COVID-19 is an inflammatory syndrome caused by novel coronavirus SARS-CoV-2. Symptoms range from mild infection to severe acute respiratory distress syndrome (ARDS) requiring ventilation and intensive care. At the time of data collection, UK cases were around 300 000 with a fatality rate of 13% necessitating over 10 000 critical care admissions; now there have been over 4 million cases. Nutrition is important to immune function and influences metabolic risk factors such as obesity and glycaemic control, as well as recovery from acute illnesses. Poor nutritional status is associated with worse outcomes in ARDS and viral infections, yet limited research has assessed pre-morbid nutritional status and outcomes in patients critically unwell with COVID-19.
Investigate the effect of body mass index (BMI), glycaemic control and vitamin D status on outcomes in adult patients with COVID-19 admitted to an intensive care unit (ICU).
Retrospective review of all patients admitted to a central London ICU between March and May 2020 with confirmed COVID-19. Electronic patient records data were analysed for patient demographics; comorbidities; admission BMI; and serum vitamin D, zinc, selenium and haemoglobin A1c (HbA1c) concentrations. Serum vitamin D and HbA1c were measured on admission, or within 1 month of admission to ICU. Primary outcome of interest was mortality. Secondary outcomes included time intubated, ICU stay duration and ICU-related morbidity.
Seventy-two patients; 54 (75%) men, mean age 57.1 (±9.8) years, were included. Overall, mortality was 24 (33%). No significant association with mortality was observed across BMI categories. In the survival arm admission, HbA1c (mmol/mol) was lower, 50.2 vs 60.8, but this was not statistically significant. Vitamin D status did not significantly associate with mortality (p=0.131). However, 32% of patients with low vitamin D (<25 IU/L) died, compared with 13% of patients with vitamin D levels >26 IU/L. Serum zinc and selenium, and vitamin B12 and folate levels were measured in 46% and 26% of patients, respectively.
Increased adiposity and deranged glucose homeostasis may potentially increase risk of COVID-19 infection and severity, possibly relating to impaired lung and metabolic function, increased proinflammatory and prothrombotic mechanisms. Vitamin D deficiency may also associate with poorer outcomes and mortality, supporting a possible role of vitamin D in immune function specific to pulmonary inflammation and COVID-19 pathophysiology. There are plausible associations between raised BMI, glycaemic control, vitamin D status and poor prognosis, as seen in wider studies; however, in this service evaluation audit during the first wave of the pandemic in the UK, with a limited data set available for this analysis, the associations did not reach statistical significance. Further research is needed into specific nutritional markers influencing critical care admissions with COVID-19.
This snapshot audit in a London ICU from the first peak of the COVID-19 pandemic adds to the body of evidence associating overweight/obesity as a risk factor of disease severity and mortality in COVID-19.
Further data suggesting baseline vitamin D status may influence severity of COVID-19 as higher mortality was observed in individuals with vitamin D deficiency (<25 nmol/L).
This highlights baseline nutritional parameters that are easily obtainable in the critical care setting and suggests additional nutritional markers worthy of further research.
As COVID-19 cases continue to rise worldwide, so does the role of nutritional factors on the agendas of government, public health, primary and secondary healthcare bodies. One aspect of clinical importance is the role of nutrition in critical care, particularly how specific baseline parameters of nutritional status may influence illness severity, morbidity and mortality. At the time of data collection and analysis in September 2020, UK cases have exceeded 300 000 with a fatality rate of ~13% necessitating >10 000 critical care admissions.
As SARS-CoV-2 is a novel coronavirus, it is not yet fully understood which specific nutritional parameters are most impactful on disease severity and outcomes. However, it is well established that nutrition plays a key role in immune function and several macronutrients and micronutrients are required for key physiological processes during infection.
The rationale for this audit was to present data on nutritional parameters of patients with COVID-19 from a real-world intensive care unit (ICU) setting. We wanted to determine whether there were any clear associations between mortality and nutritional parameters that (1) are commonly measured in clinical practice, and (2) have been identified as possible risk factors for worse outcomes in the care of inpatients with COVID-19. Parameters of particular interest were obesity, glycaemic control and vitamin D status.
It has been demonstrated that higher body mass index (BMI) and obesity are associated with poorer outcomes in COVID-19.
Other established risk factors for worse COVID-19 outcomes include cardiometabolic diseases: type 2 diabetes mellitus (T2DM), hypertension and dyslipidaemia.
Another area of interest is the role of micronutrients and how individual baseline levels may impact morbidity and mortality. Vitamin D status has attracted much attention as a secosteroid with biological roles in immune function and anti-inflammatory, antifibrotic and antioxidant effects.
