To compare CATIE, a randomized double blind study, and SOHO, a 3-year prospective non-randomized observational European study of outpatients with schizophrenia, on the Number Needed to Treat (NNT) for all-cause medication discontinuation. NNTs place data into a clinically meaningful context - the number of patients needed to be treated with one antipsychotic instead of another to prevent one negative outcome, defined here as one additional medication discontinuation for any cause.

Rate of medication discontinuation for any cause during the 18 months post initiation was calculated for patients newly initiated on olanzapine (N=4247), risperidone (N=1549), quetiapine (N=583), amisulpride (N=256), clozapine (N=274), oral typicals (N=471) or depot typicals (N=348). Cox models were employed to adjust for treatment group differences at baseline. NNTs with their 95% confidence intervals were calculated and compared with published NNTs for CATIE (Phase 1).

The NNTs for all-cause discontinuation of olanzapine vs. each studied atypical antipsychotic during the 18 month following medication initiation in SOHO were comparable to CATIE: 4.3(95% CI: 3.6–5.3) for olanzapine vs. quetiapine (5.5 in CATIE); 16.1(11.0–28.1) for olanzapine vs. risperidone (10.1 in CATIE); 6.9(5.2–10.1) for olanzapine vs. oral typicals (9.0 in CATIE for olanzapine vs. perphenazine).

The NNTs for all-cause medication discontinuation based on CATIE appeared comparable to NNTs based on SOHO. The NNTs for olanzapine therapy were consistently better when compared to each studied atypical antipsychotic (except clozapine) and when compared to typical antipsychotics. Results should be interpreted conservatively, due to the observational design of SOHO.