Association between habitual sleep duration and blood pressure and clinical implications: A systematic review

Abstract

Elucidation of the association between short sleep duration and elevated blood pressure has implications for assessing and managing hypertension in adults. Objective. To assess the relationship between sleep duration and blood pressure, and its role in the etiology of hypertension. Methods. On a systematic search from MEDLINE, EMBASE, CINAHL, PEDro, PsychINFO and grey literature were included articles with participants over 18 years, reported sleep duration, measured blood pressure or diagnosed hypertension, and the relationship between sleep duration and blood pressure was analyzed. Results. Of 2522 articles initially identified, 11 studies met the inclusion criteria. Sample sizes ranged from 505 to 8860 (aged ≥ 20–98 years). Five studies (aged ≥ 58–60 years) determined that sleep duration and blood pressure were unrelated. In younger adults, five studies reported an association between short sleep duration and hypertension before adjustment for confounding variables; only the findings from one study remained significant after adjustment. Two studies supported a sex association; women who sleep less than 5–6 h nightly are at greater risk of developing hypertension. Conclusion. Sleep duration and blood pressure are associated in both women and adults under 60 years. Controlled studies are needed to elucidate confounding factors and the degree to which sleep profiles could augment diagnosis of hypertension and sleep recommendations to prevent or manage hypertension.

Key Words::

• blood pressure
• hypertension
• sleep
• sleep quality

Introduction

High blood pressure, estimated to cause 7.1 million deaths worldwide annually (1), is a risk factor for ischemic heart disease, cerebrovascular disease, and cardiac and renal failure (2). Over a quarter of the world's adult population was estimated to have hypertension in 2000; this proportion is predicted to rise to 29% by 2025 (3). In Canada, the 6-month healthcare cost for a single patient with hypertension exceeds $3000, with drugs accounting for more than 50% of the direct cost (4).

Although the need to integrate health promotion and disease prevention strategies into biomedical care is becoming better acknowledged in healthcare, some authorities argue that a substantial gap persists between knowledge and action (5). In line with this trend, lifestyle factors and their modification are being considered in preventing and managing hypertension (2); however, these are unlikely to be maximally exploited until the body of evidence supporting them is firmly established. Given the economic and social burdens associated with hypertension, the elucidation of the role of modifiable lifestyle risk factors in prevalent conditions such as hypertension is justified and highly compelling in terms of augmenting our knowledge about effective, low risk, economical approaches to its prevention and management.

The relationship between sleep and health has become an increasing focus of scientific investigation in part because of the impact of social factors on sleep patterns. Artificial lighting and changes in stress and social demands, for example, have contributed to changes in sleep patterns over time (6); the average American sleeps 6.7 h a night (7). Although the optimal amount of sleep that people need has been debated (6), a growing body of evidence supports the key role of sleep in overall health and wellbeing. Conversely, suboptimal sleep has been associated with a myriad of health issues including coronary heart disease (8), immune dysfunction (9), type II diabetes (10,11), obesity (12) and mood disorders (13).

Short-term sleep deprivation studies support that reduced habitual sleep increases blood pressure (14,15). Related to this body of evidence are the results of several population-based studies that conclude sleep apnea is an independent risk factor for hypertension (16,17), and a causal link has been proposed (18). This relationship remains to be firmly established for the general population. If such a relationship exists, sleep prescription and sleep hygiene counseling could be adopted as a practical, non-pharmaceutical and economical intervention to prevent or mitigate hypertension as well as promote general health and wellbeing.

The primary purpose of this study was to conduct a literature review to evaluate systematically the evidence associating habitual sleep duration with blood pressure in adults. If a firm association exists, our secondary purpose was to establish whether a direct relationship exists between habitual short sleep duration and hypertension.

Methods

Search strategy

We performed a systematic search of MEDLINE (1950 to present), EMBASE (1980 to present), CINAHL (1982 to present), PEDro, PsychINFO (1800 to present), and grey literature (ProQuest, Networked Digital Library of Theses and Dissertations, clinicaltrials.gov, Papers First). Initial searching was conducted in November 2009 with a final search in February 2010. The search strategies used MeSH headings and keywords for “hypertension” or “blood pressure” and “sleep”. In EMBASE and MEDLINE, these results were filtered with keywords for sleep duration: “duration*”, “quantit*”, “hour*”, or “amount*”. The searches were limited to the English language and adults (18+or 19+, depending on the database). Studies of children were excluded from the analysis given their sleep requirements vary across developmental stages (19). Relevant articles were searched on Web of Science (1900 to present) to find more recent literature citing these articles. A total of 2522 articles were identified for screening (Figure 1).

