Moderate drinking and mortality: why the apparent benefit of a daily drink fades in better studies

Few pieces of health folklore have been as welcome as the idea that a drink a day is good for you. It had respectable science behind it. For decades, studies that followed large groups of people found that those who drank a little died later, on average, than those who drank nothing, and far later than those who drank a lot. Plotted on a graph, the risk dipped before it climbed, in a shape that gave the finding its name: the J-shaped curve.

The dip is now in serious doubt. In 2023 Jinhui Zhao, Tim Stockwell and colleagues at the University of Victoria in Canada re-examined the whole body of evidence in JAMA Network Open. Their argument is that the curve was produced less by alcohol than by a mistake in who the drinkers were compared with, and that when the mistake is corrected the benefit of light drinking shrinks to nothing that can be measured.

Where the idea that moderate drinking is healthy came from

The observation itself is not in dispute. Study after study has found lower death rates, especially from heart disease, in people who report one or two drinks a day. Biological explanations were offered, and red wine acquired a reputation of its own.

All of this evidence is observational. Nobody has randomly assigned thousands of people to drink or abstain for decades, so the comparison is always between people who chose to drink moderately and people who, for whatever reason, did not drink.

Why the non-drinkers skew studies of moderate drinking

That phrase, for whatever reason, is where the trouble lies. Non-drinkers are a mixed group. Some have never drunk. Others used to drink and gave it up, and the Victoria team note that former drinkers include many of whom cut down or stop for health reasons. Epidemiologists call them sick quitters. If those who stopped because of liver disease, heart trouble or medication are counted as abstainers, the abstainers will look unhealthy and everyone else will look good beside them.

The moderate drinkers, meanwhile, are not a random sample either. The authors point out that light and moderate drinkers are systematically healthier than current abstainers on a range of health indicators unlikely to be associated with alcohol use, and they list dental hygiene, exercise, diet, weight and income. Nobody thinks a glass of wine improves a person’s teeth. A more plausible reading is that moderate drinking is a marker of a comfortable, orderly life.

A third distortion builds with age. Heavy drinkers who die in their fifties never enter a study of seventy-year-olds, so older cohorts contain the survivors. The authors found that the apparent protection of the heart was only apparent among older cohorts that are more exposed to lifetime selection biases, meaning the cumulative filtering of who is still alive and still drinking.

What 107 studies show about moderate drinking and death

The review identified 107 studies of alcohol use and deaths from all causes published from 1980 to July 2021, covering 4.8 million people and more than 425,000 deaths. The team recorded how each study had defined its non-drinkers, how old its participants were, and how well it had handled the known sources of bias, and then built those features into the analysis.

Taken at face value, the studies showed the familiar dip. Low-volume drinkers, defined as those consuming up to about 24 grams of alcohol a day, had death rates 15% lower than abstainers, and former drinkers had rates 22% higher.

After correction, the picture changed. The advantage for low-volume drinkers shrank to 7% and could no longer be told apart from chance. For occasional drinkers, who have less than one drink a week, it was 4%. At the other end, the corrected risk rose with dose: slightly higher at 25 to 44 grams a day, 19% higher at 45 to 64 grams and 35% higher above that. For readers used to counting drinks, a US standard drink holds about 14 grams, so 45 grams is a little over three.

Women fared worse than men at every level. The abstract summarizes that increased risk was evident at higher consumption levels, starting at lower levels for women. Among women, death rates were already about a fifth higher at 25 to 44 grams a day.

The team also tried a different yardstick. Lifetime abstainers may themselves be unusual, so they reran the analysis against occasional drinkers, arguing that occasional drinkers are a more appropriate reference group because the tiny amounts they drink could hardly affect their bodies. The pattern held, with no advantage for moderate drinkers and higher risk for heavy ones.

What other large analyses of alcohol and mortality have found

The Victoria review has company. In 2018 Angela Wood and colleagues at the University of Cambridge combined individual records on nearly 600,000 current drinkers, which sidesteps the abstainer problem by leaving non-drinkers out altogether. Death rates rose with consumption, with the minimum mortality risk around or below 100 g per week, roughly seven US drinks. Compared with that level, heavier drinkers had lower life expectancy at age 40 years of approximately 6 months, 1-2 years, or 4-5 years in three rising bands of intake.

The Cambridge study also showed why alcohol’s record on the heart is mixed. More drinking went with fewer heart attacks but with more strokes, more heart failure and more deaths from high blood pressure.

The Global Burden of Disease project, which estimates the causes of illness for every country, reached the starkest conclusion the same year: the level of consumption that did least harm across health outcomes was zero. Its 2022 update was more qualified. For adults under 40 the safest amount was still at or near zero, because alcohol’s harms at that age are dominated by injuries. For older adults in most regions, small amounts carried little or no excess risk. The authors concluded that there is strong evidence to support recommendations on alcohol consumption varying by age and location.

Genetics offers a separate line of evidence. Some of us carry gene variants that lead to drinking less, and because genes are assigned at conception they cannot be the result of illness or income. Kiran Biddinger and colleagues at the Broad Institute used this approach, known as Mendelian randomization, in 371,463 UK Biobank participants. They first confirmed the confounding: light to moderate alcohol consumption was associated with healthier lifestyle factors, and adjusting for those factors weakened the apparent benefit. The genetic analysis then showed risk rising at every level, gently at first: light alcohol intake was associated with minimal increases in cardiovascular risk, whereas heavier consumption was associated with exponential increases in risk.

What health bodies say about alcohol

Official statements have hardened. In 2023 the World Health Organization’s European office declared that there is no safe amount that does not affect health. Its reasoning rests on cancer more than on the heart. Alcohol is classed as a cause of at least seven cancers, and the WHO notes that around half of all alcohol-attributable cancers in the WHO European Region occur among people whose alcohol consumption does not meet the definition of heavy drinking.

On the disputed benefits, the statement quotes Jürgen Rehm, a senior scientist in addiction research in Toronto: “Potential protective effects of alcohol consumption, suggested by some studies, are tightly connected with the comparison groups chosen and the statistical methods used, and may not consider other relevant factors”.

“No safe amount” describes whether any risk exists and says nothing about how large it is. The Cambridge and Global Burden of Disease analyses both suggest that for an adult in middle age the added risk at one drink a day or less is small. What level of drinking, if any, suits one person depends on their health and history, which is assessed by their doctor.

What the alcohol and mortality studies leave unsettled

The Victoria authors are candid about the weaknesses of their material. Self-reported alcohol consumption is underreported in most epidemiological studies, and consumption in many studies was assessed at only 1 point in time, although drinking habits change over a lifetime. Even the category of lifetime abstainer is shaky, since many who describe themselves that way turn out to have drunk in the past.

Pattern matters as well as volume. Seven drinks spread over a week and seven drinks on a Saturday night average out the same, and the authors note that a pattern of binge drinking removes the appearance of reduced health risks even when mean daily volume is low. Few studies recorded it.

Finally, death from any cause is a blunt outcome. The review’s results, in the authors’ words, only apply to the net effect of alcohol at different doses on deaths from all causes. Alcohol may nudge the risk of one disease down while pushing others up, and the balance will differ between a 30-year-old and a 70-year-old.

The best current reading is that light drinking does not lengthen life, that its harm is small at low doses, and that the old case for drinking as a health measure rested on a flawed comparison.