Objective

Antipsychotic drug efficacy may have decreased over the decades. We, therefore, present a comprehensive meta-analysis of all placebo-controlled trials in acute schizophrenia, we investigate which trial characteristics have changed over the years and which ones are moderators of drug-placebo efficacy differences.

Method

We searched multiple electronic databases, ClinicalTrials.gov and the FDA website. The outcomes were overall efficacy (primary outcome), responder rates, drop-out rates, positive, negative and depressive symptoms, quality of life, functioning, and major side-effects. Multiple potential moderators of overall efficacy were analyzed by meta-regression.

Results

167 double-blind randomized controlled trials with 28102 participants were included. The standardized mean difference (SMD) for overall efficacy was 0.47 (95% CrI 0.42,0.51), but accounting for small trial effects/publication bias reduced the SMD to 0.38. 51% in the antipsychotic group versus 30% in placebo had at least a ‘minimal’ response, and 23% versus 14% had a ‘good’ response. Positive symptoms improved more than negative symptoms and depression. There were also small-to medium sized improvements in quality of life and functioning (SMDs 0.35 and 0.34). In the analysis of response predictors, 17 of 26 trial characteristics analyzed changed over the decades. But in a multivariable meta-regression, only industry-sponsorship and increasing placebo response were significant moderators of effect sizes. Importantly, drug response remained stable over time.

Conclusions

Approximately two times more patients improved under antipsychotics compared to placebo, but only a minority experienced a good response. Industry sponsorship reduced, rather than increased effect sizes. The decrease of effect sizes over the years was caused by increasing placebo response, not by decreasing drug response. Meta-analyses need to take this confounder into account. In drug development, somewhat smaller sample sizes but better selected patients may overcome a possible vicious circle of increasing sample sizes, more variability and smaller effect sizes.