The guide

Why one drink a day looked healthy, and why it wasn't

By GyroscopeUpdated November 15, 202612 sources

For decades, study after study found that light drinkers died a little less often than non-drinkers. That’s the J-curve: risk dips at a drink or so a day, then climbs. The dip came from who got counted as a “non-drinker.” Many of them had quit because they were sick, or drank so little they barely counted. When researchers fixed the comparison, the dip went away.

The J-curve was one of the most repeated findings in health. It’s why doctors said a glass of wine with dinner was fine or even good, and why guidelines set a daily allowance instead of saying less is better. It felt right too. The people you know who drink a little tend to be relaxed, social and healthy. The people who never touch it sometimes have a reason, and that’s where the curve went wrong.

Why the curve dipped

Imagine you want to know whether jogging is safe, so you compare joggers with people who don’t jog. The non-joggers include people with bad knees, heart failure and cancer. Joggers will look amazing, and jogging might have nothing to do with it.

Alcohol studies did the same thing. They asked people how much they drank, then followed them for years. Anyone who said “none” went in the non-drinker group. That group held lifelong non-drinkers, but it also held people who had stopped because of an illness, a medicine, a liver problem or a drinking problem. Those people die sooner, and they pulled the non-drinkers’ numbers down.

In 1988 the British researcher A. G. Shaper and two colleagues saw it in their own data. They followed 7,735 middle-aged men for 7.5 years and found the classic U shape. But the U only showed up in men who already had heart disease or a related illness at the start. Men with those diagnoses had moved into the non-drinking and occasional groups. The authors wrote that the protective effect was “likely to be ill founded.” Most of the field kept using the curve for about another 30 years.

Fixing the comparison

In 2016 Tim Stockwell’s team in Canada pooled 87 studies covering 3,998,626 people and 367,103 deaths. Done the old way, light drinkers (1.3 to 25 g a day, up to almost two US drinks) had a 14% lower death rate than “abstainers.” Former drinkers had a 22% higher one, which is a clue to what was hiding in the abstainer group. After adjusting for it and for study quality, the light drinkers’ advantage shrank to 3% and was no longer statistically real.

In 2023 the same group repeated it with 107 studies, 4.8 million people and 425,564 deaths. They found that 86 of the 107 studies had former or occasional drinkers in the abstainer group. Compared with lifetime non-drinkers, occasional drinkers and drinkers of up to 24 g a day showed no lower risk of dying. Risk rose significantly from 45 g a day overall, and from 25 g a day for women.

What the genes say

Asking people how much they drink, then waiting, can’t get rid of every bias. Drinkers and non-drinkers differ in income, friends, diet and a hundred other ways. Genes offer a way around it. Some people are born with variants that make alcohol unpleasant, so they drink less for reasons that have nothing to do with their health. Comparing them with everyone else is like a natural experiment. Scientists call it Mendelian randomization.

• In 2014, a pooled analysis of 56 studies and 261,991 Europeans looked at a variant in the ADH1B gene. Carriers drank about 17% less, had lower blood pressure and smaller waists, and had 10% lower odds of coronary heart disease, at every level of drinking. The authors concluded that drinking less helps the heart “even for light to moderate drinkers.”
• In 2019, a study of 512,715 adults in China used two variants common in East Asia. When men reported their own drinking, the usual U shape appeared, with the lowest stroke and heart attack risk at about one to two drinks a day. When drinking was predicted from their genes, the U disappeared. Stroke risk rose steadily with alcohol, and the authors wrote that the protective effect against stroke was “largely non-causal.”
• In 2022, a genetic analysis of 371,463 people in the UK Biobank found light drinking nudged heart risk up a little and heavier drinking raised it fast. The light drinkers also had healthier habits overall, which explained part of their apparent benefit.
• In 2026, a study of 559,559 people and genetic data from 2.4 million found the same pattern for dementia: a U shape in what people reported, and a steady rise with no dip in the genetics.

Where the big reviews landed

The Global Burden of Disease study, which tracks causes of death in every country, looked at alcohol in 2018 using 592 studies. It found the amount that minimized harm was zero. Its 2022 update found the healthiest amount for people aged 15 to 39 was zero or close to it, while for people 40 and over the curve still dipped a little at small amounts. In January 2023 the World Health Organization said “there is no safe amount that does not affect health.”

Not everyone agrees on the last decimal. A December 2024 National Academies review found “moderate certainty” that moderate drinking went with a lower death rate than never drinking. We side with the bias-corrected and genetic studies, because they’re the ones that can tell a cause from a coincidence, and because every one of them points the same way. The US Dietary Guidelines for 2025 to 2030 dropped the old allowance of one drink a day for women and two for men, and now just say “Consume less alcohol for better overall health.”

What this means for you

The J-curve was never permission to drink. At best it said light drinkers weren’t worse off than a group that included sick people. Now that the comparison is fixed, alcohol looks like most other things we enjoy that carry a cost, with a small risk at low amounts that grows with every drink. Your risk, with and without shows the numbers, and the drink-a-day myth and the red wine myth take on the two versions people still repeat. lifespan.guide has what drinking does to life expectancy.