Is Alcohol Bad for You? The J-Curve, the French Paradox, and the Real Answer

Why this fight exploded again in 2026

No health question has flipped as violently as this one. For most of living memory, the official advice was that moderate drinking was fine — even healthy. A glass of red wine with dinner was not a vice to be managed; it was, according to decades of headlines, practically a health food. Then, in the space of three years, the world's health authorities tore that advice up. The WHO said in January 2023 that there is no safe level. Canada slashed its guidance the same month from up to 15 drinks a week for men to two. The US Surgeon General called for cancer warnings on bottles in January 2025. And in January 2026, the new US Dietary Guidelines dropped their numeric alcohol limits entirely and told Americans simply to "consume less alcohol for better overall health".

And yet the argument is not over — it is louder than ever, because the evidence genuinely points in two directions at once. In December 2024, the National Academies of Sciences, Engineering, and Medicine — the body Congress asked to review the evidence for the new dietary guidelines — concluded with moderate certainty that moderate drinking is associated with lower all-cause mortality and lower cardiovascular mortality than never drinking, while also raising the risk of breast cancer in women. Then, in 2026, a federally commissioned modelling study reached the opposite conclusion: no net protective effect at any level of drinking, and a 1-in-25 lifetime risk of an alcohol-caused death for men averaging two drinks a day. That study was then attacked in a House Oversight Committee report as "a study fraught with bias", and the health department said the published version had not been cleared by the agency that funded it. Two federally commissioned reviews, opposite conclusions, and a political fight over which one counts.

That is why this question is perfect for this series. It is not a settled question with a noisy fringe. It is a live, genuine scientific dispute in which both sides can point at real studies, real institutions, and real numbers — and in which the public has been told, in succession, that a glass a day is medicine, that no amount is safe, and that the truth depends on which committee you ask. The question people actually type into Google — "is alcohol bad for you?", "is moderate drinking healthy?", "is red wine good for your heart?" — gets millions of searches a year, and the answers are mostly useless: content-farm pages that repeat the WHO line without explaining it, wine blogs that repeat the J-curve without mentioning it was debunked, and health sites that hedge so hard they say nothing. None of them does the obvious thing, which is to lay the whole record out — the curve, the bias, the genes, the cancer numbers, the heart data, and the political fight — and let the evidence speak. That is what this article does.

Where "a glass a day is good for you" came from

The idea that moderate drinking protects the heart did not come from a single study. It came from a pattern that showed up again and again, in study after study, for decades. Researchers would follow tens of thousands of people for years, split them into non-drinkers, light drinkers, moderate drinkers and heavy drinkers, and plot their death rates. The graph almost always came out the same shape: a J. Non-drinkers died at a certain rate. Light drinkers — one drink a day, sometimes two — died at a lower rate. Then the curve turned upward, and heavy drinkers died at a much higher rate. The dip in the middle was the J-curve, and it was one of the most reproduced findings in nutritional epidemiology. It appeared in the Framingham Heart Study, in the Nurses' Health Study, in the Health Professionals Follow-up Study, and in meta-analyses pooling hundreds of thousands of people. By the 1990s it was conventional wisdom, and it shaped the official guidelines of nearly every country: a drink or two a day was not merely tolerated, it was presented as the healthy middle of the curve.

The idea got its most famous boost from television. In November 1991, the American programme 60 Minutes aired a segment on the "French paradox": the French ate diets rich in butter, cheese and fatty meat, yet had strikingly lower rates of coronary heart disease than Americans. The segment's proposed explanation was red wine. The following year, the French epidemiologist Serge Renaud published a paper in The Lancet arguing that moderate wine consumption explained the paradox, and the resveratrol hypothesis was born: red wine contains a plant compound, resveratrol, that protects the heart. Red wine sales in the United States jumped by more than 40% in the year after the broadcast. The J-curve had a story, a hero compound, and a glass in everyone's hand.

But there was always a problem with the J-curve, and it was hiding in plain sight: the comparison group. The J-curve compares drinkers with "non-drinkers" — and the people who do not drink are not a random sample of the population. They include people who never drank, certainly. But they also include people who used to drink and stopped — often because they became ill, because their drinking damaged their liver, because a doctor told them to stop, or because alcoholism had already taken its toll. They include people who do not drink because they are already sick, or because they take medication that forbids it, or because they are frail. Put the sickest people in the comparison group and the drinkers will look healthy even if the drink is doing nothing at all. This is the "sick quitter" problem, also called abstainer bias, and it took the field thirty years to take it seriously.

The French paradox, dismantled

Start with the story that sold the most wine. The French paradox has three ingredients, and all three are weaker than they looked in 1991. First, the resveratrol arithmetic does not work. A typical glass of red wine contains roughly one to two milligrams of resveratrol. The doses used in the laboratory and human studies that showed effects were hundreds of milligrams a day — typically 150 to 500 milligrams in human trials. To get a trial dose from wine you would need to drink roughly a hundred to a thousand glasses a day. The alcohol in that much wine would kill you long before the resveratrol helped you. Supplement makers solved the arithmetic by putting 250 to 1,000 milligrams in a capsule, but that is a drug dose, not a glass of wine, and the human trials of resveratrol supplements have produced mixed and inconsistent results — modest effects on blood pressure at high doses in some studies, nothing in healthy people in others. Resveratrol is not the reason red wine was ever linked to heart health, because there is not enough of it in wine to do anything.

