Does Wine Extend Lifespan? What the Evidence Says

For about thirty years the answer looked like a qualified yes. A glass of red with dinner sat inside official drinking guidelines, the J-shaped curve turned up in cardiology lectures, and moderate drinkers really did die at lower rates than abstainers in study after study. That was not a fringe belief. It was mainstream public health advice, and holding it was the reasonable position.

What changed is not a new trial. It is a correction applied to the old ones. Zhao and colleagues pooled 107 cohort studies, 724 risk estimates, 4,838,825 participants and 425,564 deaths, published between 1980 and July 2021 (JAMA Network Open, 2023). After adjusting for sampling variation, former-drinker bias and study-quality criteria, occasional drinkers (more than 0 to under 1.3 g of ethanol a day) came in at a relative risk of 0.96 against lifetime non-drinkers (95% CI 0.86 to 1.06, P = .41), and low-volume drinkers (1.3 to 24.0 g a day) at 0.93 (P = .07). Neither is statistically significant. The benefit did not shrink. It stopped being there.

Quick answer

• The mortality benefit does not survive bias correction. In the 107-study, 4,838,825-participant meta-analysis, low-volume drinking came in at RR 0.93 (P = .07) and occasional drinking at RR 0.96 (P = .41), neither significant. Risk was significantly higher at 45 to 64 g a day (RR 1.19) and at 65 g or more (RR 1.35), both P < .001.
• The resveratrol story does not survive arithmetic. A 5 oz glass of red wine carries roughly 0.3 mg of trans-resveratrol. The lowest dose that produced a positive metabolic result in a human trial was 150 mg a day. That is 534 glasses a day, and since a 5 oz glass of 12% wine is one US standard drink by definition, 534 standard drinks a day.
• Pleasure is a legitimate reason on its own. None of this is guidance on what anyone should drink, and it is not meant to be read as any. What changed is the filing, not the pleasure. The only thing that changed is the filing. If you want wine to be something other than a consumable, fine wine is a real asset market, and the honest version of that pitch includes three consecutive down years on the industry benchmark.

Where the benefit went

This is the most interesting part of the story,.

Cohort studies compare drinkers to non-drinkers. Everything depends on who counts as a non-drinker. For decades the reference group swept together people who had never drunk, people who drank occasionally, and, critically, people who used to drink and had stopped. A large share of that last group stopped because something went wrong: a diagnosis, a liver result, a doctor's instruction, a medication that does not mix with alcohol, general frailty in old age. Sick people quit drinking. Then they got counted as non-drinkers.

That one sorting error does two things at once. It loads the abstainer group with people who were already unwell, which pushes non-drinker mortality up. And it leaves the moderate-drinker group full of people healthy enough to still be drinking, which pushes drinker mortality down. The famous J-curve is what you get when those two artefacts meet in the middle. This is former-drinker bias, sometimes called sick-quitter bias, and it requires nobody to have done anything dishonest. It is a definitional problem in the reference category.

The cleanest evidence that this is the whole story comes from splitting the same literature by quality. Stockwell and colleagues took the same 107 studies and defined "higher quality" as three things: mean cohort age 55 or younger at recruitment, follow-up continuing past 55, and former and occasional drinkers excluded from the abstainer reference group (Journal of Studies on Alcohol and Drugs, 2024).

In the studies that met those criteria, low-volume drinkers came in at RR 0.98 against abstainers (95% CI 0.87 to 1.11). No benefit. In the studies that did not meet them, RR 0.84 (0.79 to 0.89). A clear benefit.

0.98 against 0.84. Same corpus, same populations, same literature. The entire protective effect lives in the gap between studies that handled the abstainer group carefully and studies that did not. The authors put it plainly: "Studies with lifetime selection biases may create misleading positive health associations."

Two more numbers are worth carrying, and the first needs the same caveat applied above. In Stockwell's nonsmoking cohorts, low-volume drinkers came in at RR 1.16 (0.91 to 1.41). That interval crosses 1, so it is not statistically significant either, and it would be dishonest to present it as though it were. Separately, in the Zhao meta-analysis, female drinkers as a group showed significantly higher all-cause mortality than female lifetime non-drinkers, RR 1.22 (P = .03).

