The alcohol J-curve did not survive its own control group

The finding that light drinkers outlive teetotallers does not survive correction for how the teetotaller group was assembled. A systematic review and meta-analysis of 107 cohort studies covering 4,838,825 participants and 425,564 deaths, published in 2023, found no significantly reduced risk of all-cause mortality among occasional or low-volume drinkers once the models adjusted for sampling variation, former-drinker bias and other prespecified study-quality criteria [s1]. Increased risk was evident at higher consumption levels, starting at lower levels for women than for men [s1]. The J-shaped curve, in other words, was largely an artefact of who ended up counted as a non-drinker.

The bias that built the curve

The problem is stated in the review's own framing: the analysis was designed to compare studies with and without some degree of control for biases affecting the distinction between abstainers and drinkers [s1]. Lifetime non-drinkers are a small and unusual group. Cohort studies that lump them together with people who stopped drinking — often because of illness, medication or advice — produce a reference group with elevated mortality for reasons that have nothing to do with abstention. Any drinking category then looks protective by comparison.

The review drew on 724 risk estimates from those 107 studies, published between 1980 and July 2021 [s1]. When the models adjusted for former-drinker bias and study quality, the apparent protection at low intake did not reach significance [s1].

Genetics points the same way

A separate approach avoids the abstainer problem entirely by using genetic variants associated with alcohol intake rather than reported drinking. In 371,463 UK Biobank participants (mean age 57.0, 46% men) consuming a mean 9.2 standard drinks a week, of whom 121,708 had hypertension, light to moderate drinking was associated with healthier lifestyle factors — and adjusting for those factors attenuated the apparent cardioprotective associations with modest intake [s2].

In linear Mendelian randomisation analyses, a one-standard-deviation increase in genetically predicted alcohol consumption was associated with 1.3-fold higher risk of hypertension (95% CI 1.2 to 1.4) [s2]. The authors' conclusion is that genetic epidemiology suggests alcohol consumption of all amounts is associated with increased cardiovascular risk, while marked risk differences exist across levels of intake, including those accepted by current national guidelines [s2].

That last clause matters and is usually cut. "All amounts carry increased risk" and "the risk at one drink is far smaller than at five" are both true statements about the same dataset.

The cancer argument, which was never J-shaped

WHO's Regional Office for Europe published a statement in The Lancet Public Health in January 2023 arguing there is no safe amount of alcohol that does not affect health [s3]. Its reasoning is carcinogenic rather than cardiovascular. Ethanol has been classified by IARC as a Group 1 carcinogen — the highest evidence category, which also contains asbestos, radiation and tobacco — and causes at least seven types of cancer including bowel cancer and female breast cancer [s3]. Because the compound causes cancer as it breaks down in the body, any beverage containing alcohol carries the risk regardless of price or quality [s3].

WHO's specific claim is about thresholds: currently available evidence cannot indicate a level below which the carcinogenic effects of alcohol switch on [s3]. It also reports that half of all alcohol-attributable cancers in the WHO European Region are caused by light and moderate drinking — defined there as less than 1.5 litres of wine, less than 3.5 litres of beer, or less than 450 millilitres of spirits per week — with that pattern responsible for the majority of alcohol-attributable breast cancers in women [s3]. Over 200 million people in the region are at risk of developing an alcohol-attributable cancer [s3].

On the trade-off that the J-curve implied, WHO is direct: there are no studies demonstrating that the potential beneficial effects of light and moderate drinking on cardiovascular disease and type 2 diabetes outweigh the cancer risk at those same levels for individual consumers [s3].

What is still contested

Not everything here is settled, and two things are worth separating. The claim that low-level drinking is protective is the one that has failed — in the corrected meta-analysis [s1], in the genetic analysis [s2] and in WHO's assessment [s3]. The claim that risk at low levels is large is not what any of these papers say; the 2023 meta-analysis locates increased mortality risk at higher consumption levels, with the threshold lower for women than men [s1].

A WHO adviser quoted in the 2023 statement notes that the suggested protective effects are tightly connected to the comparison groups chosen and the statistical methods used, and may not account for other relevant factors [s3] — which is a methodological claim, not a measurement of harm. Neither the meta-analysis nor the Mendelian randomisation study is a randomised trial, and no randomised trial of long-term moderate drinking against abstention has been completed.

This article is informational and is not medical advice. Alcohol interacts with many medications and with pregnancy, liver disease and dependence, and those are clinical questions.

Sources

• Association Between Daily Alcohol Intake and Risk of All-Cause Mortality: A Systematic Review and Meta-analyses — JAMA Network Open, 2023-03-31
• Association of Habitual Alcohol Intake With Risk of Cardiovascular Disease — JAMA Network Open, 2022-03-25
• No level of alcohol consumption is safe for our health — WHO Regional Office for Europe, 2023-01-04