The curve is real. The verdict is not. Science is still pouring.
It begins, as so many good arguments do, with a graph. Draw it out on a cocktail napkin…a line that starts high on the left, swoops elegantly downward, touches bottom somewhere around a glass of Burgundy a day, and then climbs again, steeply, toward the kind of numbers that appear on actuarial tables and ruin dinner parties. Scientists call it the J-curve. Everyone else calls it the best news they have heard from a laboratory in years.
The J-curve describes the relationship between how much a person drinks and how likely that person is to die prematurely. For almost a century, the scientific community has been aware of this J-shaped relationship between alcohol consumption and all-cause mortality. Moderate drinkers appear to live longer than both abstainers and heavy drinkers. If you drink nothing at all, the data suggest, your risk of dying sits at a level that is, paradoxically, higher than the person who has a glass of wine with dinner. If you drink a great deal, your risk climbs sharply. But somewhere in the middle, around one drink a day, perhaps two for men of a certain constitution, the curve bottoms out into what epidemiologists call the nadir, a Latin word meaning, roughly, the best possible place to be. Enobytes | ScienceDirect
Now, before anyone orders a second round, a word of caution that Sir Richard Doll himself, the great British epidemiologist who spent his career making smoking look as lethal as it is, would have readily offered: correlation is not causation, and a graph on a napkin is not a prescription.
Still, the graph is hard to ignore.
The Evidence, Considerable as It Is
The man who did more than anyone to give the J-curve scientific respectability was, improbably, the same man who proved that cigarettes kill. The J-curve first appeared, to widespread notice, in a 1994 study by Sir Richard Doll that employed the same method of surveying doctors that he had used 40 years earlier to show the association between smoking and lung cancer. Doll was a physician of ferocious precision, not a man given to false comfort. And yet, a few months before his death in 2005, he published a second paper, a 23-year prospective follow-up of 12,000 male British doctors that not only confirmed what the first had found but addressed the main objection to it.
The objection was this: perhaps the abstainers who appeared to die at elevated rates were not lifelong teetotalers at all. Perhaps they were former drinkers who had quit because illness had already overtaken them…”sick quitters,” in the parlance of the field, who skewed the comparison unfairly against sobriety. Doll tested this directly. When ex-drinkers were considered together with current drinkers and compared only against men who had been non-drinkers in both questionnaires, ischemic heart disease, respiratory disease, and all-cause mortality were still significantly lower in the drinkers. Overall mortality among current drinkers showed a relative risk of 0.81. Citation: Richard Doll Consortium
Doll concluded in a 1998 review that evidence showing small amounts of alcohol reduce the risk of vascular disease by roughly a third, and reduce total mortality in middle and old age, was, by that point, massive — and that alternative explanations for the observed inverse relationships had been ruled out. Citation: Wiley Online Library
This was not a fringe opinion. It was, and remains, the view of the mainstream of cardiovascular epidemiology.
In 1997, Michael Thun and his colleagues at the American Cancer Society, working alongside Doll and Richard Peto, published a large prospective study in the New England Journal of Medicine examining the drinking habits of 490,000 Americans followed for nine years. Alcohol consumption was found to have both adverse and beneficial effects on survival, and the researchers examined the balance of these in what was, at the time, an unusually large and well-characterized population. Citation: New England Journal of Medicine
A decade later, Eric Rimm, professor of epidemiology and nutrition at the Harvard T.H. Chan School of Public Health, helped assemble much of what the field had learned into a memorable formulation. He wrote that moderate alcohol consumption is associated with a lower risk of coronary heart disease in healthy men and women, and that this association — documented in more than 70 epidemiologic studies — is likely causal and can be explained, in part, by alcohol’s beneficial effects on serum lipids and clotting factors. Rimm would later say, with the practiced understatement of a man who has testified at congressional hearings, “There’s good evidence that, in general, moderate drinkers who average one to two drinks a day tend to live longer. Whether that is directly linked with alcohol, other lifestyle factors, or some combination is still being explored.” Citation: Pub Med Harvard Health
The plumbing behind the association has been studied with some care. Kenneth Mukamal of Harvard, in a study that tracked the drinking habits of nearly 40,000 men over 12 years, observed that men who drank moderate amounts three or more times a week had a risk of heart attack 30 to 35 percent lower than non-drinkers. Mukamal noted that alcohol raises levels of HDL, the so-called good cholesterol, while also impacting insulin sensitivity and platelet function and clotting factors. These are not mysterious mechanisms. They are among the best-understood pathways in cardiovascular medicine.
The 2011 systematic review by Paul Ronksley, Kenneth Mukamal, and colleagues in the BMJ drew on 84 prospective cohort studies and reached a similar conclusion. Light to moderate alcohol consumption was associated with a reduced risk of multiple cardiovascular outcomes, with the lowest risk of coronary heart disease mortality occurring at one to two drinks a day. Citation: Semantic Scholar
And in a meta-analysis of 34 prospective studies covering more than a million subjects and more than 94,500 deaths, a clear J-shaped relationship between alcohol intake and mortality was observed. The lowest mortality risk, a reduction of roughly 19 percent was observed at approximately six grams of alcohol per day, or about half a drink, but reduced mortality compared with abstention persisted with up to four drinks a day in men and two drinks a day in women. The J-shaped curve persisted after adjustment for many relevant confounding variables. Citation: JACC
The Skeptics, Who Are Not Wrong
One does not mention the J-curve in certain academic corridors without provoking a response. And the response, it must be said, is not without merit.
