Good Only When It's Bad: Drugs for Depression

Abstract & Commentary

By Russell H. Greenfield, MD, Editor

Synopsis: This unique meta-analysis of antidepressant drug therapy for depression focused on randomized, placebo-controlled trials of at least 6 weeks' duration and included subjects with a wide range of symptomatology, from mild to very severe. Subjects with extremely severe depression responded to antidepressant medication better than to placebo, but in all other groups drug therapy was ineffective. Of note, the studies evaluated used only two different antidepressant medications, but the results suggest a gross overreliance on antidepressant medication where it may not be indicated.

Source: Fournier JC, et al. Antidepressant drug effects and depression severity. JAMA 2010;303:47-53.

The authors of this patient-level meta-analysis (or mega-analysis) identified English-language articles of randomized, placebo-controlled trials of FDA-approved antidepressants in the treatment of the full range of patients with major or minor depressive disorder. The intervention had to be of at least 6 weeks' duration, and Hamilton Rating Scale for Depression (HRSD) scores had to be determined at the beginning and end of the intervention phase. The analysis focused on data from six large-scale randomized, placebo-controlled trials that included patients with a broad range of baseline symptom severity, even those with minor depressive disorder (only one of the six studies). The HRSD defines mild depression as a score of 8-13, moderate depression as 14-18, severe depression as 19-22, and very severe depression as 3 23. Subjects in the six trials identified had HRSD scores of 10-39.

The tricyclic antidepressant imipramine was used in three trials, while the selective serotonin reuptake inhibitor paroxetine was used in the other three studies. A total of 434 subjects received medication while 284 were given placebo.

The sample was divided into three groups based on HRSD scores: mild to moderate, HRSD score £ 18 (n = 180); severe, HRSD score of 19-22 (n = 255); and very severe, HRSD score 3 23 (n = 283). Effect sizes for patients in both the mild-to-moderate (d = 0.11) and severe (d = 0.17) score ranges were below the standard description of a small effect (d = 0.20). For patients in the very severe group, the effect size was significantly greater (d = 0.47). The cutoff for a Cohen d medium effect, however, is d = 0.50.

The authors concluded that the efficacy of antidepressant medication varies as a function of depressive symptom severity. True drug effects were essentially absent in patients with mild, moderate, and even severe symptoms, whereas they were significant for people with very severe symptoms. Only for those people with the highest levels of baseline depression severity were the effects of antidepressant medication markedly superior to placebo.


One of the perks of being editor of this newsletter is the opportunity to choose the specific articles to be reviewed each month. In this instance, the current article was chosen not only because of its importance and clinical relevance, but for a personal professional reason as well. First to the clinical relevance.

Depression remains a societal plague for which antidepressant medication has long been considered the standard medical intervention. While innumerable studies have shown antidepressant medication to be superior to placebo in the treatment of depression, they focused mainly on people with severe symptomatology. The findings ascribed to those with severe depression seem to have been extrapolated to apply equally well to people with milder disease, such that efficacy was shown, but since the majority of patients seen in clinical practice have significantly milder disease, effectiveness was simply assumed. The results of this trial suggest that a high level of depressive symptomatology is required for a significant response to antidepressant medication. Keep in mind, however, that of the myriad drugs available for the treatment of depression, the studies reviewed here focused on only two, both of them older agents that have been largely supplanted by potentially more effective pharmaceuticals.

As practitioners, some of us have been guilty of a knee-jerk response to a patient's clinical depression. The data suggested, so we thought, that medication could help and so a prescription for the antidepressant of choice (and there were many to choose from) was likely offered, and perhaps some patients benefitted. Other practitioners bucked the trend and stubbornly contended that antidepressants not only didn't help most of their patients, but introduced significant side effects. Both groups were right to some extent.

This editor has been assailed by proponents of both sides of the argument. Some said my prescription pad didn't come out often enough to help, others that I shouldn't have had any confidence in antidepressants in the first place. The results of this study support what has long been suspected and acted upon by many practitioners, including myself — that most people with mild-to-moderate and even severe depression can do well with interventions other than drug therapy, such as cognitive behavioral therapy, acupuncture regular exercise, and optimized sleep, among many others. Those who seem to lack the energy to initiate such changes or whose symptomatology is very severe (not discussing the actively suicidal patient), however, might benefit from a trial of low-dose antidepressant medication to help them get moving in the right direction, with the hope of discontinuing drug therapy within 3-9 months.

I claim no specific expertise in the treatment of depression that exceeds that of any of the readers of Alternative Medicine Alert. I take this opportunity to share the results of a trial whose conclusions are clinically important, and that also support a stance I have long held: In those depressed patients whose symptoms are so severe they cannot initiate self-help, antidepressant medication may be of benefit. In almost all other instances that is likely not the case. Vindication is a perk this editor both merits and gets to enjoy all too infrequently.