Bedroom lighting environment and incident diabetes mellitus: a longitudinal study of the HEIJO-KYO cohort

Objectives: Light information received by the brain influences human circadian timing and metabolism; low-level light at night (LAN) significantly increased body mass and led to prediabetes in mice. We hypothesized that LAN exposure increases the diabetes risk in humans. The aim of the present study was to evaluate a longitudinal association between LAN exposure and the incidence of diabetes in a general population. Methods: In our prospective cohort study, bedroom light intensity was measured at 1-min intervals in 678 elderly participants without diabetes at baseline. The average light intensity recorded between bedtimes and rise times over two consecutive nights was used in the analysis. Results: During follow-up (median, 42 months), 19 of the 678 participants (mean age, 70.6 years) developed diabetes. Poisson regression models revealed that the incidence rate for diabetes was significantly higher in the LAN group (average >= 5 lux, N = 128) than the dark group (average < 5 lux, N = 550) (incidence rate ratio, 3.74; 95% confidence interval (CI), 1.55-9.05; p = 0.003). Further propensity score adjustments in relation to LAN produced consistent results (incidence rate ratio, 3.19; 95% CI, 1.38-7.35; p = 0.007). When the cut-off value of LAN was decreased to 3 lux, the relationship remained significant (incidence rate ratio 2.74; 95% CI, 1.19-6.33; p = 0.018). Conclusions: Our findings suggest that LAN exposure increases the incidence of diabetes in a general elderly population. Further research involving a large cohort with new-onset diabetes is warranted to elucidate these findings. (C) 2019 Elsevier B.V. All rights reserved.