Figure 1. Study selection flow chart.

Inclusion and exclusion criteria

Studies were included if they (1) involved adult participants over 18 years or included separate analysis of those 18 years and older, vs those below this age; (2) reported participants’ habitual sleep duration; (3) measured blood pressure, or recorded a diagnosis of hypertension or use of antihypertensive medication; and (4) analyzed the relationship between habitual sleep duration and blood pressure.

Studies were excluded if they (1) exclusively analyzed populations with diagnosed sleep disorders, sleep disordered breathing, metabolic disease or cardiovascular disease; (2) were not available in full text; or (3) were not available in English.

Data extraction and quality assessment

All initial search results were screened independently by two reviewers for potentially relevant articles; irrelevant articles and duplicates were excluded. Full-text review was then conducted on the resulting 19 studies with respect to the specific inclusion and exclusion criteria. Each article was reviewed independently by the two reviewers who used a customized screening tool. The reviewers aimed for consensus for study selection; however, in the event of uncertainty or disagreement, a third person reviewed the article and discussed his or her assessment with the reviewers.

Methodological quality was assessed using the Downs and Black Quality Index (20). This index has been reported to be valid and reliable, and allows for assessment of both randomized and non-randomized studies. In an evaluation of this quality assessment tool (21), it scored high for internal consistency, test–retest, inter-rater reliability, criterion validity and respondent burden. The Downs and Black Quality Index is based on 27 items within five categories: study quality/reporting (10 items); external validity (three items); internal validity – bias (seven items); internal validity – confounding/selection bias (six items); and power (one item). The highest achievable score on the Downs and Black Quality Index is 32; however, when it is applied to non-randomized studies, the maximum score is reduced to 19 (20). Each review extracted independently date from articles and summarized as study, country, study size, study design, subject age, quantity of sleep and main findings.

Results

Study selection

A flow chart of the study search and selection process appears in Figure 1. Our initial search strategy yielded 2522 articles from six resources: 990 from MEDLINE; 1038 from EMBASE; 141 from CINAHL; 138 from PsycINFO; 212 from grey literature sources; and three from the Web of Science. After our stringent selection process, 11 source studies resulted.

Quality assessment

For the source studies, quality assessment scores derived from the Downs and Black Quality Index appear in Table I. The average score across all 11 studies is 15.82 ± 2.79.

Table I. Quality assessment scores based on the Downs and Black Quality Index of the source studies.

		Study quality	External validity	Internal validity	Total score
Study	Year	(/ll)	(/3)	(/13)	(/32)
Bjorvatn et al. (23)	2006	8	0	5	13
Cappuccio et al. (30)	2007	9	0	8	17
Choi et al. (24)	2008	9	2	5	16
Gangwisch et al. (29)	2006	9	2	8	19
Hall et al. (25)	2008	9	2	5	16
Kawada et al. (26)	2008	7	0	4	11
Knutson et al. (31)	2009	9	2	8	19
Lima-Costa et al. (22)	2008	7	2	3	12
Lopez-Garcia et al. (32)	2009	9	2	8	19
Stranges et al. (27)	2008	8	1	7	16
Van den Berg et al. (28)	2007	9	2	5	16

Each paper was given a score of 0/5 on the power sub-scale.

Participants and study design

Several types of study designs were reflected in the source studies, and were mostly limited to older adults (Table I). Therefore, our results reflect associations between habitual sleep duration and blood pressure across the adult lifespan.

This review incorporates data from multiple countries, included both sexes (except one study) and most had large sample sizes (Table II).

Table II. Description of the samples of the source studies.

