Impact of schizophrenia relapse definition on the comparative effectiveness of oral versus injectable antipsychotics: A systematic review and meta-analysis of observational studies

Abstract

Although relapse is an important outcome to measure the effectiveness of schizophrenia treatment, no standard definition exists. This review aimed at identifying definitions and measurements of schizophrenia relapse in observational studies of long-acting injectables (LAIs) versus oral antipsychotics (OAPs) and at determining their impact on heterogeneity of comparative effectiveness estimates. A systematic review was conducted using MEDLINE and Embase (01 January 2010–11 November 2019 [date last searched]). Pragmatic searches of gray literature and snowballing were also conducted. Search outputs were screened independently by two assessors at first stage, and full-text of potentially eligible sources at second stage. For each retained source, definition and measurement of relapse, study methods, and comparative effectiveness estimates were extracted. Heterogeneity of estimates was assessed using I2 statistic with a threshold of 50% for substantial heterogeneity. Literature search yielded 543 sources and pragmatic searches, 21, of which 35 were eligible. Twelve definitions of relapse were found based on hospitalization/emergency department (ED) data (28 studies) or clinical assessment (5 studies). No definition was provided in five studies. According to quantitative analyses, in studies defining relapse as schizophrenia-related hospitalization and/or ED visits over 1-year follow-up, LAIs were significantly more effective than OAPs. For studies measuring relapse based on all-cause hospitalization, heterogeneity was too high for pooling; yet this definition is the most frequently found in pooled estimates published in the literature. Schizophrenia relapse definitions led to substantial heterogeneity of comparative effectiveness estimates of LAIs versus OAPs. Creating study subgroups based on relapse definition effectively reduces statistical heterogeneity.