Background: Previous meta-analysis showed an inverse association between coffee consumption and all-cause mortality. However, the relationship between caffeinated and decaffeinated coffee consumption and all-cause mortality is inconsistent. We aimed to identify and review the published evidence updating the association between coffee consumption and all-cause mortality and, furthermore, to investigate the association of caffeinated and decaffeinated coffee consumption and all-cause mortality. Methods: We systematically searched PubMed and Web of Science for studies published up to 9 November 2017. Cohort studies in which authors reported relative risks (RRs) of all-cause mortality for at least three levels of coffee consumption were eligible. Random-effects models were used to estimate the pooled RR of all-cause mortality with coffee consumption. Restricted cubic splines were used to model the dose-response association. Results: We included 21 cohort study articles (10 103 115 study participants and 240 303 deaths). We found a nonlinear association between coffee consumption and all-cause mortality (Pnonlinearity < 0.001). Compared with no or rare coffee consumption, with a consumption of 3 cups day-1 , the risk of all-cause mortality might reduce 13% (RR = 0.87; 95% confidence interval = 0.84-0.89). Conclusions: The findings of the present study provide quantitative data suggesting that coffee consumption plays a role in reducing the risk of all-cause mortality. Similar inverse associations are found for caffeinated coffee and decaffeinated coffee.