 82 200,387 132.54 (17.99) 136.36 (18.1) 135.12 (18.09)
Diastolic blood pressure 82 200,549 79.73 (10.98) 83.72 (11.32) 83.72 (11.75)
Lipid markers
Total cholesterol (mmoll/l) 82 198,587 5.80 (1.14) 5.88 (1.14) 6.03 (1.12)
Non HDL-C (mmol/l) 68 106,894 4.48 (1.17) 4.57 (1.15) 4.74 (1.18)
HDL-C (mmol/l) 68 107,010 1.29 (0.37) 1.25 (0.35) 1.25 (0.39)
Loge triglyceride (mmol/l) 67 151,950 0.37 (0.52) 0.48 (0.49) 0.49 (0.52)
Apo A1 (g/l) 21 24,813 1.41 (0.27) 1.49 (0.29) 1.46 (0.25)
Apo B (g/l) 22 25,295 1.12 (0.31) 1.05 (0.26) 1.14 (0.3)
Loge Lp(a) 20 18,349 2.20 (1.18) 2.45 (1.28) 2.48 (1.23)
Inflammatory markers
Loge CRP (mg/l) 34 18,870 0.79 (1.12) 0.75 (1.14) 1.07 (1.12)
 Fibrinogen (μmol/l) 32 42,822 9.63 (2.20) 8.64 (2.11) 8.63 (2.08)
Albumin (g/l) 28 32,989 43.14 (3.68) 44.04 (3.51) 44.15 (3.68)
Loge leukocyte
count(x10^9/l) 25 34,496 1.98 (0.27) 1.87 (0.24) 1.88 (0.26)
Loge Interleukin 6 (ng/l) 9 3,646 0.57 (0.62) 0.36 (0.63) 0.21 (0.71)
SD: standard deviation. Mean and standard deviation were calculated within studies and
pooled across studies using random effects meta-analyses. *With information on the type of
tobacco used.
171
Table 5.2: Progressive adjustment of RRs of current pipe and cigar smoking compared to never smokers
a) Risk ratios for fatal and non-fatal MI
Current pipe or cigars only
smokers Current cigarettes only
Current cigarettes &
cigars/pipes
Progressive adjustment No of
events RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
Age, sex and BMI 240 1.58 (1.33, 1.87) 0 (0, 31) 3882 2.05 (1.84, 2.29) 40 (18, 56) 55 2.16 (1.62, 2.88) 0 (0, 31)
Plus systolic blood pressure 240 1.59 (1.33, 1.89) 0 (0, 31) 3882 2.09 (1.87, 2.34) 39 (17, 56) 55 2.20 (1.65, 2.93) 0 (0, 31)
Plus history of diabetes 240 1.51 (1.26, 1.80) 0 (0, 31) 3882 2.10 (1.91, 2.32) 36 (12, 54) 55 2.21 (1.65, 2.95) 0 (0, 31)
Plus non HDL-cholesterol,
HDL-cholesterol and loge triglycerides
240 1.43 (1.20, 1.71) 0 (0, 31) 3882 1.99 (1.83, 2.16) 21 (0, 44) 55 2.12 (1.58, 2.83) 0 (0, 31)
Additional adjustment
Age, sex and BMI 373 1.43 (1.27, 1.61) 0 (0, 28) 6036 2.09 (1.93, 2.25) 46 (30, 59) 129 1.99 (1.65, 2.39) 0 (0, 28)
Plus alcohol consumption 373 1.44 (1.28, 1.62) 0 (0, 28) 6036 2.14 (1.98, 2.31) 47 (30, 59) 129 2.02 (1.67, 2.43) 0 (0, 28)
Age, sex and BMI 659 1.51 (1.33 1.70) 0 (0 52) 1439 2.00 (1.83 2.18) 0 (0 52) 227 1.98 (1.71 2.30) 0 (0 52)
Plus occupation 659 1.67 (0.45 6.24) 0 (0 52) 1439 2.01 (1.84 2.21) 0 (0 52) 227 2.02 (0.37 11.15) 0 (0 52)
Plus education attainment 659 1.70 (0.46 6.28) 0 (0 52) 1439 1.98 (1.77 2.22) 0 (0 52) 227 2.04 (0.37 11.19) 0 (0 52)
Study-specific adjusted risk ratios were stratified, where appropriate, by sex and trial arm and then pooled across studies by random effects meta-analysis.
Studies with fewer than 10 events for a specific outcome were excluded from the analysis of that outcome. The dataset was restricted to studies with baseline
information on age, sex, BMI and either 1) systolic blood pressure, history of diabetes, non HDL-cholesterol, HDL-cholesterol and loge triglycerides; 2) alcohol
consumption, or 3) occupation and education attainment. Restricting the data to individuals with baseline information on all additional covariates of adjustment
at the same time (rather than in 3 subsets) would have resulted in too small a sample.
172
b) Risk ratios for all cardiovascular events
Current pipe or cigars
only smokers Current cigarettes only
Current cigarettes
& cigars/pipes
Progressive adjustment No ofevents RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
No of
events RR (95% CI)
I2 (95%CI)
%
Age, sex and BMI 340 1.36 (1.18, 1.56) 0 (0, 29) 6115 1.89 (1.71, 2.09) 49 (33, 62) 84 2.23 (1.76, 2.82) 0 (0, 29)
Plus systolic blood pressure 340 1.38 (1.20, 1.59) 0 (0, 29) 6115 1.91 (1.73, 2.10) 47 (29, 60) 84 2.25 (1.78, 2.85) 0 (0, 29)
Plus history of diabetes 340 1.34 (1.17, 1.54) 0 (0, 29) 6115 1.94 (1.77, 2.12) 45 (26, 59) 84 2.25 (1.78, 2.84) 0 (0, 29)
Plus non HDL-cholesterol,
HDL-cholesterol and loge triglycerides
340 1.32 (1.16, 1.51) 0 (0, 29) 6115 1.90 (1.76, 2.05) 33 (9, 51) 84 2.19 (1.73, 2.76) 0 (0, 29)
Additional adjustment
Age, sex and BMI 562 1.44 (1.30, 1.59) 0 (0, 27) 9322 1.98 (1.84, 2.12) 62 (51, 70) 202 2.00 (1.72, 2.33) 0 (0, 27)
Plus alcohol consumption 562 1.44 (1.30, 1.59) 0 (0, 27) 9322 2.00 (1.86, 2.15) 62 (52, 70) 202 2.05 (1.76, 2.38) 0 (0, 27)
Age, sex and BMI 867 1.37 (1.19 1.59) 0 (0 50) 2367 1.90 (1.78 2.04) 0 (0 50) 303 1.92 (1.69 2.18) 0 (0 50)
Plus occupation 867 1.28 (0.47 3.47) 0 (0 50) 2367 1.82 (1.66 1.98) 0 (0 50) 303 2.41 (0.67 8.62) 0 (0 50)
Plus education attainment 867 1.33 (0.47 3.76) 0 (0 50) 2367 1.79 (1.62 1.97) 0 (0 50) 303 2.17 (0.62 7.61) 0 (0 50)
Study-specific adjusted risk ratios were stratified, where appropriate, by sex and trial arm and then pooled across studies by random effects meta-analysis.
Studies with fewer than 10 events for a specific outcome were excluded from the analysis of that outcome. The dataset was restricted to studies with baseline
information on age, sex, BMI and either 1) systolic blood pressure, history of diabetes, non HDL-cholesterol, HDL-cholesterol and loge triglycerides; 2) alcohol
consumption, or 3) occupation and education attainment. Restricting the data to individuals with baseline information on all additional covariates of adjustment
at the same time (rather than in 3 subsets) would have resulted in too small a sample.
173
References
1. Office for National Statistics. Smoking and drinking among adults, 2007. General
Household Survey. 2007.
2. Delnevo CD. Smokers' choice: what explains the steady growth of cigar use in the
U.S.? Public Health Rep. 2006;121:116-119.
3. Jacobs EJ, Thun MJ, Apicella LF. Cigar smoking and death from coronary heart
disease in a prospective study of US men. Arch Intern Med. 1999;159:2413-2418.
4. National Institute on drug abuse. NIDA InfoFacts: Cigarettes and other tobacco
products. [http://www.nida.nih.gov/infofacts/tobacco.html#References], Accessed on
01/06/2010.
5. Baker F, Ainsworth SR, Dye JT et al. Health risks associated with cigar smoking. Stat
Med. 2002;21:1539-1558.
6. O'Connor RJ, McNeill A, Borland R et al. Smokers' beliefs about the relative safety of
other tobacco products: findings from the ITC collaboration. Nicotine Tob Res.
2007;9:1033-1042.
7. Doll R, HILL AB. The mortality of doctors in relation to their smoking habits: a
preliminary report. BMJ. 1954;328:1529-1533.
8. Iribarren C, Tekawa IS, Sidney S, Friedman GD. Effect of cigar smoking on the risk of
cardiovascular disease, chronic obstructive pulmonary disease, and cancer in men. N
Engl J Med. 1999;340:1773-1780.
9. Jacobs EJ, Thun MJ, Apicella LF. Cigar smoking and death from coronary heart
disease in a prospective study of US men. Arch Intern Med. 1999;159:2413-2418.
10. Shaper AG, Wannamethee SG, Walker M. Pipe and cigar smoking and major
cardiovascular events, cancer incidence and all-cause mortality in middle-aged British
men. Int J Epidemiol. 2003;32:802-808.
11. Tverdal A, Bjartveit K. Health consequences of pipe versus cigarette smoking. Tob
Control. 2011;20:123-130.
12. Baker F, Ainsworth SR, Dye JT et al. Health risks associated with cigar smoking. Stat
Med. 2002;21:1539-1558.
13. Easton DF, Peto J, Babiker AG. Floating absolute risk: an alternative to relative risk in
survival and case-control analysis avoiding an arbitrary reference group. Stat Med.
1991;10:1025-1035.
14. Jacobs EJ, Thun MJ, Apicella LF. Cigar smoking and death from coronary heart
disease in a prospective study of US men. Arch Intern Med. 1999;159:2413-2418.
174
15. Iribarren C, Tekawa IS, Sidney S, Friedman GD. Effect of cigar smoking on the risk of
cardiovascular disease, chronic obstructive pulmonary disease, and cancer in men. N
Engl J Med. 1999;340:1773-1780.
16. Shaper AG, Wannamethee SG, Walker M. Pipe and cigar smoking and major
cardiovascular events, cancer incidence and all-cause mortality in middle-aged British
men. Int J Epidemiol. 2003;32:802-808.
17. Pechacek TF, Folsom AR, de GR et al. Smoke exposure in pipe and cigar smokers.
Serum thiocyanate measures. JAMA. 1985;20;254:3330-3332.
18. Brunnemann KD, Hoffmann D. Chemical studies on tobacco smoke. XXIV. A
quantitative method for carbon monoxide and carbon dioxide in cigarette and cigar
smoke. J Chromatogr Sci. 1974;12:70-75.
19. Iribarren C, Tekawa IS, Sidney S, Friedman GD. Effect of cigar smoking on the risk of
cardiovascular disease, chronic obstructive pulmonary disease, and cancer in men. N
Engl J Med. 1999;340:1773-1780.
20. U.S.Department of Health and Human Services. How tobacco smoke causes disease:
the biology and behavioral basis of smoking-attributable disease: a report of the
surgeon general. 2010.
21. Doll R, Peto R. Mortality in relation to smoking: 20 years' observations on male British
doctors. Br Med J. 1976;2:1525-1536.
22. Baker F, Ainsworth SR, Dye JT et al. Health risks associated with cigar smoking. Stat
Med. 2002;21:1539-1558.
23. Tverdal A, Bjartveit K. Health consequences of pipe versus cigarette smoking. Tob
Control. 2011;20:123-130.
24. Henley SJ, Thun MJ, Chao A, Calle EE. Association between exclusive pipe smoking
and mortality from cancer and other diseases. J Natl Cancer Inst. 2004;96:853-861.
25. Iribarren C, Tekawa IS, Sidney S, Friedman GD. Effect of cigar smoking on the risk of
cardiovascular disease, chronic obstructive pulmonary disease, and cancer in men. N
Engl J Med. 1999;340:1773-1780.
26. Henley SJ, Thun MJ, Chao A, Calle EE. Association between exclusive pipe smoking
and mortality from cancer and other diseases. J Natl Cancer Inst. 2004;96:853-861.
27. Baker F, Ainsworth SR, Dye JT et al. Health risks associated with cigar smoking. Stat
Med. 2002;21:1539-1558.
28. Thun MJ, Apicella LF, Henley SJ. Smoking vs other risk factors as the cause of
smoking-attributable deaths: confounding in the courtroom. JAMA. 2000;284:706-712.
175
Section B: Smoking and smokeless tobacco in relation to the
risk of cardiovascular diseases in a developing country
Chapter 6: Description of the Pakistan Risk of Myocardial
Infarction Study
Summary
The Pakistan Risk of Myocardial Infarction Study (PROMIS) is an ongoing case-
control study collecting genetic, lifestyle and other determinants of myocardial
infarction in South Asia. In particular, detailed questionnaire based information is
available for smoking and smokeless forms of tobacco use. By March 2011, the
PROMIS had enrolled 7,905 first ever MI patients and 7,458 age and sex frequency
matched controls. After exclusion of individuals with missing information on
covariates of adjustment, 6,871 controls and 6,050 cases were included in the
analysis. Participants were recruited from six different urban hospitals located across
Pakistan and sampled from all major Pakistani ethnic groups. PROMIS will provide a
unique insight into the effect of smoking and smokeless tobacco on the risk of
myocardial infarction in a developing country setting, bearing in mind the limitations
inherent to the case-control study design.
176
6.1 Background of PROMIS
Over the next decade, the burden of CVD is projected to increase the most rapidly in
South Asia. In a country like Pakistan, it threatens to counter major health gains
achieved during the past half century, translated into the prolongation of life
expectancy from 50 to 67 years between 1960 and 2010 1. CHD is currently
responsible for 14% of all deaths in men and 12% in women in Pakistan (Figure 6.1),
accounting for 150,000 deaths a year, representing a substantial burden for a middle
low-income country of 185 million people 2. In addition, several million people of
Pakistani or other South Asian descent who live in developed countries tend to have
striking excesses of vascular diseases 3 and vascular disease mortality 4,5 compared
with natives of European ancestry. Despite the highlighted need for large scale
epidemiological studies 6, previous studies on South Asians have been characterized
by small sample sizes 7,8, conducted amongst South Asians living abroad 9 and have
not taken into account ethnicity-specific differences within the South Asian
subcontinent 10-12.
In this context, the objective of the PROMIS has been to establish an epidemiological
and genetic resource to enable reliable study of a range of determinants of MI in
South Asia, including genetic, lifestyle and other determinants 13
(http://www.cncdpk.com/projects/the-pakistan-risk-of-myocardial-infarction-study-
promis.html). The PROMIS has established 6 centres of recruitment in five urban
centres of Pakistan: Karachi, Hyderabad, Lahore, Multan and Faisalabad (Figure
6.2). Recruitment has been ongoing since 2005 and aims to achieve a total of 20,000
MI cases and 20,000 controls, at which stage it will represent the largest scale
epidemiological resource for the study of CHD in South Asia. This will enable more
detailed assessment than previously possible of the relationship between risk factors
and MI risk in Pakistan. For some lifestyle habits such as tobacco uses, which are
relatively similar across the South Asian subcontinent, findings from PROMIS may be
generalizable to other South Asian populations. It may also be informative for
Western populations with a South Asian origin who have retained some lifestyle
characteristics and a similar genetic background to Pakistanis. Comparing findings of
the PROMIS with those of studies conducted with Pakistani migrants settled in
Western countries could also shed light on the relative importance of genetic versus
behavioural and socio-economic risk factors, assuming Pakistani migrants are similar
genetically to Pakistanis and only differ in their acquisition of Western lifestyle habits.
177
As recruitment centres are located across Pakistan, major ethnic groups are all
represented within the PROMIS (Table 6.1). In Pakistan, major ethnic groups are
Punjabi (45% of population), Pathan (15%), Sindhi (14%), Sariaki (8%), Muhajirs
(also called Urdu) (8%), Balochi (4%) and other (6%) 14. Punjabis originate from the
Punjab region. Pathans are also called Pashtuns and are an indigenous group from
the land located south of the Hindu Kush in Afghanistan and West of the Indus River
in Pakistan. Sindhis are ancient people principally inhabiting the province of Sindh in
Pakistan. Baluch originate from the Baluchistan province. Muhajirs, also referred as
Urdu because of the language they speak, are a multi-ethnic group of Muslims who
emigrated after the partition of Pakistan from India. The Saraikis maintain a separate
language and culture, but are ethnically the same as the natives of Sindh and
Punjab.
6.2 Methods
6.2.1 Inclusion criteria
Patients were eligible for inclusion as MI cases if they were between 20 and 80 years
old and admitted to emergency rooms of participating hospitals fulfilling all the
following criteria: (1) sustained symptoms suggestive of MI lasting longer than 20
minutes within the previous 24 hours; (2) ECG changes indicative of MI (i.e., new
pathologic Q waves, at least 1mm ST elevation in any 2 or more contiguous limb
leads, new left bundle branch block, new persistent ST-T wave changes diagnostic of
non-Q wave MI); (3) positive troponin test; (4) no previous cardiovascular disease,
defined as self-reported history of angina, MI, coronary revascularisation, transient
ischaemic attack, stroke, or evidence of CHD on prior ECG or in other medical
records; and (5) absence of cardiogenic shock.
Controls were individuals identified concurrently in the same hospitals as index cases
frequency matched to cases by sex and by 5 year age bands, without a self-reported
history of CVD and no ECG changes consistent with a previous MI. Controls were
recruited in the following order of priority: (1) visitors of patients attending the
outpatient department; (2) patients attending outpatient departments for routine
non-cardiac complaints; (3) non-blood related visitors of index MI cases.
Participants were not recruited if they presented with (1) a history of viral or bacterial
infection in the past two weeks; (2) a documented chronic condition such as
malignancy, inflammatory disorder, hepatitis or renal failure; (3) pregnancy; (4)
refused to participate. Institutional review boards provided approval and participants
178
gave informed consent for use of samples in genetic, biochemical and other
analyses.
6.2.2 Measurement of lifestyle variables
Each PROMIS participant was administrated a questionnaire with detailed
behavioural and lifestyle information questions, including smoking, diet, socio-
economic status, ethnic group, medical and family history, physical activity and
anthropometry (List 6.1). The questionnaire was locally relevant, recording local
linguistic and ethnic groups (Urdu, Punjabi, Pathan, Balochi, Sindhi, Memon, Gujrati
and others) and regional types of tobacco use (cigarettes, bidis, huqqa, chilum,
naswar, paan, supari, gutka, chalia) (List 6.2). Individuals using smokeless products
without adding tobacco (“non tambako” users) were considered as non users. For the
purpose of the analysis, paan, gutka and supari users were grouped under the term
“chew tobacco“ and individuals using cigarettes, bidis, hookah or chilum as “smoke
tobacco”. The use of naswar was termed “snuff dipping”. “Smokeless tobacco”
includes both chewing and dipping tobacco. To assess dietary habits, participants
completed a food-frequency questionnaire tailored to Pakistani diet, which had been
developed with a local nutritionist after a pilot 24h recall study on a subset of 100
participants. For MI cases, questionnaires were administered after medical
stabilization and related to habits and characteristics before the diagnosis of acute
MI.
6.2.3 Measurement of biochemical information, weight and height
Using standardized procedures and equipment, research officers obtained
measurements of height, weight, waist and hip circumference, systolic and diastolic
blood pressure, and heart rate. Non-fasting serum, EDTA plasma, and whole blood
samples were obtained while participants were recumbent at about 450, centrifuged
within 45 minutes, and stored long-term at -80°C. In MI cases, blood samples were
drawn within 24 hours of the onset of symptoms and prior to administration of any
thrombolytic agents. Time since onset of symptoms and since last meal was
recorded. Total cholesterol and high-density lipoprotein cholesterol were measured
using a homogeneous enzymatic colorimetric method. Low-density lipoprotein
cholesterol was measured using the same techniques or computed using the
Friedewald formula.
179
6.3 The dataset
6.3.1 Descriptive characteristics
The PROMIS had recruited 7905 MI patients and 7458 controls by May 2011 from 6
different hospitals located in 5 urban centres of Pakistan: Karachi, Hyderabad,
Lahore, Multan and Faisalabad. One centre, Faisalabad, had only started recruiting
when this analysis was performed and so 11 cases and 24 controls from this centre
were excluded from the analyses. For the purpose of the analysis, individuals with
missing information on age, sex, recruitment centre, history of diabetes and
hypertension, family history of MI, LDL-c and waist to hip ratio, were excluded,
leaving 6,051 cases and 6,871 controls. All participants were recruited from large
urban centres and the majority of participants came from Karachi or Lahore. All
individuals reporting the use of hookah or chilum were also smokers of either
cigarettes or bidis.
Cases were mostly men, accounting for 83% of all cases (Table 6.2). Mean age was
53.4 years old (SD: 10.3). The proportion of self-reported diabetics was 22%, self-
reported hypertensive individuals: 49%, and individuals with a family history of MI:
21%. Average total cholesterol levels were: 5.07 (SD: 1.37), and average levels of
LDL-C levels were: 3.3 (SD: 1.18). Waist to hip ratio had a mean of 0.959 (SD:
0.056). Average number of years of education was 8 years (IQR: 0-12) and average
monthly income was 15,000 Pakistani rupees (IQR: 10,000-25,000). 27% of cases
declared cooking with ghee only, 29% with a combination of ghee and oil, and the
rest with oil only. The proportion of current tobacco users in cases was 18% of
women and 59% of men; and the proportion of ex-tobacco users was 6%.
Frequency matching by sex was imperfect and men represented 78% of controls.
Mean age of controls was 53.0 years old (SD: 9.5). Overall, 60% of controls had
never used tobacco, 7% were ex-tobacco users and 32% current tobacco users. The
main centre of recruitment was the National Institute of Cardiovascular Diseases
located in Karachi, where 48% of controls and 39% of cases came from. As Karachi
is located in an area predominantly populated by the Urdu ethnic-linguistic group, this
ethnicity was the most represented in my data (39% of controls) (Table 6.1). Out of
all controls, 14% declared having diabetes (975 individuals), 28% being hypertensive
(1901 individuals), and 15% had a family history of MI (1045 individuals). Average
total cholesterol levels were: 4.64 (SD: 1.37), and average levels of LDL-C levels:
2.87 (SD: 1.06). Waist to hip ratio had a mean of 0.942 (SD: 0.066). Average number
180
of years of education was 8 years (IQR: 0-12) and average monthly income was
12,000 Pakistani rupees (IQR: 8000-22,000). 23% of controls declared cooking with
ghee only, 27% with a combination of ghee and oil, and the rest with oil only. The
proportion of current tobacco users was 13% of women and 38% men, and the
proportion of ex-tobacco users was 7% of controls.
6.3.2 Principal component analysis of diet and socio-economic status
The principal component analysis on education, monthly income, ownership, and
employment status identified one main component labelled “socio-economic status”
and explained 14% of the variance (Appendix 2). Two dietary patterns were
identified by a principal component analysis of the food frequency questionnaire and
explained nearly 15% of the variance (Appendix 3). The first pattern was labelled
“vegetables and carbohydrate” based diet and the second pattern was labelled as
“meat, fish and sweet” based diet.
6.4 Use of the dataset in later Chapters
The analysis is restricted to controls in Chapter 7 for the investigation of the
prevalence of tobacco uses and its relationship to other CVD risk factors in
Pakistanis. In Chapter 8, cases and controls are included in the analysis which aims
to quantify the risk of tobacco users compared to non-tobacco users, taking into
account other relevant risk factors for CVD.
181
Figure 6.1: Estimated proportional mortality (%) in Pakistan, 2004
Source: Global Burden of disease 16
182
Figure 6.2: Geographical location of PROMIS recruitment centres
Recruitment centreCity Predominant
ethnicity
Punjab Institute of CardiologyLahore Punjabi
Karachi Urdu
Urdu
Multan Institute of CardiologyMultan Punjabi
Red Crescent Hospital
National Institute of CVD
Karachi Institute of Heart Diseases
Hyderabad Urdu
Faisalabad Institute of CardiologyFaisalabad Punjabi
Deewan Mushtaq Institute of CardiologyHyderabad Punjabi
Map source: U.S. Central Intelligence Agency: The World Fact book- Pakistan – accessed on
14/07/2011 15
183
Table 6.1: Number of a) cases and b) controls from the PROMIS per recruitment
centre
a) Cases
Ethnicity
Recruitment centre Urdu Punjabi Pathan Balochi Sindhi Memon Gujrati Others Total
DMIC 36 6 0 2 80 0 0 3 127
KIHD 779 99 35 12 26 19 16 39 1,025
MIC 112 188 4 2 2 0 0 290 598
NICVD 1,142 419 239 59 199 92 32 197 2,379
PIC 104 1,217 21 1 1 0 1 44 1,389
RCH 218 25 12 0 264 2 3 8 532
Total number of cases 2,391 1,954 311 76 572 113 52 581 6,050
b) Controls
Ethnicity
Recruitment centre Urdu Punjabi Pathan Balochi Sindhi Memon Gujrati Others Total
DMIC 86 3 2 0 66 0 3 1 161
KIHD 905 104 54 9 32 8 15 38 1,165
MIC 74 148 6 2 0 0 0 185 415
NICVD 1,280 582 417 114 519 67 101 246 3,326
PIC 55 1,138 18 1 6 0 0 44 1,262
RCH 274 34 8 2 208 2 0 14 542
Total number of controls 2,674 2,009 505 128 831 77 119 528 6,871
DMIC: Deewan Mushtaq Institute of Cardiology, KIHD: Karachi Institute of Heart Diseases,
MIC; Multan Institute of Heart Diseases, NICVD: National Institute of Cardiovascular
Diseases, PIC: Punjab Institute of Cardiovascular Disease, RCH: Red Crescent Hospital.
Note: Faisalabad Institute of Cardiology which had only recruited 35 individuals by the time of
the analysis was excluded from the dataset.
184
Table 6.2: Descriptive characteristics of PROMIS cases and controls by gender
Cases
% or mean (SD)
Controls
% or mean (SD)
P-value (1) P-value (2)
Male
(5,037
individuals)
Female
(1,013
individuals)
Male
(5,359
individuals)
Female
(1,512
individuals)
Age 52.7 (10.2) 56.6 (9.9) 53.0 (9.7) 53.5 (8.6) Matched1
History of hypertension 45% 67% 27% 38% <0.0001 <0.0001
History of diabetes 19% 37% 14% 17% <0.0001 0.001
Family history of MI 21% 20% 13% 12% <0.0001 0.531
Tobacco usage <0.0001 <0.0001
None 41% 82% 62% 87%
Current smoker 42% 4% 22% 2%
Current smokeless
user 10% 14% 12% 10%
Current both 7% 0% 4% 1%
Cooking medium used <0.0001 0.001
Ghee only 28% 23% 22% 27%
Oil only 44% 28% 51% 49%
LDL-C (mmol/l) 3.29 (1.68) 3.38 (2.85) 2.84 (1.64) 2.96 (2.67) <0.0001 0.015
Waist to Hip ratio
0.961
(0.090)
0.947
(0.153)
0.945
(0.088)
0.933
(0.143) <0.0001 <0.0001
(1): P-value comparing cases versus controls obtained fitting a linear regression for
continuous outcomes and a logistic (multinomial) for categorical outcomes adjusting for age,
sex, ethnicity and centre. (2): P-value comparing men versus women in controls only obtained
fitting a linear regression for continuous outcomes and a logistic (multinomial) for categorical
outcomes adjusting for age, ethnicity and centre. Means for LDL-C, and WHR were adjusted
for age equal 53 ( the mean age in controls), ethnicity and centre
185
List 6.1: Summary of questionnaire-based information collected in the Pakistan Risk
of Myocardial Infarction Study
186
List 6.2: Information collected on dietary intake and tobacco consumption
187
References
1. World Bank. Life expectancy at birth in Pakistan. Accessed online on 28-7-2011.
2. Mackay J, Mensah G, World Health Organization. The Atlas of heart disease and
stroke. 2004.
3. Anand SS, Yusuf S, Vuksan V et al. Differences in risk factors, atherosclerosis, and
cardiovascular disease between ethnic groups in Canada: the Study of Health
Assessment and Risk in Ethnic groups (SHARE). Lancet. 2000;356:279-284.
4. Balarajan R. Ethnic differences in mortality from ischaemic heart disease and
cerebrovascular disease in England and Wales. BMJ. 1991;302:560-564.
5. Forouhi NG, Sattar N, Tillin T, McKeigue PM, Chaturvedi N. Do known risk factors
explain the higher coronary heart disease mortality in South Asian compared with
European men? Prospective follow-up of the Southall and Brent studies, UK.
Diabetologia. 2006;49:2580-2588.
6. Nishtar S. Prevention of coronary heart disease in south Asia. Lancet. 2002;360:1015-
1018.
7. Ismail J, Jafar TH, Jafary FH, White F, Faruqui AM, Chaturvedi N. Risk factors for non-
fatal myocardial infarction in young South Asian adults. Heart. 2004;90:259-263.
8. Joshi P, Islam S, Pais P et al. Risk factors for early myocardial infarction in South
Asians compared with individuals in other countries. JAMA. 2007;297:286-294.
9. Forouhi NG, Sattar N, Tillin T, McKeigue PM, Chaturvedi N. Do known risk factors
explain the higher coronary heart disease mortality in South Asian compared with
European men? Prospective follow-up of the Southall and Brent studies, UK.
Diabetologia. 2006;49:2580-2588.
10. Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in 52
countries in the INTERHEART study: a case-control study. Lancet. 2006;368:647-658.
11. Iqbal R, Anand S, Ounpuu S et al. Dietary patterns and the risk of acute myocardial
infarction in 52 countries: results of the INTERHEART study. Circulation.
2008;118:1929-1937.
12. Rosengren A, Subramanian SV, Islam S et al. Education and risk for acute myocardial
infarction in 52 high, middle and low-income countries: INTERHEART case-control
study. Heart. 2009;95:2014-2022.
13. Saleheen D, Zaidi M, Rasheed A et al. The Pakistan Risk of Myocardial Infarction
Study: a resource for the study of genetic, lifestyle and other determinants of
myocardial infarction in South Asia. Eur J Epidemiol. 2009;24:329-338.
14. Central Intelligency Agency. The world factbook. 15-4-2012.
15. U.S.A Central Intelligence Agency. The World Factbook: Pakistan. CIA publications.
16. World Health Organization. Estimated proportional mortality (%), Pakistan, 2004. 17-8-
2011.
188
Chapter 7: Correlates of tobacco use in a developing country
Summary
Whilst the prevalence of smoking has decreased in Western countries over the past
few decades, the production and consumption of tobacco have been rapidly
expanding in developing countries. In contrast to industrialized countries where the
mass-produced cigarette is by far the most common form of tobacco use, cigarette
smoking coexists in low and middle income countries with a range of other practices
involving smoking, chewing or dipping tobacco. The prevalence of different types of
tobacco use and how they correlate with traditional risk factors of CHD has been little
studied in a developing country with a large population such as Pakistan.
The Pakistan Risk of Myocardial Infarction Study (PROMIS) recorded detailed
lifestyle information on over 12,000 cases and controls based in urban centres
across Pakistan. In controls, prevalence of tobacco use was higher in men, who
mostly smoked cigarettes and bidis, than women, who generally chewed paan.
Smoking forms of tobacco was correlated with a “protein and sweets” based diet, as
well as with the use of ghee rather than oil for cooking, and was most common
amongst the Punjabi ethnic group. Chewing tobacco was associated with a lower
socio-economic status, the use of oil versus ghee as a cooking medium, and was
more common amongst the Memon, the Gujrati and the Urdu ethnic groups.
Tobacco users of any type were less likely to report a history of diabetes or
hypertension. Waist to hip ratio and LDL-C did not differ significantly between users
and non users of tobacco products. In a developing country where resources are
scarce, these findings may help prioritize groups and design targets for public health
prevention of CVD. Adjustment for relevant correlates when investigating the
association with MI risk is also important and will be presented in Chapter 8.
189
7.1 Background
Of the total number of smokers worldwide, 84% live currently in developing
countries, which amounts to approximately 1.1 billion people 1. By 2030, more than
80% of tobacco related deaths are predicted to occur in lower and middle income
countries 2. However, evidence on the effect of tobacco use has been mainly derived
from studies in developed countries and, as a result; cigarette smoking has been the
main studied type of tobacco. In the developing world, and in particular in South
Asia, several forms of tobacco use coexist and include various types of tobacco
smoking, chewing and dipping 3-5. In addition to the usual cigarettes, bidis (hand-
rolled cigarettes where tobacco is wrapped in a dried temburini leaf and tied with a
string), hookah (water-pipe) and chilum (straight conical clay pipe) 6,7 are popular
smoking forms of tobacco. In Pakistan, smokeless tobacco is chewed as paan (a
mixture of areca nut, betel leaf, lime, catechu and tobacco), supari (chewed as a
digestive and made of processed areca nut, spices and flavourings), gutka (a
powdered mixture of areca nut, tobacco, catechu, lime, spices and flavourings), or as
oral snuff dipping called naswar (a moist mixture made from fresh tobacco leaves,
calcium oxide and wood ash) 8-11.
Smoking and smokeless forms of tobacco use have been gaining in popularity
amongst all sectors of the populations living in developing countries as a result of
aggressive marketing techniques by tobacco companies looking for new markets
with ineffective tobacco regulations 12, after restrictions on tobacco sale and
advertisement have been imposed in the developed world 13. There is evidence of a
rise in both overall and per capita cigarette consumption in the Eastern
Mediterranean and South Asian regions over the past decades 12,14. In Pakistan
itself, the production and consumption of cigarettes have doubled since 1970 15.
Water-pipe, which was in decline worldwide in the 1980’s, has been spreading in all
parts of the world and mainly in the Eastern Mediterranean region over the past 10
years 16. The current number of daily users of hookah and other water-pipes has
been estimated at 100 million people, and they are especially popular amongst
youths and women who perceive it as pleasant smooth smoke with a reduced harm
compared to cigarettes. Increasing use of smokeless tobacco has been reported in
India and several countries in South East Asia 17. Smokeless tobacco users in India
and Pakistan combined have been estimated to number 100 million individuals, with
an over-representation of children, adolescents and women 18,19. Chewable tobacco
is sometimes believed in these countries to have medicinal value for improvement of
toothache, headache and even stomach ache, and is often used as toothpaste
amongst children20,21.
190
As tobacco is a main preventable cause of disability and death, monitoring of its
different forms of use in developing countries is necessary. Pakistan is a country with
a large population of 165 million individuals where information on tobacco prevalence
and correlates is quasi-inexistent 22. A Pakistani based agency conducted a National
Health Study of Pakistani (NHSP) households, but individuals were enrolled in 1990-
1994 and there is no more recent information available 23. The WHO has set up a
Global Tobacco Surveillance System to evaluate tobacco control interventions and
monitor key articles of the WHO’s Framework Convention on Tobacco Control, which
is a broad treaty for global tobacco control brought into force in 2005 24. With the aim
of providing accurate documentation on prevalence of tobacco use, it recently
conducted cross-sectional Global Adult Tobacco surveys (GATS) in representative
countries. In its first phase it included a South Asian country, Bangladesh 25.
However, Pakistan is more advanced economically than Bangladesh, being
classified as a lower middle income country rather than a lower income country 26,
and has higher rates of urbanization, all factors which may influence prevalence and
type of tobacco use. INTERHEART, a recent worldwide case-control study of MI with
a South Asian component had only limited information on Pakistan itself and
aggregated data from countries at different stages of economic development and
with religious, lifestyle, ethnic and other differences which may reflect on tobacco use
(India, Pakistan, Bangladesh, Sri Lanka and Nepal) 27.
In this context the PROMIS controls represent the largest and most recent effort to
gather information on a range of lifestyle factors and biochemical factors relevant to
CVD in urban Pakistanis, including information from more than 6,000 individuals on
their smoking and smokeless habits. Controls were frequency matched to cases by
sex and by 5 year age bands, and were recruited randomly and primarily from
visitors of patients to the outpatient department. In this Chapter, I aim to better
characterize recent trends of tobacco use in urban Pakistan using the control
population of PROMIS participants, and investigate correlations of tobacco use with
a range of traditional and locally relevant risk factors of CVD.
7.2 Methods
7.2.1 Participants
The analysis of prevalence and correlates of tobacco use was restricted to the
control set of PROMIS participants. Details of the PROMIS data have been
described in Chapter 6. In particular, controls were individuals without a self-
reported history of CVD drawn from attendants of people visiting out-patients clinics
191
or non blood related attendants of cardiac patients 28. Controls were matched to
cases by hospital of recruitment and frequency matched by 5 year age bands and by
sex. Major forms of tobacco products in Pakistan have been described in Section
1.2. Participants were requested to fill in a lifestyle and medical questionnaire with a
detailed section on current and past use of the following tobacco products:
cigarettes, bidis, hookah, chilum, paan, supari, gutka, chalia and naswar. No
individuals declared smoking chalia. As all users of hookah or chilum also reported
smoking manufactured cigarettes or bidis, I was not able to study them separately
from cigarette smoking. For the assessments of correlates, tobacco use was
grouped into “smoking” (cigarettes, bidis, hookah, or chilum), “chewing” (paan,
supari, chalia or gutka) and “dipping”.
Statistical methods
Mean and standard deviations (SD) were reported for approximately normally
distributed variables. For skewed variables such as tobacco amounts, median and
inter-quartile ranges were reported. To compare tobacco usages across levels of a
categorical variable, 2 tests of independence were performed and “unadjusted p-
values” were reported. To account for the large number of tests performed, p-
values<0.01 (rather than p-values<0.05) were considered as significant. For
continuous variables, a Student t-test of equalities of the means across tobacco
usages was performed.
A multinomial logistic model adjusted for age, sex, ethnicity and centre was used to
predict tobacco status categorized into “never, “ex”, “current smoker only”, “current
chewer only”, “current naswar dipper only”, “current smoker and chewer/dipper”
(there were no chewer and dippers who were not smokers) depending on lifestyle
and medical characteristics. Multinomial logistic regression extends logistic
regression by estimating the effect of one or more exposure variables on the
probability that the outcome is in a particular category 29. Let X1 ,X2, …, Xi denote the
number that are in each of the i categories of tobacco use, in a sample of total size n,
and the probabilities of belonging to these classes p1, p2, …, pi. Let zj denote
baseline characteristics j, 1≤j≤k. The random variables X1 ,X2, …, Xi have a
multivariate distribution given by:
421 ...
!!...!!
!),...,,( 21
21
2211
n
i
nn
i
ii pppnnn
nnXnXnXP 
192
This likelihood represents an extension of the binomial distribution to the multivariate
case and is maximised under the constraints
(i) nn
i
l
l 
1
,
(ii) 1
1


i
l
lp and
(iii) for 2≤ j ≤i, j
k
j
j
l z
p
p



1
0
1
)ln(  ,
The first category of tobacco use was chosen as the reference group. Models were
adjusted for at least age, sex, recruitment centre and ethnicity.
In order to investigate the correlations between tobacco and diet, a principal
component analysis (PCA) on the 43 food items recorded in the questionnaire was
conducted on the model of the INTERHEART study 30 (see Appendix 2). Another
PCA was used to define “socio-economic status” and used the following variables:
monthly income (<10,000 Pakistani rupees, 10-20,000 Pakistani rupees, >20,000
Pakistani rupees), education (no formal education, 1-10 years, >10 years), asset
ownership (mobile phone, television, motorcycle, bicycle, home, computer, car,
radio); and occupation (unemployed, retired, housewife, unskilled labour, skilled
labour, clerical, business, professional, farmer, self-employed other) when available
(see Appendix 3). All analyses were run using STATA version 10 (StataCorp,
Texas).
7.3 Results
Descriptive characteristics of controls have been detailed in the previous Chapter.
There were 6,871 controls included in this analysis. Controls were recruited from
three sources: primarily as visitors of patients to the outpatient department of the
hospital where cases were recruited (71%), and otherwise as patients of the
outpatient department (14%) or as non blood related of MI patients (15%). Nearly
half of controls were recruited at the National Institute of CVD based in Karachi
(48%), and the rest of controls came mainly from the Punjab Institute of CVD (18%)
and the Karachi Institute of Heart Diseases (17%). Within each centre, the proportion
of women, mean age and tobacco uses was similar across different sources of
controls (Table 7.1). There were, however, some ethnic differences between
centres, which were sometimes located far apart. Unsurprisingly, the population in
the catchment area of the Punjab Institute of CVD was mainly Punjabi (90% of
193
controls) whilst centres located in Karachi mainly recruited individuals belonging to
the Urdu ethno-linguistic group (53% of Deewan Mushtaq Institute of CVD and 77%
of individuals from the Karachi Institute of Heart Diseases).
7.3.1 Tobacco use and sex
Current tobacco use was three times more common amongst men than women: 38%
versus 12% (Table 7.2). In men, the proportion of never users was 54% and of ex-
users 9%. In women, the proportion of never users was 86% and of ex-users 2%.
The most popular products amongst men were cigarettes or bidis : 26% prevalence
amongst male controls; followed by paan :8% of controls; and naswar 7% of controls.
The least popular products amongst men were gutka: 1% of controls; and supari :
0.4% of controls (Figure 7.1). 4% of men reported the use of several types of
tobacco products. Amongst men who chewed, 33% also reported smoking; and
amongst men who dipped naswar, 16% also reported smoking. Male tobacco users
consumed a median of 10 cigarettes or bidis, 5 paan, 5 gutka, 2 supari, and 1 dose
of naswar a day. Men who combined several types of tobacco use in general
reported lower amounts of each tobacco use than single users. For example men
who smoked and chewed reported an average of 4 paan per day and 7 cigarettes or
bidis per day; versus 5 paan per day for chewers only and 10 cigarettes or bidis per
day for smokers only.
The most popular product amongst women was paan, consumed by 8.1% of women.
Naswar was used by 2% of women, cigarettes or bidis by 2%, gutka by 0.4% and
supari by 0.3%. Multi-usage was rare: only 5% of women chewing also reported
smoking and/or dipping. Female current tobacco users reported an average of 5
cigarettes or bidis, 5 paan, 3 supari, 2 gutka and 1 dose of naswar a day.
In a multinomial model adjusted for age, ethnicity and centre of recruitment,
compared to never or ex tobacco users, male sex was significantly associated with
the probability of smoking (OR for women versus men: 0.05; 95% CI: 0.04-0.08) and
with dipping naswar (0.19; 0.12-0.29), but not with chewing (1.14; 0.91-1.42) (Table
7.3).
194
7.3.2 Tobacco use and age
Ex-users of tobacco were older on average than never or current tobacco users
(Table 7.2). The mean age was 56.3 years old (SD: 9.7) in ex-users, 52.3 (9.2) in
never users, and 53.1 (9.9) in current users (p-value<0.001). Younger individuals
were more likely to smoke than older individuals, but equally likely to use smokeless
products: the proportion of smokers only amongst controls aged ≤45 years old was 
21%, smokeless users only 11% and smoking and smokeless users 4%. By
comparison, amongst individuals aged >55 years old, the proportion of smokers only
was 16%, smokeless users only 13% and smoking and smokeless users 2%.
