Age-dependent changes in the association between sleep duration and impaired glucose metabolism

To investigate whether the association between sleep duration and impaired glucose metabolism varies among younger and older populations.

We reviewed data of self-reported habitual sleep duration per night, HbA1c levels, and clinically relevant factors in a cross-sectional checkup database of 75472 Japanese from the general population aged 20-79 years (51695 men and 23777 women). Associations of prediabetes (HbA1c ≥ 5.7% and/or diabetic pharmacotherapy) or diabetes (HbA1c ≥ 6.5% and/or diabetic pharmacotherapy) with short and long sleep durations compared with a reference sleep duration (7 h) were investigated by multivariate logistic regression analysis. We controlled for potential relevant confounders, including age, sex, and work duration per day according to younger and older subjects.

As age advanced, sleep duration became longer and this increase in the 40s and 50s was two times greater in men than in women. This finding was accompanied by a deterioration in HbA1c levels. In subjects aged younger than 40 years (n = 32929), HbA1c levels were inversely and linearly correlated with sleep duration in both sexes. However, in subjects aged 40 years or older (n = 42543), HbA1c levels showed a non-linear relationship against sleep duration with a nadir at 7 h. Multivariate logistic regression analysis showed that in younger subjects, short durations of sleep (≤ 5 h and 6 h) were positively associated with prediabetes (both P < 0.001), but a long duration of sleep (≥ 8 h) was inversely associated with prediabetes (P < 0.001). These associations remained significant after adjustment for relevant confounders, including age, sex, and work duration per day (ORs = 1.20, 95%CI: 1.05-1.37, P < 0.001; ORs = 1.12, 95%CI: 1.02-1.24, P < 0.05; and ORs = 0.84, 95%CI: 0.72-0.99, P < 0.05, respectively). In contrast, in older subjects, besides an association of prediabetes with a short duration of sleep (≤ 5 h) (ORs = 1.12, 95%CI: 1.03-1.21, P < 0.01), diabetes was significantly associated with a long duration of sleep (≥ 8 h) (ORs = 1.11, 95%CI: 1.02-1.25, P < 0.05).

A short sleep duration may be associated with prediabetes throughout life. However, the association between a long sleep duration and glucose metabolism can change with aging.