What the genetics say

Observational studies have a permanent problem here. People who drink a moderate amount of wine are different from people who do not, in ways that are hard to measure and easy to leave in the residual: more money, better diets, more exercise, better access to care.

Mendelian randomization gets around a lot of that. Instead of comparing drinkers to non-drinkers, it uses genetic variants associated with alcohol consumption as the exposure. Those variants are assigned at conception, long before income, diet or exercise exist, so they are far harder to confound with lifestyle.

Biddinger and colleagues ran it on 371,463 UK Biobank participants, mean age 57.0, consuming a mean 9.2 standard drinks a week (JAMA Network Open, 2022). A one-standard-deviation increase in genetically predicted alcohol intake was associated with 1.3 times the risk of hypertension (95% CI 1.2 to 1.4, P < .001) and 1.4 times the risk of coronary artery disease (95% CI 1.1 to 1.8, P = .006).

The same paper names the mechanism behind the older result directly: "Light to moderate alcohol consumption was associated with healthier lifestyle factors, adjustment for which attenuated the cardioprotective epidemiologic associations with modest intake." Its conclusion was that "alcohol consumption of all amounts was associated with increased cardiovascular risk," non-linearly, with light intake a minimal increase and heavy intake rising exponentially.

Two methods, two different failure modes, same direction of travel.

The resveratrol arithmetic

This part you can check with a calculator.

Start with the wine. The best-sourced pooled figure for red wine is a mean of 1.9 plus or minus 1.7 mg of trans-resveratrol per litre, with an observed range from non-detectable to 14.3 mg/L (Stervbo et al., Food Chemistry, 2007). The review written specifically to answer this question restates it verbatim: "The average red wine can be expected to contain ~1.9 +/- 1.7 mg trans-resveratrol/L" (Weiskirchen and Weiskirchen, Advances in Nutrition, 2016).

A US standard glass of wine is 5 fl oz, which is 147.87 mL, or 0.1479 L. So:

• 1.9 mg/L x 0.1479 L = 0.28 mg per glass (average red)
• 5.4 mg/L x 0.1479 L = 0.80 mg per glass (French Pinot noir, the highest varietal average in the pooled data)
• 14.3 mg/L x 0.1479 L = 2.1 mg per glass (the single most resveratrol-rich wine in the literature)

Now the other side of the ledger. Human resveratrol trials cluster between 150 and 1,000 mg a day, with oncology dose-finding work running to 5,000 mg. The lowest dose that produced a headline positive human metabolic result was 150 mg a day for 30 days in 11 obese men (Timmers et al., Cell Metabolism, 2011).

Divide one by the other:

• 150 mg divided by 1.9 mg/L = 78.9 L a day = 534 glasses a day
• From French Pinot noir at 5.4 mg/L: 27.8 L a day, or 188 glasses
• From the most resveratrol-rich wine ever measured: 10.5 L a day, or 71 glasses

Here is the line that closes it. A 5 oz glass of 12% ABV wine contains 147.87 mL x 0.12 x 0.789 g/mL = 14.0 g of ethanol, which is exactly one US standard drink by definition. So the glass count is the standard-drink count. The alcohol arrives in enormous quantity orders of magnitude before the resveratrol arrives in pharmacological quantity.

Weiskirchen and Weiskirchen ran the same sum against real French consumption data, 31.7 L of red wine plus 11.7 L of white a year, and got "~70 mg resveratrol/y (or 0.2 mg/d) or 5000 times less than the proposed therapeutic dose of 1 g/d." Their conclusion: "it is not possible to absorb the recommended dose of resveratrol through uptake of any of these nutrients or combinations thereof."

Two further multipliers stack on top of the dose gap. First, bioavailability: a single 5 g oral dose produced a peak plasma resveratrol of only 2.4 micromol/L, below the 5 micromol/L that in vitro chemopreventive effects require, because first-pass sulfation and glucuronidation dominate (Boocock et al., Cancer Epidemiology Biomarkers and Prevention, 2007). The shortfall is not only that there are too few milligrams. It is that the milligrams you do absorb largely do not arrive as resveratrol.

