Verified by 
Psychiatry
Compared to What: A Second Look
Peter D. Kramer revisits a posting on the uncertain benefits of medical care.
Posted June 17, 2008
Yesterday, I discussed a concept that is arousing increasing interest in medicine, the number of patients a doctor needs to treat in order to help one person who would not get better on his or her own. In that context, I took a quick glance at psychopharmacology and psychotherapy.
The posting elicited a number of thoughtful responses. Now, I have a policy about readers’ comments. I don’t reply to them. In fact, when I signed on to contribute this blog, one of the conditions — explicit, in the contract — was that I would not be expected to involve myself in what Slate calls “the fray.” I did not want to be tempted to justify myself when I could spend the same time opening up new territory.
Besides, I know how much space and effort it takes to make a simple point clearly. Blogs are necessarily “quick and dirty.” With the best will in the world, readers are bound to misunderstand blog contributors. Also, the divide between those who mostly accept the assumptions of modern psychiatry and those who mostly do not has gotten so deep and wide that to address challenges that arise from marked attitudes of mistrust would require a constant reiteration of my own assumptions.
Those are already laid out in my books, which span a good deal of the territory under dispute. For example, in Listening to Prozac, I worry over “diagnostic bracket creep,” the tendency of medical categories to expand to embrace what treatments can influence. All the same, in Against Depression, I outline the advantages of current definitions of depressive disorders, based on what research has discovered about the harm mood disorders cause. I think that these contrasting positions are compatible. But it’s hard in a blog to conjure up the multiple viewpoints that emerge in the course of a career of writing.
That said, I do think I moved through the Number Needed to Treat (NNT) material too fast. There were too many threads, and evidently the tone was not always clear. So I will try returning to the subject — and here, I would suggest to readers who have not yet looked at the prior article and the accompanying comments that they would do well to begin there.
One of my themes is that too often psychiatry is held to different standards than the rest of medicine. If doctors use finasteride as a preventative for lethal prostate cancer, and some seem headed in that direction, they will do so on the basis of data that is phenomenally indirect. Rarely do we know whether affecting one link in a chain of causality alters the final outcome — for instance, whether lowering cholesterol via medication increases longevity. The idea that finasteride, given to asymptomatic men, might save lives is not groundless — some fact and theory lend support — but it is speculative. As the Times writer points out, the critical experiment will never be undertaken. It is too expensive, and by the time it is completed, likely the field would be interested in an entirely different intervention. Medicine is an “empirical science,” where “empirical’ means only partly guided by empirical evidence.
In the light of the broad leeway given in other specialties, I tried to play with the idea of medication or psychotherapy (in this case, for active patients) as a way of preventing death due to mental illness. Play is the operative word; I was fiddling with statistics. I didn’t march all the way to the conclusion, but I suggested that one could probably put together a set of assumptions that would, on the finsteride model, justify treatment on that basis (mortality) alone. This walk-through was pure fancy; I had tried to indicate as much with such phrases as “wild speculation” and “back-of-the-envelope guesses.” As I say, my overarching point is that while psychiatrists always want more and better, even our imperfect tools fit comfortably in the medical armamentarium.
The question of acute treatment for an episode of depression or panic anxiety is a different matter, and, as one reader points out, different considerations apply. In my main example, I tried to use modest measures of efficacy, response rates of 55 per cent for the active intervention and 35 per cent for placebo. My intention in choosing these numbers was to be utterly uncontroversial so that readers would not be distracted and we could get on with the math. I find figures like an NNT of five shocking — am I really doing little good for 80 per cent of my patients? — but, again, they match what much of medicine is forced to accept.
As for the Keller Serzone-and-CBT monograph in the New England Journal of Medicine, I agree that the results for combined treatment are unusually high. That’s why I chose them — not to be representative (and not in hopes that they would be accepted uncritically), but to demarcate an extreme. To be fair on the placebo issue, I set aside the attributed recovery rate in the study and again used the 35% figure. I wanted to say that even if we took the best outcomes on record, they would suggest that we only help half our patients.
Parenthetically, one reason I was never comfortable with the NEJM article is that I found nefazodone to be an unhelpful medication for my patients. When the drug was withdrawn from the market, I may have had to switch one patient off it. Or perhaps not. It may be that I had no one on nefazodone.
But it is interesting that comments from readers are almost always about medication. Only 52 per cent of patients who completed the study did well with cognitive-behavioral therapy (CBT) when it was given without medication; if you look at all those who began in this arm of the study, only 48 per cent had “satisfactory” outcomes. Assume a placebo response rate of 35 per cent, and you would conclude that 87 per cent of patients in psychotherapy would do as well to forgo the expertise. And as one reader suggests, the criteria for response are generous; a patient whose symptom load has halved but who still has moderate depression would qualify. The remission rate (meaning absence of depression at ten or twelve weeks) was 33 per cent for psychotherapy. So one of the most optimistic studies in the literature can be seen as debunking of CBT in the absence of medication; and medication without CBT is scarcely better.
From my earliest writing – my first book, Moments of Engagement, contains a pointed critique — I have been uncomfortable with outcome studies as they are generally conducted. Given unrepresentative patient groups, manual-driven psychotherapies, and rigid medication dosing schedules, it is surprising that the studies find as much benefit as they do. (Like most clinicians, I can’t shake the belief that the utility of what we do in the office is greater than what these clumsy studies suggest.) Still, the formal justification for mental health treatments is not out of line with what is accepted in much of medicine. Critics may be looking for more certainty than is attainable . . . But the posting as I wrote it was not about making the case for treatments. It was about the sobering effect of the metric under discussion. The most upbeat studies suggest that doctors’ powers are limited.
Afterthought: Now I understand why I asked to be exempted from responding to readers’ comments, even stimulating ones. What emerges in my writing is dense, elliptical, no clearer than what came before. I see that I will need to re-read my earlier posting on the nature of blogs. Still — thanks to all who ventured an opinion.
