Efficacy and safety of antipsychotic treatments for schizophrenia: A systematic review and network meta-analysis of randomized trials in Japan.

Abstract Background We examined the efficacy and safety of using antipsychotic medication for schizophrenia using only randomized trials of antipsychotic for schizophrenia conducted in Japan to avoid the biological and environmental heterogeneities caused by pooling data from various races and ethnicities. Methods We searched for eligible studies on Embase, PubMed, and CENTRAL. Primary outcomes were improvement in Positive and Negative Syndrome Scale total score (PANSS−T) and all-cause discontinuation. Other outcomes were improvement in PANSS subscale scores, discontinuation due to adverse events or inefficacy, and the incidence of 16 adverse events. Results We calculated mean difference or risk ratios and 95% credible intervals. We identified 34 RCTs (6798 patients; mean study duration, 9.0 ± 4.24 weeks; proportion of male patients, 53.7%; mean age, 43.3 years). Besides placebo, studies included aripiprazole, asenapine, blonanserin, blonanserin-patch, brexpiprazole, clocapramine (no PANSS data), clozapine (no PANSS data), haloperidol, lurasidone, mosapramine, olanzapine, paliperidone, perospirone, quetiapine, and risperidone. Efficacy and safety profiles differed for antipsychotics used with schizophrenia in Japanese patients. All active treatments other than haloperidol and quetiapine outperformed placebo to improve PANSS−T. Asenapine, olanzapine, paliperidone, and risperidone outperformed placebo for all-cause discontinuation. Asenapine, blonanserin, blonanserin-patch, haloperidol, lurasidone, mosapramine, olanzapine, paliperidone, and risperidone outperformed placebo to improve PANSS positive subscale scores. Aripiprazole, asenapine, blonanserin, blonanserin-patch, brexpiprazole, lurasidone, olanzapine, paliperidone, perospirone, and risperidone outperformed placebo to improve PANSS negative subscale scores. The confidence in evidence of most outcomes was low or very low. Conclusion Our results are similar to those of previous network meta-analysis involving various races and ethnicities.