Association between sleep duration and hypertension in southwest China: a population-based cross-sectional study

Objective Hypertension is a major risk factor and cause of many non-communicable diseases in China. While there have been studies on various diet and lifestyle risk factors, we do not know whether sleep duration has an association to blood pressure in southwest China. This predictor is useful in low-resource rural settings. We examined the association between sleep duration and hypertension in southwest China. Design Population-based cross-sectional study. Setting This study was part of the baseline survey of a large ongoing prospective cohort study, the China Kadoorie Biobank. Participants were enrolled in 15 townships of Pengzhou city in Sichuan province during 2004–2008. Participants 55 687 participants aged 30–79 years were included. Sleep duration was assessed by a self-reported questionnaire. Main outcome measures Hypertension was defined as systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90 mm Hg, or prior physician-diagnosed hypertension in hospitals at the township (community) level or above. Results The prevalence of hypertension was 25.17%. The percentages of subjects with sleep durations of <6, 6, 7, 8 and ≥9 hours were 17.20%, 16.14%, 20.04%, 31.95% and 14.67%, respectively. In multivariable-adjusted analyses, the increased ORs of having hypertension were across those who reported ≥9 hours of sleep (men: 1.16, 95% CI 1.04 to 1.30; women: 1.19, 95% CI 1.08 to 1.32; general population: 1.17, 95% CI 1.08 to 1.26). The odds of hypertension was relatively flat until around 6.81 hours of sleep duration and then started to increase rapidly afterwards in subjects and a J-shaped pattern was observed. There was a U-shaped relationship between sleep duration and hypertension in females. Conclusion Long sleep duration was significantly associated with hypertension and a J-shaped pattern was observed among rural adults in southwest China, independent of potential confounders. However, this association was not obvious between short sleep duration and hypertension.

1) The quality of statistical reporting and data presentation was acceptable. I score 6 in a scale from 0 (poor) to 10 (very high). However, the authors could improve it. 2) Page 8, Statistical analysis sub-section: There are insufficient details in the statistical methods, and it needs to be more detail. For example: -Line 16: Replace the symbol of mean value with correct symbol and define the abbreviations. -Line 18: The expression "composition ratio or rate" is strange. Do you mean "Frequency and percentage distributions were reported for categorical variables"? -Lines 18-22: Please name the continuous and categorical variables (covariates). In addition, clarify that you are estimating the statistical significance of their associations with sleep duration. You should also estimate their associations with your main outcome, i.e. hypertension. -Line 26: Name the potential confounding factors analysed.
-Have you examined normality of the continuous covariates? 3) Your main outcome variable is hypertension (no vs yes). Help your readers and include a table where frequency and percentage distributions of hypertension are presented by sleep duration and the main covariates used for the adjustment in the multivariable regression modelling. This data helps your readers to interpret why the relationship between sleep duration and hypertension changed from the unadjusted analysis. 4) Table 2: The adjustment for age and sex changed the effect of sleep duration on the outcome. Were age and sex confounding factors? Table 1 clearly shows that age is associated with sleep duration. In the Figure 1 you report ORs stratified by age groups. In Table 2, you report ORs stratified by age and sex. However, in models 2 for males and females, the only adjusting factor is age. Further adjustments with other covariates do not seem to change the findings. How is age related with hypertension? Readers would like to see the relationship between age and hypertension. Is it age that explains the unadjusted correlation between sleep duration and hypertension? I think you should clarify this more clearly using your data. Please see my previous comment about adding a table where the distribution of hypertension is reported by sleep duration, age and some other (but not all) potential confounding factors. 5) Table 1: The title is not correct. You report distributions of sleep duration by categorical basic characteristics and descriptive statistics of continuous covariates in the sleep duration groups. In addition, divide the table more clearly to two parts. In the first part add "n (%)" to column headings and report the p-value of chisquare test. In the second part add "Mean (SD)" to column headings.

