Predicting antipsychotic drug response - Replication and extension to six weeks in an international olanzapine study

Objective: To use the degree of response after 2 weeks of treatment to predict non-response at 4 to 6 weeks. Method: Post-hoc re-analysis of a large multi-centered double-blind trial including 1996 patients with schizophrenia using receiver-operator curves and logistic regression analyses to predict non-response at 4 weeks and at 4-6 weeks from the percentage BPRS change at weeks 1 and 2. The primary non-response criterion was a less than 25% BPRS reduction from baseline. Results: A 0% BPRS reduction at 2 weeks predicted non-response at 4 weeks with a positive predictive value of 77.1%; and sustained non-response at weeks 4, 5 and 6 with a positive predictive value of 75.8%. In a secondary last-observation-carried forward-analysis a less stringent cutoff of ≤ 15% BPRS reduction was associated with an acceptable positive predictive value (75%), with even higher sensitivity (76%). Conclusions: Those patients who showed little to no reduction of symptoms at week 2 were unlikely to show even minimal response at weeks 4 to 6. There is increasing evidence that such patients may benefit from a change in treatment.