Comment on Baranova et al. Causal Associations Between Basal Metabolic Rate and COVID-19. Diabetes 2023;72:149–154

We read with interest the article in Diabetes by Zhang and colleagues (1), which claims to provide evidence that higher basal metabolic rate (BMR) causally increases susceptibility to and severity of COVID-19.

This conclusion is, unfortunately, undermined by a fundamental methodological problem. The Mendelian randomization approach on which this study is based relies on the availability of a strong and valid genetic instrument for the exposure, in this case BMR. The genetic instrument used in this article is derived from a genome-wide association study (GWAS) of BMR in the UK Biobank study (2). However, BMR was not measured directly in any participants in the UK Biobank study. Rather, it was predicted on the basis of parameters entered or measured in the course of estimation of body composition by segmental impedance; interestingly, the equation used in this case is not available even on the UK Biobank Resource website (https://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=23105), making it difficult to fully reproduce any results.

BMR is expensive and time-consuming to measure directly, so it is rarely assessed in large population studies. In contrast, there are a host of readily available equations to predict BMR that typically use a combination of age, sex, and measures of body size or body composition, parameters that are frequently available in large cohort studies like UK Biobank. The availability of predicted BMR in such large studies alongside genetic data explains why a number of researchers have conducted similar Mendelian randomization studies that use genetic associations derived from predicted BMR, for example, with cancer as an outcome (3).

All such analyses are unsound because it is likely that the genetic variations they rely on are not specific to BMR itself but reflect instead the components of the equations used to predict BMR. The variance in BMR not explained by body size, age, or sex is the interesting and relevant element for a Mendelian randomization study of BMR. Unfortunately, this cannot be captured by such an analysis, and no amount of adjustment for confounding, including for BMI, can deal with the problem that the GWAS that has been undertaken is one of a predicted rather than a measured phenomenon.

An alternative interpretation of the results presented is that fat mass increases COVID-19 susceptibility, and there are considerable supportive data for this (4). What is purported in this article to be a causal link between BMR and COVID-19 is instead a result of the mistaken use of a predicted value of BMR that strongly relies on body composition, explaining why FTO and MC4R are the lead genetic variants for BMR in the GWAS used in this study (2,5).