Do smoking and alcohol behaviours influence risk of type 2 diabetes? A Mendelian randomisation study.

Background: Previous studies suggest that smoking and higher alcohol consumption are both associated with greater risk of type 2 diabetes (T2D). However, studies examining whether these associations reflect causal relationships are limited and do not consider continuous glycaemic traits. The aim of the study was to determine whether there are causal effects of smoking and alcohol consumption on T2D risk and related glycaemic traits. Methods and Findings: We conducted both two-sample and one-sample MR to examine the effects of lifetime smoking index (LSI) and alcoholic drinks per week on T2D and continuous traits (fasting glucose, fasting insulin and glycated haemoglobin, HbA1c). For two-sample MR we used results from genome-wide association studies (GWAS) of LSI (N=462,690), alcohol consumption (N=941,280), T2D (N= 148,726 cases and 965,732 controls) and continuous traits (N=149,289 to 209,605). We used inverse variance weighting (IVW) for our main analyses and conducted several sensitivity analyses to explore violation of MR assumptions. We compared two-sample MR to one-sample MR results for alcohol effects on T2D and HbA1c in UK Biobank (N=336,984). Only these analyses were conducted to avoid sample overlap and due to data availability. The main IVW two-sample MR results suggested possible causal effects of higher LSI on T2D risk (OR per 1SD higher LSI=1.42, 95% CI=1.22 to 1.64); however, sensitivity analyses did not consistently support this finding, and there was evidence of potential horizontal pleiotropy. There was no robust evidence that higher drinks per week influenced risk of T2D from our main IVW two-sample MR analyses (OR per 1 SD higher log-transformed drinks per week=1.04, 95% CI=0.40 to 2.65), despite evidence of causal effects on higher fasting glucose (difference in mean fasting glucose in mmol/l per 1SD higher log-transformed drinks per week=0.34, 95% CI=0.09 to 0.59). One-sample MR results suggested a possible causal effect of higher drinks per week on T2D risk (OR per 1 SD higher log-transformed drinks per week=1.71, 95% CI: 1.24 to 2.36), but in contrast, lower HbA1c levels (difference in mean SD of log transformed HbA1c (mol/mol) per 1 SD higher log-transformed drinks per week=-0.07, 95% CI: -0.11 to -0.02). Key limitations include limited generalisability of results due to analyses being conducted in European populations, and potential selection bias in UK Biobank influencing results. Conclusion: Our results suggest effective public health interventions to prevent and/or reduce smoking and alcohol consumption are unlikely to reduce the prevalence of T2D.