Associations of midpoint of sleep and night sleep duration with type 2 diabetes mellitus in Chinese rural population: the Henan rural cohort study

The present study was conducted using the baseline data of the Henan Rural Cohort Study. Details of the study have been reported elsewhere [23]. Baseline data collection was obtained from July 2015 to September 2017. In total, 39,259 adults aged 18–79 years completed the baseline survey, including questionnaire, anthropometric and laboratory tests. After excluding participants who were shift workers (N = 1553), T1DM patients (N = 4), had missing information on sleep assessment and T2DM diagnosis (N = 99), 37,276 were retained in the present study for analysis.

This research protocol was approved by the Life Science Ethics Committee of Zhengzhou University. All participants signed informed consent.

Information on demographics, lifestyle characteristics, and family disease history was collected by a structured questionnaire. According to previous classification criteria [6], marital status was grouped into two groups: married or cohabitating unmarried and unmarried, divorced or widowed. Education levels were divided into primary school or below, junior high school, and senior high school or above. The average monthly income was categorized into less than 500, 500-, and 1000 or above. Adequate vegetables and fruit intake (> 500 g/day consumption of vegetables and fruit) and high-fat diet (> 75 g/day consumption of meat from livestock and poultry) were defined separately. Smoking and alcohol drinking were grouped into never, former, and current. International Physical Activity Questionnaire (IPAQ) [24] was used to evaluated physical activity (light, moderate, and vigorous). Anthropometric tests were conducted by trained staff according to a standard guideline. Body mass index (BMI) was calculated from the weight and height obtained.

Sleep variables (sleep timing, sleep latency snoring, and so on) were obtained by a face-to-face interview based on the Pittsburgh Sleep Quality Index (PSQI). The PSQI consists of 19 items, divided into 7 components, each with a score of 0–3 and a total score of 21, which is used to assess sleep quality [25]. In the present study, the subjects reported bedtime and wake-up time (the time they usually went to bed and got up) and sleep latency (the minutes it took to fall asleep) in the past month. Using these data, sleep time was computed as bedtime plus sleep latency. The midpoint of sleep was calculated as the halfway point between sleep time and wake-up time and was classified into three categories by the 25th and 75th percentile according to the distribution: Early, Intermediate (reference), and Late [26]. Sleep duration was also calculated from sleep time and wake-up time and grouped as < 5 h, 5- h, 6- h, 7- h (reference), 8- h, 9- h, and ≥ 10 h [27]. In addition, information on napping frequency and napping duration were also obtained.

Subjects were defined as T2DM passing the following criteria: (1) the fasting blood glucose ≥ 7.0 mmol/L or glycated hemoglobin ≥ 6.5%; (2) self-reported being diagnosed with T2DM by a physician or taking hypoglycemic drugs in the past 2 weeks.

Continuous variables are presented as the mean ± SD and categorical variables are presented as the mean ± SD. The t-test (continuous variables) and chi-square test (categorical variables) were used for comparing baseline characteristics of the participants among different T2DM groups. We performed a multiplicative interaction term between gender and the midpoint of sleep or night sleep duration in the regression models and found potential gender interactions (P < 0.05). Hence the results were presented by gender (men and women). Moreover, to explore the associations of the midpoint of sleep and night sleep duration with T2DM, logistic regression models were performed by controlling for age, gender, physical activity, marital status, smoking status, drinking status, educational level, average monthly income, BMI, high fat diet, adequate vegetable and fruit intake, napping duration, family history of diabetes, sleep quality, snoring, midpoint of sleep or night sleep duration. The dose-response associations between the midpoint of sleep and night sleep duration with T2DM were evaluated by the three-knots restricted cubic spline. The three knots were 25th, 50th, and 95th percentiles for the midpoint of sleep and 15th, 50th, and 85th percentiles for the night sleep duration according to the distribution. Besides, stratified analyses were performed to assess the potential interaction. Furthermore, night sleep duration was regrouped into < 6 h, 6- h, 7- h, 8- h, ≥ 9 h to investigate the combined effects of the midpoint of sleep and night sleep duration on T2DM by taking the Intermediate group and 7- h of night sleep duration as the reference. Data were analyzed by using the SAS 9.1 (SAS Institute) and R Language software (version 4.0.2). Statistical significance was based on two-tailed P < 0.05.

