Dynamic prediction of treatment response in late-life depression

Abstract

Objective: To identify actionable predictors of remission to antidepressant pharmacotherapy in depressed older adults and to use signal detection theory to develop decision trees to guide clinical decision making.

Method: We treated 277 participants with current major depression using open-label venlafaxine XR (up to 300 mg/day) for 12 weeks, in an NIMH-sponsored randomized, placebo-controlled augmentation trial of adjunctive aripiprazole. Multiple logistic regression and signal detection approaches identified predictors of remission in both completer and intent-to-treat samples.

Results: Higher baseline depressive symptom severity (odds ratio [OR]: 0.86, 95% confidence interval [CI]: 0.80-0.93; p <0.001), smaller symptom improvement during the first two weeks of treatment (OR: 0.96, 95% CI: 0.94-0.97; p <0.001), male sex (OR: 0.41 95% CI: 0.18-0.93; p = 0.03), duration of current episode ≥2 years (OR: 0.26, 95% CI: 0.12-0.57; p <0.001) and adequate past depression treatment (ATHF ≥3) (OR: 0.34, 95% CI: 0.16-0.74; p = 0.006) predicted lower probability of remission in the completer sample. Subjects with Montgomery Asberg (MADRS) decreasing by greater than 27% in the first 2 weeks and with baseline MADRS scores of less than 27 (percentile rank = 51) had the best chance of remission (89%). Subjects with small symptom decrease in the first 2 weeks with adequate prior treatment and younger than 75 years old had the lowest chance of remission (16%).

Conclusion: Our results suggest the clinical utility of measuring pre-treatment illness severity and change during the first 2 weeks of treatment in predicting remission of late-life major depression.

Trial registration: ClinicalTrials.gov NCT00892047.

Keywords: Late-life depression; decision trees; dynamic remission prediction.