Investigating in more detail consumption of smokeless tobacco, prevalence of
chewing was similar across ages while prevalence of naswar use was greater in
older age groups: 5% of individuals ≤45 years old were dipping naswar, compared to 
7% of individuals aged >55 years old. After adjustment for conventional risk factors,
individuals >55 years old had 52% higher probability of dipping naswar than
individuals ≤45 years old (p-value: 0.01), 28% lower probability to chew (p-value: 
0.01) and 18% lower probability to smoke (p-value: 0.02) (Table 7.3).
7.3.3 Tobacco use and ethnicity
Tobacco users varied according to ethnic groups. Tobacco use was most common
amongst Pathan and least common amongst Punjabi (Table 7.2). Amongst men,
around 1 in 3 Punjabi (32%) smoked, compared to 1 in 4 Urdu and Sindh (24% and
25% respectively) and fewer than 1 in 6 Pathan (15%) (Figure 7.2) Naswar was
most popular amongst Pathan (36% of men) and minority ethnic groups (11% of
men), whilst it was relatively uncommon in other major ethnic groups: <5% in
Punjabi, Urdu and Sindh men. Chewable tobacco was favoured by Urdu (14% of
men current users or paan, supari or gutka) with low prevalence in Punjabi, Pathan
and Sindh (3%, 3% and 6% respectively). In minority ethnic groups, 10% reported
being current chewers. Prevalence of all forms of tobacco use was much lower in
women, independently of ethnic background. Women from the region of Sindh had
the highest prevalence of smoking: 8%, whilst smoking was not practiced by women
from other groups. Naswar was used by 5% of women from Sindh, 4% of women
from minority ethnicities and 2% belonging to the Pathan ethnic group. Regarding
chewable tobacco, nearly 1 in 5 Urdu women reported using chewable products
(17%), while the proportion was 10% of women from minorities, 6% of Sindhi
women, 3% of Pathan women and 2% of Punjabi women. In a multinomial model,
after adjustment for age, sex and centre of recruitment, Pathan ethnicity was
195
positively associated with naswar use, and Urdu ethnicity with the use of chewable
forms of tobacco. Both Pathan and Urdu ethnicities had a reduced probability of
smoking, while Punjabi were the most likely to smoke (Table 7.3).
There was no significant correlation between centre of recruitment and type of
tobacco use, apart from the Multan Institute of Heart Diseases (MIC) where the
probability of chewing was lower compared to other centres, even after adjusting for
sex, age and ethnic group (Table 7.3). The MIC is the only recruitment centre
located in the geographical centre of the country which recruited individuals with an
ethnic composition different from that of the other centres (45% of controls and 49%
of cases were classified as “Others” ethnic groups, probably Multani ethnicity).
7.3.4 Tobacco use and conventional risk factors of MI
There was a higher proportion of never or past tobacco users amongst individuals
diagnosed with diabetes or hypertension than in the general population of the
PROMIS controls: 73% of diabetics were non tobacco users versus 67% of non
diabetics, and the contrast was 74% versus 66% of controls for hypertension (Table
7.4). After adjustment for age, ethnicity, sex and centre of recruitment, individuals
with self reported diabetes or hypertension remained less likely to declare
themselves as smokers (P-value: 0.002 and <0.0001 respectively), but were equally
likely to chew tobacco or to dip naswar (p-values>0.01) than to be non users of
tobacco (Table 7.5). Compared to never users of tobacco, individuals with a history
of diabetes were marginally more likely to have stopped smoking than non-diabetics
(adjusted multinomial OR 1.28; 95%CI: 1.00-1.63; p-value: 0.048); and individuals
with hypertension to have stopped chewing than non-hypertensives (adjusted
multinomial OR 1.48; 95%CI: 1.02-2.15; p-value 0.034). Tobacco use was not
significantly influenced by family history of MI. Sensitivity analyses further adjusting
for socio-economic status and diet did not modify these results. Mean levels of lipids
did not significantly differ across different types of tobacco use and across status
(current, ex or never user) (Figure 7.3). There was also no evidence of a relationship
between waist to hip ratio and the use of tobacco (Figure 7.4).
7.3.5 Tobacco use and socio-economic status
Low socio-economic status was associated with higher tobacco consumption.
Amongst individuals earning ≥20,000 Pakistani rupees a month (~230 US dollars), 
the proportion of current tobacco users was 25%, while it was 36% amongst
196
individuals earning <10,000 Pakistani rupees a month (~115 US dollars) (Table 7.6).
Similar differences were observed between individuals reporting >10 years of formal
education, compared to no formal education (25% versus 36% of current users). The
proportion of current users of tobacco in the top third of the socio-economic gradient
was 25% versus 38% in the lowest third. In particular naswar dippers and chewers
were over-represented in lower socio-economic groups: 70% of naswar and 47% of
chewers were in the bottom third of socio-economic status. Onaverage, smokers in
the lowest third of socio-economic gradient reported smoking 10 cigarettes or bidis
per day (IQR: 5-20), compared to 12 (5-20) in the top third of socio-economic status
(p-value: 0.034).
After adjustment for age, sex, recruitment centre and ethnicity, earning ≥20,000 
Pakistani rupees a month was associated with a 37% lower likelihood to smoke, 55%
lower likelihood to chew and 70% lower probability to dip naswar compared to
individuals earning <10,000 Pakistani rupees a month (Table 7.7). There was a 53%
reduction in the probability of smoking, 61% reduction in the probability of chewing
and 90% reduction in the probability of dipping naswar for individuals reporting >10
years of formal education compared to individuals with no formal education. For
socio-economic status a reduction of 42%, a 63% and an 85% in the likelihood of
smoking, chewing and dipping was observed when comparing top versus bottom
third of the socio-economic gradient.
7.3.6 Tobacco use and diet
Individuals who cooked with ghee only were more likely to report current tobacco use
than individuals who cooked with oil (p-value<0.001) (Table 7.8). Amongst never or
ex-users of tobacco the prevalence of ghee users, exclusively or in combination with
oil, was 49%, while it was 42% amongst chewers, 53% amongst current smokers (or
multi-users) and 62% amongst naswar dippers. In a multinomial adjusted model, the
relationship between tobacco and ghee use remained significant: ghee only users
had twice the probability to be also naswar dippers and 57% higher probability to
smoke than oil only users (p-values<0.001) (Table 7.9). The association with
tobacco chewing was non significant.
Individuals in the top fifth of the distribution of the “vegetables and carbohydrate”
based diet were more likely to use tobacco than individuals consuming “vegetables
and carbohydrates” less frequently. The proportion of current tobacco users was
lowest in the middle categories of the “protein and sweets” based diet, whilst
extremes reported a higher prevalence of smokers. Amongst individuals scoring in
197
the top fifth for the “high meat, fish and sweet diet” dietary pattern, the proportion of
smokers was 26%, chewers 7% and naswar dippers 3%; compared to 19% smokers,
4% chewers and 4% naswar dippers in the middle quintile. However, in a multinomial
model adjusted for demographic variables, there was no significant evidence of a
correlation between tobacco use and dietary patterns (Table 7.9).
7.4 Discussion
7.4.1 Tobacco use and gender
The high prevalence of tobacco use amongst PROMIS participants is comparable to
other published findings. Smoking is most prevalent in men from developing
countries and South Asia is the region of the world with the highest prevalence of
smokeless tobacco users 1. Tobacco in its smoking, chewing or snuff forms was
used by nearly 2 in every 5 men and more than 1 in every 10 women of PROMIS
control participants. The proportion of current smokers was 26% men and 2%
women. The use of smokeless types of tobacco attained 9% of chewers and 7% of
naswar dippers in men; versus 9% and 2% in women. Rates of tobacco cessation
were low: only 9% of men and 2% of women declared being ex-tobacco users. The
patterns of tobacco use differed between sexes in PROMIS. The majority of male
tobacco users smoked while women were more likely to chew paan. Combining
smoking, chewing and dipping, tobacco use was more common amongst men than
women. After adjustment for other demographic characteristics, smoking and dipping
naswar were predominant amongst men, while chewing was not associated with a
specific gender.
These findings are in broad agreement with previous studies in South Asian
populations. The National Health Study of Pakistani (NHSP) conducted a decade
ago in Pakistan reported that 34% of men and 12.5% of women in Pakistan use
tobacco regularly 31. In the GATS survey of Bangladeshis established by WHO,
higher rates of both smoking and smokeless tobacco use were observed compared
to PROMIS. Bangladeshi men aged ≥15 years old reported a 47% current smoking 
rate, including 25% daily manufactured cigarette smoking, and a 26% current
smokeless tobacco rate 33. In women, rates were 1.5% current smoking, including
0.2% daily cigarette smoking and 28% current smokeless use. Regarding cessation
rates, rates in PROMIS were lower than in INTERHEART 34 (proportions of male ex-
smokers were 13% and 22% among young and old men respectively) and than in
198
GATS (17% in men and 41% in women). As INTERHEART and GATS were partially
or wholly set in South Asian countries other than Pakistan, these differences may
reflect differences in countries’ prevalence and reinforce the need for country specific
statistics in order to effectively monitor tobacco control in South Asia. Lower rates of
tobacco cessation in PROMIS compared to the rest of South Asia also raise the
alarm of a lack of effectiveness of public health campaigns intended to discourage
tobacco use in Pakistan relative to the rest of South Asia 35,36.
7.4.2 Tobacco use and age and socio-economic characteristics
Smoking was more likely amongst younger age groups, a pattern also observed by
INTERHEART 37. This raises concerns that younger age groups in Pakistan and
more generally in developing countries see smoking as part of the adoption of a
modern and westernized lifestyle which accompanies poor diet, low levels of physical
activity and is currently held responsible for an epidemic of obesity 38. Punjabi and
Sindhi favoured smoking, Pathan reported mainly using naswar, and Urdu both
smoked and chewed tobacco. In the NHSP, the highest prevalence of smoking was
also seen amongst urban Punjabi (30% amongst men) and Sindhi (24% amongst
men) compared to other ethnic groups and rural populations 39. Tobacco use was
more common in my study amongst individuals in lower socio-economic groups
characterized by short education, low income, poor asset ownership, over-
representation in jobs such as unskilled labour, and had a relatively large proportion
of unemployed individuals. In the NHSP study, illiteracy and low levels of education
were also significantly associated with tobacco use, but there was no significant
association with monthly income 40.
7.4.3 Tobacco use and conventional risk factors of MI
Regarding medical and biochemical risk factors for CVD, hypertension and diabetes
were highly prevalent amongst controls. Diabetes and hypertension are spreading
with urbanization in Pakistan. In the NHSP, 21% of urban compared to 16% of rural
participants reported being hypertensive 41. In the Pakistan National Diabetes Study,
the prevalence of diabetes was 10.8% in rural areas compared to 11.9% in urban
areas 42. In comparison, in PROMIS, 28% of controls had self reported hypertension
and 14% had diabetes, among which a quarter used tobacco products. There was a
slightly higher proportion of never or past tobacco users amongst diabetics and
hypertensive than in the general population of the PROMIS controls, maybe
reflecting a better awareness of the dangers of tobacco use in this group of people.
199
There was no significant difference in LDL-C and total cholesterol levels in current
versus ex or never tobacco users. This is in contrast to results from a large scale
meta-analysis of Western prospective cohort studies presented in Chapter 3 which
found a modest association with lipid levels. Similarly, no correlation was observed
between waist to hip ratio and tobacco use in PROMIS, which could be due to
measurement error in either waist to hip ratio or the recording of tobacco use,
confounded by factors not adjusted for in the linear regression of waist to hip ratio on
tobacco use such as physical activity, or a real lack of association between
anthropometric measurements and tobacco use in South Asian populations. These
findings would need to be replicated using a prospective design in order to confirm
the direction of causality.
7.4.4 Tobacco use and diet
Tobacco users were more likely to use ghee. Traditional South Asian diet is known to
be rich in fried food and, rather than being boiled, vegetables and meat are usually
fried in either ghee, which is clarified butter and is a saturated fat containing
cholesterol oxides, or vegetable oil 43. Ghee use has been associated with higher
adipose tissue levels of trans-fatty acids 44 than oil for cooking. The lifestyle in
Pakistan has been rapidly changing over the past 30 years, as urbanization has
been accompanied by higher consumption of meat, sweets and fat-rich food and a
preference for cigarettes over chewable forms of tobacco 45,46 However, the
association of tobacco use with the two main dietary patterns, “high vegetables and
carbohydrate diet” and “high meat, fish and sweet diet”, was not significant after
adjustment for demographic factors and ghee use.
7.4.5 Strengths and limitations
My study contains several strengths and limitations. Recruitment for the PROMIS
was performed in 6 centres located in 5 urban centres across Pakistan, enabling
investigation of tobacco use in all major ethnic groups of Pakistan. The PROMIS
represents the largest dataset in Pakistan with detailed information on several
smoking and smokeless forms of tobacco use, as well as other lifestyle and
biochemical factors for CVD. In comparison, INTERHEART included ~2,000 controls
from South Asian countries, and less than 1,000 specifically from Pakistan 47. The
NHSP included 9442 individuals ≥15 years old 48; however its recruitment started
more than a decade ago 49.
200
The PROMIS controls were not recruited to form a representative sample of the
population of Pakistan, but rather to be age and sex frequency matched to cases
without a self-reported history of CVD, and mainly identified as visitors of patients.
Nevertheless, prevalence of tobacco use was in broad agreement with previous
estimates 50. To remedy confounding by the demographic structure of my data, such
as an over-representation of men and Urdu in PROMIS compared to the overall
population of Pakistan, multinomial logistic regression were fitted adjusting at least
for age, sex, ethnicity and recruitment centre. Socio-economic status was assessed,
taking into account several social and economic dimensions including monthly
income, education, employment status, asset ownership of nine types of common
household possessions, and type of job categorized into six categories. Assessment
of diet included a question on the use of ghee, as well as a 43 food item
questionnaire on meals and food locally relevant to Pakistanis. Information was
sparse on amount amongst chewers and naswar dippers and was not collected
separately for cigarettes and for bidis. Data on specific types of tobacco such as
huqqa, chilum, paan, supari and gutka were sparse and these types could not be
investigated on their own but had to be grouped into broader categories such as
“smoking tobacco” and “chewing tobacco”. Epidemiological studies with designs
favouring enrolment of users of smokeless tobacco, for example set up in an Urdu
population, would be of interest in this respect, allowing the analysis of correlates of
specific subtypes of smokeless tobacco.
7.5 Conclusion
Tobacco use was widespread in the control set of PROMIS participants, with high
prevalence in all ethnic groups. Younger age groups were more likely to smoke
cigarettes or chew tobacco, whilst older age groups had a higher probability of
dipping naswar. Men had increased likelihood of smoking cigarettes or bidis, often in
combination with chewing or dipping tobacco; and women had increased likelihood
of chewing paan. Stopping smoking was rare, even amongst populations at high risk
of MI such as diabetics and hypertensive individuals. Current tobacco use was more
common amongst lower socio-economic groups and was correlated with a poor diet
characterized by the use of ghee rather than vegetable oil for cooking. These
findings emphasize the need for accurate monitoring of tobacco trends in Pakistan,
in order to better implement tobacco control.
201
Figure 7.1: Current tobacco use amongst 5,359 men and 1512 women controls
0
5
10
15
20
25
a) Men b) Women
Smoking cigarettes or bidis
Chewing paan
Chewing gutka
Chewing supari
Dipping naswar
Legend:
P
er
ce
nt
ag
e
(%
)
202
Figure 7.2: Current tobacco use amongst controls by ethnic group amongst 5,359 men and 1512 women controls
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
a) Punjabi, Men b) Punjabi, Women c) Urdu, Men d) Urdu, Women
e) Pathan, Men f) Pathan, Women g) Sindhi, Men h) Sindhi, Women
i) Others, Men j) Others, Women
P
er
ce
nt
ag
e
(%
)
P
er
ce
nt
ag
e
(%
)
P
er
ce
nt
ag
e
(%
)
Smoking cigarettes or bidis
Chewing paan
Chewing gutka
Chewing supari
Dipping naswar
Legend:
203
Figure 7.3: Adjusted lipid levels and 95%confidence intervals by tobacco use after multiple
adjustment
4.4
4.6
4.8
5
5.2
5.4
M
ea
n
to
ta
lc
ho
le
st
er
ol
(m
m
ol
/l)
Never/ex Dip naswar Chew Smoke
(and dip/chew)
2.4
2.6
2.8
3
3.2
3.4
M
ea
n
LD
L-
ch
ol
es
te
ro
l(
m
m
ol
/l)
Never/ex Dip naswar Chew Smoke
(and dip/chew)
Tobacco use
4.91
(0.06)
4.86
(0.09) 4.81
(0.09)
4.96
(0.06) 2.92
(0.07)
2.90
(0.07)
2.95
(0.05)2.94
(0.04)
Tobacco use
a) Total cholesterol b) LDL-cholesterol
Means and 95% CI were computed by fitting a linear regression of lipid levels over categories of
tobacco use, adjusting for age, sex, centre of recruitment and ethnicity. Adjusted means were
obtained as predicted coefficients for age 50 years old, Male, Urdu and averaging the effects of
centre. Adjusted means (standard deviations) are given for each category.
204
Figure 7.4: Levels of waist to hip ratio by tobacco use after multiple adjustment
M
ea
n
w
ai
st
to
hi
p
ra
tio
Never/ex Dip naswar Chew Smoke (and dip/chew)
Tobacco use
.9
0.95
1
0.964
(0.002)
0.959
(0.004) 0.963
(0.004)
0.963
(0.003)
Means and 95% CI were computed by fitting a linear regression of lipid levels over categories of
tobacco use, adjusting for age, sex, centre of recruitment and ethnicity. Adjusted means were
obtained as predicted coefficients for age 50 years old, Male sex, Urdu and averaging the effects of
centre. Adjusted means (standard deviations) are given for each category
205
Table 7.1: Prevalence of demographic factors by source of controls and by recruitment centre
DMIC
centre
(n=161)
KIHD centre
(n=1,165)
MIC
centre
(n=415)
NICVD centre
(n=3,326)
PIC centre
(n=1,2363)
RCH centre
(n=542)
Variable
Visitors
of MI
patients
(n=137)
Visitors
of OPD
patients
(n=1027)
Visitors
of MI
patients
(n=158)
OPD
patients
(n=869)
Visitors
of OPD
patients
(n=1826)
Visitors
of MI
patients
(n=307)
Visitors
of OPD
patients
(n=956)
Visitors
of MI
patients
(n=352)
Visitors
of OPD
patients
(n=188)
Age (years)
50.3
(9.2)
54.2
(7.5)
55.2
(9.1)
50.7
(7.9)
58.1
(8.3)
52.7
(9.9)
48.3
(9.6)
50.4
(8.9)
52.5
(3.5)
51.6
(98.3)
Female sex 18% 14% 19% 15% 25% 29% 18% 25% 30% 21% 27%
Major sub ethnicities
Punjabi 2% 8% 9% 36% 13% 17% 20% 88% 91% 5% 10%
Urdu 53% 80% 77% 18% 41% 42% 34% 7% 4% 53% 46%
Pathan 1% 3% 5% 1% 11% 14% 13% 2% 1% 1% 2%
Sindhi 41% 3% 3% 0% 20% 18% 17% 0% 1% 39% 37%
Others 2% 6% 6% 45% 14% 8% 16% 4% 3% 2% 6%
Tobacco use
Never user 53% 49% 52% 67% 61% 63% 54% 65% 71% 75% 76%
Ex-user 11% 18% 13% 7% 6% 5% 8% 5% 6% 1% 1%
Dipping only 1% 1% 4% 1% 11% 9% 8% 3% 1% 1% 2%
Chewing only 8% 12% 12% 0% 5% 5% 8% 1% 1% 3% 4%
Smoking only 21% 16% 16% 21% 14% 16% 18% 25% 20% 18% 17%
Smoking & dipping/chewing 6% 4% 3% 2% 4% 1% 4% 2% 1% 1% 2%
N: Number of individuals; SD: standard deviation; DMIC: Deewan Mushtaq Institute of Cardiology; FIC: Faisalabad Institute of Cardiology; KIHD: Karachi Institute of
Heart Diseases; MIC: Multan Institute of Heart Diseases; NICVD: National Institute of Cardiovascular Diseases; PIC: Punjab Institute of Cardiovascular Disease;
RCH: Red Crescent Hospital. OPD: Outpatient Department. Information on source of controls was missing from 473 controls from NICVD and 1 control from KIHD.
1 individual from MI and 2 individuals from RCH who were OPD patients were excluded from this Table. Columns are column percentages
206
Table 7.2: Description of tobacco use by demographic characteristics
Tobacco use
Median
number
cigarettes
/ bidis per
day
Never/Ex-
user
(n=4675)
Dip
naswar
(n=348)
Chew
(n=450)
Smoke
(and
dip/chew)
(n=1416) P-value
Sex <0.0001
Male 62% 6% 6% 26% 10
Female 88% 2% 9% 2% 5
Age, mean (SD) 53.0 (9.4) 55.5 (9.9) 53.1 (9.9) 52.2 (9.8) <0.0001
≤45 years 
old 66% 4% 6% 24% <0.0001 10
46-55 years
old 69% 5% 6% 20% 12
>55 years
old 68% 7% 7% 19% 10
Major ethnic groups <0.0001
Punjabi 72% 3% 1% 24% 12
Urdu 67% 2% 12% 19% 10
Pathan 60% 26% 2% 12% 6
Sindhi 70% 4% 4% 22% 10
Other 63% 8% 7% 22% 10
Centre of recruitment <0.0001
DMIC 64% 1% 8% 27% 10
FIC 79% 0% 0% 21% 5
KIHD 65% 4% 12% 19% 10
MIC 74% 1% 0% 24% 10
NICVD 64% 8% 8% 20% 10
PIC 75% 2% 1% 22% 12
RCH 77% 1% 3% 19% 10
N: Number of individuals; SD: standard deviation; DMIC: Deewan Mushtaq Institute of
Cardiology; FIC: Faisalabad Institute of Cardiology; KIHD: Karachi Institute of Heart
Diseases; MIC: Multan Institute of Heart Diseases; NICVD: National Institute of
Cardiovascular Diseases; PIC: Punjab Institute of Cardiovascular Disease; RCH: Red
Crescent Hospital. Percentages correspond to row percentages. P-value from a 2 test of
independence between row and column variables. The median number of cigarettes or bidis
per day is reported amongst current controls who smoke.
207
Table 7.3: Demographic determinants of tobacco use after multiple adjustment
Dip naswar Chew Smoke (and dip/chew)
Basic adjustment
P-value
Basic adjustment
P-value
Basic adjustment
P-valueOR (95% CI) OR (95% CI) OR (95% CI)
Sex
Male 1 - 1 - 1 -
Female 0.19 (0.12;0.29) <00001 1.14 (0.91;1.42) 0.3 0.05 (0.04;0.08) <0.001
Age, mean (SD) 1.02 (1.00;1.03) 0.01 0.99 (0.98;1) 0.01 0.99 (0.99;1) 0.09
≤45 years old  1 - 1 - 1 -
46-55 years old 1.28 (0.92; 1.78) 0.14 0.75 (0.58; 0.97) 0.03 0.88 (0.75; 1.03) 0.11
>55 years old 1.52 (1.10; 2.10) 0.01 0.72 (0.55; 0.93) 0.01 0.82 (0.70; 0.97) 0.02
Major ethnic groups
Punjabi 1 - 1 - 1 -
Urdu 0.55 (0.36;0.85) 0.007 4.22 (2.76;6.46) <0.001 0.69 (0.57;0.83) <0.001
Pathan 6.27 (4.23;9.3) <0.001 0.67 (0.31;1.46) 0.3 0.49 (0.35;0.67) <0.001
Sindhi 0.99 (0.60;1.62) 1 1.57 (0.92;2.69) 0.1 0.73 (0.57;0.93) 0.01
Other 2.45 (1.61;3.71) <0.001 3.42 (2.11;5.54) <0.001 0.84 (0.67;1.06) 0.1
Centre of recruitment
DMIC 1 - 1 - 1 -
KIHD 2.66 (0.62;11.34) 0.2 1.3 (0.7;2.4) 0.4 0.71 (0.47;1.07) 0.1
MIC 0.48 (0.09;2.49) 0.4 0.06 (0.01;0.27) <0.001 0.59 (0.38;0.94) 0.03
NICVD 3.28 (0.79;13.61) 0.1 1.17 (0.64;2.14) 0.6 0.78 (0.53;1.14) 0.2
PIC 1.00 (0.22;4.52) 1 0.2 (0.08;0.5) 0.001 0.62 (0.4;0.95) 0.03
RCH 0.94 (0.2;4.54) 0.9 0.31 (0.14;0.67) 0.003 0.62 (0.41;0.96) 0.03
OR: Odds ratio; 95% CI: 95% confidence interval. Odds ratios were computed using a
multinomial logistic regression with never or ex tobacco users chosen as the reference group.
Models were adjusted for age as a continuous variable (except when looking at the effect of
categories of age, when age as a continuous covariate was omitted from the model), sex,
major ethnic groups and centre of recruitment. Coefficients for Faisalabad Institute of
Cardiology were not estimable because of low numbers.
208
Table 7.4: Conventional CVD risk factors and tobacco use
Tobacco use
Never/Ex-
user Dip naswar Chew
Smoke
(and
dip/chew)
P-value(n=4675) (n=348) (n=450) (n=1416)
Self reported diabetes 0.0004
No 67% 5% 7% 21%
Yes 73% 4% 6% 16%
Self reported hypertension <0.0001
No 66% 5% 6% 23%
Yes 74% 4% 7% 15%
Family history of CAD 0.009
No 68% 5% 6% 21%
Yes 69% 3% 7% 20%
Biochemical information
Total cholesterol (mmol/l),
mean (SD))
4.66
(1.35)
4.37
(1.26)
4.37
(1.41) 4.69 (1.42)
LDL-C (mmol/l), mean (SD)
2.88
(1.05)
2.77
(1.08)
2.78
(1.08) 2.87 (1.11)
Waist to hip ratio, mean (SD)
0.943
(.0671)
0.937
(.0616)
0.936
(.0665)
0.946
(.0631)
N: Number of individuals; SD: standard deviation; cig: cigarettes. Percentages correspond to
row percentages. P-value from a 2 test of independence between row and column
variables. The median number of cigarettes or bidis per day is reported amongst current
controls who smoke.
209
Table 7.5: Conventional CVD risk factors and tobacco use after multiple adjustment
Dip naswar only
(n=348)
Chew only
(n=450)
Smoke
(and dip/chew)
(n=1416)
Basic
adjustment
P-value
Basic
adjustment
P-value
Basic
adjustment
P-valueOR (95% CI)
OR (95%
CI) OR (95% CI)
Diabetes
0.75
(0.53;1.07) 0.115
0.83
(0.62;1.11) 0.203
0.74
(0.61;0.89) 0.002
Hypertension
0.75
(0.57;0.99) 0.043
0.9
(0.72;1.12) 0.351 0.69 (0.59;0.8) <0.0001
Family history
of MI
0.94
(0.63;1.41) 0.773
1.23
(0.93;1.64) 0.147
0.92
(0.77;1.11) 0.387
Multinomial logistic regression adjusted for age, sex, major ethnic groups and centre of
recruitment. Never/ex tobacco users were chosen as the reference group.
210
Table 7.6: Association of socio-economic status with tobacco use
Tobacco use
Median
number of
cig./bidis
per day
Never/Ex-
user
(n=4675)
Dip
naswar
only
(n=348)
Chew
only
(n=450)
Smoke
(and
dip/chew)
(n=1416) P-value
Income <0.0001
Low (<10,000 Pakistani
rupees/month) 64% 7% 7% 21% 10
Middle (10-20,000
Pakistani rupees/month) 68% 5% 7% 20% 10
High (≥20,000 Pakistani
rupees/month) 75% 2% 3% 20% 12
Education <0.0001
No formal education 64% 9% 8% 19% 10
1-10 years 65% 5% 7% 23% 10
>10 years 75% 1% 4% 20% 10
Socio-economic gradient <0.0001
Lower third 62% 9% 7% 21% 10
Middle third 67% 5% 8% 20% 10
Top third 75% 1% 3% 21% 12
N: Number of individuals; SD: standard deviation; Percentages correspond to row
percentages; cigs: cigarettes. P-value from a 2 test of independence between row and
column variables. The median number of cigarettes or bidis per day is reported amongst
current controls who smoke.
211
Table 7.7: Socio-economic status and tobacco use after multiple adjustment
Dip naswar only (n=348) Chew only (n=450)
Smoke (and dip/chew)
(n=1416)
OR (95% CI)
P-
value OR (95% CI)
P-
value OR (95% CI)
P-
value
Income
Low (<10,000
Pakistani
rupees/month) 1 - 1 - 1 -
Middle (10-20,000
Pakistani
rupees/month)
0.79
(0.61; 1.04) 0.091
0.91
(0.72; 1.16) 0.474
0.88
(0.76; 1.02) 0.101
High (≥20,000
Pakistani
rupees/month)
0.32
(0.21; 0.48) 0
0.45
(0.33; 0.63) 0
0.73 (0.62;
086) 0
Education
No formal education 1 - 1 - 1 -
1-10 years
0.53
(0.40; 0.69) 0
0.93
(0.73; 1.19) 0.556
0.75
(0.64; 0.88 0.001
>10 years
0.10
(0.06; 0.17) 0
0.39
(0.29; 0.52) 0
0.47
(0.40; 0.56) 0
Socio-economic gradient
Lower third 1 - 1 - 1 -
Middle third
0.5
(0.38;0.66) 0
0.93
(0.73;1.18) 0.56
0.78
(0.67; 0.91) 0.003
Top third
0.16
(0.1;0.24) 0
0.37
(0.27; 0.51) 0
0.58
(0.50; 0.69) 0
Multinomial models were adjusted for age, sex, ethnicity and recruitment centre.
212
Table 7.8: Diet and tobacco use
Tobacco use Median
number
of
cig./bidis
per day
Never/Ex-
user
(n=4675)
Dip
naswar
only
(n=348)
Chew
only
(n=450)
Smoke
(and
dip/chew)
(n=1416) P-value
Cooking fat <0.0001
Oil 70% 4% 7% 20% 10
Oil & ghee 67% 5% 7% 21% 10
Ghee only 65% 7% 3% 25% 10
Dietary pattern 1
“High vegetables and carbohydrates” 0.0009
Quintile 1 70% 4% 4% 22% 10
Quintile 2 69% 4% 6% 21% 12
Quintile 3 70% 4% 6% 20% 10
Quintile 4 71% 5% 6% 18% 10
Quintile 5 65% 6% 7% 23% 10
Dietary pattern 2
“High mean, fish and sweets” <0.0001
Quintile 1 69% 6% 6% 19% 10
Quintile 2 69% 5% 6% 20% 10
Quintile 3 72% 4% 5% 19% 10
Quintile 4 69% 4% 7% 20% 10
Quintile 5 64% 3% 7% 26% 10
N: Number of individuals; SD: standard deviation; Percentages correspond to row
percentages; cigs: cigarettes. P-value from a 2 test of independence between row and
column variables. The median number of cigarettes or bidis per day is reported amongst
current controls who smoke.
213
Table 7.9: Diet and tobacco use after multiple adjustment
Dip naswar only (n=348) Chew only (n=450)
Smoke (and dip/chew)
(n=1416)
Basic adjustment P-
value
Basic adjustment
P-value
Basic adjustment
P-valueOR (95% CI) OR (95% CI) OR (95% CI)
Cooking fat
Oil 1 - 1 - 1 -
Oil & ghee 1.48 (1.06;2.06) 0.021 1.36 (1.05;1.76) 0.02 1.09 (0.93;1.29) 0.286
Ghee only 2.11 (1.52;2.93) 0 0.88 (0.59;1.32) 0.542 1.57 (1.31;1.89) 0
Dietary pattern 1
“High vegetables and carbohydrates”
Quintile 1 1 - 1 - 1 -
Quintile 2 1.14 (0.72;1.82) 0.576 1.62 (1.08;2.45) 0.021 1.05 (0.85;1.32) 0.638
Quintile 3 1.26 (0.79;2.03) 0.336 1.72 (1.13;2.62) 0.011 1.08 (0.86;1.36) 0.519
Quintile 4 1.37 (0.9;2.08) 0.143 1.77 (1.19;2.62) 0.005 0.95 (0.77;1.19) 0.68
Quintile 5 1.64 (1.09;2.46) 0.017 1.59 (1.07;2.34) 0.021 1.21 (0.98;1.5) 0.073
Dietary pattern 2
“High mean, fish and sweets”
Quintile 1 1 - 1 - 1 -
Quintile 2 0.94 (0.65;1.36) 0.749 0.95 (0.67;1.34) 0.767 1.16 (0.94;1.43) 0.159
Quintile 3 0.79 (0.53;1.18) 0.246 0.72 (0.49;1.06) 0.095 0.98 (0.79;1.22) 0.849
Quintile 4 0.79 (0.52;1.21) 0.281 1.06 (0.74;1.53) 0.734 1.02 (0.81;1.28) 0.868
Quintile 5 0.54 (0.34;0.88) 0.012 1.05 (0.73;1.53) 0.78 1.41 (1.13;1.76) 0.003
n: Number of individuals. Multinomial models adjusted for age, sex, ethnicity, recruitment
centre, type of ghee used for cooking and dietary patterns.
214
References
1. World Health Organization. Tobacco: deadly in any form or disguise. 2006.
2. Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002
to 2030. PLoS Med. 2006;3:e442.
3. Changrani J, Gany FM, Cruz G, Kerr R, Katz R. Paan and Gutka Use in the United
States: A Pilot Study in Bangladeshi and Indian-Gujarati Immigrants in New York City. J
Immigr Refug Stud. 2006;4:99-110.
4. Dobe M, Sinha DN, Rahman K. Smokeless tobacco use and its implications in WHO
South East Asia Region. Indian J Public Health. 2006;50:70-75.
5. Imam SZ, Nawaz H, Sepah YJ, Pabaney AH, Ilyas M, Ghaffar S. Use of smokeless
tobacco among groups of Pakistani medical students - a cross sectional study. BMC
Public Health. 2007;7:231.:231.
6. Nisar N, Qadri MH, Fatima K, Perveen S. A community based study about knowledge
and practices regarding tobacco consumption and passive smoking in Gadap Town,
Karachi. J Pak Med Assoc. 2007;57:186-188.
7. Anjum Q, Ahmed F, Ashfaq T. Knowledge, attitude and perception of water pipe
smoking (Shisha) among adolescents aged 14-19 years. J Pak Med Assoc.
2008;58:312-317.
8. Ali NS, Khuwaja AK, Ali T, Hameed R. Smokeless tobacco use among adult patients
who visited family practice clinics in Karachi, Pakistan. J Oral Pathol Med.
2009;38:416-421.
9. Akhtar S, Sheikh AA, Qureshi HU. Chewing areca nut, betel quid, oral snuff, cigarette
smoking and the risk of oesophageal squamous-cell carcinoma in South Asians: A
multicentre case-control study. Eur J Cancer. 2011.
10. Zakiullah, Saeed M, Muhammad N et al. Assessment of potential toxicity of a
smokeless tobacco product (naswar) available on the Pakistani market. Tob Control.
2011.
11. Mazahir S, Malik R, Maqsood M et al. Socio-demographic correlates of betel, areca
and smokeless tobacco use as a high risk behavior for head and neck cancers in a
squatter settlement of Karachi, Pakistan. Subst Abuse Treat Prev Policy.
2006;1:10.:10.
12. World Health Organization. WHO report on the global tobacco epidemic; the MPOWER
package. World Health Organization . 2008.
13. Tong EK, Glantz SA. ARTIST (Asian regional tobacco industry scientist team): Philip
Morris' attempt to exert a scientific and regulatory agenda on Asia. Tob Control.
2004;13 Suppl 2:ii118-24.:ii118-ii124.
14. Shafey O, Dolwick S, Guindon EG, World Health Organization, American Cancer
Society, International Union Against Cancer. Tobacco control country profile. 2nd ed.
2003.
15. World Health Organization. Country profiles, Pakistan. 25-8-2011.
16. Maziak W. The global epidemic of waterpipe smoking. Addict Behav. 2011;36:1-5.
215
17. Dobe M, Sinha DN, Rahman K. Smokeless tobacco use and its implications in WHO
South East Asia Region. Indian J Public Health. 2006;50:70-75.
18. Imam SZ, Nawaz H, Sepah YJ, Pabaney AH, Ilyas M, Ghaffar S. Use of smokeless
tobacco among groups of Pakistani medical students - a cross sectional study. BMC
Public Health. 2007;7:231.:231.
19. Gupta PC, Ray CS. Smokeless tobacco and health in India and South Asia.
Respirology. 2003;8:419-431.
20. Imam SZ, Nawaz H, Sepah YJ, Pabaney AH, Ilyas M, Ghaffar S. Use of smokeless
tobacco among groups of Pakistani medical students - a cross sectional study. BMC
Public Health. 2007;7:231.:231.
21. Gupta PC, Ray CS. Smokeless tobacco and health in India and South Asia.
Respirology. 2003;8:419-431.
22. The World Bank: Pakistan. http://data.worldbank.org/country/pakistan accessed
04/07/2012..
23. Ahmad K, Jafary F, Jehan I et al. Prevalence and predictors of smoking in Pakistan:
results of the National Health Survey of Pakistan. Eur J Cardiovasc Prev Rehabil.
2005;12:203-208.
24. www.who.int/fctc/en. WHO Framework Convention on Tobacco Control. Accessed
24/08/2012.
25. Giovino GA, Mirza SA, Samet JM et al. Tobacco use in 3 billion individuals from 16
countries: an analysis of nationally representative cross-sectional household surveys.
Lancet. 2012;380:668-679.
26. Chartsbin (http://chartsbin.com/view/2438) and The World Bank. Country Income
Groups (World Bank Classificatin). Accessed 21/08/2012.
27. Joshi P, Islam S, Pais P et al. Risk factors for early myocardial infarction in South
Asians compared with individuals in other countries. JAMA. 2007;297:286-294.
28. Saleheen D, Zaidi M, Rasheed A et al. The Pakistan Risk of Myocardial Infarction
Study: a resource for the study of genetic, lifestyle and other determinants of
myocardial infarction in South Asia. Eur J Epidemiol. 2009;24:329-338.
29. Kirkwood B, Sterne J. Logistic regression: controlling for confounding and other
extensions. In: Blackwell Science, Medical statistics. 2 ed. Massachusetts, USA:
2003:212-3.
30. Iqbal R, Anand S, Ounpuu S et al. Dietary patterns and the risk of acute myocardial
infarction in 52 countries: results of the INTERHEART study. Circulation.
2008;118:1929-1937.
31. Nanan, D. and White, F. PROCOR: 7/25/99: The National Health Survey of Pakistan:
Review and discussion of report findings pertaining to selected risk factors for
cardiovascular disease.
32. Ahmad K, Jafary F, Jehan I et al. Prevalence and predictors of smoking in Pakistan:
results of the National Health Survey of Pakistan. Eur J Cardiovasc Prev Rehabil.
2005;12:203-208.
33. Giovino GA, Mirza SA, Samet JM et al. Tobacco use in 3 billion individuals from 16
countries: an analysis of nationally representative cross-sectional household surveys.
Lancet. 2012;380:668-679.
216
34. Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in
52 countries in the INTERHEART study: a case-control study. Lancet. 2006;368:647-
658.
35. IRIN. Pakistan: anti-smoking law have little impact. IRIN Humanitarian news and
analysis. 2004.
36. Ali NS, Khuwaja AK, Ali T, Hameed R. Smokeless tobacco use among adult patients
who visited family practice clinics in Karachi, Pakistan. J Oral Pathol Med.
2009;38:416-421.
37. Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in
52 countries in the INTERHEART study: a case-control study. Lancet. 2006; 368:647-
658.
38. Raymond SU, Leeder S, Greenberg HM. Obesity and cardiovascular disease in
developing countries: a growing problem and an economic threat. Curr Opin Clin Nutr
Metab Care. 2006;9:111-116.
39. Nasir K, Rehan N. Epidemiology of cigarette smoking in Pakistan. Addiction.
2001;96:1847-1854.
40. Alam AY, Iqbal A, Mohamud KB, Laporte RE, Ahmed A, Nishtar S. Investigating socio-
economic-demographic determinants of tobacco use in Rawalpindi, Pakistan. BMC
Public Health. 2008;8:50.:50.
41. Nanan, D. and White, F. PROCOR: 7/25/99: The National Health Survey of Pakistan:
Review and discussion of report findings pertaining to selected risk factors for
cardiovascular disease. 25-7-1999.
42. Shera AS, Rafique G, Khawaja IA, Baqai S, King H. Pakistan National Diabetes
Survey: prevalence of glucose intolerance and associated factors in Baluchistan
province. Diabetes Res Clin Pract. 1999;44:49-58.
43. Lip GY, Malik I, Luscombe C, McCarry M, Beevers G. Dietary fat purchasing habits in
whites, blacks and Asian peoples in England--implications for heart disease prevention.
Int J Cardiol. 1995;48:287-293.
44. Misra A, Khurana L, Isharwal S, Bhardwaj S. South Asian diets and insulin resistance.
Br J Nutr. 2009;101:465-473.
45. Singh RB, Niaz MA, Ghosh S et al. Association of trans fatty acids (vegetable ghee)
and clarified butter (Indian ghee) intake with higher risk of coronary artery disease in
rural and urban populations with low fat consumption. Int J Cardiol. 1996;56:289-298.
46. Hakeem R, Thomas J, Badruddin SH. Food habits and nutrient density of diets of
Pakistani children living in different urban and rural settings. J Health Popul Nutr.
2002;20:255-263.
47. Yusuf S, Hawken S, Ounpuu S et al. Effect of potentially modifiable risk factors
associated with myocardial infarction in 52 countries (the INTERHEART study): case-
control study. Lancet. 2004;364:937-952.
48. Ahmad K, Jafary F, Jehan I et al. Prevalence and predictors of smoking in Pakistan:
results of the National Health Survey of Pakistan. Eur J Cardiovasc Prev Rehabil.
2005;12:203-208.
49. Nanan, D. and White, F. PROCOR: 7/25/99: The National Health Survey of Pakistan:
Review and discussion of report findings pertaining to selected risk factors for
cardiovascular disease. 25-7-1999.
217
50. Nanan, D. and White, F. PROCOR: 7/25/99: The National Health Survey of Pakistan:
Review and discussion of report findings pertaining to selected risk factors for
cardiovascular disease. 25-7-1999.
218
Chapter 8: Smoking and smokeless tobacco use and the risk of
myocardial infarction in Pakistan
Summary
South Asians are particularly susceptible to premature CHD in their home countries
and compared to their host population, when they migrate to Western industrialised
countries. The epidemic of CHD amongst South Asians has been attributed to the
increasing prevalence of known risk factors such as cigarette smoking, and some
have hypothesized the importance of locally relevant factors such as the use of
smokeless forms of tobacco. However, epidemiological studies on cigarette smoking
have rarely been conducted in South Asian settings, and the relationship between
alternative forms of tobacco use and MI risk remains uncertain. PROMIS represents
the largest epidemiological resource for the study of indigenous forms of tobacco use
and the risk of CHD, with over 12,000 first ever MI cases and age and sex frequency
matched controls recruited in 5 urban centres across Pakistan. Odds ratios for non-
fatal MI were 3.37 (3.05; 3.71) with current smoking only, 1.71 (1.46; 2.00) with
current chewing only, and 1.46 (1.21; 1.78) with current snuff dipping only, when
compared to never tobacco users. Combining several types of tobacco use conferred
a risk of 3.91 (3.21; 4.76) with respect to never tobacco users. There was evidence
of a dose-response relationship between both smoking and smokeless tobacco with
the risk of MI, with the sharpest increase in risk experienced below 5 doses a day.
