Moderate drinkers only looked healthier because 'abstainers' included the sick

For decades, large studies of drinking and death rates have traced a J-shaped curve: people who drink a little appear to live longer than people who drink nothing at all. That finding shaped public advice and gave the nightly glass of wine a health halo. But the comparison only works if the non-drinkers are a fair benchmark. Many people classed as abstainers are actually former drinkers, and people often stop drinking precisely because their health is failing. This study set out to test whether that misclassification, along with other quality problems in the underlying research, explains the apparent benefit of light drinking.

The researchers conducted a systematic review and meta-regression of prospective studies on alcohol use and all-cause mortality. In total they analysed 87 studies covering 3,998,626 individuals, among whom 367,103 deaths were recorded. Drinkers were grouped by intake, with low-volume drinking defined as 1.3 to 24.9 g of ethanol per day and occasional drinking as less than 1.3 g per day. The analysis then adjusted for abstainer biases, such as former and occasional drinkers being misclassified as abstainers, and for other quality-related characteristics of the studies.

Without adjustment, the pooled data replicated the classic J-shaped curve: low-volume drinkers showed a relative risk of 0.86 for death from any cause, occasional drinkers 0.84, and former drinkers an elevated 1.22. After adjusting for abstainer biases and study quality, the apparent protection evaporated: low-volume drinkers had a relative risk of 0.97, which was not significantly different from lifetime abstainers. Analyses restricted to higher-quality, bias-free studies also failed to find any reduced mortality risk for light drinkers. Risk estimates for occasional drinkers were similar to those for low and medium-volume drinkers, which itself undermines the idea of a dose-dependent benefit.

This work directly challenges earlier pooled analyses that reported a protective effect of light drinking, arguing the effect was manufactured by how studies defined their comparison group rather than by anything alcohol does. It fits a broader pattern of scrutiny in which supposed benefits of moderate drinking shrink or disappear as methods improve. Because all of the underlying research follows people rather than assigning them to drink, it cannot settle what alcohol causes; it can only show that the correlation favouring light drinkers is fragile. The authors note implications for low-risk drinking guidelines and for how future studies should classify abstainers.

If part of the appeal of a modest drinking habit was the belief that it protects your heart or lengthens your life, this analysis is a reason to let that belief go. The people light drinkers were being compared against included former drinkers who had often quit because they were already unwell, which flattered the drinkers. Once that is corrected, light drinkers and lifetime abstainers die at similar rates. This is not a finding that light drinking is dangerous; it is a finding that it is not protective, which changes the calculation from a health decision to a lifestyle one.