Postscript: Cymbalta and the FDA

A follow-up on my postings about antidepressants and the studies that pharmaceutical houses submitted to the Food and Drug Administration . . . Readers will recall that there are two issues: whether the medicines work and whether the efficacy trials had been reported correctly.

On the second point, Eli Lilly had claimed that, contrary to a critique in the New England Journal of Medicine, the studies on Lilly's product Cymbalta (duloxetine) had been published in a timely manner. In my commentary, I discussed whether the main outcome measure—effect size—had been calculated using all the results, including those from studies that were inconclusive. I could not be certain, especially as the effect size arrived at in the articles Lilly points to does not line up with the effect size in the analysis in the New England Journal. A researcher who worked on the first summary study (one Lilly cites as timely publication) has emailed me to say that, yes, the calculations did include the entire data set.

My correspondent also commented on the first point, why (in the early studies) the placebo response rates were so high that they all but drowned out any "signal" from the active drug. My impression has been that the problem was poor subject selection—people without major depression made their way into the drug trials, and these relatively healthy subjects recovered from their scattered symptoms on their own. My correspondent suggested a different explanation: The subjects got so much attention from psychiatrists, psychologists, assistants, and secretaries that the research provided social support that acted like psychotherapy. In effect, these early trials were contrasting medication plus psychotherapy with placebo plus psychotherapy, so that any benefits of the medications became hard to demonstrate. It's an interesting thought—though of course, the same might be true of many outcome studies.

One further thought: The dosing recommendations for Cymbalta seem to arise from the research in only approximate fashion. The manufacturer's suggested dosing is 40 to 60 milligrams. But in the trials, 40 milligrams looks relatively ineffective. The 60 milligram dose worked much better; and the lowest rate of the major side effects—dizziness, nausea, and somnolence—was on a once-a-day 60 milligram dose.