These effects were not modified upon adjustment for local diet, use of ghee for
cooking and socio-economic status. My analysis represents the first robust
epidemiological evidence of a harmful effect of smoking and smokeless tobacco in
the South Asian context. These findings reinforce the fact that urgency is needed to
tackle the tobacco epidemic in Pakistan, and discourage the use of both smoking
and smokeless forms of tobacco in all segments of the population, in order to
alleviate the exploding burden of CHD in the region.
219
8.1 Background
Low and middle income countries now shoulder 80% of the burden of chronic
diseases; and the South Asian continent, which hosts a quarter of the world’s
population, carries a substantial share of this burden 1,2. In Pakistan, coronary heart
diseases represent about 25% of all deaths and account for 150,000 deaths a year 1.
They threaten to counter major health gains achieved in the country over the past
decades which have led to a prolongation of the life expectancy from 50 to 67 years
between 1960 and 2009 3. They also represent a substantial economic cost for a
middle low-income country of 165 million people 4.
The reasons for the disproportionately high burden of CHD in developing countries
are multiple and complex. Compared to their Western hosts, South Asian migrants
have been reported to shoulder a 40% to 60% higher risk of CVD 5 and experience
higher vascular mortality 6,7 at younger ages : 5 to 10 years earlier than their Western
counterparts on average 8. Higher levels of conventional risk factors of CHD in the
Pakistani population is likely to play a role 9. South Asian migrants living in Western
industrialised countries have a higher prevalence of diabetes and lower HDL-C levels
than their host populations 9. In Pakistan itself, it is estimated that a third of
Pakistanis more than 45 years old have hypertension and the country ranks in the
top 10 world nations for the prevalence of diabetes 10.
Smoking is another one of the “Western” risk factors which have spread to
developing countries over the past decades and contributed to the epidemic of
chronic disease. Already in 1983, Crofton was writing that “many countries are well
on the way to adding a formidable epidemic of smoking related diseases to their
already overwhelming health problems... All of us should feel a responsibility towards
helping to prevent the tragedy.” 11. In South Asia, the per-capita consumption of
manufactured cigarettes has been continuously increasing since the 1970s 12.
Smoking has traditionally been done using hand-rolled cigarettes called bidis, and
water pipes such as hookah and chilum, and these local types of tobacco remain
extremely popular 13. In addition, over the past decades, tobacco companies have
developed a large panel of manufactured tobacco products suited to South Asian
taste such as gutka, by adding tobacco to traditional chewing and snuffing mixtures
of betel and areca nut called paan 14 (see Chapter 1 Section 1.2).
Despite the need for large scale epidemiological studies investigating the relationship
between these multiple forms of tobacco use and MI risk 2, previous studies on
220
South Asians have been characterized by small sample sizes, generally less than
1,000 cases and controls15,16-18, conducted amongst South Asians living abroad 19 or
have not taken into account ethnic specific differences within the South Asian
subcontinent 20-22. Most studies have focused on manufactured rather than hand-
rolled cigarettes (bidis) which are popular in South Asian countries. Smokeless
tobacco has been associated with CVD risk in Western countries 23,24, where
practices and compositions of tobacco products are likely to differ from the types
which are used in Pakistan. Most of the studies were based in Sweden or the USA
where the main type of smokeless tobacco used is a type of snuff called snus with a
specific heating procedure of tobacco leaves destined to limit its carcinogenic effects.
In comparison, chewing is more popular amongst Pakistanis than snuff dipping and
the type of snuff used differs in composition and is called naswar 25.
In this context, there is a need to investigate the association between different
smoking and smokeless types of tobacco use with the risk of cardiovascular
diseases in South Asia. PROMIS represents the largest case-control study of first-
onset acute MI conducted in South Asia with a detailed lifestyle questionnaire
enquiring about types of tobacco use in Pakistan as well as biochemical information.
The present analysis aims to better assess than the published literature the effect of
cigarette and indigenous modes of tobacco use in the context of traditional and
locally relevant biochemical, medical and lifestyle factors on the risk of MI in urban
Pakistanis. As tobacco usages are relatively similar across South Asia 26, findings
from this analysis could have important public health implications for prevention of
CVD in the whole region.
8.2 Method
8.2.1 Participants
Details of PROMIS participants have been described in Chapter 6. The analysis
included 6,051 first ever MI cases and 6,871 age and sex frequency matched
controls enrolled by March 2011 who were recruited from 6 centres located in 5 cities
in Pakistan. Tobacco use was categorized as “never user”, “ex-user”, “current user of
dipping tobacco only (naswar)”, “current user of chewing tobacco only (paan, gutka
or supari)”, “current smoker only (manufactured cigarettes or bidis) and “current
smoker and smokeless user (chewing or dipping tobacco)”. To investigate the dose-
response relationship, number of manufactured or hand rolled cigarettes per day was
categorised as 0 “non smoker”, <5, 5-10, 10-15, 15-20 and ≥20 per day for current 
221
smokers. Smokeless tobacco amount was categorised as “non smokeless user”, <5,
5-10, 10-15 and ≥15 doses per day for current smokeless users.  
8.2.2 Statistical methods
Odds ratios (ORs) were calculated using unconditional logistic regression as cases
and controls were frequency matched by sex and 5 year age bands rather than
individually matched. The general form of a logistic regression is:
kk xxxp
p
 

...)
)1(
ln( 22110
Where p represents the probability of experiencing an MI,
)1( p
p

represents the
odds of experiencing an MI, x1...xk are k exposure variables and k ....1 are
regression coefficients associated with the k exposure variables. In this model,
0 corresponds to the odds of disease in the baseline group
and )exp( i corresponds to the odds ratio for exposure i (i>0).
Odds ratios were adjusted at least for matching and demographic variables (age,
sex, recruitment centre and ethnicity); and when indicated, also adjusted for
conventional risk factors of MI: history of diabetes or hypertension, family history of
MI, waist to hip ratio and LDL-C. ORs investigating the relationship between smoking
amount (respectively smokeless amount) and MI risk were also adjusted for
smokeless (respectively smoking) tobacco use categorized as “current”, “past” or
“never” user. Further analyses investigated the effect of progressive adjustment for
diet (high protein and sweet diet; and high vegetable and carbohydrate diet), the use
of ghee – clarified butter used in traditional cooking and high in trans-fatty acids 27 -
versus oil for cooking; and socio-economic status categorized into low, middle and
high social class. Dietary patterns and socio-economic status were identified by
performing principal component analyses (see Appendices 2 & 3). The dietary
analysis was done on 43 questions from a locally validated food frequency
questionnaire capturing dietary habits of Pakistanis; and the analysis of socio-
economic status was done using monthly income, ownership of household items and
formal education.
For ORs of tobacco uses with MI risk, never tobacco users were chosen as the
reference group. For the computation of ORs with smoking/smokeless amount, never
smokers/smokeless users were chosen as reference groups and past tobacco users
were excluded.
222
To characterize shapes of associations, ORs were calculated within pre-defined
categories of smoking and smokeless amount and were plotted against mean values
within each quintile or category. Graphs representing odds ratios used log-linear
scale to allow graphical assessment of the linearity of dose-response relationships.
To enable graphical comparison of any two ORs and not only with the reference
group, 95% confidence intervals were drawn using “floating absolute” variances,
which are extracted after transforming the matrix of variance-covariance of the
coefficients under the constraint of nullity of the transformed covariances 28.
Effect-modification was investigated graphically and analytically for age group (≤50 
years old compared to >50 years old), sex, ethnicity, socio-economic status, dietary
patterns, medical history and metabolic factors. Forest plots displayed ORs for
current and ex versus never tobacco users by subgroups. Never tobacco users who
were also either ≤50 years old, Men, Punjabi, in the bottom third of LDL-C, in the 
bottom third of WHR, in the bottom third of socio-economic status, below median for
the consumption of vegetables and carbohydrate diet, or below median for the
consumption of meat and protein diet, were chosen as reference groups. Formal
tests of effect modification were performed by adding interaction terms to logistic
regression models and then tested for significance. To maximize power, continuous
variables in their original forms (rather than divided around the median or into thirds)
were used for the tests of interaction. To take into account the large number of tests,
p-values of interaction <0.001 were emphasized. Because there were relatively few
women smoking cigarettes or bidis (41 cases and 30 controls), it was not possible to
investigate the effect of smoking (only or in combination with smokeless products) by
sex.
Sensitivity analyses explored heterogeneity across centres of recruitment and by
source of controls. In a first analysis, ORs were computed within each centre and
pooled by meta-analysis rather than adjusted for centre in a single model. A second
analysis was performed when controls in a specific centre were recruited from
different sources (>100 controls for a specific source). In that case, ORs comparing a
specific type of controls to cases in each were computed, to allow comparison of
ORs depending on the source of controls. Heterogeneity levels between estimates
for different centres or sources were assessed by reporting I2 values and confidence
intervals 29. All analyses were conducted using STATA v10 (StataCorp, Texas) and
used code purposely developed.
223
8.3 Results
Cases and controls have been described in Chapter 6; and correlations between
tobacco use and other lifestyle and medical risk factors have been described in
Chapter 7. The average age of cases was 53.3 years old (SD: 10.2) and 83% were
men. Amongst cases, 41% reported being never tobacco users, 6% ex-users, 4%
snuff dippers only, 6% current chewers only, 36% current smokers only; and 6%
combined smoking and smokeless use.
8.3.1 Association of smoking with MI risk
Current cigarette or bidi smoking was most strongly associated with MI risk. Current
cigarette and bidi smokers versus never tobacco users had an OR for non-fatal MI of
3.37 (3.05; 3.70), after adjustment for demographic and conventional risk factors
(Figure 8.1a). Associations were unchanged upon further adjustment for diet and for
socio-economic status (Figure 8.1b). There was a non linear positive association
between the number of cigarettes or bidis smoked per day and MI risk which was not
altered upon adjustment (Figure 8.2a & 8.3a). Compared to never users, individuals
smoking <5 cigarettes or bidis per day had an OR (95% CI) of 1.76 (1.45; 2.13),
smokers of 5-10 cigarettes or bidis per day an OR of 2.51 (2.03; 3.10), smokers of
10-15 cigarettes or bidis per day an OR of 2.63 (2.24; 3.07), smokers of 15-20
cigarettes or bidis per day an OR of 2.75 (2.05; 69) and smokers of ≥20 cigarettes or 
bidis per day an OR of 4.33 (3.84; 4.87).
The effect of current smoking of cigarettes or bidis was higher in younger age groups
(≤50 years old) compared to older age groups (>50 years old): ORS of 3.93 (3.46; 
4.46) versus 2.88 (2.58; 3.22), compared to never tobacco users (Figure 8.4). The
association did not differ according to sex or ethnic group (p-value >0.001). Smoking
was associated with MI risk in individuals independently of the presence or absence
of a history of hypertension (Figure 8.5). Individuals in the top third of the distribution
of LDL-C who smoked had an OR of 6.08 (5.24; 7.05) compared to never smokers in
the bottom third of LDL-C, and individuals in the top 3rd of WHR had an excess risk
of 5.41 (4.62; 6.34) compared to never smokers in the bottom 3rd of the distribution of
WHR (Figure 8.6). Smoking was most harmful for individuals with low socio-
economic status compared to individuals with high socio-economic status: 4.31
(3.78; 4.91) versus 3.11 (2.72; 3.57) (Figure 8.7). There was no evidence of effect
modification by diet.
224
There was no significant heterogeneity in ORs between recruitment centre and
depending on the source of controls (p-value of heterogeneity: 0.64) (Figure 8.8).
ORs were comprised of values between 2.8 and 4.6 with overlapping confidence
intervals and I2 was equal to 0. Pooled RRs from fixed effects and random effects
meta-analyses of centre specific ORs were similar and did not differ from ORs when
using a one-step rather than a two-step approach (Figure 8.9).
8.3.2 Association of smokeless tobacco with MI risk
Ninety percent of the individuals who chewed tobacco declared using paan rather
than other chewing products (supari and gutka). Only one type of snuff, called
naswar, was recorded in PROMIS. Compared to never-consumers of tobacco, ORs
for non-fatal MI were 1.46 (1.20; 1.77) with current snuff dipping and 1.71 (1.46;
2.00) with current chewing after adjusting for demographic and conventional risk
factors (Figure 8.1a). Individuals who currently used smokeless tobacco on top of
smoking had a greater risk than individuals who currently only chewed or dipped and
their risk was slightly higher than that of current smokers alone: ORs were 3.91
(3.21; 4.76) for individuals using smoking and smokeless tobacco, versus 3.36 (3.05
to 3.71) for smokers alone. The associations were independent of known lifestyle
and medical risk factors of MI, and were minimally affected by adjustment (Figure
8.1b).
Looking at the shape of association in relation to smokeless amount, there was a
positive association up to 10 smokeless products a day and a non significant
decrease in risk afterwards. ORs were 1.32 (1.16; 1.52) for users of <5 smokeless
products a day, 1.77 (1.43; 2.20) for 5-10 products a day, 1.74 (1.26; 2.41) for 10-15
products a day and 1.49 (1.01; 2.19) for ≥15 products a day (Figure 8.2 & 8.3).
ORs for chewing and dipping tobacco did not significantly differ across age groups,
by sex, and by ethnic group (p-values>0.001) (Figure 8.4). Individuals who dipped
tobacco and were hypertensive had an OR of 3.12 (2.30; 4.24) compared to
normotensive and never users of tobacco (Figure 8.5). There was evidence
suggestive of a stronger increase in MI risk when using smokeless tobacco and
belonging to the top third of socio-economic status compared to the bottom third
(Figure 8.7).
There was evidence of heterogeneity across centre and source of controls for the
effect of smokeless tobacco use (chewing and dipping tobacco combined) versus
never users: I2 was equal to 54.8% and ORs for non-fatal MI ranged from 1.05 to
225
5.32 (Figure 8.8). However, random and fixed effects meta-analyses produced
similar ORs, which were significant and only slightly lower than the 1-step estimate:
pooled OR by random effect meta-analysis was 1.53 (1.14; 2.05) and by fixed effects
meta-analysis it was 3.38 (3.04; 3.76) (Figure 8.9). Level of heterogeneity was low
for the ORs comparing smoking and smokeless tobacco users versus never users:
I2: 37.5% and p-values: 0.099.
8.3.3 Ex-tobacco users
Ex-users of all forms of tobacco experienced a slightly increased risk compared to
never users, which was significantly lower than current users of any type of tobacco.
The OR for past versus never users was 1.22 (1.04; 1.42) (Figure 8.1). There was
no significant effect modification according to age, sex or ethnicity (Figure 8.4). The
increase in risk due to past tobacco use was significant in individuals who otherwise
had a relatively low risk of CVD: non diabetics, non hypertensive, individuals without
a history of MI; whilst it did not carry extra-risk in individuals who were already at
high risk because they were diabetics, hypertensive or reported a family history of MI
(Figure 8.5). The level of heterogeneity was below 50% for ex versus never users
(Figure 8.8). In a sensitivity analysis, pooling centre-specific ORs gave a pooled
estimate of 1.31 (1.08; 1.58) by random effects meta-analysis, and 1.30 (1.10; 1.52)
by fixed effects meta-analysis (Figure 8.9).
8.4 Discussion
My analysis represents the first robust epidemiological evidence of a harmful effect
of smoking and smokeless tobacco on the risk of CVD in the South Asian context.
The findings contained in this Chapter should alert South Asian governments to the
urgency needed to tackle both the smoking and smokeless tobacco epidemics
happening in their countries. They may also be generalisable to individuals of South
Asian origins who currently live in Western countries such as the USA and the UK
and who have retained the tobacco habits of their countries of origin.
8.4.1 Strengths of the analysis
I studied smoking in relation to MI risk in a large Pakistani urban population using a
case-control study design. This analysis included 6,050 cases and 6,871 controls
from Pakistan, making it more than 3 times as large as the recent INTERHEART
case-control study on risk factors for MI risk in South Asians 8. INTERHEART also
did not investigate the effect of dipping tobacco, only that of chewing tobacco.
Studies on dipping tobacco have all been conducted in Northern Europe and North
226
America, where products and toxic contents differ from South Asian products 23,24.
PROMIS individuals had detailed lifestyle and biochemical information on a range of
risk factors for CVD, including locally relevant factors such as dietary habits and
ghee consumption. Analyses were adjusted for a large number of potential
confounders and I performed tests of effect modifications of both smoking and
smokeless tobacco with other cardiovascular risk factors. Alcohol was not included in
the list of potential confounders. This is because alcohol consumption is religiously
prohibited to Muslims and as 95% of PROMIS individuals declared being Muslim
alcohol consumption may be subject to misreporting.
Individuals were drawn from all major ethnic groups of Pakistan, enabling
comparison across ethnic groups, and making my results applicable to neighbouring
South Asian countries where these ethnicities are represented. PROMIS was
conducted in 5 urban centres across Pakistan. Currently 37% of the Pakistani
population and 30% of the South Asian population live in urban areas 30,31. Urban
Pakistanis tend to favour smoking products, in particular manufactured cigarettes,
while Pakistanis living in rural areas have especially high rates of smokeless tobacco
and overall higher rates of tobacco prevalence than urban areas 32,33. Urbanization is
happening at a fast rate in Pakistan and in South Asia in general and therefore
smoking habits and the use of manufactured cigarettes are likely to keep expanding
with economic development, unless drastic political measures are taken to curb the
trend 34.
On the other side, smokeless tobacco has a long history and is unlikely to regress in
the immediate future in South Asia 13,35. It is reportedly more acceptable than
smoking amongst children, teenagers and women 35. Low socio-economic status
groups favour smokeless tobacco, and especially snuff dipping, which tends to be
cheaper than manufactured cigarettes. The commercial production and marketing of
smokeless tobacco products has also promoted a rapid increase in sale over the
past decades. In India, per capita smokeless tobacco consumption has increased
among the poor since the 1960s in both rural and urban areas and the total number
of smokeless tobacco users in India and Pakistan combined has been estimated to
number 100 million individuals 18,19. The rate of growth of gutka , a manufactured
product with long shelf life, has overtaken that of smoking forms of tobacco in India
26. In this context, the findings of this study have wider implications for the whole of
the South Asian region where similar smoking and smokeless products coexist.
8.4.2 Smoking tobacco and MI risk
227
First, I have confirmed an association between cigarette smoking and the risk of MI
in a South Asian population. Current smokers experienced an OR of 3.36 (3.05;
3.70) compared to never tobacco users. The strength of association was similar to
that of previous reports from case-control studies 20, and was relatively higher than
the summary estimate of a doubling in risk in Western populations I obtained
analysing the ERFC (see Chapter 4). Several factors may account for this observed
difference. ORs were estimated here because of the retrospective nature of the
PROMIS study, whilst HRs were estimated in the ERFC dataset which uses a
prospective design. Cases enrolled in PROMIS may be subject to recall bias and
may be more likely to report their smoking history than controls, inflating the ORs;
whilst the prospective nature of the ERFC ensures information on exposure is
collected before individuals experience their first cardiovascular event. This
difference could also be due to the pooling of cigarettes with bidi smokers in
PROMIS, whilst only cigarettes smokers were considered in the ERFC. This
difference could finally reflect real differences in susceptibility to CVD in Pakistanis
compared to Western European population when smoking.
There was evidence of a dose-response relationship between the number of
cigarettes or bidis smoked per day and MI risk. The relationship was non linear,
which is in agreement with findings from the ERFC dataset, with the highest relative
increase experienced for smokers of 1-5 cigarettes per day versus never smokers
corresponding to ~80% higher risk. By comparison, INTERHEART reported an
approximately linear increase of 6% higher risk per additional cigarette 20. The
association was independent of conventional risk factors as well as economic status
and diet, showing that clustering of poor health records and unhealthy diet and low
socio-economic status cannot account for this contrast.
ORs were higher in younger individuals. This is in agreement with published findings
that younger age groups are at higher risk when they smoke 37 and with my findings
in developed populations (see Chapter 3). There was similarly no effect modification
according to sex, in contrast to a literature based meta-analysis which found a ~25%
greater increase in risk for women versus men 5; and in agreement with ERFC
results and findings from INTERHEART.
8.4.3 Smokeless tobacco use and MI risk
Secondly, these results provide a novel insight into the epidemiology of smokeless
tobacco in relation to CHD in a South Asian context. Smokeless tobacco is known to
be responsible for cancer and tobacco use has been estimated to account for about
228
50% of oral cancers in India 25. With the growth of gutka, the incidence of oral
submucous fibrosis has reached epidemic proportions in India among individuals
below 35 years old. Because tobacco is not burned and therefore there is no
production of carbon monoxide, smokeless tobacco is often thought of as having no
cardiovascular effect, and it is commonly used as a breath freshener or as
toothpaste 38,39.
In my analysis, both chewing and dipping tobacco were strongly associated with MI
risk. The magnitude of the estimate for chewing tobacco is in agreement with that of
INTERHEART but with a narrower confidence interval in my analysis which may be
attributable to greater homogeneity of chewing products used in Pakistan compared
to the rest of the world 20. Indeed, INTERHEART aggregated data from participants
located in 52 countries located in all continents around the world and chewing
tobacco is sold in various forms and shapes which are likely to differ in toxicity. A
recent meta-analysis of chewing and dipping products combined, where most of the
evidence came from Northern American studies of chewable tobacco, also
demonstrated a significant association with smokeless tobacco of a relatively lower
magnitude: the RR for MI was 1.13 (1.03; 1.21) for all smokeless products23.
An association between snuff dipping and MI risk is being debated in the literature
and most of the evidence comes from studies conducted in northern Europe 40. A
recent meta-analysis selecting only Swedish data found a non significant increase in
MI risk for snuff users (HR of 1.04; 95% CI 0.93 to 1.17) 41. Possible explanations for
the presence of a significant association in Pakistan and non significant association
in Western countries with regard to MI risk include true heterogeneity of results by
geographic area due to differences in the composition of the products used in the
USA, Northern Europe and South Asia. The major components of tobacco snuff are
alkaloids, with nicotine as the main compound (85-95% of total alkaloids) 25. During
product manufacturing, tobacco leaves, stems and other ingredients are blended to
achieve a specific nicotine content, pH, taste, flavour and aroma. The pH strongly
affects the concentration of bioavailable nicotine, whereas the nitrite content affects
nitrosamine concentrations in the product. The major form of smokeless tobacco
used in Sweden and the US is snus which is made of air-cured and fire-cured
tobacco, flavoured and powdered into fine particles, containing 20-55% moisture by
weight. Products used in the USA have higher nitrosamine content than those in
Northern Europe and, as a result, an increased risk of oral cancer for use of
smokeless tobacco has been reported in the USA and not in Nordic countries 25. The
Pakistani form of tobacco dipping, naswar, contains tobacco which undergoes a
229
different curing process than snus, and is typically less moist and has higher levels of
nitrosamines 26.
The magnitude of excess risk when chewing or dipping tobacco was relatively strong
in my dataset, close to 50% increased risk for naswar dipping and over 70%
increased risk for chewing tobacco. This was smaller than the OR associated with
smoking products, but still highly significant. Individuals who combined the use of
smokeless products with that of smoking products did not experience any benefit
compared to using smoking products alone in terms of MI risk. A study conducted in
the US found that the use of smokeless tobacco may lead to subsequent cigarette
smoking. Young males who were not smokers but regularly used smokeless tobacco
were more than three times as likely as never users to become current smokers
within the next four years 42. In PROMIS, individuals belonging to the control group
who were current chewers were also more likely to smoke than never chewers (24%
versus 20%), and individuals who had stopped chewing were also more likely to
have stopped smoking (30% of ex-chewers were ex-smokers, compared to 5% of
current chewers) (p-value χ2 of independence between smoking and chewing status
<0.001). Education on the dangers of all forms of tobacco use should encourage
individuals to stop smoking and using smokeless forms of tobacco, rather than
(partially) switch from one form to another, without any real benefit on cardiovascular
health.
These findings also suggest a role for toxins that are intrinsic to tobacco itself, and
not just confined to the smoked form. Animal experiments and studies in human
have found short term effects of smokeless tobacco on the hemodynamic system.
Nicotine, which is present in all forms of tobacco use, has been shown in animal
experiments to produce arrhythmias and ventricular fibrillations (see Chapter 1
Section 1.4.1) and may be one mediator of smokeless tobacco effect on MI risk.
Other mechanisms include acute elevation of blood pressure, chronic hypertension
and acute activation of the sympathetic nervous system 24. Higher blood pressure
levels have been found in smokeless tobacco users as well as a higher prevalence
of hypertensive individuals, in a Swedish cross-sectional study including more than
30,000 individuals 43. Smokeless tobacco products also have high amounts of
sodium chloride, which could contribute to inflammation and tumour promotion 25. In
my dataset, individuals who were diabetics or hypertensive showed a non significant
increase in risk of MI when they used smokeless tobacco products. Some of this
attenuation could be due to recall bias in non diabetics and normotensive, especially
as these variables were self-reported.
230
8.4.4 Past use of tobacco and MI risk
Ex-users experienced a slightly increased risk, which was non significant amongst
younger age groups. The magnitude of the OR for ex-users was below the OR for
current users of any tobacco type. Stopping the use of tobacco needs to be more
actively encouraged in Pakistan and more largely in South Asia. In PROMIS, ex-
smokers represented between 5% and 10% of controls depending on ethnic group,
three or four times less than the proportion of current tobacco users in these groups.
In women controls, <2% indicated having stopped tobacco use, and the proportion in
men was below 10% (9%). Implementing tobacco regulations has not been really
successful so far in the South Asian subcontinent. In India, legislation mandating
pictorial warnings on smokeless and smoking tobacco packaging is not effective
because of inappropriate pictures chosen and irrelevant text messages 44. In
Pakistan, anti-smoking laws which ban consumption in public areas and storage near
educational institutions were passed in 2009 45, but are not being enforced 46. This
analysis provides evidence supporting public health campaigns to encourage
cessation of all forms of tobacco use.
8.4.5 Limitations
Despite its strengths, this analysis contains several limitations. Controls came from 3
different sources, with nearly ¾ of controls recruited amongst visitors of patients from
the outpatient department. However, estimates were relatively homogeneous across
different types of controls within the same centre and across different centres.
Heterogeneity levels were relatively low, and estimates using a one step approach
were in broad agreement with estimates obtained from a two-step approach
combining centre-specific ORs by random and fixed effects meta-analyses. Case-
control studies are prone to recall bias due to MI cases being more likely to
overestimate their past smoking and smokeless tobacco use, and MI controls being
more likely to under-estimate it, leading to an over-estimation of ORs. As a result,
case-controls are unable to prove causality and a prospective design would be
needed to strengthen the evidence of a causal association between smokeless
tobacco and MI risk.
Information was not available on the type of tobacco used by past users. The brevity
of the questionnaire on tobacco use meant previous information was not asked such
as duration of tobacco use, age starting and, in ex-users, age stopped and number
of years since stopped. I was also not able to investigate cigarettes and bidies
231
separately, as amount was not available separately for manufactured and hand
rolled cigarettes. Smokeless products can be homemade or sold in pouches and
quantities of tobacco are not standardized across products and brands. Therefore, it
was not possible to accurately investigate the relationship between tobacco dose of
smokeless tobacco and risk of MI. I had to group all chewing types of tobacco (paan,
supari and gutka) together into a single category “chewing tobacco” to maximize
power. Finally, analyses were adjusted for a large number of conventional and locally
relevant risk factors. In particular, I adjusted analyses for local dietary pattern, ghee
consumption and several variables representing socio-economic status. However,
some risk factors were not available for these analyses. Blood pressure was for
example measured but considered as unreliable in cases that had just been
administrated stabilizing drugs after experiencing an MI. Self reported history of
hypertension was used instead.
8.5 Conclusion
In Pakistan, all forms of tobacco use are hazardous to cardiovascular health. The
increase in risk was independent of conventional risk factors, diet and socio-
economic markers. The effect of chewing and dipping tobacco was intermediate
between that of non tobacco users and that of smoking. These findings should help
devise strategies to address the increasing burden of tobacco related CVD in South
Asia.
232
Figure 8.1: Odds ratios for myocardial infarction with tobacco usage a) with
adjustment for conventional factors, and b) showing progressive adjustment
Adjusted for age, sex, recruitment centre, ethnicity
Further adjusted for socio-economic status,
dietary patterns and cooking oil
Ne
ve
r u
se
r
Ex
-u
se
r
Di
p
on
ly
(n
as
wa
r)
Ch
ew
on
ly
(P
aa
n/
Su
pa
ri/
Gu
tka
)
Sm
ok
e
on
ly
(C
iga
re
tte
s/B
ee
die
s)
Sm
ok
e
an
d
dip
/ch
ew
1
1.
2
1.
5
2
2.
5
3
4
5
O
R
(9
5%
C
I)
1
1.
2
1.
5
2
3
4
5
O
R
(9
5%
C
I)
Ne
ve
r u
se
r
Ex
-u
se
r
Di
p
on
ly
(n
as
wa
r)
Ch
ew
on
ly
(P
aa
n/
Su
pa
ri/
Gu
tka
)
Sm
ok
e
on
ly
(C
iga
re
tte
s/B
ee
die
s)
Sm
ok
e
an
d
dip
/ch
ew
Further adjusted for LDL-c, WHR, diabetes,
hypertension and family history of MI
a) b)
OR: Odds ratio; 95% CI: 95% Confidence interval plotted using “floating absolute variances”.
“Conventional risk factors” of adjustment include age, sex, ethnicity, recruitment centre; self
reported diabetes or hypertension, LDL-C levels, WHR and family history of MI. In figure a),
ORs (95% CI) compared to never users were 1.22 (1.04; 1.42) for ex-users, 1.46 (1.20; 1.77)
for dipper only, 1.71 (1.46; 2.00) for chewers only, 3.36 (3.05; 3.71) for smokers only and
3.91 (3.21; 4.76) for individuals who dipped or chewed tobacco and also smoked. For figure
b), the dataset was restricted to 5,365 cases and 5,557 controls with information on all the
covariates of adjustment.
233
Figure 8.2: Odds ratios for myocardial infarction with a) smoking and b) smokeless
tobacco amount
1
1.
2
1.
5
2
3
4
5
O
R
(9
5%
C
I)
0 5 10 15 20 25 30
Number of cigarettes per day
1
1.
2
1.
5
2
3
4
5
O
R
(9
5%
C
I)
0 5 10 15 20 25
Number of smokeless products per day
a) Smoking tobacco amount b) Smokeless tobacco amount
OR: Odds ratio; 95% CI: 95% Confidence interval plotted using “floating absolute variances”.
Models were adjusted for age, sex, ethnicity, recruitment centre; self reported diabetes or
hypertension, LDL-C levels, WHR, family history of MI and a) chewing and dipping status or
b) smoking status. ORs were 1.32 (1.16; 1.52) for users of <5 smokeless products a day,
1.77 (1.43; 2.20) for 5-10 products a day, 1.74 (1.26; 2.41) for 10-15 products a day and 1.49
(1.01; 2.19) for ≥15 products a day.  
234
Figure 8.3: Progressive adjustment of the associations of a) smoking and b)
smokeless tobacco amount with risk of myocardial infarction
1
1.
2
1.
5
2
3
4
5
O
R
(9
5%
C
I)
0 5 10 15 20 25 30
1
1.
2
1.
5
2
3
4
5
O
R
(9
5%
C
I)
0 5 10 15 20 25
a) Smoking tobacco amount b) Smokeless tobacco amount
Number of cigarettes per day Number of smokeless products per day
Adjusted for age, sex, recruitment centre, ethnicity
Further adjusted for socio-economic status, dietary patterns and cooking oil
Further adjusted for LDL-c, WHR, diabetes, hypertension and family
history of MI and smoking (smokeless amount)
OR: Odds ratios and 95% confidence interval. Dataset restricted to 5,365 cases and 5,557
controls with information on all the covariates of adjustment. Odds ratios were computed
within categories of number of cigarettes or bidis per day smoked by current smokers (0: non
smokers, <5, 5-10, 10-15, 15-20 and ≥20) or categories of smokeless tobacco (0: non 
smokers, <5, 5-10, 10-15, and ≥15)) and plotted against the arithmetic mean within each 
category.
235
Figure 8.4: Association of tobacco use with MI risk by socio-demographic
characteristics
≤50  
≤50  
Urdu
Others
Others
Punjabi
Punjabi
>50
Women
Women
Women
Punjabi
Others
≤50  
Urdu
Urdu
Men
Punjabi
>50
Men
≤50  
Men
Women
Others
Men
Urdu
Men
>50
≤50  
≤50  
>50
>50
Urdu
Women
Urdu
Punjabi
Others
Punjabi
>50
Others
43% / 20%
36% / 61%
41% / 59%
5% / 5%
33% / 16%
5% / 7%
45% / 22%
44% / 60%
4% / 2%
80% / 86%
3% / 2%
2% / 3%
5% / 7%
4% / 5%
30% / 16%
2% / 2%
34% / 53%
4% / 2%
5% / 6%
43% / 22%
3% / 4%
4% / 6%
2% / 2%
41% / 58%
5% / 6%
7% / 3%
7% / 9%
5% / 2%
8% / 3%
6% / 7%
8% / 9%
7% / 6%
12% / 12%
12% / 9%
8% / 8%
42% / 65%
7% / 4%
2% / 1%
32% / 17%
9% / 11%
3.93 (3.46, 4.46)
1.00 (0.90, 1.11)
1.26 (1.15, 1.38)
1.98 (1.45, 2.71)
3.54 (3.06, 4.09)
1.22 (0.93, 1.60)
3.89 (3.41, 4.44)
0.96 (0.88, 1.04)
3.18 (1.85, 5.49)
0.88 (0.80, 0.97)
1.24 (0.72, 2.13)
1.51 (0.97, 2.34)
1.12 (0.85, 1.48)
1.28 (0.97, 1.69)
3.74 (3.29, 4.26)
1.73 (1.17, 2.57)
1.00 (0.94, 1.07)
4.28 (2.76, 6.63)
1.37 (1.10, 1.70)
3.35 (3.09, 3.62)
1.42 (1.02, 1.98)
1.42 (1.18, 1.70)
1.50 (0.78, 2.89)
1.09 (0.98, 1.21)
1.75 (1.48, 2.08)
4.97 (3.76, 6.58)
1.22 (1.05, 1.41)
3.28 (2.54, 4.25)
4.69 (3.56, 6.19)
1.90 (1.50, 2.42)
1.14 (0.96, 1.36)
1.56 (1.29, 1.89)
2.04 (1.71, 2.43)
1.48 (1.13, 1.94)
1.72 (1.39, 2.12)
1.00 (0.89, 1.12)
4.02 (2.97, 5.44)
2.15 (1.27, 3.64)
2.88 (2.58, 3.22)
1.54 (1.24, 1.92)
0.09
0.95
0.42
1 1.5 2 3 4 5 6
Category Subgroup % or mean
cases/controls
OR (95% CI) P-value
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
By age groups
By sex
By ethnicity
Current chew only
Current smoke only
Current dip only
Ex-user
Never user
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
Odds ratio (95% CI)
OR: Odds ratios; 95% CI: 95% confidence interval, P-value: p-value of interaction. Confidence
intervals were plotted using “floating absolute” variances for tobacco use. Models were
adjusted for conventional risk factors: age, sex, ethnicity, centre of recruitment, LDL-C, WHR,
history of diabetes or hypertension and family history of MI. P-values of interaction are derived
from tests of significance of the interaction terms added to the model. The proportion of women
smoking and dipping or chewing was too small to enable investigation of potential interactions
with sex on MI risk. Black boxes correspond to the reference group (age 50 and below, Men,
Punjabi); whilst blue and red boxes indicate other groups.
236
Figure 8.5: Association of tobacco use with MI risk according to medical history
.
No
No
Yes
No
No
Yes
Yes
Yes
Yes
No
Yes
No
No
No
Yes
Yes
Yes
Yes
No
No
Yes
No
Yes
No
6% / 3%
36% / 18%
35% / 18%
40% / 20%
4% / 4%
4% / 5%
41% / 60%
38% / 60%
7% / 2%
6% / 7%
50% / 64%
6% / 7%
31% / 13%
7% / 9%
7% / 7%
8% / 3%
42% / 61%
7% / 10%
8% / 6%
34% / 59%
5% / 5%
3% / 4%
24% / 14%
4% / 5%
5% / 7%
53% / 63%
6% / 7%
6% / 7%
4% / 2%
5% / 3%
7% / 7%
40% / 18%
2% / 3%
5% / 2%
6% / 7%
6% / 9%
4.00 (3.28, 4.88)
3.42 (3.15, 3.72)
5.48 (4.53, 6.63)
3.58 (3.27, 3.91)
2.81 (1.82, 4.32)
1.52 (1.25, 1.85)
1.00 (0.94, 1.07)
1.00 (0.94, 1.07)
12.16 (7.14, 20.72)
1.25 (1.07, 1.46)
2.87 (2.63, 3.12)
1.59 (1.36, 1.87)
9.47 (8.19, 10.95)
2.60 (2.11, 3.20)
3.20 (2.30, 4.46)
4.72 (3.82, 5.84)
1.65 (1.45, 1.89)
1.68 (1.25, 2.25)
2.16 (1.81, 2.58)
1.00 (0.93, 1.08)
1.77 (1.43, 2.19)
3.12 (2.30, 4.24)
4.64 (3.76, 5.72)
1.50 (1.24, 1.82)
3.19 (2.51, 4.05)
2.06 (1.83, 2.31)
1.54 (1.28, 1.86)
1.36 (1.16, 1.60)
7.06 (4.92, 10.13)
3.52 (2.88, 4.31)
3.60 (2.55, 5.07)
3.65 (3.36, 3.96)
2.20 (1.27, 3.81)
7.70 (4.61, 12.86)
1.72 (1.47, 2.03)
1.77 (1.27, 2.47)
0.08
0.002
<0.0001
1 1.5 2 3 4 5 6
Category Subgroup % or mean
cases/controls
OR (95% CI) P-value
By history of diabetes
By history of hypertension
By family history of MI
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
No
Yes
Yes
Yes
No
No
No
No
Yes
Yes
No
Yes
Odds ratio (95% CI)
OR: Odds ratios; 95% CI: 95% confidence interval. Confidence intervals were plotted using
“floating absolute” variances for tobacco use. Models were adjusted for conventional risk
factors: age, sex, ethnicity, centre of recruitment, LDL-C, WHR, history of diabetes or
hypertension and family history of MI. P-values of interaction are derived from tests of
significance of the interaction terms added to the model. Black boxes correspond to the
reference group (individuals without a medical history of either diabetes, hypertension of family
history of MI); whilst red boxes indicate other groups (individuals with a history of diabetes, a
history of hypertension or a history of MI).
237
Figure 8.6: Association of tobacco use with MI risk according to levels of LDL-C and
Waist to Hip Ratio
.
6% / 3%
43% / 60%
5% / 6%
6% / 7%
6% / 6%
7% / 7%
5% / 5%
7% / 7%
6% / 3%
6% / 7%
3% / 5%
34% / 17%
39% / 58%
8% / 8%
6% / 9%
37% / 18%
5% / 3%
37% / 19%
37% / 16%
6% / 2%
32% / 18%
7% / 9%
9% / 7%
6% / 3%
6% / 3%
6% / 7%
41% / 57%
4% / 4%
37% / 18%
42% / 62%
4% / 6%
4% / 5%
42% / 62%
41% / 62%
5% / 6%
4% / 5%
4.13 (3.07, 5.57)
1.66 (1.47, 1.87)
2.32 (1.72, 3.13)
2.06 (1.59, 2.67)
1.40 (1.08, 1.82)
1.93 (1.52, 2.45)
3.16 (2.38, 4.19)
1.99 (1.57, 2.53)
3.32 (2.43, 4.54)
1.62 (1.26, 2.08)
1.84 (1.29, 2.63)
3.44 (2.99, 3.96)
1.22 (1.11, 1.34)
1.20 (0.94, 1.53)
1.79 (1.38, 2.32)
4.32 (3.81, 4.91)
6.16 (4.24, 8.94)
6.08 (5.24, 7.05)
5.00 (4.39, 5.69)
8.31 (5.73, 12.04)
2.84 (2.45, 3.29)
1.45 (1.15, 1.82)
1.91 (1.50, 2.43)
4.99 (3.66, 6.80)
4.73 (3.44, 6.50)
1.83 (1.43, 2.35)
1.00 (0.89, 1.13)
3.33 (2.41, 4.62)
5.41 (4.62, 6.34)
1.80 (1.61, 2.03)
1.21 (0.89, 1.64)
1.80 (1.32, 2.47)
1.00 (0.89, 1.12)
1.24 (1.13, 1.36)
1.97 (1.48, 2.61)
1.65 (1.23, 2.23)
0.048
0.7
1 1.5 2 3 4 5 6
Category Subgroup % or mean
cases/controls
OR (95% CI) P-value
By thirds of LDL-c
By thirds of Waist to Hip Ratio
Odds ratio (95% CI)
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
OR: Odds ratios; 95% CI: 95% confidence interval. Confidence intervals were plotted using
“floating absolute” variances for tobacco use. Models were adjusted for conventional risk
factors: age, sex, ethnicity, centre of recruitment, LDL-C, WHR, history of diabetes or
hypertension and family history of MI. P-values of interaction are derived from tests of
significance of the interaction terms added to the model. Black boxes correspond to the
reference group (lowest third of LDL-C, lowest third of WHR); whilst blue and red boxes indicate
other groups (middle and top thirds of LDL-C and WHR).
