By Tamar Haspel
As a moderate drinker, I desperately want to believe that moderate drinking is good for you.
January, when a lot of us (including me) cut back on alcohol, seems like a good time to see if it actually is.
For a while there, it was looking good.
The idea that some alcohol improved some health outcomes was widely accepted for quite some time. A meta-analysis published in the BMJ a decade ago is typical: "Light to moderate alcohol consumption is associated with a reduced risk of multiple cardiovascular outcomes," it concludes.
The association is remarkably consistent across studies and populations. People who drink moderately (the definition varies; it's usually one to two drinks per day) have better health outcomes than either people who don't drink at all or who drink a lot. Graph that, and the curve looks like a J.
Lately, though, scientists and researchers are beginning to doubt. So am I, for two reasons.
First, there are questions about the reliability of all that evidence, most of which is observational.
Second, there's a new way to study this, and it's not looking good.
Let's look at the observational evidence first.
Because alcohol is addictive, and undoubtedly unhealthy in large quantities, running trials in which you give it to people is problematic.
There have been some trials, but most of the evidence we have comes from asking people how much they drink and then seeing what happens to them.
But in 2013, a Norwegian psychologist named Hans Olav Fekjaer, who studies addiction and substance abuse, widened the lens.
He published an overview of many things whose risks were said to be decreased by moderate drinking, and it was a very long list.
Lots of the items on his list jibe with the ideas at the time: cardiovascular health, cancer mortality, stroke. But then there's hearing loss. Asthma. Gallstones.
The common cold, really?
Although I have a friend who asserts that tequila does indeed ward off colds, the rest of us probably balk at the idea that a glass of wine with dinner is protective.
This is a clue that maybe those associations aren't about the alcohol.
They're about the lifestyle of people who have a drink or two.
Those of you who come here often know that I am sceptical of observational research, so I wanted to hear from someone to make the best case in its favour. Deirdre Tobias, assistant nutrition professor at Harvard University's T.H. Chan School of Public Health, is an epidemiologist and researcher I admire because she both calls out epidemiological overreach and is trying to improve methods from the inside.
She's also willing to engage on these issues in public, civilly and constructively, even with irritating journalists who disagree with her.
If you're on Twitter, she's a great follow (@deirdre_tobias).
The first thing she pointed out was the sheer volume of the data.
There are many studies around the world that maintain data on large groups of people, and "there have been hundreds," Tobias told me, that look at alcohol consumption.
"We have the statistical power to tease out a relationship, if there is one." And we consistently see lower heart disease risk with moderate drinking.
She also points out that, if you look at the trials that give people a moderate amount of alcohol, results there are consistent, too: HDL cholesterol goes up, and triglycerides go down.
She is, however, quick to add that HDL, the "good" cholesterol, isn't considered as protective as it once was, so "you have to take this with a grain of salt."
Then I asked her about the second reason to doubt the protective effect of moderate drinking: We have a whole new way to study this. It involves genes.
Imagine a gene for liking alcohol. One gene. It affects how much you drink, but only that. It's randomly distributed in the population, so people who have it are exactly like people who don't in every other way.
If there were such a gene, all you have to do is look at the risk of heart disease in people who have it versus people who don't. Voila! A perfect way to suss out what drinking does to you.
We don't have that gene, but we have a couple of decent approximations in the form of gene variants that make drinking uncomfortable in some way (by governing the way our bodies metabolise alcohol).
We know that people who have these variants drink less, and we can study the difference between those people and those without the variants.
This technique is called Mendelian randomization (MR), named for Gregor Mendel, whose famous pea plant experiments probably made your seventh-grade biology curriculum, and there have been several studies of the connection between alcohol and heart disease using it.
Michael Holmes, a researcher at the University of Oxford, has co-authored many MR papers, including one from 2019, done in China, that tries to answer the question of whether moderate drinking is protective.
The answer is no. Drinking 280 grams of alcohol per week (about 20 average-size drinks in the U.S., on the high end of moderate) is associated with a 38% increase in stroke risk, although only a 5% increase in heart disease risk.
In fact, Holmes wrote to me, the body of MR work fails to confirm the J-shaped curve consistently found in observational research.
The China study is particularly compelling for a couple reasons.
First, when you find a variant that's associated with reduced drinking, you have to wonder what else that variant might do to change your risk.
So the people with the variant might drink less, but they also might have something going on metabolically that would reduce risk.
Luckily, in the areas of China the study dealt with, the fact that women drink very little, no matter their genes, offered a kind of control group. In women, the variant made no difference to disease risk, which gives us a pretty good idea that, in men, it was the drinking.
Second, if the variant in question is overrepresented in particular ethnic groups, you would have to wonder whether the culture, diet or lifestyle of that group is responsible.
One of the variants in question is common in Ashkenazi Jews, for example, who may be different from other groups in other ways.
(Although I'll note that the variant in question didn't stop my Ashkenazi ancestors from brewing slivovitz in the bathtub.) Doing the study in a relatively homogenous population reduces that kind of confounding.
MR is not perfect. Tobias notes that people who drink more sometimes also smoke more - just because they're drinking more, so smoking could be attributable to the gene in question.
In the China study, smoking was no different in the groups with different variants, but we can't just assume that the variants in question affect only alcohol.
Tobias is unpersuaded, but I am sorry to report that I am. I find the MR research conclusion compelling, and that I no longer believe that moderate drinking is good for me.
And there's even some more recent observational research that agrees; a study of global alcohol consumption, published in 2018, concluded that "the safest level of drinking is none."
I think the observational data steered us wrong for so long because moderate drinking is a marker for the ability to be moderate.
Although researchers try to correct for things like diet quality, exercise and smoking, a prudent, moderate lifestyle is hard to quantify.
What if moderate drinkers are better at managing stress, spend less time in front of screens or sleep better?
I'm betting they also get fewer speeding tickets, break fewer bones and almost never fall off ladders.
But the news isn't all bad. No, drinking isn't helpful, but it's not that harmful either. The risks are small enough that we'll still be opening a bottle of wine most nights at our house.
I'll make up for it by managing my stress, spending less time in front of screens and sleeping better.
Pass the pinot noir.