Abstract Aims To determine the association between the sleep duration and the risk of all-cause mortality in adults. Methods A search strategy was conducted in the MEDLINE, CENTRAL, EMBASE and LILACS databases. Searches were also conducted in other databases and unpublished literature. Cohort studies were included without language, time or setting restrictions. The risk of bias was evaluated with a modified Cochrane Collaboration's tool. An analysis of random effects was conducted. The primary outcome was all-cause mortality. The measure of the effect was the risk difference (RD) with a 95% confidence interval (CI). The planned comparisons were 7–9 h of sleep v. <7 h and the same reference v. >9 h. Results Thirty-nine studies were included in our qualitative analysis, regarding the quantitative analysis, 19 studies were included in <7 v. 7–9 h analysis, and 18 studies in the >9 v. 7–9 h. A low risk of bias was shown for most of the study items. The overall RD for all-cause mortality was 0.09 (95% CI 0.07–0.11) favouring the >9 h group compared with our reference. In contrast, no differences were found between the <7 h and the reference sleep duration groups (RD 0.00, 95% CI 0.00–0.01). Conclusion We found a probable association of long sleep duration and higher mortality; however, it could reflect an underlying systemic or neurological disease that cause sleep fragmentation, deterioration in quality and micro-awakenings.