A Meta-analysis of Factors Impacting Detection of Antidepressant Efficacy in Clinical Trials: The Importance of Academic Sites

• Authors: Boadie W Dunlop, Emory University; Michael E Thase, University of Pennsylvania; Chuan-Chuan Wun, Pfizer Inc; Rana Fayyad, Pfizer Inc; Christine J Guico-Pabia, Pfizer Inc; Jeff Musgnung, Pfizer Inc; Philip T Ninan, (Formerly of) Pfizer Inc
• Journal Title: Neuropsychopharmacology
• Volume: Volume 37, Number 13
• Publisher: Nature Publishing Group | 2012-12, Pages 2830-2836
• Type of Work: Article | Post-print: After Peer Review
• Publisher DOI: http://doi.org/10.1038/npp.2012.153
• Subjects: Psychology, Behavioral; Health Sciences, Pharmacology
• Keywords: Academic Medical Centers; Antidepressive Agents; Clinical Pharmacology/Clinical Trials; Depression; Unipolar/Bipolar; Drug Discovery/Development; Ethics; Mood/Anxiety/Stress Disorders; Placebo; Signal Detection; Venlafaxine; placebo; academic medical centers; signal detection; antidepressants; venlafaxine; desvenlafaxine

Abstract

Variability in placebo response greatly complicates the design, conduct, and interpretation of clinical trials of antidepressant medications. To identify factors that impact detection of antidepressant–placebo differences, we conducted a meta-analysis of all relevant phase II–IV clinical trials for major depressive disorder conducted by the manufacturer of venlafaxine and desvenlafaxine completed by March 2011. We examined 15 factors potentially relevant to trial outcomes, using the standardized mean difference on the Hamilton Rating Scale for Depression (HAM-D17) score as the primary outcome. Thirty trials comprising 8933 patients were included. In univariate analyses, antidepressant efficacy (ie, drug vs placebo difference) was predicted most strongly (β=3.74, p=0.0002) by the proportion of patients in the trial enrolled from academic sites. Other factors predicting larger drug–placebo differences included lower participant completion rate, fewer post-baseline study visits, earlier year of study, and study drug (venlafaxine>desvenlafaxine). In multivariate meta-regression modeling, only the proportion of patients from academic sites maintained statistical significance as a predictor of drug–placebo separation for both HAM-D17 continuous score change (β=2.24, p=0.034) and response rate (β=2.26, p=0.035). Including a higher proportion of academic sites may increase the ability to detect differences between active drug and placebo in clinical trials of major depressive disorder.