Do heart and metabolic conditions cause worse COVID-19?
Of 17 heart and metabolic risk factors tested with genetic methods, only higher BMI showed evidence of causally raising the risk of being hospitalized with COVID-19.
Source
Cardiometabolic risk factors for COVID-19 susceptibility and severity: A Mendelian randomization analysis
Study at a glance
- Design
- Mendelian randomisation — Two-sample MR using published GWAS summary statistics for 17 cardiometabolic exposures and COVID-19 Host Genetics Initiative outcome GWAS (round 4), with pleiotropy-robust sensitivity analyses and multivariable MR
- N
- Summary-level data; primary susceptibility outcome 14,134 cases vs 1,284,876 population controls, primary severity outcome 6,406 hospitalized cases vs 902,088 controls, all of European ancestry
- Population
- Participants of mostly European ancestry in 22 cohorts contributing to the COVID-19 Host Genetics Initiative, plus large exposure GWAS consortia
- Outcome
- COVID-19 susceptibility (testing positive) and severity (hospitalization), plus five secondary COVID-19 outcome definitions
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The researchers used Mendelian randomization, which treats genetic variants that shift a risk factor as a natural experiment, to test 17 cardiometabolic exposures such as type 2 diabetes, BMI, blood pressure, lipids, coronary disease and kidney disease. Genetic effects on each exposure came from large published genome-wide studies, and genetic effects on COVID-19 came from the COVID-19 Host Genetics Initiative. The two primary outcomes were testing positive (14,134 cases) and being hospitalized (6,406 cases), each compared with population controls. They then ran sensitivity methods robust to pleiotropy and multivariable analyses to see whether BMI acted through obesity-related diseases.
What they found
Only BMI passed the strict multiple-testing threshold for severity: each 1 kg/m² genetically higher BMI was linked to 14% higher odds of COVID-19 hospitalization (odds ratio 1.14) and 6% higher odds of testing positive, though the latter did not meet the corrected threshold. None of the other exposures, including type 2 diabetes and blood pressure, showed significant associations. When coronary disease, stroke, kidney disease or type 2 diabetes were each added alongside BMI in multivariable models, BMI's direct effect disappeared, hinting that its effect may run through these conditions.
The limits
What it doesn't show
Genetic instruments explained only a small share of variance in each exposure, so modest causal effects of the other risk factors cannot be ruled out; null results are not proof of no effect. Pleiotropy-robust methods such as MR-Egger were weaker, so the BMI result could still be partly biased by pleiotropy. Participants were mostly European and non-randomly selected into biobanks, which risks collider/selection bias, and 'controls' were anyone not recorded as a case, so undetected infections could dilute effects. The analysis assumed linear effects and could not check differences by sex.
Key terms
- Mendelian randomization
- A method that uses inherited genetic variants linked to a risk factor as a natural experiment to test whether that risk factor causes a disease.
- Pleiotropy
- When one genetic variant affects several traits, which can violate MR assumptions if it influences the outcome through a route other than the exposure.
- Inverse-variance weighted (IVW) estimate
- The main MR estimate, combining single-variant estimates weighted by their precision.
- Multivariable MR
- An MR analysis that includes several exposures at once to estimate the direct effect of each.
- Genome-wide association study (GWAS)
- A study scanning many genetic variants across the genome for associations with a trait or disease.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
What does Mendelian randomization use as a proxy for an exposure?
Common questions
Why use genes instead of just comparing sick and healthy people?
Observational comparisons can be confounded because risk factors cluster together. Genes are assigned at conception, so genetically predicted differences in a trait are less likely to be tangled with lifestyle confounders.
Does this mean diabetes doesn't matter for COVID-19?
Not necessarily. The genetic instruments may have been too weak to detect a modest effect, and diabetes can still be a useful marker for identifying high-risk patients even if not itself causal.
What does it mean that BMI's effect vanished in multivariable models?
It suggests that part of BMI's effect on COVID-19 may be mediated by obesity-related diseases like diabetes or kidney disease, though shared genetic variants also make this hard to disentangle.
More on Infectious disease