Intended for healthcare professionals

Editorials

Alcohol and cardiovascular disease

BMJ 2014; 349 doi: https://doi.org/10.1136/bmj.g4334 (Published 10 July 2014) Cite this as: BMJ 2014;349:g4334
  1. M Maria Glymour, associate professor
  1. 1Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, CA 94115, USA
  1. mglymour{at}epi.ucsf.edu

New research tools will help us untangle this enigmatic association, eventually

Many observational studies report that people who drink moderate amounts of alcohol have lower cardiovascular risk than non-drinkers,1 but they cannot conclusively answer the most important question: will moderate drinking reduce cardiovascular risk? The distinction between the answers we derive from conventional observational studies (which are predictive) and the answers we need for improving health (which are causal) is fundamental. Mistaking predictive associations for cause and effect could, for example, lead us to confiscate matchbooks as a strategy to prevent lung cancer, because carrying matchbooks predicts future risk of lung cancer.

Recent decades have seen major breakthroughs in methods to evaluate cause and effect from observational studies.2 Despite their potential scientific value, the proposed advances are controversial.3 4 The paper by Holmes and colleagues (doi:10.1136/bmj.g4164) perfectly illustrates why.5 Holmes and colleagues try to evaluate whether moderate drinking helps prevent cardiovascular disease using a mendelian randomisation design.6 The paper explores only the direction of effects—whether drinking increases or decreases cardiovascular risk—because they omit the usual analysis in mendelian randomisation studies (instrumental variables analyses) which would estimate the magnitude of alcohol’s effects.7 8

Mendelian randomisation studies treat genetic background as “quasi-experiments”, assuming that inheriting the A-allele of the ADH1B locus is equivalent …

View Full Text

Log in

Log in through your institution

Subscribe

* For online subscription