238
Figure 8.7: Association of tobacco use with MI risk by socio-economic status and
diet
≤ median
> median
≤ median
> median
> median
≤ median
≤ median
≤ median
> median
≤ median
> median
> median
6% / 6%
7% / 6%
6% / 7%
7% / 3%
7% / 7%
36% / 19%
6% / 8%
5% / 2%
5% / 6%
4% / 5%
43% / 63%
37% / 17%
4% / 2%
6% / 8%
4% / 4%
42% / 63%
3% / 4%
8% / 3%
38% / 19%
8% / 7%
6% / 5%
40% / 59%
40% / 60%
35% / 17%
1.61 (1.27, 2.04)
2.93 (2.37, 3.62)
1.27 (1.02, 1.58)
2.89 (2.22, 3.76)
1.31 (1.06, 1.62)
4.58 (4.10, 5.12)
1.49 (1.20, 1.84)
5.20 (3.75, 7.23)
1.46 (1.14, 1.87)
1.05 (0.81, 1.38)
1.00 (0.92, 1.09)
3.65 (3.24, 4.10)
3.52 (2.51, 4.94)
0.84 (0.68, 1.03)
1.60 (1.21, 2.12)
1.00 (0.92, 1.09)
2.37 (1.75, 3.20)
6.42 (4.96, 8.31)
3.31 (2.95, 3.70)
1.46 (1.18, 1.79)
1.80 (1.42, 2.28)
1.39 (1.28, 1.52)
0.72 (0.66, 0.79)
2.62 (2.34, 2.94)
0.13
0.56
1 1.5 2 3 4 5 6
Category Subgroup % or mean
cases/controls
OR (95% CI) P-value
By Carbohydrate and vegetable consumption
By protein and meat consumption
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
≤ median
> median
≤ median
> median
> median
≤ median
≤ median
≤ median
> median
≤ median
> median
> median
High
Medium
Low
1% / 1%
39% / 17%
6% / 7%
6% / 9%
6% / 8%
6% / 7%
49% / 67%
34% / 18%
7% / 3%
41% / 60%
7% / 8%
4% / 4%
5% / 3%
6% / 2%
34% / 55%
8% / 8%
35% / 18%
6% / 3%
2.25 (1.28, 3.95)
4.31 (3.78, 4.91)
1.80 (1.39, 2.33)
1.32 (1.04, 1.69)
1.06 (0.82, 1.37)
1.37 (1.06, 1.77)
1.08 (0.98, 1.19)
3.11 (2.72, 3.57)
4.52 (3.29, 6.20)
1.48 (1.34, 1.63)
1.50 (1.18, 1.90)
2.57 (1.87, 3.52)
2.46 (1.77, 3.41)
4.18 (2.93, 5.96)
1.00 (0.90, 1.11)
2.82 (2.24, 3.56)
4.88 (4.27, 5.58)
5.86 (4.25, 8.08)
0.0008
By socio-economic status
Current chew only
Current smoke only
Current smoke & dip/chew
Current dip only
Ex-user
Never user
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
High
Medium
Low
Odds ratio (95% CI)
OR: Odds ratios; 95% CI: 95% confidence interval, P-value: p-value of interaction. Confidence
intervals were plotted using “floating absolute” variances for tobacco use. Models were
adjusted for conventional risk factors: age, sex, ethnicity, centre of recruitment, LDL-C, WHR,
history of diabetes or hypertension and family history of MI. P-values of interaction are derived
from tests of significance of the interaction terms added to the model. To investigate diet,
dietary patterns were created using principal component analysis of 43 food items from a food
frequency questionnaire. The two principal components were chosen after orthogonal rotation
of the matrix of components and labelled respectively “Carbohydrate and vegetable diet” and
“protein and sweet diet”, because of their high loadings for questions rich in protein, sweet,
carbohydrate or vegetable. These patterns were then dichotomised at the median. Black boxes
correspond to the reference group (lowest third of socio-economic status, below median
consumption of a carbohydrate and vegetable diet, below median of a protein and meat diet);
whilst blue and red boxes indicate other groups.
239
Figure 8.8: Association of tobacco use with MI risk by recruitment centre and by
source of controls
NOTE: Weights are from random effects analysis
.
.
.
.
Ex versus never tobacco users
Smokeless tobacco only versus never use
Smoking tobacco only versus never user
Smoking & smokeless tobacco versus never use
Tobacco use
/ Centre Source of controls
0.30 (0.08, 1.12)
0.96 (0.57, 1.60)
1.47 (1.10, 1.98)
2.39 (1.31, 4.35)
1.39 (0.65, 2.96)
1.76 (1.19, 2.61)
1.00 (0.76, 1.33)
1.18 (0.63, 2.20)
1.06 (0.71, 1.60)
1.91 (0.38, 9.54)
0.52 (0.06, 4.71)
1.14 (0.31, 4.15)
1.30 (0.74, 2.30)
1.17 (0.89, 1.54)
5.32 (2.06, 13.72)
2.04 (1.26, 3.32)
2.38 (1.83, 3.08)
1.68 (1.39, 2.04)
1.30 (0.59, 2.85)
2.10 (1.15, 3.85)
1.47 (0.52, 4.13)
1.05 (0.24, 4.47)
3.07 (1.41, 6.70)
3.03 (1.79, 5.12)
2.88 (2.24, 3.70)
4.42 (3.08, 6.34)
4.64 (2.78, 7.74)
3.60 (2.83, 4.58)
3.27 (2.73, 3.91)
3.04 (2.16, 4.28)
3.81 (3.00, 4.83)
2.84 (1.88, 4.28)
3.25 (1.73, 6.12)
0.80 (0.15, 4.36)
2.99 (1.13, 7.94)
3.84 (2.38, 6.18)
5.18 (2.02, 13.26)
4.68 (1.94, 11.28)
11.45 (5.66, 23.16)
3.22 (2.38, 4.36)
3.00 (1.09, 8.24)
3.43 (1.72, 6.84)
7.66 (1.99, 29.40)
3.31 (0.61, 17.99)
OR (95% CI)
2.53
10.37
16.95
8.63
6.26
13.67
17.46
8.19
13.27
1.73
0.95
2.65
9.12
16.50
4.47
10.82
17.00
18.96
5.99
8.45
3.87
2.16
1.42
3.13
13.76
6.66
3.31
14.78
26.94
7.38
15.33
5.13
2.16
2.79
7.05
16.48
7.46
8.22
11.01
21.99
6.69
11.28
4.21
2.81
Weight (%)
1 29.4
DMIC
KIHD
KIHD
MIC
NICVD
NICVD
NICVD
PIC
PIC
RCH
RCH
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors OPD patients
Visitors MI patients
OPD patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Subtotal (I-Squared = 37.5%, p=0.099)
Subtotal (I-Squared = 0%, p=0.64)
Subtotal (I-Squared = 54.8%, p=0.01)
Subtotal (I-Squared = 44.5%, p=0.059)
DMIC
KIHD
KIHD
MIC
NICVD
NICVD
NICVD
PIC
PIC
RCH
RCH
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors OPD patients
Visitors MI patients
OPD patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
DMIC
KIHD
KIHD
MIC
NICVD
NICVD
NICVD
PIC
PIC
RCH
RCH
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors OPD patients
Visitors MI patients
OPD patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
DMIC
KIHD
KIHD
MIC
NICVD
NICVD
NICVD
PIC
PIC
RCH
RCH
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors OPD patients
Visitors MI patients
OPD patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
Visitors MI patients
Visitors OPD patients
OR (95% CI)
OPD: Outpatients Department. Models adjusted for age, sex, ethnicity, LDL-C, Waist to hip
ratio, history of diabetes or hypertension and family history of MI. ORswere computed within
centres and for each type of controls versus all cases of the centre separately, when at least
100 controls for a specific type were available. DMIC: Deewan Mushtaq Institute of Cardiology;
FIC: Faisalabad Institute of Cardiology; KIHD: Karachi Institute of Heart Diseases; MIC: Multan
Institute of Heart Diseases; NICVD: National Institute of Cardiovascular Diseases; PIC: Punjab
Institute of Cardiovascular Disease; RCH: Red Crescent Hospital; MI: Myocardial infarction.
240
Figure 8.9: Association of tobacco use with MI risk by recruitment centre with fixed
and random effect meta-analyses
.
.
.
Ex versus never tobacco users
KIHD
MIC
NICVD
PIC
Random effects pooled estimate
Fixed effects pooled estimate
One-step estimate
One step estimate
One step estimate
Once step estimate
Smokeless tobacco only versus never users
KIHD
NICVD
PIC
Random effects pooled estimate
Fixed effects pooled estimate
Smoking tobacco only versus never user
DMIC
KIHD
MIC
NICVD
PIC
RCH
Random effects pooled estimate
Fixed effects pooled estimate
Smoking & smokeless tobacco versus never use
KIHD
NICVD
Random effects pooled estimate
Fixed effects pooled estimate
Tobacco use
/ Centre
125
46
127
75
131
404
43
47
316
261
772
600
182
77
192
No case
159
28
225
76
189
537
31
34
188
89
549
263
96
33
116
No controls
1.36 (1.02, 1.80)
2.14 (1.20, 3.80)
1.21 (0.94, 1.56)
1.11 (0.76, 1.64)
1.31 (1.08, 1.58)
1.30 (1.10, 1.52)
1.22 (1.05, 1.43)
1.61 (1.42, 1.83)
3.37 (3.06, 3.72)
3.92 (3.22, 4.77)
1.18 (0.90, 1.54)
1.71 (1.45, 2.01)
1.97 (1.14, 3.37)
1.53 (1.14, 2.05)
1.57 (1.37, 1.80)
3.07 (1.41, 6.70)
2.84 (2.23, 3.61)
4.29 (3.01, 6.11)
3.44 (2.95, 4.01)
3.69 (2.96, 4.59)
2.86 (1.95, 4.19)
3.38 (3.04, 3.76)
3.38 (3.06, 3.75)
3.63 (2.32, 5.69)
3.94 (3.02, 5.14)
3.86 (3.07, 4.85)
3.86 (3.07, 4.85)
OR (95% CI)
32.35
9.89
37.68
20.08
100.00
36.30
44.97
18.73
100.00
1.80
18.31
8.59
41.97
21.94
7.39
100.00
25.93
74.07
100.00
Weight*
1 6.7
(Heterogeneity: I-squared = 21.6%, p = 0.281)
(Heterogeneity: (I-squared = 67.1%, p = 0.048)
(Heterogeneity: I-squared = 3.5%, p = 0.394)
(Heterogeneity: I-squared = 0.0%, p = 0.756)
Odds ratio (95% CI)
*Weights are from random effects meta-analysis. Models adjusted for age, sex, ethnicity, LDL-
C, Waist to hip ratio, history of diabetes or hypertension and family history of MI. ORswere
computed within centres, when at least 100 controls for a specific type were available. Random
effects pooled estimate was obtained using an estimate of heterogeneity being taken from the
inverse-variance fixed-effect model. The one-step estimate corresponds to ORs adjusting for
conventional risk factors and centre of recruitment.
241
References
1. Mackay J, Mensah G, World Health Organization. The Atlas of heart disease
and stroke. 2004.
2. Nishtar S. Prevention of coronary heart disease in south Asia. Lancet.
2002;360:1015-1018.
3. World Bank. Life expectancy at birth in Pakistan. 28-7-2011.
4. U.S.Department of State. Background Note: Pakistan. 6-10-2011.
5. Anand SS, Yusuf S, Vuksan V et al. Differences in risk factors,
atherosclerosis, and cardiovascular disease between ethnic groups in Canada: the
Study of Health Assessment and Risk in Ethnic groups (SHARE). Lancet.
2000;356:279-284.
6. Balarajan R. Ethnic differences in mortality from ischaemic heart disease
and cerebrovascular disease in England and Wales. BMJ. 1991;302:560-564.
7. Forouhi NG, Sattar N, Tillin T, McKeigue PM, Chaturvedi N. Do known risk
factors explain the higher coronary heart disease mortality in South Asian
compared with European men? Prospective follow-up of the Southall and Brent
studies, UK. Diabetologia. 2006;49:2580-2588.
8. Joshi P, Islam S, Pais P et al. Risk factors for early myocardial infarction in
South Asians compared with individuals in other countries. JAMA. 2007;297:286-
294.
9. Bhopal R. Epidemic of cardiovascular disease in South Asians. BMJ.
2002;324:625-626.
10. Ghaffar A, Reddy KS, Singhi M. Burden of non-communicable diseases in
South Asia. BMJ. 2004;328:807-810.
11. Crofton J. The gathering smoke clouds: a worldwide challenge. Int J
Epidemiol. 1984;13:269-270.
12. Guindon EG, Boisclair D. Past, current and future trends in tobacco use.
Economics of Tobacco Control. 2003;1-45.
13. Maziak W. The global epidemic of waterpipe smoking. Addict Behav.
2011;36:1-5.
14. Mazahir S, Malik R, Maqsood M et al. Socio-demographic correlates of betel,
areca and smokeless tobacco use as a high risk behavior for head and neck
cancers in a squatter settlement of Karachi, Pakistan. Subst Abuse Treat Prev
Policy. 2006;1:10.
15. Ismail J, Jafar TH, Jafary FH, White F, Faruqui AM, Chaturvedi N. Risk
factors for non-fatal myocardial infarction in young South Asian adults. Heart.
2004;90:259-263.
16. Rahman MA, Zaman MM. Smoking and smokeless tobacco consumption:
possible risk factors for coronary heart disease among young patients attending a
tertiary care cardiac hospital in Bangladesh. Public Health. 2008;122:1331-1338.
17. Pais P, Fay MP, Yusuf S. Increased risk of acute myocardial infarction
associated with beedi and cigarette smoking in Indians: final report on tobacco
risks from a case-control study. Indian Heart J. 2001;53:731-735.
242
18. Pais P, Pogue J, Gerstein H et al. Risk factors for acute myocardial infarction
in Indians: a case-control study. Lancet. 1996;348:358-363.
19. Forouhi NG, Sattar N, Tillin T, McKeigue PM, Chaturvedi N. Do known risk
factors explain the higher coronary heart disease mortality in South Asian
compared with European men? Prospective follow-up of the Southall and Brent
studies, UK. Diabetologia. 2006;49:2580-2588.
20. Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial
infarction in 52 countries in the INTERHEART study: a case-control study. Lancet.
2006;19;368:647-658.
21. Iqbal R, Anand S, Ounpuu S et al. Dietary patterns and the risk of acute
myocardial infarction in 52 countries: results of the INTERHEART study.
Circulation. 2008;118:1929-1937.
22. Rosengren A, Subramanian SV, Islam S et al. Education and risk for acute
myocardial infarction in 52 high, middle and low-income countries: INTERHEART
case-control study. Heart. 2009;95:2014-2022.
23. Boffetta P, Straif K. Use of smokeless tobacco and risk of myocardial
infarction and stroke: systematic review with meta-analysis. BMJ. 2009;339.
24. Yatsuya H, Folsom AR. Risk of incident cardiovascular disease among users
of smokeless tobacco in the Atherosclerosis Risk in Communities (ARIC) study.
Am J Epidemiol. 2010;172:600-605.
25. Boffetta P, Hecht S, Gray N, Gupta P, Straif K. Smokeless tobacco and
cancer. Lancet Oncol. 2008;9:667-675.
26. World Health Organization. Tobacco: deadly in any form or disguise. 2006.
27. Singh RB, Niaz MA, Ghosh S et al. Association of trans fatty acids (vegetable
ghee) and clarified butter (Indian ghee) intake with higher risk of coronary artery
disease in rural and urban populations with low fat consumption. Int J Cardiol.
1996;56:289-298.
28. Easton DF, Peto J, Babiker AG. Floating absolute risk: an alternative to
relative risk in survival and case-control analysis avoiding an arbitrary reference
group. Stat Med. 1991;10:1025-1035.
29. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis.
Stat Med. 2002;21:1539-1558.
30. The World Bank. The World Bank: Pakistan.
http://data.worldbank.org/country/pakistan accessed 04/07/2012 . 2012.
31. The World Bank. Regional Data: South Asia. 2010.
32. Alam AY, Iqbal A, Mohamud KB, Laporte RE, Ahmed A, Nishtar S.
Investigating socio-economic-demographic determinants of tobacco use in
Rawalpindi, Pakistan. BMC Public Health. 2008;8:50.:50.
33. Nasir K, Rehan N. Epidemiology of cigarette smoking in Pakistan. Addiction.
2001;96:1847-1854.
34. World Health Organization. WHO report on the global tobacco epidemic; the
MPOWER package. 2008.
35. Dobe M, Sinha DN, Rahman K. Smokeless tobacco use and its implications
in WHO South East Asia Region. Indian J Public Health. 2006;50:70-75.
243
36. Dhirendra SN. Report on oral tobacco use and its implications in South East
Asia. Patna, India: 2004.
37. Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial
infarction in 52 countries in the INTERHEART study: a case-control study. Lancet.
2006;%19;368:647-658.
38. Khawaja MR, Mazahir S, Majeed A et al. Chewing of betel, areca and
tobacco: perceptions and knowledge regarding their role in head and neck
cancers in an urban squatter settlement in Pakistan. Asian Pac J Cancer Prev.
2006;7:95-100.
39. Gupta PC, Ray CS. Smokeless tobacco and health in India and South Asia.
Respirology. 2003;8:419-431.
40. Foulds J, Ramstrom L, Burke M, Fagerstrom K. Effect of smokeless tobacco
(snus) on smoking and public health in Sweden. Tob Control. 2003;12:349-359.
41. Hansson J, Galanti MR, Hergens MP et al. Use of snus and acute myocardial
infarction: pooled analysis of eight prospective observational studies. Eur J
Epidemiol. 2012.
42. Tomar SL. Is use of smokeless tobacco a risk factor for cigarette smoking?
The U.S. experience. Nicotine Tob Res. 2003;5:561-569.
43. Bolinder GM, Ahlborg BO, Lindell JH. Use of smokeless tobacco: blood
pressure elevation and other health hazards found in a large-scale population
survey. J Intern Med. 1992;232:327-334.
44. Oswal KC, Raute LJ, Pednekar MS, Gupta PC. Are current tobacco pictorial
warnings in India effective? Asian Pac J Cancer Prev. 2011;12:121-124.
45. Government of Pakistan and Ministry of Health. Statutory Notifications
containing Rules and Orders issued by all Ministries and Divisions of the
Government of Pakistan and their Attached and Subordinate Offices and the
Supreme Court of Pakistan. M-302. 23-10-2009.
46. IRIN. Pakistan: anti-smoking law have little impact. IRIN Humanitarian
news and analysis. 2004.
244
Chapter 9: Discussion
Summary
Tobacco use remains the main preventable cause of death and accounts for 1 in 10
CVD events. Worldwide, smoking prevalence is highest in developed countries at
around one fourth of the population, but, when considering only men, developing
countries are at the top of the list in terms of both smoking and smokeless tobacco
rates. In the ERFC, prevalence of tobacco use was higher than current estimates in
the developed world and there was evidence of a gap between men and women. In
PROMIS controls, tobacco prevalence was relatively low compared to previous
reports. All forms of tobacco use investigated in this thesis, namely cigarettes, pipes,
cigars, and South Asian chewing and dipping tobacco, were associated with a
relatively strong increase in CVD risk. These findings lend more weight to public
health campaigns which address all forms of tobacco rather than cigarette smoking
alone, and limit the scope for safe alternatives to cigarette smoking. Future work
includes harvesting genetic information already available in PROMIS to investigate
genetic determinants of tobacco dependence and genetic susceptibility to CVD
amongst tobacco users. In the longer term, additional studies on smoking and
smokeless tobacco using a prospective design would help strengthen the evidence
of a causal relationship between all forms of tobacco use and CVD and help
disentangle the source of the additional CVD risk experienced by tobacco users.
245
9.1 Introduction
CVDs account for 30% of all deaths and 20% of global Disability Adjusted Life Years
(DALYs) in individuals aged 30 years and older, being the leading single cause of
death and disability worldwide; and this dominance is set to increase as populations
are ageing worldwide 1. Tobacco use remains the main preventable cause of CVD
and death. By 2030, if current trends continue, smoking will be killing 9 million people
annually 2. Most tobacco users now live in developing countries and South Asia
accounts for a large proportion of them. India itself is thought to contain more than
10% of worldwide smokers 2. At the recent high-level meeting of the United Nations
on non-communicable diseases, in September 2011, delegates asserted that “the
increasing global crisis in non-communicable diseases is a barrier to development
goals including poverty reduction, health equity, economic stability, and human
security … The most urgent and immediate priority is tobacco control.” 3
It has been estimated that reducing death rates from chronic diseases (which include
CVD, cancers, chronic respiratory diseases and diabetes) by 2% from 2006 to 2015
would avert 36 million deaths, of which 28 million lives would be saved in low-income
and middle-income countries 4. Almost half of these averted deaths would be in men
and women younger than 70 years old. The experience of high income countries
shows what can be achieved with sustained interventions, in particular sensitizations
to the dangers of smoking. Over the past three decades, the chronic diseases death
rate has been reduced by between 1% and 3% every year in the developed world,
translating into a fall of up to 70% in the death rate of CHD in some countries
(Australia, Canada, Japan, the UK, and the USA). During this period, the number of
CVD deaths averted has been estimated to be 14 million in the USA, 8 million in
Japan and 3 million in the UK. Reasons for these decreases remain partially
understood but have been largely attributed to the halving in smoking prevalence
which happened during the same period in these countries.
In this thesis, I have investigated in more detail than previously possible the
relationship between tobacco use and the risk of CVD. In particular, I have
strengthened the evidence concerning pipe and cigar smoking in relation to MI and
stroke using a prospective design, and showed a relatively strong association
between South Asian chewing and dipping tobacco with MI risk using a case-control
design. I have also investigated in more detail than previous large scale studies or
meta-analyses have been able to regarding the relationship between cigarette
246
smoking and CVD risk. In this Discussion, I review my main findings, their public
health relevance and envision future work.
9.2 Tobacco prevalence in developed and developing countries
9.2.1 Smoking tobacco products
The developed world has been successfully implementing strategies to reduce
smoking since the 1960’s, and this reflects in prospective studies included in the
ERFC. Early studies started in the 1960’s and 1970’s reported a higher proportion of
baseline current smokers, especially amongst men, than later studies which began
enrolment in the 1990’s and 2000’s and had prevalence compatible with estimates of
20-25% in both men and women. In developing countries, and particularly in South
Asia, the picture is vastly different. Firstly, the gap between men and women in
smoking prevalence remains present with a much higher prevalence in men.
Secondly, the prevalence of smoking is still high and rising with time. The share of
cigarette production and consumption of developing was 40% of worldwide
production in 1970, whilst it is over 70% nowadays 2. In Pakistan, the production and
consumption of cigarettes has doubled since 1970 5. Thirdly, cigarettes coexist with
other popular smoking and smokeless products. In the PROMIS controls, nearly a
quarter of controls smoked, and more than 1 in 10 used smokeless products.
Smoking was associated with the use of ghee, a traditional cooking medium which is
cheaper than oil and has been shown to be associated with increased risk of CVD,
demonstrating a clustering of harmful habits in tobacco users 6. All forms of tobacco
use were inversely associated with socio-economic status, indicating that it is mainly
the poorest section of the population that is attempting to emulate the bad habits of
the West.
9.2.2 Smokeless tobacco products
The use of traditional smokeless tobacco products appears stable in the USA at
around 3% of the population, representing 8.1 million individuals 7, 8. Smokeless
tobacco had fallen out of fashion in the Western world during the 1st half of the 20th
century but there has been a resurgence since the 1970’s as smokeless tobacco
was rebranded as a new, cheap and trendy alternative to cigarettes 9. An example
includes dipping snuff which does not require spitting: It is called snus and is already
popular in Northern Europe, especially Sweden and Norway. It has variants in the
US with higher nitrosamine and carcinogenic content than the Swedish product;
while it remains forbidden in other countries of the European Union 10. With
globalization, there is a risk of developed countries importing products already
247
popular in developing countries. Migrations of population from developing to
developed countries, the internet, and tourism make these products more accessible
and more visible 11. As snus has becomes the dominant form of tobacco used by
Swedish men 12, this popularity may serve to promote other dipping products such as
the South Asian version called naswar. By contrast, betel quid chewing and dipping
have been part of local South Asian cultures for centuries, even millennia, and
tobacco has been added to chewing and dipping mixtures in the 17th and 18th century
at the time of colonization. Smokeless tobacco was traditionally hand prepared, but
over the past half century, industrialized pouches such as gutka in Pakistan have
been developed and have been so successful that they have overridden traditional
products (such as paan and supari). In India, there is evidence that per capita
smokeless tobacco consumption has increased among the poor since the 1960s in
both rural and urban areas, and the total number of smokeless tobacco users has
been estimated at around 100 million people 13, 14. In the PROMIS controls,
smokeless tobacco use was correlated with traditional Pakistani diet high in
carbohydrates and vegetables and was especially favoured by women. Dipping
tobacco, called naswar, was mainly favoured by older individuals and lower socio-
economic groups.
9.3 Association of tobacco with cardiovascular risk
9.3.1 Smoking tobacco use and CVD risk
Cigarette smoking caused an increased risk of CVD even at low intensities. The risk
doubled from 0 to 5 cigarettes per day; and, to double again, a smoker had to
increase their intake from 5 to more than 25 cigarettes per day. Inversely, stopping
smoking was rapidly associated with a reduced risk of CVD, and ex-smokers had
CVD risk generally below 15%. This would indicate that most of the effect of smoking
on CVD is relatively short-term and reversible, promoting plaque rupture and
thrombosis. In that respect, cigarette smoking has been shown to increase
myocardial workload, reduce oxygen-carrying capacity of the blood, cause coronary
vasoconstriction, increase catecholamine release and induce a hypercoagulable
state 9. However, risk for ex-smokers did not go back to that of never smokers until
20 years after cessation, which means that some components of smoking durably
affect vessel walls, and are probably involved in the early processes of
atherosclerosis and in plaque building by promoting degradation of the vessel wall
and lipid uptake by macrophages.
248
Smoking cigarettes was associated with a doubling in risk of MI in the ERFC and a
tripling in risk of MI in the PROMIS participants. It has been suggested that South
Asians are more susceptible to traditional CVD risk factors and that this could
account for the fact that they experience CVD on average 10 years earlier than
Western individuals 18. However, the difference I observed in risk ratios for smoking
could also be an artefact resulting from the difference in study designs. Indeed, the
ERFC is a collection of prospective studies and allowed estimation of hazard ratios,
whilst PROMIS used a case-control design and only allowed estimation of odds
ratios, which are prone to over-estimate risk ratios, and are subject to recall bias 19.
Studies included in the ERFC mostly started enrolment before the 1980’s, when
smoking prevalence was high and therefore rates of passive smoking in never
smokers were probably high as well, with few legislations at the time to limit its
impact. Smoke-free policies, even if they are becoming more common 2, are still
rarely enforced in developing countries. In Pakistan, legislation on tobacco use is not
implemented 20 and rates of passive smoking in non-tobacco users are likely to be
elevated as a result. Passive smoking has been shown to increase the risk in non-
smokers by up to a factor of 3.7 21. Therefore, if passive smoking was to exert its
effect mainly in never smokers as has been shown previously, ERFC and PROMIS
estimates could both be conservative and the true increase in risk caused by
smoking and smokeless tobacco use may be even higher.
The association between pipe or cigar smoking with CVD was strong and
intermediate between that of never smokers and that of cigarette smokers, in
agreement with previous findings 15-17. RRs for all CVD events were 1.31 (1.19; 1.44)
for cigar only smokers, 1.68 (1.56; 1.81) for pipe only smokers; and 1.97 (for
cigarette only users). At high intensities (≥15 cigarettes equivalent per day), pipes or 
cigars experienced close to a doubling in risk: RR was 1.86 (1.54; 2.25). By contrast,
cigarette smokers smoking 15-20 cigarettes per day had a RR of 2.50 (2.34; 2.66),
and >25 cigarettes per day a RR of 2.93 (2.63; 3.26). Smoke from pipes and cigars
contains the same toxic substance as cigarette smoke and individuals who switched
from cigarettes to pipes or cigars are more likely to inhale tobacco than never
cigarettes users 15. In this context, the reason that cigarette smoking caused a
greater increase in the risk of CVD compared to pipes and cigars at similar
intensities remains unclear and would need to be further investigated in studies
where full smoking history is available for all smoking types, including doses and
duration.
249
The investigation of interplay between smoking and other risk factors on CVD risk
showed that individuals with lower absolute risk, such as younger age groups,
women, non-diabetics and non-hypertensives had a higher increase in risk when
they smoked than older age groups, men, diabetics and hypertensive individuals.
These statistically significant synergies remain difficult to interpret. They could be
attributed to a relative absence of competing risks in individuals with lower baseline
risk. It is a clear public health message for women, non-diabetics, non-hypertensive
and younger age groups that the added impact of smoking may counterbalance their
relatively better CVD prospect than groups with higher baseline risk such as men,
older individuals, hypertensive and diabetics. Nevertheless, this interaction on a
relative scale did not translate into a substantially higher burden of smoking related
CVD in these low risk groups compared to the higher risk groups. In absolute terms,
more CVD events were still caused by smoking amongst older age groups, men,
diabetics and hypertensive. The interactions with body mass index and total
cholesterol were non-significant.
9.3.2 Smokeless tobacco use and CVD risk
The effect of South Asian smokeless tobacco was significant and intermediate
between that of a never tobacco user and that of a current smoker of cigarettes or
bidis in PROMIS participants. These estimates were more precise and broadly
comparable to the published literature 22. ORs for myocardial infarction were 1.31
(1.39-1.88) with current chewing of tobacco, and 1.35 (1.12-1.63) with current snuff
dipping, compared to never tobacco users. Individuals who chewed or dipped
tobacco on top of smoking seemed to be at higher risk than smokers alone (3.91
versus 3.36), even if the difference was not statistically significant.
My results regarding chewable tobacco were in broad agreement with a worldwide
case-control study 22 as well as with recent meta-analyses of the effect of
smokeless tobacco in Western population where most of the evidence came from a
Northern American study on chewable rather than snuff tobacco 23. Regarding
tobacco dipping, my results differ from a recent meta-analysis selecting only Swedish
data which found a non-significant increase in MI risk for snuff users 24 and has led
some to argue that Swedish snuff called snus is an acceptable alternative to smoking
25, 26. However, Pakistani snuff (naswar) and Swedish snus have different processes
of fabrication and Swedish snus is especially low in nitrosamines 26. Therefore
caution should remain when generalizing results from Swedish studies to other
populations and promoting liberalization of snuff dipping outside of Northern Europe.
250
As pointed out by Last in his Dictionary of Epidemiology, the presence of a dose-
response relationship is another argument in favour of a causal relationship between
smokeless tobacco and MI 27. There was evidence in the PROMIS data of a dose-
response relationship below 10 smokeless products a day. Individuals consuming 5-
10 doses a day had an OR of 1.77 (1.43; 2.20) risk of MI, which is a considerable
increase in risk. The increased risk of a similar amount of cigarettes was 2.56 (2.07;
3.16).
These findings may help better understand the aetiology of tobacco use in relation to
CVD and suggest a role for toxins that are intrinsic to tobacco itself. The
pathogenesis of CVD as a result of smokeless tobacco use is not well understood.
The proportion of nicotine is very variable and chemical composition differs between
products. Unlike smoking, which produces rapid peaks and troughs, smokeless
tobacco use causes more prolonged, sustained levels of nicotine, often lasting for 1
hour 9. The rate of absorption may vary among different forms of smokeless tobacco
depending on the pH level of the product, amount of nicotine and size of the tobacco
cutting. When testing a range of smokeless tobacco products, venous concentrations
have been shown to be higher than for cigarette smoking 28. In addition to nicotine,
smokeless tobacco has been shown to contain nitrosamines, nitrosamine acids,
polycyclic aromatic hydrocarbons, aldehydes and metals, whose effects are known
to be cancerous and remain unknown relative to CVD 29. Blood pressure levels are
thought to be affected by the high sodium content of smokeless tobacco as well as
by nicotine, and acute cardiovascular effects, similar to those caused by cigarette
smoking, are seen with the use of smokeless tobacco such as an increase in heart
rate and blood pressure levels which is strong enough to activate the sympathetic
nervous system 9. In particular, chewing betel quid has been shown to increase
blood pressure and the chance of being hypertensive 29.
9.4 Public health relevance of my findings
Cigarettes account for 96% of global sales of manufactured tobacco by value, and
global cigarette production continues to increase dramatically 2, 30. Around a billion
people are addicted to nicotine in the form of cigarettes and many have no
immediate plans to quit. The concept of “harm reduction” has been coined to refer to
the objective of minimising the net damage to health of continuing tobacco users and
the general population by substituting less harmful tobacco products for cigarettes,
for example other smoking products such as pipes or cigars or even smokeless
251
tobacco 15, 30. I explore below whether, in light of my findings, this concept is
applicable to either developed or developing countries.
Smokeless tobacco as a “harm reduction” tool, and snus in particular for which
evidence regarding CVD risk remains inconclusive, has been advocated in
developed countries 31. In Sweden, consumption of snus is thought to play a role in
the relatively low prevalence of smoking and low rates of MI compared to other
European countries 12. However, extending the experience to other developed
countries remains controversial 32. Another argument against the use of snus, which
is also valid for cigars and pipes, includes the many unintended consequences that
promoting snus as a substitute to cigarettes could have Individuals that never
intended to smoke cigarettes could initiate tobacco use.. Some individuals could only
partially rather than entirely replace cigarette smoking with use of another tobacco
product. In the ERFC, individuals who smoked pipes or cigars on top of cigarettes
had a RR for CVD of 1.95 (1.81; 2.10), nearly equivalent to that of cigarette only
users who experienced a RR of 1.97 (1.85; 2.11). Individuals who initiate use of
cigars, pipes or smokeless tobacco may also be tempted to later switch to cigarettes.
Snus and cigarettes are sold by the same companies 33 and cigars are increasingly
made to look like cigarettes in shape and format 17. In the US, young men who were
not smokers but regularly used smokeless tobacco were more than three times as
likely as never users to have become current smokers four years later 34. Cigar and
pipe smoking, which are very similar forms of tobacco use compared to cigarette
smoking, could have even higher switch rates. In this context, the steady growth of
cigar35 and smokeless tobacco use9 over the past decades in the US is a source of
concern, and I believe that the best public health policy in developed countries
remains to encourage quitting all forms of tobacco use. One example which I was not
able to investigate in my thesis because of the lack of data available, and soon to
compete in prevalence with pipes and cigars in developed countries, is the e-
cigarette. E-cigarettes are the fastest growing smokeless tobacco industry in
developed countries: the industry is already worth £150m in the UK and £520m in
the US 36. The e-cigarette is an electronic nicotine delivery device 37 promoted as a
safe alternative to cigarette smoking which produces no passive smoking 38. It is not
legislated in EU countries and is sometimes used on pharmacy counters as a
smoking cessation product. Evidence regarding the effect of e-cigarettes on health is
unknown and it is questionable whether it should belong in the category of nicotine
replacement products or is a novel form of smokeless tobacco.
252
In developing countries, and in particular in South Asia, smoking is not yet the
dominant culturally accepted form of tobacco use. In my analysis of urban
Pakistanis, smoking cigarettes coexisted with bidies smoking and a large panel of
smokeless products (paan, gutka, supari and naswar). Smoking was preferred by
men whilst smokeless tobacco was particularly favoured by women and younger
individuals. All these forms of tobacco use were significantly associated with the risk
of MI. Current users of both smoking and smokeless products experienced a risk
ratio for non-fatal MI higher than smokers alone; and their risk was nearly four times
the risk of a never tobacco user. In this context, applying different public health
policies to smoking and smokeless tobacco and encouraging smokers to switch to
smokeless products as advocated by the “harm reduction” approach would have
several harmful effects which may counter any beneficial impact 39. Firstly, in
developing countries, tobacco control is in its infancy 24, and this policy risks
damaging educational effort teaching that tobacco is harmful to cardiovascular
health. In the PROMIS participants, only 9% of men and 2% of women were past
tobacco users, indicating that most smokers have probably never attempted to quit.
In this context, attempting to quit, rather than switching to other tobacco products,
needs to remain the first step for someone willing to curtail his or her CVD risk.
Secondly, it would leave out women who are already using smokeless tobacco.
Thirdly, there is evidence that smokeless tobacco is a gateway to smoking. In South
Asia, smokeless products are particularly favoured by children who later on will start
smoking in teenagehood and adulthood 14. Fourthly, smokeless products in
developing countries have an unknown toxicity. In my study, in contrast to findings
regarding snus use in developed countries, snuff dipping in the form of naswar was
significantly associated with MI risk. In the absence of regulatory control that can
successfully address the toxicities introduced by the practices of small local
producers, any shift from smoking to smokeless tobacco use is likely to be to the
indigenous forms of high toxicity rather than to the Swedish-style products
manufactured to have low toxicity 39.
For all the reasons listed above, the findings of this thesis encourage a public health
policy addressing all forms of tobacco use, rather than cigarettes alone. The
Framework Convention on Tobacco Control (FCTC) was adopted in 2003 by 170
countries members of WHO 40. It advocates a world free of all forms of tobacco use
with a target prevalence <5% worldwide. I have shown that CVD is the cause of
around 1/3 of smoking deaths in middle-age Western individuals. In the ERFC, the
CVD risk of cigarette smoking was rapidly reversible in past smokers, with an 80%
reduction in risk over the first 5 years while the decrease was much slower for the
253
risk of lung cancer. It is plausible that stopping using smokeless tobacco will also
rapidly result in a decreased risk. Among PROMIS participants, past users of all
tobacco products had a non-significant increased risk of MI. In this context, WHO
predicts that full implementation of the FCTC would avert 5.5 million deaths over 10
years in 23 low-income and middle-income countries with a high burden of non-
communicable diseases, including Pakistan 3. Most of the deaths averted would be
because of a reduction in the burden of CVD rather than cancer or other non-
communicable diseases 3.
9.5 Future work
This thesis contains several shortcomings which could be addressed either using the
dataset already available, or in the longer term by creating new appropriate
resources.
Firstly, case-control studies are unable to establish causality, and my estimates are
subject to recall bias 27. To prove causality of smokeless tobacco on CVD risk
beyond doubt, a randomized control trial would be needed, but would prove ethically
challenging, given the evidence established in this thesis. A prospective study with a
long follow-up and set in a developing country such as Pakistan where the
prevalence of these practices is relatively high (>1 in 10 men and women used
smokeless tobacco in PROMIS controls) would provide invaluable information on the
relationship between smokeless tobacco and cardiovascular outcomes. Information
on cotinine levels, a biomarker of nicotine ingested and therefore of smoking
intensity, and tobacco and other contents of the smokeless product would also be
useful as they would enable the analyst to address the question of dose-response
relationship and help unravel aetiological mechanisms by which smokeless tobacco
leads to CVD, in particular whether nicotine plays a role in inducing CVD. It could
also help inform strategies regarding other nicotine products such as e-cigarettes.
Secondly, genetic information from a genome wide association study (GWAS) done
using Illumina 660 Quad array is available in PROMIS participants. This could allow
me (1) to investigate genetic determinants of smoking and smokeless tobacco use,
and (2) to test whether some tobacco users have a genetic predisposition to CVD.
Before the development of high-throughput genotyping platforms, candidate gene
studies investigated whether targeted genes, selected because of their biological
relevance to the atherosclerotic process, had an effect on CVD and whether this
effect depended on smoking status. Candidate gene studies have generally been
254
small, underpowered and poorly replicated. Genes found to interact with smoking
status on the risk of CVD include the C-allele of CYP1A1, an enzyme present in the
lungs that is known to activate smoke carcinogens 41; eNOS4a, an uncommon
polymorphism of the endothelial enzyme eNOS involved in the synthesis of nitric
oxide 42; the p53 protein, a transcription factor that suppresses growth and triggers
apoptosis 43; the glycoprotein IIIa P1(A2) polymorphism 44; the factor V Arg506 Gln
mutation known to reduce the anticoagulation effect of activated protein C;
glutathione S-transferases M1 and T1 45; PON1192Arg polymorphism in the
paroxanase gene; and the L-gene promoter polymorphism of the serotonin
transporter gene. More recently, GWASs have identified single nucleotide
polymorphisms (SNPs) associated with addiction, initiation and cessation of smoking
(Table 9.1). The most highly associated locus with number of cigarettes per day has
been rs1051730, located on chromosome 15q25 and the overlapping three genes
CHRNA5/CHRNA3/CHRNA4 known to encode neuronal nicotinic acetylcholine
receptor subunits 46-49. This locus has also been associated with lung cancer and
peripheral vascular diseases but not yet with CVD. Complementary to GWASs of
DNA variation, large scale epigenetic studies have identified a locus cg03636183
located in F2RL3 with altered methylation patterns in current versus never smokers
50. This locus happens to be lying in a gene coding for a potential drug target of
cardiovascular importance. An analysis of PROMIS participants, with replication of
my findings in other populations to enhance power, could yield novel genetic markers
of smoking and smokeless tobacco dependence and help us understand how
tobacco use causes CVD in the context of a South Asian population.
Thirdly, a source of passive smoking other than “being in the presence of a smoker”
has been recently introduced as “third-hand smoking” and relates to bio-persistent
cigarette smoke residue 51. Nicotine has been shown to persist in ceiling tiles for up
to 30 years 52. Particulates from cigarette smoke deposit on indoor surfaces and can
have toxic effect: for example nicotine residue can react with ambient nitrous acid to
produce carcinogenic compounds. The effect of third-hand smoking on health and, in
particular, on the risk of CVD remains unknown. It is also unknown whether other
forms of tobacco use, such as chewing and dipping also produce residue (for
example on the table where they are prepared, often by children) which persist for a
long time and whether these would have an effect on health. Investigating the effect
of third-hand smoking would require first defining more precisely the exposure of
interest and second devising reliable methods to measure it on a large scale before
being able to assess its effect on population health.
255
9.7 Conclusion
Tobacco use remains a danger to cardiovascular health. Cigarette smoking is the
most popular form of tobacco in developed countries whilst it coexists with a range of
smokeless products in a developing country such as Pakistan. Cigarette or bidi
smoking carried the greatest risk, whereas pipes, cigars, chewable products and
dipping snuff were all associated with a substantially increased risk of MI. Enforcing
strict regulations on both smoking and smokeless tobacco use and encouraging
individuals to quit any type of tobacco use remain preferable, compared to a policy
which advocates replacement of cigarettes with other tobacco products with weak
associations with CVD risk.