Comparisons of characteristics according to T2DM status are presented in Table 1. Of the 37,276 participants, 3580 subjects suffered from T2DM. Compared with those without T2DM, individuals with T2DM were older, and more likely to be women, unmarried, and had lower education level, lower average monthly income, less physical activity, higher BMI, higher PSQI scores and a family history of T2DM. Meanwhile, they tend to have long night sleep and napping durations and early midpoint of sleep. The study participants according to midpoint of sleep showed significant differences for baseline characteristics. The mean midpoint of sleep among the 3 categories were1:05 AM ±23 min, 1:56 AM ±14 min, and 2:57 AM ±34 min, respectively. Details of the baseline characteristics of participants are provided in Supplementary Table 1.

The differences in bedtime, wake-up time and sleep duration (Supplementary Table 2.) across 3 categories of the midpoint of sleep were statistically significant. People with an early midpoint of sleep woke 85 min earlier and went to bed 117 min earlier than people with a late midpoint of sleep. Compared with the Intermediate group (7 h 43 min), the Early group had a longer night sleep duration (8 h 11 min) and the Late group had a shorter night sleep duration (7 h 19 min). In addition, differences were found between men and women.

To further explore the associations between midpoint of sleep/night sleep duration and T2DM, logistic regression models were performed for odds ratios (ORs) and 95% confidence intervals (CIs) (Table 2). Compared with the Intermediate group, people with early midpoint of sleep and late midpoint of sleep had higher odds of T2DM, the ORs (95% CIs) were 1.13 (1.04–1.22) and 1.14 (1.03–1.26) after adjustment for age, gender, marital status, educational levels, average income per month, high vegetables and fruits intake, high fatty diet, smoking status, drinking status, physical activity, body mass index, family history of diabetes, napping duration, sleep quality, snoring and night sleep duration. Stratification by gender revealed that the association between the midpoint of sleep and T2DM was not significant in both men and women in the adjusted model.

Compared with individuals who slept 7- h, longer (≥ 10 h) night sleep duration was relevant to higher odds of T2DM in the adjusted model, the OR was 1.28 (1.08–1.51). Gender stratification showed that the association was more pronounced in women but not statistically significant in men. The ORs and 95%CI were 1.51 (1.07–2.14) and 1.31 (1.07–1.61) for women who slept < 5 h and ≥ 10 h per night.

Furthermore, potential associations between midpoint of sleep/night sleep duration and T2DM were also assessed by restricted cubic spline (Fig. 1). A U-shaped dose-response relationship between the midpoint of sleep and T2DM was found after controlling for potential confounders (P for nonlinearity < 0.001). Gender stratification showed that the U-shaped dose-response association was significant in both men (P for nonlinearity =0.0194) and women (P for nonlinearity =0.0036). As for night sleep duration, the U-shaped dose-response relationship was only found in women (P for nonlinearity =0.0302). In total participants and men, restricted cubic spline showed that P for overall association < 0.05 and P for nonlinearity > 0.05 (P for nonlinearity =0.5338 for total participants and P for nonlinearity =0.1068 for men).

The results of the associations between midpoint of sleep/night sleep duration and T2DM stratified by age, BMI, family history of diabetes, physical activity, smoking and drinking status are presented in Supplementary Table 3. The effects of the midpoint of sleep on T2DM were significantly modified by age. Participants less than 60 years had statistically stronger associations than the others with early midpoint of sleep (OR 1.32, 95%CI 1.15–1.52 vs. OR 1.11, 95%CI 1.00–1.23), but participants ≥ 60 years had statistically stronger associations than the others with late midpoint of sleep (OR 1.20, 95%CI 1.04–1.39 vs. OR 0.92, 95%CI 0.80–1.05). Furthermore, current smoking and past drinking had stronger associations with T2DM in both the Early and Late groups.

Figure 2 displays a combined analysis of the midpoint of sleep and night sleep duration with T2DM. Compared with those with intermediate midpoint of sleep and 7- h of night sleep duration, individuals with a combined late midpoint of sleep and longer night sleep duration (OR 1.38, 95%CI 1.10–1.74) had the highest odds of T2DM. Gender stratification showed that this combined association was enhanced in women, which showed a notably U relationship with T2DM.