Wednesday, March 05, 2008

Nemeroff Confirms Kirsch: SSRIs Offer Little Benefit

This post will discuss how the latest meta-analysis claiming to show public health benefits for Effexor actually also showed that antidepressants aren't up to snuff. Part 1 detailed how the study authors found a very small advantage for Effexor over SSRIs, which they then suggested meant that Effexor offered significant benefits for public health over SSRIs. Ghostwriters, company statisticians, questions about transparency, etc. Even the journal editor jumped on board. All the usual goodies.

Bad News for SSRIs: But now, on to part deux. Remember that the authors used a Hamilton Depression Rating Scale of 7 or less as indicative of remission, which was the one and only outcome measure of import in their analysis. In their database of studies analyzed in the meta-analysis, there were nine studies that had an Effexor group, an SSRI group, and a placebo group. In these studies, there was a 5.5% difference in remission rates for SSRIs versus placebo. Read it again: there was a 5.5% difference in remission rates for SSRIs versus placebo. You should be shaking your head, perhaps cursing under your breath or even aloud. Using the number needed to treat statistic that the authors used in their analysis of Effexor versus SSRIs, that means you would have to treat 18 people with SSRI instead of a placebo to get one additional remission that you would not get if all 18 had received a placebo. Damn -- that is pathetic! In these same nine trials, the difference between Effexor and SSRIs was 13%, for a number needed to treat of 8. One might conclude that Effexor was more than twice as effective as SSRIs based on these figures, but one would be wrong. Please see my prior post for why depression remission should absolutely not be used as the only judgment of a drug's efficacy. Granted, the numbers for SSRIs were based on nine trials, which limits the generalizability of the findings, but the findings sure fit well with the Kirsch series of meta-analyses that found only a small difference for SSRIs over placebo in all but the most severe cases.

If you told most people that you would have to treat 18 depressed patients with a SSRI rather than a placebo to get one additional remission in depressive symptoms, you'd get laughed out of the room, but that is exactly what Nemeroff et al found. Do the authors conclude with: "The findings confirm earlier work by Kirsch and colleagues showing that the benefits of SSRIs over placebo are quite modest"? Not exactly. Here is their interpretation:
To achieve one remission more than with placebo, 8 patients would need to be treated with venlafaxine (NNT = 8) compared with 18 patients who would need to be treated with an SSRI (NNT = 18). From this perspective, the magnitude of the advantage of SSRIs versus placebo in the placebo-controlled dataset (NNT=18) is similar to the advantage of venlafaxine relative to SSRIs in the combined data set (NNT = 17).
This is right after the authors wrote about how a NNT of 17 was possibly important to public health (see part 1), which was about the time I fell out of my chair laughing. A more plausible interpretation is that SSRIs yielded very little benefit over placebo and that Effexor, in turn, yielded very little benefit (in fact, a statistically significant benefit over only Prozac) over SSRIs. But that sort of interpretation does not lead to good marketing copy or press releases that tout the benefits of medication well beyond what is reasonable. What if the press releases for this study read: "Nemeroff confirms findings of Kirsch: Antidepressants offer very little benefit over placebo." That would have been refreshing.

Sidebar: Here is my standard statement about antidepressants -- they work. Huh? Yeah, the average person (surely not everyone) on an antidepressant improves by a notable amount. The problem is that the vast majority (about 80%) of such improvement is due to the placebo effect and/or the depression simply getting better over time. Give someone a pill and that person will likely show some improvement, but nearly all of the improvement is due to something other than the drug. If most improvement is due to the placebo effect, couldn't we usually get such improvement using psychotherapy, exercise, or something else, which might avoid some drug-induced side effects? Moving on...

Key Opinion Leaders: But notice how this Wyeth/Advogent authored piece featuring Charles Nemeroff as lead author (as well as Michael Thase as last author) throws down a major spin job regarding the efficacy of antidepressants. As reported previously, their measure of efficacy was quite arbitrary. It could have been supplemented with other measures, as Wyeth is in possession of such relevant data, but such analyses were not conducted. But even using their questionable measure of efficacy, antidepressants put on a poor performance. Similarly, Effexor's advantage over SSRIs was meager. Yet the authors (remember, three medical writers worked on this paper) conclude that venlafaxine offers a public health benefit over SSRIs. Maybe the authors were afraid of being sued for writing anything negative in their paper? Or perhaps they just know who is buttering their bread. It is also possible that the authors truly cannot envision the idea that SSRIs offer such a meager advantage over placebo and that Effexor yields very little (if any) benefit over SSRIs. And that is the problem. The "key opinion leaders" are all stacked on one side of the aisle -- drugs are highly effective and each new generation of medications is better than the last. So plug in the name of the next drug here, and you'll see a key opinion leader along with a team of medical writers rushing out to show physicians that the latest truly is the greatest. Since we don't really train physicians to understand clinical trials or statistics particularly well, you can also expect many physicians targeted by such marketing efforts to simply lap up unsupported claims of "public health benefit."

Hey, is there a counter-detailer in the room somewhere?


Radagast said...

Well, I took fluoxetine, and while I know that "anecdotal" patient reporting isn't given a whole lot of weight by the "scientific" community (probably because it can't be contradicted), I can confirm that the only difference between the drug and a pile of dung is that one has the benefit of putting the dung on one's roses, when one decides that one doesn't wish to ingest it.


soulful sepulcher said...

speaking of counter-detailing, what are your thoughts about Carlat's adventures?

CL Psych said...


I think counter-detailing is fantastic. Imagine if even a small fraction of the key opinion leaders who stump for various drugs were to switch to the counter-detailing side. Though I suppose they'd get disowned by Pharma so quickly that they would no longer be considered KOLs.

Counter-detailing doesn't solve all ills, but any step in the right direction is appreciated greatly.