Background Despite accumulating evidence of an association between stressful life events and psychosis relapse, the extent to which this is a causal relationship remains unclear. We aimed to examine the association between exposure to, and number of, stressful life events after initial psychosis onset and psychosis relapse.Methods In this 2-year prospective observational study, we recruited individuals with first-episode psychosis, aged 18-65 years, who presented to psychiatric services in south London, UK. Participants were assessed via interview, with additional data obtained from electronic clinical records. Stressful life events were recorded at psychosis onset and during the 2-year follow-up using a brief questionnaire that assesses 12 major life events. Psychosis relapse was defined as inpatient admission because of symptom exacerbation within 2 years from psychosis onset. We examined the time to first psychosis relapse and the number and length of relapses using survival and binomial regression analyses. We used fixed-effects regression and cross-lagged path analysis to examine the directionality of effects and control for unmeasured confounders. Findings Between April 12, 2002, and July 26, 2013, 256 individuals with first-episode psychosis (100 [39%] female and 156 [61%] male; 16 [6%] Asian, 140 [55%] Black African or Caribbean, 86 [34%] White, and 14 [6%] mixed ethnicity) were recruited, with a mean age of onset of psychosis of 28 center dot 06 years (SD 8 center dot 03; range 17 center dot 21-56 center dot 03). 93 (36%) participants experienced at least one relapse during the 2-year follow-up. 253 individuals had all relevant data and were included in analyses. For people exposed to stressful life events after the onset of psychosis, the adjusted hazard (hazard ratio [HR] 2 center dot 60, 95% CI 1 center dot 63-4 center dot 16, p<0 center dot 0001), incidence (incidence rate ratio [IRR] 1 center dot 87, 1 center dot 24-2 center dot 80, p=0 center dot 0026), and length (IRR 2 center dot 53, 1 center dot 40-4 center dot 67, p=0 center dot 0011) of relapse were greater than for those who were unexposed. These relationships were dose dependent (HR 1 center dot 36; 1 center dot 09-1 center dot 69, p=0 center dot 0054; incidence IRR 1 center dot 26, 1 center dot 02-1 center dot 53, p=0 center dot 023; length IRR 1 center dot 52, 1 center dot 12-2 center dot 12, p=0 center dot 0028). Adjusted fixed-effects models showed a higher (odds ratio [OR] 3 center dot 82, 1 center dot 82-8 center dot 00, p=0 center dot 0004) and dose-dependent (OR 1 center dot 62, 1 center dot 18-2 center dot 21, p=0 center dot 0028) risk of relapse when stressful life events preceded relapse compared with the period when they did not. Cross-lagged path analysis confirmed an effect of stressful life events on the number of subsequent relapses (beta=0 center dot 66, p=0 center dot 0055) that was dose dependent (beta=0 center dot 29, p=0 center dot 029), but it did not show an effect of relapses on subsequent risk or number of stressful life events.Interpretation These results provide converging evidence of a causal effect of stressful life events on the risk of relapse in psychosis. They suggest that there is a need to develop interventions at the individual and health-service level that could mitigate the harmful effects of stressful life events.