Second, the paradox itself may be an artefact. The French eat more saturated fat than Americans, but they also eat differently in almost every other way: smaller portions, more vegetables, more fish, longer meals, less snacking, more walking. And the heart-disease statistics that underpinned the paradox were partly a recording problem: France historically under-certified coronary heart disease on death certificates relative to other countries, so some of the "paradox" was a paperwork gap, not a biological one. When researchers adjusted for these differences, the paradox shrank. Third, even if red wine drinkers in France did have healthier hearts, wine drinkers are not a random group: in study after study, moderate drinkers are richer, better educated, more physically active, less likely to smoke, and more likely to have health insurance than abstainers. The wine may have been a marker of a healthy life, not the cause of it. The French paradox is a good story about a good meal. It is not evidence that alcohol protects the heart.

The sick-quitter problem: the study that broke the J-curve

The decisive blow to the J-curve came in March 2023, in a systematic review and meta-analysis published in JAMA Network Open by Junco Zhao, Tim Stockwell, Timothy Naimi and colleagues. It was the largest analysis of its kind ever conducted: 107 cohort studies, more than 4.8 million participants, and 425,564 deaths. And it did something no earlier meta-analysis had done systematically: it separated the studies by quality, asking which ones had guarded against the biases that make light drinking look protective, and which ones had not.

The result is the punchline of the modern literature. When all 107 studies were pooled naively, the familiar protective dip appeared: low-volume drinkers — up to about 1.3 to 24 grams of alcohol a day, under two US drinks — had a 15% lower risk of death than non-drinkers, a relative risk of 0.85. But when the analysis was restricted to the studies that had guarded against the biases — using lifetime abstainers as the comparison group instead of a mixed group of abstainers and former drinkers, adjusting for cohort age and sex, and properly controlling for smoking — the benefit vanished. The relative risk rose to 0.93, and it was no longer statistically significant. Light-to-moderate drinkers did not have a meaningful survival advantage over lifetime non-drinkers. Meanwhile the harm at higher intake stayed clear: the risk of death from all causes was significantly raised for women drinking 25 grams a day or more, and for men drinking 45 grams or more. The authors' conclusion was blunt: "low-volume alcohol drinking was not associated with protection against death from all causes".

The reason is in the numbers. Of the 107 studies, 86 — four out of five — had included former drinkers, or occasional drinkers, in the "abstainer" reference group. Only 21 studies were free of both abstainer biases. Former drinkers have significantly elevated mortality compared with lifetime abstainers, because people often quit drinking when their health fails. Put them in the comparison group and the group looks sicker than it is; compare light drinkers with that group and the light drinkers look healthier than they are. The J-curve, in this reading, was largely a measurement artefact: the dip in the middle was not the alcohol protecting anyone, it was the comparison group being full of people who had already been harmed by it. The 2023 analysis also found that the apparent J-shape was strong and significant in lower-quality studies and absent in the small number of high-quality ones — exactly the pattern you would expect if the effect were a bias rather than a benefit.

What the genes say: the experiment nature already ran

There is a second, independent line of evidence against the J-curve, and it does not depend on how studies classified their abstainers at all. It comes from genetics. In East Asia, two common genetic variants — in the genes ALDH2 and ADH1B — strongly affect how much alcohol people drink: people carrying them flush, feel unwell, and drink far less. Because genes are allocated randomly at conception, this is nature's version of a randomised trial: compare people with the variants against people without them, and you can see what alcohol itself does, free of the lifestyle differences that confound ordinary studies. The method is called Mendelian randomisation.

The largest such study, published in The Lancet in 2019 by Iona Millwood and colleagues, followed 512,715 adults in the China Kadoorie Biobank for about ten years, genotyping more than 160,000 of them. In the conventional analysis — the same kind of analysis that produced the J-curve — men who reported drinking about 100 grams of alcohol a week, roughly one to two drinks a day, had lower risks of ischaemic stroke, intracerebral haemorrhage and heart attack than non-drinkers or heavier drinkers. The J-shape appeared, exactly as expected. But in the genetic analysis, the U-shape vanished entirely. Genotype-predicted alcohol intake had a continuously positive, log-linear association with stroke risk: the more alcohol the genes predicted, the higher the risk, with no dip anywhere. Every extra four drinks a day raised the risk of stroke by about 35%, and even one to two drinks a day raised it by 10 to 15%. For heart attack, the genetic analysis found no significant association either way — no protection, no harm. The authors' interpretation was direct: "genetic epidemiology shows that the apparently protective effects of moderate alcohol intake against stroke are largely non-causal".