Second, the mouse study everyone is remembering is thinner than the memory. Baur and colleagues fed middle-aged mice on a 60% fat diet resveratrol at 5.2 and 22.4 mg/kg/day, and the high dose "reduced the risk of death from the HC diet by 31% (hazard ratio = 0.69, P = 0.020)" (Nature, 2006). Convert that to a human equivalent by body surface area, the method proposed in direct response to press coverage of those very papers (Reagan-Shaw et al., FASEB Journal, 2008), and 22.4 mg/kg/day in a mouse becomes about 109 mg a day for a 60 kg adult. That is the most generous number available anywhere in this literature, and it still needs 57.4 L, or 388 glasses, of average red wine a day.

The honest caveats run in both directions. Baur restored obese mice toward the survival curve of normal-diet mice; it did not push any group past normal lifespan. The 2008 follow-up was titled "Resveratrol Delays Age-Related Deterioration and Mimics Transcriptional Aspects of Dietary Restriction without Extending Life Span." The NIA Interventions Testing Program found no survival effect from resveratrol at all in genetically heterogeneous mice across three sites (Miller et al., 2011). And the supplement side is unresolved, not vindicated: the largest and best-controlled human trial in this set, 1,500 mg a day for four weeks in 24 obese men, was null on insulin sensitivity, blood pressure, energy expenditure and ectopic fat, and its authors wrote that the result "raises doubt about the justification of resveratrol as a human nutritional supplement in metabolic disorders" (Poulsen et al., Diabetes, 2013). None of this is a case for a pill. It is a case that the wine number and the trial number were never in the same universe.

Where the French Paradox came from

The phrase is French, and older than most people assume. Three INSERM epidemiologists, Richard, Cambien and Ducimetiere, used it in La Nouvelle Presse Medicale in March 1981, and Richard put it in a paper title in 1987: French coronary mortality was moderate and noticeably lower than in most industrial European or English-speaking countries despite saturated fat consumption comparable to high-risk nations.

The version everyone remembers is American and arrived a decade later. On 17 November 1991, CBS 60 Minutes ran a segment by Morley Safer featuring Serge Renaud of Bordeaux and INSERM. Safer's on-air conclusion was that the explanation of the paradox "may lie in this inviting glass" of red wine. Audience estimates range from over 20 million viewers to 33 million depending on the source, so treat that number as approximate. US red wine sales rose about 39 to 40 percent in the year that followed, per Wine Spectator and the historian Rod Phillips. That figure comes from wine-trade and popular-history reporting rather than audited market data, and I could not find a primary industry dataset behind it.

Chronology matters here, because it usually gets reversed. The Renaud and de Lorgeril paper in The Lancet appeared on 20 June 1992, seven months after the broadcast, not before it. And what it argued was narrower than the legend: that the paradox "may be attributable in part to high wine consumption," and that since HDL concentrations are not higher in France than elsewhere, moderate alcohol was probably acting "rather through a haemostatic mechanism," meaning platelets. It was an argument about alcohol, with wine as the French delivery vehicle, built on cross-population correlations and food supply data. It was not an argument about resveratrol at all.

Then the paradox itself came apart, in three independent ways.

Death-certificate coding. Law and Wald attributed about 20 percent of the France versus UK coronary gap to under-certification of coronary heart disease on French death certificates (BMJ, 1999). Ferrieres, writing from the French side, showed what the correction actually does: Toulouse at 53 per 100,000 by official statistics against 91 per 100,000 once WHO MONICA registries validated events under standardised criteria, versus Glasgow at 365 and Belfast at 279 (Heart, 2004). Correcting the coding narrows the gap and France still lands in the low-risk group, so this is a partial explanation, not the whole one. It also exposed a north to south gradient inside France, Lille above Strasbourg above Toulouse, which is awkward for any single national wine story.

Time lag. Also Law and Wald: cross-country coronary mortality tracks animal fat consumption and serum cholesterol from roughly 30 years earlier far better than it tracks recent levels. French animal fat intake and serum cholesterol rose only recently; Britain's rose decades earlier. On that reading France was not paradoxical, it was simply earlier on the same curve. Marion Nestle had proposed the time-lag explanation first, in a 1992 letter to The Lancet, and Law and Wald acknowledged the priority in BMJ afterwards.

