ABSTRACT
OBJECTIVE: This post hoc analysis assessed the predictive value of improvement in depressive scores at early time points for treatment outcomes at week 8 in patients with major depressive disorder treated with desvenlafaxine 50 mg/d or placebo. METHODS: Pooled data from 6 double-blind, fixed-dose studies in adult patients with major depressive disorder. Patients were randomly assigned to desvenlafaxine or placebo. Primary end point was change in 17-item Hamilton Rating Scale for Depression (HAM-D17) scores from baseline to week 8 (or last observation carried forward). Optimal thresholds of improvement (percent change from baseline HAM-D17) at weeks 2 and 3 for predicting 4 levels of treatment success (≥ 45%, ≥ 50%, and ≥ 65% decrease from baseline HAM-D17, HAM-D17 ≤ 7) at week 8 (last observation carried forward) were determined using receiver operating characteristic analysis. Odds ratios of the predictability of improvement thresholds were computed from a logistic regression model adjusting for significant baseline predictors. RESULTS: Desvenlafaxine 50 mg/d (n = 1207) had significantly greater rates of treatment success for each level of treatment success at 8 weeks compared with placebo (n = 1067). Optimal early improvement thresholds for weeks 2 (20%-30%) and 3 (28%-41%) were highly predictive of all 4 levels of treatment success after adjusting for significant baseline predictors (odds ratios, 0.951-0.960; all P < 0.0001). Negative predictive value of early improvement increased, and positive predictive value decreased, for increasingly stringent definitions of treatment success at week 8. CONCLUSIONS: Clinical observations of patients' early response to desvenlafaxine 50 mg/d may have clinical value in predicting treatment success and guiding patient management.