Context

Many observational studies find associations between ‘moderate’ alcohol use and reduced mortality risk compared with abstention. Low doses of alcohol have also been associated with improvements in some biomarkers of cardiovascular health in short-term randomised studies. However, the absence of any randomised clinical trial of any morbidity or mortality outcome, and contradictory Mendelian randomisation studies,1 cast doubt on the veracity of associations found in observational studies. Goulden lists alternative explanations for the classic J-shaped curve for alcohol use and mortality risk, including selection bias, misclassification of drinkers as abstainers and residual confounding. We comment here as authors of a meta-analysis2 raising identical concerns and reporting similar findings.