Sleep Duration and Hypertension: Analysis of > 700,000 Adults by Age and Sex
The objective of this study was to evaluate the cross-sectional relationship between sleep duration and hypertension in a large, nationally-representative dataset that spans 10 years. This analysis may provide detailed information with high resolution about how sleep duration is related to hypertension and how this differs by demographic group.
Data were aggregated from the 2013 Behavioral Risk Factor Surveillance System (n = 433,386) and the combined 2007–2016 National Health Interview Surveys (n = 295,331). These data were collected by the Centers for Disease Control and Prevention from nationally-representative samples. Surveys were combined, and survey-specific weights were used in all analyses. Sleep duration was assessed with the item, “On average, how many hours of sleep do you get in a 24-hour period?” in both surveys. Hypertension was assessed as self-reported history. Covariates were assessed identically in both datasets and included, age (in 5-year groupings), sex, race/ethnicity, and employment status.
In adjusted analyses, compared to 7 hours, increased risk of hypertension was seen among those sleeping ≤ 4 hours (odds ratio [OR] = 1.86, P < .0005), 5 hours (OR = 1.56, P < .0005), 6 hours (OR = 1.27, P < .0005), 9 hours (OR = 1.19, P < .0005), and ≥ 10 hours (OR = 1.41, P < .0005). When stratified by age, sex, and race/ethnicity groups, short sleep was associated with increased risk for all age groups < 70 years, and long sleep (≥ 10 hours only) was associated with risk for all except < 24 years and > 74 years. Findings for short sleep were relatively consistent across all race/ethnicities, although findings for long sleep were less pronounced among Black/African-American and Other/Multiracial groups. A significant sleep by 3-way sleep × age × sex interaction (P < .0005) suggests that the relationship depends on both age and sex. For both men and women, the OR of having hypertension associated with short sleep decreases with increasing age, but there is a higher association between short sleep and hypertension for women, throughout the adult lifespan.
Both short and long sleep duration are associated with increased hypertension risk across most age groups. The influence of covariates is stronger upon long sleep relationships. Relationships with short sleep were stronger among younger adults and women.
Grandner M, Mullington JM, Hashmi SD, Redeker NS, Watson NF, Morgenthaler TI. Sleep duration and hypertension: analysis of > 700,000 adults by age and sex. J Clin Sleep Med. 2018;14(6):1031–1039.
Hypertension is a leading cause of cardiovascular death and disability, with 1.3 billion people suffering from this condition worldwide.1 In the United States alone, 75 million adults, or about one-third of the population, have hypertension with only about one-half achieving desired therapeutic blood pressure control.2 Thus, hypertension is a major public health problem contributing to stroke, renal disease, ischemic heart disease and heart failure. The Centers for Disease Control and Prevention (CDC) partnered with two professional sleep societies, the American Academy of Sleep Medicine (AASM) and the Sleep Research Society (SRS), to create the National Healthy Sleep Awareness Project. The partnership has formed alliances with multiple stakeholders in sleep health, including regional sleep societies, primary care organizations, and public affairs specialists with the goal of developing and implementing a comprehensive, multi-year, nationwide sleep awareness program to promote healthy sleep. The program's primary goal is to help achieve the sleep health objectives of Healthy People 2020. One of the goals of Heathy People 2020 is to increase the proportion of adults who get sufficient sleep, and one of the health benefits anticipated is to reduce population hypertension prevalence.
The traditional focus for hypertension prevention and management has been on reducing obesity and dietary sodium, increasing physical activity, prescribing antihypertensive medications, and treating sleep-disordered breathing. However, epidemiological evidence, including studies reviewed for the recently published consensus statements from the AASM and the SRS on recommendations for healthy sleep durations for adults,3,4 suggest that sleep duration less than 7 hours contributes to hypertension development.
