It seems as if the media are finally coming around to questioning the over-medicalization of American psychiatry and mental health services in general. I recently wrote a piece debunking Ritalin and related stimulants for treating Attention Deficit Disorder. On Sunday, 60 Minutes on CBS discussed Irving Kirsch's work showing that anti-depressants are no better than placebos for all but the most severe cases of depression.
A key part of the show dealt--without using the terms--with two statistical concepts that might have been confused in the minds of some viewers. These are significance level and effect size. Significance level refers to the likelihood that an experimental result could be due to chance--that is, that a study using different people would have shown no effect. In general, if there is less than a 5% chance that the results are a coincidence, then the finding is called significant. (That is, even when a drug has no effect, an experiment might by chance occasionally seem to produce one--so results of all experiments, positive and negative, need to be considered.) If the probability is much lower--say one in a thousand--then it is called highly significant.
This is a very different use of the word significant from everyday speech, where it means important or substantial. In statistics, a highly significant finding may be of no importance at all. Suppose that an experiment shows that a pill improves mildly depressed people's mood by 1%. If there is only a one in a million chance that that result was a coincidence, then the finding is statistically highly significant. However, it is clinically meaningless, since neither the depressed individuals nor others close to them would be able to detect a 1% improvement.
So when a drug company says that studies of the effectiveness of their antidepressant--the ones they allowed to be published--obtained "highly significant results" that doesn't necessarily tell you what you want to know. The important information is effect size. Once you know that the results are statistically significant, you need to quantify the treatment impact. For example, how much lower on a scale of depression do people score, or how many weeks less does their depression last?
There are statistical ways of combining findings from different studies; and Kirsch was making two important points. The first is that when you combine the suppressed studies with the published ones, statistical significance disappears for most people. The second is that, even when statistically significant results are obtained, only for the severely depressed is the effect size large enough to justify taking antidepressants.
What does work for depression? The same thing that makes the placebo effect work--anything that believably encourages hope and gets individuals to expect that their life will improve.
Structured therapies (e.g., cognitive and behavioral therapies) focus on reversing the downward spiral of depression by catching up on work and other tasks left undone, by recognizing and changing self-defeating thoughts and actions, and by building on strengths, interests, and individual and social assets. Studies comparing antidepressants and focused talk therapies have also shown therapy to be more effective in preventing relapses--presumably because individuals learn ways of thinking and acting that make depression less likely in the first place, and also help them to cope better when they feel down.
In addition, physical activity and exercise--probably by combating lethargy and rumination, and by increasing the oxygen supply to the brain--have also been shown to be helpful.
Prescription placebos used in research and practice
NIH, objects donated by Elaine and Arthur Shapiro
Visit my website: www.jeffersonfish.com