256
Table 9.1: Genome-wide association studies reporting on the association with number of cigarettes per day and other smoking phenotypes
Genetic information
Smoking
phenotype
Effect size Reference
Chr SNP Position(Mb)
Coded
allele
Non
coded
allele
Gene Effectsize se P-value
Study
population
Sample
size Journal Year First author
1 rs910696 30,236,689 MA intergenic CPD ABS NA 3.00E-06 CGEMS (PLCO & NHS) 4.5K PlosOne 2009 Caporaso
3 rs6437740 108,948,507 MA BBX CPD
available
by study NA 2.40E-07 CGEMS (PLCO & NHS) 4.5K PlosOne 2009 Caporaso
4 rs5522 149576925 NR3C2 CPD 1.52E-05 15K Molecular Psychiatry 2008 Berrettini
4 rs5525 149576925 NR3C2 CPD 3.78E-05 15K Molecular Psychiatry 2008 Berrettini
5 rs2645339 178348669 GRM6 CPD 0.000272 15K Molecular Psychiatry 2008 Berrettini
7 rs215605 32,303,490 G T intergenic CPD 0.26 0.04 5.4x10-9 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
7 rs215614 32,313,860 G A intergenic CPD 0.22 0.04 2.1x10-7 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
7 rs7804771 136783133 CGKI CPD 9.81E-05 15K Molecular Psychiatry 2008 Berrettini
8 rs13280604 42,678,743 A G CPD 0.31 0.05 1.3x10-8 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
8 rs6474412 42,669,655 T C CHRNB3\CHRNA6 CPD 0.29 0.05 1.4x10-8 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
9 rs10869409 149576925 RORB CPD 5.53E-05 15K Molecular Psychiatry 2008 Berrettini
9 rs13293006 76326716 RORB CPD 0.000592 15K Molecular Psychiatry 2008 Berrettini
9 rs7846903 149576925 RORB CPD 0.000122 15K Molecular Psychiatry 2008 Berrettini
9 rs7873840 76340109 RORB CPD 0.001698 15K Molecular Psychiatry 2008 Berrettini
10 rs1028936 93,349,297 C A EGLN2  CPD −0.4464 0.074 1.29 × 10−9 Ox-GSK, TAG, ENGAGE 74K Nature Genetics 2015 TAGC
10 rs1329650 93,347,620 T G EGLN2 CPD −0.3673 0.059 5.67 × 10−10 
Ox-GSK, TAG,
ENGAGE 74K Nature Genetics 2014 TAGC
15 rs1051730 76,681,394 A G CHRNA5/CHRNA3/CHRNB5 CPD 0.8 0.05 2.4x10-69 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
15 rs1051730 76,681,394 G CHRNA5/CHRNA3/CHRNB5 CPD 1.71x10-66 Ox-GSK, TAG, ENGAGE 41K Nature Genetics 2010 Liu
15 rs1051730 76,681,394 T G
CHRNA5/CHRNA3/
CHRNB5 CPD 0.095 6x10-20
deCODE, Zaragoza,
the Netherlands 16K Nature 2008 Thorgeirsson
15 rs1051730 G A CHRNA5/CHRNA3/CHRNB5 CPD −1.0209 0.056 2.8x10-73 Ox-GSK, TAG,ENGAGE 74K Nature Genetics 2010 TAGC
15 rs12439738 90336555 SLCO3A1 CPD 0.000531 15K Molecular Psychiatry 2008 Berrettini
15 rs12439765 90336606 SLCO3A1 CPD 0.000625 15K Molecular Psychiatry 2008 Berrettini
15 rs16969968 76,669,980 G CHRNA5/CHRNA3/CHRNB5 CPD 4.29x10*65 Ox-GSK, TAG, ENGAGE 41K Nature Genetics 2010 Liu
15 rs16969968 G A CHRNA5/CHRNA3/CHRNB5 CPD −1.0029 0.056 5.57 × 10−72 Ox-GSK, TAG, ENGAGE 74K Nature Genetics 2011 TAGC
15 rs4932597 90338621 SLCO3A1 CPD 0.000245 15K Molecular Psychiatry 2008 Berrettini
15 rs4932598 90338849 SLCO3A1 CPD 0.000192 15K Molecular Psychiatry 2008 Berrettini
15 rs55853698 76,652,480 T
CHRNA5/CHRNA3/
CHRNB5 CPD 1.74x10-3
Ox-GSK, TAG,
ENGAGE 41K Nature Genetics 2010 Liu
15 rs6495308 76694711 CHRNA5/CHRNA3/CHRNB5 CPD 6.9x10-5 15K Molecular Psychiatry 2008 Berrettini
15 rs6495308 76,694,711 T CHRNA5/CHRNA3/CHRNB5 CPD 5.82x10-44 Ox-GSK, TAG, ENGAGE 41K Nature Genetics 2010 Liu
16 rs7260329 46,213,478 G A CYP2B6 CPD 0.2 0.04 5.5x10-6 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
17 rs758642 3,733,656 MA CAMKK1 CPD ABS NA 7.30E-06 CGEMS (PLCO & NHS) 4.5K PlosOne 2009 Caporaso
19 rs10411195 19,897,176 MA ZNF505 CPD ABS NA 5.80E-06 CGEMS (PLCO & NHS) 4.5K PlosOne 2009 Caporaso
19 rs1801272 46,046,373 A T CPD 0.68 0.18 1.1x10+4 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
19 rs4105144 46,050,464 C T CYP2A6 CPD 0.39 0.06 2.2x10-12 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
19 rs7937 45,994,546 T C RAB4B CPD 0.24 0.04 2.4x10-9 Ox-GSK, TAG, ENGAGE 85K Nature Genetics 2010 Thorgeirsson
X rs7050529 110,961,378 MA TRPC5 CPD ABS NA 6.20E-06 CGEMS (PLCO & NHS) 4.5K PlosOne 2009 Caporaso
MA: Minor allele; ABS: available by study. Legend: Chr: Chromosome; SNP: Single Nucleotide Polymorphism; CPD: Number of Cigarettes smoked per day; SI:
Smoking initiation. Literature search conducted on PubMed in January 2011. (search: (“smoking” OR “tobacco” OR “cigarettes”) AND (“genome wide” OR “whole
genome” OR “GWA” OR “SNP” OR “single nucleotide polymorphism“) as well as retrieving papers through manual inspections of citation lists.
257
References
(1) Strong K, Mathers C, Leeder S, Beaglehole R. Preventing chronic diseases: how
many lives can we save? Lancet 2005 October 29;366(9496):1578-82.
(2) World Health Organization. WHO report on the global tobacco epidemic. the
MPOWER package. 2008.
(3) Beaglehole R, Bonita R, Horton R et al. Priority actions for the non-communicable
disease crisis. Lancet 2011 April 23;377(9775):1438-47.
(4) World Health Organization. The Tobacco Atlas. 2002.
(5) World Health Organization. Country profiles, Pakistan. 25-8-2011.
(6) Ismail J, Jafar TH, Jafary FH, White F, Faruqui AM, Chaturvedi N. Risk factors for
non-fatal myocardial infarction in young South Asian adults. Heart 2004 March
1;90(3):259-63.
(7) Saunders C, Geletko K. Adolescent Cigarette Smokers' and Non-cigarette Smokers'
Use of Alternative Tobacco Products. Nicotine Tob Res 2012 February 7.
(8) Piano MR, Benowitz NL, Fitzgerald GA et al. Impact of smokeless tobacco products
on cardiovascular disease: implications for policy, prevention, and treatment: a policy
statement from the American Heart Association. Circulation 2010 October
12;122(15):1520-44.
(9) Gupta R, Gurm H, Bartholomew JR. Smokeless tobacco and cardiovascular risk.
Arch Intern Med 2004 September 27;164(17):1845-9.
(10) Loukas A, Batanova MD, Velazquez CE et al. Who uses snus? A study of Texas
adolescents. Nicotine Tob Res 2012 May;14(5):626-30.
(11) Peeters S, Gilmore AB. How online sales and promotion of snus contravenes current
European Union legislation. Tob Control 2012 January 21.
(12) Foulds J, Ramstrom L, Burke M, Fagerstrom K. Effect of smokeless tobacco (snus)
on smoking and public health in Sweden. Tob Control 2003 December;12(4):349-59.
(13) Dobe M, Sinha DN, Rahman K. Smokeless tobacco use and its implications in WHO
South East Asia Region. Indian J Public Health 2006 April;50(2):70-5.
(14) Gupta PC, Ray CS. Smokeless tobacco and health in India and South Asia.
Respirology 2003 December;8(4):419-31.
(15) Wald NJ, Watt HC. Prospective study of effect of switching from cigarettes to pipes or
cigars on mortality from three smoking related diseases. BMJ 1997 June
28;314(7098):1860-3.
(16) Tverdal A, Bjartveit K. Health consequences of pipe versus cigarette smoking. Tob
Control 2011 March;20(2):123-30.
(17) Baker F, Ainsworth SR, Dye JT et al. Health risks associated with cigar smoking. Stat
Med 2002 June 15;21(11):1539-58.
(18) Forouhi NG, Sattar N, Tillin T, McKeigue PM, Chaturvedi N. Do known risk factors
explain the higher coronary heart disease mortality in South Asian compared with
European men? Prospective follow-up of the Southall and Brent studies, UK.
Diabetologia 2006 November;49(11):2580-8.
258
(19) Kirkwood B, Sterne J. Measure of association and impact. Medical statistics. 2003. p.
447-57.
(20) IRIN. Pakistan: anti-smoking law have little impact. IRIN Humanitarian news and
analysis 2004 July 14.
(21) Whincup PH, Gilg JA, Emberson JR et al. Passive smoking and risk of coronary heart
disease and stroke: prospective study with cotinine measurement. BMJ 2004 July
24;329(7459):200-5.
(22) Teo KK, Ounpuu S, Hawken S et al. Tobacco use and risk of myocardial infarction in
52 countries in the INTERHEART study: a case-control study. Lancet 2006
August;19;368(9536):647-58.
(23) Boffetta P, Straif K. Use of smokeless tobacco and risk of myocardial infarction and
stroke: systematic review with meta-analysis. BMJ 2009 August 18;339:
(24) Hansson J, Galanti MR, Hergens MP et al. Use of snus and acute myocardial
infarction: pooled analysis of eight prospective observational studies. Eur J Epidemiol
2012 June 22.
(25) Nitzkin JL. Re: "Risk of incident cardiovascular disease among users of smokeless
tobacco in the Atherosclerosis Risk in Communities (ARIC) Study". Am J Epidemiol
2011 March 1;173(5):593-4.
(26) Gartner CE, Hall WD, Vos T, Bertram MY, Wallace AL, Lim SS. Assessment of
Swedish snus for tobacco harm reduction: an epidemiological modelling study.
Lancet 2007 June 16;369(9578):2010-4.
(27) Last J. A Dictionary of Epidemiology. 4th edition ed. Oxford University Press; 2001.
(28) Benowitz NL, Porchet H, Sheiner L, Jacob P, III. Nicotine absorption and
cardiovascular effects with smokeless tobacco use: comparison with cigarettes and
nicotine gum. Clin Pharmacol Ther 1988 July;44(1):23-8.
(29) Heck JE, Marcotte EL, Argos M et al. Betel quid chewing in rural Bangladesh:
prevalence, predictors and relationship to blood pressure. Int J Epidemiol 2012
April;41(2):462-71.
(30) Warner KE. Tobacco harm reduction: promise and perils. Nicotine Tob Res 2002;4
Suppl 2:S61-71.:S61-S71.
(31) Foulds J, Kozlowski L. Snus--what should the public-health response be? Lancet
2007 June 16;369(9578):1976-8.
(32) Boffetta P, Hecht S, Gray N, Gupta P, Straif K. Smokeless tobacco and cancer.
Lancet Oncol 2008 July;9(7):667-75.
(33) Gartner CE, Hall WD, Chapman S, Freeman B. Should the health community
promote smokeless tobacco (snus) as a harm reduction measure? PLoS Med 2007
July;4(7):e185.
(34) Tomar SL. Is use of smokeless tobacco a risk factor for cigarette smoking? The U.S.
experience. Nicotine Tob Res 2003 August;5(4):561-9.
(35) State-specific prevalence and trends in adult cigarette smoking--United States, 1998-
2007. MMWR Morb Mortal Wkly Rep 2009 March 13;58(9):221-6.
(36) Burnettes K. Sales of e-cigarettes soar, but do they really work? The Guardian 12
A.D. January 10;www.guardian.co.uk.
259
(37) Cobb NK, Abrams DB. E-cigarette or drug-delivery device? Regulating novel nicotine
products. N Engl J Med 2011 July 21;365(3):193-5.
(38) Grana RA, Glantz SA, Ling PM. Electronic nicotine delivery systems in the hands of
Hollywood. Tob Control 2011 June 9.
(39) yo-Yusuf OA, Burns DM. The complexity of 'harm reduction' with smokeless tobacco
as an approach to tobacco control in low-income and middle-income countries. Tob
Control 2012 March;21(2):245-51.
(40) www.who.int/fctc/en. WHO Framework Convention on Tobacco Control. 24-8-2012.
(41) Wang XL, Greco M, Sim AS, Duarte N, Wang J, Wilcken DE. Effect of CYP1A1 MspI
polymorphism on cigarette smoking related coronary artery disease and diabetes.
Atherosclerosis 2002 June;162(2):391-7.
(42) Wang XL, Sim AS, Badenhop RF, McCredie RM, Wilcken DE. A smoking-dependent
risk of coronary artery disease associated with a polymorphism of the endothelial
nitric oxide synthase gene. Nat Med 1996 January;2(1):41-5.
(43) Wang XL, Wang J, Wilcken DE. Interactive effect of the p53 gene and cigarette
smoking on coronary artery disease. Cardiovasc Res 1997 August;35(2):250-5.
(44) Barakat K, Kennon S, Hitman GA et al. Interaction between smoking and the
glycoprotein IIIa P1(A2) polymorphism in non-ST-elevation acute coronary
syndromes. J Am Coll Cardiol 2001 November 15;38(6):1639-43.
(45) Li R, Boerwinkle E, Olshan AF et al. Glutathione S-transferase genotype as a
susceptibility factor in smoking-related coronary heart disease. Atherosclerosis 2000
April;149(2):451-62.
(46) Thorgeirsson TE, Gudbjartsson DF, Surakka I et al. Sequence variants at CHRNB3-
CHRNA6 and CYP2A6 affect smoking behavior. Nat Genet 2010 May;42(5):448-53.
(47) Thorgeirsson TE, Geller F, Sulem P et al. A variant associated with nicotine
dependence, lung cancer and peripheral arterial disease. Nature 2008 April
3;452(7187):638-42.
(48) Liu JZ, Tozzi F, Waterworth DM et al. Meta-analysis and imputation refines the
association of 15q25 with smoking quantity. Nat Genet 2010 May;42(5):436-40.
(49) Genome-wide meta-analyses identify multiple loci associated with smoking behavior.
Nat Genet 2010 May;42(5):441-7.
(50) Breitling LP, Yang R, Korn B, Burwinkel B, Brenner H. Tobacco-Smoking-Related
Differential DNA Methylation: 27K Discovery and Replication. Am J Hum Genet 2011
April 8;88(4):450-7.
(51) Kuschner WG, Reddy S, Mehrotra N, Paintal HS. Electronic cigarettes and thirdhand
tobacco smoke: two emerging health care challenges for the primary care provider.
Int J Gen Med 2011 February 1;4:115-20.:115-20.
(52) Schick S. Thirdhand smoke: here to stay. Tob Control 2011 January;20(1):1-3.
260
APPENDIX 1: List of publications authored during PhD
Published
1. Genetic Variants in Novel Pathways Influence Blood Pressure and
Cardiovascular Disease Risk. International Consortium for Blood Pressure
Genome-Wide Association Studies. Nature 2011 Sep 11. 478(7367);103-9.
2. Genetic determinants of major blood lipids in Pakistanis compared with
Europeans. Saleheen D, Soranzo N, Rasheed A, Scharnagl H, Gwilliam R,
Alexander M & alii. Circ Cardiovasc. Genet. 2010 Aug;3(4):348-57.
3. Association of the 9p21.3 locus with risk of first-ever myocardial infarction in
Pakistanis: case-control study in South Asia and updated meta-analysis of
Europeans. Saleheen D, Alexander M & alii. J. Arterioscler Thromb Vasc Biol.
2010 Jul;30(7):1467-73. [Note: Saleheen D. and Alexander M. are joint first
authors].
In preparation
4. Behavioural, biological and socio-economic risk factors for myocardial infraction
in South Asia: case-control study in urban Pakistan. Alexander M, Saleheen D,
Di Angelantonio E, Kee Ho W, Johnson L, Danesh J & alii.
5. Smoking cessation and access to tobacco outlet: English experience. Han T,
Niggebrugge A, Alexander M, Battersby J, Hollands G, Marteau T.
6. Interplay of cigarette smoking with metabolic risk factors on the incidence of
major vascular morbidity; an individual participant meta-analysis of 1 million
people. Alexander M, The Emerging Risk Factors Collaboration.
261
Appendix 2: Socio-economic status and the risk of myocardial
infarction in Pakistan
Background
In developed countries, lower socioeconomic groups have been shown to carry more CVD
risk factors and to be at increased risk of MI 1,2, but results have been inconsistent in other
parts of the world 3. In Pakistan, a developing country with a narrower middle class and a
bigger proportion of poor individuals compared to developed countries4, the importance of
socio-economic status in relation to CVD risk has not yet been investigated. PROMIS
collected information on several markers of socio-economic status (income, education,
household possessions, employment status and category of job), lifestyle, medical and
biochemical information on CVD risk factors in more than 12,000 first ever MI cases and age
and sex frequency matched controls. This large dataset provides a unique opportunity to
investigate more precisely than has been previously possible the relationship between multi-
faceted socio-economic status and the risk of MI in Pakistan 5.
Methods
A principal component analysis was used to define socio-economic status using variables of
income, education, asset ownership and occupation. As occupation was added later in the
questionnaire and available only on a subset of 4826 cases and 5417 controls, sensitivity
analyses were done to compare the PCA results with and without occupation on individuals
with non missing information. ORs were computed using unconditional logistic regression,
and adjusted for at least the “basic covariates”: age, sex, recruitment centre, sub-ethnicity,
tobacco use, history of diabetes or hypertension, family history of CAD, waist to hip ratio and
LDL-cholesterol. Participants were excluded if they did not provide information on all the
covariates for basic adjustment, retaining 6057 cases and 6889 controls. To characterize
shapes of associations, ORs calculated within quintiles or within pre-defined categories were
plotted against mean values within each quintile or category; 95% confidence intervals were
represented using “floating absolute variances” which enables graphical comparisons of ORs
between every two categories, and not only with the reference group 6. I investigated the
possibility of effect-modification by age, sex and other relevant subgroups fitting interaction
terms and representing the shapes of association within subgroups. Analyses were
conducted using STATA v10 (StataCorp, Texas).
262
Results
Amongst controls, 77% were employed and 8% were unemployed, the rest declaring to be
retired. The main occupations in Men were professional - 15%, skilled and unskilled labour -
respectively 16% and 20% -; while nearly 9 out of 10 women declared being housewives
(Table A2.1). The median self-reported monthly income was 12,000 Pakistani rupees, which
corresponds to approximately 140$ US. The most commonly owned items were in order a
television, a mobile phone, a home, a radio, a motorcycle, a bicycle, a computer, air
conditioning and land. Around two thirds of men (69%) reported having received a formal
education and 50% reported more than 10 years of education, while 51% percent of women
reported no formal education.
The principal component analysis of socio-economic indices identified one main gradient
labelled “socio-economic status” which explained 14% of the variance of all socio-economic
indices (Figure A2.1 & A2.2), and was approximately normally distributed (Figure A2.3).
In the top third of socio-economic status (labelled “upper class”), the median income was
25,000 Pakistani rupees, 14 years of education and 6 household possessions; in the middle
third it was 12,000 Pakistani rupees, 8 years of education and 4 household possessions; and
in the bottom third it was 7,000 Pakistani rupees, no formal education and 3 household
possessions (Table A2.2). In the upper class, professionals were largely over-represented,
and 72% reported more than 10 years of education, compared to 21% of the middle class and
2% in the lower class. Individuals belonging to the upper class were more likely to be never
smokers and to cook using oil rather than ghee. Women, Pathan and Balochi ethnic groups
were the most socio-economically deprived groups, with more than 50% of the individuals in
these groups belonging to the bottom third of socio-economic status, labelled ‘lower class’;
while Men, Sindhi and Urdu ethnic groups were the most privileged groups (Table A2.3).
The middle and upper classes of the “socio-economic status” were at lower risk of MI risk
than the middle class (Figure A2.4). The protection conferred by belonging to the upper class
was attenuated amongst individuals aged ≤50 years old (Figure A2.5).
Discussion
In developed countries a low socio-economic position has been shown to increase MI risk 1,2
whilst results have been inconsistent in middle and low-income regions 3. In PROMIS,
individuals in the middle range of socio-economic status were at increased risk of MI while
individuals in top third of the socio-economic gradient were protected. The middle group was
more likely to smoke, use ghee, to have a self-reported history of diabetes or hypertension,
and to have a family history of CAD than individuals belonging to the lower group.
Urdu were over-represented amongst the upper class and were also more likely to report a
family history of MI -20% of Urdu versus 16% of Punjabi, 6% of Pathan and 5% of Sindhi -
and had higher LDL-cholesterol levels (p-value<0.0001). Urdu speakers have also been
263
shown to have higher age adjusted prevalence of hypertension than Punjabi and Sindhi
individuals 7. However, adjustment for these covariates as well as ethnicity in the model did
not alter the strength of the association between socio-economic status and MI risk. There
may be some residual confounding such as amount and duration of tobacco use, but better
access to health care and better awareness of the danger of tobacco use in the upper class
group is most likely to explain their protection against MI risk. In this respect, lack of
awareness of the danger of smoking and smokeless tobacco have been correlated with lower
education levels in previous studies in Pakistan 8,9. In a cross-sectional study among
Pakistani students of a private medical university with equal proportion of men and women,
the average income of the household of students was 50,000 Pakistani rupees per months,
meaning most would be classified in the top third of the socio-economic gradient and labelled
“upper class”, and only 7% of students were tobacco users, compared to 34% of men and
12% of women in the general Pakistani population 10,11.
Measurement error in socio-economic indices resulting from mis-reporting of income and
asset ownership, errors in recall or mis-categorization into type of jobs, may also have
resulted in misclassification of individuals 12. In addition, socio-economic position determined
by principal component analysis of asset-ownership has been shown to correlate poorly in
developing countries with consumption expenditure 13, and therefore may not be an
appropriate risk factor. Finally, the first component of the principal component analysis
labelled as “socio-economic gradient” only explained 14% of the overall variance of socio-
economic indices, and therefore only captured a small part of the multidimensional effects of
income, education, asset ownership, employment and type of job.
INTERHEART reported a weaker association of socio-economic status in low income
countries compared to high income countries, and a weak association worldwide of income
with MI risk; whilst education was the marker most consistently associated with risk for acute
MI globally 14. In this context and in light of my findings, the use of a single dimension of
socio-economic status rather than a composite variable, namely education, may be more
appropriate when assessing MI risk in the developing world and more specifically in
Pakistanis. In my study as in INTERHEART, education was continuously and log linearly
associated with a decrease in MI risk. OR for ≥20 years of education versus no education: 
was 0.73 (0.65; 0.82).
Conclusion
Compared to lower socio-economic groups, Individuals classified as middle class on the basis
of income, education, household possessions, job and employment status, were at increased
risk of MI in a Pakistani urban population, whilst the upper class was protected.
264
Figure A2.1: Principal component analysis showing second versus first component
Ownership Income
Education Occupation and employment status
Retaining two components without rotation of the matrix of loadings.
Home
Car
Motorcycle
Bicycle
Radio
Television
Computer
Mobile_phone
Income<10K_PAK_rupees
Income_10-20K_PKR
Income>20K PKR
No formal education
Education_1-10 years
Education>10 yearsProfessional
Skilled_labour
Unskilled_labour
Housewife
Farmer Business
Clerical
Self employed
Other
Unemployed
Retired
-.4
-.2
0
.2
.4
S
ec
on
d
pr
in
ci
pa
lc
om
po
ne
nt
-.4 -.2 0 .2 .4
First principal component
265
Figure A2.2: Scree plot of Eigen values over the number of principal components
0
1
2
3
4
E
ig
en
va
lu
es
0 5 10 15 20 25
Number
266
Figure A2.3: Distribution of the socio-economic gradient by case-control status
Graph by case-control status
0
.1
.2
.3
-5 0 5 -5 0 5
Control Case
D
en
si
ty
Scores for component 1
Graphs by Case-control status
267
Figure A2.4: Association of socio-economic status with MI risk
.7
5
1
1.
5
2
3
O
R
(9
5%
C
I)
Low Middle High
Adjusted for age, sex, ethnicity and centre, diabetes, hypertension, family history of CAD,
LDL, tobacco use, WHR
Further adjusted for socio-economic status, dietary patterns and cooking oil
Note: Dataset reduced to 4521 cases and 4603 controls with information on all the covariates
Odds ratios were computed within fifths of income, education and number of household
possessions; and were plotted against the arithmetic mean within each fifth; except for
income and education where the geometric mean was used to account for the skewness of
the distributions. The size of the box is proportional to the inverse of the variance of OR.
268
Figure A2.5: Association with socio-economic status by age groups and by sex
.5
.7
5
1
1.
5
2
3
O
R
(9
5%
C
I)
Low Middle High
Male
Female
≤50 years old
>50 years old
.5
.7
5
1
1.
5
2
3
O
R
(9
5%
C
I)
Low Middle High
a) By age groups b) By sex
OR: Odds ratios. Models adjusted for gender, ethnicity and centre, tobacco use, LDL-C,
WHR, history of diabetes and hypertension
269
Table A2.1: Descriptive characteristics of socio-economic status by status and sex
Cases Controls Overall
P-value
of
difference
cases
versus
controls
% Male / %
Female
(n=5503/n=1456)
P-value
Male v
Female
% Male / %
Female (n=4643
/ n=922)
P-value
Male v
Female
Formal education <0.0001 <0.0001 <0.0001
No 31% / 64% 30% / 60%
≤10 years 40% / 29% 35% / 26%
>10 years 30% / 7% 35% / 14%
Income 0.8 0.003 <0.0001
<10K PKR 26% / 25% 34% / 36%
10-20K PKR 34% / 36% 33% / 36%
  ≥20K PKR 40% / 39%   33% / 28%     
Occupation1 <0.0001 <0.0001 <0.0001
Professional 9% / 4% 13% / 7%
Skilled labour 17% / 3% 14% / 3%
Unskilled labour 15% / 6% 16% / 5%
Housewife 0% / 25% 0% / 47%
Farmer 2% / 0% 2% / 0%
Business 12% / 0% 10% / 1%
Clerical 6% / 0% 7% / 0%
Self-employed 6% / 1% 6% / 0%
Other 12% / 3% 10% / 3%
Unemployed 6% / 50% 5% / 30%
Retired 15% / 7% 16% / 4%
Ownership
Home 80% / 80% 0.7 72% / 72% 0.6 <0.0001
Car 17% / 14% 0.03 19% / 14% <0.0001 0.2
Motorcycle 42% / 38% 0.01 39% / 30% <0.0001 <0.0001
Bicycle 25% / 22% 0.02 27%/ 22% <0.0001 0.06
Radio 41% / 38% 0.1 47% / 42% 0.0003 <0.0001
Television 89% / 88% 0.6 83% / 82% 0.5 <0.0001
Land 12% / 7% <0.0001 16% / 9% <0.0001 <0.0001
Computer 18% / 17% 0.26 23% / 15% <0.0001 <0.0001
Mobile phone 88% / 84% 0.001 75% / 69% <0.0001 <0.0001
1: For occupation, in cases information was available in 4391 males and 248 females
provided information; and in controls in 4682 males and 662 women; % corresponds to a
column percentage. PKR: Pakistani rupees.
270
Table A2.2: Socio-economic indices across quintiles of the socio-economic gradient
Socio-economic indices Q1 Q2 Q3 Q4 Q5 P-value
Monthly income, median (PKR) 6,000 8,000 13,000 20,000 30,000 <0.0001
Formal education, median (years) 0 0 8 12 14 <0.0001
Number of household possessions,
median 2 3 4 5 6 <0.0001
Occupation/employment (row %)
Professional 1 2 10 25 62 <0.0001
Skilled labour 15 25 27 22 11
Unskilled labour 50 29 12 7 2
Housewife 22 28 29 16 5
Farmer 18 29 22 23 9
Business 3 5 17 29 45
Clerical 4 11 27 36 22
Self employed 10 23 29 24 13
Other occupation 15 23 20 22 21
Unemployed 40 26 19 10 5
Retired 16 24 21 23 16
Qi: Quintile number i. P-value of linear trend for income, education and number of household
possessions; and p-value from a Chi2 test of independence for occupation/employment.
271
Table A2.3: Correlates of socio-economic gradient
'Socio-economic gradient'
Low Middle High P-value
Demography
Male sex 73% 77% 86% <0.0001
Age, mean (SD) 54.2 (9.83) 53.2 (9.45) 51.3 (8.98) <0.0001
Major ethnic groups <0.0001
Punjabi 26% 33% 31%
Urdu 35% 39% 41%
Pathan 13% 7% 2%
Sindhi 14% 9% 14%
Biochemistry, medical and familial
historyDiabetes 13% 15% 15% 0.08
Hypertension 23% 29% 31% <0.0001
Family history of MI 8% 15% 22% <0.0001
LDL-cholesterol (mmol/l), mean (SD) 2.76 (1.07) 2.87 (1.05) 2.99 (1.06) <0.0001
Waist to hip ratio, mean (SD) .934 (.0659) .946 (.0657) .95 (.0638) <0.0001
Tobacco use <0.0001
Never/ex 62% 67% 75%
Dip naswar only 9% 5% 1%
Chew only 7% 8% 3%
Smoke (and dip/chew) 21% 20% 21%
Diet
Cooking fat <0.0001
Oil 42% 47% 59%
Oil & ghee 20% 29% 32%
Ghee only 38% 25% 9%
Dietary pattern 1 <0.0001
Low 30% 34% 36%
Middle 35% 37% 28%
High 34% 29% 36%
Dietary pattern 2 <0.0001
Low 53% 33% 16%
Middle 30% 40% 30%
High 17% 27% 55%
Note: Column percentages. Low corresponds to the bottom third, middle to the middle third
and high to the top third of the distribution of the socio-economic gradient. *P-values derive
from Chi2 tests of independence for categorical variables and from a Fisher test of equality of
the means (computed as a test of nullity of all the coefficients in the regression of the
continuous row variable - for example LDL-C - over categories of the column variable - for
example tobacco use).
272
References
1. Avendano M, Kunst AE, Huisman M et al. Socioeconomic status and ischaemic heart
disease mortality in 10 western European populations during the 1990s. Heart.
2006;92:461-467.
2. Mackenbach JP, Stirbu I, Roskam AJ et al. Socioeconomic inequalities in health in 22
European countries. N Engl J Med. 2008;358:2468-2481.
3. Manrique-Garcia E, Sidorchuk A, Hallqvist J, Moradi T. Socioeconomic position and
incidence of acute myocardial infarction: a meta-analysis. J Epidemiol Community Health.
2011;65:301-309.
4. Ravaillon, M. The developing world's bulging (but vulnerable) "middle class". Policy
Research Working Paper 4816. The World Bank. 4816, 1-28. 2009. Washington DC,
USA.
5. Dennis B, Aziz K, She L et al. High rates of obesity and cardiovascular disease risk
factors in lower middle class community in Pakistan: the Metroville Health Study. J Pak
Med Assoc. 2006;56:267-272.
6. Easton DF, Peto J, Babiker AG. Floating absolute risk: an alternative to relative risk in
survival and case-control analysis avoiding an arbitrary reference group. Stat Med.
1991;10:1025-1035.
7. Jafar TH, Levey AS, Jafary FH et al. Ethnic subgroup differences in hypertension in
Pakistan. J Hypertens. 2003;21:905-912.
8. Alam AY, Iqbal A, Mohamud KB, Laporte RE, Ahmed A, Nishtar S. Investigating socio-
economic-demographic determinants of tobacco use in Rawalpindi, Pakistan. BMC
Public Health. 2008;8:50.:50.
9. Khawaja MR, Mazahir S, Majeed A et al. Chewing of betel, areca and tobacco:
perceptions and knowledge regarding their role in head and neck cancers in an urban
squatter settlement in Pakistan. Asian Pac J Cancer Prev. 2006;7:95-100.
10. Nisar N, Qadri MH, Fatima K, Perveen S. A community based study about knowledge
and practices regarding tobacco consumption and passive smoking in Gadap Town,
Karachi. J Pak Med Assoc. 2007;57:186-188.
11. Nanan, D. and White, F. PROCOR: 7/25/99: The National Health Survey of Pakistan:
Review and discussion of report findings pertaining to selected risk factors for
cardiovascular disease. 25-7-1999.
12. Onwujekwe O, Hanson K, Fox-Rushby J. Some indicators of socio-economic status may
not be reliable and use of indices with these data could worsen equity. Health Econ.
2006;15:639-644.
13. Howe LD, Hargreaves JR, Huttly SR. Issues in the construction of wealth indices for the
measurement of socio-economic position in low-income countries. Emerg Themes
Epidemiol. 2008;5:3.:3.
14. Rosengren A, Subramanian SV, Islam S et al. Education and risk for acute myocardial
infarction in 52 high, middle and low-income countries: INTERHEART case-control study.
Heart. 2009;95:2014-2022.
273
Appendix 3: Diet and the risk of myocardial infarction in
Pakistan
Background
A “prudent” pattern of diet high in fruit, vegetables and fish intake, has been shown to reduce
prospective CVD risk in Western populations 1-4. However, there is sparse evidence on the
importance of diet as a risk factor for CVD in other populations 5,6. In addition, whilst Western
diets are well characterized, South Asian diet remains little studied in indigenous settings. In
studies set amongst immigrants, South Asians reported higher mean total energy intake,
including higher intakes of total fat, polyunsaturated fat and protein, than White Europeans 7,8
and Americans 9. However, migrants varied in their adherence to traditional diet and were
shown to have adopted some Western food, such as snacks 10.
In this context, PROMIS represents the largest scale epidemiological resource with detailed
dietary information. PROMIS recorded frequency of 43 food items locally relevant to Pakistani
diet in more than 12,000 first ever MI cases and matched controls recruited in 5 urban
centers across Pakistan, with additional information on demography, lifestyle, as well as
medical and family history.
Methods
A locally relevant food frequency questionnaire was developed following a 24 hour recall and
a 7-days food diary on a subset of 200 healthy adults as described previously 11. Questions
related to food rather than drink consumption were used in this analysis. Categories of food
frequency were converted into weekly consumption (Table A3.1).
Principal component analysis on the converted food items with rotation of the matrix of
loading was conducted to reveal main dietary patterns. Major components were determined
after looking at the scree-plot (Figure A3.1), which shows the drop in variance explained by
each additional component; and the interpretability of the components obtained (List A3.1).
Sensitivity analyses were run which (1) excluded individuals with self reported hypertension
or diabetes (2568 case and 2528 controls); (2) defined quintiles of dietary patterns within sex;
(3) computed the matrix of loadings in controls only and inferred scores for cases and
controls; (4) ran the principal component analysis of food groups rather than food items.
Analyses were conducted using STATA v10 (StataCorp, Texas).
Odds ratios (ORs) were computed using unconditional logistic regression, and adjusted for at
least the “basic covariates”: age, sex, recruitment centre and sub-ethnicity, tobacco use,
history of diabetes or hypertension, family history of CAD, waist to hip ratio and LDL-
cholesterol. To characterize shapes of associations, ORs calculated within quintiles or within
274
pre-defined categories were plotted against mean values within each quintile or category. For
skewed variables such as income, the geometric mean was reported rather than the
arithmetic mean.
When assessing the shape of association with MI, 95% confidence intervals were derived
from “floating absolute variances” to enable graphical comparisons of ORs between every
two categories, and not only with the reference group 12. I investigated the possibility of
effect-modification by age, sex and other relevant subgroups fitting interaction terms and
representing the shapes of association within subgroups. To avoid false-negative findings
and because of the large number of tests performed, only p-values <0.001 were emphasized.
Analyses were conducted using STATA v10 (StataCorp, Texas).
Results
Principal component analysis of the food frequency questionnaire identified two main dietary
patterns: a “high vegetables and carbohydrates diet” (with loadings >0.2 for cooked and raw
vegetables, and potatoes) and a “high meat, fish and sweets pattern” (with loadings >0.2 for
chicken, beef, mutton, fish and sweets), each pattern accounting for 9% of the total variance
of the dietary data (Figure A3.2).
There was considerably greater consumption of raw and cooked vegetables (but only slightly
higher intake of carbohydrates) across the fifths of values of the first principal component
(Figure A3.3 & Figure A3.4). Controls scoring in the top third of the “high vegetables and
carbohydrates diet” compared to the bottom third were more likely to be women, Urdu, with a
self-reported history of hypertension, oil rather than ghee users and to score high for the
“high meat, fish and sweet diet”. Controls above the median score for “high meat, fish and
sweet diet” were in average younger, more likely be men, Urdu, with a family history of MI, to
smoke, to belong to the upper class, to use oil rather than ghee for cooking; and less likely to
be diabetics.
The “high vegetables and carbohydrates” diet association with MI risk was non linear (Figure
A3.5), with a positive relationship below the median and a negative relationship above the
median. The association did not attenuate upon adjustment for socio-economic status, the
other dietary patterns and cooking oil; upon adjustment for individual food items scoring high
for this component (cooked and raw vegetables, pulses and nut, potatoes and rice) (Figure
A3.6). There was no evidence of an effect modification by age and sex (Figure A3.7).
The “high meat, fish and sweet” pattern was positively and approximately log-linearly
associated with MI risk. Individuals scoring in the top 5th of the distribution had an OR of 1.79
(1.54; 2.08) compared to the bottom 5th. Further adjustment for the other dietary pattern,
cooking oil and socio-economic status only modestly attenuated the ORs. Sensitivity
analyses identified similar dietary patterns, yielding similar shapes and strength of
association with MI risk.
275
Discussion
Over the past 30 years, South Asian countries have been experiencing a dietary transition,
which has been accompanying the rural exodus and rapid urbanization 13,14. The adoption of
a westernized diet especially rich in meat and sweets has been linked to urbanization 15,16
and is also happening amongst Pakistani immigrants settling in Western countries 17.
The nutrition transition has been marked by a shift from diets high in carbohydrates and
based on indigenous staple grains or starchy roots, locally grown legumes, vegetables and
fruits and limited foods of animal origins; toward more fat-rich diets than include more pre-
processed food of animal origin, more sweets and sugar and more fried food cooked in ghee
and oil 13,18,19. In PROMIS, traditional staples correlated positively with fruit and vegetable
consumption on the first dietary pattern; while meat consumption correlated positively with
sweet snacks on the second dietary pattern.
In Western populations, adherence to a diet high in red and processed meat has been shown
to increase MI risk whilst a “prudent diet” high in fruit, vegetables and fish intake was
protective 1-4. In this study, a “high meat, fish and sweets” diet was associated with a positive
and approximately log linear relationship with MI risk (adjusted OR for top 5th versus bottom
5th: 1.79 (95% CI: 1.54; 2.08)). Individuals who scored in the top 3rd of this dietary pattern and
who also cooked with ghee had more than triple the risk of MI compared to individuals in the
bottom 3rd of the high meat diet and cooking with oil only (adjusted OR equal to 2 (2.4; 4.3)).
Pakistan is the country with the highest per capita consumption of meat in all of South Asia 20,
being twice as high as in India, and therefore public health campaigns highlighting to the
health effects of both high meat consumption and cooking with ghee could be considered.
In contrast, the “high vegetables and carbohydrates” pattern was not associated with a
reduction in risk, and the shape of association was an inverted “J”. Individuals scoring around
the median of this dietary pattern experienced an increased risk, even after multiple
adjustments. Rather than boiled or steamed, vegetables are usually fried in Pakistan 8 and
higher intake of vegetables may mask a greater consumption of edible fat. Quantities of
edible fat were not reliably recorded in this dataset and it was not possible to adjust for this
variable. In addition, individuals in the middle group of the “high vegetables and
carbohydrates” pattern were likely to belong to the middle class, a socio-economic group who
has been the most rapid in adopting a Westernized lifestyle of low physical activity and low-
quality diet. Tthe lower class has been slower in adopting these practices and the upper class
has access to better quality produces, and has been shown to practice leisure time physical
activity 19.
This study has several limitations. The dietary questionnaire enquired about the frequency of
food consumption and did not capture the quantity of food consumed. As a result, I was
unable to adjust for total energy intake. In addition, food frequency questionnaires are prone
to measurement error due to imperfect recall and within person variability. This may result in
276
misclassification of individuals, and is more likely to happen amongst individuals reporting
food frequencies not far from the average, such as individuals in the middle tertile of the
middle tertile of the distribution of each dietary pattern.
Conclusion
As part of a strategy to reduce the epidemic burden of CVD in Pakistan and more generally in
South Asia, dietary programs could be considered which target individuals with a high
consumption of meat and sweets.
277
Figure A3.1: Scree plot of Eigen values over number of principal components
0
1
2
3
4
5
E
ig
en
va
lu
es
0 10 20 30 40
Number
278
Figure A3.2: Plot of the first and second principal components labelled as “dietary pattern 1” and “dietary pattern 2”
Khameri_Nan
Paratha_Puri Roti(safed_Aaata)
Roti(Laal)
Roti(mixed_Aaata)
Beef_mutton_biryani
Kata_kut_organ_meat
Beef_salan
Mutton_salan
Beef_boti_tikka_kabab
Paya
Nehari
Chicken_biryani
Chicken_salan
Chicken_boti_tikka_kabab
Fried_chicken, breast chicken
Dark_green_leafy_vegetables
(cooked)
Cruciferous_vegetables
(cooked)
Other vegetables (cooked)
Milk
Makkhan_
MargarineLassi_Curd
Eggs
Fish_salan
Fried_fish
Fruits
Nimko_other_fried items
Pickled_vegetables
Potatoes
Fried_potatoes_French_fries
Daal_Lobia_Channa_Cholay
Nuts_seeds
Dark_green_leafy_vegetables(salad)
Cruciferous_vegetables
(salad)Other_vegetables(salad)
Tomatoes
Raw_onion
Rice
Gurh_Shakkar_
honey_jam_
marmelade
Bagarkhani_Papay_cakes_biscuits
Kheer_custard_
milkshakes_ice_cream
Halwa_Mithai_Jalabe_chocolate
-.1
0
.1
.2
.3
.4
D
ie
ta
ry
pa
tte
rn
2
-.1 0 .1 .2 .3 .4
Dietary pattern 1
The first component explains 10.2% of the variance, the 2nd component 8.1% and the 3rd component 4.5% of the variance.
279
Figure A3.3: Median frequency of weekly intake of food groups across fifths of the two dietary
patterns
a) By Fifths of the first dietary pattern labelled “high carbohydrates and vegetables” diet
0
1
5
(Once
a day)
10
15
20
(3 times
a day)
25
M
ed
ia
n
w
ee
kl
y
fre
qu
en
cy
Be
ef
or
mu
tto
n Ce
rea
ls
Ch
ick
en
Co
ok
ed
ve
ge
tab
les
Da
iry
Eg
gs
Fis
h
Fru
its Oi
ly
bre
ad Ot
he
r
Bre
ad Pic
kle
s
Pu
lse
s
& n
uts
Po
tat
oe
s
Ra
w
ve
ge
tab
les
Ric
e
Sn
ac
ks
Sw
ee
ts
Q1
Q2
Q3
Q4
Q5
First dietary pattern
b) By fifths of the second dietary pattern labelled “high meat and sweets” diet
0
1
5
10
15
20
25
M
ed
ia
n
w
ee
kl
y
fre
qu
en
cy
Be
ef
or
mu
tto
n Ce
rea
ls
Ch
ick
en
Co
ok
ed
ve
ge
tab
les
Da
iry
Eg
gs
Fis
h
Fru
its
Ot
he
r
Bre
ad
Pic
kle
s
Pu
lse
s
& n
uts
Po
tat
oe
s
Ra
w
ve
ge
tab
les
Ric
e
Sn
ac
ks
Sw
ee
ts
Oil
y
bre
ad
(Once
a day)
(3 times
a day)
Q1
Q2
Q3
Q4
Q5
Second dietary pattern
280
Figure A3.4: Ratio of carbohydrate to vegetable intake across fifths of the two dietary patterns
0
1
2
3
4
5
R
at
io
of
ca
rb
oh
yd
ra
te
s
to
ve
ge
ta
bl
es
Q1 Q2 Q3 Q4 Q5
Quintiles of first dietary pattern
0
1
2
3
4
5
R
at
io
of
ca
rb
oh
yd
ra
te
s
to
ve
ge
ta
bl
es
Q1 Q2 Q3 Q4 Q5
Quintiles of second dietary pattern
a)
b)
Q: Quintile. The ratio of carbohydrates to vegetables was defined as the ratio of the following food
items converted into weekly frequencies: (rice + fried potatoes + potatoes + roti safed Aaata +
khameri + rotil laal + roti mixed Aaata) divided by (cooked dark green vegetables + cooked
cruciferous vegetables + other cooked vegetables + dark green vegetables in salad + cruciferous
vegetables in salad + other vegetables in salad + tomatoes + raw onions).
