If you’ve been following medical research news and look to the Mayo Clinic or Harvard School of Public Health for advice on alcohol use, you might get a mild whiplash.
While neither website outright “encourages” moderate alcohol use, both come pretty close. Harvard’s site states that “moderate drinking seems to be good for the heart and circulatory system, and probably protects against Type 2 diabetes and gallstones.”
The Mayo Clinic site states that alcohol consumption may reduce risk of heart disease, stroke and diabetes. The Harvard site echoes this and concludes with oddly ambiguous advice: “If you don’t drink, there’s no need to start. You can get similar benefits with exercise … or healthier eating.”
Wait, you might say, didn’t a massively influential study not long ago find there is no safe level of alcohol use?
In 2018 the elite British medical journal Lancet published a meta-study surveying nearly 600 studies and people in 195 countries. It concluded, “the risk of all-cause mortality, and of cancers specifically, rises with increasing levels of (alcohol) consumption, and the level of consumption that minimises health loss is zero.”
No safe level? That headline ricocheted around the world. Yet four years later two of the websites of America’s preeminent health institutions, Mayo and Harvard School of Public Health, read as if it never happened.
The ground is shifting, and the notion that there is any benefit, or truly safe level of alcohol consumption, appears increasingly antiquated.
The decadeslong consensus that a little alcohol is better than none is oft-repeated in prominent places. But the ground is shifting, and the notion that there is any benefit, or truly safe level of alcohol consumption, appears increasingly antiquated.
Welcome to the battle of the J-curve — the latest in the simmering alcohol wars that have run hot and cold in America since the early years of the 19th century. But this time the battlefield is global, and what’s at stake is the health of billions who are looking to science to answer whether it’s truly safe to imbibe.
Graph the letter J on an X-Y axis. Moving up on the graph means greater health damage. Moving to the right is higher alcohol use. The J-curve suggests that a little alcohol is better than none. Sometime in the early 1990s that curve became cemented in elite and public opinion. Hundreds of studies on alcohol’s effect on cardiovascular health to dementia seemed to find that light or moderate drinking was better than abstinence.
This consensus persisted virtually unchallenged for years, a rare case of a seemingly vindicated vice.
But the science was never settled. In 2009, a seven-year study of 1.2 million women found that even moderate drinking increased breast, liver, rectum, mouth, throat and esophagus cancers. The researchers estimated that 13% of cancers among these women stemmed from alcohol. “There were no minimum levels of alcohol consumption that could be considered to be without risk,” Naomi Allen, a cancer researcher at the University of Oxford, told WebMD.
But the J-curve still held its ground. When Anya Topiwala, a psychiatry professor at Oxford University in England, launched a large study of alcohol and dementia, she expected to confirm the protective effect of light drinking. Topiwala and her co-authors used the UK Biobank, an ongoing study of 500,000 Britons that began in 2006. The Biobank paired alcohol use data with brain scans, allowing the researchers to look inside the brain for subtle effects long before dementia became visible in behavior.
Published in 2017 in the British Medical Journal, the surprise results found serious early brain harm among even moderate drinkers and no protective effect for light drinkers.
“We found that even small amounts of alcohol seem to be associated with literally less brain volume across nearly the whole brain,” Topiwala told me. They also found weaker “cabling of the brain fibers” and less communication between brain regions. “These brain measures can be picked up years before the onset of dementia,” Topiwala told me. “Subtle changes, maybe years before they’re going to develop any memory problems.”
The 2017 British Medical Journal study concludes with some policy implications, endorsing recent reductions in U.K. alcohol guidelines and questioning laxer U.S. standards, “which suggest that up to 24.5 units a week is safe for men, as we found increased odds of hippocampal atrophy at just 14-21 units a week, and we found no support for a protective effect of light consumption on brain structure.”
The surprise results found serious early brain harm among even moderate drinkers and no protective effect for light drinkers.