First author	Year	Country	Study design	Study size	Subject age	Population source
Bjorvatn et al. (23)	2007	Norway	Cross-sectional	8860	40–5	Hordaland Health Study data (population-based, all born in Hordaland county 1953–1957)
Cappuccio et al. (30)	2007	UK	Cross-sectional Longitudinal (mean 5-year follow-up)	5766; Men: 4199; Women: 1567 3691; Men: 2686; Women: 1005	35–55	Whitehall II Study, white British civil servants; Excluded: ethnic groups other than white (n = 612) Normotensive population from cross-sectional study
Choi et al. (24)	2008	Korea	Cross-sectional	4222	>20	Portion of Korean National Health and Nutrition Survey (nationally representative survey); Excluded: those on medications for metabolic syndrome, missing data on sleep or fasting time, not enough time fasting
Gangwisch et al. (29)	2006	USA	Longitudinal (8-10-year follow-up)	4810	25-74	National Health and Nutrition Examination Survey (civilian, non-institutionalized US citizens); Excluded: those with diagnosed hypertension or blood pressure > 140/90 mmHg at baseline
Hall et al. (25)	2008	USA	Cross-sectional	1214	30-54	University of Pittsburgh's Adult Health and Behavior (AHAB) registry (Southwestern Pennsylvania, principally Allegheny county); Excluded: those with atherosclerosis, chronic kidney or liver disease, neurological disorder, schizophrenia or psychotic illness, pregnancy and the use of insulin, glucocorticoid, antiarrhythmic, psychotropic or weight-loss medications
Kawada et al. (26)	2008	Japan	Cross-sectional	4941	36-60	Male employees of an electronics company; Excluded: past medical history of hypertension, dyslipidemia, and/or diabetes mellitus
Knutson etal. (31)	2009	USA	Cross-sectional Longitudinal (5-year follow-up)	578 505 (535 for incident hyper-tension analysis)	33–45	Chicago participants of CARDIA (Coronary Artery Risk Development in Young Adults) Study, white and African American sub-populations; Excluded: pregnant women, those missing actigraphy data, taking antihypertensive medications at baseline, or missing a measure of DBP at baseline or follow-up Excluded: as above + those on antihypertensive medications at follow-up
Lima-Costa et al. (22)	2008	Brazil	Cross-sectional	1423	60-95	Bambui Health Aging Study (community-based cohort in Bambui city; 81.7% of population >60 years old)
Lopez-Garcia et al. (32)	2009	Spain	Cross-sectional Longitudinal (2-year follow-up)	3686 890	>60	Population-based cohort representative of non-institutionalized Spanish population >60 years old Normotensive population from cross-sectional study
Stranges et al. (27)	2008	UK; USA	Comparative cross-sectional	UK: 6472; USA: 3027	UK: 45–69; USA: 35–79	Whitehall II Study (white British civil servants); Western NY Health Study (Erie or Niagara County, NY; white population subset); Excluded: incomplete data, ethnic groups other than white
van den Berg et al. (28)	2007	Netherlands	Cross-sectional	5058	58–98	Rotterdam Study (population-based cohort study, all >55 in district of Rotterdam invited to participate)

Measurement of habitual sleep and blood pressure

Table III compares studies with respect to measurement of habitual sleep and blood pressure measurement. All 11 source studies collected data on sleep quantity from self-report through means of interview or questionnaire. Two studies used wrist actigraphy in addition to self-report to measure sleep (28,31). Studies were inconsistent regarding attention to the time of day for blood pressure measurement.

Table III. Measurement of sleep and blood pressure variables.