An association between vitamin D status and respiratory tract infections, tuberculosis and chronic obstructive pulmonary disease is long-established and supplementation in individuals with low vitamin D levels has improved outcomes in interventional studies.
Several other micronutrients are essential to healthy immune responses and many may have a role in COVID-19 outcomes, particularly in deficiency. The practice of measuring serum levels of other micronutrients varies greatly between individuals. Nonetheless, for completeness of data regarding nutritional parameters, we also wanted to present recorded levels of zinc, selenium, magnesium, vitamin B12 and folate for all patients in whom these were measured.
We aimed to investigate the relationship between BMI, glycaemic control and vitamin D status, respectively, on outcomes in patients admitted to the ICU with COVID-19 pneumonia. Primary measure was mortality. Secondary measures included time intubated, ICU stay duration and ICU-related morbidity.
We undertook a retrospective review of all patients admitted to a West London ICU between 18 March and 18 May 2020 with confirmed COVID-19, as confirmed clinically and by a positive oropharyngeal or nasopharyngeal SARS-CoV-2 PCR test. No patients were excluded from analysis.
Data were extracted from the electronic patient database. The following data were collected: patient demographics; comorbidities; BMI on admission (using ICU bedscale or patient database); serum vitamin D concentration; mean weekly random plasma glucose and plasma haemoglobin A1c (HbA1c). Serum vitamin D concentration and HbA1c were taken on admission, or within 1 month of admission to ICU. The serum C reactive protein (CRP) was also recorded at the same time point as vitamin D. The requirement for insulin, steroid, antibiotic, intravenous water-soluble vitamin concentrate (the preparation used in this department was Pabrinex, which contains thiamine, riboflavin, pyridoxine hydrochloride, ascorbic acid, nicotinamide and glucose) and renal replacement therapy at any time point during the admission was also recorded. The primary outcome measure was mortality. Secondary outcome measures included the time intubated, the length of ICU stay and ICU-related morbidity.
Statistical analysis was undertaken using SPSS V.24 (IBM). Statistical significance was set as p<0.05. Continuous parametric data sets were compared by the t-test, and non-parametric data sets by the Mann-Whitney U test. Categorical data were compared using Χ2 test. Data are presented as mean (±SD) or median (IQR).
In total, 72 patients were included in the service evaluation (see
Mean BMI was 29.5 kg/m2 (SD ±5.7) for the survival group and 28.2 kg/m2 (SD ±4.8) for the non-survivors. No patients were admitted with a BMI <18.5 kg/m2. No significant difference in mortality was observed across different BMI categories, p=0.589 (
BMI categories and primary outcome of survival
BMI, kg/m2 | Survived, n=48 | Died (n=24) | P value |
8 (61.5) | 5 (38.4) | 0.589 | |
22 (64.7) | 12 (35.3) | ||
10 (71.4) | 4 (28.6) | ||
4 (57.1) | 3 (42.9) | ||
4 (100.0) | 0 (0.0) |
BMI, body mass index.
Admission HbA1c was determined by either electronic patient records from the 1-month period prior to admission or, if unavailable, was taken at admission. Mean HbA1c levels were less overall in the survival group than in the non-survivors, 50.2 mmol/mol (SD ±17.0) vs 60.8 mmol/mol (SD ±34.7), but this was not statistically significant, p=0.284. There was no statistical difference in outcome between patients who took insulin prior to admission and those with diet-controlled or tablet-controlled T2DM.
Weekly mean plasma glucose control was also assessed across the cohort of 71 (one individual’s data were not fully available and therefore were omitted) and indicated no statistically significant difference between survivors and non-survivors. Overall, mean plasma glucose was deemed well controlled across both groups with mean glucose levels of 8.1 mmol/L (SD ±1.9) and 8.9 mmol/L (SD ±2.3), respectively (
Mean plasma glucose control during the total length of stay
Week | Mean plasma glucose, mmol/L (SD) | ||
Survived | Died | P value | |
1 (n=71) | 8.0 (2.2) | 9.0 (2.7) | 0.125 |
2 (n=66) | 8.3 (2.2) | 9.3 (2.4) | 0.132 |
3 (n=47) | 7.9 (1.9) | 9.1 (2.4) | 0.063 |
4 (n=32) | 8.0 (1.8) | 9.3 (2.7) | 0.112 |
5 (n=19) | 8.2 (2.1) | 8.8 (2.2) | 0.618 |
6 (n=7) | 7.9 (2.3) | 11.0 | 0.267 |
7 (n=5) | 7.4 (1.7) | 8.7 | 0.551 |
8 (n=2) | 7.4 (1.8) | – | – |
9 (n=1) | 8.4 | – | – |
8.1 (1.9) | 8.9 (2.3) | 0.083 |
A limited analysis of vitamin D status was possible as 51 (71%) of the 72 patients included had a recorded vitamin D level either on admission or within 1 month of admission. If this was taken on admission, it was ensured that a correlating CRP was taken to aid interpretation. We subcategorised this measure into three groups as per UK national cut-offs for deficient, insufficient and sufficient serum concentration of 25-hydroxyvitamin D (25(OH)D) (
Subcategories of vitamin D levels measured
25-hydroxyvitamin D (nmol/L) | Frequency | ||
Survived, n=40 | Died, n=11 | P value | |
17 | 8 | 0.131 | |
15 | 3 | ||
8 | 0 |
Χ2 test.