The core objection is methodological. Mounting evidence suggests these associations might be due to systematic biases that affect many observational studies. The sick-quitter problem, though addressed by Doll’s careful design, has not been put to rest universally. Moderate drinkers, as a group, tend to be employed, socially integrated, and possessed of the kind of orderly lifestyle habits that also happen to predict a long life. Isolating the contribution of the wine from the contribution of the dinner party, the regular sleep, the gym membership, the job with health insurance…this is the eternal challenge of epidemiology. Citation: PubMed Central
More recently, researchers have turned to Mendelian randomization, a genetic method that sidesteps some of the confounding in traditional observational studies by using inherited variants in alcohol-metabolizing genes as a proxy for consumption. Genetic analysis has not provided evidence for a nonlinear association, failing to support observational studies suggesting benefits from modest alcohol intakes. While the greatest mortality risks are seen with high intakes, the authors concluded that even moderate intake poses a risk. Citation: Oxford Academic
This is a genuine scientific disagreement, not a political one, though it sometimes gets treated as the latter.
Mukamal and Rimm have acknowledged that many people should not drink at all — pregnant women, people recovering from alcoholism, and those who cannot control how much they drink. For many cancers, the picture is different still. Drinking modest amounts raises the risk of breast, colon, and esophageal cancer. The heart may benefit from what the liver, in other contexts, does not. Citation: Harvard Magazine
The Trial That Was, and the One That Must Come
What the field has lacked, for all these decades of painstaking observation, is a randomized controlled trial, the kind of experiment in which participants are assigned by chance to drink or not drink, and followed long enough for an answer to emerge. Observational studies, however large and well-designed, watch the world as it is. They cannot watch the world as it might have been if half the participants had drunk a glass of Merlot every evening for six years.
Such a trial was designed. The Moderate Alcohol and Cardiovascular Health Trial known, with the bleak creativity of acronym-hungry researchers, as MACH15, would have enrolled approximately 7,800 adults over 50, assigned them to either one standard drink daily or abstention, and followed them for an average of six years. Its primary objective was to determine the effects of one standard serving of alcohol daily compared with no alcohol intake on the risk of incident cardiovascular disease among adults at above-average cardiovascular risk. The primary outcome was a composite of non-fatal heart attack, non-fatal ischemic stroke, hospitalization for angina, coronary or carotid revascularization, or all-cause mortality. Citation: clinicaltrials
The trial was terminated in 2018, not because of anything wrong with its science, but because of what investigators at the New York Times reported: improper contacts between the National Institute on Alcohol Abuse and Alcoholism and the alcohol industry during the early development of the study. An internal investigation, prompted by a dogged New York Times report, uncovered inappropriate interactions between the alcohol industry and the NIAAA in the execution of MACH15. The science was sound. The politics were not. Undark Magazine
Mukamal and Rimm have blamed a lack of funding and polarized attitudes about alcohol consumption for the difficulty in pursuing large randomized studies on drinking. Rimm has put the frustration plainly: “The biggest frustration for me is the implication that the science is settled so clearly and securely that we don’t actually need to get the right answer to this.” Citation: Harvard Magazine
The right answer, as of this writing, does not exist. What exists is the J-curve, carved out of decades of observation, challenged by genetics, complicated by cancer, and stubbornly persistent in the cardiovascular data. What also exists, building quietly in the background, is the expectation of a new randomized trial…and more carefully designed, more insulated from industry contact, more likely to survive the scrutiny that felled MACH15. If and when that trial reports results, perhaps sometime around 2029, the conversation will finally have the evidence it deserves.
Until then, the curve sits there on the page, graceful and ambiguous, making an argument that neither the glass nor the abstainer has yet definitively won.
Those who enjoy a drink in the meantime are neither vindicated nor condemned. They are simply, like the rest of us, waiting for data.
—
The key observational studies behind the J-curve:
- Doll R, Peto R, Boreham J, Sutherland I. “Mortality in relation to alcohol consumption: a prospective study among male British doctors.” Int J Epidemiol. 2005;34(1):199–204.
- Thun MJ, Peto R, Lopez AD, et al. “Alcohol Consumption and Mortality Among Middle-Aged and Elderly U.S. Adults.” N Engl J Med. 1997;337:1705–1714.
- Di Castelnuovo A, et al. Meta-analysis of 34 prospective studies (>1M subjects). 2006.
- Ronksley PE, Brien SE, Turner BJ, Mukamal KJ, Ghali WA. “Association of alcohol consumption with selected cardiovascular disease outcomes: a systematic review and meta-analysis.” BMJ. 2011;342:d671.
- Rimm EB, Moats C. “Alcohol and Coronary Heart Disease: Drinking Patterns and Mediators of Effect.” Ann Epidemiol. 2007.
