Baseline characteristics of participants were summarized and compared by subgroups of habitual bedtime. Continuous variables were expressed as mean with SD and compared by 1-way analysis of variance or Kruskal-Wallis test. Categorical variables were presented as frequency (percentage) and compared by 2-way χ2 test. To investigate the associations between sleep timing behaviors (ie, bedtime and wake-up time) and specific obesity types, multilevel logistic regression models were applied to calculate adjusted odds ratios (AORs) and the 95% CIs, using bedtime 8 pm to 10 pm or wake-up time 4 am to 6 am as the reference. Random effects for centers were used to account for the clustering within centers (which also accounts for country and region). The fully adjusted models for the association with bedtime included age, sex, education, location (urban or rural area), country income status (high, middle, or low), smoking status (current or former vs never smoker), drinking status (drinker vs nondrinker), family history of cardiovascular conditions (defined as father, mother, or siblings having diabetes, stroke, coronary heart disease, or high blood pressure), history of diabetes (defined as self-reported diabetes or using antidiabetic medications), depression, physical activity (expressed as MET-min/wk), nocturnal sleep duration (hours), total energy intake (kilocalories), and habitual naps (yes or no). We examined collinearity between potential confounding variables using variance inflation factors (VIFs), and all VIFs were less than 5, indicating a lower chance of collinearity problem between these covariables.

We examined the normality for the timing and length of nocturnal sleep, and they were not strictly normally distributed. We then classified nocturnal sleep duration into 7 subgroups (<5 hours, 5-6 hours, 6-7 hours, 7-8 hours, 8-9 hours, 9-10 hours, ≥10 hours) and napping duration into 3 subgroups (0, 0-1 hour, ≥1 hour) to examine the associations with the key outcomes by taking sleep duration of 7 or more to less than 8 hours and napping 0 hours as the reference. Additionally, we calculated midsleep time (MST) as a measurement of sleep phase, using noon as the reference and formula bedtime + (nocturnal sleep duration / 2), and treated MST as a continuous variable when examining the association with obesity types.

Given potential differences in the definition of abdominal obesity between ethnic groups, we performed sensitivity analyses by redefining ethnicity-specific abdominal obesity (ethnic groups were classified based on the self-reported ethnicity of the participants) according to the current recommended waist circumference thresholds for abdominal obesity prescribed by a Joint Interim Statement25; however, we only redefined general obesity in the sensitivity analyses for Chinese individuals according to the recommended criteria for Chinese population (ie, BMI≥28) because the WHO has no clear guideline for Asian populations.26 Missing data of sleep owing to no response (eFigure in the Supplement) were imputed using SAS statistical software Multiple Imputation Procedure (SAS Institute) to reanalyze the data as the sensitivity analysis. Subgroup analyses were performed according to categories of nocturnal short sleep (ie, <6 h/d vs ≥6 h/d), nocturnal sleep deprivation (ie, <5 h/d vs ≥5 h/d), and napping behavior (yes vs no) to examine if the associations of bedtime with obesity outcomes were modified by short sleep, sleep deprivation, or napping behavior. Consistency of associations of bedtime and MST with specific obesity types was also assessed across subgroups by introducing a multiplicative term in the full models for sex (men vs women), age (ie <65 years vs ≥65 years), physical activity (low, moderate, or high) and total energy intake (tertiles), and further stratified analysis was conducted only if the multiplicative interaction test was statistically significant. P values were 2-sided, and statistical significance was set at P < .05. All multivariate analyses were performed based on the complete data of 136 652 participants. Data analysis occurred from October 2020 through March 2021.

Note: The content above has been extracted from a research article, so it may not display correctly.



Q&A
Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.



We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.