Further analysis calculated the mean BMI of each group subcategorised by vitamin D status (
Mean BMI as per vitamin D status
25-hydroxyvitamin D (nmol/L) | ||||
<25 (n=25) | 26–50 (n=18) | >50 (n=8) | P value | |
30.7 (5.9) | 30.0 (5.7) | 26.7 (5.6) | 0.238 |
BMI, body mass index.
We also analysed the correlation between CRP when serum concentration of 25(OH)D and found no significant correlation (
Vitamin D status (25(OH)D) in nmol/L versus CRP (ng/L). Pearson correlation: −0.053 (p=0.718). 25(OH)D, 25-hydroxyvitamin D; CRP, C reactive protein.
In this group, selenium and zinc were measured in 33 of 72 (46%) patients. For selenium, the mean serum level was 0.88 µmol/L (range 0.64–1.6 µmol/L; normal reference range). For zinc, the mean serum level was 10.2 µmol/L (range 3.0–20.1 µmol/L; normal reference range).
Folate and vitamin B12 levels were measured in 19 of 72 (26%) patients. The mean vitamin B12 level was 803 ng/L (range 285–2000 ng/L; normal reference range). The mean serum folate was 5.4 µg/L (range 2.1–20 µg/L; normal reference range).
Our patient demographics for this audit aligned with the national picture in ICU at the time of data collection comparing with the Intensive Care National Audit & Research Centre (ICNARC) data. This was comparable for age and gender and can be considered nationally representative in terms of ethnicity, with similar proportions of white, Asian and black patients
Patient demographics versus ICNARC data
Characteristic | Audit sample (n=72) | ICNARC (n=10 624) | ||
n (%) | Median | n (%) | Median | |
Age | 57–60 | 60 | ||
Gender (%) | ||||
Male | 75 (66.7*) | 70.2 (58.3*) | ||
Female | 25 (66.6*) | 29.8 (64.9*) | ||
Ethnicity | ||||
White | 68.1 | 66.2 | ||
Asian | 9.5 | 15.6 | ||
Black | 11.8 | 9.6 | ||
Mixed/other | 10.6 | 9.6 | ||
BMI (kg/m2) (%) | ||||
<18.5 | 0 | 0.8 | ||
18.5–24.9 | 18.0 (61.5*) | 25.6 (59.5*) | ||
25.0–29.9 | 47.2 (64.7*) | 34.4 (57.9*) | ||
30.0–40.0 | 29.2 (66.7*) | 31.4 (63.6*) | ||
≥40.0 | 5.6 (100*) | 7.9 (65.3*) |
*Survived critical care.
BMI, body mass index; ICNARC, Intensive Care National Audit & Research Centre.
A high percentage of our cohort were both older and men observing a higher case mortality rate in male patients, consistent with data from across China, Europe and the USA.
Addressing the notion associating male sex with development of more severe disease has been observed in much larger studies analysed in a global meta-analysis which observed increased rate of ICU admissions in men seen as a proxy for disease severity.
Our audit data observed a higher proportion of patients in the overweight than obese category in keeping with the trend seen in the ICNARC data. However, our data were less representative of individuals within BMI range 18.6–24.9 kg/m2 and while acknowledging this as a limit there remained comparable trends to national data sets as further outlined. We noted that survivorship data were comparable across all BMI groups when cross-referencing to the ICNARC data set (
While the results may be at odds with other recently published data on the relationship of severe obesity and disease severity, our data are in keeping with the ICNARC data and suggest obesity may increase risk of ICU admission but is not necessarily an individual marker for poor prognosis when used in isolation.
First, there may have been substantial differences in the proportion of patients who died in the normal weight compared with overweight/obese categories, without there being an independent association between BMI and mortality, if controlling for the effect of metabolic and inflammatory derangements we know to be associated with worse disease outcomes.
