Sleep duration and risk of diabetes: Observational and Mendelian randomization studies

Inadequate sleep could contribute to type 2 diabetes, but observational studies are inconsistent and open to biases, particularly from confounding. We used Mendelian randomization (MR) to obtain an unconfounded estimate of the effect of sleep duration on diabetes, fasting glucose (FG) and hemoglobin A1c (HbA1c), and an observation study to assess differences by sex. Using MR, we assessed the effects of genetically instrumented sleep on diabetes, based on 68 single nucleotide polymorphisms (SNPs), applied to the DIAbetes Genetics Replication and meta-analysis case (n = 26,676)-control (n = 132,532) study and on FG and HbA1c, based on 55 SNPs, applied to the Meta-Analyses of Glucose and Insulin-related traits Consortium (MAGIC) study of FG (n = 122,743) and HbA1c (n = 123,665). In the population-representative Hong Kong Chinese “Children of 1997” birth cohort we assessed whether associations of sleep duration at ~17.5 years with FG and HbA1c differed by sex. Using inverse variance weighting with multiplicative random effects, sleep duration was not associated with diabetes (odds ratio (OR) 0.85 per hour of sleep, 95% confidence interval (CI) 0.64 to 1.13), FG (−0.032 mmol/l per hour of sleep, 95% CI −0.126 to 0.063) or HbA1c (−0.022% per hour of sleep, 95% CI −0.069 to 0.024). In “Children of 1997”, the associations of sleep duration with FG differed by sex (p for interaction 0.05) but not with HbA1c. Overall sleep duration does not appear to be related to diabetes, FG or HbA1c, but the possibility of sex differences merits investigation.