Ecological fallacy. Evans made the methodological point directly: correlating national wine supply with national coronary rates establishes nothing at the individual level, and inferring that an unmeasured factor must explain a roughly fourfold gradient because the measured risk factors did not is precisely the reasoning pattern that manufactures false causes (International Journal of Epidemiology, 2011). Ferrieres, who reported wine drinkers showing a 24 to 31 percent decrease in all-cause mortality against beer or spirits drinkers, argued the same thing from the data: a cluster of behaviours, not a molecule.

There is one more episode, and it needs care, because it is usually told wrong. On 11 January 2012 the University of Connecticut Health Center announced the conclusion of a three-year research misconduct investigation into Dipak K. Das, director of its Cardiovascular Research Center and a prominent resveratrol researcher. The report ran roughly 60,000 pages, covered more than seven years of laboratory activity, and concluded that Das was guilty of 145 counts of fabrication and falsification of data, centred on manipulated Western blot images. UConn notified 11 journals, froze externally funded research in his laboratory, declined $890,000 in federal grants already awarded, and began dismissal proceedings. Retractions reached 20 by March 2014. Das and his attorney disputed all charges from the outset, press reports in 2013 said he intended to sue the university, and I could not verify the outcome of any such litigation. He died on 19 September 2013, aged 67. The 145-count finding is UConn's institutional finding; I found no public federal Office of Research Integrity finding.

What the Das case did not do is debunk the French Paradox. His fabricated work was mechanistic animal and cell research on resveratrol and cardioprotection. It was not the epidemiology, and it was not the Renaud and de Lorgeril platelet hypothesis. The paradox was undercut by coding corrections, the time-lag argument and ecological-fallacy criticism, all of which are independent of him. The larger erosion of confidence in resveratrol came from independent null results: Semba and colleagues measured 24-hour urinary resveratrol metabolites in older community-dwelling adults and found no association with inflammatory markers, cardiovascular disease or cancer, and no prediction of all-cause mortality, with death rates across quartiles of 34.4, 31.6, 33.5 and 37.4 percent (JAMA Internal Medicine, 2014).

So the myth's survival is not one mistake. It is a French observation that was real, a mechanism attached to it that was plausible, a television segment that made it famous, a molecule that arrived afterwards and fit the shape of the story, and a reference-group definition that had been quietly manufacturing the effect in the underlying data the whole time.

If you love wine anyway

Nothing above is an argument for drinking less. Pleasure is a legitimate reason to do something, and a bottle opened with people you like does not need a hazard ratio to justify it. The only thing that changed is the filing. Wine is a thing you enjoy. It is not a thing you do for your healthspan, and the evidence stopped supporting that a while before most of us noticed.

Which leaves a more interesting question for anyone genuinely into wine: if the health frame is gone, is there a better one? One answer is ownership. Fine wine is a real asset market with real prices, and unlike almost everything else in the longevity category, the thing you buy is also the thing you can drink.

Vinovest is one platform for doing that without running your own bonded cellar. Here is what it actually is, taken only from its own published pages.

Vinovest fees, minimums, shipping costs and terms in this section were verified on vinovest.co on 1 August 2026. Fees move, so check them before committing.

Disclosure, and it belongs here rather than in a footer because this section is aimed at people committing real money: the Vinovest link in this piece is my personal referral link. Vinovest's published referral term is up to four months of free storage and insurance for both sides once the referred account is funded, so if you sign up through it and fund an account, I benefit. Whatever the exact figure lands at on my account, treat everything in this section as written by someone with an interest in your answer. That is not an affiliate commission, and Lifespan Vault is not paid by Vinovest for this article. Vinovest's own published referral term is "up to 4 months of free storage and insurance" for both sides once the referred person funds, it is legally structured as a purchase rebate rather than cash, and it is "void where prohibited," which means residents of some US states may get nothing. The signup page itself publishes none of those terms, which is why they are stated here. I am not a licensed adviser and none of this is investment advice. The link: vinovest.co.

What you own. Direct personal ownership of physical bottles and casks, not a fund, a share or a security. Vinovest issues ownership certificates viewable in the Activity tab and states that "these legal documents prove that you are the sole owner of the bottles." Its How It Works page: "you own your bottles and casks outright, while we handle the logistics, from sourcing and storage to authentication and insurance." On insolvency it says "your wines remain your property, fully insured and securely stored in world-class facilities," and that storage partners would contact you to transfer, ship or auction the collection.

