Does moderate drinking, especially red wine, protect the heart and help you live longer?
Claim attributed to A decades-old J-curve / French paradox narrative in popular media and older epidemiology, amplified by alcohol-industry-funded research. , Not one named person but an industry-adjacent narrative; the flagship NIH MACH15 trial meant to prove it was largely alcohol-company funded and cancelled in 2018.
Once you correct for who ends up in the non-drinker group, the survival advantage of moderate drinkers evaporates, and genetic evidence shows alcohol's effect on the heart is linear and harmful. The protective J-curve is largely a study-design artifact, and the special red-wine story rests on resveratrol doses you cannot reach by drinking.
Moderate drinkers do tend to outlive abstainers in old data, but mostly because dying ex-drinkers were filed as non-drinkers; correct that and the benefit vanishes, while the cancer risk does not.
What it’s supposed to target
- HDL cholesterol (claimed)
- Resveratrol / polyphenols (claimed)
- Cancer + cardiovascular risk (real)
- Study design / confounding
The longevity case for moderate drinking rests on a proposed mechanism: a little alcohol nudges up HDL (good) cholesterol and reduces clotting, and red wine adds polyphenols like resveratrol, so the story goes that a daily glass protects the heart. This is the famous “J-curve,” in which light drinkers appeared to outlive both heavy drinkers and teetotalers, dressed up further by the “French paradox.”
The mechanism is mostly a mirage. The J-curve largely reflects study-design flaws, above all the sick-quitter effect (people who stopped drinking because they were already ill get counted as abstainers) and the fact that moderate drinkers tend to be healthier and wealthier to begin with. When analyses correct for this, the apparent benefit shrinks or vanishes, and genetic (Mendelian-randomization) studies show alcohol's harm rising roughly linearly, with cancer risk climbing even at low intake. The resveratrol in wine is far too dilute to matter, and much of the favorable research was industry-funded. A comforting theory the cleaner evidence does not support.
Mechanism is theory, not proof. A plausible pathway explains why something might work, not whether it does. The verdict rests on the evidence below, not the elegance of the theory.
What would have to be true
The mortality dip at low intake would have to be real rather than produced by misclassifying sick ex-drinkers as non-drinkers: this link FAILS, 86 of 107 cohorts carried that bias and the benefit disappears in the 21 clean ones.
Moderate drinkers would have to be comparable to abstainers on wealth, activity and care: this link FAILS, confounding is well documented.
Genetic instruments would have to show a protective threshold: this link FAILS, MR shows linear harm.
Red wine's resveratrol dose would have to be biologically meaningful: this link FAILS, achievable doses are far too low.
What the evidence actually shows
Bias-corrected epidemiology erases the benefit
The largest bias-corrected meta-analysis (Zhao/Stockwell, JAMA Network Open 2023; 107 cohorts, ~4.8 million people, ~425,000 deaths) found no significant all-cause mortality benefit for low-volume drinkers (1.3-24 g/day): RR 0.93 (95% CI 0.85-1.01). The classic protective dip came from the sick-quitter bias: 86 of 107 studies wrongly placed former or occasional drinkers into the non-drinker reference group, and only 21 were free of both biases. Harm at higher intake was clear (women >=65 g/day RR 1.61; men >=65 g/day RR 1.34).
Genetics and global burden point to harm
A Mendelian randomization study (Lankester 2021, PLOS One; UK Biobank, ~337,000 people; ADH1B rs1229984 instrument) found alcohol was not protective for any outcome, with each extra drink/day raising systolic blood pressure (+2.65 mmHg), hemorrhagic stroke (OR 2.25) and atrial fibrillation (OR 1.26). The Global Burden of Disease 2016 analysis (Lancet) concluded the consumption level that minimizes health loss is zero, and the WHO (2023) states there is no safe level: ethanol is an IARC Group 1 carcinogen causing at least seven cancers, with risk from the first drink.
Studies, graded, and who paid
Bias-corrected meta-analysis of 107 cohorts finds no significant benefit at low intake (RR 0.93, CI 0.85-1.01). Tested and not supported.
Mendelian randomization (genetic, unconfounded) shows linear harm: higher blood pressure, stroke, atrial fibrillation. No protective threshold.
Resveratrol is present only in trace amounts; lab-relevant doses are unreachable by drinking. No robust human benefit beyond the ethanol's harms.
