key: cord-0717737-nl5cr61t
authors: Bhaskaran, Krishnan
title: New insights into the association between body-mass index and severe COVID-19
date: 2021-04-28
journal: Lancet Diabetes Endocrinol
DOI: 10.1016/s2213-8587(21)00109-1
sha: 429f991afe9865ab916219e115678fa955b7cb51
doc_id: 717737
cord_uid: nl5cr61t

nan

The emerging SARS-CoV-2 pandemic captured worldwide attention in the first months of 2020. 1 Most people who become infected have mild disease or are asymptomatic, but many have severe and often fatal disease. Older age quickly emerged as the most important risk factor for poor outcomes, but obesity was one of a number of other factors rapidly identified as being associated with COVID-19-associated mortality. 2 In The Lancet Diabetes & Endocrinology, Min Gao, Carmen Piernas, and colleagues 3 present detailed research on the risk of severe COVID-19 outcomes in people with high BMI. This comprehensive study builds on earlier evidence by examining a range of outcomes across the full spectrum of BMI. Using the QResearch database of electronic primary care records in England, UK, linked to SARS-CoV-2 testing results, hospital admissions, and death registration data, Gao and colleagues identified over 6·9 million individuals with at least one BMI record aged 20 years and older who had available data on the QResearch system between Jan 24 and April 30, 2020, among whom 13 503 hospital admissions, 1601 intensive care unit (ICU) admissions, and 5479 deaths due to COVID-19 occurred within the study period (data cutoff April 30, 2020). J-shaped associations were observed between BMI and COVID-19 hospital admissions and deaths; for hospital admissions, risk began to increase linearly above a BMI of 23 kg/m², whereas the increase in risk of death began to increase linearly at a slightly higher BMI of 28 kg/m², in keeping with the association between BMI and a wide range of cause-specific mortality outcomes. 4 A strong age interaction was observed, with unit increase in BMI being associated with large increases in risk for the youngest age group (20-39 years), and no association in the oldest age group (≥80 years).

The authors also identified an increased risk of admission to hospital and death due to COVID-19 at the lowest BMI levels, which is a new insight not addressed by early COVID-19 research that focussed on obesity. A limitation of the analysis presented by Gao and colleagues is that interactions were only assessed in those with a BMI of 23 kg/m² or higher, so that the association between underweight and risk in older age groups (in whom low BMI is common) could not be discerned. Interestingly, by contrast with other outcomes, BMI was approximately linearly associated with ICU admission throughout the BMI range, with no increase in risk at low BMIs. This observation is not straightforward to interpret. An ICU admission reflects not only severe disease, but also the complex clinical decision making process around whether a patient is likely to tolerate and benefit from intensive treatment. 5 The reduced risk of ICU admission in underweight individuals is likely to reflect the most atrisk underweight patients being deemed unsuitable for intensive care.

Another important limitation of the study is that over one million individuals had no BMI record available, and were excluded from the main analyses. The availability of many millions of routinely collected BMI records is a strength of UK primary care data sources, but missing data are a common problem, especially because BMI might be more likely to be recorded if outside the healthy weight range, violating the so-called missing at random assumption needed for popular multiple imputation approaches. 6 Gao and colleagues should be commended for providing descriptive data on these individuals, among whom a higher proportion were young (aged 20-39 years) and male than in the population with BMI measurements. Restriction of analyses to those with complete data can be a reasonable approach even if they are not fully representative of the broader patient population, provided the association between BMI and outcomes is correctly represented by those included. 7 A further important limitation is that investigating severe COVID-19 outcomes in a general population-based cohort does not allow one to disentangle whether BMI is associated with risk of infection, progression to severe disease once infected, or both. Mechanistic hypotheses to date have suggested that an imbalance in pro-inflammatory and anti-inflammatory cytokines at higher BMI might promote more severe COVID-19 disease than in those with a lower BMI. 8 However, any epidemiological study using available large-scale data is unlikely to be able to reliably isolate the association Shutterstock between BMI and outcomes in infected individuals because this would require large-scale representative testing data to identify both asymptomatic and symptomatic infection. In reality, COVID-19 testing in the UK and most other settings has been systematically targeted towards those with symptoms, making a positive test an insensitive and selective marker of infection, and those testing positive a biased sample in which to examine severe disease outcomes. For similar reasons, Gao and colleagues rightly avoided a focus on outcomes among patients already in hospital. 9 People with a BMI of more than 40 kg/m² were recognised early on in public health guidance as being at increased risk of severe COVID-19 outcomes, and BMI is now being used in risk prediction tools that are informing vaccine prioritisation, 10 highlighting the importance of obtaining a detailed understanding of the association between BMI and COVID-19 outcomes to inform policy. Key future research priorities will be to establish whether BMI affects vaccine efficacy, and to understand whether people outside the BMI range considered to be healthy (18·5-24·9 kg/m²) are at increased risk of post-COVID-19 sequelae. Further careful epidemiological study of these and other emerging questions will inform the ongoing public health response to this new disease that is likely here to stay. 

World Health Organization

Factors associated with COVID-19-related death using OpenSAFELY

Associations between body-mass index and COVID-19 severity in 6·9 million people in England: a prospective, community-based cohort study

Association of BMI with overall and cause-specific mortality: a population-based cohort study of 3·6 million adults in the UK

Decision-making around admission to intensive care in the UK pre-COVID-19: a multicentre ethnographic study

What is the difference between missing completely at random and missing at random?

Bias and efficiency of multiple imputation compared with complete-case analysis for missing covariate values

Obesity, adipokines and COVID-19

Collider bias undermines our understanding of COVID-19 disease risk and severity

Living risk prediction algorithm (QCOVID) for risk of hospital admission and mortality from coronavirus 19 in adults: national derivation and validation cohort study