Study	BP measurement technique	Calculation of BP	Definition of htn or high BP	Method of sleep data collection	Specific question posed and/or technique used to analyze sleep data
Bjorvatn et al. (23)	1. Relaxed sitting position following 2 min rest; 2. Three automated BP measurements; 3. No talking during measurements	Averaged final 2 of 3 readings	Defined high BP as SBP >140 mmHg orDPB >90 mmHg	Self-report via self-administered questionnaire	Sleep duration calculated as total time in bed (from bedtime to rise time) minus self-reported sleep latency; Calculated sleep during workweek and during free time independently
Cappuccio et al. (30)	1. Three BP measures using standard mercury manometer; 2. Onset of 1st and 5th Korotkoff sounds used to represent SBP and DBP	Averaged final 2 of 3 readings	Defined htn as SBP/ DBPa 140/90 mmHg, or regular use of antihypertensive medication	Self-report questionnaire	"How many hours of sleep do you have on an average week night?"
Choi et al. (24)	1. In seated position following 10 min rest; 2. Two SBP and 2 DBP measurements with a 5 min interval between readings	Average of 2 readings	Defined htn as SBP or DBP > 130/85 mmHg	Self-report questionnaire	N/A
Gangwisch et al. (29)	1. Three BP measurements taken at the same sitting; 2. No reading taken if maximum cuff inflation >160 mmHg	Measured 3 times	Defined htn as SBP > 140 mmHg or DPB > 90 mmHg, or self-report of physician diagnosis, hospital record, or reported cause of death	Self-report questionnaire	“How many hours of sleep do you usually get a night (or when you usually sleep)?”
Hall et al. (25)	1. Two measurements taken consecutively on morning after a 12-h fast;2. BP measured in a seated position following 10 min rest; 3. Mercury sphygmomanometer used on right arm	Average of 2 readings	Defined high BP as SBP > 130 mmHg, or DBP > 85 mmHg, or use of antihypertensive medication	Self-report through in-person interview conducted by trained interviewers	Hours per night slept over the past 7 days; asked separately about sleep duration on weeknights (Sunday-Thursday) and weekend nights (Friday, Saturday); Weekly average reported sleep duration calculated as the weighted average of weeknight and weekend values
Kawada et al. (26)	1. Right arm in seated position after 5 min rest; 2. Used oscillometric sphygmomanometer; 3. 1st and 5th Korotkoff sounds to indicate SBP and DBP, respectively	1. If initial reading was under 130/85 mmHg, then BP was not re-measured; 2. Measured twice after 3 min interval if initial readings 130/85 mmHg; lowest reading retained	Defined high BP as SBP > 130 mmHg and/or a 85 mmHg	Self-report	N/A
Knutson etal. (31)	1. Readings taken three times after 5 min rest; 2. Used Hawksley random-0 sphygmomanometer and digital BP monitor which were calibrated against each other for consistency	Averaged final 2 of 3 readings	Defined htn as SBP >140 mmHg, or DPB >90 mmHg, or use of antihypertensive medication	Self-report during clinical exam and wrist actigraphy monitor	1. Actigraphy for three consecutive days on 2 occasions approximately 1 year apart; 2. Self-reported questions about sleep duration and quality during clinical exam
Lima-Costa et al. (22)	N/A	Average of 2 out of 3 readings	Defined htn as SBP >140 mmHg, or DBP >90 mmHg, or use of antihypertensive medication	Self-report	1. “During the last month, in the workdays and without considering the weekends, at what time did you: lay down to sleep; get asleep; wake up; leave the bed?”; 2. Weekend sleeping habits determined by use of similar question to above; 3. Usual sleep duration determined by formula
Lopez-Garcia et al. (32)	1. Measured 6 times at level of heart in right arm at 2 min intervals; 2. Measured after 5 min rest in seated position using mercury sphygmomanometer	Average of 6 readings	Defined htn as SBP >140 mmHg, or DBP >90 mmHg, or use of antihypertensive medication, or self-reported physician diagnosis	1. Self-report through home based personal interview at baseline; 2. Phone interview at follow-up	“How many hours do you usually sleep per day (including sleep at night and during the day)?”
Stranges et al. (27)	1. Sitting position using mercury sphygmomanometer; 2. 1st and 5 th Korotkoff sounds used for SBP and DBP, respectively	Averaged final 2 of 3 readings	Defined htn as SBP/ DBP > 140/90 mmHg, or regular use of antihypertensive medication	Self-report questionnaire	Analyzed data from two sources where the following questions were posed: 1. Whitehall II study: “How many hours of sleep do you have on an average weeknight?”; 2. Western NY study: “On the average, how many hours did you sleep each night during the last 5 weekday nights (Sunday-Thursday)?”
van den Berg et al. (28)	1. Two blood pressure measurements in sitting at right brachial artery with random-0 sphygmomanometer	Average of 2 readings	Defined htn as SBP> 160 mmHg, or DBP> 100 mmHg, or current use of antihypertensive medication	Self-report by way of in-person home interview, and wrist actigraphy	1. “During the past month, how many hours of actual sleep did you get at night?”; 2. Wrist actigraphy over 5-7 consecutive nights; actigraphy algorithm used on raw data

BP, blood pressure; SBP, systolic blood pressure; DBP, diastolic blood pressure.

There were also marked inconsistencies across studies regarding the qualifications and experience of personnel taking blood pressure, the body position and resting state of the subjects prior to taking blood pressure, arm the blood pressure was taken, and whether blood pressures were averaged (Table III).

Table III also shows detailed information on the blood pressure and sleep data as well as hypertension definitions. Hypertension was defined inconsistently across the source studies.

Main findings

There was marked variability among the source studies in how they represented and reported their results (Table IV). Studies generally compared sleep duration with a reference group whose nightly sleep was approximately 7 h, though the exact parameters varied. Most studies adjusted their findings for multiple confounding variables. The significance of the relationship between sleep duration and blood pressure often changed with adjustments suggesting a role for mediating variables.

Table IV. Adjusted variables and main findings of the source studies.