Experimental studies in both hypertensive and normotensive human subjects show increases in blood pressure and sympathetic tone during sleep restriction.5–8 One preliminary study has found that blood pressure is lowered by a behavioral intervention that includes extending sleep duration.9 Cross-sectional10–33 and longitudinal epidemiological studies13,34–37 suggest a relationship between sleep duration and the incidence or prevalence of hypertension, but the relationship is complex, and there is limited information about presumed important independent variables such as sex and age.38 Existing literature is also limited by a high degree of variability in sample sizes and designs. For example, study size varies between less than several hundred subjects to up to 82,130 subjects, in the cross-sectional data reported by Gangwisch et al.17 Sleep duration measurement in these studies predominantly included self-report with occasional objective measures, and various cutoff points have been used to define short and long sleep duration. Risk of incident hypertension associated with objective short sleep duration based on actigraphic measurement had an odds ratio (OR) of 1.37.24 Although some studies suggest variations in the association between sleep duration and hypertension based on sex,13,31 they had limited statistical power and available measurements have been barriers to understanding the interplay of these possible clinical, demographic, and socioeconomic factors.
There is a need to identify those who are at highest risk for hypertension associated with short or prolonged sleep duration. Better understanding of this relationship will assist in targeting personalized and public health interventions to address optimal sleep needed to support primordial health. However, little is known about the specific populations who are at highest risk for hypertension based on sleep duration. The purpose of this study, an analysis of data obtained from the National Health Interview Survey (NHIS) and the Behavioral Risk Factor Surveillance System (BRFSS), is to evaluate the association between prevalence of hypertension and self-reported sleep duration in a large survey population, and to conduct focused subgroup analyses addressing the associations between sleep duration and prevalence of hypertension in groups stratified by age, sex, race, employment status, and health-related variables (obesity, smoking, alcohol).
The NHIS, conducted annually by the National Center for Health Statistics within the CDC, is a cross-sectional, in-person, interviewer-administered household health survey of the non-institutionalized civilian population that provides nationally-representative data on a broad range of topics. NHIS study populations from 2007 to 2016 were combined and included a total sample of 295,331 adults aged ≥ 18 years with complete data on self-reported sleep duration, self-reported high blood pressure, and other covariates for these analyses. Response rates ranged from 54.3% in 2016 to 67.8% in 2007. Data were weighted to account for the stratified, complex sampling design of the survey, and incorporated final sample weights adjusted for nonresponse and calibrated to population control totals to ensure generalizability to the civilian adult population aged ≥ 18 years.
The BRFSS is an annual survey of adults aged ≥ 18 years that is administered by state health departments in collaboration with the CDC. The BRFSS is a random-digit-dialed telephone survey (landline and cell phone) of non-institutionalized adults and includes questions about respondents' health and risk factors. In 2013, there were 433,386 adults who provided complete information on sleep duration, self-reported high blood pressure, and other characteristics for these analyses. The median response rate for the 50 states and the District of Columbia in 2013 was 46.4% and ranged from 29.0% to 60.3%. Data were weighted to take into account non-response, the complex sampling design, and to be representative of each state population.
The NHIS and BRFSS datasets were combined for all analyses. Although the methodology for the two datasets is slightly different, this was done to generate the largest and most representative sample possible. Of note, the respondents are non-overlapping and weighted. Also, all items needed to be measured in nearly the same way (see below) to be considered for inclusion. This allowed for a total sample size of 728,717, spanning 10 years, with the majority of participants assessed in 2013.
Sleep duration was assessed with a single survey item, “On average, how many hours of sleep do you get in a 24-hour period?” in both the NHIS and BRFSS datasets. Results were coded as whole numbers. Because of fewer numbers at the extremes, and to maintain consistency with previous literature, sleep duration was categorized as ≤ 4 hours, 5 hours, 6 hours, 7 hours (which for our analyses, we regarded as the reference sleep duration), 8 hours, 9 hours, or ≥ 10 hours.
Hypertension was assessed with the item, “Have you ever been told by a doctor, nurse, or other health professional that you…had hypertension, also called high blood pressure?” (NHIS) or “Have you ever been told by a doctor, nurse, or other health professional that you…have high blood pressure?” (BRFSS). Responses were coded as “yes” or “no.”
