Shao · Journal of stroke and cerebrovascular diseases : the official journal of National Stroke Association 2021 · systematic review and meta-analysis of cohort studies · n=21 studies (30 cohorts, >2,400,000 participants)

Coffee Consumption and Stroke Risk: Evidence from a Systematic Review and Meta-Analysis of more than 2.4 Million Men and Women.

Cited 48 times in the scientific literature.

Level 3 - non-randomized controlled study

Systematic review and meta-analysis of observational cohort studies

PubMed 33188952 · doi:10.1016/j.jstrokecerebrovasdis.2020.105452 · record verified 2026-08-28

What was done

Authors conducted a systematic review and meta-analysis assessing the association between coffee consumption and stroke risk across 21 studies with 30 independent cohorts encompassing over 2.4 million participants. Random-effects models pooled relative risks (RRs) for highest versus lowest consumption categories, and a one-stage robust error meta-regression model evaluated dose-response relationships.

What was found

The pooled RR for the highest versus lowest coffee consumption was 0.87 (95% CI 0.80–0.94) with moderate heterogeneity (I² = 32.0%). Dose-response meta-regression showed a nonlinear U-shaped association (p = 0.0002), with the strongest stroke risk reduction (21% lower risk) at 3–4 cups per day and no further risk reduction at higher consumption levels. Publication bias was detected by Egger's test (p = 0.006), though trim-and-fill adjustment did not noticeably alter the findings.

Why it matters

This large meta-analysis provides consolidated observational evidence that moderate daily coffee intake is linked to lower stroke risk, establishing a dose-response plateau at 3 to 4 cups per day.

Limits

All included studies were observational cohorts subject to residual confounding and exposure misclassification from self-reported intake. The abstract does not report stroke subtypes (ischemic vs. hemorrhagic), coffee preparation methods, or differences between caffeinated and decaffeinated coffee, and significant publication bias was present.

Cited by