Placebo power contd ...
Coming clean about the margins
Around the same time as the arthroscopic knee study was published in July, University of Connecticut psychologist Irving Kirsch and co-researcher Thomas Moore released findings from a study in the e-journal Prevention and Treatment, which suggested that antidepressants work only slightly better than placebos.
Kirsch and Moore reviewed 47 studies used by the US Food and Drug Administration as grounds for the approval of six antidepressants prescribed most widely between 1987 and 1999. The researchers relied on a Freedom of Information Act request to secure the data, which showed that the antidepressants worked only 18 percent better than placebo's across the 47 studies.
While this constitutes a statistically significant difference, they said, it was "not meaningful for people in clinical settings". Kirsch told USA Today: "More than half of the 47 studies found that patients on antidepressants improved no more than those on placebos. They should have told the American public about this. The drugs have been touted as much more effective than they are."
But Janet Woodstock of the FDA Center for Drugs came back fighting: "We make sure these drugs work before we put them on the market," she said, insisting that trials were confounded by selection criteria which skewed the experimental populations towards those who would be most inclined to benefit from a placebo.
Comparing the numbers
With all this positive press, I thought, the placebo effect should be able to confidently claim that it is only a hair's breadth away from directly challenging antidepressants as effective.
But a scorching new book from Elio Frattaroli, entitled Healing the Soul in the Age of the Brain, argues that we may have been looking wrongly at the numbers all along. And in the process, the placebo may be getting far less recognition than it deserves.
Frattaroli's argument goes something like this:
Assume that the sum of all double-blind antidepressant studies shows that 70 percent of those on the antidepressant are significantly improved during the study. Assume that 35 percent of those receiving placebo pills get better. Does this constitute evidence that the placebo is a less effective treatment than the active ingredients of the antidepressant?
Assuming that all extraneous variables are controlled for, if 35 percent of the placebo group gets better, we can assume that treatment of any population with a pill purported to relieve the symptoms of depression will result in a significant improvement in around 35 percent of the group.
So, what if that pill is a real antidepressant? Does this negate the effect of the placebo condition? Statistically speaking, no. Thirty-five percent can be expected to get better simply because they received a pill they were told would make a difference. So, when 70 percent of the antidepressant group in a clinical drug trial improves, around 35 of this 70 percent can be accounted for by the placebo effect. The implications: The most commonly prescribed antidepressants are no more effective on account of their active ingredients the substances that are allegedly being tested in such a trial -- than are sugar pills.
Frattaroli's idea puzzled me. Was it fair to contend that the placebo effect was something which should be excluded from the effect observed in the experimental group of drug trials? I couldn't quite decide. Nor could Frattaroli, by the sounds of it. Nor, interestingly, could the editor of the American Journal of Psychiatry who, when Frattaroli submitted a paper outlining for debate his discovery of the number crunching error, wrote: "The present state of our knowledge indicates that the Placebo effect may be the result of a neurophysiologic response by way of neurotransmitters." Huh?
Let the people have opinions
But I thought there must be some real scientist who could fathom the hole in this placebo puzzle. And, acting on this belief, the placebo puzzle went out in a fortnightly column of South Africa's oldest newspaper, the Natal Witness, inviting its 100,000 readers to shoot it down. Only one argument came back. Alex Chernavsky, writing from Rochester New York, picked up the article on the web and wrote:
"I'm confused about the fundamental point you're trying in make in that article. In particular, I'm puzzled by these two sentences: In short, when 70% of the antidepressant group in a clinical drug trial improves, around 35% of this 70 percent can be accounted for by the placebo effect.
The implications: the most commonly prescribed antidepressants are no more effective in the treatment of depression than sugar pills.
I don't understand your reasoning here. To me, the implications are that antidepressants are twice as good as placebos, because twice as many people improve on antidepressants as compared to placebos.
That's what the common reading of such results is. However, if 35 percent of people in a control group improve on account of the placebo effect, it is fair to assume that 35 percent of any other group will improve for the same reason. If so, this would also be the case in the experimental group, who receive the antidepressant. So, half of the 70 percent who improve in the antidepressant group can be assumed to be attributable to the placebo. Which means antidepressants are responsible for 35 percent, which is equivalent to the impact of the placebo. which means antidepressants are as effective as placebo, not more effective.
Alex insisted that Frattaroli's logic was still lost on him:
I guess I still don't see the reasoning as being valid. By implication, then, if antidepressants also had a 35 percent effectiveness rate (same as placebo), would Frattaroli then claim that antidepressants were less effective than placebos, because 35 percent minus 35 percent is zero? I'm very sympathetic to the argument that antidepressants are virtually useless, but I don't find Frattaroli's calculations to be convincing. I also think that in a typical clinical trial of antidepressants, significantly more than 35 percent of the people improve on placebo.
And I replied:
Yes, statistically speaking. I wonder what difference might be made by substituting the following variables:
1. For depression, substitute a key marker of physiological development, eg. height gain or weight gain.
2. For placebo, substitute a normal, balanced diet.
3. For antidepressant, insert hormonal or steroidal supplements.
If the height (or weight, or whatever measure you choose to study) gain of the control group is 3cm in a year and that of the experimental group is 5cm, we wouldn't say that all 5cm are to be attributed to the hormonal supplements. We'd be more inclined to say that, all things being equal, 3 of the 5 cm are probably due to the nutrition supplied by the normal balanced diet and developmental processes governed by normal human development. If both groups gained 3cm, then we'd be inclined to say the supplements made no difference, i.e. had no effect.
Can you find a problem with this analogy?
I must assume Alex could find no problem, if his ensuing silence is anything to go by. But his "termination" of our discussion left me with a few nagging questions. What if there is no hole in Frattaroli's argument? Does the average psychotherapist, aware of the recent frontal attack on the efficacy of antidepressants, simply carry on as usual? If not, what changes are prompted? Does he or she continue to prescribe antidepressant medication for clients or refer to a prescribing practitioner? What about the 100,000 plus readers of the newspaper column and the thousands of others who make their way through Frattaroli's book? If they are at least thinking that there's something puzzling about all this placebo business, then I reckon the placebo might well be the next big idea in the treatment of depression.
Copyright © New Therapist