A second Mendelian randomisation study, by Biddinger and colleagues in 2022, using data from more than 370,000 people in the UK Biobank, reached the same conclusion for cardiovascular disease in general: light-to-moderate drinking was associated with higher cardiovascular risk, not lower, and there was no protective sweet spot. The genetic evidence is not perfect — it mostly covers East Asian and European populations, and it cannot capture every kind of drinking pattern — but it is independent of the sick-quitter problem, independent of self-reported drinking, and independent of the lifestyle differences between drinkers and abstainers. And it points the same way as the bias-corrected meta-analysis: the benefit was an artefact.

Cancer: the part nobody can argue away

If the heart evidence is contested, the cancer evidence is not. Alcohol has been classified as a Group 1 carcinogen by the International Agency for Research on Cancer for decades — the highest risk category, shared with tobacco, asbestos and radiation — and it is causally linked to at least seven types of cancer: breast, colorectum, oesophagus, liver, mouth, throat and voice box. The mechanism is understood, not merely statistical. When the body breaks down alcohol, it produces acetaldehyde, a chemical that damages DNA and prevents cells from repairing it. Alcohol also acts as a solvent, helping other carcinogens penetrate tissues, and it raises oestrogen levels, which helps explain the breast cancer link. This is not a correlation looking for a cause. It is a cause with a mechanism.

The scale of it was laid out in the US Surgeon General's advisory of 3 January 2025, issued by Dr Vivek Murthy. Alcohol, the advisory said, is the third leading preventable cause of cancer in the United States, after tobacco and obesity. It is responsible for about 100,000 cancer cases and 20,000 cancer deaths every year in the US — more than the roughly 13,500 alcohol-related traffic crash deaths a year. For breast cancer specifically, 16.4% of all cases are attributable to alcohol. And the risk starts low: for breast, mouth and throat cancers, the advisory noted, risk "may start to increase around one or fewer drinks per day". The type of drink does not matter — beer, wine and spirits carry the same risk, because the carcinogen is the alcohol itself, not the beverage. Yet only 45% of Americans knew alcohol could cause cancer at all.

The WHO's January 2023 statement made the same point for the world: at least half of all alcohol-attributable cancers in the WHO European Region occur among people whose drinking does not meet the definition of heavy drinking. The statement's central claim was that the available evidence cannot identify any threshold at which the carcinogenic effects of alcohol "switch on" — the risk does not start at some number of drinks and stay at zero below it. As Dr Carina Ferreira-Borges of the WHO Regional Office for Europe put it: "We cannot talk about a so-called safe level of alcohol use. It doesn't matter how much you drink — the risk to the drinker's health starts from the first drop of any alcoholic beverage. The only thing that we can say for sure is that the more you drink, the more harmful it is — or, in other words, the less you drink, the safer it is."

None of this means one glass of wine gives you cancer. The absolute risk from light drinking is small — a small increase on a small baseline is still small. A woman drinking one glass a day raises her breast cancer risk by roughly 5 to 10% relative to a non-drinker, which on a lifetime baseline risk of about 12% means an extra half to one percentage point of absolute risk. That is a real increase, and it is the kind of increase that matters enormously across a population of millions, but it is not the kind of increase that should terrify an individual who drinks occasionally. The cancer argument is strongest at the population level and weakest as a scare tactic — and it is precisely because the risk is real but small that the debate over light drinking is so bitter. Both sides are, in a sense, right about the size of the risk. They disagree about whether a small, avoidable increase in cancer risk is worth a pleasure that has no health benefit.

Cancer type	What the evidence says
Breast (women)	Risk rises from around one drink a day or less; 16.4% of US breast cancer cases are attributable to alcohol (US Surgeon General, 2025)
Colorectum	Risk rises with intake; one of the most common alcohol-related cancers worldwide
Oesophagus	Strong dose-response relationship; risk amplified sharply when combined with smoking
Liver	Risk rises with intake, partly through cirrhosis; acetaldehyde directly damages liver cells
Mouth, throat, voice box	Risk may start rising at one or fewer drinks a day; alcohol acts as a solvent for other carcinogens
Stomach and pancreas	Evidence growing; the WHO and IARC list at least seven cancers as causally linked, with more under study

The heart: the last refuge of the defenders

The heart is where the defenders of moderate drinking make their last stand, and it is the one place where their case is not empty. The December 2024 report from the National Academies of Sciences, Engineering, and Medicine — requested by Congress to inform the US Dietary Guidelines — reviewed the evidence published between 2010 and 2024 and concluded, with moderate certainty, that moderate drinking is associated with lower all-cause mortality and lower cardiovascular mortality than never drinking, and with a lower risk of non-fatal heart attack and non-fatal stroke. It also concluded, with moderate certainty, that moderate drinking is associated with a higher risk of female breast cancer. The committee was careful: it excluded studies in which the comparison group mixed never-drinkers with former drinkers, and none of its conclusions reached high certainty, which the committee said was unlikely without a randomised controlled trial. But the report is real, it is recent, it was written by nine experts, and it contradicts the WHO's position. The defenders are not making it up.

