Peter Kramer is an American psychiatrist specializing in depression, and author of the best-selling book Listening to Prozac.  In a long op-ed piece in Sunday’s New York Times “In defense of antidepressants,” Kramer argues that recent attacks in the press on the efficacy of antidepressants (particularly Marcia Angell’s scathing two-part critique of medical psychiatry in the New York Review of Books here and here), are misguided and may be inimical to the treatment of depression.  (I’ve discussed my take on Angell’s review on this website.)  Kramer’s thesis is this:

Antidepressants work — ordinarily well, on a par with other medications doctors prescribe. Yes, certain researchers have questioned their efficacy in particular areas — sometimes, I believe, on the basis of shaky data. And yet, the notion that they aren’t effective in general is influencing treatment.

I’ve been working my way through the books Angell reviewed, particularly Irving Kirsch’s The Emperor’s New Drugs: Exploding the Antidepressant Myth, to see how much of Angell’s critique was accurate (I’ve also been looking at the primary literature on antidepressants).  I don’t claim to have anywhere near Kramer or Angell’s knowledge about antidepressants, and so take this analysis of Kramer’s article in the same way I read Angell’s critique: as one seen through the eyes of a scientist who has a better-than-average knowledge of the literature on depression and antidepressants.

Nevertheless, I found Kramer’s piece, well, depressing: scattered, lacking cohesion, hard to follow, and even a bit misleading.  He’s not all gung-ho on antidepressants, but anyone who reads his editorial should also read Kirsch’s book, for Kramer makes several arguments that were already addressed by Kirsch, and neglects some of Kirsch’s important points, like the difference between “active” and “inert” placebos.

Here are what I see as Kramer’s main arguments:

• Some studies show that antidepressants work.  He quotes an article which appears to come from Lancet Neurology (Chollet et al. 2011 10:123-130) showing that fluoxetine (the generic name for Prozac) helped significantly improve motor function in French stroke patients.

I have read the original study, which was done in a double-blind trial with placebos, and involved 118 patients. All patients were also given psychotherapy.  The study did indeed show a significant drug-induced improvement in motor function over the three months of the study.  My quibbles with this analysis are threefold:

1. It did not measure long-term improvement.

2. It did not test whether or not the drug (an SSRI, or “selective serotonin reuptake inhibitor”) had any effect on mental functions, so we don’t know whether its effects—and I don’t deny it had them—involved alleviating depression, making the patients thereby more responsive to treatment, or acting more directly on motor function itself (after all, SSRIs affect neurotransmitters).  Thus, the reader might be misled when Kramer follows the description of this study with the statement, “Antidepressants are good at treating post-stroke depression and good at preventing it.”  That may be the result of other studies (he doesn’t mention any), but certainly not from the one he describes.  It’s also possible that the recovery of motor function itself (and not the drug’s effect on the brain) is what makes the patients less depressed.

3.  The placebos were probably not “active” placebos, i.e., those placebos that have side effects, but rather “inert” placebos—those with no side effects, like sugar pills.  I suspect that the placebos in this study weren’t active simply because they are not described as active.

Kirsch has made the important point that the “effects” of antidepressants in double-blind trials (which are much smaller than most people assume, often neither statistically nor clinically significant) might well stem from patients’ perception that they are indeed taking the real drug because they experience the common side effects of antidepressants. Inert placebos have no side effects.  In other words, when a patient perceives side effects, he concludes that he’s getting a drug that might be efficacious, and improves via an “enhanced placebo effect.”

It’s already well known, and nobody denies, that purely inert placebos, without side effects, themselves have a huge effect in alleviating depression—the famous “placebo effect”.  In fact, the “beneficial” effects of antidepressants are usually much smaller than the placebo effect. But when one uses “active” placebos, with side effects, the true drug effect often disappears.

Kirsch argues, then, that inert placebos without side effects are simply not the right controls in studies of depression, since perception of physiological effects enables patients to “break the blind,” and conclude that they’re in the “drug” group rather than the control. Any improvement could occur simply because patients have sussed out that they’re getting treatment.