First author, year	Study design	Sleep reference (h/day)	Data analysis/adjusted variables	Main findings
Bjorvatn et al. (23)	Cross-sectional	7-7.99	Hierarchical linear regression analysis: Step 1: unadjusted sleep duration; Step 2: gender, smoking, sleep duration free time; Step 3: BMI; Crude logistic regression analyses using sleep duration (workweek) as the predictor for systolic blood pressure (>140 mmHg vs si40 mmHg) and diastolic blood pressure (>90 mmHg vs £90 mmHg); Adjusted logistic regression analysis controlling for gender, smoking, BMI	Unadjusted, both SBP and DBP were higher with shorter sleep durations; Controlled, these associations were no longer significant; Sleep duration 5-5.99 h and 6-6.99 h became significantly negatively related to SPB; SBP: no relationship between sleep duration and SBP >140 mmHg; DBP: <5 h significantly related to DBP >90 mmHg; SBP: 5-5.99 h and 6-6.99 h associated with decreased risk of high SBP; DBP: no duration significantly related
Cappuccio et al. (30)	Cross-sectional Longitudinal (5-year follow-up)	7–<8	Reduced model: adjusted for age and employment; Fully adjusted: as above plus alcohol consumption, smoking, physical activity, BMI, cardiovascular disease drugs, Short Form-36 mental and physical health score, depression, use of hypnotics As above	No consistent pattern of association in men; Significantly higher prevalence of htn in women sleeping £5 h in fully adjusted model (OR= 1.72 [95% CI: 1.07-2.75],p = 0.037), with a significant inverse linear trend across decreasing hours of sleep (p = 0.003); Consistent, significant inverse associations (p <0.05) between duration of sleep and either SBP or pulse pressure only among women, in fully adjusted models No consistent pattern of association in men; Significantly higher incidence of htn in women sleeping £5 h in unadjusted and reduced model analysis (6 h: OR= 1.56 [95% CI 1.07–2.27], <5 h: OR= 1.94 [95% CI 1.08–3.50]; Results not significant after adjustment for cardiovascular risk factors and psychiatric co-morbidities
Choi et al. (24)	Cross-sectional	7	Unadjusted	U-shaped relationship between sleep and htn for subjects under age of 60, lowest incidence at 7 h; BP significantly differed (p = 0.009) with sleep duration £5 h for subjects under 60; No difference for subjects over 60 0 = 0.623)
Gangwisch et al. (29)	Longitudinal (8-10-year follow-up)	7–8	Model 1: Unadjusted; Model 2: daytime sleepiness, depression, physical activity, alcohol consumption, salt consumption, smoking, pulse rate, gender; Model 3: model 2 + education, age, ethnicity; Model 4: model 3 + obesity, diabetes	Unadjusted: significantly increased risk of diagnosis in all sleeping £5 h with younger age group (32-59) (HR=2.10 [95% CI 1.58–2.79]); htn in older adults 60-86 associated with sleeping >9 h HR= 1.54 [1.03–2.30]; Fully adjusted: relationship attenuated but still significant for £5 h, especially young (32–59) HR= 1.60 [1.19–2.14]
Hall et al. (25)	Cross-sectional	7–8	Adjusted for age, sex, race, educational attainment, smoker, physical activity, low-density lipoprotein cholesterol and symptoms of depression	No relationship between sleep and mean BP in unadjusted or adjusted analysis
Kawada et al. (26)	Cross-sectional	<6 vs >6	Adjusted for drinking, physical activity, sleep duration, eating breakfast, snacking frequency, BMI	Sleeping >6h not significantly associated with high BP (OR= 1.12 [0.90–1.31])
Knutson et al. (31)	Cross-sectional Longitudinal (5 years)	Continuous variable	Partial adjustment: age, sex, race; Full: partial + snoring, day time sleepiness, income, education, smoking status, BMI, physical activity, alcohol use, AND 5-year change in smoking, alcohol, BMI, and physical activity As above	Partial adjustment: shorter sleep durations significantly associated with higher SBP (p = 0.006) and DBP (p = 0.001); Full adjustment: no significant association between sleep duration and SBP (p = 0.07) or DBP O = 0.09) Partial adjustment: short sleep associated with smaller decreases or increases in SBP (p = 0.02), DBP (p= <0.001), and increased odds of incident htn (OR=1.37 [95% CI 1.05–1.78]): 1-h sleep reduction = 37% increase in odds; Full adjustment: sleep duration not associated with change in SBP (p = 0.09); significantly associated with change in DBP (p = 0.03); increased odds of htn attenuated (OR=1.3 [0.96–1.75])
Lima-Costa et al. (22)	Cross-sectional	7-<8	Model 1: unadjusted; Model 2: age, gender; Model 3:2 + skin color, diabetes mellitus, depressive symptoms, body mass index, and hypnotic or sedative meds	Usual sleep duration not associated with prevalent htn in unadjusted or adjusted models
Lopez-Garcia et al. (32)	Cross-sectional Longitudinal (2-year follow-up)	7	Adjusted for age, sex, physical activity, BMI, smoking, alcohol consumption, coffee, consumption, education level, number of social ties, perceived health, depression, number of chronic disease, arousal from sleep at night, anxiolytic intake As above	No association between sleep duration and prevalent htn, before or after adjustment, or using a htn cut-off of 160/100 mmHg or systolic htn only; No association between sleep duration and SBP, DBP, or pulse pressure No association between sleep duration and incident htn; Result unchanged when analysis repeated for men and women separately
Stranges et al. (27)	Comparative cross-sectional	6-8	Analyzed correlates of sleep duration including marital status, socioeconomic status, BMI, current smoker, current alcohol, low vs. high physical activity, Short Form-36 score, depressive symptoms, diabetes mellitus); adjusted odds ratio for age and sleep	Significant relationship between hypertension and short sleep duration (<6h) in women in theWNYH study in models adjusting for only age and sex OR: 1.80 [1.32–2.47] p= <0.001; In fully adjusted models, significant association of short sleep duration and htn in women, in both studies OR =1.70 [WNY 1.07-2.70 £ = 0.02; Whitehall 1.13-2.56 £ = 0.01]
van den Berg et al. (28)	Cross-sectional	7–< 8 for self-report; 6-<7 for actigraphy group	Model 1: age, gender; Model 2: model 1+BMI, smoking, depression symptoms, sleep medication use, diabetes mellitus, myocardial infarction, stroke, daytime napping	No association between sleep and htn, even using a lower cut-off value for htn; In multivariable linear analysis, sleep duration was not associated with SBP or DBP

