Tea intake shows no genetic causal effect on stroke: A two-sample Mendelian randomization study based on STROBE-MR guidelines

Abstract

Observational studies have reported inconsistent links between tea intake and stroke risk. We applied two-sample Mendelian randomization (MR) to clarify whether the association is causal. Following STROBE-MR guidelines, we extracted genome-wide association study (GWAS) summary statistics for tea intake (UK Biobank, n = 447,485; GWAS ID ukb-b-6066) and stroke (UK Biobank, n = 462,933; GWAS ID ukb-b-6358), both of European ancestry. Thirty-six independent single-nucleotide polymorphisms (SNPs) meeting genome-wide significance (P < 5 × 10-8), linkage disequilibrium threshold r² < 0.001 (10,000 kb), and F > 10 were selected as instrumental variables (IVs). Causal estimates were calculated using inverse variance weighting (IVW) as the primary analysis and weighted median, MR-Egger, simple mode, and weighted mode as complementary methods. Heterogeneity (Cochran Q and Rücker Q), horizontal pleiotropy (MR-Egger intercept), and leave-one-out sensitivity analyses were performed to evaluate robustness. IVW indicated no causal effect of tea intake on stroke (odds ratio [OR] = 0.997, 95% confidence interval [CI]: 0.991-1.004, P = .381). Findings were consistent across MR-Egger (OR = 0.984, 95% CI: 0.953-1.015, P = .306), weighted median, simple mode, and weighted mode analyses. No significant heterogeneity was detected (Cochran Q P = .176; Rücker Q P = .181), and the MR-Egger intercept showed no evidence of horizontal pleiotropy (P = .386). Leave-one-out and funnel plot assessments confirmed result stability and minimal bias. Genetic evidence from this large, two-sample MR study does not support a causal relationship between habitual tea consumption and stroke risk. Public health strategies aimed at stroke prevention should therefore not rely on promoting tea intake alone.

Keywords: Mendelian randomization; prognosis; single-nucleotide polymorphism; stroke; tea intake.

Figure 1.

Forest map of two-sample MR results. MR = Mendelian randomization.

Figure 2.

Scatter plot of MR results for 2 samples. MR = Mendelian randomization.

Figure 3.

Excluding sensitivity analysis results one by one using the “leave-one-out”.

Figure 4.

Funnel plot of MR results for 2 samples. MR = Mendelian randomization.