Standard versus reduced dose of antipsychotics for relapse prevention in multi-episode schizophrenia: a systematic review and meta-analysis of randomised controlled trials

Background Dose reduction of antipsychotic maintenance treatment in individuals with schizophrenia could be desirable to minimise adverse effects, but evidence for this strategy is unclear. We aimed to compare risks and benefits of reduced versus standard doses of antipsychotics. Methods We searched Embase, Medline, PsycINFO, and the Cochrane Library from database inception until June 17, 2020, for randomised trials in adults with schizophrenia or schizoaffective disorder lasting at least 24 weeks, including individuals clinically stable at baseline, and comparing at least two doses of the same antipsychotic, excluding trials in first-episode psychosis or treatment-resistant schizophrenia. We compared low-dose (within 50–99% of the lower limit of the standard dose) and very-low dose (less than 50% of the lower limit) with standard dose, defined as doses higher than the lower limit of the treatment dose recommended by the International Consensus Study. Data from published reports on number of participants, treatment, sex, age, number of events, and changes in psychopathology scores were extracted independently by at least two authors. Investigators or sponsors were contacted by email to obtain missing information regarding outcomes. Co-primary outcomes were relapse and all-cause discontinuation. Study-level data were meta-analysed using random-effects models, calculating risk ratios (RRs) for dichotomous data, and Hedges' g for continuous data. The protocol was registered with OSF registries. Findings 7853 references were identified in the database search and one additional reference from a manual review of relevant studies. 5744 abstracts were assessed for eligibility, and 101 references were assessed for full-text review. Of these, 79 were excluded for a variety of reasons, resulting in 22 studies being included in the meta-analysis, reporting on 24 trials and 3282 individuals. Study participants had a median age of 38 years (IQR 36–40) with 2166 (65·9%) males and 1116 (34·0%) females. Compared with standard dose, low dose increased the risk of relapse by 44% (16 trials, 1920 participants; RR 1·44, 95% CI 1·10–1·87; p=0·0076; I2=46%) and the risk of all-cause discontinuation by 12% (16 trials, 1932 participants; RR 1·12, 1·03–1·22; p=0·0085; I2=0%). Very low dose increased the risk of relapse by 72% (13 trials, 2058 participants; RR 1·72, 95% CI 1·29–2·29; p=0·0002; I2=70%) and all-cause discontinuation by 31% (11 trials, 1866 participants; RR 1·31, 1·11–1·54; p=0·0011; I2=63%). Compared with low dose, very low dose did not significantly increase the risk of relapse (five trials, 686 participants; RR 1·31, 95% CI 0·96–1·79; p=0·092; I2=51%) or all-cause discontinuation (five trials 686 participants; RR 1·11, 95% CI 0·95–1·30; p=0·18; I2=43%). Subgroup analyses comparing double-blind versus open-label studies, first-generation versus second-generation antipsychotics, and oral versus long-acting injectable antipsychotics were consistent with the overall results. Most studies were classified as having some concerns in the risk of bias assessment, which was mainly caused by absence of publicly available study registrations. Interpretation During maintenance treatment in multi-episode schizophrenia, antipsychotic doses should probably not be reduced below the standard dose range recommended for acute stabilisation, because reducing the dose further is associated with an increased risk of both relapse and all-cause discontinuation. Funding None