htn, hypertension; SBP, systolic blood pressure; DBP, diastolic blood pressure; OR, odds ratio; HR, hazard ratio; CI, confidence interval.

Age-related findings. Several studies exclusively analyzed older adults (22,28,32). Each of these studies concluded that sleep duration is not associated with prevalent hypertension in this population before or after adjustment for confounding variables (Table IV). Two studies analyzed age groups separately and provided data for those over 60 years of age (24,29). One reported no association between sleep duration and elevated blood pressure for subjects over 60 years of age (24). The other study reported an increased likelihood of hypertension in older adults (60–85 years of age), sleeping 9 h or more (hazard ratio (HR): 1.54; 95% confidence interval, CI 1.03–2.30) (29). This relationship was no longer significant, however, when the model was adjusted for daytime sleepiness, depression, physical activity, alcohol and salt consumption, smoking, pulse rate and gender (29).

One study reported that the risk of being diagnosed with hypertension was twice as high in subjects under 60 years of age who slept 5 or fewer hours nightly (HR = 2.10; 95% CI 1.58–2.79) (29). Adjusting for variables modestly attenuated this relationship but it remained significant for subjects between 32 and 59 years of age who slept 5 or fewer hours nightly (HR = 1.60; 95% CI 1.19–2.14). In an unadjusted analysis, another study reported that the prevalence of elevated blood pressure was significantly higher (p = 0.009) with sleep duration ≤5 h for subjects between 20 and 59 years of age (24). In a study of adults between 33 and 45 years of age, short sleep duration was associated with higher blood pressure when adjusted for age, sex and race (31). In a fully adjusted model, the association was no longer significant. At a 5-year follow-up, short sleep duration was associated with incident hypertension and each hour of sleep reduction increased the odds by 37%. Similarly, the relationship was attenuated with full adjustment. Interestingly, a significant association with changes in diastolic blood pressure remained (p = 0.03).

In an unadjusted analysis for a population between 40 and 45 years of age, sleep duration of <5 h was positively related to blood pressure (23). After full adjustment, however, sleep durations of 5–5.99 h and 6–6.99 h became negatively associated with systolic blood pressure.

One study reported no relationship between short sleep (<6 h vs 7–8 h) and elevated blood pressure in participants who were between 30 and 54 years of age (25). Another study compared <6 h with ≥6 h for men 36–60 years of age and reported no significant difference with respect to exhibiting higher blood pressure (26). Two studies reported no relationship between sleep and blood pressure in participants under 60 years of age (25,26).

Sex-related findings. One study reported that short sleep duration (≤5 h) was associated with hypertension only in women in a fully adjusted model (odds ratio, OR = 1.72; 95% CI 1.07–2.75) (30). This relationship remained significant after a 5-year follow-up in unadjusted analyses, and when adjusted for age and employment (6 h a night OR = 1.56; 95% CI 1.07–2.27; 5 h per night OR = 1.94; 95% CI 1.08–3.50). However, the associations were no longer significant after accounting for cardiovascular risk factors and psychiatric co-morbidity (OR = 1.42; 95% CI 0.94–2.16; OR = 1.31; 95% CI 0.65–2.63, respectively). No association between sleep duration and blood pressure was detected in men in cross-sectional or longitudinal analysis.

