Background: The presence or absence of anxiety has traditionally determined choice of an antidepressant ("activating" or "sedating"); however, there is little scientific support for the construct. We conducted a meta-analysis to determine whether comorbid anxiety affected efficacy or predisposed patients to specific adverse events.
Method: Data were evaluated from 19 randomized, double-blind clinical trials comparing fluoxetine with placebo or a tricyclic antidepressant (TCA) or both in 3,183 patients with major depression (Cochran-Mantel-Haenszel, Mantel-Haenszel incidence difference, analysis of variance, and Pearson's chi-square methods). On the basis of the anxiety/somatization factor within the 21-item Hamilton Rating Scale for Depression (HAM-D21), patients were characterized as anxious (score > or = 7) or nonanxious (score < 7).
Results: Fluoxetine was significantly (p < or = .05) more effective than placebo in treating both anxious and nonanxious major depression (mean improvement in HAM-D21 total score, 11.0 versus 8.1 and 8.1 versus 5.5, respectively). Fluoxetine was also statistically significantly more effective than placebo in reducing the HAM-D21 anxiety/somatization factor score (anxious, all patients). The efficacy of fluoxetine and TCAs was comparable (all measures, all groups). Discontinuations for adverse events were statistically significantly higher with fluoxetine than placebo (anxious, 15.2% versus 3.8%; nonanxious, 13.2% versus 6.7%) and with TCAs than fluoxetine (anxious, 28.9% versus 15.5%; nonanxious, 34.1% versus 19.0%). Among events suggestive of "activation" or "sedation," incidence rates ranged from 0.0% to 32.9%; discontinuation rates were low (0.0% to 4.1%), except for somnolence with TCAs (9.4% to 13.2%).
Conclusion: Anxiety in major depression does not appear to affect response to an antidepressant. Therefore, antidepressant selection should be determined by an agent's overall risk:benefit ratio, not the presence or absence of anxiety.