OBJECTIVES: This systematic review and meta-analysis combined published study-level data and unpublished individual-participant data with the aim of quantifying the relation between long working hours and the onset of depressive symptoms. METHODS: We searched PubMed and Embase for published prospective cohort studies and included available cohorts with unpublished individual-participant data. We used a random-effects meta-analysis to calculate summary estimates across studies. RESULTS: We identified ten published cohort studies and included unpublished individual-participant data from 18 studies. In the majority of cohorts, long working hours was defined as working ≥55 hours per week. In multivariable-adjusted meta-analyses of 189 729 participants from 35 countries [96 275 men, 93 454 women, follow-up ranging from 1–5 years, 21 747 new-onset cases), there was an overall association of 1.14 (95% confidence interval (CI) 1.03–1.25] between long working hours and the onset of depressive symptoms, with significant evidence of heterogeneity (I=45.1%, P=0.004). A moderate association between working hours and depressive symptoms was found in Asian countries (1.50, 95% CI 1.13–2.01), a weaker association in Europe (1.11, 95% CI 1.00–1.22), and no association in North America (0.97, 95% CI 0.70–1.34) or Australia (0.95, 95% CI 0.70–1.29). Differences by other characteristics were small. CONCLUSIONS: This observational evidence suggests a moderate association between long working hours and onset of depressive symptoms in Asia and a small association in Europe.