Results of stroke risk and coffee consumption are inconclusive. This study aimed to provide an updated systematic review and meta-analysis of the association between coffee consumption and stroke risk.
Random-effects models were used to pool relative risk estimates (RRs) with 95% confidence intervals (CIs). The highest versus the lowest categories of coffee consumption as well as dose-response analysis with a one-stage robust error meta-regression model (REMR) were assessed for stroke risk.
In total, 21 studies including 30 independent cohorts that comprised more than 2.4 million participants were included. The pooled RR with 95% CI for the highest versus the lowest categories of coffee consumption was 0.87 (0.80–0.94) with moderate heterogeneity (I2 = 32.0%). Sensitivity analysis suggested that the influence of each individual data set to an overall result was not significant. As suggested by Begg’s funnel plots and Egger tests (p=0.006), some evidence for publication bias was observed. Further analysis with the trim-and-fill method indicated no noticeable harm to our results was generated by any potential bias. Dose-response analysis suggested a nonlinear relationship (U-shape) between stroke risk and coffee (p = 0.0002). The strongest association for stroke (21% lower risk) was found for coffee consumption of 3–4 cups/day and no further reduction in stroke risk was observed with increasing levels of coffee consumption beyond this amount.
Our study provided evidence of a significant inverse association between coffee consumption and risk of stroke. Future large prospective studies with excellent design are warranted to confirm our findings and provide a more definitive conclusion.