Does early non-improvement predict treatment failure in pharmacotherapy for obsessive-compulsive disorder? A diagnostic test accuracy meta-analysis with individual participant data.

Background: In the treatment of obsessive-compulsive disorder (OCD) with antidepressant medication, the earliest reliable indication of treatment failure remains uncertain. We investigated if non-improvement following 4 weeks of treatment predicts nonresponse at the end of the trial.

Methods: We conducted a random-effects bivariate diagnostic accuracy study using individual patient data from industry-sponsored short-term trials of adults with OCD receiving selective serotonin reuptake inhibitors or clomipramine, submitted for marketing approval. The primary outcome was accuracy of non-improvement (<25% reduction on the Yale-Brown Obsessive Compulsive Scale [YBOCS] after 4 weeks) in predicting nonresponse (<35% YBOCS reduction at trial endpoint [10-13 weeks]). Secondary outcomes were accuracy of non-improvement after 6 weeks, nonresponse after 8 weeks, and inclusion of Clinical Global Impression Scale - Improvement in definitions of improvement and response. We performed meta-regressions for sex, age, severity, trial duration, dosing regimen, and compound.

Results: In 11 studies totaling 1,753 patients, non-improvement at week 4 predicted subsequent nonresponse (positive predictive value, PPV) in 86% of cases (95% confidence interval [CI] = 83-88%). Sensitivity was 78%, specificity was 70%, and the negative predictive value was 60%. Secondary outcomes showed similar PPV after 6 weeks and a PPV of 93% for nonresponse after 8 weeks. Predictive accuracy was significantly higher in men relative to women (β = -0.64, 95% CI = -1.12 to -0.16, p = 0.0089).

Conclusions: Patients with OCD who do not improve after 4 weeks of antidepressants will likely not respond to short-term treatment. Thus, a change in strategy should be considered after 4 weeks without treatment benefits.