Abstract
AIMS/HYPOTHESIS: Baseline sleep duration has a U-shaped relationship with type 2 diabetes, but little research examines the associated changes. We examined long-term changes in sleep duration and concomitant changes in diet, physical activity, weight and subsequent diabetes. METHODS: The cohort includes 59,031 women aged 55-83 years in the Nurses' Health Study without diabetes in 2000. Change in sleep duration is the difference between self-reported 24 h sleep duration in 1986 and 2000. Diet, physical activity and covariates were updated every 2-4 years. Self-reported diabetes was confirmed via validated questionnaires. Cox regression models were adjusted for 1986 sleep duration and 1986 values of diabetes risk factors, including BMI, and subsequently for change in covariates from 1986 to 2000. RESULTS: We documented 3,513 incident diabetes cases through to 2012. Compared with no change, decreases in sleep duration were adversely associated with changes in diet quality and physical activity, while increases were associated with greater weight gain. After adjustment for 1986 covariates, HRs (95% CI) for ≤ -2, > -2 to < 0, > 0 to < 2 and ≥ 2 h/day changes in sleep duration (vs no change) were 1.09 (0.93, 1.28), 1.10 (1.001, 1.12), 1.09 (1.00, 1.18) and 1.30 (1.14, 1.46), respectively. Additional adjustment for diet and physical activity did not appreciably alter the results. Increases in sleep duration ≥ 2 h/day remained adversely associated with diabetes (HR [95% CI]: 1.15 [1.01, 1.30]) after adjustment for change in covariates, including BMI. CONCLUSIONS/INTERPRETATION: Increases in sleep duration among middle-aged and older women were modestly associated with risk of diabetes; changes in diet, physical activity and BMI did not explain associations.