Abstract
OBJECTIVE: To investigate the potential causal effects of long term plasma caffeine concentrations on adiposity, type 2 diabetes, and major cardiovascular diseases. DESIGN: Two sample mendelian randomisation study. SETTING: Genome-wide association study summary data for associations of two single nucleotide polymorphisms associated with plasma caffeine at the genome-wide significance threshold (rs2472297 near the CYP1A2 gene and rs4410790 near the AHR gene) and their association with the outcomes. PARTICIPANTS: Primarily individuals of European ancestry participating in cohorts contributing to genome-wide association study consortia. MAIN OUTCOME MEASURES: Outcomes studied were body mass index, whole body fat mass, whole body fat-free mass, type 2 diabetes, ischaemic heart disease, atrial fibrillation, heart failure, and stroke. RESULTS: Higher genetically predicted plasma caffeine concentrations were associated with lower body mass index (beta -0.08 standard deviation (SD) (95% confidence interval -0.10 to -0.06), where 1 SD equals about 4.8 kg/m(2) in body mass index, for every standard deviation increase in plasma caffeine) and whole body fat mass (beta -0.06 SD (-0.08 to -0.04), 1 SD equals about 9.5 kg; P<0.001) but not fat-free mass (beta -0.01 SD (-0.02 to -0.00), 1 SD equals about 11.5 kg; P=0.17). Higher genetically predicted plasma caffeine concentrations were associated with a lower risk of type 2 diabetes in two consortia (FinnGen and DIAMANTE), with a combined odds ratio of 0.81 ((95% confidence interval 0.74 to 0.89); P<0.001). Approximately half (43%; 95% confidence interval 30% to 61%) of the effect of caffeine on type 2 diabetes was estimated to be mediated through body mass index reduction. No strong associations were reported between genetically predicted plasma caffeine concentrations and a risk of any of the studied cardiovascular diseases. CONCLUSIONS: Higher plasma caffeine concentrations might reduce adiposity and risk of type 2 diabetes. Further clinical study is warranted to investigate the translational potential of these findings towards reducing the burden of metabolic disease.