Posted by viridis on September 7, 2002, at 3:24:04
In reply to Antidepressants and placebo...the truth, posted by ZyprexaNumbTongue on September 6, 2002, at 16:19:23
Putting aside any assumptions about biases of psychologists vs. psychiatrists etc., there are other ways that a drug which is actually very effective for a substantial proportion of patients might appear similar to placebo, even if the trials are carried out in a totally fair, controlled, unbiased way. It depends on various factors, including the proportion of patients who respond favorably to the drug; the proportion whose condition is worsened by the drug; and the magnitude of the placebo effect.
To illustrate this, I invented a completely hypothetical scenario that builds in several guesses and assumptions. Suppose that a pharmaceutical company is testing a new antidepressant (call it AD "X"), and that they use some imaginary rating scale for depression that ranges from 0 (not depressed at all) to 100 (most severe depression possible). They round up a group of test subjects who each score around 60 on this scale -- pretty seriously depressed. They divide the subjects into two equal-sized groups, one of which receives the real drug, the other of which gets an identical-appearing "sugar pill".
Now, assume that the drug is actually quite effective for 30% of patients, cutting their level of depression in half (from 60 to 30 on the scale). It does nothing for 45% of patients, and it actually increases depression level to, say, a score of 80 for 25% of the subjects. (For example, suppose that for some it doesn't help depression, but increases anxiety and/or has various other side effects that are intolerable for this 25%, probably a conservative assumption based on current estimates for many real ADs).
In the placebo group, suppose that 65% of subjects have no response either way, 25% experience some improvement due to "placebo effect", from a score of 60 to 50, and 10% experience an increase in depression score from 60 to 70 for whatever reason (progression of illness, negative placebo effect, random factors, etc.).
Using these totally cooked-up (but probably not ridiculous) numbers, the average depression score at the end of the trial for the treatment group would be 56, and that for the placebo group would be 58.5 -- less than a 3% difference. Yet, about a third of patients in the treatment group had their level of depression cut in half by the drug. Maybe this isn't fantastic, but it's a big help for this 30% of patients who are strong responders, especially if some don't respond to other ADs. Depending on sample sizes, this difference between treatment and placebo groups might well be statistically non-significant. And of course, one could easily tweak the hypothetical numbers to make them come out exactly the same.
The point is not that these numbers represent any real drug trial, or that statistics are inherently misleading. And, the model I've used is obviously an oversimplification -- just a thought experiment. Furthermore, I'm not trying to justify any particular pharmaceutical company's research or marketing strategy. I'm just pointing out the complexities that go into assessing the effectiveness of a given medication, and that a drug which could be of real benefit to a lot of people could appear no better than placebo even in a properly-conducted experiment. Of course, given numbers like the ones above, you probably wouldn't want to start with AD X. But one implication is that it may take some trial and error to find the best medication, and for treatment-resistant cases especially, it may be worth trying even drugs that look iffy on paper if others have failed.
I do suspect that, as ZNT suggested, the differences between treatment and placebo groups might become more pronounced if AD X was tried on the most severely depressed patients. But still, it would depend on which of the factors discussed above were operating, how strongly and at what frequency.
poster:viridis
thread:119067
URL: http://www.dr-bob.org/babble/20020906/msgs/119151.html