Genetically determined body mass index mediates the effect of smoking on type 2 diabetes risk, but not coronary artery disease risk

Clinical observations have linked tobacco smoking with increased type 2 diabetes risk (1-5), a major public health concern (6). Mendelian randomization analysis has recently suggested smoking may be a causal risk factor for type 2 diabetes (7). However, this initial association could be mediated by additional causal risk factors correlated with smoking behavior, which have not been investigated to date. We hypothesized that body mass index (BMI) could explain the association between smoking and diabetes risk. First, we confirmed previous reports that genetically determined smoking behavior increased risk for both type 2 diabetes (OR=1.21, 95% CI: 1.15-1.27, P=1x10-12) and coronary artery disease (CAD; OR=1.21, 95% CI: 1.16-1.26, P=2x10-20). Additionally, a 2-fold increased smoking risk is positively associated with body mass index (BMI; [~]0.8 kg/m2, 95% CI: 0.54-0.98 kg/m2, P=1.8x10-11). In multivariable Mendelian randomization analysis, including BMI accounted for nearly all of the risk of smoking on type 2 diabetes (OR 1.06, 95% CI: 1.01-1.11, P=0.03). In contrast, the independent association between smoking and CAD persisted (OR 1.12, CI: 1.08-1.17, P=3x10-8) despite controlling for BMI. Causal mediation analyses agreed with these estimates. Our findings support a model whereby smoking initiation increases obesity, which in turn increases type 2 diabetes risk, with minimal if any direct effects from smoking on diabetes risk. Patients should be advised to stop smoking to limit both type 2 diabetes and CAD risk, and therapeutic efforts should consider pathophysiology relating smoking and obesity.