281
Figure A3.5: Odds ratios for myocardial infarction with dietary patterns
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
First dietary pattern
“carbohydrates and vegetables”
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
Second dietary pattern
“meat and sweets”
a)
b)
0
5
10
15
20
25
30
Q1 Q2 Q3 Q4 Q5
0
1
2
3
4
Q1 Q2 Q3 Q4 Q5
Quintile of first dietary pattern
“Carbohydrate and vegetables” diet
Quintile of second dietary pattern
“meat and sweets” diet
Carbohydrates
Raw vegetables
Cooked vegetables
Sweets
Beef of mutton
Chicken
Fish
M
ed
ia
n
w
ee
kl
y
fre
qu
en
cy
of
fo
od
co
ns
um
pt
io
n
M
ed
ia
n
w
ee
kl
y
fre
qu
en
cy
of
fo
od
co
ns
um
pt
io
n
Panel a) OR: Odds Ratio, CI: confidence intervals, Q: Quintiles of the exposure variable. All
analyses were adjusted for age, sex, self-reported ethnicity, recruitment centre, tobacco use,
diabetes status, hypertension status; LDL-cholesterol levels, waist-to-hip ratio, and family history
of MI. Confidence intervals were calculated using “floating absolute variances”. Participants in
lowest fifths are referents. Dietary patterns were calculated using principal component analyses
(PCA). Panel b) across fifths of each dietary pattern, in the control participants, median weekly
consumption of major food groups that have a loading value of ≥0.20 on PCA analyses.   
282
Figure A3.6: Progressive adjustment of the association between dietary patterns and MI risk
Adjusted for conventional risk factors
Further adjusted for socio-economic status
Further adjusted for the other dietary pattern (food groups for fruit and veg. consumption) and cooking oil
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
First dietary pattern
“carbohydrates and vegetables”
Second dietary pattern
“meat and sweets”
Quintile Quintile
OR: Odds Ratio, CI: confidence intervals, Q: Quintiles of the exposure variable.
283
Figure A3.7: Association between dietary patterns and MI risk by age groups and by sex
Male
Female
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
First dietary pattern
“carbohydrates and vegetables”
P-value of interaction<0.0001
Second dietary pattern
“Meat and sweets”
P-value of interaction: 0.6
≤50 years old
>50 years old
.4
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
.4
.6
.8
1
1.
2
1.
5
2
O
R
(9
5%
C
I)
Q1 Q2 Q3 Q4 Q5
First dietary pattern
“carbohydrates and vegetables”
P-value of interaction: 0.5
Second dietary pattern
“Meat and sweets”
P-value of interaction: 0.0002
a) By age groups
b) By Sex
Models adjusted for gender, age, ethnicity and centre, LDL-C, tobacco use, WHR, family history
of CAD, history of hypertension or diabetes.
284
Table A3.1: Conversion of food frequency questions into weekly consumption
Nominal Level Coded as Converted to
weekly consumption
Converted to daily
consumption
More than 4 times per day 1 4X7=28 4
Between 2 and 4 times per day 2 3X7=21 3
Once per day 3 1X7=7 1
More than 3 times per week 4 5 5/7 = .71428571
2 to 3 times per week 5 2.5 2.5/7 = .35714286
Once per week 6 1 1/7 = .14285714
Less than once per week 7 1/2.5=0.4 1/14 = .07142857
Once per month 8 1/4=0.25 1/30 = .03333333
Less than once per month 9 1/12=0.083333 1/80 = .01250000
During Ramadan only 10 . .
Never 11 0 .
Everything else * . .
285
Table A3.2: Correlates of dietary patterns in controls
Dietary pattern 1 Dietary pattern 2
Low
(n=1817)
Middle
(n=1684)
High
(n=2326) P-value
Low
(n=2167)
Middle
(n=1940)
High
(n=1720) P-value
Demography
Male sex 82% 73% 79% <0.0001 74% 75% 85% <0.0001
Age, mean (SD)
53.2
(9.49)
52.47
(9.33)
53.17
(9.49) 0.02
54.91
(9.28)
53.22
(9.37)
50.71
(9.19) <0.0001
Major ethnic groups <0.0001 <0.0001
Punjabi 30% 39% 25% 33% 37% 24%
Urdu 38% 35% 41% 33% 35% 46%
Pathan 6% 7% 8% 8% 7% 6%
Sindhi 10% 11% 15% 14% 9% 12%
Biochemistry, medical and familial history
Diabetes 16% 13% 14% 0.1 17% 16% 11% <0.0001
Hypertension 25% 28% 32% <0.0001 26% 30% 29% 0.002
Family history of MI 16% 18% 14% 0.002 11% 17% 19% <0.0001
LDL-cholesterol
(mmol/l), mean (SD)
2.86
(1.03)
2.92
(1.08)
2.83
(1.06) 0.03
2.85
(1.1)
2.85
(1.01)
2.91
(1.05) 0.1
Waist to hip ratio,
mean (SD)
0.95
(0.07)
0.95
(0.06)
0.94
(0.06) 0.3
0.94
(0.07)
0.95
(0.07)
0.95
(0.06) 0.02
Tobacco use <0.0001
Never/ex 69% 70% 66% 68% 71% 66%
Dip naswar
only 4% 4% 6% 0.001 6% 5% 3%
Chew only 5% 6% 7% 6% 5% 7%
Smoke (and
dip/chew) 22% 20% 22% 21% 20% 24%
Socio-economic status <0.0001 <0.0001
Low 29% 34% 33% 50% 29% 16%
Middle 34% 37% 29% 33% 40% 27%
High 37% 29% 38% 17% 31% 57%
Diet
Cooking fat <0.0001 <0.0001
Oil 43% 54% 53% 44% 51% 55%
Oil & ghee 30% 20% 31% 23% 25% 33%
Ghee only 27% 26% 16% 34% 24% 12%
Dietary pattern 1 <0.0001 <0.0001
Low 100% 0% 0% 39% 35% 26%
Middle 0% 100% 0% 33% 36% 31%
High 0% 0% 100% 28% 29% 43%
Dietary pattern 2 <0.0001 <0.0001
Low 39% 33% 28% 100% 0% 0%
Middle 35% 36% 29% 0% 100% 0%
High 26% 31% 43% 0% 0% 100%
Note: Column percentages. *P-value derives from Chi2 tests of independence for categorical
variables and from a Fisher test of equality of the means (computed as a test of nullity of all the
coefficients in the regression of the continuous row variable - for example LDL-C - over
categories of the column variable - for example tobacco use).
286
List A3.1: Grouping of food items into food groups
Colour Food Groups Food items
1
Beef and mutton Beef or mutton biryani; Kata kut or organ meat; beef salan; mutton
salan; beef boti, tikka, kabab, beef shawarma and others; Nehari; Paya
2
Breads Khameri or Nan (refined bread); Paratha/Puri (oily bread); Roti safed
Aaata (unrefined bread); Roti Laal Aaata/chakki (refined bread); Roti
mixed Aaata (refined bread)
3 Cereals Daliya
4
Chicken
Chicken biryani; Chicken salan; Chicken boti, tikka, kabab, chicken roll;
chicken shawarma and others; Chicken fried, chicken breast
5
Cooked vegetables Dark green leafy vegetables and yellow vegetables; cruciferous
vegetables (Gobi, phool gobi, bang gobi, sursoon, others); other
vegetables excluding potatoes
6 Dairy products Milk; Makkhan, Margarine; Lassi; Curd
9 Eggs
10 Fish Fish Salan, Fried Fish
11 Fruits
Nimkod
12 Pickled vegetables
Potatoes French fries, potatoes
13 Pulses and nuts Daal, lobia, channa, cholay; nuts and seeds
14
Raw vegetables (salad) Dark green leafy vegetables and yellow vegetables; cruciferous
vegetables (Gobi, phool gobi, bang gobi, sursoon, others); other
vegetables excluding potatoes; tomatoes as salad; raw onion
15 Rice
8
Sweet snacks & desserts Bagarkhani; Papay; Cakes; Biscuits; Gurrh; Sakkar; honey; jam;
marmelade; Kheer; custard; Milkshakes; Ice cream; Halwa, Mithai,
Jalabe, chocholate
287
List A3.2: Definition of local food items
Bagarkhani: Slightly sweet and salty phyllo puff-pastry
Bang gobi: Cabbage
Biryani: A means of cooking rice, where rice is mixed with a curry including meat or fish, eggs and
vegetablesBoti: Pieces of meat
Channa: Chick peas
Cholay: Chick peas
Curd: Yoghurt
Daal: Lentil
Ghee: Clarified butter
Gurrh: Unrefined sugar
Halwa: Confectionary generally made from grain flour (typically semolina), oil and sugar
Hool gobi: Cauliflower
Jalabe chocolate: Confectionary made by deep-frying a kind of pretzel and then soak it in syrup
Katakut: Combination of spices, brains, liver, kidney and other organ meats
Khameri: type of Nan
Khees: Generally corn in milk gravy
Lassi: Yoghurt drink made by blending yogurt with water or milk and adding spices
Lobia: Black eyed beans
Makkhan: Butter
Mitahi: Confectionary made from unrefined sugar and milk/butter fried
Nan: Flat bread, stone baked in a clay oven (tandoor)
Nehari: Curry made usually from beef shank and more rarely lamb
Nimko: Mixture of spicy dried ingredients, which may include fried lentils, peanuts, chickpea flour
noodles, corn, vegetable oil, chickpeas, flaked rice, fried onion and curry leaves. These are flavoured
with salt and a blend of spices that may include coriander and mustard seed
Paratha: Unleavened flat bread made of whole-wheat and fried
Paya: Cow hoof
Puri: Unleavened bread prepared from aata (stone-ground whole-wheat flour) deep fried in ghee or
vegetable oil.
Roti chakki: Bread whole-wheat flour
Roti Laal Aaata: Bread whole-wheat flour
Roti mixed aaata – bread, mixed flour
Roti safed Aaata: Bread, white flour
Sakkar: Fried flour sweet
Salan: a means of cooking curry
288
References
1. Oliveira A, Rodriguez-Artalejo F, Gaio R, Santos AC, Ramos E, Lopes C. Major habitual
dietary patterns are associated with acute myocardial infarction and cardiovascular risk
markers in a southern European population. J Am Diet Assoc. 2011;111:241-250.
2. Heidemann C, Schulze MB, Franco OH, van Dam RM, Mantzoros CS, Hu FB. Dietary
patterns and risk of mortality from cardiovascular disease, cancer, and all causes in a
prospective cohort of women. Circulation. 2008;118:230-237.
3. Brunner EJ, Mosdol A, Witte DR et al. Dietary patterns and 15-y risks of major coronary
events, diabetes, and mortality. Am J Clin Nutr. 2008;87:1414-1421.
4. Hu FB, Rimm EB, Stampfer MJ, Ascherio A, Spiegelman D, Willett WC. Prospective
study of major dietary patterns and risk of coronary heart disease in men. Am J Clin
Nutr. 2000;72:912-921.
5. Iqbal R, Anand S, Ounpuu S et al. Dietary patterns and the risk of acute myocardial
infarction in 52 countries: results of the INTERHEART study. Circulation.
2008;118:1929-1937.
6. Radhika G, Sudha V, Mohan SR, Ganesan A, Mohan V. Association of fruit and
vegetable intake with cardiovascular risk factors in urban south Indians. Br J Nutr.
2008;99:398-405.
7. Donin AS, Nightingale CM, Owen CG et al. Nutritional composition of the diets of South
Asian, black African-Caribbean and white European children in the United Kingdom: the
Child Heart and Health Study in England (CHASE). Br J Nutr. 2010;104:276-285.
8. Lip GY, Malik I, Luscombe C, McCarry M, Beevers G. Dietary fat purchasing habits in
whites, blacks and Asian peoples in England--implications for heart disease prevention.
Int J Cardiol. 1995;48:287-293.
9. Yagalla MV, Hoerr SL, Song WO, Enas E, Garg A. Relationship of diet, abdominal
obesity, and physical activity to plasma lipoprotein levels in Asian Indian physicians
residing in the United States. J Am Diet Assoc. 1996;96:257-261.
10. Simmons D, Williams R. Dietary practices among Europeans and different South Asian
groups in Coventry. Br J Nutr. 1997;78:5-14.
11. Saleheen D, Zaidi M, Rasheed A et al. The Pakistan Risk of Myocardial Infarction
Study: a resource for the study of genetic, lifestyle and other determinants of
myocardial infarction in South Asia. Eur J Epidemiol. 2009;24:329-338.
12. Easton DF, Peto J, Babiker AG. Floating absolute risk: an alternative to relative risk in
survival and case-control analysis avoiding an arbitrary reference group. Stat Med.
1991;10:1025-1035.
13. Drewnowski A, Popkin BM. The nutrition transition: new trends in the global diet. Nutr
Rev. 1997;55:31-43.
289
14. Pingali, P. and United Nations. Westernization of Asian Diets and the transformation of
food systems: Implications for research and policy. EAS Working Paper No.04-
17.Agricultural and Development Economics Division.The Food and Agriculture
Organization of the United Nations . 2004.
15. Singh RB, Niaz MA, Ghosh S et al. Association of trans fatty acids (vegetable ghee)
and clarified butter (Indian ghee) intake with higher risk of coronary artery disease in
rural and urban populations with low fat consumption. Int J Cardiol. 1996;56:289-298.
16. Hakeem R, Thomas J, Badruddin SH. Food habits and nutrient density of diets of
Pakistani children living in different urban and rural settings. J Health Popul Nutr.
2002;20:255-263.
17. Wandel M, Raberg M, Kumar B, Holmboe-Ottesen G. Changes in food habits after
migration among South Asians settled in Oslo: the effect of demographic, socio-
economic and integration factors. Appetite. 2008;50:376-385.
18. Misra A, Singhal N, Sivakumar B, Bhagat N, Jaiswal A, Khurana L. Nutrition Transition
in India: Secular Trends in Dietary Intake and their Relationship to Diet-related Non-
communicable Diseases. J Diabetes. 2011;10-0407.
19. Popkin BM. The nutrition transition: an overview of world patterns of change. Nutr Rev.
2004;62:S140-S143.
20. Speedy A. Animal source foods to improve micronutrient nutrition in developing
countries. Journal of Nutrition. 2003;4048S-4054S.
290
Appendix 4: Association of the 9p21.3 locus with risk of first-
ever myocardial infarction in Pakistan
+Danish Saleheen1,2, +Myriam Alexander2, Asif Rasheed1, David Wormser2, Nicole Soranzo3,4,
Naomi Hammond3, Adam Butterworth2, Moazzam Zaidi1, Philip Haycock2, Suzannah
Bumpstead3, Simon Potter3, Hannah Blackburn3, Emma Gray3, Emanuele Di Angelantonio2,
Stephen Kaptoge2, Nabi Shah1, Maria Samuel1, Ahmedyar Janjua1, Nasir Sheikh2, Shajjia Razi
Haider1, Muhammed Murtaza1, Usman Ahmad1, Abdul Hakeem1, Muhammad Ali Memon1,
Nadeem Hayat Mallick5, Muhammad Azhar5, Abdus Samad6, Syed Zahed Rasheed6, Ali Raza
Gardezi7, Nazir Ahmed Memon8, Abdul Ghaffar8, Fazal-ur-Rehman9, Khan Shah Zaman10,
Assadullah Kundi10, Zia Yaqoob10, Liaquat Ali Cheema10, Nadeem Qamar10, Azhar Faruqui10,
Rashid Jooma11, Jawaid Hassan Niazi11, Mustafa Hussain1, Kishore Kumar1, Asim Saleem1,
Kishwar Kumar1, Muhammad Salman Daood1, Fatima Memon1, Aftab Alam Gul1, Shahid
Abbas1, Junaid Zafar1, Faisal Shahid1, Zehra Memon1, Shahzad Majeed Bhatti1, Waleed
Kayani1, Syed Saadat Ali1, Muhammad Fahim1, Muhammad Ishaq6, Philippe Frossard1,
*Panos Deloukas3, *John Danesh2 ,
+ These authors have contributed equally to this work; * These authors have contributed equally to this
work.
1Center for Non-Communicable Diseases (CNCD) Karachi Pakistan, 2Department of Public Health and
Primary Care, University of Cambridge, UK; 3Wellcome Trust Sanger Institute, Hinxton, Cambridge, UK;
4Department of Twin Research & Genetic Epidemiology, King's College London, St Thomas' Hospital
Campus, Lambeth Palace Rd, London, UK; 5Punjab Institute of Cardiology Lahore Pakistan; 6Karachi
Institute of Heart Diseases, Karachi, Pakistan; 7 Ch. Perwaiz Elahi Institute of Cardiology, Multan,
Pakistan; 8Deewan Mushtaq CCU, Liaquat University of Medical and Health Sciences, Civil Hospital;
9Red Crescent Institute of Cardiology, Hyderabad, Pakistan; 10National Institute of Cardiovascular
Diseases, Karachi, Pakistan. 11Jinnah Post-graduate Medical Centre, Karachi, Pakistan
Arterioscle Thromb Vasc Biol. 2010;30:1467-1473.
Correspondence:
Dr. Danish Saleheen
Center for Non-Communicable Diseases (CNCD),
Karachi, Pakistan &
Department of Public Health and Primary Care
University of Cambridge
Strangeways Research Laboratory
Cambridge CB1 8RN, UK
danish.saleheen@cncdpk.com,
danish.saleheen@medschl.cam.ac.uk
Text: 2188 words
Abstract: 233 words
Total number of figures and tables: 3 tables and 2 figures
291
Summary
We studied variants at the 9p21 locus in a case-control study of acute myocardial infarction (MI)
in Pakistan, and did an updated meta-analysis of published studies in people of European
ancestry to help contextualize these data. A total of 1851 patients with first-ever confirmed MI
and 1903 controls were genotyped for 89 tagging SNPs at 9p21, including the lead variant
(rs1333049) identified by the Wellcome Trust Case Control Consortium. Minor allele
frequencies and extent of linkage disequilibrium observed in Pakistanis were broadly similar to
those seen in Europeans. In the Pakistani study, six variants were associated with MI (P<10-2)
in the initial sample set, as well as in an additional 741 cases and 674 controls in whom further
genotyping was done for these variants. For Pakistanis, the odds ratio for MI was 1.13 (95% CI:
1.05-1.22; P=2 X 10-3) for each copy of the C allele at rs1333049. By comparison, meta-
analysis of studies in Europeans yielded a higher odds ratio of 1.31 (1.26-1.37) for the same
variant (P=1x10-3 for heterogeneity). Meta-analyses of 23 variants, in up to 38,250 cases and
84,820 controls, quantified odds ratios for CAD generally yielding higher odds ratios in
Europeans than in Pakistanis. This study has provided the first demonstration that variants at
the 9p21 locus are significantly associated with MI risk in Pakistanis. Association signals at this
locus were, however, considerably weaker in Pakistanis than those previously reported in
Europeans.
292
Background
Variants at the 9p21.3 locus have been established as among the strongest common genetic
factors associated with the risk of coronary artery disease (CAD) in people of European
continental ancestry.1-5 These variants are in high linkage disequilibrium (LD) and span a 58 kb
region that has multiple neighbouring genes (CDKN2A, CDNK2B and MTAP), without
annotating to any single protein coding sequence.6 An RNA coding gene, ANRIL, has been
identified that overlaps with the risk locus associated with CAD, suggesting a regulatory role in
gene expression.7 Although associations of variants at 9p21.3 with CAD have been established
in several non-European populations (e.g., East Asians), they have not been well-studied in
South Asians, populations at high risk of vascular disease.8
We report the first large-scale study of variants at the 9p21 locus in relation to risk of acute
myocardial infarction (MI) in Pakistan. This study involved 1851 patients with confirmed
diagnoses of first-ever myocardial infarction (MI) and 1903 controls from the Pakistan Risk of
Myocardial Infarction Study9 (PROMIS). Genotyping involved 89 tagging single nucleotide
polymorphisms (SNPs) at the 9p21.3 locus, including the lead variant (rs1333049) identified by
the Wellcome Trust Case Control Consortium in association with MI.10,11 To place our findings
in context, we also report a literature-based meta-analysis of relevant studies, encompassing
information on 23 variants at the 9p21 locus in up to 38,250 CAD cases and 84,820 controls.
The current meta-analysis substantially updates a previous relevant review12, involving data
from an additional 82,117 participants and 20 additional variants.
Methods
Study design This paper follows the reporting recommendations of STREGA.13 PROMIS is a
case-control study of acute first-ever MI in urban Pakistan9. MI cases had: (1) symptoms within
24 hours of hospital presentation; (2) typical ECG characteristics (e.g., 1 mm or more ST
elevation in any two or more contiguous limb leads, new onset left bundle branch block); and
(3) a positive troponin test (>1ng/ml). Controls were individuals without a history of
cardiovascular disease frequency-matched to cases by sex and age in 5 year age bands, and
concurrently identified in the same hospitals as the index cases by virtue of being: (1) visitors
of patients attending the outpatient department; (2) patients attending the outpatient
department for routine non-cardiac complaints, or (3) non-blood related visitors of index MI
cases. Controls with recent illnesses or infections were excluded. A locally-piloted and -
validated epidemiological questionnaire was administered to participants by medically qualified
research officers that sought >200 items of information in relation to: ethnicity, demographic
characteristics, lifestyle factors (e.g., tobacco and alcohol consumption, dietary intake and
physical activity); personal and family history of cardiovascular disease; and medication usage.
Non-fasting blood samples were drawn from each participant and centrifuged within 45 minutes
of venepuncture. Samples were stored at -80 ºC. The study has received approval from
293
relevant institutional review boards in each recruitment centre and the Center for Non-
Communicable Diseases (CNCD) Karachi, Pakistan. Informed consent was obtained from all
the participants (including consent to use samples in genetic, biochemical and other analyses).
Genotyping DNA was extracted from leucocytes using a reference phenol-chloroform protocol.9
Genotyping was performed at the Wellcome Trust Sanger Institute, Hinxton, UK. To minimise
any systematic biases arising from plate- or batch-specific genotyping error, genotyping plates
contained a mixture of cases and controls, including negative and positive controls. 1851 cases
and 1903 controls were genotyped using version 1 of the IBC array of about 2000 candidate
genes. 169 SNPs tagged the 9p21.3 locus at r2 > 0.8 and were available in the current
analyses. This array employed a cosmopolitan tagging approach, using information from the
Hap Map Caucasian (CEU), East Asian (Han Chinese and Japanese), and African (Yoruba)
populations14. SNPs were excluded if: (1) the call rate was <95% (2 SNPs); (2) there was
evidence of departure from Hardy-Weinberg Equilibrium in controls at a P-value of <0.05 (17
SNPs); or (3) the minor allele frequency (MAF) was <1% (63 SNPs), with most of such
omissions due to SNPs relevant for Africans being monomorphic in Pakistanis. Seven
individuals were excluded either because self-reported gender did not match chromosomal sex
status, evidence of cryptic relatedness or more than 2% missing genotypes. After such quality
control, 89 SNPs remained. The six SNPs most significantly associated with MI risk were
genotyped in a further 741 cases and 674 controls (iPLEX: Sequenom).
Statistical analysis Analyses involved PLINK 1.0615 and STATA version 10.0 (StataCorp).
Assuming an additive model, associations with each SNP were tested fitting a logistic
regression model adjusted for age, sex and the first two principal components (calculated using
all 45,000 SNPs genotyped in the array, as described previously).16,17 Effect-modification was
investigated by tests of interaction in fully adjusted models. LD was assessed using Haploview,
with blocks graphically identified from the LD intensity expressed in D’.18 Haplotype association
analyses were performed with the THESIAS software implementing the Stochastic-EM
algorithm,19 enabling simultaneous estimation of haplotypes frequencies and their effects on
MI, again assuming additive effects. Using a parsimonious approach, the most informative
tagging SNPs were chosen for haplotype analyses that accounted for at least 85% of the
haplotypic block variability.20 To compare LD patterns in Pakistanis with those in other ethnic
groups, data were downloaded from the HapMap website for Caucasians, East Asians (Han
Chinese and Japanese), and Africans (Yoruba), and drawn using Haploview.21
Systematic review We sought studies reporting on associations between variants at the 9p21
locus and risk of CAD before January 2010 (Figure A4.1). Electronic searches, not limited to
the English language, used the MEDLINE database and involved search terms related to the
locus (e.g., chromosome 9, CDKN2A, CDNK2B, MTAP, ANRIL, rs-numbers of variants
previously reported) and coronary artery disease (eg, coronary heart disease, myocardial
294
infarction, atherosclerosis, coronary stenosis). These searches were supplemented by
scanning reference lists, hand searching relevant journals, and discussion with authors. Two
investigators independently extracted the following: genotype frequencies; unadjusted additive
odds ratios; ethnicity; geographical location; CAD definition; study type (e.g., genomewide
association study, candidate replication study); genotyping platform; study design; and source
of controls. Additive odds ratios were computed for each study using genotype counts and were
compared with reported odds ratios, when available. For prospective cohorts, hazard ratios
were assumed to approximate odds ratios. Summary odds ratios and 95% confidence interval
(CI) for each SNP were calculated fitting a random-effects model that included between-study
heterogeneity. Heterogeneity was assessed by the I2 statistic22 and the Q statistic, and
investigated by pre-specified study-level characteristics, notably: study size, case definition,
study design, and genotyping platform used.
Results
Analyses in Pakistanis As would be expected, baseline levels of conventional risk factors were
significantly higher in MI cases than controls (Table A4.1). Genetic similarity did not correlate
strongly with self-reported ethnicity among the 8 ethnic and linguistic groups studied in
PROMIS (Figure A4.2) and LD patterns at the 9p21 locus were similar among the four major
Pakistani sub-ethnicities (Figure A4.3). Of the 89 relevant SNPs assessed, 6 were associated
with MI at p<10-2 (Figure A4.4 & Table A4.2), including the lead variant (rs1333049) identified
by Wellcome Trust Case Control Consortium.23 Odds ratios for MI with each of these 6 SNPs
were about 1.13 (p<10-3 for each; Table A4.3). In the case of rs1333049, the odds ratio for MI
was 1.12 (95% CI: 1.04-1.20; P=2x10-3) per C allele. Odds ratios for MI were not significantly
different under a range of circumstances (Figures A4.5), though there was limited power to
evaluate potential effect-modification (e.g., it was not possible to confirm or refute possible
differences noted in men and women and between ethnic groups for some SNPs).
There were 5 distinct LD blocks at the 9p21 locus in Pakistanis, each having strong intra-block
LD (Figure A4.6). All 6 variants and haplotypes associated with MI were located within block 3.
Using a parsimonious model, haplotypic associations observed in block 3 could be explained
by only two tagging SNPs (i.e., rs1412832 and rs1333049; Table A4.4). Haplotype analyses
involving these 2 tagging SNPs generated 4 haplotypes, each with a frequency greater than
2%, and the global test of association with MI was significant (p=0.028). The odds ratio for MI
with the AG haplotype was 0.85 (0.76-0.95; P=0.003) compared with the most frequent AC
haplotype.
Three SNPs (rs7865618, rs1292136, rs7044859) located in the genomic region of block 2 had
previously been found to modulate the effect of SNP rs1333049 on the risk of MI in Europeans,
with the two most frequent haplotypes being in “yin yang”.24 To evaluate these findings in
Pakistanis, we analysed proxy SNPs (rs518394, rs10965212 and rs7049105) based on the
295
HapMap CEU data (0.69≤r2≤0.90). We did not find any significant evidence for a 
heterogeneous effect of the rs1333049-C allele according to the haplotypic backgrounds
generated by these three proxy SNPs, or SNPs tagging block 2 (data available upon request).
Meta-analysis We identified 26 relevant studies of 23 variants at the 9p21 locus, comprising
38,250 cases and 84,820 controls (Table A4.5). 19 studies were based in Europe or the USA25-
29,30-43, 6 in East Asia44-49 and 1 in North Africa50. 5 were prospective in design51,52,53,54,55, 2
were nested case-control studies56,57 and 19 were retrospective case-control studies. 7 studies
included only MI patients, while 20 studies included patients with MI or coronary stenosis
(defined by intervention procedures). Of the 23 variants studied, 21 were associated with CAD
risk in Europeans and 6 with CAD risk in East Asians (Figure A4.7 & Figures A4.8a-w). For
most variants, there was null to moderate heterogeneity in ethnic-specific combined odds ratios
(e.g., I2 values typically ranged from 0% to 20%, with 6 SNPs with I2>50%). For Europeans
and East Asians, odds ratios for CAD were 1.31 (1.26-1.37; P<10-35) and 1.25 (1.13–1.39;
P<10-5), respectively, per C allele of rs1333049. Excluding the original report, the
corresponding odds ratio was 1.29 (1.23-1.36) in Europeans. Of the 23 variants studied, data
were available on 12 in PROMIS (Figure A4.7). Odds ratios for MI with variants at this locus
were generally lower in Pakistanis than Europeans (e.g., heterogeneity P=1x10-3 for
rs1333049), although the allelic frequencies in Pakistanis and Europeans were similar (Figure
A4.7). HapMap suggested a high degree of LD between rs1333049 and other variants studied
at the 9p21 locus, with a similar pattern of LD in Europeans and Pakistanis (Figure A4.9).
Discussion
In the first large-scale genetic study of MI in Pakistan, we investigated 89 SNPs spanning 350
kb at the 9p21.3 locus and identified 6 SNPs significantly associated with risk of MI. In an
updated literature-based meta-analysis of 38,250 CAD cases and 84,820 controls, we
confirmed associations of 21 variants at this locus with CAD in Europeans (as well as
confirming associations of 6 variants at this locus with CAD in East Asians). We observed that
odds ratios for MI with variants at the 9p21.3 locus were generally lower (typically by about half
as much) in Pakistanis than those reported in studies of people of European continental
ancestry. As discussed below, however, further studies are needed to determine whether such
differences are mainly related to ethnicity or study design features or both.
The LD structure in PROMIS was similar to the one previously observed in Europeans
(HapMap CEU) and somewhat stronger than in East Asians and Africans. The 6 SNPs
associated with MI in PROMIS were located within one block that had high intra-block LD
spanning 53kb. Two haplotype tagging SNPs (rs1333049 and rs1412832) were sufficient to
explain the observed association with MI in Pakistanis. This block partially overlapped with the
region of association previously identified in Europeans.58 Although we did not observe in
Pakistanis the previously reported significant modulation in Europeans of the effect of
296
rs1333049 on MI by either proxy or tag SNPs of haplotype block 259, this null finding may have
been due to limited statistical power and/or inadequate proxy markers in the Pakistan study.
For Pakistanis, Europeans, and East Asians, analyses of the present data and HapMap
indicated that CAD-related variants at the 9p21.3 were in high LD with one another and
localized within a region that is devoid of a single protein coding sequence.60 This locus
coincides with the ANRIL gene, a recently discovered antisense non-coding RNA postulated to
enhance gene expression.61,62,63 There are multiple neighbouring protein coding genes in this
region, including CDKN2A, CDKN2B and MTAP, proposed to play important roles in cell-cycle
progression, cellular proliferation, apoptosis and cellular senescence.64,65,66
Although allelic frequencies for most variants and LD pattern at the 9p21.3 locus were similar in
Pakistanis and Europeans, odds ratios for CAD with most variants at this locus were lower in
Pakistanis. There are previous reports of cardiometabolic diseases having differential genetic
effects by ethnicity, such as stronger associations of a LTA4H haplotype with MI in Africans
than Europeans,67 and TCF7L2 being relevant to type 2 diabetes in West Africans but not East
Asians.68
Nevertheless, further study is needed to determine whether the apparently different odds ratios
for CAD with variants at the 9p21.3 locus noted in this study of Pakistanis versus those of
Europeans are due mainly to non-ethnic factors, such as differences in case definitions, age of
disease onset, and epidemiological design. In the absence of ethnically-mediated differences in
disease susceptibility, however, the present Pakistani study might have been expected to yield
higher odds ratios than most previous studies in Europeans owing to this Pakistani study’s
presumed enrichment for genetic signals (e.g., through use of strict phenotyping and inclusion
of early-onset first-ever MI cases with high degrees of familial clustering). Such considerations,
therefore, reinforce the need for further studies of MI in Pakistanis to help validate and discover
its population-relevant genetic determinants.
Conclusion
This study has provided the first demonstration that variants at the 9p21 locus are significantly
associated with MI risk in Pakistanis. Association signals at this locus were, however,
considerably weaker in Pakistanis than those previously reported in Europeans.
297
Acknowledgements
We would like to acknowledge the contributions of the following individuals:
Epidemiological fieldwork in Pakistan: Zeeshan Ozair, Fahad Shuja, Mustafa Qadir Hameed,
Imad Hussain, Hamza Khalid, Ali Memon, Kamran Shahid, Ali Kazmi, Sana Nasim, Muhammad
Ahsan Javed, Zahir Hussain, Kanwal Aamir, Mazhar Khan, Muhammad Zafar, Faisal Majeed,
Madiha Ishaq, Turkey Hussain Marmoos, Faud Khurshid, Farhat Abdul Muntaqim, Sarosh
Fatima, Rehan Ahmed, Muhammad Nabeel, , Syed Shazad Hussain, Muhammad Zuhair Yusuf,
Nadeem Baig, Madad Ali Ujjan, Parveen Sultan, Asghar Ali, Ayaz Ali, Mir Alam, Hassan Zaib,
Abdul Ghafoor, Saeed Ahmed, Muhammad Riazuddin, Muhammad Waqar Khan, Muhammad
Irshad Javed, Jabir Furqan, Abdul Ghaffar, Muhammad Shahid, Tanveer Baig Mirza,
Muhammad Naeem, Afzal Hussain, Abdul Hakeem, Zahid Hussain, Tanveer Abbas,
Muhammad Khurram Shahzad, Iqbal Ali, Muhammad Imran Nisar, Altaf Hussain, Muhammad
Shazad, Mehmood Jafree and Ayeesha Kamal.
Laboratory assays: Sadaf Raza, Naeem Khan, Asad Ali Shah, Sobia Naz, Farina Hanif,
Shaheen Khanum, Aisha Nazir, Aisha Sultana, Mehwish Jabar, Zahid Hussain, Madiha
Yameen, Nadir Khan, Inosh Hasan.
Data management: Sarfaraz Sher Ali, Touqeer Ahmed, Syed Waqas Ahmed, Azfar Hussain,
Matthew Walker and Sarah Watson
We are grateful to Dr. David Tregouet for providing help in the haplotype analyses.
Administration: Kashif Saleheen, Hannah Sneath and Sarah Drummond.
Source of funding
Epidemiological fieldwork in PROMIS has been supported by unrestricted grants to
investigators at the University of Cambridge. Genotyping for this study was funded by the
Wellcome Trust. The British Heart Foundation has supported some biochemical assays. The
Yousef Jameel Foundation supports Dr Saleheen. The Marie Curie Bloodomics 2 scholarship
supports Myriam Alexander. The cardiovascular disease epidemiology group of Professor
Danesh is underpinned by programme grants from the British Heart Foundation, the UK
Medical Research Council and the Wellcome Trust.
Disclosure
None.
298
Figure A4.1: Literature search strategy used in the current meta-analyses
PubMed search performed
on 07 January 2010
• 909 articles
Exclusion of, in order,:
• 409 articles published before Jan 2007
• 26 articles involving non Human participants
• 56 articles performing linkage, functional or expression analysis
• 172 articles investigating the risk of quantitative trait or of
other outcomes than CHD (subclinical CHD, T2 diabetes, stroke…)
• 2 articles not population based
• 27 articles without information on chromosome 9p21
• 200 irrelevant articles (reviews, protocols, statistical
methodology…)
Extraction of data for:
• 23 articles from PubMed search
• 6 articles found by scanning
references and hand searching
Exclusion of:
• 1 article because of no available data on chromosome 9p21
(Larson 2007 BMC Med. Gen.)
• 2 articles because data overlap entirely with another published
study: ARIC study (Brautbar 2009 Circ. CVD Gen.) and Chaoyang
study (Yang 2009 Clin. Biochem.)
26 articles finally included in the
analysis:
• 3 GWA
• 25 replication studies
23 SNPs finally included in the meta-
analyses, which had been genotyped
in a total of:
• 3 GWA
• 23 replication studies
• 38,250 cases and 84,820 controls
299
Figure A4.2: Scatter plot of the first two principal components and self reported ethnicities in
PROMIS control participants
C1
C
2
Scatter plot of the first two principal components identified by principal component analysis of
the identity-by-state matrix in PROMIS IBC data on 1851 cases and 1903 controls genotyped
on 45,000 SNPs. The colours of points refer to self reported-ethnicities in PROMIS participants.
C1 and C2 axis represent to the first and second principal components.
300
Figure A4.3: Linkage disequilibrium patterns in four major Pakistani ethnic groups
There were 302 Pathan, 588 Punjabi, 280 Sindhi and 17510 Urdu. Regions of strong LD are in red, of moderate LD in light red, of very moderate LD in blue
and of no LD in white.
Pathan Punjabi
Sindhi Urdu
301
rs
96
32
88
4
Figure A4.4: Association of variants at 9p21.3 locus with myocardial infarction in PROMIS participants
a) Regional plot of association for all the 89 variants at 9p21.3 locus genotyped in PROMIS participants
(1851 cases and 1903 controls) based on r 2 values between SNPs and rs1333049. The dotted line
represents the nominal threshold of significance (P-value<10 -2). The association between SNPs and MI
was tested fitting an additive model adjusting for age, sex and the first two principal components. The
lowest part of the panel represents the exons-introns structure of annotated genes. Gene information
was downloaded from www.ucsc.edu. CDKN2BAS is also called ANRIL.
b) LD plot in PROMIS participants for the region highlighted in panel (a). SNPs associated with MI are
highlighted in black boxes and the tagging SNPs selected for haplotype analyses are in red boxes.
b)
rs
15
37
37
2
rs
28
91
16
8
rs
49
77
57
4
rs
96
32
88
5
rs
14
12
83
2
rs
10
75
72
74
rs
13
33
04
9
rs
78
55
16
2
rs
10
11
62
77
rs
10
96
52
28
rs
13
33
04
0
rs
10
75
72
72
rs
13
33
04
2
rs
10
51
17
01
rs
23
83
20
6
rs
10
96
52
35
rs
94
47
97
rs
15
37
37
5
rs
17
76
14
46
rs
13
33
04
8
C9orf53
CDKN2BAS
CDKN2B
-lo
g1
0(
p-
va
lu
e)
0
.5
1
1.
5
2
2.
5
21.9 Mb 22 Mb 22.1 Mb 22.2 Mb
r2<0.5
0.5≤r2<0.7
0.7≤r2<0.8
0.8≤r2
See b)a)
CDKN2A
302
Figure A4.5a: Effect modification of variant rs10757274 by different factors in PROMIS
participants
Subtotal (I-squared = 51.4%, p = 0.151)
Subtotal (I-squared = 0.0%, p = 0.623)
Subtotal (I-squared = 3.4%, p = 0.404)
Subtotal (I-squared = 0.0%, p = 0.980)
Subtotal (I-squared = 36.8%, p = 0.206)
Subtotal (I-squared = 0.0%, p = 0.520)
Subtotal (I-squared = 28.4%, p = 0.237)
Subtotal (I-squared = 0.0%, p = 0.372)
1.16 (1.07, 1.27)
1.00 (0.83, 1.21)
1.19 (1.05, 1.36)
1.08 (0.93, 1.26)
1.13 (0.96, 1.33)
1.22 (1.09, 1.37)
1.10 (0.91, 1.32)
1.18 (0.85, 1.64)
1.03 (0.62, 1.72)
0.92 (0.70, 1.20)
1.10 (0.67, 1.80)
1.58 (0.63, 3.94)
0.94 (0.75, 1.19)
1.13 (1.03, 1.24)
1.15 (0.92, 1.44)
1.16 (0.84, 1.59)
1.21 (1.07, 1.37)
1.28 (0.95, 1.73)
1.05 (0.93, 1.19)
1.15 (1.06, 1.26)
1.07 (0.88, 1.31)
1.15 (1.06, 1.26)
1.01 (0.83, 1.24)
1.10 (0.96, 1.28)
1.04 (0.91, 1.20)
1.24 (1.02, 1.51)
.15
.31
.58
.3
.82
.51
.07
.66
.4
.06
.87
.87
.65
.11
.49
.23
.58
.28
.53
.53
.54
.53
.52
.54
.52
.57
.55
.45
.52
.62
.50
.53
.53
.51
.54
.55
.52
.53
.52
.53
.53
.52
.52
.55
.49
.52
.49
.51
.49
.49
.5
.54
.55
.47
.50
.51
.52
.50
.50
.48
.49
.49
.51
.50
.51
.50
.53
.50
.50
.50
1.8 1.2 1.4 1.61.8 2
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
RAF
Case/cont
Group OR (95% CI) P-value
Male
Female
Urdu
Punjabi
Pathan
Balochi
Sindhi
Memon
Gujrati
Other
Recruitment centre
NICVD & JI
KIHD
PIC
Tobacco use
Never
Ex
Current
Sex
Age
<50 years old
50-60 years old
>60 years old
Ethnic group
History of diabetes
Yes
No
History of MI
Yes
No
Waist to hip ratio
Lowest third (<.92)
Middle third (0.92-.96)
Top third (>0.96)
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref: reference group; P-value: P-value of difference from the reference group; Interaction
between each variable and the genotype was assessed by introducing an interaction term in
the additive model adjusting for age, gender and self reported ethnicity. For variables with
more than two categories, an overall test of heterogeneity is reported below the estimates for
all subgroups of this variable.
303
Figure A4.5b: Effect modification of variant rs1333049 by different factors in PROMIS
participants
Ref: reference group; P-value: P-value of difference from the reference group; Interaction
between each variable and the genotype was assessed by introducing an interaction term in
the additive model adjusting for age, gender and self reported ethnicity. For variables with more
than two categories, an overall test of heterogeneity is reported below the estimates for all
subgroups of this variable.