VERSION 1 -AUTHOR RESPONSE
Reviewer: 1 Dr. Weihong Chen, Huazhong University of Science and Technology Comments to the Author: The authors have analyzed the association between sleep duration and hypertension in southwest China with a cross-sectional study. Overall, the sample size is large and the age group is more, which could provide more information. But some comments need to be addressed.
(1) The sleep duration is 6, 7, 8, or 6-7, 7-8, 8-9? Please clarify Responds: At the baseline survey, participants were asked to report the number of hours they slept a day during the last 12 months. Sleep duration was assessed by a self-reported questionnaire with the following question: "On average, how many hours do you typically sleep per day (including daytime naps)?". Respondents could report in only 1-hour increments. In this study, sleep duration was categorized as five groups of <6, 6, 7, 8 and ≥ 9 hours. Therefore, the sleep duration is 6, 7, 8 hours, not 6-7, 7-8, 8-9 hours. We have revised the description of sleep duration in the part of "subject and methods".
(2) The dose response relationship with cure spline is needed in the result. Responds: Done. Restricted cubic splines regression was used to visualize the relationship of sleep duration with hypertension. Results of restricted cubic splines regression among different gender and age subgroups were displayed in Figure 2.
(3) Please add the relationship of sleep duration with sbp and dbp Responds: Done. We used multiple linear regression model to analyse the relationship between sleep duration and blood pressure. Results of the association for sleep duration with sbp and dbp were displayed in Table 4. Age and gender-specific linear regression coefficients of sleep duration for SBP and DBP were calculated by multiple linear regression model.
(4) The English language is need to improved: a) The "risk" of hypertension should be avoided, as it is a cross-sectional study. b) Long sleep duration of longer sleep duration？ c) …… Responds: a) Yes, have changed throughout the manuscript. b) It means long sleep duration. We have changed throughout the manuscript. Reviewer: 2 Dr. Pentti Nieminen, University of Oulu Comments to the Author: This review is primarily a statistical one, with recommendations and specific major and minor points.
1) The quality of statistical reporting and data presentation was acceptable. I score 6 in a scale from 0 (poor) to 10 (very high). However, the authors could improve it. Responds: According to the comments and suggestions of reviewers and editor, we have revised our manuscript.
2) Page 8, Statistical analysis sub-section: There are insufficient details in the statistical methods, and it needs to be more detail. For example: -Line 16: Replace the symbol of mean value with correct symbol and define the abbreviations. Responds: Done. Have changed at the beginning of statistical analysis sub-section. -Line 18: The expression "composition ratio or rate" is strange. Do you mean "Frequency and percentage distributions were reported for categorical variables"? Responds: Yes. Have changed throughout the manuscript. -Lines 18-22: Please name the continuous and categorical variables (covariates). In addition, clarify that you are estimating the statistical significance of their associations with sleep duration. You should also estimate their associations with your main outcome, i.e. hypertension. Responds: Done. Have changed throughout the manuscript. In statistical analysis sub-section, we named the continuous and categorical variables (covariates) and clarified that the statistical significance of their associations with sleep duration was estimated. Results were presented in Table 1. We also estimated associations of the continuous and categorical variables with hypertension. We added two tables to display the prevalence of hypertention in different covariates subgroups. Chisquare tests and the Cochran-Armitage tests for trend were used to examine the relationships between covariates and hypertension. Results were presented in Table 2 and Table 3.

3) Your main outcome variable is hypertension (no vs yes). Help your readers and include a table
where frequency and percentage distributions of hypertension are presented by sleep duration and the main covariates used for the adjustment in the multivariable regression modelling. This data helps your readers to interpret why the relationship between sleep duration and hypertension changed from the unadjusted analysis. Responds: Done. We added two tables to display the prevalence of hypertention and percentage distributions of hypertension in different sleep duration groups and the main covariates used for the adjustment in the multivariable regression modelling subgroups. Results were presented in Table 2 and Table 3. 4) Table 2: The adjustment for age and sex changed the effect of sleep duration on the outcome. Were age and sex confounding factors? Table 1 clearly shows that age is associated with sleep duration. In the Figure 1 you report ORs stratified by age groups. In Table 2, you report ORs stratified by age and sex. However, in models 2 for males and females, the only adjusting factor is age. Further adjustments with other covariates do not seem to change the findings. How is age related with hypertension? Readers would like to see the relationship between age and hypertension. Is it age that explains the unadjusted correlation between sleep duration and hypertension? I think you should clarify this more clearly using your data. Please see my previous comment about adding a table where the distribution of hypertension is reported by sleep duration, age and some other (but not all) potential confounding factors. Responds: Done. We added a table where the distribution of hypertension is reported by sleep duration, age and some other potential confounding factors. We also analyzed the relationships between sleep duration and hypertension, SBP and DBP by sex and age groups by restricted cubic splines and multiple linear regression model. 5) Table 1: The title is not correct. You report distributions of sleep duration by categorical basic characteristics and descriptive statistics of continuous covariates in the sleep duration groups. In addition, divide the table more clearly to two parts. In the first part add "n (%)" to column headings and report the p-value of chi-square test. In the second part add "Mean (SD)" to column headings. Responds: Done. We have revised the title of Table 1. We added "n (%)" and "Mean (SD)" to column headings of Table 1. Have changed throughout the manuscript.

GENERAL COMMENTS
The authors addressed the main concerns from the reviews, the revised version of the manuscript appears to be good. The authors addressed the main concerns from the reviews, the revised version of the manuscript appears to be good.
I have some minor comments: