TIP: If no statistically significant difference is found between two treatments, we cannot conclude that they work equally well or equally poorly. Without a placebo control group, it is not possible to say whether this particular study in this particular set of patients had the ability to detect a meaningful effect or difference. This concept is called “assay sensitivity” and is analogous to a control window in a home pregnancy test that indicates that the test is working correctly. In clinical trials, if one or more of the treatments is superior to placebo, this indicates that the “assay” is working.
However, in the STAR*D study, citalopram(Drug information on citalopram) plus bupropion was found to be better than citalopram plus buspirone(Drug information on buspirone) on several secondary outcome measures.4
TIP: Why do researchers differentiate between primary and secondary outcome measures? Because hypothesis testing is based on probability, and the more statistical tests of comparison that are done, the greater the likelihood that one test will be “significant” by chance alone. Secondary outcome measures should either be corrected for multiple comparisons (a simple way for readers to do this is to divide 0.05 by the number of comparisons and use that P value as the cutoff for statistical significance) or be considered tentative and in need of confirmation.
In the STAR*D study, the only difference that met this lower cutoff was number of dropouts because of intolerance (P < .0009) (ie, patients receiving citalopram+buspirone dropped out because of adverse effects more often than those receiving citalproam+bupropion). Notably, the proportion of patients who had a response (50% or greater reduction) on self-rated depression did not differ between the two groups. Thus, we cannot conclude with confidence from this study that the combination of an SSRI and bupropion is efficacious on the basis of either response or remission as the outcome.
A recent, small clinical trial from India randomized 60 patients with MDD who had not responded adequately to a 4-week trial of an SSRI to the addition of either bupropion or placebo.5 The study found that add-on bupropion was significantly superior to add-on placebo in reducing the severity of depression. However, a few serious problems with the study design prevent us from drawing confident conclusions.
First, the trial was conducted in a single-blind rather than a double-blind manner, which is not a minor problem. Remember that investigators are likely to have conscious and unconscious biases in favor of the treatment they are evaluating, perhaps even more so than patients. This bias not only can affect investigators’ ratings but also may be inadvertently communicated to the patients. Therefore, double-blind is not an optional feature but rather is crucial to the conduct of scientifically valid research.
Second, it would be more appropriate to set the minimum duration of the trial of the SSRI at 6 (or even 8) weeks before patients are considered to have an inadequate response to the antidepressant.
Third, the SSRI trial was received before patients entered this study. Thus, their adherence and the trajectory of improvement could not have been reliably determined. It is now standard to conduct a prospective trial of the antidepressant; all patients also receive a placebo to determine whether patients have an adequate response to the SSRI.