Dementia risk in the population over time: potential for primary prevention and intervention Holly Qing Tian Bennett

Dementia prevalence and incidence has decreased over time in the UK. This will in part be due to changes in dementia risk. Although dementia risk has been studied extensively, there is relatively little literature on change in dementia risk over time and current evidence is conflicting. If the risk factor profile of dementia changed this would impact prevention and intervention strategies as well as future health service and care needs. This thesis explores the change in dementia risk factor profiles in the UK population, changes across time and implications.

Risk factor prevalence and their association with dementia were examined using the two Cognitive Function and Ageing Studies (CFAS I and II), both large UK population based studies. Baseline interviews for CFAS I began in 1991 and for CFAS II in 2008 with identical study dementia diagnosis to provide reliable estimates of dementia risk. Inverse probability weights were used to adjust for initial non-response and multiple imputation was used to account for item non-response. Relative risk of incident dementia was measured by Poisson regression accounting for person-years. Changes in prevalence and relative risk would result in changes to Population Attributable Fractions (PAFs) of dementia that measure the proportion of incident dementia cases associated with risk factors. Three combined PAF models were considered. An early/midlife model, a health condition model and a proximal model. The full model included the risk factors from all three models. To explore whether dementia prevalence will continue to decline in the future taking into account current risk factor trends, number of people with dementia and dementia prevalence in the UK were forecasted until 2040. Considering the possibility of population interventions, higher education was used as a case example. A systematic review was first conducted, followed by analysis on lifelong education and cognition using Structural Equation Modelling in a further population based study that focused on healthy ageing (the Cambridge Centre for Ageing and Neuroscience (CamCAN) study, initiated 2010).

The risk results suggested that having an unskilled occupation compared to a semi-skilled occupation and currently smoking compared to never smoking were associated with increased risk of dementia in CFAS II but not CFAS I. Other associations remained stable with stroke, loneliness and functional impairment associated with increased dementia risk and higher education with decreased dementia risk in both studies. There were more changes in prevalence of risk factors over time, the largest changes being in prevalence of education, smoking, hypertension and diabetes. The fully combined PAF model was associated with a greater proportion of incident dementia cases in CFAS II than in CFAS I. This was mainly due to an increase in the proportion of incident dementia cases associated with proximal risk factors. Early/midlife risk factors and health condition risk factors were associated with similar proportions of incident dementia cases in both studies. In the future, number of people with dementia and dementia prevalence are expected to increase. However, increases in education and prevention of midlife obesity and stroke could greatly attenuate expected future dementia cases. Finally, current literature on lifelong education and cognition provide support for a role of higher education as an intervention. The CamCAN analysis addressed some of the gaps in the literature and suggested that education in later life in addition to higher education in young adulthood was associated with better cognition.

These findings add to the growing literature on dementia risk. The dementia risk profile has changed over time indicating a shift towards proximal risk factors for prevention. Throughout, higher education has been highlighted as an important protective factor and increasing higher education could potentially attenuate future expected dementia cases. To more accurately estimate the impact of increasing higher education on future dementia, longitudinal models accounting for mortality based on data from the UK are essential.