Is Frequent Drinking Really Good for the Heart?

Classic Confusion of Correlation with Causation

by John French

re: Heart Aided by Frequent Drinking, Study Finds

posted at DrugWar.com Jan. 10, 2003

Here is another story of good science led astray.

The abstract of an article in today's NEJM  describes a study of "the *association* [my emphasis] of alcohol consumption with the risk of myocardial infarction among 38,077 male health professionals who were free of cardiovascular disease and cancer at base line."

Their conclusion appropriately said, "Among men, consumption of alcohol at least three to four days per week was inversely *associated* with the risk of myocardial infarction. Neither the type of beverage nor the proportion consumed with meals substantially altered this *association.* Men who increased their alcohol consumption by a moderate amount during follow-up had a decreased risk of myocardial infarction."
http://content.nejm.org/cgi/content/short/348/2/10

So far, so good, even though the abstract does not mention the internal validity problem of self selection.

Now, look at how the study is described by the BBC Online, at
http://news.bbc.co.uk/1/hi/health/2638399.stm

"Not only does alcohol help to ward off a heart attack, a regular daily drink is probably a good idea, researchers have found."
....
'Researcher Dr Kenneth Mukamal said: "Even relatively modest amounts of alcohol *may* be protective if consumed frequently.'
....
"The researchers examined the effect of drinking red and white wine, beer and spirits. They found all *appeared* to cut the risk of heart attack."

And so on.....

Of course the study has no basis for assigning causation, and the researchers carefully qualify everything they say to the press. The BBC also neglected to mention these male subjects were all professionals, hence, higher education, SES, etc., thus sharply restricting the population studied.

And further, no mention was made in either the abstract or the news story about the fact that the men who drank at the cited levels were self selected, and could have been predisposed to drink because of their healthy hearts! (Unlikely maybe, but possible.)

You can't make assumptions about any individual based on aggregate data. There will be members of the "three times a week" group that will die due to the effects of alcohol.

Why should we care, you might ask. Well, the problem of treating alcoholics is difficult enough without feeding them misinformation that gives them great opportunities for denial and rationalization. And of course the same goes for the legislators who are asked to fund good treatment.

Still, it looks like a good study with serious implications :-)