Thursday, February 28, 2008

Prozac vs the Placebo Redux

I'm no expert on clinical trials of drugs but I do have an old copy of Goodman & Gilman's The Pharmacological Basis of Therapeutics (Fifth Edition, 1975). It includes a quick overview of the development and testing of new drugs as it was done then:
What constitutes an adequate controlled clinical trial necessarily varies, depending on the drug effect being evaluated. The more important general requirements for all trials are an appropriate and sensitive method of evaluation, an adequate number of subjects, lack of bias, concurrent comparison of the new drug with a reference drug over a range of doses and appropriate statistical validation. Many clinical trials must be conducted under so-called blind conditions. In a blind experiment, the nature of the medication is concealed from the patient (single blind) or from both the patient and all persons associated with conduct and evaluation of the trial (double blind). Blind conditions are essential for trials in which subjective effects of medication are being studied; they may also be necessary in evaluation of objective drug effects if these are under voluntary control or otherwise easily biased. In addition to a ... reference drug, a new compound is often compared with inert dummy medication, to serve as a control for placebo effects and coincident effects, such as spontaneous remission of symptoms.(Original emphasis)

Over the past 24 hours the newspapers, the broadcast media and the blogosphere have run wild with reports of a new study, led by Professor Irving Kirsch, showing that in most patients, the "most popular" anti-depressants are no more effective than a placebo. A lot of opinion pieces have also spewed forth, much of that equally predictable, give or take the odd moralising rant. People are being wrong on the internet again. In Malcom Farr's case, bizarrely so.

Evidently my shrink, the estimable Doctor Ziggy Stellenstaub, was right a few weeks ago when he suggested that it was unreasonable to think that I could give up this blogging business. I'd rather not have been angered by reading the report of the study in yesterday's Age, angered more by reading the misinformed (or just plain ignorant) posturing that quickly followed on MSM blogs and then totally pissed off after I used Google, and various other techniques to do some background checking that seems to have been omitted by every journalist who has reported on this "new" study. For example, Laura Blue, published at TIME

There are plenty of studies about antidepressants. What makes this one so important — the results were front-page news across the U.K. on Tuesday — is that the researchers were able to track down comprehensive unpublished trial results from the drug makers themselves before the drugs were authorized for sale in the U.S., and include them in their review of the literature. The U.S. Food and Drug Administration (FDA) must receive records of all relevant pharmaceutical-company trials, both published and unpublished, before it will approve a drug. Under the Freedom of Information Act, the researchers writing in PLoS Medicine were recently able to obtain those FDA records of industry-sponsored clinical trials. They yield data, they believe, that lets them avoid a bias that often plagues reviews of previous research: the tendency for conclusive positive results to be published, sometimes more than once, and thus over-represented, while mediocre results can be ignored or even swept under the rug.
Important? Unlikely. Although Blue's article suggests the exciting possibility that the lid has finally been blown off a scandalous cover-up by Big Pharma, Professor Kirsch's new study is not the first of its kind. The University of Connecticut web-site has this 1998 report:

The effectiveness of antidepressants is mainly in the placebo effect of treatment, not in the medication itself, according to a new study by Irving Kirsch.

Seventy-five percent of the response to medication for depression was a result of the patient being in treatment, while at the most 25 percent of the response was a true drug effect, asserts the study by Kirsch, a professor of psychology, and former UConn graduate student Guy Sapirstein.

"This means that for a typical patient, 75 percent of the benefit obtained from the active drug would also have been obtained from an inactive placebo," Kirsch says. "Whether the remaining 25 percent of the drug response is a true effect of the drug or a psychologically triggered response to side effects alone cannot yet be determined."

...

The study analyzed the possibility that antidepressants serve as active placebos, which produce side effects but do not cause any actual drug effect on the problem.

"Data from other studies indicate that most participants in studies of antidepressant medication are able to deduce whether they have been assigned to the drug condition or the placebo condition," according to the study.

"This study suggests that antidepressants might function as active placebos, in which the side effects amplify the placebo effect by convincing patients that they are receiving a potent drug," Kirsch says. "So if a patient takes a pill that causes side effects, he or she feels better because they believe they have been given an actual antidepressant and that the pill must be working."


In 2002, USA Today reported:

Antidepressants work only slightly better than dummy pills, and the Food and Drug Administration has not informed physicians of how little benefit most of these drugs offer, suggests a study to be released next week.

Through a Freedom of Information Act request, two psychologists obtained 47 studies used by the FDA for approval of the six antidepressants prescribed most widely between 1987-99.