How can two careful reviews of the same literature reach opposite conclusions? Partly because they asked different questions. The Zhao meta-analysis asked about all-cause mortality across all levels of drinking, and pooled 107 studies going back decades. The National Academies asked specifically about moderate drinking, looked only at studies from 2019 to 2023 for the mortality question, and used stricter inclusion criteria — only eight studies were eligible for the all-cause mortality conclusion. Partly because the evidence itself is genuinely mixed: the heart-protection signal survives in some analyses and vanishes in others, depending on which studies are included, how the comparison group is defined, and whether the analysis adjusts for the fact that moderate drinkers tend to be wealthier, healthier and more active. And partly because the effect, if it exists at all, is small and sits inside a much larger pattern of harm. Even the National Academies' own report found that moderate drinking raises breast cancer risk — and breast cancer is one of the commonest cancers in the world.

There is one more piece of heart evidence that the defenders rarely mention, because it cuts against them. Alcohol is a well-established cause of atrial fibrillation — the irregular heartbeat sometimes called "holiday heart" after the cases that show up in emergency departments after weekends and holidays of heavy drinking. In 2020, a randomised trial published in the New England Journal of Medicine took regular drinkers with atrial fibrillation and randomly assigned half of them to abstain from alcohol. The abstainers had significantly fewer recurrences of the arrhythmia and went longer between episodes. It is one of the very few randomised trials in this whole debate — most of the evidence is observational — and it found that removing alcohol helped the heart. Alcohol also raises blood pressure in a dose-dependent way, and the genetic study in China found that the rise in blood pressure was the main route by which alcohol raised stroke risk. A drug that raises blood pressure, causes arrhythmia, and is a Group 1 carcinogen is a strange candidate for a heart medicine, even at one glass a day.

The 1-in-25 study and the political fight over it

The newest and most explosive piece of evidence is also the most contested. In 2026, the Journal of Studies on Alcohol and Drugs published the Alcohol Intake and Health Study, led by Katherine Keyes of Columbia University and Kevin Shield of the University of Toronto. It was commissioned by a federal body — the Interagency Coordinating Committee on the Prevention of Underage Drinking — to inform the US Dietary Guidelines. The team reviewed more than 7,200 scientific papers, picked the strongest risk estimates for each alcohol-related condition, and applied them to US health, census and death data. The comparison group was people who never drank, and former drinkers were excluded — the sick-quitter problem, handled by design. The result: no net protective effect of alcohol at any level of drinking, and a lifetime risk of dying from an alcohol-caused condition of about 1 in 25 — 4% — for men averaging 14 drinks a week, the level that had been the old US daily ceiling for men. For women the risk was lower but still rose from the first drink.

The study was immediately attacked. In January 2026, the House Oversight Committee released a staff report titled "A Study Fraught with Bias", arguing that the study duplicated work Congress had assigned to the National Academies, was run in a way inconsistent with federal law, and drew on a review panel whose members and process showed bias. The Department of Health and Human Services said the published version had not been cleared by SAMHSA, the agency that originally funded the work. The study's defenders replied that it was a modelling exercise, not a trial, and that its assumptions were stated openly: it assumes the observational risk estimates it used are causal, and sums the deaths across conditions. Both criticisms are fair, and both miss the point. A model is only as good as its inputs, and the inputs are the same observational studies the whole field argues about. But the model did something no single study could: it added up every known alcohol-caused condition, used never-drinkers as the comparison, and asked what the total risk looks like. The answer — no benefit anywhere, and a 1-in-25 lifetime risk at two drinks a day for men — is the most concrete number the "no safe level" camp has ever had. It is also a number that depends on the assumption that the risks are causal, which is exactly what the other camp disputes. The fight over the study is really a fight over that assumption, and it will not be settled by another committee report.

Meanwhile the official guidance has quietly converged on the same message, even where the numbers differ. In January 2026, the 2025–2030 US Dietary Guidelines dropped the numeric limits that had stood for decades — two drinks a day for men, one for women — and replaced them with a single line: "consume less alcohol for better overall health". The National Academies' report, which had found a mortality benefit from moderate drinking, was not the basis for that change; the modelling study was the one that informed the guidelines' direction. The message from the authorities is now consistent across countries, even if the numbers are not: there is no amount that is good for you, and less is always safer than more.

