Abstract P206: Coffee Intake Was Not Causally Associated With Type 2 Diabetes Risk in Asians: a Mendelian Randomization Based on the Singapore Chinese Health Study

Abstract

Introduction: Despite the largely reported inverse association between coffee intake and type 2 diabetes (T2D) risk by observational studies, the causality of such association is not well established. Hypothesis: We assessed the hypothesis that the observationally inverse association between coffee intake and T2D risk is causal. Methods: In a nested case-control study within the Singapore Chinese Health Study cohort, we included 2436 T2D cases and 2436 matched controls. At biospecimen collections (1999-2004), participants were free of diagnosed T2D, cardiovascular disease and cancer. Cases were participants who reported to have physician-diagnosed T2D at follow-up visits during 2006-2010. Controls were randomly selected among those who remained free of T2D and were matched to the index cases by age, sex, dialect group and date of biospecimen collection. We 1) tested the association between coffee intake and T2D risk using conditional logistic regression analysis; 2) tested the association of the nine single-nucleotide polymorphisms (SNPs) with coffee intake or risk of T2D; 3) used the instrumental variable (IV) estimators to quantify the strength of the causal association of coffee intake and risk of T2D. The IV estimator (β IV ) which is identical to that derived by the widely used two-stage least squares method, was calculated through dividing the β of the regression coefficients for SNP-T2D by SNP-coffee associations. Results: Median intake (inter quartile range) for coffee intake was 1.00 (0.36-2.51) cups/day in this population. Coffee intake was observationally associated with lower risk of T2D and OR (95% CI) was 0.93 (0.88-0.98) in multivariate adjusted model. We found three SNPs that were significantly associated with higher coffee intake with β coefficients (95% CI) as follows: 0.73 (0.23, 1.24) for rs4410790 (P=0.004), 0.72 (0.22, 1.22) for rs6968554 (P=0.005), and 1.60 (0.28, 2.91) for rs7800944 (P=0.017). Among these tree SNPs, two SNPs were marginally associated with higher T2D risk with OR (95% CI) identically 1.08 (0.99, 1.18) for 7:17284577 and 7:17287106, and the other one was insignificantly associated with lower risk (OR 0.94; 95% CI 0.75, 1.18). The estimated β IV coefficients (95% CI) was not statistically significant for any of the three SNPs: 0.10 (-0.04, 0.25) for rs4410790 (P=0.16), 0.11 (-0.04, 0.25) for rs6968554 (P=0.15), and -0.04 (-0.19, 0.11) for rs7800944 (P=0.60). Conclusions: High coffee intake was observationally associated with lower T2D risk, but our finding does not support a causal association.