Does physical activity reduce the risk of psychosis? A systematic review and meta-analysis of prospective studies

•This review identified 4 studies examining the longitudinal relationships between physical activity and risk for psychosis.•Crude analyses reveal that physical activity may be prospectively associated with a decreased risk of incident psychosis.•Analyses including 2 studies that adjusted the effects for confounding factors did not find protective effects of PA on incident psychosis/schizophrenia.•It is unclear if PA is associated with future psychosis/schizophrenia risk and future robust longitudinal studies are required to clarify this potential relationship. Longitudinal prospective cohorts have suggested that physical activity (PA) may be a protective factor against psychosis and schizophrenia. However, no meta-analysis has been conducted. The study aims to examine the prospective relationship between PA and incident psychosis/schizophrenia. Major databases were searched from inception to July 2019 for prospective studies that calculated the odds ratio (OR) or the adjusted odds ratio (AOR) of incident psychosis/schizophrenia in people with higher PA against people with lower PA. Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS). A random-effects meta-analysis was conducted, for OR and AOR, separately. Across 4 cohorts (N = 30,025 median males = 50%, median follow-up = 32 years), people with high self-reported PA (versus low PA) were at reduced odds of developing psychosis/schizophrenia (OR = 0.73, 95%CI 0.532 to 0.995, p = 0.047). Analysis including 2 cohorts presenting AOR were not statistically significant (AOR = 0.59, 95%CI 0.253 to 1.383, p = 0.226). Overall study quality was high (mean NOS = 7.0). The literature on the topic is scarce, whilst crude analysis suggests that PA may be a protective factor against the emergence of psychosis/schizophrenia, but when adjusting for covariates, the association is no longer significant. Further studies with objective physical activity and adjustment for confounders are needed.