The Skinny: The evidence is unreliable and it is being misread. The evidence actually suggests that even moderate alcohol consumption is bad for you. But the evidence is of the most unreliable sort.
Those of us who like to drink have no doubt taken notice of reports that up to four alcoholic drinks per day is good for overall health (women, it is suggested, should drink a little less than men). Alcohol, it appears, reduces risk of death from the most common killers of modern civilization: “Renaud reported a 30% reduction in death rates from all ailments from two to three glasses of wine a day, a 35% reduction from cardiovascular disease, and an 18% to 24% reduction from cancer.”
Here’s the problem with these reports. The health improvements from alcohol consumption cited here – and just about everywhere – take as their baseline those who consume NO alcohol. They then compare the baseline to the health outcomes of people who consume alcohol more frequently – grouping them together according to how many grams of alcohol the people generally consume in a day. This leads to the creation of a J-shaped curve – where the verticle axis is mortality and the horizontal access is quantity of alcohol consumed. It looks like this:
As you can (sort of) see, those who consume up to four drinks a day have the same overall mortality rate as those in the no-alcohol group (all the way to the left, jammed up against the vertical axis).
What you can also see is that those who consume very-little-but-some-alcohol (like 1-3 drinks per WEEK) have the lowest overall mortality by far. Which is an odd finding, suggesting there is something very wrong with the no-alcohol group. Which there probably is.
While there are some healthy people who choose to never drink, many in the no-alcohol group are undoubtably recovering alcoholics. Recovering alcoholics tend to be in very poor health. And, it is plausible, that some others in this group don’t drink precisely because they are sickly and find alcohol overwhelming.
A less problematic comparison can be made using a different baseline – very light drinkers, the 1-3 drinks per week group. This elminates people who are in recovery, and people who are just too sick to drink at all.
When light-drinkers are the baseline, the J-curve turns into a straight linear relationship – the more you drink, the higher your overall mortality for the period under study. Alcohol consumption is associated with increased risk of death from stroke, heart disease, cancer, and “external causes” (like car crashes and bar fights).
Now, since this is epidemiological evidence, we should not take it too seriously anyway. No causal relationship has been demonstrated between alcohol and increased mortality rates, and other factors could sway results – for instance, less affluent people tend to drink more, and less affluent people tend to have higher mortality rates.
But this is an odd case where the evidence – regardless of how we assess the validity of the evidence – most strongly suggests an interpretation that is the opposite of the popular interpretation.