Male
Female
(Heterogeneity I-squared = 64.3%, p = 0.094)
(Heterogeneity I-squared = 0.0%, p = 0.723)
Urdu
Punjabi
Pathan
Balochi
Sindhi
Memon
Gujrati
Other
(Heterogeneity I-squared = 0.0%, p = 0.860)
Recruitment centre
NICVD & JI
KIHD
PIC
(Heterogeneity I-squared = 0.0%, p = 0.724)
Tobacco use
Never
Ex
Current
(Heterogeneity I-squared = 49.3%, p = 0.139)
(Heterogeneity I-squared = 0.0%, p = 0.360)
(Heterogeneity I-squared = 0.0%, p = 0.364)
(Heterogeneity I-squared = 0.0%, p = 0.591)
1.16 (1.07, 1.27)
0.98 (0.82, 1.18)
1.18 (1.04, 1.34)
1.12 (0.96, 1.31)
1.09 (0.92, 1.28)
1.19 (1.06, 1.34)
1.06 (0.89, 1.27)
1.16 (0.85, 1.58)
1.23 (0.68, 2.21)
1.02 (0.76, 1.36)
1.16 (0.70, 1.92)
1.38 (0.60, 3.16)
0.99 (0.77, 1.26)
1.11 (1.02, 1.22)
1.24 (0.97, 1.58)
1.14 (0.84, 1.56)
1.21 (1.07, 1.37)
1.28 (0.95, 1.72)
1.03 (0.91, 1.17)
1.15 (1.06, 1.26)
1.04 (0.86, 1.27)
1.14 (1.05, 1.25)
1.03 (0.84, 1.27)
1.14 (0.98, 1.32)
1.05 (0.91, 1.21)
1.18 (0.98, 1.42)
.1
.6
.41
.27
.85
.91
.31
.9
.64
.18
.34
.86
.71
.07
.34
.35
.44
.75
.52
.51
.53
.51
.50
.53
.51
.55
.55
.45
.49
.6
.50
.51
.54
.49
.53
.53
.50
.52
.51
.52
.52
.51
.50
.53
.48
.51
.49
.49
.49
.48
.49
.51
.50
.45
.46
.54
.50
.49
.49
.47
.48
.47
.50
.49
.50
.48
.51
.48
.49
.49
RAF
Case/cont
1.8 1.2 1.4 1.6 1.8 2
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
./
./
/
/
/
/
/
/
/
/
Group OR (95% CI) P-value
Sex
Age
<50 years old
50-60 years old
>60 years old
Ethnic group
History of diabetes
Yes
No
History of MI
Yes
No
Waist to hip ratio
Lowest third (<.92)
Middle third (0.92-.96)
Top third (>0.96)
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref
304
Figure A4.5c: Effect modification of variant rs1537372 by different factors in PROMIS
participants
Ref: reference group; P-value: P-value of difference from the reference group; Interaction
between each variable and the genotype was assessed by introducing an interaction term in
the additive model adjusting for age, gender and self reported ethnicity. For variables with more
than two categories, an overall test of heterogeneity is reported below the estimates for all
subgroups of this variable.
Subtotal (I-squared = 35.0%, p = 0.215)
Subtotal (I-squared = 27.0%, p = 0.254)
Subtotal (I-squared = 24.4%, p = 0.235)
Subtotal (I-squared = 0.0%, p = 0.913)
Subtotal (I-squared = 59.4%, p = 0.085)
Subtotal (I-squared = 1.6%, p = 0.313)
Subtotal (I-squared = 0.0%, p = 0.429)
Subtotal (I-squared = 0.0%, p = 0.377)
1.15 (1.06, 1.26)
1.01 (0.84, 1.22)
1.21 (1.06, 1.38)
1.03 (0.89, 1.19)
1.14 (0.96, 1.34)
1.23 (1.10, 1.39)
1.09 (0.91, 1.30)
1.08 (0.80, 1.45)
1.25 (0.67, 2.35)
0.88 (0.68, 1.14)
1.11 (0.68, 1.81)
1.50 (0.64, 3.48)
0.93 (0.74, 1.17)
1.13 (1.03, 1.24)
1.16 (0.93, 1.46)
1.07 (0.81, 1.44)
1.21 (1.08, 1.37)
1.33 (0.97, 1.80)
1.02 (0.91, 1.16)
1.16 (1.06, 1.26)
1.04 (0.86, 1.26)
1.14 (1.05, 1.25)
1.05 (0.85, 1.28)
1.13 (0.98, 1.30)
1.04 (0.91, 1.20)
1.23 (1.02, 1.49)
.21
.09
.52
.22
.37
.96
.03
.64
.51
.04
.79
.73
.51
.05
.3
.41
.43
.41
.50
.50
.52
.49
.49
.52
.49
.53
.52
.41
.51
.59
.48
.50
.51
.48
.52
.52
.49
.51
.50
.50
.51
.49
.50
.52
.47
.5
.47
.49
.46
.46
.47
.52
.48
.44
.49
.49
.50
.47
.48
.47
.47
.45
.49
.47
.49
.47
.5
.47
.48
.47
0
1.8 1.2 1.4 1.6 1.8 2
Male
Female
Urdu
Punjabi
Pathan
Balochi
Sindhi
Memon
Gujrati
Other
Recruitment centre
NICVD & JI
KIHD
PIC
Tobacco use
Never
Ex
Current
Sex
Age
<50 years old
50-60 years old
>60 years old
Ethnic group
History of diabetes
Yes
No
History of MI
Yes
No
Waist to hip ratio
Lowest third (<.92)
Middle third (0.92-.96)
Top third (>0.96)
RAF
Case/cont
Group OR (95% CI) P-value
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref
305
Figure A4.5d: Effect modification of variant rs2891168 by different factors in PROMIS
participants
Ref: reference group; P-value: P-value of difference from the reference group; Interaction
between each variable and the genotype was assessed by introducing an interaction term in
the additive model adjusting for age, gender and self reported ethnicity. For variables with more
than two categories, an overall test of heterogeneity is reported below the estimates for all
subgroups of this variable.
Subtotal (I-squared = 44.9%, p = 0.178)
Subtotal (I-squared = 0.0%, p = 0.522)
Subtotal (I-squared = 7.9%, p = 0.369)
Subtotal (I-squared = 0.0%, p = 0.995)
Subtotal (I-squared = 52.5%, p = 0.122)
Subtotal (I-squared = 0.0%, p = 0.417)
Subtotal (I-squared = 42.0%, p = 0.189)
Subtotal (I-squared = 0.0%, p = 0.403)
1.17 (1.07, 1.27)
1.01 (0.84, 1.22)
1.21 (1.06, 1.37)
1.08 (0.93, 1.25)
1.13 (0.96, 1.34)
1.23 (1.09, 1.38)
1.10 (0.92, 1.32)
1.18 (0.85, 1.63)
1.07 (0.63, 1.80)
0.91 (0.70, 1.19)
1.11 (0.67, 1.82)
1.60 (0.63, 4.04)
0.94 (0.75, 1.19)
1.14 (1.04, 1.24)
1.14 (0.92, 1.43)
1.15 (0.84, 1.58)
1.22 (1.08, 1.38)
1.32 (0.97, 1.79)
1.04 (0.92, 1.18)
1.16 (1.06, 1.27)
1.06 (0.87, 1.29)
1.16 (1.06, 1.27)
1.00 (0.82, 1.22)
1.13 (0.98, 1.30)
1.05 (0.91, 1.21)
1.24 (1.02, 1.51)
.17
.24
.53
.29
.78
.58
.06
.66
.38
.05
.95
.92
.56
.07
.39
.19
.49
.38
.53
.53
.54
.53
.52
.54
.52
.57
.54
.44
.51
.62
.51
.53
.53
.51
.54
.55
.52
.53
.52
.53
.53
.52
.52
.55
.49
.52
.49
.51
.49
.49
.50
.54
.53
.47
.49
.50
.53
.50
.50
.48
.49
.48
.51
.50
.51
.49
.53
.49
.50
.50
1.8 1.2 1.4 1.61.8 2
RAF
Case/cont
Group OR (95% CI) P-value
Male
Female
Urdu
Punjabi
Pathan
Balochi
Sindhi
Memon
Gujrati
Other
Recruitment centre
NICVD & JI
KIHD
PIC
Tobacco use
Never
Ex
Current
Sex
Age
<50 years old
50-60 years old
>60 years old
Ethnic group
History of diabetes
Yes
No
History of MI
Yes
No
Waist to hip ratio
Lowest third (<.92)
Middle third (0.92-.96)
Top third (>0.96)
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref
306
Figure A4.5e: Effect modification of variant rs4977574 by different factors in PROMIS
participants
Ref: reference group; P-value: P-value of difference from the reference group; Interaction
between each variable and the genotype was assessed by introducing an interaction term in
the additive model adjusting for age, gender and self reported ethnicity. For variables with
more than two categories, an overall test of heterogeneity is reported below the estimates for
all subgroups of this variable.
Subtotal (I-squared = 46.9%, p = 0.170)
Subtotal (I-squared = 0.0%, p = 0.544)
Subtotal (I-squared = 13.8%, p = 0.322)
Subtotal (I-squared = 0.0%, p = 0.991)
Subtotal (I-squared = 48.7%, p = 0.143)
Subtotal (I-squared = 0.0%, p = 0.431)
Subtotal (I-squared = 39.9%, p = 0.197)
Subtotal (I-squared = 0.0%, p = 0.406)
1.16 (1.07, 1.27)
1.01 (0.84, 1.21)
1.20 (1.06, 1.37)
1.08 (0.93, 1.25)
1.13 (0.95, 1.33)
1.23 (1.09, 1.38)
1.10 (0.91, 1.32)
1.18 (0.85, 1.63)
1.00 (0.61, 1.64)
0.90 (0.70, 1.18)
1.13 (0.68, 1.89)
1.60 (0.63, 4.04)
0.94 (0.75, 1.19)
1.13 (1.04, 1.24)
1.14 (0.92, 1.43)
1.15 (0.84, 1.58)
1.22 (1.08, 1.38)
1.30 (0.96, 1.76)
1.04 (0.92, 1.18)
1.16 (1.06, 1.26)
1.06 (0.87, 1.29)
1.16 (1.06, 1.26)
1.00 (0.82, 1.22)
1.12 (0.97, 1.29)
1.05 (0.91, 1.21)
1.24 (1.02, 1.50)
.17
.26
.5
.27
.78
.42
.05
.74
.38
.05
.92
.9
.62
.07
.41
.2
.51
.37
.53
.53
.54
.53
.52
.54
.52
.57
.54
.44
.52
.62
.51
.53
.53
.51
.54
.55
.52
.53
.52
.53
.53
.52
.52
.55
.49
.52
.49
.51
.49
.49
.50
.54
.55
.47
.49
.50
.53
.50
.50
.48
.49
.49
.51
.50
.51
.50
.53
.49
.51
.50
1.8 1.2 1.4 1.61.8 2
RAF
Case/cont
Group OR (95% CI) P-value
Male
Female
Urdu
Punjabi
Pathan
Balochi
Sindhi
Memon
Gujrati
Other
Recruitment centre
NICVD & JI
KIHD
PIC
Tobacco use
Never
Ex
Current
Sex
Age
<50 years old
50-60 years old
>60 years old
Ethnic group
History of diabetes
Yes
No
History of MI
Yes
No
Waist to hip ratio
Lowest third (<.92)
Middle third (0.92-.96)
Top third (>0.96)
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref
307
Figure A4.5f: Effect modification of variant rs9632885 by different factors in PROMIS
participants
Subtotal (I-squared = 75.7%, p = 0.042)
Subtotal (I-squared = 27.5%, p = 0.252)
Subtotal (I-squared = 31.5%, p = 0.176)
Subtotal (I-squared = 0.0%, p = 0.838)
Subtotal (I-squared = 75.9%, p = 0.016)
Subtotal (I-squared = 0.0%, p = 0.760)
Subtotal (I-squared = 0.0%, p = 0.771)
Subtotal (I-squared = 0.0%, p = 0.937)
1.16 (1.07, 1.27)
0.95 (0.80, 1.13)
1.23 (1.08, 1.40)
1.04 (0.90, 1.20)
1.17 (0.98, 1.39)
1.23 (1.10, 1.38)
1.17 (0.96, 1.42)
0.93 (0.72, 1.21)
1.11 (0.63, 1.95)
0.92 (0.71, 1.18)
0.90 (0.60, 1.34)
1.35 (0.64, 2.85)
0.95 (0.75, 1.20)
1.12 (1.02, 1.22)
1.10 (0.89, 1.37)
1.24 (0.88, 1.75)
1.25 (1.10, 1.41)
1.30 (0.96, 1.76)
0.99 (0.88, 1.11)
1.14 (1.04, 1.24)
1.10 (0.89, 1.35)
1.13 (1.04, 1.24)
1.10 (0.88, 1.36)
1.10 (0.95, 1.26)
1.10 (0.95, 1.28)
1.14 (0.95, 1.37)
.05
.09
.61
.61
.07
.7
.06
.18
.74
.06
.89
.49
.74
.01
.74
.75
.97
.71
.57
.56
.58
.57
.56
.58
.58
.58
.58
.49
.52
.63
.54
.57
.57
.58
.58
.59
.56
.57
.57
.57
.57
.56
.57
.58
.53
.57
.53
.56
.52
.53
.54
.60
.55
.51
.55
.55
.56
.54
.55
.53
.53
.53
.57
.54
.55
.54
.55
.54
.54
.55
0
1.8 1.2 1.4 1.6 1.8 2
Male
Female
Urdu
Punjabi
Pathan
Balochi
Sindhi
Memon
Gujrati
Other
Recruitment centre
NICVD & JI
KIHD
PIC
Tobacco use
Never
Ex
Current
Sex
Age
<50 years old
50-60 years old
>60 years old
Ethnic group
History of diabetes
Yes
No
History of MI
Yes
No
Waist to hip ratio
Lowest third (<.92)
Middle third (0.92-.96)
Top third (>0.96)
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
./
./
/
/
/
/
/
/
/
/
RAF
Case/cont
Group OR (95% CI) P-value
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref
Ref: reference group; P-value: P-value of difference from the reference group; Interaction
between each variable and the genotype was assessed by introducing an interaction term in
the additive model adjusting for age, gender and self reported ethnicity. For variables with
more than two categories, an overall test of heterogeneity is reported below the estimates for
all subgroups of this variable.
308
Figure A4.6: Linkage disequilibrium in Pakistanis
1 2 3 4 5
rs
13
33
04
9
rs
96
28
84
rs
49
77
75
6
rs
10
75
72
81
rs
49
77
76
1
rs
78
54
62
9
rs
10
75
72
88
rs
10
95
19
7
rs
10
70
36
65
6
rs
32
18
02
0
Block 1 extended from rs1095197 to rs107036656, block 2 from rs3218020 to rs4977756, block 3 from rs9632884 to rs1333049, block 4 from rs10757281 to
rs4977761 and block 5 from rs7854629 to rs10757288. The number of selected tag SNPs were 6 for block 1 (rs10965197 , rs1077261, rs7041637, rs3731246,
rs2811708, rs3731239), 5 for block 2 (rs3218002, rs2069418, rs10738604, rs11790231, rs10965224), 8 for block 3 (rs9632885, rs1412832, rs10965228,
rs1333040, rs10757274, rs1537372, rs1333042 and rs1333049), 9 for block 4 (rs10757281, rs10965241, rs10965243, rs10965244, rs2383208, rs7045889,
rs10217762, rs10811661, rs11791416, rs4977761) and 6 for block 5 (rs7854629, rs2065505, rs215283, rs7856219, rs10965256, rs7853123). Colours of the
blocks are blue or bright red if D’=1 and white or shades of red otherwise.
309
Figure A4.7: Association with MI in Pakistanis, Europeans and East Asians for 12 variants
genotyped in PROMIS
rs518394
rs10116277
rs1333040
rs10757272
rs10757274
rs4977574
rs2891168
rs1333042
rs2383206
rs1333048
rs1333049
rs10757283
PROMIS
White
PROMIS
White
PROMIS
White
PROMIS
White
PROMIS
White
East Asian
PROMIS
White
PROMIS
White
PROMIS
White
PROMIS
White
East Asian
PROMIS
White
PROMIS
White
East Asian
PROMIS
White
1
3
1
6
1
6
1
3
1
14
2
1
7
1
8
1
5
1
12
3
1
9
1
7
2
1
3
1851
6916
1848
8838
1851
8838
1849
6916
2591
11829
728
2584
11528
2586
8839
1847
8912
1851
10893
2088
1850
13079
2587
8455
1283
1793
7089
1903
8950
1902
17210
1903
17210
1903
8963
2575
43841
397
2576
16289
2576
10531
1902
10339
1903
28714
1757
1902
16956
2573
11006
1857
1853
9511
0.98 (0.88, 1.08)
1.22 (1.16, 1.28)
1.10 (1.00, 1.21)
1.25 (1.20, 1.30)
1.08 (0.99, 1.19)
1.23 (1.18, 1.28)
1.13 (1.03, 1.24)
1.30 (1.22, 1.39)
1.13 (1.05, 1.23)
1.27 (1.18, 1.35)
1.38 (1.12, 1.69)
1.13 (1.05, 1.23)
1.29 (1.24, 1.34)
1.14 (1.05, 1.23)
1.23 (1.14, 1.33)
1.12 (1.02, 1.23)
1.32 (1.26, 1.39)
1.11 (1.02, 1.22)
1.27 (1.19, 1.36)
1.26 (1.05, 1.50)
1.12 (1.02, 1.23)
1.29 (1.25, 1.33)
1.13 (1.05, 1.22)
1.31 (1.26, 1.37)
1.25 (1.13, 1.39)
1.01 (0.93, 1.11)
1.07 (0.92, 1.24)
6x10-01
8x10-16
4x10-02
1x10-25
9x10-02
2x10-22
1x10-02
5x10-17
1x10-03
2x10-12
1x10-03
1x10-03
9x10-38
1x10-03
11x10-07
2x10-02
7x10-30
2x10-02
1x10-13
5x10-03
1x10-02
2x10-50
2x10-03
4x10-35
1x10-05
7x10-01
2x10-01
<1x10-3
3x10-2
2x10-2
1x10-2
3x10-2
8x10-2
4x10-3
2x10-1
2x10-3
2x10-3
2x10-1
6x10-3
1x10-3
1x10-1
6x10-1
Ethnicity
(N studies)
N cases / controls P-value RAF
P-value
het
Odds ratio
(95% CI)
1 1.2 1.4 1.6
0.72
0.55
0.58
0.45
0.61
0.52
0.56
0.48
0.50
0.47
0.45
0.50
0.47
0.50
0.48
0.59
0.49
0.54
0.50
0.47
0.52
0.48
0.51
0.48
0.49
0.58
0.45
Odds ratio (95% CI)
P-het: P-value for heterogeneity; RAF: Risk allele frequency in controls (in %). Odds ratios in
PROMIS were computed using an additive model adjusted for age, sex and the first two
principal components. Analyses for rs1333409, rs10757274, rs4977574 and rs2891168 are
adjusted for age, gender and self-reported ethnicity. Effect estimates in Europeans and East
Asians are derived from literature based meta-analyses The P-value heterogeneity
corresponds to a heterogeneity test comparing effect estimates in the different ethnic groups.
Individual plots for each meta-analysis are presented in efigures 6a-w.
310
Figures A4.8: Meta-analyses of the 23 SNPs previously published in the literature
Subtotal (I-squared = 0.0%, p = 0.971)
Ye - Bruneck
PROMIS
name
Schunkert (CardiogG) PRIME
Ghazouani - Tunisia
Study
Schunkert (CardioG) LMDS
Hiura - Suita study
Broadbent PROCARDIS
Hinohara - Tokyo
Hinohara - SMC2
Schunkert (CardioG) AtheroG
Samani - WTCCC
Samani - German MI
56
2587
cases
525
296
N
483
589
4251
604
679
370
1926
844
713
2573
cont
520
324
N
442
2475
4443
1151
706
345
2938
1605
1.39 (0.96, 2.00)
1.13 (1.05, 1.22)
ES (95% CI)
1.26 (1.06, 1.50)
1.11 (0.89, 1.38)
1.21 (1.12, 1.31)
1.34 (1.11, 1.62)
1.16 (1.02, 1.31)
1.27 (1.19, 1.36)
1.31 (1.14, 1.52)
1.18 (1.02, 1.38)
1.29 (1.23, 1.36)
1.30 (1.05, 1.60)
1.37 (1.27, 1.49)
1.33 (1.18, 1.51)
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs1333049 (C)
0.51
controls
0.52
0.51
% ra
0.48
0.49
0.48
0.49
0.48
0.46
0.47
0.48
East-Asian
Subtotal (I-squared = 50.9%, p = 0.026)
Subtotal (I-squared = 62.0%, p = 0.048)
Subtotal (I-squared = 42.2%, p = 0.159)
Other/mixed
Assimes (ADVANCE) Black
McPherson - ARIC
McPherson - OHS-1
Assimes (ADVANCE) Hispanic
Zee - PHS
Abdullah - GeneQuest
Broadbent PROCARDIS
McPherson - CCHS
Talmud - NPHS2
name
Zhang - Chaoyang
Shen - SMC2
McPherson - OHS-3
PROMIS
Study
Zhou - Wuhan
Assimes (ADVANCE) East Asian
McPherson - DHS
Shen - VHP
Assimes (ADVANCE) MB
Assimes (ADVANCE) MNB
Assimes (ADVANCE) White
McPherson - OHS-2
333
8532
312
335
560
4443
9053
2430
cont
430
294
842
1903
1360
103
527
308
110
1046
326
1.03 (0.75, 1.42)
1.15 (1.05, 1.25)
1.63 (1.31, 2.04)
1.79 (1.15, 2.80)
1.13 (0.74, 1.72)
1.61 (1.33, 1.97)
1.26 (1.18, 1.35)
1.12 (1.04, 1.21)
1.17 (0.98, 1.40)
ES (95% CI)
1.44 (1.18, 1.74)
1.30 (1.10, 1.60)
1.24 (1.07, 1.44)
1.11 (1.02, 1.22)
1.25 (1.18, 1.33)
1.30 (1.12, 1.52)
1.13 (1.02, 1.25)
1.68 (1.13, 2.49)
1.30 (1.01, 1.68)
1.30 (1.00, 1.60)
0.92 (0.57, 1.49)
1.15 (0.89, 1.48)
1.09 (0.78, 1.52)
1.33 (1.18, 1.50)
1.45 (1.16, 1.82)
0.44
0.51
0.52
0.51
0.56
0.50
0.51
0.47
0.51
controls
0.41
0.44
0.52
0.54
% ra
0.47
0.49
0.51
0.56
0.45
0.53
0.51
0.51
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs2383206 (G)
96
1154
322
108
335
310
4251
1525
270
cases
432
611
651
1851
N
1360
117
154
416
72
215
1201
304
82
N
60
Other/mixed
East-Asian
GWA
White
White
GWA
Subtotal (I-squared = 0.0%, p = 0.405)
a)
b)
311
Subtotal (I-squared = 0.0%, p = 0.795)
White
Samani - German MI
Schaefer - PopGen 2
Helgadottir - Durham
name
Helgadottir - Atlanta
Samani - WTCCC
PROMIS
Helgadottir - Philadelphia
Broadbent PROCARDIS
Study
Schaefer - PopGen 1
Helgadottir - Iceland A+B
844
1096
1104
cases
582
1924
1850
557
4251
N
823
1898
1622
733
708
cont
859
2934
1902
486
4443
N
671
4500
1.33 (1.18, 1.50)
1.26 (1.11, 1.44)
1.22 (1.07, 1.40)
ES (95% CI)
1.25 (1.08, 1.45)
1.27 (1.23, 1.32)
1.36 (1.25, 1.48)
1.12 (1.02, 1.23)
1.36 (1.14, 1.62)
1.28 (1.19, 1.37)
1.38 (1.20, 1.60)
1.22 (1.13, 1.32)
0.48
0.45
0.49
controls
0.51
0.49
0.52
0.50
0.49
% ra
0.47
0.45
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs1333048 (C)
SNP rs10757274 (G)
Subtotal (I-squared = 64.1%, p = 0.001)
Other/mixed
Subtotal (I-squared = 0.0%, p = 0.530)
East-Asian
Subtotal (I-squared = 14.6%, p = 0.319)
Assimes (ADVANCE) White
McPherson - OHS-3
Assimes (ADVANCE) Hispanic
Assimes (ADVANCE) MNB
Paynter - WGHS
McPherson - OHS-1
Assimes (ADVANCE) Black
Shen - VHP
McPherson - DHS
Shen - SMC2
Study
Broadbent PROCARDIS
Talmud - NPHS2
name
PROMIS
McPherson - OHS-2
Zee - PHS
Assimes (ADVANCE) MB
Abdullah - GeneQuest
Assimes (ADVANCE) EA
Deghan - Rotterdam
McPherson - CCHS
Zhang - Chaoyang
McPherson - ARIC
1201
647
108
215
469
322
96
416
154
611
N
4251
270
cases
2591
304
335
72
310
117
588
1525
432
1037
1046
847
82
110
10660
312
333
308
527
294
N
4443
2430
cont
2575
326
335
60
560
103
5251
9053
430
7743
1.33 (1.18, 1.49)
1.34 (1.15, 1.55)
1.61 (1.04, 2.51)
1.25 (1.17, 1.33)
1.24 (0.89, 1.72)
1.15 (1.01, 1.31)
1.66 (1.33, 2.08)
1.14 (0.79, 1.64)
1.25 (1.01, 1.55)
1.34 (1.04, 1.74)
1.30 (1.10, 1.60)
1.38 (1.21, 1.57)
1.29 (1.20, 1.38)
1.28 (1.07, 1.53)
ES (95% CI)
1.20 (0.97, 1.49)
1.13 (1.05, 1.23)
1.44 (1.15, 1.80)
0.84 (0.58, 1.22)
0.87 (0.53, 1.42)
1.73 (1.42, 2.10)
1.66 (1.12, 2.45)
1.03 (0.90, 1.18)
1.17 (1.08, 1.26)
1.40 (1.15, 1.70)
1.16 (1.06, 1.28)
0.49
0.48
0.49
0.47
0.49
0.24
0.52
0.48
0.44
% ra
0.48
0.48
controls
0.50
0.49
0.47
0.33
0.47
0.50
0.45
0.41
0.49
1.8 1.2 1.4 1.6 1.8 2 2.2
GWA
GWA
White
c)
d)
312
GWA
White
PROMIS
Subtotal (I-squared = 0.0%, p = 0.501)
name
Broadbent PROCARDIS
Samani - German MI
Samani - WTCCC
Study
cases
4251
1851
743
1922
N
cont
4443
1903
1575
2932
N
ES (95% CI)
1.20 (1.12, 1.28)
0.98 (0.88, 1.08)
1.26 (1.11, 1.43)
1.23 (1.13, 1.33)
1.21 (1.14, 1.29)
controls
0.56
0.72
0.55
0.54
% ra
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs518394 (G)
Subtotal (I-squared = 58.6%, p = 0.025)
White
Samani - WTCCC
Study
Broadbent (PROCARDIS) Germany
Schaefer - PopGen 2
Broadbent (PROCARDIS) UK
Broadbent (PROCARDIS) Sweden
Schaefer - PopGen 1
Broadbent (PROCARDIS) Italy
name
PROMIS
Samani - German MI
1926
N
436
1096
325
3010
818
480
cases
2586
748
2938
N
524
733
571
2829
777
519
cont
2576
1640
1.35 (1.24, 1.46)
1.21 (1.11, 1.31)
1.26 (1.02, 1.55)
1.30 (1.14, 1.49)
1.26 (1.00, 1.59)
1.28 (1.18, 1.39)
1.07 (0.93, 1.21)
1.36 (1.11, 1.67)
ES (95% CI)
1.14 (1.05, 1.23)
1.03 (0.91, 1.17)
0.48
% ra
0.44
0.52
controls
0.50
0.47
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs2891168 (G)
Subtotal (I-squared = 20.4%, p = 0.285)
White
Study
Broadbent PROCARDIS
Helgadottir - Atlanta
Helgadottir - Philadelphia
Helgadottir - Iceland B
Helgadottir - Durham
PROMIS
Helgadottir - Iceland A
name
N
4251
596
582
665
1137
1851
1607
cases
N
4443
1284
504
3533
718
1903
6728
cont
1.23 (1.16, 1.30)
1.22 (1.14, 1.31)
1.39 (1.19, 1.62)
1.33 (1.11, 1.59)
1.18 (1.05, 1.33)
1.13 (0.98, 1.30)
1.08 (0.99, 1.19)
1.23 (1.14, 1.33)
ES (95% CI)
% ra
0.56
0.57
0.58
0.50
0.59
0.61
0.49
controls
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs1333040 (T)
GWA
GWA
e)
f)
g)
313
Other/mixed
Subtotal (I-squared = 74.5%, p = 0.020)
Subtotal (I-squared = 0.0%, p = 0.720)
Helgadottir - Durham
Study
Helgadottir - Philadelphia
PROMIS
name
Samani - WTCCC
Helgadottir - Iceland A+B
Helgadottir - Atlanta
MIGC - MIGen
MIGC - PennCATH-MedSTAR
MIGC - deCODE
Broadbent PROCARDIS
Samani - German MI
1132
N
569
2584
cases
1924
2215
577
2971
4251
860
714
N
495
2576
cont
2937
4806
1254
3014
4443
1640
1.23 (1.08, 1.41)
1.38 (1.16, 1.64)
1.13 (1.05, 1.23)
1.37 (1.20, 1.56)
ES (95% CI)
1.35 (1.24, 1.46)
1.25 (1.17, 1.35)
1.27 (1.22, 1.33)
1.29 (1.12, 1.48)
1.25 (1.16, 1.34)
1.64 (1.37, 1.96)
1.34 (1.20, 1.49)
1.23 (1.11, 1.36)
1.36 (1.21, 1.53)
0.47
% ra
0.49
0.50
controls
0.48
0.44
0.50
0.49
0.47
1.6 .8 1.2 1.4 1.61.8 2 2.2
SNP rs4977574 (G)
GWA
Subtotal (I-squared = 20.0%, p = 0.290)
White
Samani - German MI
name
Study
Samani - WTCCC
Broadbent PROCARDIS
Schaefer - PopGen 2
PROMIS
Schaefer - PopGen 1
827
cases
N
1921
4251
1096
1847
817
1573
cont
N
2920
4443
733
1902
670
1.38 (1.22, 1.55)
ES (95% CI)
1.32 (1.23, 1.41)
1.34 (1.23, 1.45)
1.22 (1.11, 1.36)
1.33 (1.16, 1.51)
1.12 (1.02, 1.23)
1.42 (1.23, 1.63)
0.48
0.49
0.50
0.45
0.59
0.46
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs1333042 (G)
White
Subtotal (I-squared = 41.8%, p = 0.190)
GWA
PROMIS
Broadbent PROCARDIS
name
Samani - WTCCC
Samani - German MI
Study
1849
4251
cases
1910
755
N
1903
4443
cont
2884
1636
N
1.13 (1.03, 1.24)
1.22 (1.10, 1.35)
ES (95% CI)
1.33 (1.23, 1.45)
1.28 (1.15, 1.42)
1.36 (1.20, 1.54)
0.56
0.49
0.49
0.47
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs10757272 (T)
White
GWA
h)
i)
j)
% ra
controls
% ra
controls
314
Subtotal (I-squared = 55.7%, p =
White
Helgadottir - Iceland A+B
Helgadottir - Atlanta
Helgadottir - Philadelphia
Helgadottir - Durham
2099
564
567
1070
2350
1193
486
687
1.10 (1.01, 1.21)
1.34 (1.13, 1.58)
1.38 (1.12, 1.68)
1.21 (1.09, 1.35)
1.16 (0.99, 1.35)
.69
.71
.72
.71
1. .8 1.2 1.4 1. 1.8 2 2.2
SNP rs6475608 (C)
Study name No
cases
No
controls
ES
(95% CI)
Subtotal (I-squared = 81.5%, p = 0.000)
Study
Helgadottir - Iceland A+B
Helgadottir -
Broadbent PROCARDIS
Helgadottir - Durham
Helgadottir - Philadelphia
name
N
2204
709
4251
1125
556
cases
N
4761
1247
4443
697
491
cont
1.36 (1.22, 1.52)
1.26 (1.17, 1.35)
1.76 (1.55, 2.01)
1.28 (1.20, 1.37)
1.27 (1.11, 1.45)
1.36 (1.14, 1.62)
ES (95% CI)
%
0.4
0.4
0.4
0.4
0.4
controls
1.6 .8 1.2 1. 1.6 1.8 2 2.
SNP rs10738607
Whit
Subtotal (I-squared = 67.7%, p = 0.026)
name
Helgadottir -
Study
Helgadottir - Iceland A+B
Helgadottir - Durham
Helgadottir - Atlanta
case
465
N
1866
1093
569
cont
410
N
2113
694
834
ES (95% CI)
1.48 (1.22, 1.79)
1.20 (1.09, 1.31)
1.22 (1.08, 1.38)
1.04 (0.90, 1.19)
1.26 (1.09, 1.47)
controls
0.42
% ra
0.37
0.44
0.41
1.6 . 1. 1.4 1. 1.8 2 2.2
SNP rs10811650 (G)
White
p)
q)
r)
315
Subtotal (I-squared = 0.0%, p = 0.640)
GWA
Subtotal (I-squared = 57.2%, p = 0.013)
Other/mixed
Subtotal (I-squared = 46.7%, p = 0.131)
White
Assimes (ADVANCE) H
Lemmens - Caregene
Zhang - Chaoyang
Abdullah - GeneQuest
Assimes (ADVANCE) Black
Helgadottir - Iceland B
Assimes (ADVANCE) EA
Helgadottir - Atlanta
Assimes (ADVANCE) White
name
Helgadottir - Iceland A
Shen - VHP
Assimes (ADVANCE) MNB
Broadbent PROCARDIS
Helgadottir - Durham
Shen - SMC2
Deghan - Rotterdam
Helgadottir - Philadelphia
Chen - China
Study
108
926
432
72
310
96
665
117
596
1201
cases
1607
416
215
4251
1110
611
588
582
212
N
82
828
430
60
560
333
3533
103
1284
1046
cont
6728
308
110
4443
693
294
5251
504
232
N
1.61 (1.05, 2.48)
1.34 (1.18, 1.54)
1.28 (1.05, 1.57)
1.32 (1.09, 1.61)
1.30 (1.22, 1.39)
1.08 (0.64, 1.85)
1.75 (1.43, 2.14)
1.30 (0.89, 1.88)
1.32 (1.13, 1.54)
1.56 (1.05, 2.32)
1.32 (1.15, 1.52)
1.32 (1.17, 1.48)
ES (95% CI)
1.30 (1.18, 1.43)
1.28 (1.00, 1.60)
1.19 (0.86, 1.65)
1.27 (1.19, 1.37)
1.27 (1.11, 1.46)
1.29 (1.10, 1.60)
1.03 (0.90, 1.18)
1.39 (1.17, 1.66)
1.45 (1.23, 1.71)
1.85 (1.42, 2.40)
0.48
0.48
0.42
0.28
0.45
0.22
0.44
0.49
0.48
0.47
controls
0.43
0.50
0.46
0.47
0.46
0.46
0.47
0.41
% ra
1.8 1.2 1.4 1.6 1.8 2 2.2
SNP rs10757278 (G)
Assimes (ADVANCE) MB
East Asian
East-Asian
White
GWA
Study
Broadbent PROCARDIS
WTCCC - WTCCC
Yang - Chaoyang
name
N
4251
1925
432
cases
N
4443
2937
430
cont
1.07 (0.96, 1.18)
0.97 (0.90, 1.06)
1.08 (0.87, 1.35)
ES (95% CI)
% ra
0.47
0.47
0.23
controls
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs7027989 (A)
Subtotal (I-squared = 0.0%, p = 0.808)
White
GWA
name
Samani - WTCCC
Study
Broadbent PROCARDIS
Samani - German MI
cases
1926
N
4251
761
cont
2938
N
4443
1636
ES (95% CI)
1.27 (1.17, 1.38)
1.21 (1.13, 1.30)
1.23 (1.08, 1.40)
1.22 (1.14, 1.29)
controls
0.56
% ra
0.58
0.58
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs564398 (T)
s)
t)
u)
316
% ra: frequency of the risk allele in controls, N cont: number of controls, N cases: number of
cases; SNP names are followed by the name of the risk allele. Weights are from a random
effect meta-analysis. Assimes H: Assimes Hispanic; Assimes MB: Assimes Mixed Blacks;
Assimes MNB: Assimes Mixed non Blacks; Assimes EA: Assimes East Asians. Study
acronyms: ADVANCE: Atherosclerosis Disease Vascular FuNction & GenetiC Epidemiology;
PROCARDIS; AMC-PAS: Academic Medical Centre Amsterdam Premature Atherosclerosis
Study; ECTIM: Etude Cas-Temoin sur l’Infarctus du Myocarde; EPIC: European Prospective
Investigation into Cancer and Nutrition Study; GerMIFS: German MI Family Study; LURIC:
Ludwigshafen Risk and Cardiovascular Heatlh; PopGen: PopGen biobank; UKMI: UK MI; SMC:
Samsung Medical Centre; OHS: Ottawa Heart Study,; ARIC: Atherosclerosis Risk in
Community; CCHS: Copenhagen City Heart Study; DHS: Dallas Heart Study; AMI Gene: Acute
MI Gene; VHS: Verona Heart Study; MAHI: Mid America Heart Institute IFS: Irish Family Study;
PennCATH: University of Pennsylvania Catheterization Study; MedSTAR: Washington based
Study; VHP: Verona Heart Project; NPHS: Northwick Park Health Study; WTCCC: Wellcome
Trust Case Control Consortium
Subtotal (I-squared = 10.9%, p = 0.290)
GWA
White
Samani - WTCCC
Study
Samani - German MI
Karvanen - MORGAM-CC
name
1924
N
859
900
cases
2936
N
1639
2107
cont
1.29 (1.19, 1.40)
1.20 (1.07, 1.36)
1.15 (1.05, 1.25)
1.10 (0.98, 1.23)
ES (95% CI)
0.55
% ra
0.57
0.61
controls
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
SNP rs7865618 (A)
SNP rs6475606 (T)
Subtotal (I-squared = 0.0%, p = 0.941)
White
GWA
Broadbent PROCARDIS
Samani - WTCCC
Samani - German MI
name
Study
4251
1926
860
cases
N
4443
2938
1636
cont
N
1.27 (1.20, 1.35)
1.27 (1.19, 1.36)
1.37 (1.26, 1.48)
1.28 (1.13, 1.44)
ES (95% CI)
0.48
0.47
0.47
controls
% ra
1.6 .8 1.2 1.4 1.6 1.8 2 2.2
v)
w)
317
Figure A4.9: Linkage disequilibrium patterns in Pakistanis, Europeans, Chinese and West African ethnicities
PAK: Pakistani from the PROMIS controls; CEU: Utah residents with Northern and Western European ancestry from the CEPH collection, YRI: Yoruba in
Ibadan Nigeria, CHB+JPT: Han Chinese in Beijing, China and Japanese in Tokyo, Japan. Regions of strong LD are in red, of moderate LD in light red, of very
moderate LD in blue and of no LD in white. PROMIS blocks are superposed to LD patterns in CEU, CHB+JPT and YRI. Only common variants genotyped in
PROMIS as well as in HapMap CEU, CHB+JPT and YRI populations are used in these plots.
YRI
PAK
CHB+JPT
CEU
318
Table A4.1: Baseline characteristics of the participants from PROMIS in 2,592 cases
and 2,577 controls
Cases Controls P-value
Age, years 54.2 (10.6) 52.9 (10.1) <0.001
Female (%) 16% 20% <0.001
History of diabetes 22% 15% <0.001
Family history of MI 21% 12% <0.001
Tobacco users <0.001
Never 35% 51%
Ex 12% 11%
Current 54% 38%
Ethnic group <0.001
Urdu (%) 46% 44%
Punjabi (%) 23% 21%
Pathan 7% 9%
Balochi 2% 2%
Sindhi 6% 9%
Memon 3% 2%
Gujrati 1% 4%
Other 11% 8%
Waist / Hip ratio 0.941 (0.001) 0.933 (0.001) <0.001
Data are mean (standard deviation) or percentages; P-values come from a t-test of
differences between the means for continuous variables and a χ2 test of independence for 
categorical variables. Information on family history of MI was only available in 5,068
participants for history of MI and history of diabetes, 5,041 for tobacco use and 4,509 for
waist to hip ratio. All participants had information on age, sex and ethnicity.