The observation in patients in the highest BMI category is in line with previous meta‐analyses showing inverse associations with mortality in ICU patients with severe obesity.
A number of physiological mechanisms may provide an explanation for these observations. Central obesity as measured by abdominal girth versus BMI status is a causal factor in atelectasis, which can have negative effects on pulmonary compliance, functional residual capacity and, ultimately, arterial oxygenation.
Overall, there was a significantly higher proportion of patients with known T2DM in the group who died (42%) compared with those who survived (19%). On average, HbA1c levels were lower in those who survived but this difference did not reach statistical significance. This is in line with national findings whereby the largest UK-based population study found individuals with T2DM had increased odds of in-hospital death with COVID-19 (OR 2.03) compared with those without.
The pathophysiological interplay between COVID-19 and T2DM is not fully understood; however, it is well recognised that infection itself causes metabolic complications associated with adverse clinical outcomes. COVID-19 can precipitate and worsen acute metabolic complications associated with both T1DM and T2DM, such as diabetic ketoacidosis and hyperglycaemia, which may impact mortality rates in this group.
Targets for glycaemic control within the ICU setting are controversial with no definitive consensus on what constitutes ‘optimal control’. There is agreement when understanding the detrimental effects of both hyperglycaemia and hypoglycaemia but not as to what targets most optimally minimise this.
In this audit, mean plasma glucose levels were within a well-controlled range for the duration of both groups of ICU stay: 8.1 mmol (SD 1.9) in the survival group and 8.9 mmol (SD ±2.3) in the non-survivors (p=0.083) with an overall mortality rate of 33%. HbA1c levels were lower overall in the survival group (50.2 mmol/mol vs 60.8 mmol/mol), but this association did not reach statistical significance. NHS England data comparing COVID-19 outcomes in patients with higher HbA1c (>86 mmol/mol) versus those with HbA1c between 48 and 53 mmol/mol found mortality increased with higher HbA1c in patients with both T1DM and T2DM.
In our cohort, we observed a negative correlation between vitamin D status and BMI, though this association did not reach statistical significance. Beyond the association with BMI, vitamin D deficiency (<25 nmol/L) was seemingly associated with increased mortality in the group, whereby 8 of the 25 patients with vitamin D deficiency died. Conversely, all patients with vitamin D status >50 nmol/L survived, posing the question as to whether an improved vitamin D status contributes to better outcomes in the ICU.
Multiple analyses of UK Biobank data have found a higher frequency of vitamin D deficiency in patients with COVID-19 when compared with population-based controls.
The association of low vitamin D status with higher BMI could be explained by physiological, behavioural, genetic and even geographical determinants.
Vitamin D deficiency in critical illness is known to be associated with poorer outcomes and supplementation has largely been demonstrated as safe; however, defining optimum target ranges, clinical assays and measurable metabolites has proven difficult.
Retrospective meta-analyses from American and European cohorts have indicated low vitamin D status to be associated with higher rates of ICU admission in patients with COVID-19. Similar to trends observed in our analysis, this translated to poorer prognosis, while also highlighting significantly lower vitamin D status in ICU compared with patients receiving ward-based care.
This increasing depth of observational data has resulted in calls for more clinical trials to evaluate a potential therapeutic role of vitamin D in COVID-19. The results of the first of these pilot studies have recently been published from Spain, which assessed oral supplementation of high-dose calcifediol (25(OH)D) alongside combination drug therapy in patients with COVID-19, demonstrating a reduction in critical care admissions in those supplemented.
In this audit, we collected all micronutrient data that were measured in this cohort. Serum selenium and zinc were measured in only 46% of patients, and vitamin B12 and folate measured in even fewer, despite these being important micronutrients for immune function and their measurements being commonly available laboratory tests.
The latest clinical guidelines from the European Society for Clinical Nutrition and Metabolism on the nutritional management of patients with COVID-19 highlight the importance of ensuring sufficient levels of essential micronutrients to ‘potentially reduce disease negative impact’.
One harsh reality is that, since the majority of patients hospitalised with COVID-19 will be elderly and with other comorbidities, many will not be offered more intensive treatments available in ICU. Therefore, clinical audit and trials from outside the ICU setting are also crucially important. From a practical perspective, it would be wise to frame sensible clinical nutrition interventions such as ensuring micronutrient adequacy as the responsibility of all healthcare providers caring for patients with COVID-19 and people at risk.
In the absence of robust interventional data, the nutritional management of patients with COVID-19 in critical care remains consensus guided and based on foundational principles. Broadly, these are: routine screening for malnutrition; supplementation for insufficiency and deficiency; a ‘food first’ approach, including dietary modification for nutritional content and texture; oral nutritional supplementation, and escalation to enteral, then parenteral, nutrition where appropriate.