What the world's guidelines actually say

Authority	Year	What it says
World Health Organization	2023	No safe amount can be established; risk starts from the first drop; less is safer
Canada (CCSA)	2023	Continuum of risk: 0 drinks = no risk; 1–2 a week = low risk; 3–6 = moderate risk; 7+ = increasingly high risk. Replaced the old limits of 10 a week for women and 15 for men
United Kingdom (Chief Medical Officers)	2016	No more than 14 units a week (112 g of alcohol), spread over three or more days; no level is completely safe
United States (Dietary Guidelines)	2026	Numeric limits dropped; "consume less alcohol for better overall health"
United States (old limits)	2020	Up to 2 drinks a day for men, 1 for women — the limits the 2026 guidelines abandoned
Australia (NHMRC)	2020	No more than 10 standard drinks a week and no more than 4 on any one day, to reduce lifetime harm
US Surgeon General	2025	Alcohol is the third leading preventable cause of cancer; called for cancer warnings on labels

The spread in those numbers is itself the story. Canada says two drinks a week is the low-risk line; the old American guidance said two drinks a day for men. That is a seven-fold difference, and it is not because Canadians and Americans have different bodies. It is because the two countries answered a different question: Canada asked "at what level does the risk become clearly measurable?", and the old American guidance asked "at what level does the risk become clearly dangerous?". The WHO's answer — no safe level — is the most honest statement of the science, and the least useful as a personal rule, because it gives a person no number to aim at. The Canadian continuum is the most useful: it does not pretend there is a safe line, it shows the risk rising, and it lets a person choose how much risk they are willing to carry. The old American numbers were the most permissive, and they are the ones that have now been abandoned.

At a glance: what alcohol does, and what it does not

Claim	What the evidence says	Verdict
A glass of wine a day protects your heart	The J-curve appeared in decades of observational studies, but the 2023 Zhao meta-analysis of 107 studies and 4.8 million people found the benefit vanished once former drinkers were removed from the comparison group. Genetic studies in China and the UK found no protective sweet spot either	Mostly false — the benefit was largely a measurement artefact
Red wine is healthy because of resveratrol	A glass of red wine holds roughly 1–2 mg of resveratrol; the trials that showed effects used 150–500 mg a day. You would need roughly a hundred glasses a day to reach a trial dose. Human resveratrol trials have been mixed and inconsistent	False — the arithmetic does not work
Alcohol causes cancer	Group 1 carcinogen, causally linked to at least seven cancers. The US Surgeon General (2025): about 100,000 cases and 20,000 deaths a year in the US. Risk for breast, mouth and throat cancers may start rising at one or fewer drinks a day	True
Moderate drinking lowers your risk of dying	Zhao 2023: no significant protection once bias is corrected. The National Academies 2024: lower all-cause mortality with moderate certainty, using stricter criteria and only eight eligible studies. The two reviews disagree — and the disagreement is the story	Disputed — the best reviews disagree
Alcohol raises stroke risk even at low intake	Millwood 2019 (512,715 people, genetics): stroke risk rises log-linearly from the first drink; 1–2 drinks a day raised it 10–15%, four a day raised it about 35%. No protective effect at any level	True
Alcohol is the leading killer of young adults	GBD 2018: alcohol was the leading risk factor for death and disability worldwide in people aged 15–49, causing 12.2% of male deaths and 3.8% of female deaths in that age group, and 2.8 million deaths in total in 2016	True
Two drinks a day carries a 1-in-25 lifetime risk of an alcohol-caused death	Keyes/Shield modelling study, 2026: about 4% lifetime risk for men at 14 drinks a week, no net protective effect at any level. It is a model, not a trial, and it is politically contested — a House Oversight report called it biased	Plausible but contested — a model, not a measurement
Drinking helps you sleep	Alcohol is a sedative, so it can make you fall asleep faster — but it fragments sleep, suppresses REM sleep, and worsens snoring and sleep apnoea, so the sleep you get is worse	False as a sleep aid
Beer is safer than spirits	The carcinogen is the alcohol, not the beverage. A standard drink of beer, wine or spirits contains the same amount of ethanol, and the cancer risk is the same per gram of alcohol	False — only the dose matters
Quitting drinking reverses the damage	Much of it, yes. Blood pressure falls within weeks, liver fat clears within weeks to months, and cancer risk falls gradually over years after stopping — though some risk, and some liver scarring, is permanent	Mostly true — and the best reason the debate matters

Claim vs evidence: the arguments, scored

Argument	Who makes it	Strength of the evidence
"No safe level of alcohol consumption can be established"	WHO, Anderson et al., Lancet Public Health 2023	Strong — the cancer evidence has no identifiable threshold, and the statement is endorsed by the global alcohol research community
"The J-curve benefit disappears when bias is corrected"	Zhao et al., JAMA Network Open 2023	Strong — 107 studies, 4.8 million people; the benefit vanished in the bias-corrected analyses
"Genetics show no protective effect against stroke"	Millwood et al., The Lancet 2019	Strong — 512,715 people, 10 years; the genetic analysis is independent of the sick-quitter problem
"Alcohol causes about 100,000 cancers a year in the US"	US Surgeon General, January 2025	Strong — third leading preventable cause of cancer, after tobacco and obesity
"Moderate drinking is associated with lower mortality"	National Academies, December 2024	Moderate — real finding, moderate certainty, but based on eight eligible studies and contradicted by the larger Zhao analysis
"1 in 25 lifetime risk at two drinks a day for men"	Keyes/Shield, Journal of Studies on Alcohol and Drugs 2026	Moderate — a careful model built on 7,200 papers, but a model, not a trial, and politically contested
"Resveratrol makes red wine healthy"	Wine industry; the French paradox	Weak — a glass holds 1–2 mg against trial doses of 150–500 mg; human trials are mixed
"My grandad drank every day and lived to 94"	Every family argument ever	Weak — survivorship bias: the grandads who drank every day and died at 60 are not at the table to argue
"Alcohol was the leading killer of people aged 15–49"	GBD 2016 Alcohol Collaborators, The Lancet 2018	Strong — 2.8 million deaths in 2016; leading risk factor in that age group worldwide