This is not just speculation, but is supported by several pieces of evidence, including a significantly higher rate of”drug-induced” improvement in studies where all patients get drugs compared to patients who get drugs in placebo studies: the only difference between the two studies is that patients in the first type know they’re getting an antidepressant, and presumably experience side effects.  There’s also a huge correlation (0.96 for Prozac!) between the severity of side effects of an antidepressant and alleviation of depression (this is not proof of the ‘side effect’ hypothesis, but supports it).  Further, when the placebos are “active”—when they’re substances like atropine that have side effects but aren’t used to treat depression—the difference in improvement caused by antidepressant versus placebo is much smaller (Kirsch reports that only two of nine studies using active placebos showed a significant effect of the drug in alleviating depression.)  Kirsch cites other data as well; you can find it on pp. 15-21 of his book.

Kramer also cites another study which, he claims, shows that “elderly female cardiac patients who had emergency operations and were given antidepressants experienced less depression, shorter hospital stays, and fewer deaths in the hospital.”  I’ve read that study (Hata et al. Surgery Today 41: pp. 791-794), and as far as I can see, it used no placebos at all. Here’s the relevant line from the methods:

We divided the patients into two groups: group I consisted of 58 patients who had undergone surgery from January 2006 and were prophylactically prescribed 10 mg paroxetine as an SSRI from the first or second postoperative day; group II comprised 59 previous patients who had undergone surgery before December 2005 and had not been administered an SSRI after surgery.

The word “placebo” does not appear in the report, either.

If no placebos were given, not even inert ones (and we already know that inert placebos themselves improve depression far more than no drug at all, with the drug effect being substantially smaller than the placebo effect), then one can’t conclude anything from this study about the effects of the drug.  It could all be placebo effect.

It can’t be emphasized strongly enough that double-blind trials (preferably with active placebos) are far better than plain clinical observation in determining whether a drug is efficacious, for the effects of placebos on alleviating depression are strong and well documented.  If a doctor gives a patient an antidepressant and she improves, we have no idea whether she’s responding to the drug, to perceiving the drug’s side effects that makes her think it’s “working”, or simply to the fact that she’s being treated by a professional and taking a pill.  The testimony of doctors that their patients “appear to improve” when they take an antidepressant are worthless—and by “worthless” I mean they don’t say anything about whether the drug per se is alleviating depression.  It’s clear that many doctors do see their patients improve when given antidepressants. And lots of people taking antidepressants have reported improvement (many on my previous thread).  Those patients might have felt just as much better had they been taking placeobs.

Given the side effects of antidepressants, and any possible long-term effects on the brain, doctors might start treatment by giving patients placebos, which will often cause big improvement.  Unfortunately, I don’t think doctors are allowed to prescribe placebos—and you’d have to hide that fact from the patient, which could be considered unethical.

• The studies on mild depression submitted to the FDA by the drug companies were inadequate and flawed.

Kramer pinpoints several flaws: “quick and sloppy trials” (it seems to me, though, that if a drug really works, then “quick and sloppy trials” would be less likely to demonstrate a drug effect), the fact that patients may exaggerate their initial symptoms to get into a drug study, and the fact that some “patients” might not have been depressed at all, and that these of course wouldn’t show a drug effect.

Kramer might be right here, but I find this an odd criticism.  After all, the studies he’s citing were the ones used by the FDA (American Food and Drug Administration) to show that the drugs worked and to support the drugs’ approval.  If those studies were flawed, then we shouldn’t be using those drugs at all!

• Other types of studies—not the traditional “double blind” ones—also show positive effects of antidepressants.  Here’s Kramer’s description of one type of study:

One reason the F.D.A. manages to identify useful medicines is that it looks at a range of evidence. It encourages companies to submit “maintenance studies.” In these trials, researchers take patients who are doing well on medication and switch some to dummy pills. If the drugs are acting as placebos, switching should do nothing. In an analysis that looked at maintenance studies for 4,410 patients with a range of severity levels, antidepressants cut the odds of relapse by 70 percent. These results, rarely referenced in the antidepressant-as-placebo literature, hardly suggest that the usefulness of the drugs is all in patients’ heads.

