ABSTRACT: To meta-analytically assess the efficacy and tolerability of nonsteroidal anti-inflammatory drugs (NSAIDs) vs placebo in schizophrenia.Searching PubMed, PsycINFO, ISI Web of Science, and the US National Institute of Mental Health clinical trials registry from database inception to December 31, 2012, we conducted a systematic review/meta-analysis of randomized placebo-controlled studies assessing the efficacy of adjunctive NSAIDs. Primary outcome was the change in Positive and Negative Syndrome Scale (PANSS) total score. Secondary outcomes included change in PANSS positive and negative subscores, all-cause discontinuation, and tolerability outcomes. Random effects, pooled, standardized mean changes (Hedges' g) and risk ratios were calculated.Across 8 studies, including 3 unpublished reports (n = 774), the mean effect size for PANSS total score was -0.236 (95% CI: -0.484 to 0.012, P = .063, I(2) = 60.6%), showing only trend-level superiority for NSAIDs over placebo. The mean effect sizes for the PANSS positive and negative scores were -0.189 (95% CI: -0.373 to -0.005, P = .044) and -0.026 (95% CI: -0.169 to 0.117, P = .72), respectively. The relative risk for all-cause discontinuation was 1.13 (95% CI: 0.794 to 1.599, P = .503). Significant superiority of NSAIDs over placebo regarding PANSS total scores was moderated by aspirin treatment (N = 2, P = .017), inpatient status (N = 4, P = .029), first-episode status (N = 2, P = .048), and (in meta-regression analyses) lower PANSS negative subscores (N = 6, P = .026).These results indicate that adjunctive NSAIDs for schizophrenia may not benefit patients treated with first-line antipsychotics judged by PANSS total score change. NSAIDs may have benefits for positive symptoms, but the effect was minimal/small. However, due to a limited database, further controlled studies are needed, especially in first-episode patients.