What people actually say: how divided the argument really is

The alcohol debate is not a polite disagreement between public health agencies. It is one of the most furious arguments on the internet, and the popular-opinions section is where that fury lives. What follows is paraphrased from the recurring positions in real threads — on Reddit, in health forums, in YouTube comments and on Quora — not invented quotes. No usernames, no vote counts, no fabricated comments: just the arguments people actually make, in the shape they actually make them.

The grandad camp. The loudest camp in every thread is the one with a personal counterexample. Their argument runs: my grandfather drank whisky every night and lived to 94, my uncle smoked and drank and outlived the teetotallers, the French drink wine with every meal and live longer than Americans. Their weakness is survivorship bias, and it is total: the grandfathers who drank every night and died at 58 are not alive to post. The grandad who lived to 94 is the exception that proves the rule, not the rule itself — and he is usually also the grandad who walked everywhere, ate vegetables from a garden, and never smoked. But their strength is that they are pointing at something real: population statistics do not feel like an answer to a personal question, and a rule that says "your grandfather's nightly whisky was poisoning him" sounds, to them, like an insult to a man they loved. The science is on the other side. The emotion is on theirs.

The wine-is-medicine camp. The second camp is the one that has read about the French paradox and will not let it go. Their argument runs: red wine has resveratrol, resveratrol is an antioxidant, the French drink wine and have healthy hearts, therefore wine is medicine, and the WHO is a bunch of joyless bureaucrats. They post pictures of vineyards and quotes from doctors who drink a glass a day. Their weakness is the arithmetic: a glass of wine holds one or two milligrams of resveratrol against trial doses of hundreds of milligrams, and the French paradox shrinks every time someone adjusts for the rest of the French lifestyle. Their strength is that they are defending something real too — the pleasure of a meal, the social ritual, the glass shared with friends — and they are right that a public health message that ignores pleasure will be ignored. The wine is not medicine. The meal around it might be.

The sober-curious camp. The third camp is the fastest-growing one, and it is the one that has changed the argument most in the last few years. These are the people who did Dry January, noticed they slept better, lost weight, and had more energy, and never went back — or went back to drinking much less. Their argument runs: I stopped and I feel better, my skin is better, my anxiety is lower, my mornings are mine again, and I do not miss it. They post before-and-after stories, they trade alcohol-free beer recommendations, and they are genuinely evangelical about it. Their weakness is that personal experience is not evidence — feeling better after quitting proves that quitting helped you, not that a glass a day was harming you — and some of them slide from "I feel better" to "everyone who drinks is poisoning themselves", which is the same moralising they object to in the other direction. Their strength is that they are the only camp with a test anyone can run: stop for a month and see. It is the one experiment in this debate that does not require a meta-analysis.

The "it's all prohibitionist junk science" camp. The fourth camp is the one that treats the whole "no safe level" position as a political project rather than a scientific one. Their argument runs: the WHO is captured by temperance ideology, the Zhao study was written by researchers who want prohibition, the 1-in-25 study was so biased that Congress investigated it, and the National Academies — the careful, Congress-commissioned review — found that moderate drinking is associated with lower mortality. They post the National Academies report, they post the House Oversight report, and they accuse the other side of cherry-picking. Their weakness is selectivity of their own: they quote the National Academies on mortality and skip the same report's finding that moderate drinking raises breast cancer risk; they quote the House Oversight report and skip the fact that it was a partisan staff report, not a scientific review. Their strength is that they are right about one thing: the 1-in-25 study is a model with contestable assumptions, the Zhao analysis is disputed by serious researchers, and the National Academies did reach the opposite conclusion on mortality. The disagreement is real, and pretending it is not only makes the sceptics louder.

The pragmatists. And then there is the largest camp of all, which never posts: the people who drink a glass of wine with dinner on Friday, have a beer at a barbecue, and have never thought about the question until a relative sent them a headline saying no amount is safe. They are the audience this article is written for. They do not need a lecture, and they do not need a cartoon. They need the record, stated plainly, with the gaps left in — and a practical answer to the only question that matters to them: how much does this actually matter, for me, at the amount I drink?

The verdict

So: is alcohol bad for you? On the evidence available in 2026, the answer is yes — with a qualification that matters more than the yes. The WHO's position is the best summary of the science: no safe level can be established, the risk starts from the first drink, and less is always safer than more. The cancer evidence is the firmest part of the case: alcohol is a Group 1 carcinogen, it causes at least seven cancers, and the risk rises from low levels. The J-curve — the idea that a glass a day protects your heart — has largely collapsed: the largest bias-corrected meta-analysis found no significant benefit, and the genetic studies found no protective sweet spot at all. The Global Burden of Disease study found alcohol the leading risk factor for death and disability in people aged 15 to 49. That is the case against drinking, and it is strong.