Ethanol is an IARC Group 1 carcinogen; WHO and GBD 2016 find the health-loss-minimizing level is zero.
| # | Study | Type | Size | Funding / COI | Key limitations |
|---|---|---|---|---|---|
| 1 | Zhao/Stockwell 2023, bias-corrected meta-analysis | Systematic review and meta-analysis of cohorts | 107 studies; ~4.8M people; ~425,564 deaths | Independent Canadian Centre on Substance Use and Addiction subcontract on a Health Canada grant; no alcohol-industry money. Stockwell disclosed expert-witness fees. | Observational; relies on correcting abstainer bias, not on randomized trials. |
| 2 | Lankester 2021, Mendelian randomization (UK Biobank) | Mendelian randomization | ~337,484 in MR analysis | Independent US NIH NHLBI and NIDDK grants. Co-author Ingelsson disclosed prior GlaxoSmithKline employment (pharma, not alcohol). | MR assumes no pleiotropy; ADH1B variant rare in some populations; some estimates attenuate in multivariable models. |
| 3 | GBD 2016 Alcohol Collaborators (Lancet 2018) | Comparative risk assessment | 195 countries; 694 data sources; 592 risk studies | Independent Gates Foundation / academic (IHME, University of Washington); no industry funding. | Modeled relative risks pooled across heterogeneous sources. |
| 4 | WHO 2023 statement / Lancet Public Health | Position statement synthesizing evidence | Policy statement | Independent World Health Organization; no industry funding. | Statement, not primary data; rests on IARC carcinogen classification. |
| 5 | NIH MACH15 cancellation (STAT 2018) | News report on a cancelled RCT | Trial planned ~7,800 participants (terminated) | Industry-funded ~$100M largely from liquor companies via the NIH Foundation; cited as evidence of conflict, not of an alcohol effect. | Trial never ran; design judged too short to detect cancer harms. |
Three independent lines (bias-corrected cohorts, genetics, and global-burden modeling) converge on the same answer, which is unusual and strengthens confidence.
Unproven ≠ disproven
The narrow resveratrol-specific red-wine benefit is better called unproven than disproven: it is untested at reachable doses rather than tested and failed.
Where claim and evidence diverge
No large, long, unconflicted randomized trial of moderate drinking versus abstention has ever been completed, so even the corrective verdict rests on observational and genetic evidence.
The money trail
The flagship trial designed to prove benefit, NIH MACH15 (~$100M), was largely funded by liquor companies via the NIH Foundation and was cancelled in 2018 after staff improperly solicited industry money and framed the study to show benefit.
The studies that overturned the J-curve were independently funded (Health Canada, US NIH, Gates Foundation).
The honest read
The survival advantage of moderate drinkers is mostly an accounting error about who counts as a non-drinker, and the red-wine angle does not survive the dose math; there is no safe level for overall health.
What would change this verdict
A large, long, industry-independent randomized trial showing lower all-cause mortality at low intake versus true lifelong abstainers.
Mendelian-randomization evidence across diverse populations showing a genuine protective threshold rather than linear harm.
Sources
- Zhao J, Stockwell T, et al. Association Between Daily Alcohol Intake and Risk of All-Cause Mortality. JAMA Network Open. 2023;6(3):e236185.
- Lankester J, et al. Alcohol use and cardiometabolic risk in the UK Biobank: A Mendelian randomization study. PLoS One. 2021;16(8):e0255801.
- GBD 2016 Alcohol Collaborators. Alcohol use and burden for 195 countries, 1990-2016. Lancet. 2018;392(10152):1015-1035 (PMID 30146330).
- World Health Organization (Europe). No level of alcohol consumption is safe for our health. 4 January 2023.
- NIH to end funding for the MACH15 moderate-drinking trial. STAT News, 15 June 2018.
- University of Washington / IHME. No safe level of alcohol, scientific study concludes (GBD 2016 summary).
- Union for International Cancer Control (UICC). Cancer and alcohol (IARC Group 1 classification).
People also ask
- Is red wine actually good for your heart?
- No. Red wine's resveratrol is present only in trace amounts, and the doses studied in labs are unreachable by drinking. Genetic (Mendelian randomization) evidence shows alcohol's effect on the heart is linear and harmful, with higher blood pressure, stroke, and atrial fibrillation.
- Why do moderate drinkers seem to live longer in studies?
- Mostly a study-design artifact. Sick people who quit drinking often get counted as non-drinkers, making that group look unhealthy. A bias-corrected meta-analysis of 107 cohorts found no significant survival benefit at low intake (RR 0.93, CI 0.85-1.01).
- Is there a safe amount of alcohol to drink?
- No safe level exists for overall health. Ethanol is an IARC Group 1 carcinogen, and both the WHO and the GBD 2016 analysis concluded the health-loss-minimizing level of consumption is zero. Harm is present even at low intake.
- What is the J-curve for alcohol and mortality?
- The J-curve is the old idea that light drinkers outlive both abstainers and heavy drinkers. Once you correct for who ends up in the non-drinker group, that survival advantage largely evaporates. The protective dip is mostly an accounting error, not a real biological effect.
Part of our guide: Longevity diets, fact-checked
Caveat is journalism, not medical advice. We check public claims against published evidence; we don’t diagnose, treat, or tell you what to take.