IMPORTANCEThe doses of antipsychotic drugs needed for relapse prevention in schizophrenia is a debated issue.OBJECTIVE To examine dose-response findings in a meta-analysis of randomized clinical trials.DATA SOURCES Studies were identified through the Cochrane Schizophrenia Group's Study-Based Register of Trials (March 9, 2020), PubMed (January 1, 2021), and previous reviews. First authors and/or pharmaceutical companies were contacted for additional information.STUDY SELECTION Two reviewers independently selected randomized clinical trials that compared fixed doses of a second-generation antipsychotic, haloperidol, or fluphenazine for relapse prevention in patients with stable schizophrenia. DATA EXTRACTION AND SYNTHESIS Using the Preferred Reporting Items for SystematicReviews and Meta-analyses guideline, all parameters in duplicate were extracted and frequentist dose-response random-effects meta-analyses were conducted.MAIN OUTCOMES AND MEASURES Study-defined relapse (primary outcome), rehospitalization, Positive and Negative Syndrome Scale or Brief Psychiatric Rating Scale total score reduction from baseline, all-cause discontinuation, and dropouts due to adverse events. RESULTSEvidence from 72 dose arms from 26 studies with 4776 participants was analyzed. The efficacy-related dose-response curves had a hyperbolic shape meaning that the probability to relapse decreased rapidly with doses of up to 5-mg/d risperidone equivalent (relative relapse risk, 0.43; 95% CI, 0.31-0.57; standardized mean difference for Positive and Negative Syndrome Scale total score reduction, −0.55; 95% CI, −0.68 to −0.41), but flattened thereafter. In contrast, dropouts due to adverse events continued to increase beyond this dose (relative risk at 5 mg/d, 1.38; 95% CI, 0.87-2.55; relative risk at 15 mg/d, 2.68; 95% CI, 1.49-4.62). In a subgroup analysis of patients in remission, a plateau was reached earlier, at approximately 2.5-mg/d risperidone equivalent. CONCLUSIONS AND RELEVANCEThe findings of this meta-analysis suggest that doses higher than approximately 5-mg/d risperidone equivalent may provide limited additional benefit for relapse prevention but more adverse events. For patients in remission or who are receiving high-potency first-generation antipsychotics, doses as low as 2.5-mg/d risperidone equivalent may be sufficient. However, caution is needed at this low dose end when further decreases of dose may be accompanied by a disproportionally higher relapse risk. Moreover, the observations are averages, and factors such as slow or rapid metabolism, age, illness stage, comorbidities, and drug-drug interactions suggest that individual patients will often need higher or lower doses.

Background Weight gain is among the most important side-effects of antipsychotics. It is, however, unclear whether it is associated with antipsychotic doses. We aimed to fill this gap with a dose-response meta-analysis. Methods We searched multiple electronic databases (last update search June 2021) for all fixed-dose studies that investigated 16 second-generation antipsychotics and haloperidol in adults with acute exacerbation of schizophrenia or with negative symptoms. We estimated the dose-response curves by conducting random-effects dose-response meta-analyses. We used the restricted cubic spline to model the dose-response relationship. The primary outcome was mean weight gain in kg from baseline to endpoint, the secondary outcome was the number of patients with clinically important weight gain. Findings Ninety-seven studies with 333 dose arms (36 326 participants) provided data for meta-analyses. Most studies were short-term with median duration of 6 weeks (range 4 to 26 weeks). In patients with acute exacerbation, amisulpride, aripiprazole, brexpiprazole, cariprazine, haloperidol, lumateperone, and lurasidone produced mild weight gain in comparison to placebo (mean difference at any dose≤1 kg), while more significant weight gain was observed by all other drugs. For most drugs, dose-response curves showed an initial dose-related increase in weight which plateaued at higher doses, while for others there was no plateau and some even had bell-shaped curves, meaning less weight gain to be associated with higher doses. Interpretation Second-generation antipsychotics do not only differ in their propensity to produce weight gain, but also in the shapes of their dose-response curves. This information is important for dosing decisions in clinical practice.

Dose–response models express the effect of different dose or exposure levels on a specific outcome. In meta-analysis, where aggregated-level data is available, dose–response evidence is synthesized using either one-stage or two-stage models in a frequentist setting. We propose a hierarchical dose–response model implemented in a Bayesian framework. We develop our model assuming normal or binomial likelihood and accounting for exposures grouped in clusters. To allow maximum flexibility, the dose–response association is modelled using restricted cubic splines. We implement these models in R using JAGS and we compare our approach to the one-stage dose–response meta-analysis model in a simulation study. We found that the Bayesian dose–response model with binomial likelihood has lower bias than the Bayesian model with normal likelihood and the frequentist one-stage model when studies have small sample size. When the true underlying shape is log–log or half-sigmoid, the performance of all models depends on choosing an appropriate location for the knots. In all other examined situations, all models perform very well and give practically identical results. We also re-analyze the data from 60 randomized controlled trials (15,984 participants) examining the efficacy (response) of various doses of serotonin-specific reuptake inhibitor (SSRI) antidepressant drugs. All models suggest that the dose–response curve increases between zero dose and 30–40 mg of fluoxetine-equivalent dose, and thereafter shows small decline. We draw the same conclusion when we take into account the fact that five different antidepressants have been studied in the included trials. We show that implementation of the hierarchical model in Bayesian framework has similar performance to, but overcomes some of the limitations of the frequentist approach and offers maximum flexibility to accommodate features of the data.