The findings of this retrospective analysis are time and location specific with limitations in terms of generalisability across different geographical and demographic contexts. We have referenced national and international data sets for comparisons to guide applicability and generalisability to similar regions. The small sample size and retrospective nature of the data collection were never powered to answer a specific clinical question. Logistical barriers such as unprecedented workload and cumbersome personal protective equipment requirements likely impacted accuracy of some measurements, particularly accurate weight and height. Indeed, for many parameters, for example vitamin D, a proportion of patients had no measurement recorded, which poses a high risk of bias to the results in such a small sample. The gaps in our data also made it difficult to account for probable confounders (age, sex, ethnicity, socioeconomic status) when statistically analysing associations between parameters and outcomes. Shared access to electronic community healthcare records provided access to recently recorded data where these were not immediately available, for example, height and weight, vitamin D level and HbA1c. Selection bias was minimised by including all individuals during the time period, but information bias could not be fully mitigated against.
Collectively, these limitations also restrict our ability to establish any causality (consistency, temporality, specificity, coherence, experiment and analogy) to define relationships between nutritional measures and COVID-19 outcomes. If more nutritional data from similar settings are reported, this could allow judgements of consistency in relationships across studies in the future.
Baseline nutritional measurements, including capillary glucose, haemoglobin A1c (if type 2 diabetes mellitus present) and serum micronutrients such as vitamin D, selenium and zinc can be readily made in a critical care context in this pandemic.
Obtaining accurate measures of weight and correlating with body mass index may help as part of risk stratification when assessing clinical prognosis.
This intensive care unit department achieved adequate control of capillary blood glucose levels despite significant logistical challenges and with increasing use of dexamethasone may warrant further investigation.
Vitamin D status in the critically ill patient is easily obtainable and treatment to address deficiency remains in keeping with consensus guidelines.
This was a single-centre service evaluation of clinical and biochemical data relating to nutritional status from patients with COVID-19 admitted to ICU during the first peak of the 2020 pandemic. We were able to measure markers of nutritional status such as serum micronutrients, vitamin D levels and blood glucose more accurately when compared with anthropometric measures such as height and weight, owing to the unstable nature of this patient group and increased clinical pressures during this phase of the pandemic. There was a high prevalence of T2DM and vitamin D insufficiency/deficiency in this ICU cohort and T2DM was more prevalent in those who died from COVID-19. Of these, low vitamin D is more straightforward to remedy in the acute setting, which supports the notion that active detection and correction of vitamin D deficiency is warranted in patients with, and at greater risk of, COVID-19. However, our data suggest that BMI in isolation may not be an accurate predictor of mortality in COVID-19 without consideration of a fuller clinical picture.
With our evolving understanding of COVID-19 and the interplay with nutrition, this presented audit helps to signpost to important markers for consideration and advocates for better standardisation in nutrition-related measures for future audits. Alongside BMI, further anthropometric data collection may help refine our risk stratification and subsequent public health advice. Equally by measuring HbA1c as routine would also help to risk stratify patients in and outside of the critical care setting as steroid therapy is common practice which subsequently impacts glycaemic control and may help to mitigate additional complications. Further nutritional measures of micronutrient status specifically a panel implicated in immune function and inflammatory mediation (vitamins C, D, E, zinc and selenium) could help to identify individuals at greater risk. This would also help to establish if simple, low-risk interventions such as targeted micronutrient supplementation confer clinical benefit and could help to explore interventions which are not exclusive to the critical care setting. Furthermore, prospective data capture, larger observational studies and robust, well-designed interventional trials are required to better establish best practice for clinical nutrition approaches and the management of nutritional and metabolic comorbidities in patients with COVID-19.
We would like to thank the NNEdPro COVID-19 Taskforce and the wider clinical team at Imperial College Healthcare NHS Trust for the support during the data collection process and the write up of this service evaluation.
@DrTimothyEdenRD, @ShaneMacZ
TE, JN and SR planned the above service evaluation and determined the data collection. TE and JN collected the relevant data and JN and SR provided the statistical analysis. TE, SM and DC contributed to the report writing and discussion of this service evaluation. SR supervised the undertaking of the work and contributed to the overall revisions prior to submission by TE.
The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
None declared.
Not commissioned; externally peer reviewed by Marcello Scopazzini, NHS Lothian, Edinburgh, UK.
This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
All data relevant to the study are included in the article or uploaded as supplemental information. All relevant data and statistical analyses have been included within the submission. For further information and additional data, please contact the first author Timothy Eden (timothy.eden@nhs.net).
Not required.