Thomas A. M. Kramer, MD


Medscape General Medicine. 2003;5(4) 

In This Article


Recently, there has been a great deal of attention paid in the lay press to randomized, placebo-controlled clinical trials. There have been concerns about whether selective serotonin reuptake inhibitors (SSRIs) in general or paroxetine in particular can cause suicidal behavior or ideation, particularly in children. There have been other discussions as to whether these drugs are even effective in children. These issues have been looked at with this research methodology, and this has made the national news. Clinicians need to be conversant and understand what randomized, placebo-controlled clinical trials do and do not tell us. This is important not only to assist in clinical decision-making but also to be able to answer the inevitable questions from patients and their families. While I by no means intend to have the following be a comprehensive discussion of how to evaluate the quality of a clinical trial, here are a few things to think about when you look at data and try to decide how seriously to take them.

Clinical trials by their very nature compare things. These comparisons are between drugs and placebo or between different drugs. What is often overlooked in the discussion is exactly how they are being compared. We see how often one causes something to happen compared with the other, but we tend not to pay adequate attention to exactly what that "something" is. Frequent readers of this column may recall a previous discussion in which the distinction was made between recovery and response. Most clinical trials measure the rate of response between drugs and placebo. They usually define "response" as a 50% reduction in symptoms. That may mean that the drug actually does something, but it certainly does not tell us whether it would be useful in clinical practice, where our goal is remission, not response, and ultimately recovery. Michael Thase has recently published and presented data[1,2] in which he re-examined clinical trial data comparing different antidepressants and found that while they may not vary much in their response rate, there were some significant differences in their remission rates, specifically that venlafaxine may have higher remission rates than SSRIs. We might speculate that drugs that affect multiple neurotransmitter systems may be more likely to promote recovery than drugs that only affect, for the most part, a single neurotransmitter. Whatever the explanation for these findings might be, it is nice that somebody is analyzing clinical trial data to see if drugs are not just doing something, but doing what we want them to do.