The gap in that. Vinovest publishes no statement that customer wine is legally segregated from its corporate balance sheet, and names no bailment or trust structure. It also publishes no statement that holdings are not FDIC-insured or SIPC-protected. Neither protection applies, because wine is not a deposit and not a security, but the absence of the statement is itself worth noticing. There is no published registration status as an investment adviser or broker-dealer.

Minimum. $2,000 for a managed portfolio. Self-directed trading has no minimum. Both figures appear on Vinovest's pricing and how-it-works pages.

Fees. Annual management fee by tier, from the pricing page: Starter at a $2,000 minimum is 2.85%, Plus at $10,000 is 2.7%, Premium at $50,000 is 2.5%, Grand Cru at $250,000 is 2.25%. The fee covers "insurance, storage, authentication, and active management of your portfolio," prorated across the year and charged only on invested capital, with leftover cash not charged. Card processing runs 2.9% plus 30 cents on US cards, with an extra 1.5% on international cards. Selling outside the ideal window carries a 1.5% listing fee, waived for anything sold after five years. Unpaid storage and insurance fees accrue 1.5% interest after a month. Turning on auto-invest gets 5% off storage and insurance.

The clause most reviews skip. From the terms: "After 180 days of nonpayment of management fees, Vinovest has the right to sell the wines and whiskeys in the account portfolio at a 15% discount." Fees are due on the 1st of each month and must be paid within one month.

Storage. Bonded facilities in the UK, US, France and Belgium. Whiskey casks sit in temperature and humidity controlled bonded warehouses, Scotch in Scotland and American casks in Wyoming. Wines in storage carry "an insurance policy protecting them at full current market value, with appraisals updated annually by third-party experts." Vinovest names no warehouse operator and publishes no copy of the insurance policy, and its own terminology drifts between a fee that covers storage and separate "Management and Storage Fees" due monthly.

Can you drink it. Yes, with friction. The homepage says "we'll even ship your bottles to your doorstep if you'd like to drink them." The help centre's actual mechanics: email support with the wines and the address, wait for a quote, full cases only, no single bottles and no mixed cases, quarterly shipping schedules with cutoff dates, and shipping "for 1 case begins at £215" plus local taxes and duties you are responsible for. That figure is a GBP floor, not a price. Taking delivery removes the wine from bond, which is the moment the duty and VAT liability triggers and the bonded-storage tax advantage is given up.

The returns claims, handled carefully. Vinovest publishes individual exit examples with annualized returns on its track-record page, and different annualized figures for some of the same wines on its homepage. Dom Perignon Luminous 2010 appears at 30.7% on the homepage and 70.71% on the track-record page. Domaine Anne-Francoise Gros appears at 43.1% and 193.04%. Either the label "annualized" is being applied to different things or the datasets differ, and I am not quoting either as a return figure until that is reconciled. Both pages carry disclaimers: "past exits are not indicative of future results," "past performance is no guarantee of future results," and on how-it-works, "all images and return and projection figures shown are for illustrative purposes only."

The risks, in Vinovest's own words. "All items purchased from Vinovest are speculative in nature. The price of the wine and or whiskey you purchase may decrease or lose its value completely." Disputes go to binding arbitration under California law. Vinovest provides no tax advice, publishes no guaranteed buyback, no redemption window and no time-to-sale estimate. The clearest signal it gives about the holding period is that five-year sale-fee waiver.

And the market itself, from Vinovest's own blog: "the Liv-ex Fine Wine 1000 peaked in September 2022 and declined approximately 25-30% through mid-2025 before showing signs of stabilization and early recovery." That is the sentence to read next to any upside figure. The benchmark Liv-ex Fine Wine 100 fell 14.1% in 2023, 9.1% in 2024 and 2.5% in 2025, three consecutive down years, leaving it about 25% below the September and October 2022 peak.

Two risks that do not come from Vinovest. Fine wine is not a regulated financial instrument in the UK or the US, so there is no FCA or SEC recourse and no compensation scheme. And provenance risk is real: Rudy Kurniawan was convicted of wine fraud in the US and imprisoned, and the authentication specialist who testified against him estimates that $592m to $683m of his fakes remain in circulation. Her wider estimate that roughly 20% of fine wine traded is counterfeit is disputed within the trade and comes from someone who sells authentication services, so treat it as an attributed estimate rather than a fact.

