In 2019, a paper published in The Lancet made headlines when researchers questioned wine's protective effects against heart disease and instead found that any level of alcohol consumption increased blood pressure and the possibility of heart attack or stroke.
But two London-based professors recently took a deeper dive into those claims, and found that the Lancet study is based on "unsound methodology."
The 2019 research, conducted by a team at Oxford, enrolled over 500,000 adults from 10 different areas of China and recorded alcohol consumption and cardiovascular disease incidence over the course of 10 years. Researchers used a type of genetic epidemiological analysis, known as Mendelian randomization, to examine the data. Genetic epidemiology looks not just at who gets sick, but at genetic factors, to try and understand what role genes play and what role environmental factors play (in this case alcohol consumption).
Multiple studies have found a J-shaped relationship between alcohol and stroke risk. Non-drinkers faced a slightly higher risk of heart attack and stroke, heavy drinkers faced a much higher risk that increased with each drink, and moderate drinkers fell at the bottom of the J—they faced the lowest risk.
But the authors of the 2019 study wrote that in their genetic epidemiological analysis using Mendelian randomization, they saw a linear association. The risk of cardiovascular disease rose with each sip of alcohol the subjects consumed. The study was touted by many as evidence that no level of drinking was safe. "Claims that wine and beer have magical protective effects were not borne out,” said lead author Richard Peto, in a statement released when the research was published.
Professor Sir Nicholas Wald of University College London and professor Chris Frost of London School of Hygiene and Tropical Medicine have prior experience with methods that are parallel to genetic analysis, and were struck by the study's strong claims. "Professor Wald sought my advice about the methodological approach used in the [Lancet] paper because there are some links between the approach used by the authors and earlier work that I had carried out," Frost told Wine Spectator.
Wald and Frost dug deeper, and in their new study, published in Oxford University Press's International Journal of Epidemiology, they write that they have uncovered a fundamental weakness in the 2019 analysis that potentially concealed an underlying J-shaped curve.
Frost and Wald found that when a relationship between an outcome and exposure is non-monotonic (such as J-shaped), there is a risk of systematically distorting the true relationship of risk factors and disease. They ran a hypothetical genetic analysis with results based on the methods used in the 2019 paper and assumed non-drinkers had a 15 percent chance of stroke, while light drinkers (one glass per day) had a 10 percent chance and heavy drinkers (five glasses per day) had a 20 percent chance. That's a J shape. Frost and Wald wanted to show that it could be distorted into a straight-line relationship using the genetic analysis from the 2019 paper.
How did the Lancet team end up with a linear association? Frost offered an explanation: Suppose that people either drink no alcohol, one unit of alcohol per day or two units of alcohol per day, and that individuals of these three types have a risk of an outcome over a period of 15 percent (for those who drink nothing), 10 percent (for those who drink one unit of alcohol per day) and 35 percent (for those who drink two units of alcohol per day). If one makes a composite group whose mean consumption is one unit of alcohol per day then it is possible that the risk in this group could be as low as 10 percent (if all participants drink one unit per day) or as high as 25 percent.
The authors of the 2019 study created such composite groups as they separated participants by genotype.
"What we have shown is that when the underlying relationship between an exposure and an outcome is J-shaped, then the type of analysis carried out in the 2019 Lancet paper can potentially turn this J-shaped relationship into a relationship that is linear," Frost said. "We use simple examples to demonstrate this, so I am confident that this could have happened in the 2019 Lancet paper."
Frost says that further research into statistical methods and the Chinese data analyzed in the Lancet paper are needed. "Our work does not prove that moderate drinking is safe, only that the claims that it is harmful are based on unsound methodology."
"We think that the scientific position on alcohol consumption and vascular disease is as it was before the publication of the 2019 Lancet paper," he added.
Want to learn more about how wine can be part of a healthy lifestyle? Sign up for Wine Spectator's free Wine & Healthy Living e-mail newsletter and get the latest health news, feel-good recipes, wellness tips and more delivered straight to your inbox every other week!
