Sleep And Cardiovascular Disease Association Studies

Correlations between sleep pattern and cardiovascular diseases in a Chinese middle-aged population--the application of multilevel models in cardiovascular disease researchBackground and PurposeCardiovascular (CV) diseases is the result of the interaction of genes and environmental factors. Residents in the same cluster tend to have similar CV risk factors, which lead to a similar CV prevalence. Multistage cluster sampling is useful in large-scale, population-based epidemiological studies. The data derived from multistage cluster sampling is hierarchically structured, in which the cluster effect and individual effect is inseparable. Multilevel regression, but not traditional linear or logistic regression, is appropriate for the hierarchically structured data. Epidemiological and animal studies have suggested an association between habitual sleep pattern with CV events, but the results are still controversial. Therefore, the aims of this study to investigate the relationships between habitual sleep patterns with CV disease based on Prescriptive Urban & Rural Epidemiology Study (PURE)-China study using multilevel logistic regression, prior to that, how to choose multilevel logistic regression when data with more than two levels involved was explored using simulation method.MethodsProspective Urban Rural Epidemiology (PURE) China study was used to explore the association between sleep patterns and CV events, in which a multistage sampling was used,12 centers (county) and 270 communities were selected. There were 46,285 participants, aged 35-70 years olds, were recruited in this study. Habitual sleep patterns and CV events are self-reported. Multlevel logistic regression was used in our analysis. On the other hand, three levels data (County-Community-individual) with different intraclass correlation coefficients (ICC), which were evaluated based on PURE-China study, was simulated, and then bootstrap method was used to extract sub-datasets with different sample size and data structure. Accuracy of parameters estimation was used to evaluate the performance of two-level and three-level logistic regression in different scenarios base on sample size, ICC and numbers of County or Community.ResultsIn this study,39,515 participants, aged 35-70 years, were eligible in our analysis, including 23,345(59.1%) women, and 16,170(40.9%) men. Mean self-reported sleep duration was 8.7±1.1 h per day and 8.2±1.0 h per night.14.2% of participants reported sleeping 7-8 h per day,46.8% of participants sleeping≥9 h per day, only 0.7% sleeping<6 h per day; and 22.7% participants reported sleeping 7-8 h per night,29.4% participants sleeping ≥9 h per night,1.1% sleeping <6 h per night. After adjusting age, sex, body mass index, education attainment, smoking, drinking, exercise, urban/rural, total cholesterol, diabetes, hypertension and depression, comparing 7-8 h per day, the participant with sleeping≥ 9 h per day increased CV risk by 16% (OR= 1.16,95% CI:1.03-1.29, P= 0.0267). Naps increased CV risk by 27% (OR= 1.27, 95% CI:1.18-1.36, P<0.0001), Meanwhile, the CV risk increased with increasing nap duration, and men might suffer more from the longer naps duration. The participant with sleeping≥9 h per day increased stroke risk by 37% (OR=1.37,95% CI:1.12-1.62, P=0.0127); naps increased stroke risk by 47% (OR= 1.47,95% CI: 1.30-1.64, P<0.0001). The participant with sleeping< 6 h per day increased coronary artery disease(CAD) risk by 58% (OR= 1.58,95% CI:1.13-2.03, P= 0.0478); naps increased CAD risk by 23% (OR=1.23,95% CI:1.12-1.34, P= 0.0001). Participants with just 7-8 hour per night reduced CV risk by 24% (OR= 0.76,95% CI:0.60-0.92, P= 0.0008) comparing other sleep patters(all participants were divided three groups:just 7-8 hour per night group,7-8 hours per day group and other sleep patters group), reduced the risk of stroke by 33% (OR=0.67,95% CI: 0.36-0.98, P=0.0130), reduced CAD risk by 20% (OR=0.80,95% CI:0.63-0.97, P =0.0130). The participants who don’t takes daytime naps and sleeps <6 hours per night increased CV risk by 76% (OR=1.76,95% CI:1.25-2.27, P= 0.0298).Overall, the accuracy of β using three-level logistic regression increased with increasing sample size. Accuracy of β using two-level logistic regression is associated with total sample size and number of clusters in different levels. The accuracy of β increased with increasing the number of clusters in level 3 for same total sample size. The sample size is too small, such as less than 3000, two-level logistic regression is better; and when the sample size is increased to 18,000, the three-level logistic model may obtain a more accurate estimation of β.β value itself influent the accuracy, a small sample size meets an accurate estimation when β value is large.ConclusionNaps, long and short duration of habitual sleep may increase the risk of CV events. Participants with just sleeping 7-8 hours per night, have lowest prevalence of CV disease. In addition, the simulation study provided more information for selecting multi-level logistic regression in future study.