|
Although principles such as falsifiability do render them somewhat more responsive to observed data.
Take, for example, the research done on Prozac. Negative findings were buried, and the treatment effect sizes were estimated to be much larger than subsequent research showed them to be, particularly after some of the negative findings were "excavated" from the FDA files via FOIA.
From Wikipedia:
{Irving} Kirsch’s analysis of the effectiveness of antidepressants was an outgrowth of his interest in the placebo effect. His studies in this area are primarily meta-analyses, in which the results of previously conducted clinical trials are aggregated and analyzed statistically. His first meta-analysis was aimed at assessing the size of the placebo effect in the treatment of depression.<3> The results not only showed a sizable placebo effect, but also indicated that the drug effect was surprisingly small. This led Kirsch to shift his interest to evaluating the antidepressant drug effect. Kirsch’s first meta-analysis was limited to published clinical trials. The controversy surrounding this analysis led him to obtain files from the U.S. Food and Drug Administration (FDA) containing data from trials that had not been published, as well as those from published trials. Kirsch’s analyses of the FDA data showed that the difference between antidepressant drugs and placebos is not clinically significant, according to the criteria used by the National Institute for Health and Clinical Excellence (NICE), which establishes treatment guidelines for the National Health Service (NHS) in the United Kingdom.<4>
|