Overall, antidepressant pills worked 18% better than placebos, a statistically significant difference, "but not meaningful for people in clinical settings," says University of Connecticut psychologist Irving Kirsch. He and co-author Thomas Moore will release their findings July 15 in Prevention and Treatment, an e-journal of the American Psychological Association.


"Whether the remaining 25 percent of the drug response is a true effect of the drug or a psychologically triggered response to side effects alone cannot yet be determined." This statement, from Kirsch, suggests that in 1998, he preferred to believe that Prozac had no therapeutic benefits whatever - you might also infer at least a hint of bias from the 1998 article's title "Listening to Prozac but Hearing Placebo: A Meta-Analysis of Antidepressant Medication". The title of the 2002 study was "The emperor's new drugs: An analysis of antidepressant medication data submitted to the U.S. Food and Drug Administration."

As I noted yesterday, the language of this "new" study indicates again that Kirsch and his collaborators would prefer to explain away what other researchers would take as an indication of therapeutic benefit in at least some patients. It's worth noting that the emphasis in these alternative explanations is always on the benefits of the placebo effect, rather than the other possibilities that my old Goodman and Gilman points to, such as spontaneous remission - that is patients just getting better on their own. So where the new study says:

Thus, the increased benefit for extremely depressed patients seems attributable to a decrease in responsiveness to placebo, rather than an increase in responsiveness to medication.
It might be more precise to say:
Thus, the increased benefit for extremely depressed patients seems attributable to a decrease in responsiveness to placebo and spontaneous remission rates, rather than an increase in responsiveness to medication.
But I could be wrong about that - my copy of Goodman and Gilman is over 30 years old and it's possible that since publication, medical researchers have discovered that there's no such thing as spontaneous remission - sick people just don't get better on their own. Ever.

So far, all we've got from the publication of this news is a replay - hopefully a brief one - of a standard public brou-ha-ha over whether depression is really a medical problem, with the usual know-nothings - the people who have never experienced the delights of suicidal ideation - getting on their high horses about how depressives would be so much better off if they'd just exercise, accept Jesus as their personal saviours or do something to release the ancient trauma of being imprisoned inside a volcano by an evil inter-galactic warlord.

The sooner it's over the better as far as I'm concerned - especially when people are writing and publishing guff like this:
What will the impact of this new research be? Is it a case of recognising that the Prozac emperor never had any clothes? Or, on the contrary, of acknowledging the power of placebo and finding new ways of working with it?
Disclosure: Gummo Trotsky takes 225mg of Effexor - a "popular anti-depressant" - daily. Some days he forgets to. Today wasn't one of them.

Update (1 Mar): More from Slate. Thanks to The Unknown Commenter.

Wednesday, February 27, 2008

Prozac vs the Placebo or teh Spin vs Science?

Thus, the increased benefit for extremely depressed patients seems attributable to a decrease in responsiveness to placebo, rather than an increase in responsiveness to medication.

Today's Age reports (link via Blogocracy) that a recent study has found that most patients diagnosed with depression would get as much benefit from a daily sugar pill as they would from a dose of Prozac, Aropax or Effexor. You can access the original report here.

The report is helpfully provided in three versions - an abstract, an editor's summary, and the full report. In all three, the authors make their opinion that the reason anti-depressants appear to be clinically effective in cases of severe depression is because severe depressives are less responsive to placebos clear, in case anyone reading only one of the first two misses this very important point.
The relationship between initial severity and antidepressant efficacy is attributable to decreased responsiveness to placebo among very severely depressed patients, rather than to increased responsiveness to medication. (abstract)

The findings also show that the effect for these [severely depressed] patients seems to be due to decreased responsiveness to placebo, rather than increased responsiveness to medication. (editor's summary)

These statements are not particularly scientific - they're pure spin. What the authors are telling us, in plain English, is that anti-depressants do work better than placebos on severely depressed patients, but they would prefer to explain it away. It's not particularly rational - let's see what happens when we make a few substitutions:

Thus, the increased benefit of analgesics for patients with chronic severe pain seems attributable to a decrease in responsiveness to placebo, rather than an increase in responsiveness to medication.

The relationship between initial severity and analgesic efficacy is attributable to decreased responsiveness to placebo among very severely depressed patients, rather than to increased responsiveness to medication.

The findings also show that the effect for these patients [with chronic severe pain] seems to be due to decreased responsiveness to placebo, rather than increased responsiveness to medication.


Starts to look a little woolly-headed, doesn't it? More like the sort of thing that you would write if you had some deep-seated conviction that the use of analgesics to relieve pain was just plain wrong than an objective assessment of the evidence.