Contradictory studies continue to appear, Topiwala said, but they generally are much smaller in scale and do not use brain imaging. And her results are confirmed in a new study with even larger data that is now pending peer review.
So how do we explain all those studies that supported the J-curve?
One theory is that many abstainers are actually already unwell. “If you look at the more recent meta-analyses that account for ‘sick quitters,’” said Dr. Amy Justice, a clinical epidemiologist at the Yale School of Medicine, “people who quit drinking in midlife or later because they got into trouble with their alcohol, then you get a very different result.”
Topiwala also points to U.K. data suggesting that moderate drinkers are socioeconomically better off than abstainers. “At least in the U.K.,” she said, “abstainers tend to be more female, lower in socioeconomic class, with higher levels of heart disease and diabetes.”
Either way, many researchers now think the J-curve is more likely explained by “confounding variables.”
By 2017 the J-curve was under attack from those studying alcohol’s effect on cancer and dementia. But there were still many widely cited studies showing better heart health and lower stroke risk, among other benefits of light alcohol use.
The J-curve was not yet dead.
But in 2018 the U.S. National Institutes of Health made a mess of things. As the year began, the NIH were prepared to launch an ambitious study with 7,800 volunteers who would either abstain from alcohol or imbibe one daily drink over the next 10 years. The study was largely funded by alcohol trade groups. Then the roof caved in.
In March, as volunteers were already joining the study, The New York Times ran a blistering exposé of unethical fundraising efforts and implied promises to the liquor industry. In a PowerPoint presentation to one alcohol industry group, the researchers argued that a “definitive clinical trial” was needed to “show that moderate alcohol consumption is safe and lowers the risk of common diseases. That level of evidence is necessary if alcohol is to be recommended as part of a healthy diet” (emphasis added).
“Of course they would pay for it,” the Times quoted Dr. Michael Siegel, a professor of community health sciences at Boston University School of Public Health: “They’re admitting the trial is designed to provide a justification for moderate drinking. That’s not objective science.” In June, the NIH yanked support for the study.
Then in August 2018, the landmark Lancet study shook the J-curve at its foundations. This massive meta-study of nearly 600 smaller studies collected data from 195 countries and considered 23 health factors, including car accidents, suicides, tuberculosis, liver disease, cardiovascular disease and cancers. The study found no safe level of alcohol use.
To be fair, the study did peg the harm of one drink a day at a very low level, increasing “alcohol-related health problems slightly, to 918 per 100,000 people from 914 per 100,000.” But there was still no safe level, and certainly no benefit.
However messy and contested, medical science has a way of settling over time. And, at some point, biomarkers at the cellular level may answer even more questions. In the meantime, research remains indirect, epidemiological and vulnerable. Ethics and cost challenges make large-scale controlled research unlikely.
But there are some alternate weapons at the disposal of researchers. One is Mendelian randomization, a complicated but precise tool that relies on the random distribution of certain inherited genes, genes that decisively influence certain behaviors or biological factors.
For example, many people of East Asian descent have a gene that prevents them from metabolizing alcohol, causing inflammation and flushing in the face, and severe discomfort. People with that gene tend not to use alcohol. Because the gene is random and not crossed by any other variables, any differences between those who have the gene and those who do not can be attributed to alcohol. It’s as if nature itself created a controlled experiment.
Researchers at Stanford recently used Mendelian randomization with data from the UK Biobank, isolating a couple of genes connected to alcoholism. Their results, they wrote, add “to the mounting evidence using (Mendelian randomization) that alcohol use does not improve cardiovascular health even in moderate amounts and likely worsens it when all other factors are considered.”
In light of other proven harms, the authors recommend that “it is time to reconsider current public health recommendations in the U.S. and other countries which suggest that up to two drinks/day for men and one drink/day for women is not harmful and possibly beneficial to cardiovascular health.”