One study conducted a cross-cultural comparison between the Whitehall II study (UK) and the Western New York Health (WNYH) study for correlates of sleep duration (27). They reported a relationship between hypertension and short sleep duration (<6 h) in women in the WNYH study in models adjusting for only age and sex (OR = 1.80; 95% CI 1.32–2.47, p < 0.001). In the fully adjusted models, there was an association in both studies (OR = 1.70; WNYH 95% CI 1.07–2.70, p = 0.02; Whitehall II 95% CI 1.13–2.56, p = 0.01). There was no association reported for men.

A Japanese study focusing on metabolic syndrome examined the relationship between sleep duration and hypertension in young male workers (between 36 and 60 years of age) (26). This study used <6 h of sleep as its reference group, and observed that sleeping ≥6 h a day was not associated with hypertension (OR = 1.12; 95% CI 0.90–1.31).

Several studies analyzed data from elderly men and women and reported that sleep duration and blood pressure were not associated in either sex (28,32). In a young middle-aged population (between 33 and 45 years of age), one study reported that the association between sleep and blood pressure was comparable for men and women (31).

Discussion

Among the studies we reviewed, there was heterogeneity in sample sizes, methods of measuring blood pressure and sleep duration, data collection, and the control of confounding variables. Despite these discrepancies, two relationships emerged that highlighted age- and sex-specific effects.

With respect to the age-related association between habitual short sleep duration and blood pressure, the evidence supported an association between the likelihood of hypertension among individuals <60 years of age who slept ≤5 h each night (24,29,31). In adjusted models, short sleep duration and hypertension were unrelated in people ≥60 years old (22,24,28,29,32). Table I shows that four of the five studies (24,28,29,32) reporting age-related findings (either before or after attenuation of variables) scored either 16 (24,28) or 19 (29,32) on the Downs and Black Quality Index, which were among the highest scores for quality assessment, while one scored 12 (22). The quality of these studies adds credibility to their results.

Mechanisms have been proposed for an age-related association between habitual sleep duration and blood pressure. First, lifestyle changes associated with retirement in high-income countries from where the source studies originated may allow for more opportunities for relaxation to compensate for short night-time sleep durations (28). Second, individuals with disorders such as hypertension, obesity and diabetes are less likely to survive into their later years (29).

The second main relationship was sex-specific, where women who slept <5–6 h each night were at higher likelihood of hypertension (17,30). The precise mechanism explaining this sex-specific relationship is not known. One explanation may be the role of hormonal and psychosocial changes associated with menopause (30). Studies have shown that there is no relationship between short sleep duration and hypertension in women >60 years (28,32). In our systematic review, the strongest relationship between habitual short sleep duration (≤5 h or <6 h) and hypertension seems to be among younger women between 30 and 60 years of age.

Our findings are of interest because they are relevant to diverse populations with chronic conditions. Given the variability among studies, our results may be applicable to adults living in a range of countries. Most studies collected data from populations in the northern hemisphere; thus, other populations are underrepresented. Also, the relationship between short sleep duration and blood pressure is consistent with research supporting that short sleep duration has effects on lifestyle-related conditions such as metabolic syndrome and obesity (23–26,30). Hence, these findings could inform the practices of a number of health care practitioners who could address sleep as a modifiable risk factor in mitigating these prevalent and costly chronic conditions.

The relationship between habitual sleep duration, hypertension and other systemic conditions is complex; isolating one condition from another when interpreting data is difficult. Sleep-disordered breathing is associated with short sleep duration, and hypertension is highly co-morbid with sleep apnea (34,35). The metabolic syndrome and its components also have been associated with sleep duration (23,24), and each component is likely interrelated; the relationship between obesity and sleep duration is generally well established (36). Obesity and hypertension, however, are often related co-morbid conditions (37). Furthermore, some conditions associated with habitual short sleep duration may only be detected after a period of latency. Given the complexity of these relationships, to determine directionality warrants detailed study.

Although the effect of sleep deprivation on neural cardiovascular control, especially muscle sympathetic nerve activity, is controversial (14), several mechanisms can be proposed by which habitual short sleep duration could elevate blood pressure. First, short-term sleep deprivation has been reported to increase the activity of the sympathetic nervous system. The synthesis of catecholamines is augmented via activation of the suprachiasmatic nucleus (15,38). This, in turn, leads to vessel constriction, subsequent increases in blood pressure and adverse vascular remodeling (39). Second, blood pressure and heart rate naturally follow a diurnal pattern. They are lower during sleep; thus, prolonged periods of wake time could predispose an individual to unnatural increases in blood pressure and heart rate with less reprieve time (40). More waking time may also lead to greater opportunity to engage in unhealthy behaviors that influence blood pressure (e.g. caffeine consumption, poor dietary habits, exposure to emotional stress), and that stimulate arousal and negatively impact the cardiovascular and cerebrovascular systems (40,41).