319
Table A4.2: Association of all the 89 variants at the 9p21.3 locus genotyped in 1851 cases
and 1903 controls using the IBC-array
SNP Position
Minor
allele
Major
allele
Cases
(MM/Mm/mm)
Controls
(MM/Mm/mm)
HWE p-
value in
controls Odds ratio (95%CI)
P-
value
R2 value
with
rs1333049
rs7023954 21806758 T C 113/676/10 122/695/10 0.46 1.02 (0.91; 1.13) 0.75 0.00
rs4382559 21934818 A G 8/219/1623 11/235/165 0.36 1.02 (0.85; 1.23) 0.85 0.00
rs10965197 21938666 T C 200/781/86 193/779/93 0.12 0.96 (0.87; 1.06) 0.45 0.03
rs2518722 21942926 T C 36/370/144 30/388/148 0.44 0.96 (0.83; 1.1) 0.53 0.00
rs10757261 21944953 T C 198/774/87 195/812/89 0.60 1.01 (0.91; 1.11) 0.85 0.03
rs7041637 21951866 T G 334/836/67 339/889/67 0.12 0.99 (0.9; 1.08) 0.76 0.22
rs3731257 21956221 A G 303/832/71 315/879/70 0.14 1 (0.91; 1.1) 0.98 0.18
rs3731246 21961989 G C 10/198/164 12/205/168 0.06 1 (0.82; 1.2) 0.96 0.01
rs2811708 21963422 T G 69/499/128 61/538/130 0.52 1.01 (0.89; 1.14) 0.89 0.00
rs3731239 21964218 C T 116/672/10 116/690/10 0.62 1 (0.89; 1.11) 0.93 0.10
rs2811709 21970151 A G 10/192/164 10/195/169 0.13 0.95 (0.78; 1.15) 0.59 0.00
rs3731217 21974661 G T 21/346/148 30/350/152 0.07 1.06 (0.91; 1.23) 0.47 0.01
rs3731201 21978896 G A 14/243/159 14/243/164 0.16 0.94 (0.79; 1.12) 0.48 0.01
rs7036656 21980457 C T 68/517/126 62/550/129 0.69 1 (0.89; 1.13) 0.96 0.00
rs3218020 21987872 T C 320/851/67 328/887/68 0.17 0.97 (0.89; 1.07) 0.55 0.29
rs2811712 21988035 G A 28/377/144 34/379/148 0.10 1.02 (0.88; 1.17) 0.81 0.02
rs3218002 21990841 T C 27/367/145 34/369/150 0.05 1.02 (0.89; 1.18) 0.75 0.01
rs1063192 21993367 G A 131/719/99 134/746/10 0.95 1.02 (0.92; 1.13) 0.73 0.19
rs2069418 21999698 G C 148/757/94 149/770/98 0.95 0.99 (0.89; 1.1) 0.83 0.15
rs545226 22002422 C T 331/858/64 360/882/64 0.05 0.99 (0.9; 1.09) 0.87 0.26
rs10811641 22004137 G C 332/848/66 343/884/67 0.08 0.98 (0.89; 1.07) 0.66 0.28
rs643319 22007836 A C 242/843/76 258/840/80 0.10 1.01 (0.92; 1.11) 0.78 0.20
rs518394 22009673 C G 152/764/92 149/781/97 0.69 0.98 (0.88; 1.08) 0.68 0.15
rs10757264 22009732 A G 299/885/65 303/878/71 0.25 0.97 (0.89; 1.07) 0.57 0.18
rs10965212 22013795 A T 259/873/71 283/872/74 0.28 1.03 (0.94; 1.13) 0.53 0.24
rs10738604 22015493 A G 325/852/67 330/888/68 0.15 0.97 (0.88; 1.06) 0.51 0.28
rs7049105 22018801 A G 257/874/71 285/873/74 0.26 1.04 (0.94; 1.14) 0.46 0.24
rs10965215 22019445 G A 262/878/71 286/876/74 0.31 1.03 (0.94; 1.13) 0.54 0.24
rs2151280 22024719 G A 266/876/70 289/880/73 0.36 1.03 (0.94; 1.14) 0.49 0.24
rs1333035 22034059 C T 24/370/145 31/375/149 0.18 1.03 (0.89; 1.19) 0.66 0.02
rs11790231 22043591 T C 36/392/142 16/390/149 0.10 0.85 (0.74; 0.98) 0.03 0.09
rs10120688 22046499 G A 257/847/74 282/870/75 0.26 1.06 (0.97; 1.17) 0.22 0.26
rs10965224 22057276 T A 120/694/10 123/731/10 0.81 1.04 (0.94; 1.16) 0.45 0.23
rs4977756 22058652 C T 121/694/10 124/731/10 0.81 1.04 (0.94; 1.16) 0.45 0.23
rs9632884 22062301 C G 267/818/76 297/872/73 0.16 1.1 (1; 1.2) 0.05 0.43
rs9632885 22062638 C T 370/853/62 407/923/57 0.33 1.13 (1.03; 1.24) 0.009 0.57
rs7855162 22064793 C T 8/219/1622 13/238/165 0.16 1.14 (0.94; 1.37) 0.18 0.06
rs1412832 22067543 G A 98/643/111 102/673/11 0.90 1.04 (0.93; 1.16) 0.47 0.19
rs10116277 22071397 G T 302/882/66 339/924/64 0.89 1.1 (1; 1.21) 0.05 0.55
rs10965228 22072380 C T 6/183/1659 5/206/1687 0.83 1.12 (0.91; 1.37) 0.29 0.06
rs1333040 22073404 G A 263/867/72 302/888/71 0.36 1.08 (0.99; 1.19) 0.09 0.48
rs10757272 22078260 C T 333/876/64 373/945/58 0.82 1.13 (1.03; 1.24) 0.01 0.68
rs10757274 22086055 A G 424/892/53 464/968/47 0.46 1.14 (1.04; 1.25) 0.005 0.86
rs4977574 22088574 A G 425/888/53 464/966/47 0.52 1.14 (1.04; 1.25) 0.006 0.86
rs2891168 22088619 T C 426/888/53 465/965/47 0.55 1.14 (1.04; 1.25) 0.006 0.87
rs1537372 22093183 A C 491/880/48 425/956/52 0.75 0.88 (0.8; 0.96) 0.006 0.78
rs1333042 22093813 A G 281/869/70 321/915/66 0.81 1.12 (1.02; 1.23) 0.02 0.61
rs10511701 22102599 T C 316/881/65 355/940/60 0.82 1.12 (1.02; 1.23) 0.02 0.69
rs2383206 22105026 T C 367/890/59 401/939/55 0.89 1.11 (1.02; 1.22) 0.02 0.78
rs10965235 22105105 A C 6/203/1630 10/202/168 0.14 1.05 (0.86; 1.28) 0.64 0.06
rs944797 22105286 A G 364/891/59 401/939/55 0.89 1.12 (1.02; 1.22) 0.02 0.78
rs1537375 22106071 T C 316/873/65 352/941/60 0.74 1.12 (1.02; 1.23) 0.02 0.69
rs17761446 22108102 G T 6/197/1647 10/202/169 0.14 1.08 (0.89; 1.32) 0.43 0.06
rs1333048 22115347 A C 401/886/56 437/952/51 0.93 1.12 (1.02; 1.23) 0.01 0.89
rs1333049 22115503 C G 440/904/50 487/964/45 0.58 1.13 (1.03; 1.24) 0.01 1.00
rs1333050 22115913 C T 407/886/55 444/911/54 0.09 1.09 (0.99; 1.19) 0.07 0.38
rs10757281 22117613 T C 59/513/127 60/512/133 0.21 0.98 (0.87; 1.1) 0.71 0.00
rs12379111 22118180 C G 15/258/157 12/263/162 0.62 0.98 (0.83; 1.17) 0.84 0.01
rs10811658 22118600 T C 126/707/10 142/709/10 0.15 1 (0.9; 1.11) 0.95 0.03
rs7020996 22119579 T C 29/361/146 19/368/151 0.63 0.97 (0.84; 1.13) 0.71 0.00
rs10965241 22119594 G C 53/492/130 66/514/132 0.07 1.02 (0.91; 1.16) 0.70 0.04
rs10965243 22120065 C T 23/342/148 16/336/155 0.79 0.95 (0.81; 1.1) 0.48 0.00
rs10965244 22120389 T A 53/501/129 67/523/131 0.11 1.02 (0.9; 1.16) 0.71 0.04
rs2383208 22122076 G A 36/416/139 29/443/143 0.48 1.01 (0.88; 1.16) 0.85 0.01
rs7045889 22123251 C T 333/848/66 346/882/67 0.07 1 (0.92; 1.1) 0.94 0.07
rs10217762 22123645 C T 337/875/63 335/912/65 0.57 1.02 (0.93; 1.12) 0.72 0.02
rs10811659 22123716 G A 71/553/122 69/584/125 0.94 0.99 (0.88; 1.11) 0.87 0.02
rs10811661 22124094 G A 35/401/141 32/438/143 0.92 1.06 (0.93; 1.22) 0.39 0.00
rs10757283 22124172 T C 339/875/63 335/913/65 0.60 1.01 (0.93; 1.11) 0.75 0.02
rs1333051 22126489 T A 11/243/159 13/286/160 0.88 1.17 (0.99; 1.39) 0.07 0.01
rs11791416 22128105 C T 101/665/10 117/687/10 0.49 1.03 (0.92; 1.15) 0.61 0.03
rs10757284 22128458 G C 281/819/74 272/900/73 0.88 1.01 (0.92; 1.11) 0.83 0.05
rs4977761 22128762 T C 317/838/69 291/940/67 0.21 1 (0.91; 1.1) 0.98 0.05
rs2065501 22130224 T G 106/637/11 128/688/10 0.20 1.1 (0.99; 1.22) 0.08 0.00
rs7854629 22131034 C T 149/749/95 173/775/95 0.38 1.04 (0.94; 1.15) 0.42 0.01
rs2065505 22131790 G A 92/619/113 99/671/113 1.00 1.05 (0.94; 1.17) 0.37 0.01
rs6475610 22131894 G A 240/805/80 231/869/80 0.88 0.99 (0.9; 1.09) 0.90 0.00
rs10811664 22132907 A G 13/275/156 14/278/161 0.54 0.97 (0.82; 1.15) 0.73 0.00
rs2151283 22134305 T G 271/858/72 281/886/73 0.59 1.01 (0.92; 1.11) 0.82 0.00
rs7853656 22134530 C A 135/729/98 149/722/10 0.16 1.01 (0.91; 1.12) 0.88 0.00
rs7856219 22140261 G A 382/895/57 374/928/60 0.64 0.98 (0.9; 1.08) 0.70 0.00
rs7047414 22141412 T G 24/315/151 19/317/156 0.50 0.93 (0.8; 1.08) 0.35 0.00
rs10965256 22141465 A G 40/437/137 49/459/139 0.14 1.06 (0.93; 1.2) 0.40 0.00
rs7853123 22143360 T C 205/758/88 211/812/88 0.25 1.02 (0.93; 1.13) 0.63 0.00
rs10965258 22143663 G A 71/533/124 72/533/129 0.07 1 (0.89; 1.13) 1.00 0.00
rs1333052 22147908 G T 378/909/56 382/940/58 0.96 1.01 (0.92; 1.1) 0.91 0.00
rs12238587 22148168 T A 20/265/156 16/302/158 0.67 1.03 (0.87; 1.21) 0.72 0.00
rs10122243 22148924 C T 300/859/69 313/910/68 0.78 1.02 (0.93; 1.12) 0.72 0.00
rs10757288 22149416 G A 189/752/90 188/815/90 0.87 1.02 (0.93; 1.13) 0.66 0.00
M: major allele; m: minor allele according to the frequency in the overall population, r2:
Linkage disequilibrium between rs1333049 and other SNPs; MAF: Minor allele frequency.
The odds ratio represent the per major allele increase in risk of MI adjusted for age, sex and
the first two PCs. HWE: Hardy Weinberg Equilibrium test was performed in control
participants only.
320
Table A4.3: Association of SNPs at the 9p21.3 locus significantly associated with MI in PROMIS
SNP Position
Allele
m/M RAF
N cases
(MM/Mm/mm)
N controls
(MM/Mm/mm) OR (95%CI)
P-
value
HWE
p-
value R2
rs10757274 22086055 A/G 0.50 742/1253/596 633/1302/640 1.13 (1.05;1.23) 2.E-03 0.57 0.86
rs1333049 22115503 G/C 0.51 697/1273/617 609/1290/674 1.13 (1.05;1.22) 2.E-03 0.86 1.00
rs1537372 22093183 A/C 0.47 670/1237/683 718/1273/581 1.13 (1.04;1.22) 2.E-03 0.71 0.78
rs2891168 22088619 T/C 0.50 746/1243/597 634/1298/644 1.14 (1.05;1.23) 1.E-03 0.69 0.86
rs4977574 22088574 A/G 0.50 746/1242/596 636/1298/642 1.13 (1.05;1.23) 1.E-03 0.69 0.86
rs9632885 22062638 C/T 0.54 857/1223/502 772/1240/560 1.12 (1.04;1.21) 3.E-03 0.14 0.56
M: major allele; m: minor allele according to the frequency in the overall population; RAF: Risk allele frequency in controls; N cases: Number of cases; OR: Odds ratio
representing the increased risk of MI per risk allele adjusted for age, sex and self reported ethnicity.R2: Linkage disequilibrium between rs1333049 and other SNPs;
HWE: Hardy Weinberg Equilibrium test was performed in control participants only. The risk allele corresponds to the minor allele for rs1537372 and to the major allele
for the other SNPs
321
Table A4.4: Haplotype analyses of SNPs rs1333049 and rs1412832
Polymorphisms Haplotype Frequencies Haplotypic Odds Ratio [95% CI]
rs1412832 rs1333049 Controls Cases Model 1 Model 2
G G 0.201 0.205
0.949 [0.841 – 1.072] 0.937 [0.828 – 1.060]
P-value = 0.403 P-value = 0.304
G C 0.028 0.022
0.748 [0.523 – 1.068] 0.778 [0.541 – 1.120]
P-value = 0.110 P-value = 0.177
A G 0.307 0.276
0.843 [0.755 – 0.940] 0.846 [0.757 – 0.946]
P-value = 0.002 P-value = 0.003
A C 0.464 0.497 reference reference
Test of haplotypic association
X2 = 9.07 with 3 df X2 = 9.08 with 3 df
P-value = 0.028 P-value = 0.028
Reference: Reference group for the computation of the odds ratios; df: degree of freedom; X2: Chi-square statistic for the test of haplotypic association. Model1 was
unadjusted and model 2 was adjusted for age, sex and the first two principal components. Haplotypic odds ratios represent the odds ratio for haplotypes GG, GC and
AC versus the most common reference group AC.
322
Table A4.5: Characteristics of studies included in the current literature based meta-analyses
Author Consortium and studies
names
Type Design Source
controls
Location N cases / N
controls
Diagnostic of CAD Genotyping platform SNP included in the literature
based meta-analysis
Abdullah GeneQuest Rep. PC* Population USA 310/560 CAD/MI TaqMan Assay,
Applied Biosystems
rs2383206, rs10757274
Assimes ADVANCE: White, Black,
East Asian, Hispanic, Mixed
(Black) and Mixed (non
Black)
Rep. CC Health care
system
USA 3618/3468 MI/PTCA/CABG/angina TaqMan Assay,
Applied Biosystems
rs2383206, rs10757274
Broadbent PROCARDIS: Germany,
Italy, Sweden, UK
Rep. CC Hospital Europe 4251/4443 MI/PTCA/CABG/angina Sequenom iPLEXTM rs518394, rs1333040,rs1333042,
rs1333048, rs1333049, rs2383206,
rs2891168, rs4977574, rs10116277,
rs10757272, rs10757274, rs10757283
CAD Consortium AMC-PAS, ECTIM, EPIC-
Norfolk, GerMIFS$,
KORA/GOC$, LURIC,
MORGAM$, UKMI$,
PopGen$
Rep. CC Mixed Europe 11550/11205 MI/Stenosis
(>50%)/PTCA/
CABG/Angina
Sequenom iPLEXTM rs1333049
Chen China Study Rep. CC Health care
system
China  232/212 MI/Stenosis (≥50%) TaqMan Assay, 
Applied Biosystems
rs10757278, rs2383207
Dehghan Rotterdam study Rep. PC Population Netherlands 588/5251 MI/CABG/PTCA TaqMan Assay,
Applied Biosystems
rs10757274
Ghazouani Tunisia study Rep. CC Population Tunisia 296/324 Stenosis/MI TaqMan Assay,
Applied Biosystems
rs1333049
Helgadottir Iceland A GWA CC Mixed Europe, USA 4587/12767 MI Infinium
HumanHap300
(Illumina, USA)
rs1333040, rs10116277
Iceland B, Atlanta, Durham,
Philadelphia (PennCath)
Rep. CC Hospital MI/CABG/PTCA Centaurus
(Nanogen) platform
rs1333040, rs1333048, rs4977574,
rs10116277
Hinohara SMC2 & Tokyo study Rep. CC NF Japan  1283/1857 MI/Stenosis (≥50%) TaqMan Assay, 
Applied Biosystems
rs1333049
Hiura Suita study Rep. CC Population Japan 589/2475 MI TaqMan Assay,
Applied Biosystems
rs1333049
Karvanen MORGAM$ Rep. NCC Population Europe 1050/1878 MI/angina/CABG/PCI Sequenom iPLEXTM rs1333049, rs10757283
Lemmens Caregene Ref CC Population Belgium 926/648 MI/angina/CABG/PCI TaqMan Assay,
Applied Biosystems
rs10757278
McPherson OHS-1 GWA CC Hospital Canada,
USA, Europe
4306/20119 MI/PTCA/CABG/CHD
death
Perlegen Sciences rs2383206, rs10757274
ARIC, CCHS, DHS, OHS-2,
OHS-3
Rep. CC** Population
and hospital
MIGC deCODE, MIGen, AMI
Gene-VHP-MAHI-IFS£-
GerMIFS£ II-
INTERHEART£ and
PennCATH£-MedSTAR£
Rep. CC Population,
blood donor
Europe, USA NF MI Affymetrix GeneChip
Human Mapping
500KArray Set or
Affymetrix 6.0
GeneChip
rs4977574
Paynter Women's Genome Health
Study
Rep. PC Population USA 469/10660 MI/PTCA/CABG/CHD
death
Luminex100 xMAP
microspheres
(Luminex, USA)
rs10757274
323
Samani German MI study Rep. CC Population Germany 844/1605 early MI and family
history of MI
Affymetrix GeneChip
Human Mapping
500K Array Set
rs518394, rs1333042, rs1333048,
rs1333049, rs2891168, rs10757272
Schaefer PopGen 1, PopGen 2 Rep. CC Blood
donors
Germany Stenosis
(>70%)/MI/PTCA/
CABG
SNPlex and TaqMan
GenotypingSystem
(Applied Biosystems,
USA)
rs1333042, rs1333048, rs2891168
Schunkert CARDIOGENICS:
AtheroGene, German MI II$,
LMDS, MONICA/KORA$,
PopGen$, PRIME, UKMI$
Rep. CC Mixed Europe 8912/9828 MI/angina TaqMan Assay,
Applied Biosystems
rs1333049
Shen SMC2 Rep. CC Hospital Korea 611/294 MI/stenosis/CABG/PCI TaqMan Assay,
Applied Biosystems
rs2383206, rs10757274
Shen VHP Rep. CC Hospital Italy 416/308 MI TaqMan Assay,
Applied Biosystems
rs2383206, rs10757274
Talmud NPHS2 Rep. PC Population UK 264/2430 MI/PTCA/CABG TaqMan Assay,
Applied Biosystems
rs2383206, rs10757274
WTCCC WTCCC GWA CC Population
and blood
donors
UK 1926/2938 MI or CABG/PTCA Affymetric GeneChip
Human Mapping
500K Array Set
rs518394, rs1333042, rs1333048,
rs1333049, rs2891168, rs10757272,
rs10757283
Ye Bruneck study Rep. PC Population Italy 56/713 MI/CABG/PTCA TaqMan Assay,
Applied Biosystems
rs1333049
Zee Physician's Health Study Rep. NCC Population
(physicians)
USA 335/335 MI NF rs2383206, rs10757274
Zhang Chaoyang study Rep. CC Hospital China 432/430 MI Orchid BioSciences
(GenomLab
SNPstream
genotyping platform)
rs2383206, rs10757274
Zhou Wuhan study Rep. CC Population China 1360/1360 stenosis (>=
50%)/MI/CABG/PCI
TaqMan Assay,
Applied Biosystems
rs2383206
Rep: Replication study; GWA: genome wide association study; PC: prospective cohort study; CC: Case-control study; NCC: Nested case-control study; MI:
Myocardial Infarction; NF: Not found. *: analysed as a case-control study; **: except ARIC, a prospective cohort study; £ The following studies from MIGC were
not included in the analysis for rs4977574 because they overlapped partially with previously published results. For prospective cohorts, the number of controls
corresponds to the number of individuals who had not developed the disease at the end of follow-up. Study acronyms: ADVANCE: Atherosclerosis Disease
Vascular FuNction & GenetiC Epidemiology; PROCARDIS; AMC-PAS: Academic Medical Centre Amsterdam Premature Atherosclerosis Study; ECTIM:
Etude Cas-Temoin sur l’Infarctus du Myocarde; EPIC: European Prospective Investigation into Cancer and Nutrition Study; GerMIFS: German MI Family
Study; LURIC: Ludwigshafen Risk and Cardiovascular Heatlh; PopGen: PopGen biobank; UKMI: UK MI; SMC: Samsung Medical Centre; OHS: Ottawa Heart
Study,; ARIC: Atherosclerosis Risk in Community; CCHS: Copenhagen City Heart Study; DHS: Dallas Heart Study; AMI Gene: Acute MI Gene; VHS: Verona
Heart Study; MAHI: Mid America Heart Institute IFS: Irish Family Study; PennCATH: University of Pennsylvania Catheterization Study; MedSTAR:
Washington based Study; VHP: Verona Heart Project; NPHS: Northwick Park Health Study; WTCCC: Wellcome Trust Case Control Consortium.
324
References
1. WTCCC. Genome-wide association study of 14,000 cases of seven common
diseases and 3,000 shared controls. Nature 2007;447:661-678.
2. Samani NJ, Erdmann J, Hall AS, Hengstenberg C, Mangino M, Mayer B, Dixon RJ,
Meitinger T, Braund P, Wichmann HE, Barrett JH, Konig IR, Stevens SE, Szymczak S,
Tregouet DA, Iles MM, Pahlke F, Pollard H, Lieb W, Cambien F, Fischer M, Ouwehand W,
Blankenberg S, Balmforth AJ, Baessler A, Ball SG, Strom TM, Braenne I, Gieger C, Deloukas
P, Tobin MD, Ziegler A, Thompson JR, Schunkert H. Genomewide association analysis of
coronary artery disease. N Engl J Med 2007;357:443-453.
3. McPherson R, Pertsemlidis A, Kavaslar N, Stewart A, Roberts R, Cox DR, Hinds DA,
Pennacchio LA, Tybjaerg-Hansen A, Folsom AR, Boerwinkle E, Hobbs HH, Cohen JC. A
common allele on chromosome 9 associated with coronary heart disease. Science
2007;316:1488-1491.
4. Helgadottir A, Thorleifsson G, Manolescu A, Gretarsdottir S, Blondal T, Jonasdottir A,
Jonasdottir A, Sigurdsson A, Baker A, Palsson A, Masson G, Gudbjartsson DF, Magnusson
KP, Andersen K, Levey AI, Backman VM, Matthiasdottir S, Jonsdottir T, Palsson S, Einarsdottir
H, Gunnarsdottir S, Gylfason A, Vaccarino V, Hooper WC, Reilly MP, Granger CB, Austin H,
Rader DJ, Shah SH, Quyyumi AA, Gulcher JR, Thorgeirsson G, Thorsteinsdottir U, Kong A,
Stefansson K. A common variant on chromosome 9p21 affects the risk of myocardial infarction.
Science 2007;316:1491-1493.
5. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
6. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
7. Jarinova O, Stewart AF, Roberts R, Wells G, Lau P, Naing T, Buerki C, McLean BW,
Cook RC, Parker JS, McPherson R. Functional Analysis of the Chromosome 9p21.3 Coronary
Artery Disease Risk Locus. Arterioscler Thromb Vasc Biol 2009.
8. Nishtar S. Prevention of coronary heart disease in south Asia. Lancet 2002;360:1015-
1018.
9. Saleheen D, Zaidi M, Rasheed A, Ahmad U, Hakeem A, Murtaza M, Kayani W, Faruqui
A, Kundi A, Zaman KS, Yaqoob Z, Cheema LA, Samad A, Rasheed SZ, Mallick NH, Azhar M,
Jooma R, Gardezi AR, Memon N, Ghaffar A, Fazal uR, Khan N, Shah N, Ali SA, Samuel M,
Hanif F, Yameen M, Naz S, Sultana A, Nazir A, Raza S, Shazad M, Nasim S, Javed MA, Ali
SS, Jafree M, Nisar MI, Daood MS, Hussain A, Sarwar N, Kamal A, Deloukas P, Ishaq M,
Frossard P, Danesh J. The Pakistan Risk of Myocardial Infarction Study: a resource for the
study of genetic, lifestyle and other determinants of myocardial infarction in South Asia. Eur J
Epidemiol 2009;24:329-338.
10. WTCCC. Genome-wide association study of 14,000 cases of seven common diseases
and 3,000 shared controls. Nature 2007;447:661-678.
325
11. Samani NJ, Erdmann J, Hall AS, Hengstenberg C, Mangino M, Mayer B, Dixon RJ,
Meitinger T, Braund P, Wichmann HE, Barrett JH, Konig IR, Stevens SE, Szymczak S,
Tregouet DA, Iles MM, Pahlke F, Pollard H, Lieb W, Cambien F, Fischer M, Ouwehand W,
Blankenberg S, Balmforth AJ, Baessler A, Ball SG, Strom TM, Braenne I, Gieger C, Deloukas
P, Tobin MD, Ziegler A, Thompson JR, Schunkert H. Genomewide association analysis of
coronary artery disease. N Engl J Med 2007;357:443-453.
12. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
13. Little J, Higgins JP, Ioannidis JP, Moher D, Gagnon F, von EE, Khoury MJ, Cohen B,
vey-Smith G, Grimshaw J, Scheet P, Gwinn M, Williamson RE, Zou GY, Hutchings K, Johnson
CY, Tait V, Wiens M, Golding J, van DC, McLaughlin J, Paterson A, Wells G, Fortier I,
Freedman M, Zecevic M, King R, Infante-Rivard C, Stewart A, Birkett N. STrengthening the
REporting of Genetic Association studies (STREGA)--an extension of the STROBE statement.
Eur J Clin Invest 2009;39:247-266.
14. Keating BJ, Tischfield S, Murray SS, Bhangale T, Price TS, Glessner JT, Galver L,
Barrett JC, Grant SF, Farlow DN, Chandrupatla HR, Hansen M, Ajmal S, Papanicolaou GJ,
Guo Y, Li M, Derohannessian S, de Bakker PI, Bailey SD, Montpetit A, Edmondson AC, Taylor
K, Gai X, Wang SS, Fornage M, Shaikh T, Groop L, Boehnke M, Hall AS, Hattersley AT,
Frackelton E, Patterson N, Chiang CW, Kim CE, Fabsitz RR, Ouwehand W, Price AL, Munroe
P, Caulfield M, Drake T, Boerwinkle E, Reich D, Whitehead AS, Cappola TP, Samani NJ, Lusis
AJ, Schadt E, Wilson JG, Koenig W, McCarthy MI, Kathiresan S, Gabriel SB, Hakonarson H,
Anand SS, Reilly M, Engert JC, Nickerson DA, Rader DJ, Hirschhorn JN, Fitzgerald GA.
Concept, design and implementation of a cardiovascular gene-centric 50 k SNP array for large-
scale genomic association studies. PLoS One 2008;3:e3583.
15. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, Maller J, Sklar
P, de Bakker PI, Daly MJ, Sham PC. PLINK: a tool set for whole-genome association and
population-based linkage analyses. Am J Hum Genet 2007;81:559-575.
16. WTCCC. Genome-wide association study of 14,000 cases of seven common diseases
and 3,000 shared controls. Nature 2007;447:661-678.
17. Sabatti C, Service SK, Hartikainen AL, Pouta A, Ripatti S, Brodsky J, Jones CG, Zaitlen
NA, Varilo T, Kaakinen M, Sovio U, Ruokonen A, Laitinen J, Jakkula E, Coin L, Hoggart C,
Collins A, Turunen H, Gabriel S, Elliot P, McCarthy MI, Daly MJ, Jarvelin MR, Freimer NB,
Peltonen L. Genome-wide association analysis of metabolic traits in a birth cohort from a
founder population. Nat Genet 2009;41:35-46.
18. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and
haplotype maps. Bioinformatics 2005;21:263-265.
19. Tregouet DA, Garelle V. A new JAVA interface implementation of THESIAS: testing
haplotype effects in association studies. Bioinformatics 2007;23:1038-1039.
20. Tregouet DA, Ricard S, Nicaud V, Arnould I, Soubigou S, Rosier M, Duverger N, Poirier
O, Mace S, Kee F, Morrison C, Denefle P, Tiret L, Evans A, Deleuze JF, Cambien F. In-depth
haplotype analysis of ABCA1 gene polymorphisms in relation to plasma ApoA1 levels and
myocardial infarction. Arterioscler Thromb Vasc Biol 2004;24:775-781.
21. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and
haplotype maps. Bioinformatics 2005;21:263-265.
326
22. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med
2002;21:1539-1558.
23. WTCCC. Genome-wide association study of 14,000 cases of seven common diseases
and 3,000 shared controls. Nature 2007;447:661-678.
24. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
25. WTCCC. Genome-wide association study of 14,000 cases of seven common diseases
and 3,000 shared controls. Nature 2007;447:661-678.
26. Samani NJ, Erdmann J, Hall AS, Hengstenberg C, Mangino M, Mayer B, Dixon RJ,
Meitinger T, Braund P, Wichmann HE, Barrett JH, Konig IR, Stevens SE, Szymczak S,
Tregouet DA, Iles MM, Pahlke F, Pollard H, Lieb W, Cambien F, Fischer M, Ouwehand W,
Blankenberg S, Balmforth AJ, Baessler A, Ball SG, Strom TM, Braenne I, Gieger C, Deloukas
P, Tobin MD, Ziegler A, Thompson JR, Schunkert H. Genomewide association analysis of
coronary artery disease. N Engl J Med 2007;357:443-453.
27. McPherson R, Pertsemlidis A, Kavaslar N, Stewart A, Roberts R, Cox DR, Hinds DA,
Pennacchio LA, Tybjaerg-Hansen A, Folsom AR, Boerwinkle E, Hobbs HH, Cohen JC. A
common allele on chromosome 9 associated with coronary heart disease. Science
2007;316:1488-1491.
28. Helgadottir A, Thorleifsson G, Manolescu A, Gretarsdottir S, Blondal T, Jonasdottir A,
Jonasdottir A, Sigurdsson A, Baker A, Palsson A, Masson G, Gudbjartsson DF, Magnusson
KP, Andersen K, Levey AI, Backman VM, Matthiasdottir S, Jonsdottir T, Palsson S, Einarsdottir
H, Gunnarsdottir S, Gylfason A, Vaccarino V, Hooper WC, Reilly MP, Granger CB, Austin H,
Rader DJ, Shah SH, Quyyumi AA, Gulcher JR, Thorgeirsson G, Thorsteinsdottir U, Kong A,
Stefansson K. A common variant on chromosome 9p21 affects the risk of myocardial infarction.
Science 2007;316:1491-1493.
29. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
30. Abdullah KG, Li L, Shen GQ, Hu Y, Yang Y, MacKinlay KG, Topol EJ, Wang QK. Four
SNPS on chromosome 9p21 confer risk to premature, familial CAD and MI in an American
Caucasian population (GeneQuest). Ann Hum Genet 2008;72:654-657.
31. Assimes TL, Knowles JW, Basu A, Iribarren C, Southwick A, Tang H, Absher D, Li J,
Fair JM, Rubin GD, Sidney S, Fortmann SP, Go AS, Hlatky MA, Myers RM, Risch N,
Quertermous T. Susceptibility locus for clinical and subclinical coronary artery disease at
chromosome 9p21 in the multi-ethnic ADVANCE study. Hum Mol Genet 2008;17:2320-2328.
32. Broadbent HM, Peden JF, Lorkowski S, Goel A, Ongen H, Green F, Clarke R, Collins
R, Franzosi MG, Tognoni G, Seedorf U, Rust S, Eriksson P, Hamsten A, Farrall M, Watkins H.
Susceptibility to coronary artery disease and diabetes is encoded by distinct, tightly linked
SNPs in the ANRIL locus on chromosome 9p. Hum Mol Genet 2008;17:806-814.
327
33. Samani NJ, Deloukas P, Erdmann J, Hengstenberg C, Kuulasmaa K, McGinnis R,
Schunkert H, Soranzo N, Thompson J, Tiret L, Ziegler A. Large scale association analysis of
novel genetic loci for coronary artery disease. Arterioscler Thromb Vasc Biol 2009;29:774-780.
34. Dehghan A, van HM, Sijbrands EJ, Oostra BA, Hofman A, van Duijn CM, Witteman JC.
Lack of association of two common polymorphisms on 9p21 with risk of coronary heart disease
and myocardial infarction; results from a prospective cohort study. BMC Med 2008;6:30.:30.
35. Karvanen J, Silander K, Kee F, Tiret L, Salomaa V, Kuulasmaa K, Wiklund PG, Virtamo
J, Saarela O, Perret C, Perola M, Peltonen L, Cambien F, Erdmann J, Samani NJ, Schunkert
H, Evans A. The impact of newly identified loci on coronary heart disease, stroke and total
mortality in the MORGAM prospective cohorts. Genet Epidemiol 2009;33:237-246.
36. Lemmens R, Abboud S, Robberecht W, Vanhees L, Pandolfo M, Thijs V, Goris A.
Variant on 9p21 strongly associates with coronary heart disease, but lacks association with
common stroke. Eur J Hum Genet 2009.
37. Kathiresan S, Voight BF, Purcell S, Musunuru K, Ardissino D, Mannucci PM, Anand S,
Engert JC, Samani NJ, Schunkert H, Erdmann J, Reilly MP, Rader DJ, Morgan T, Spertus JA,
Stoll M, Girelli D, McKeown PP, Patterson CC, Siscovick DS, O'Donnell CJ, Elosua R, Peltonen
L, Salomaa V, Schwartz SM, Melander O, Altshuler D, Ardissino D, Merlini PA, Berzuini C,
Bernardinelli L, Peyvandi F, Tubaro M, Celli P, Ferrario M, Fetiveau R, Marziliano N, Casari G,
Galli M, Ribichini F, Rossi M, Bernardi F, Zonzin P, Piazza A, Mannucci PM, Schwartz SM,
Siscovick DS, Yee J, Friedlander Y, Elosua R, Marrugat J, Lucas G, Subirana I, Sala J, Ramos
R, Kathiresan S, Meigs JB, Williams G, Nathan DM, MacRae CA, O'Donnell CJ, Salomaa V,
Havulinna AS, Peltonen L, Melander O, Berglund G, Voight BF, Kathiresan S, Hirschhorn JN,
Asselta R, Duga S, Spreafico M, Musunuru K, Daly MJ, Purcell S, Voight BF, Purcell S,
Nemesh J, Korn JM, McCarroll SA, Schwartz SM, Yee J, Kathiresan S, Lucas G, Subirana I,
Elosua R, Surti A, Guiducci C, Gianniny L, Mirel D, Parkin M, Burtt N, Gabriel SB, Samani NJ,
Thompson JR, Braund PS, Wright BJ, Balmforth AJ, Ball SG, Hall AS, Schunkert H, Erdmann
J, Linsel-Nitschke P, Lieb W, Ziegler A, Konig I, Hengstenberg C, Fischer M, Stark K,
Grosshennig A, Preuss M, Wichmann HE, Schreiber S, Schunkert H, Samani NJ, Erdmann J,
Ouwehand W, Hengstenberg C, Deloukas P, Scholz M, Cambien F, Reilly MP, Li M, Chen Z,
Wilensky R, Matthai W, Qasim A, Hakonarson HH, Devaney J, Burnett MS, Pichard AD, Kent
KM, Satler L, Lindsay JM, Waksman R, Epstein SE, Rader DJ, Scheffold T, Berger K, Stoll M,
Huge A, Girelli D, Martinelli N, Olivieri O, Corrocher R, Morgan T, Spertus JA, McKeown P,
Patterson CC, Schunkert H, Erdmann E, Linsel-Nitschke P, Lieb W, Ziegler A, Konig IR,
Hengstenberg C, Fischer M, Stark K, Grosshennig A, Preuss M, Wichmann HE, Schreiber S,
Holm H, Thorleifsson G, Thorsteinsdottir U, Stefansson K, Engert JC, Do R, Xie C, Anand S,
Kathiresan S, Ardissino D, Mannucci PM, Siscovick D, O'Donnell CJ, Samani NJ, Melander O,
Elosua R, Peltonen L, Salomaa V, Schwartz SM, Altshuler D. Genome-wide association of
early-onset myocardial infarction with single nucleotide polymorphisms and copy number
variants. Nat Genet 2009;41:334-341.
38. Paynter NP, Chasman DI, Buring JE, Shiffman D, Cook NR, Ridker PM. Cardiovascular
disease risk prediction with and without knowledge of genetic variation at chromosome 9p21.3.
Ann Intern Med 2009;%20;150:65-72.
39. Schaefer AS, Richter GM, Groessner-Schreiber B, Noack B, Nothnagel M, El Mokhtari
NE, Loos BG, Jepsen S, Schreiber S. Identification of a shared genetic susceptibility locus for
coronary heart disease and periodontitis. PLoS Genet 2009;5:e1000378.
40. Shen GQ, Rao S, Martinelli N, Li L, Olivieri O, Corrocher R, Abdullah KG, Hazen SL,
Smith J, Barnard J, Plow EF, Girelli D, Wang QK. Association between four SNPs on
chromosome 9p21 and myocardial infarction is replicated in an Italian population. J Hum Genet
2008;53:144-150.
41. Talmud PJ, Cooper JA, Palmen J, Lovering R, Drenos F, Hingorani AD, Humphries SE.
Chromosome 9p21.3 coronary heart disease locus genotype and prospective risk of CHD in
healthy middle-aged men. Clin Chem 2008;54:467-474.
328
42. Zee RY, Ridker PM. Two common gene variants on chromosome 9 and risk of
atherothrombosis. Stroke 2007;38:e111.
43. Ye S, Willeit J, Kronenberg F, Xu Q, Kiechl S. Association of genetic variation on
chromosome 9p21 with susceptibility and progression of atherosclerosis: a population-based,
prospective study. J Am Coll Cardiol 2008;52:378-384.
44. Chen Z, Qian Q, Ma G, Wang J, Zhang X, Feng Y, Shen C, Yao Y. A common variant
on chromosome 9p21 affects the risk of early-onset coronary artery disease. Mol Biol Rep
2009;36:889-893.
45. Hinohara K, Nakajima T, Takahashi M, Hohda S, Sasaoka T, Nakahara K, Chida K,
Sawabe M, Arimura T, Sato A, Lee BS, Ban JM, Yasunami M, Park JE, Izumi T, Kimura A.
Replication of the association between a chromosome 9p21 polymorphism and coronary artery
disease in Japanese and Korean populations. J Hum Genet 2008;53:357-359.
46. Hiura Y, Fukushima Y, Yuno M, Sawamura H, Kokubo Y, Okamura T, Tomoike H, Goto
Y, Nonogi H, Takahashi R, Iwai N. Validation of the association of genetic variants on
chromosome 9p21 and 1q41 with myocardial infarction in a Japanese population. Circ J
2008;72:1213-1217.
47. Shen GQ, Li L, Rao S, Abdullah KG, Ban JM, Lee BS, Park JE, Wang QK. Four SNPs
on chromosome 9p21 in a South Korean population implicate a genetic locus that confers high
cross-race risk for development of coronary artery disease. Arterioscler Thromb Vasc Biol
2008;28:360-365.
48. Zhang Q, Wang XF, Cheng SS, Wan XH, Cao FF, Li L, Chen XD, Liu WJ, Yang XC, Jin
L. Three SNPs on chromosome 9p21 confer increased risk of myocardial infarction in Chinese
subjects. Atherosclerosis 2009.
49. Zhou L, Zhang X, He M, Cheng L, Chen Y, Hu FB, Wu T. Associations between single
nucleotide polymorphisms on chromosome 9p21 and risk of coronary heart disease in Chinese
Han population. Arterioscler Thromb Vasc Biol 2008;28:2085-2089.
50. Ghazouani L, Khalifa SB, Abboud N, Perret C, Nicaud V, Khalfallah AB, Alamawi WY,
Cambien F, Mahjoub T. Association of three polymorphisms selected from a genome-wide
association study with coronary heart disease in the Tunisian population. J Thromb
Thrombolysis 2009.
51. Abdullah KG, Li L, Shen GQ, Hu Y, Yang Y, MacKinlay KG, Topol EJ, Wang QK. Four
SNPS on chromosome 9p21 confer risk to premature, familial CAD and MI in an American
Caucasian population (GeneQuest). Ann Hum Genet 2008;72:654-657.
52. Dehghan A, van HM, Sijbrands EJ, Oostra BA, Hofman A, van Duijn CM, Witteman JC.
Lack of association of two common polymorphisms on 9p21 with risk of coronary heart disease
and myocardial infarction; results from a prospective cohort study. BMC Med 2008;6:30.:30.
53. Paynter NP, Chasman DI, Buring JE, Shiffman D, Cook NR, Ridker PM. Cardiovascular
disease risk prediction with and without knowledge of genetic variation at chromosome 9p21.3.
Ann Intern Med 2009;%20;150:65-72.
54. Talmud PJ, Cooper JA, Palmen J, Lovering R, Drenos F, Hingorani AD, Humphries SE.
Chromosome 9p21.3 coronary heart disease locus genotype and prospective risk of CHD in
healthy middle-aged men. Clin Chem 2008;54:467-474.
55. Ye S, Willeit J, Kronenberg F, Xu Q, Kiechl S. Association of genetic variation on
chromosome 9p21 with susceptibility and progression of atherosclerosis: a population-based,
prospective study. J Am Coll Cardiol 2008;52:378-384.
329
56. Karvanen J, Silander K, Kee F, Tiret L, Salomaa V, Kuulasmaa K, Wiklund PG, Virtamo
J, Saarela O, Perret C, Perola M, Peltonen L, Cambien F, Erdmann J, Samani NJ, Schunkert
H, Evans A. The impact of newly identified loci on coronary heart disease, stroke and total
mortality in the MORGAM prospective cohorts. Genet Epidemiol 2009;33:237-246.
57. Zee RY, Ridker PM. Two common gene variants on chromosome 9 and risk of
atherothrombosis. Stroke 2007;38:e111.
58. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
59. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
60. Schunkert H, Gotz A, Braund P, McGinnis R, Tregouet DA, Mangino M, Linsel-Nitschke
P, Cambien F, Hengstenberg C, Stark K, Blankenberg S, Tiret L, Ducimetiere P, Keniry A,
Ghori MJ, Schreiber S, El Mokhtari NE, Hall AS, Dixon RJ, Goodall AH, Liptau H, Pollard H,
Schwarz DF, Hothorn LA, Wichmann HE, Konig IR, Fischer M, Meisinger C, Ouwehand W,
Deloukas P, Thompson JR, Erdmann J, Ziegler A, Samani NJ. Repeated replication and a
prospective meta-analysis of the association between chromosome 9p21.3 and coronary artery
disease. Circulation 2008;117:1675-1684.
61. Jarinova O, Stewart AF, Roberts R, Wells G, Lau P, Naing T, Buerki C, McLean BW,
Cook RC, Parker JS, McPherson R. Functional Analysis of the Chromosome 9p21.3 Coronary
Artery Disease Risk Locus. Arterioscler Thromb Vasc Biol 2009.
62. Broadbent HM, Peden JF, Lorkowski S, Goel A, Ongen H, Green F, Clarke R, Collins
R, Franzosi MG, Tognoni G, Seedorf U, Rust S, Eriksson P, Hamsten A, Farrall M, Watkins H.
Susceptibility to coronary artery disease and diabetes is encoded by distinct, tightly linked
SNPs in the ANRIL locus on chromosome 9p. Hum Mol Genet 2008;17:806-814.
63. Amaral PP, Dinger ME, Mercer TR, Mattick JS. The eukaryotic genome as an RNA
machine. Science 2008;319:1787-1789.
64. Jarinova O, Stewart AF, Roberts R, Wells G, Lau P, Naing T, Buerki C, McLean BW,
Cook RC, Parker JS, McPherson R. Functional Analysis of the Chromosome 9p21.3 Coronary
Artery Disease Risk Locus. Arterioscler Thromb Vasc Biol 2009.
65. Hannon GJ, Beach D. p15INK4B is a potential effector of TGF-beta-induced cell cycle
arrest. Nature 1994;371:257-261.
66. Pomerantz J, Schreiber-Agus N, Liegeois NJ, Silverman A, Alland L, Chin L, Potes J,
Chen K, Orlow I, Lee HW, Cordon-Cardo C, DePinho RA. The Ink4a tumor suppressor gene
product, p19Arf, interacts with MDM2 and neutralizes MDM2's inhibition of p53. Cell
1998;%20;92:713-723.
67. Helgadottir A, Manolescu A, Helgason A, Thorleifsson G, Thorsteinsdottir U,
Gudbjartsson DF, Gretarsdottir S, Magnusson KP, Gudmundsson G, Hicks A, Jonsson T, Grant
SF, Sainz J, O'Brien SJ, Sveinbjornsdottir S, Valdimarsson EM, Matthiasson SE, Levey AI,
Abramson JL, Reilly MP, Vaccarino V, Wolfe ML, Gudnason V, Quyyumi AA, Topol EJ, Rader
330
DJ, Thorgeirsson G, Gulcher JR, Hakonarson H, Kong A, Stefansson K. A variant of the gene
encoding leukotriene A4 hydrolase confers ethnicity-specific risk of myocardial infarction. Nat
Genet 2006;38:68-74.
68. Helgason A, Palsson S, Thorleifsson G, Grant SF, Emilsson V, Gunnarsdottir S,
Adeyemo A, Chen Y, Chen G, Reynisdottir I, Benediktsson R, Hinney A, Hansen T, Andersen
G, Borch-Johnsen K, Jorgensen T, Schafer H, Faruque M, Doumatey A, Zhou J, Wilensky RL,
Reilly MP, Rader DJ, Bagger Y, Christiansen C, Sigurdsson G, Hebebrand J, Pedersen O,
Thorsteinsdottir U, Gulcher JR, Kong A, Rotimi C, Stefansson K. Refining the impact of TCF7L2
gene variants on type 2 diabetes and adaptive evolution. Nat Genet 2007;39:218-225.