Of course, personalized biomarkers that might show early evidence of harm at the cellular level could tell us more. Blood tests already isolate liver enzymes that indicate heavy drinking and incipient cirrhosis, but often those results are revealed too late after the damage is done. What’s needed is something like Topiwala’s brain scans — but for the rest of the body and used in routine screening.
“It is time to reconsider current public health recommendations” (regarding alcohol consumption). — Stanford researchers
“We don’t ask diabetic patients what their glycosylated hemoglobin is,” Yale’s Amy Justice told me. “We check it, and then we have a conversation about it. But we ask people how much alcohol they drink, and we expect them to have a clear sense of it. And that’s a little absurd.”
There is one such marker already available: phosphatidylethanol. It only lasts for three to four weeks in the bloodstream, but it would not be complicated for a doctor to add it to a blood panel, Justice said. And she sees it as a pretty reliable measure of heavy alcohol exposure, especially if the test were repeated over time.
Whether there is any safe level of “light” or “moderate” drinking is one question. Defining those terms is another.
In heavy drinking Canada, a 2020 study led by Adam Sherk, a health researcher at the University of Victoria, caused a stir. The Toronto Globe and Mail headlined its resulting editorial: “Let’s face it Canada, even drinking in moderation can be dangerous.”
The editorial concludes that Canada needs “to make it clear as vodka that it is impossible to drink regularly, even in a way officially deemed moderate, and not risk serious health problems.”
But what is moderate?
Canada currently has lax guidelines: for women, no more than 10 drinks a week or two a day; for men no more than 15 a week or three a day. By comparison, in 2016 the U.K. lowered its recommendations by 33% for men, down to about five pints of beer a week. The U.K. uses a different measuring system, but two British units are close to one American drink.
Canada’s guidelines are set to change this fall, and a key player in that change is Sherk, who now serves on the decisive committee. Sherk has previously argued for a limit of roughly one drink a day.
The new guidelines, he told me, will define “zones of risk,” rather than imply a safe level. U.S. guidelines, meanwhile, remain stuck. Issued by the Department of Agriculture, they are the result of political jockeying by stakeholders, including the alcohol industry.
U.S. guidelines were revised in 2020 but remained at two drinks per day for men and one for women. This surprised many scientists, as the scientific advisory panel had recommended that guidelines fall to one drink per day for both men and women.
However, the new guidelines do strike another blow at the J-curve, stating that “even drinking within the recommended limits may increase the overall risk of death from various causes, such as from several types of cancer and some forms of cardiovascular disease. Alcohol has been found to increase risk for cancer, and for some types of cancer, the risk increases even at low levels of alcohol consumption (less than one drink in a day).”
During the recent pandemic, we became accustomed to the conclusory phrase, “The science says. …” Scientists and journalists often present science as a monolith of inevitable progress and rationality. It’s a comfortable myth, for consumers and experts. But “the science” is often messy by design.
As noted earlier, 2018 was an embarrassing year for the National Institutes of Health. In September of that year, Michael Siegel at Boston University called out the already-battered NIH to retract and apologize for implying that there is a safe level of alcohol use. At the time, the NIH website stated that “drinking too much alcohol can increase your risk of developing certain cancers” (emphasis added). Siegel’s challenge was subtle but meaningful: The phrase “too much” implied, he argued, if not a benefit, at least a no-risk point on the curve for alcohol-linked cancer.
Four years later, that same NIH web page reads differently: “There is a strong scientific consensus that alcohol drinking can cause several types of cancer. … Even those who have no more than one drink per day … have a modestly increased risk of some cancers.”
Thomas Kuhn famously observed that scientific revolutions are often personalized battles — with egos, turf and careers at stake — fought over many years until one theory is driven off and another enthroned. Often these battles are in the trenches contesting seemingly arcane language. Move by move, one side gets shoved across the map.
This revolution, the battle of the J-curve, is not over, but the insurgents now have the momentum. It will be interesting to see how much longer Mayo Clinic and Harvard School of Public Health hold out.