ABSTRACT:
Background:
The association between habitual coffee consumption and incident stroke remains uncertain.
Methods:
We analyzed 490 889 UK Biobank participants without prevalent stroke. Average self-reported coffee intake across available assessments was categorized as ≤0.5, 0.6 to 1.67, 1.68 to 3.0, and >3.0 cups/d. Incident stroke was ascertained through linkage to hospital and death registries and self-reported physician diagnosis. Multivariable Cox models and restricted cubic splines characterized associations. Analyses assessed effect modification by CYP1A2 genotype and stroke polygenic risk score and used multiple-imputation and competing-risk sensitivity analyses.
Results:
During follow-up, 22 579 incident strokes occurred. Compared with ≤0.5 cups/d, adjusted hazard ratios were 0.86 (95% CI, 0.83-0.89) for 0.6 to 1.67 cups/d, 0.92 (95% CI, 0.89-0.96) for 1.68 to 3.0 cups/d, and 0.98 (95% CI, 0.94-1.02) for >3.0 cups/d. Splines showed a nonlinear J-shaped association, with the lowest estimated risk at ∼2 to 3 cups/d. Associations were broadly consistent across CYP1A2 and polygenic-risk strata, with no significant interaction. Moderate intake was associated with lower risk among participants reporting sugar addition, whereas high decaffeinated-coffee intake was associated with higher risk.
Conclusions:
Moderate habitual coffee consumption was associated with lower incident stroke risk, with the lowest observed risk at ∼2 to 3 cups/d. These observational findings should not be interpreted as evidence to prescribe coffee for stroke prevention.