We psychiatrists have a problem with antidepressants; we have been so committed to giving them for so long that it is scary to think that they might be harmful. Side effects are bad enough, but suicide seems incredible to imagine. Yet the FDA came to that conclusion about 4 years ago.

Since then a slew of studies have appeared in psychiatric journals challenging the FDA view; a recent WHO study put them all together, and could be interpreted as showing that clinicians were right all along: antidepressants are safe, at least in adults (the WHO study continues to find risk in children, but less in young adults, unlike the FDA analysis, which found risks of suicide in young adults as well).

My problem is that I agree with both sides on this debate, and I would like to see if we cannot agree to think of antidepressants as harmful, but not useless.

The debate really cannot be discussed validly unless we discuss statistics, the bane of most doctors, and of the math-phobic American public. (I have a new book coming out soon on statistics for math-phobes). There are two key concepts that will help clarify this debate.

First, doctors' observations are faulty; all of our observations are faulty; this is because many factors influence what happens in the real world besides what we think is going on. The classic example is coffee causing cancer; it is highly associated with cancer, but it does not cause cancer, because smokers tend to be coffee drinkers. Smoking is a "confounding" factor, and this is called confounding bias. All of our observations are confounded.

We get rid of these confounding factors by "randomization": randomly getting treated or not ensures that all other factors are equal between groups and thus they can be ignored. Only in randomized studies can we believe the results at face value. Randomized studies are thus more valid, more likely to be true, than observational studies.

The FDA review was based on randomized studies; the WHO review on observational studies. The former is more valid than the latter, and thus it still should be believed.

This is a relatively simple matter, widely agreed upon by those who use statistics, so I do not think we can ignore it. Other arguments about the limits of the FDA analysis also fail: The studies in that analysis were not designed to assess suicide - in fact, they excluded highly suicidal people. So much the worse for antidepressants: if suicidality is seen even in those at low risk for it, then we really have a problem. Also, the claim that the FDA studies wrongly defined suicidality has already been assessed  by a reanalysis of those data assessing truly suicidal behavior, conducted carefully by researchers at Columbia University, and the suicide risk persists.

I think the link is real, though small, and not inherently due to something evil about antidepressants. The problem may be in us, rather than the drugs; it could be that we doctors are just using these drugs excessively or incorrectly. Many of these children and young adults, in my view, may have undiagnosed bipolar disorder, in which antidepressants can produce mixed manic states, highly associated with suicidality.

As always, it is not a matter of simply saying one or the other extreme is correct, but rather one of holding two opposite views in our heads at the same time.

You are reading

Mood Swings

He's abnormal; she's normal: That's (a reason) why Trump won

How the temperament criticism of Trump backfired

Not "Depression": Manic-Depression and Robin Williams

A deadly disease, whether we accept it or not