Several variables were used both as covariates in adjusted models and as variables for stratification. Sociodemographic characteristics included age (categorized as 18–24, 25–29, 30– 34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, 75–79, and ≥ 80 years), sex (male or female), race/ethnicity (categorized as non-Hispanic White, Black/African-American, Hispanic/Latino, Asian/Pacific Islander, or Multiracial/Other), and employment (currently employed or not). Note that more detailed information on employment was not consistently asked across both the NHIS and BRFSS. Health-related variables examined included smoking (current, former, or never), alcohol use (non-drinker, light drinker, or heavy drinker), and body mass index (BMI) based on self-reported height and weight (normal < 27, overweight 27–29.9, obese 30–39.9, and morbidly obese ≥ 40 kg/m2).
Statistical Analyses and Interpretation
Descriptive statistics for all variables were calculated for the complete sample, for NHIS and BRFSS datasets separately, and separately for those with and without hypertension. These analyses were conducted using the provided sample weights. Regarding the latter, pairwise comparisons using t tests for continuous variables and chi-square tests for categorical variables were evaluated.
To determine whether sleep duration was associated with hypertension, weighted binary logistic regression analyses were examined, with hypertension as the outcome and sleep duration as the predictor variable. Adjusted analyses included age category, sex, race/ethnicity, employment, smoking, alcohol, and BMI as covariates. These were computed for the combined sample and for the NHIS and BRFSS samples separately.
To evaluate whether this relationship differs according to levels of the covariates, these analyses were repeated, stratified for age group, sex, race/ethnicity, employment, smoking category, alcohol use category, and BMI category. To evaluate whether relationships differ by both age and sex, age groups were collapsed, and the following groups were assessed: Males and females 18–34, 35–49, 50–64, and ≥ 65 years. These analyses were adjusted for all covariates except for the one on which the results were stratified.
All results are reported as OR and 95% confidence intervals (CI), as well as P values. P ≤ .05 was considered nominally statistically significant and results of all tests are provided for heuristic value. Given the large number of statistical tests performed, results should be interpreted with appropriate caution. All analyses were performed using STATA 14.0 (STATA Corp, College Station, Texas, United States).
Characteristics of the Sample
Sample characteristics are reported in Table 1. This includes samples combined from NHIS 2007–2016, as well as BRFSS 2013. Differences between the samples are reported in Table S1 in the supplemental material. Values in the tables are weighted using the sample weight of the respective surveys/ years. Approximately 3% of the sample reported extreme short sleep, of ≤ 4 hours, with approximately 6% and 21% reporting 5 and 6 hours, respectively. Among the normative sleep duration groups, 30% and 31% reported 7 and 8 hours, respectively. Regarding long sleep, 5% and 4% reported 9 and ≥ 10 hours, respectively. Age groups were distributed such that sufficient representation was available in all groups. The sample was approximately 52% female, 77% non-Hispanic White, and 60% employed. 19% reported being current smokers and 5% reported being heavy drinkers. When characteristics were stratified by hypertension status, there were statistically significant differences between all characteristics, but many of differences were quantitatively quite small.
Sleep Duration and Hypertension
Results of the regression analyses examining relationships between sleep duration and hypertension are reported in Table 2 and displayed graphically in Figure 1. Compared to 7 hours, unadjusted analyses showed that hypertension was reported in all sleep duration groups, with extreme short sleepers (≤ 4 hours) and long sleepers (≥ 10 hours) having higher odds of hypertension (OR = 2.02 and 2.06, respectively; P ≤ .05). After adjustment for age, sex, race/ethnicity, employment status, alcohol, smoking, and BMI, the increased odds of having hypertension was sustained across those who reported ≤ 4 hours, 5, 6, 9 and ≥ 10 hours of sleep, while 8 hours of sleep was no longer associated with increased risk of hypertension.
Stratification by Sociodemographics: Age, Sex, Race/Ethnicity, and Employment
Overall, short sleep effects were reliably observed for all age groups except 70–74, 75–79, and ≥ 80 years (Figure 2). Regarding long sleep, consistent relationships with ≥ 10 hours of sleep were seen for ages 35–74 years. Relationships with long sleep were less evident in those < 24 and > 74 years. Findings for short sleep were relatively consistent across all race/ethnicities, although findings for long sleep were less pronounced among Black/African-American and Other/Multiracial groups (Figure 3). Relationships were consistent for both employed and unemployed individuals, with little difference between groups (Figure 3).