Fine wine against the alternatives

The longest clean study of wine returns is Dimson, Rousseau and Spaenjers, "The Price of Wine" (Journal of Financial Economics, 2015), a repeat-sales regression on five long-established Bordeaux wines using auction and dealer prices from 1900 to 2012. Real annualized returns, in GBP:

Wine returned 5.3% before storage and insurance and 4.1% after, so the carry took 1.2 points a year, roughly 23% of the gross return. Over 113 years it was more volatile than equities and returned less. The authors also measured wine's market-model beta to equities at 0.44, or 0.57 excluding 1941 to 1948, which is positive and material, so the uncorrelated-diversifier pitch does not hold either. They cap their own result explicitly: First Growth returns are "best considered as an upper bound on the long-term investment performance of wine more generally," and accounting for transaction costs "would lower the relative performance of wine investments even further, especially over short horizons."

The recent window is not close either.

Against that, the S&P 500 returned 11.78% annualized over the five years to 30 June 2026, roughly +74% cumulative. Two caveats: the Liv-ex indices are GBP and price-only while the S&P figure is USD and total return, so a dollar-based wine buyer is carrying an unhedged currency position on top of the asset, and wine pays no dividend, coupon or rent to offset the carry.

Three structural points the index numbers hide:

• The printed price is not your exit. Dimson and colleagues, at 2012 Christie's London rates, calculated that "a seller may only receive about 75% of the amount that the winning bidder pays out." Sotheby's wine and spirits buyer's premium is now a flat 24% of hammer price, so the round-trip wedge is wider, not narrower. Consignment to cleared funds realistically runs two to six months, and settlement after the hammer typically takes 30 to 35 business days.
• Dispersion. In 2024 only 11 of the Liv-ex Power 100 brands rose, while 343 of 410 qualifying brands, 83.7%, fell. Index-level numbers hide the fact that most individual holdings lost money. Burgundy fell 27.8% from its October 2022 peak, and the First Growths, the classic blue chip of the category, are the worst five-year performer in the table above at -22.5%.
• Survivorship. Dimson and colleagues are blunt that poorly stored bottles leak, oxidize and get damaged, that auction houses would not now sell bottles of questionable provenance, and that the returns "have therefore probably been realized only by investors who stored their wines properly." An index of wines that still trade is, by construction, an index of the ones that survived.

Carry is regressive by bottle value. UK bonded storage runs broadly £12 to £15 per case a year excluding VAT with insurance at landed value, which is roughly 2.3% a year on a £600 case and under 0.5% on a £3,000 case. In a flat market, storage alone makes a cheap position lose money.

For context outside wine, the Knight Frank Luxury Investment Index fell 0.4% in 2025 after two down years and is up 38.6% over the decade. Its wine leg has historically been a curated icons basket rather than a whole-market measure, and the 2026 report quotes the Liv-ex 100 instead, which is the kind of change in the underlying series worth knowing about before leaning on any decade figure.

Bottom line

Wine does not extend lifespan, and the reason the opposite held for thirty years is more interesting than the conclusion ever was. It was not fraud and it was not stupidity. It was a reference group that quietly filled up with people who had stopped drinking because they were already ill, and once you take them out the protective effect leaves with them: RR 0.98 in studies that handled the abstainer group properly, RR 0.84 in studies that did not. The genetics point the same way. The resveratrol dose gap was never within three orders of magnitude of closing.

None of which is a reason to pour anything out. It is a reason to stop counting a glass of Burgundy as a health behaviour and start counting it as a pleasure, which is what it always was. If you want wine to do something else as well, ownership is a defensible frame: $2,000 gets you into a managed Vinovest portfolio at 2.85% a year, with the fee falling to 2.25% only at the $250,000 tier, you own the bottles outright with certificates, and you can have them shipped from £215 a case if you would rather drink them than sell them. Go in with the real numbers in front of you. Three consecutive down years on the Liv-ex 100. Roughly a quarter of round-trip value lost to auction friction. 4.1% real annualized over 113 years against 5.2% for equities. No segregation-of-assets disclosure, and no regulator behind any of it. Wine is a good thing to love. It was never a good thing to count on.

• Ryan, Founder