A heated debate over how well psychiatric medications actually work has led some authorities in our field to suggest that psychiatry is currently experiencing a “crisis of confidence” (Nierenberg AA et al, J Clin Psychiatry 2011;72(1):27–33). This debate has even begun to spill over into the mainstream media. Most recently, a 60 Minutes exposé suggested that antidepressants may not even work for the majority of depressed patients for whom they’re prescribed. Although you may be tempted to shrug off such criticism as another fringe assault on psychiatry, be warned that you do so at your own risk: the critics make a compelling case. After all, if they didn’t, our leaders wouldn’t be calling this a “crisis” right now. So what exactly is going on?

Irving Kirsch and the Placebo Response

The 60 Minutes piece focused primarily on the work of Irving Kirsch, a psychologist currently at Harvard Medical School who has been studying the placebo response for more than three decades. In it, Kirsch asserted that depressed patients get better primarily because of the placebo effect, not because of the antidepressant medication itself. In response, the psychiatric community offered some rather weak rebuttals. Michael Thase, a prominent psychiatric researcher from the University of Pennsylvania, said that Kirsch “confused the results of studies with what goes on in practice” (ie, ‘let’s not let science get in the way of what we know to be true’). He concluded by confessing: “I wish our antidepressants were stronger.”

Tom Laughren, director of the FDA’s division of psychiatry products, admitted to 60 Minutes that “the difference between improvement in drug and placebo is rather small.” So small, in fact, that Great Britain has begun to train a new legion of psychotherapists after concluding that antidepressants don’t work for most cases of mild to moderate depression.

Kirsch first dropped the placebo bombshell on psychiatry in 1998, in an online journal article entitled “Listening to Prozac, but Hearing Placebo.” He and his co-author analyzed the results of 19 published trials involving 2,318 patients, and found that the vast majority (75%) of benefit from antidepressant medication could be attributed either to spontaneous remission (the natural ups and downs of depression) or to the placebo response. They concluded at the time that the benefits of antidepressant medications were probably overstated. In 2008, Kirsch published a second meta-analysis that now included unpublished data obtained from pharmaceutical companies using the Freedom of Information Act. Combining the results of 35 published and unpublished trials (ie, trials where results usually didn’t look so good), Kirsch reached a more damning conclusion: antidepressants offer almost no discernible benefit over placebo—other than in cases of severe depression (Kirsch I et al, PLoS Med 2008;5(2):e45).

Kirsch’s two main points can be summarized as follows: (1) The majority (perhaps 50% to 75%) of benefit that subjects report from antidepressants is attributable to placebo or passage of time, and far less to a true drug effect; and (2) When the results of unpublished trials are taken into account, this difference falls well below the consensus opinion of what is generally accepted as clinically meaningful (except for the most severely depressed patients).

These findings have been corroborated by others (see for example, Fournier JC et al, JAMA 2010;303(1):47–53), and Kirsch’s conclusions seem to have been accepted by leaders in the field. In a recent conference organized by Harvard’s Andrew Nierenberg in response to Kirsch’s work, a panel of experts concluded that clinicians should “expect greater antidepressant efficacy in patients with more severe depression than in patients with mild or moderate depression,” and recommended psychotherapy for patients with mild or moderate depression (Nierenberg AA et al, J Clin Psychiatry 2011;72(1):27–33). Although some very recent meta-analyses do show modest benefit in less severe depression (for example, Gibbons RD et al, Arch Gen Psychiatry 2012;online ahead of print) the evidence supporting the short-term efficacy of antidepressant medications is unfortunately meager at best.

Reconciling Clinical Data with Everyday Experience

Even though clinical trials show only modest support for the efficacy of antidepressants, we’ve all seen how well they work in clinical practice. How do we account for this apparent discrepancy? Critics would say it’s because we fail to appreciate how powerful the placebo effect is, and that most depressions usually remit on their own even without treatment. On both these counts they are almost undoubtedly correct—for example, in one study we found that as many as 85% of depressed patients spontaneously recover within one year, even without medications (Posternak MA et al, J Nerv Ment Dis 2006;194(5);324–329). But are there any other explanations?

One possibility is that the methodology used in antidepressant trials is flawed. In an article depicting the history of clinical trial methodology, my colleagues and I showed how clinical trial methods have remained almost completely unchanged over the past 50 years (Posternak MA et al, Am J Psychiatry 2002;159(2):191–200). Despite widespread recognition of the numerous flaws in these studies, why would pharmaceutical companies continue to use inefficient trial methods? The reason is actually quite simple: the primary goal of drug-funded studies is less about advancing science than about bringing an antidepressant to market. And the surest way to do that is for companies to do what the FDA requires of them—and no more. Changing trial design would mean going against a precedent that has arguably worked adequately for its intended purposes for decades.

Here’s one example of a 1960s-era trial design flaw: subjects are given an antidepressant or a placebo for six weeks. They are then asked to come back every week to check in so that researchers can chart their progress. A typical drug trial, right? But that’s a lot of contact with empathic listeners—and all that contact might help everyone in the trial get better, thereby increasing the placebo response. When we studied the impact of these frequent visits, by comparing these trials to others in which visits weren’t so frequent, we found that they accounted for as much as 40% of the placebo response (Posternak MA and Zimmerman M, Br J Psychiatry 2007;190:287–292). These trial design flaws would make it a lot harder for an effective medication to look much better than placebo. As a result, these studies may actually underestimate how effective antidepressants are.

What Should We Do with This Information?

A few takeaway points can be summarized:

1. The efficacy of antidepressants in the short term is meager, but treatment involves more than just the short term. Antidepressants almost always perform only slightly better than placebo in short-term trials—just well enough to bring them to market. But there’s a lot of other stuff out there showing that antidepressants work (eg, long-term discontinuation studies that show higher rates of relapse even months after patients have been switched to placebo).


2. The way we study antidepressants over the years may not be very good. The continued reliance on the outdated Hamilton Depression Rating Scale (Zimmerman M et al, J Clin Psychopharmacol 2005;25(2):105–110); the financial pressure to rush subjects through studies; the use of research assistants rather than experienced clinicians; the frequent personal contact with subjects; and the lack of inter-rater reliability assessments are just a few (of many) design flaws that likely make antidepressants look worse than they actually are.


3. Antidepressant efficacy trials have been conducted almost exclusively by the pharmaceutical industry. With few exceptions, drug trials are designed to get FDA approval. Because antidepressants only need to show slight superiority to placebo, it should not come as a surprise that that’s all that has been accomplished so far.


4. The jury is still out. Kirsch has provided a wake up call to the field, but bear in mind that his work only focuses on a tiny sliver of a very large pie.

As a result, where we go from here is anything but clear. It is disconcerting to admit that 50 years since their introduction, we still don’t even know how well antidepressants work.

TCPR Verdict: 
Antidepressants may not be as wonderful as we’ve all come to believe, but might be better than the research has suggested. As of now, it seems premature to dramatically change the way we use them. It may be valuable, however, to keep in mind that most cases of depression do resolve on their own—and a little empathic support can go a long way.