But the qualification is the size of the risk at low levels, and the honest account has to state it. A person drinking one or two drinks a week faces a small absolute increase in risk — a fraction of a percentage point on most outcomes. The National Academies, reviewing the same literature with stricter criteria, still found moderate drinking associated with lower all-cause mortality, and that disagreement is genuine: the best reviews disagree, and anyone who tells you the matter is settled is selling you a conclusion, not the evidence. The 1-in-25 figure for men drinking two a day is a model, not a measurement, and it is contested. What is not contested is the direction: every credible analysis agrees that risk rises with intake, and that the only amount with no risk is none.

The practical advice follows from that. There is no health reason to start drinking, and none to keep drinking for your heart — the resveratrol is not there in any meaningful amount, the J-curve was mostly an artefact, and the same heart benefits, if they exist, are available from exercise, which has no carcinogen attached. If you drink, drink less: the Canadian continuum is the most useful guide ever published — zero is no risk, one to two a week is low risk, three to six is moderate risk, and seven or more is increasingly high risk. Keep alcohol-free days in the week, because they break the habit loop and let your sleep and blood pressure recover. Do not drink to help you sleep — it sedates you and then wrecks the sleep it started. Never drink during pregnancy or when trying to conceive: there is no known safe amount there, and that is one place where even the two camps agree. And if you have symptoms — a liver that aches, blood pressure that will not come down, sleep that never refreshes, a drinking habit that has quietly grown — talk to a clinician rather than to a comment section. The same pattern runs through the other food and drink fights in this series: the red meat debate and the seed oils fight both turned on the same mistake — taking a real signal and applying it to the wrong dose. Here the signal is real, the dose matters, and the honest answer is the unglamorous one: alcohol is bad for you, a little is only a little bad, and the less you drink, the safer you are.

Frequently asked questions

Is alcohol bad for you even in small amounts?

Yes, in the sense that no level has been shown to be safe. The WHO's January 2023 statement concluded that no safe amount of alcohol consumption for cancers and health can be established, because the evidence cannot identify a threshold below which the carcinogenic effects switch off. But the absolute risk from one or two drinks a week is small — a small increase on a small baseline. The honest framing is a continuum of risk, not a cliff: less is always safer than more.

Is moderate drinking actually healthy?

Probably not, and the evidence for it has weakened sharply. The old J-curve — light drinkers dying less than non-drinkers — appeared in decades of studies, but the largest bias-corrected meta-analysis (Zhao 2023, 107 studies, 4.8 million people) found the benefit disappeared once former drinkers were removed from the comparison group. Genetic studies found no protective sweet spot either. The National Academies' 2024 review is the main counterweight: it found moderate drinking associated with lower all-cause mortality, with moderate certainty. The two reviews disagree, and the disagreement is genuine.

Is red wine good for your heart?

No — not in any way that survives the arithmetic. Red wine contains resveratrol, but a glass holds roughly one to two milligrams, while the trials that showed effects used 150 to 500 milligrams a day. You would need roughly a hundred glasses a day to reach a trial dose. The "French paradox" also shrinks once you account for the rest of the French lifestyle and for differences in how heart disease was recorded. Wine drinkers in studies also tend to be wealthier, more active and less likely to smoke, which explains much of the apparent benefit.

How much alcohol is safe to drink per week?

No authority can now name a safe amount, but the clearest practical guide is Canada's 2023 continuum: zero drinks is no risk, one to two a week is low risk, three to six is moderate risk, and seven or more is increasingly high risk. The UK advises no more than 14 units a week (112 grams of alcohol), spread over three or more days. The 2025–2030 US Dietary Guidelines dropped numeric limits entirely and say only to "consume less alcohol for better overall health". The old US limits — two drinks a day for men, one for women — have been abandoned.

Does alcohol really cause cancer?

Yes. Alcohol is a Group 1 carcinogen, causally linked to at least seven cancers: breast, colorectum, oesophagus, liver, mouth, throat and voice box. The US Surgeon General's January 2025 advisory put the toll at about 100,000 cancer cases and 20,000 cancer deaths a year in the United States, making alcohol the third leading preventable cause of cancer after tobacco and obesity. The mechanism is understood: alcohol breaks down into acetaldehyde, which damages DNA. The risk holds for beer, wine and spirits alike, because the carcinogen is the alcohol itself.

Is it too late to benefit from cutting down or quitting?

No. Blood pressure falls within weeks of cutting down, liver fat clears within weeks to months, and sleep improves within days. Cancer risk falls gradually over years after stopping, though some risk and some liver scarring is permanent. A 2020 randomised trial found that regular drinkers with atrial fibrillation who abstained had significantly fewer recurrences of the irregular heartbeat. The best time to cut down was years ago; the second-best time is now.