The quality of our results directly reflects the quality of the source studies. Use of the Downs and Black Quality Index for quality assessment provided a profile of each study alerting the reviewers to its methodological strengths and weaknesses (20). In interpreting the data, higher credence was given to higher scoring articles. While this index is useful in examining both observational and experimental studies, the scores of the studies in our review were low; the index has questions pertaining to experimental methods, however we reviewed strictly observational studies. In the assessment of longitudinal, non-randomized studies, the maximum achievable score is 19 (20). Common among the studies was the lack of randomization, and blinding of researchers and participants. Ultimately, these methodological limitations negatively impacted scores on the Downs and Black Quality Index.

All the source studies were cross-sectional and/or longitudinal in design. Cross-sectional studies allow for comparison of variables at the same time points, while longitudinal studies allow for the reporting of more observations over time. Given their observational nature, however, they are more susceptible to bias than experimental studies and cannot establish causality.

The methodology across source studies was inconsistent. Commonly, blood pressure was measured by trained professionals using standard sphygmomanometers. The timing and averaging of measurements, however, varied across studies. Studies also differed with respect to their definitions of high blood pressure or hypertension (Table III); this limitation could have affected the recorded prevalence or incidence of high blood pressure or hypertension. Another limitation across studies was the use of self-report as a measure of habitual sleep duration. However, good agreement has been reported in the literature between self-reported sleep duration, and actigraphic monitoring or the gold standard polysomnography (40,43). There were also discrepancies among studies as to the confounding variables for which investigators chose to control.

Blood pressure measurement was consistent in one respect and that was related to blood pressure being taken by someone other than the subject. However, that home and ambulatory blood pressures are more valid is a consideration for future studies (44).

One final limitation is reflected in the interpretation of the data. In eight of the 11 articles reviewed, approximately 21% of study sample sizes comprised the extremes (<6 h and >8 h) of reported habitual sleep duration (22,23,25,28–32). Thus, the applicability of our main findings of this systematic review needs to be interpreted cautiously.

Growing evidence supports that short habitual sleep duration is associated with metabolic disorders such as obesity, diabetes and hypertension (33). If sleep duration is a modifiable risk factor for hypertension and subsequently cardiovascular disease, then formal assessment of sleep quality and quantity warrants being assessed and potentially managed as conscientiously as an individual's functional capacity. Given there are compelling explanations associating reduced habitual sleep duration and blood pressure, interventions involving sleep hygiene whilst visiting a health care professional could serve to blunt this rising trend of high blood pressure and its sequelae, not to mention the importance of restorative sleep on health and function.

Representative prospective studies with adequate sample size and long-term duration are indicated. These should include repeated objective measures of blood pressure and habitual sleep duration, and control for confounders to help build evidence of causality. Although conducting randomized controlled trials that impose short sleep duration on healthy individuals is unethical, observation of the effect of longer sleep durations in sleep-deprived individuals with hypertension could be useful to help elucidate this relationship.

The following variables are known to affect blood pressure: age, sex, race, physical activity, obesity, diet, caffeine intake, alcohol intake, salt and potassium intake, type II diabetes mellitus, smoking, socioeconomic status, mental status (as measured by Short Form-36), family history of hypertension (45). Future research needs to standardize and control these variables. Various exclusion criteria should also be considered when designing future studies related to sleep. The following are recommended for exclusion from future studies as they may interfere with identifying a directional relationship: sleep disorders (e.g. dyssomnias and parasomnias), sleep disorders related to mental and/or neurological conditions (e.g. psychoses and dementia), sleep disordered breathing (e.g. apneas), psychological disorders, cancer, and cardiovascular, respiratory or metabolic diseases (45).

We recommend that future studies examining the relationship between habitual sleep duration and hypertension incorporate consistent methodology. Blood pressure measures, for example, should be taken at a consistent time each day using a standardized technique. Sleep duration, if not measured objectively, should be assessed in a standardized manner such as with a self-report sleep diary.

To conclude, habitual sleep duration and blood pressure are associated specifically in women and adults under 60 years of age. Controlled studies are needed to elucidate the factors that apparently confound this relationship, and the degree to which profiles of patients’ habitual sleep could augment the diagnosis of hypertension and inform sleep recommendations to prevent or manage hypertension.

Acknowledgments

The investigators acknowledge the support and assistance of Charlotte Beck, Dr Michael Bodner, Dr Lara Boyd, Dr Stanley Coren, Dean Giustini, Dr Teresa Liu-Ambrose and Dr Darlene Reid.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.