Stratification by Both Age and Sex
A significant sleep by age by sex interaction term (P < .0005) suggests that the relationship between sleep duration and hypertension depends on both variables. Figure 4 and Table S2 in the supplemental material shows the relative risk of hyper-tension for age groups (18–34, 35–49, 50–64 and ≥ 65 years) by sex with covariate adjustments (for race/ethnicity and employment, smoking, alcohol intake and body mass index). For both men and women, the OR of having hypertension associated with short sleep decreases with increasing age, but there is a higher association between short sleep and hypertension for women, throughout the adult lifespan. The largest difference between men and women is in the youngest and oldest groups examined. Women between 18–34 and over 65 years have close to 0.6 higher OR of hypertension associated with ≤ 4 hours of sleep compared with men. Women were also at greater risk (about 0.5 OR) of hypertension than men due to long sleep (≥ 10 hours) in the 35–49 age category. Of note, in examining the confidence intervals, the OR for hypertension among those sleeping 4 or fewer hours is statistically significantly higher among women ages 18–34, 35–49, and ≥ 65 years and the OR for those sleeping 10 or more hours was also higher among women ages 35–49 and 50–64 years (P < .05).
Stratification by Body Mass Index, Smoking, and Alcohol
Table 3 displays results of covariate-adjusted analyses stratified by health factors. Both short and long sleep were associated with elevated hypertension prevalence for all BMI groups, though the relationships were nominally stronger among normal-weight individuals and for short sleep in general (Figure 5). Short and long sleep were associated with hypertension across smoking status groups, though results were nominally stronger among those who never smoked (Figure 5). Among those who never drank or were light drinkers, both short and long sleep were associated with elevated hypertension risk. However, among heavy drinkers, only the relationship with short sleep was present (Figure 5).
The present study combined data from the 2007–2017 NHIS and 2013 BRFSS to evaluate cross-sectional relationships between habitual sleep duration and prevalence of hypertension. Further, exploratory analyses examined how this relationship was modified by stratifying across several sociodemographic and health factors. Overall, the findings show a reverse J-shaped relationship between sleep duration and hypertension, with the strongest and most consistent associations with habitual short sleep, and observed (but less consistent) relationships with long sleep.
While other epidemiological studies have shown a stronger relationship between short sleep and hypertension in women than men,39,40 the current study is novel in stratifying the relationship between hypertension and sleep by age as well as sex. This analysis provided evidence that the reverse J-shaped relationship seen in the adult population as a whole is dampened with increasing age, particularly in men, where it evolved into a dampened U-shaped curve. Moreover, women showed a stronger risk for hypertension associated with short sleep, across the lifespan.
For adults of all ages combined, across most race/ethnicity groups, across employment categories, and across levels of health risk factors, sleep duration showed a reverse J-curvilinear relationship with incident hypertension, with highest risk associated with short sleep. Of note, associations were attenuated among the most obese and current smokers, and the association with long sleep was attenuated among heavy drinkers.
Younger men and women show little risk, and some apparent benefit, from sleep durations over 7 hours. The recent consensus statement from the AASM and SRS recommending 7 or more hours of sleep on a regular basis to support optimal health in adults did not put an upper limit on healthy sleep duration for a number of reasons. First, increased sleep duration is appropriate for young adults and those recovering from sleep debt or medical illness, thus sleep duration recommendations with an upper limit might discourage people in these circumstances from getting the sleep that they need. Second, the epidemio-logical data showing associations between sleep duration and untoward health outcomes is inconsistent in longer sleepers, whereas no such inconsistencies exist for short sleepers. Third, there is an absence of experimental studies showing adverse effects of sleep extension on human physiology, and there is at least one study showing that sleep extension may decrease 24-hour blood pressure9; whereas an abundance of studies show the adverse effects of experimental sleep restriction on human physiology and performance.3
For these reasons, some have hypothesized that associations between long sleep and poor health outcomes such as hypertension, when observed, represent confounding by health status. As previously mentioned, most studies addressing the sleep duration/health issue, including the present study, use self-report measures of sleep. Individuals with functional limitations due to chronic diseases live more sedentary lifestyles than healthy individuals41 and may spend more time inactive in bed, in a resting state. This time in bed could be misreported as sleep, thus confounding questionnaire-based studies.
