OBJECTIVE: To demonstrate that the progressive elevation of the threshold for definition of treatment response elucidates the greater likelihood of patients with schizophrenia responding to the novel antipsychotic olanzapine (OLZ) as compared to haloperidol (HAL). METHODS: Data was analyzed post-hoc from the acute phase of a large, prospective, randomized (OLZ versus HAL, mean modal dose = 13.2 versus 11.8 mg/day, respectively), double-blind trial, conducted in 17 countries with 1996 patients who met the DSM-III-R criteria for schizophrenia, schizophreniform disorder, or schizoaffective disorder. The cumulative proportion of patients achieving a priori defined response criteria at each of three thresholds was determined. Thresholds for clinical improvement were 20% or greater, 40% or greater, and 65% or greater reduction in endpoint to baseline Brief Psychiatric Rating Scale (BPRS) total scores. At each week, chi-square tests were used to compare the proportion of OLZ-treated patients versus the proportion of HAL-treated patients that had responded to therapy at each of the three thresholds defined. Six-week response-curves were compared using log-rank tests. RESULTS: As the threshold for classifying a patient as a responder increased, the relative divergence between drug-response curves increased with the OLZ treatment group consistently attaining higher proportions of responders than the HAL treatment group. At a minimal threshold for response (?20%), 77% of OLZ versus 70% of HAL-treated patients responded by week 6 (p=.002). At a high bar threshold for response (?65%), 25.9% of OLZ versus 15.6% of HAL-treated patients responded by week 6 (p<.001). Furthermore, a separation of response rates in favor of OLZ could be seen as early as week 2. CONCLUSION: Rigorously as compared to minimally defined thresholds for response clearly differentiate the greater likelihood of patients achieving superior improvement on the novel antipsychotic OLZ as compared to HAL.