Sources

• Anderson BO, Berdzuli N, Ilbawi A, Kestel D, Kluge HP, Krech R, et al. — "Health and cancer risks associated with low levels of alcohol consumption", The Lancet Public Health 2023;8(1):e6–e7 (WHO statement: no safe amount of alcohol consumption for cancers and health can be established) (WHO, 4 January 2023)
• Zhao J, Stockwell T, Naimi T, Churchill S, Clay J, Sherk A — "Association Between Daily Alcohol Intake and Risk of All-Cause Mortality: A Systematic Review and Meta-analyses", JAMA Network Open 2023;6(3):e236185 (107 cohort studies, 4,838,825 participants, 425,564 deaths; low-volume benefit not significant after bias correction; risk significantly raised at 25 g/day for women and 45 g/day for men) (Drug Policy Facts record of the Zhao et al. findings)
• GBD 2016 Alcohol Collaborators — "Alcohol use and burden for 195 countries and territories, 1990–2016", The Lancet 2018;392:1015–1035 (2.8 million deaths in 2016; alcohol the leading risk factor for death and disability in people aged 15–49) (The Lancet)
• US Surgeon General — Advisory on Alcohol and Cancer Risk, 3 January 2025 (alcohol the third leading preventable cause of cancer in the US; about 100,000 cases and 20,000 deaths a year; 16.4% of breast cancer cases attributable; only 45% of Americans aware of the link) (University of Utah / Huntsman Cancer Institute summary)
• Millwood IY, Walters RG, Mei XW, et al. — "Conventional and genetic evidence on alcohol and vascular disease aetiology: a prospective study of 500 000 men and women in China", The Lancet 2019;393:1831–1842 (512,715 adults, 10 years; genetic analysis found no protective effect of moderate drinking against stroke — risk rose log-linearly) (PubMed 30955975)
• National Academies of Sciences, Engineering, and Medicine — "Review of Evidence on Alcohol and Health", December 2024 (moderate drinking associated with lower all-cause mortality and lower cardiovascular mortality, moderate certainty; higher risk of female breast cancer, moderate certainty) (National Academies)
• George S, Naimi TS, Keyes K, et al.; Shield K — "Alcohol Intake and Health Study", Journal of Studies on Alcohol and Drugs 2026;87(4):621–638 (modelling study of 7,200+ papers; no net protective effect at any level; about 1 in 25 lifetime risk of an alcohol-caused death for men at 14 drinks a week) (summary of the study and the dispute around it)
• Canadian Centre on Substance Use and Addiction — "Canada's Guidance on Alcohol and Health", January 2023 (continuum of risk: 0 drinks no risk; 1–2 a week low risk; 3–6 moderate risk; 7+ increasingly high risk; replaced the 2011 limits of 10 a week for women and 15 for men) (CCSA final report, January 2023)
• Renaud S, de Lorgeril M — "Wine, alcohol, platelets, and the French paradox for coronary heart disease", The Lancet 1992;339:1523–1526 (the French paradox paper; popularised by the 1991 60 Minutes segment)
• Biddinger KJ, Emdin CA, Haas ME, et al. — "Association of Habitual Alcohol Intake With Risk of Cardiovascular Disease", JAMA Network Open 2022;5(3):e223849 (UK Biobank, 371,463 people; Mendelian randomisation found light-to-moderate drinking associated with higher cardiovascular risk, no protective effect)
• Voskoboinik A, Marcus FI, et al. — "Alcohol Abstinence in Drinkers with Atrial Fibrillation", New England Journal of Medicine 2020;382:20–28 (randomised trial: abstinence reduced recurrence of atrial fibrillation in regular drinkers)
• UK Chief Medical Officers — "Alcohol Guidelines Review", 2016 (no more than 14 units a week, spread over three or more days; no level of drinking is completely safe)
• US Dietary Guidelines for Americans, 2025–2030 (January 2026) — numeric alcohol limits dropped; "consume less alcohol for better overall health"
• International Agency for Research on Cancer — alcohol classified as a Group 1 carcinogen; at least seven cancers causally linked (breast, colorectum, oesophagus, liver, mouth, pharynx, larynx)

More from the Unanswered Questions series: If Poison Expires, Does It Become Safe? · Is MSG Bad for You? · Is Organic Food Actually Healthier? · Are Seed Oils Toxic? · Is Fluoride in Tap Water Safe? · Does Red Meat Really Cause Cancer? · Does Sugar Really Make Kids Hyperactive? · Is Aspartame Safe? · Is Vaping Really Safer Than Smoking? · Are Eggs Bad for Your Cholesterol? · Are GMOs Safe to Eat? · Is Bottled Water Safer Than Tap Water? · Is Breakfast Really the Most Important Meal of the Day? · Do Vaccines Cause Autism? · Is Intermittent Fasting Actually Dangerous? · Is Microwaving Food Bad for You? · Is 5G Dangerous? · Is Soy Bad for Men?

Disclaimer: this article is general information, not medical advice. It summarises published research and the positions of public health agencies as of November 2026. If you have a health concern, speak to a qualified clinician.