Unless those placebos are active placebos, the studies are subject to the caveat that the patient’s perception of side effects, not the effects of the drug itself, is an important part of improvement. If the side effects go away because the placebo is inert, the patient may well relapse because of the absence of side effects.  Also, the withdrawal of antidepressants often causes pronounced and distressing physical symptoms which could easily make a patient more depressed.

• The news media uncritically focus on studies showing that antidepressants are ineffective.  Kramer says:

As for the news media’s uncritical embrace of debunking studies, my guess, based on regular contact with reporters, is that a number of forces are at work. Misdeeds — from hiding study results to paying off doctors — have made Big Pharma an inviting and, frankly, an appropriate target. (It’s a favorite of Dr. Angell’s.) Antidepressants have something like celebrity status; exposing them makes headlines. . .

. . . My own beliefs aside, it is dangerous for the press to hammer away at the theme that antidepressants are placebos. They’re not. To give the impression that they are is to cause needless suffering.

I’m glad Kramer agrees about the nefarious acts committed by Big Pharma with regard to antidepressants. (I’ve mentioned one of them in my earlier piece: drug companies often don’t submit the trials that don’t work to the FDA. They need submit only two “successful” trials of a drug for approval; and to get those they could conduct as many as they wanted—and keep them secret and unpublished— before you come up with the requisite two.). Too, Kirsch notes that drug companies publish the successful studies in medical journals far, far more often than they do the unsuccessful ones, giving physicians who read journals a biased view of the drugs’ efficacy.

But I think the news media glom onto this story not just because the drugs have celebrity status, but because so many Americans are taking them (antidepressants are the drugs prescribed most often in the U.S.). If you’re not taking them yourself, I bet you know someone who is.   And so any questions about their efficacy will naturally be of interest to many people.  And I do think the evidence for their efficacy is thin—certainly thinner than most Americans think.  Given the huge placebo effects involved in treating depression, the testimony of clinicians that the drugs “work” on their patients seems to me largely worthless.  Certainly most of the effects of antidepressants appear to come from the placebo effect accompanying being given a drug, so it’s not quite kosher for Kramer to assert that “antidepressants are not placebos.”

We need to know—and the drug companies and regulatory agencies have not been overly forthcoming about this—whether or not antidepressants work and, just as important, how well and how often they work.  That is what the press, and critics like Kirsch, are good for.  To say that the press’s digging into the science “causes needless suffering” is not only an unjustified threat, but a red herring—and a dangerous one itself.  It was that sort of digging that exposed the problems with the drugs and with Big Pharma.

Kramer broaches his own theory of depression, which comes from his 2003 book Against Depression:

In 2003, in “Against Depression,” I highlighted research that suggested antidepressants influence mood only indirectly. It may be that the drugs are “permissive,” removing roadblocks to self-healing. That model might predict that in truth the drugs would be more effective in severe disorders. If antidepressants act by usefully perturbing a brain that’s “stuck,” then people who retain some natural resilience would see a lesser benefit.

I haven’t read his book, but his summary above doesn’t make much sense to me.  And, at any rate, Kirsch argues (with evidence supporting him) that the effects of antidepressants aren’t really greater in more serious cases, but that the effects of the placebo controls are smaller.  That makes the drug-placebo difference (the drug’s “effect”) look larger in cases of more severe depression.

In the end, Kramer does take a more judicious stance.  He initially treats depressive patients with psychotherapy and says “I am to use drugs sparingly. They have side effects, some of them serious.”  Indeed, for the side effects of psychotherapy are nil.

I’d recommend reading Kirsch’s book, which is not a tirade but a popular presentation of evidence, with lots of citations and references. You can check the data yourself (and I’ll be continuing to do so).  If he’s right, then we almost need to go back to square one in assessing the value of antidepressants.  That won’t happen, of course, because those drugs are already approved, Big Pharma makes billions of dollars from them, and doctors see that they “work”.   But the placebo effect is powerful, and perhaps something should be done about harnessing that as well.  After all, taking placebos, at least inert ones, won’t cause you long-term damage.