The association between sleep and hypertension is addressed in the analysis in two ways. First, for the overall analysis following covariate adjustment, substantial attenuation of the long sleep/hypertension association is observed with smaller OR, with little to no attenuation observed for the short sleep/ hypertension relationship (Table 2 and Figure 1). In fact, OR increased following covariate adjustment for the short sleepers in the 6-hour and 5-hour sleep duration groups. To the extent that addressing confounders brings us closer to the truth in any analysis, this finding strengthens the short sleep/hypertension association and weakens the long sleep/hypertension association. Second, short sleep associations with hypertension held up better than long sleep associations in stratified analyses. Looking further, only unemployed long sleepers showed an association with hypertension, whereas nearly all employment categories showed an association between short sleep and hypertension (with the “unable to work” group being the only exception). For physical health, again the short sleep/hypertension associations remained regardless of health status, whereas for long sleepers, the OR generally increased as number of days of poor physical health increased. Lastly, the short sleep/ hypertension relationship existed regardless of functional status, whereas relationships with hypertension were more robust in the longer sleeping group with functional limitations supporting the notion that long sleep is not the cause but rather the consequence of deteriorated health.
This study has a number of limitations. First, these conclusions are based on subjective reports of sleep duration from a single question and not objective data, such as that obtained from supervised monitoring. Although self-reported sleep duration approximates objective measures of sleep length, some studies suggest it may be biased by overestimation.40 In addition, because this is a report based on self-report data from two cross-sectional surveys, precision and generalizability is compromised such that we cannot determine causal relationships or rule out bidirectional relationships. Lastly, healthy sleep requires not only adequate duration, but also good quality, occurrence at night, night-to-night regularity, and the absence or treatment of sleep disorders. The extent to which any of these additional factors is associated with hypertension is not ascertained in this analysis.
CONCLUSIONS AND FUTURE DIRECTIONS
We found very strong associations between subjective short and long sleep and hypertension in a large sample of United States adults. Adjusted analyses and stratified subanalyses showed the relationship between short sleep and hypertension to be stronger and more consistent than the association between long sleep and hypertension. Furthermore, short sleep was consistently more closely related to risk for hypertension in women compared with men, across the adult lifespan. Future prospective studies objectively assessing sleep duration, quality, timing, and regularity in individuals without sleep disorders and incident, objectively-measured hypertension would be important to provide additional information regarding the meaning of these associations. Future studies should also address the impact of circadian factors, as well as duration and quality on cardiovascular risk indicators. Technological advancements involving consumer sleep technologies and blood pressure assessment capable of assessing these factors over time in the participant's typical home environment are moving the field in this direction.42
Funding for this project was provided by the American Academy of Sleep Medicine and Sleep Research Society, and by cooperative agreement 1U50DP004930-04 from the Centers for Disease Control and Prevention (CDC). The findings and conclusions in this report are those of the authors and do not necessarily represent the official views of the CDC. The authors report no conflicts of interest. Ms. Hashmi is employed by the American Academy of Sleep Medicine.
1 World Health OrganizationGlobal Health Observatory (GHO) Data.
Accessed October 2016http://www.who.int/gho/ncd/risk_factors/blood_pressure_prevalence/en/.
3 Joint consensus statement of the American Academy of Sleep Medicine and Sleep Research Society on the recommended amount of sleep for a healthy adult: methodology and discussion. Sleep; 2015;388:1161-1183, 26194576.
10 Associations between sleep disorders, sleep duration, quality of sleep, and hypertension: results from the National Health and Nutrition Examination Survey, 2005 to 2008. J Clin Hypertens; 2011;1310:739-743.
14 Relationship between sleep duration and the metabolic syndrome: Korean National Health and Nutrition Survey 2001. Int J Obes; 2008;327:1091-1097.
30 Correlates of short and long sleep duration: a cross-cultural comparison between the United Kingdom and the United States: the Whitehall II Study and the Western New York Health Study. Am J Epidemiol; 2008;16812:1353-1364, 18945686.