Are Sleep Symptoms Predictors of Resistant Hypertension in a Population-Based Sample? Findings From the National Health and Nutritional Examination Survey
Harneet Walia, MD, Assistant Professor, 11100 Euclid Avenue, Wearn 612, Cleveland, OH 44106
J Clin Hypertens (Greenwich). 2012;00:000–000. ©2012 Wiley Periodicals, Inc.
The aim of this study was to test the association of self-reported sleep symptoms to those identified with severe hypertension in a nationally representative sample of adults. Self-reported and study-measured health and sleep characteristics were collected by the National Health and Nutrition Examination Survey (NHANES) from 2005 to 2008. Of 10,526 individuals with completed sleep surveys participating in the study, the authors identified 379 patients with severe hypertension defined as those treated with ≥3 antihypertensive medications including a diuretic; 110 of these had resistant hypertension (RHTN) despite therapy, while 269 were controlled for severe hypertension (CSHTN). Patients with RHTN were more likely to be married, less educated, smoke, and self-report unsatisfactory health and diabetes when compared with patients with CSHTN. Multivariate analyses showed that poorly controlled diabetes (glycated hemoglobin >7%) was the strongest predictor of RHTN (odds ratio, 3.0; 95% confidence interval, 1.2–7.9). Unsatisfactory health (odds ratio, 1.7; 95% confidence interval, 1.7–2.7) was also associated with RHTN. Poorly controlled diabetes and self-reported unsatisfactory heath showed significant association with RHTN. Contrary to expectations, there was no significant association between self-reported snoring/snorting and RHTN, when other factors were examined. The association between poorly controlled diabetes and RHTN warrants further emphasis on strict control of diabetes in these individuals.
Hypertension (HTN) is a highly prevalent chronic condition (29%–31%) in the United States and is reported as insufficiently controlled in up to two thirds of individuals.1,2 Resistant HTN (RHTN) is defined as blood pressure (BP) values in excess of 140/90 mm Hg or 130/80 mm Hg in the presence of diabetes or chronic kidney disease despite treatment with ≥3 antihypertensive medications including a diuretic.3 The prevalence of RHTN in the US adult hypertensive clinic population is estimated to be as high as 15% to 20%.4 Known risk factors associated with RHTN include low socioeconomic status and behavioral factors including a potentially complex medical regimen.5 Obesity, a high-salt diet, physical inactivity, and heavy alcohol intake may alone or in combination contribute to poorly controlled high BPs. In one study that switched from a high- to a low-salt diet was associated with an average reduction in office BP of 23/9 mm Hg in patients with RHTN.6 Hence, identifying and addressing not only nonphysiologic as well as clinically modifiable factors may be important.
The role of sleep symptoms including snoring, snorting, and poor or insufficient sleep in the etiology of RHTN is of interest for a number of reasons. First, obstructive sleep apnea (OSA) or insufficient sleep both commonly present in the population and can be a predictor of future HTN in normotensive individuals.7–9 Second, both OSA and its clinical marker of chronic, frequent snoring have been shown to be associated with RHTN in hospital-based case series; both of these entities are readily treatable.10–12
The National Health and Nutritional Examination Survey (NHANES) was designed as a public health survey tool to detect and monitor disease and illness over time. Questions have been recently added to assess the potential relationships among sleep disorders, sleep duration, sleep quality, and prevalent conditions such as HTN and diabetes. Prior overview surveys found increased risk in hypertensive vs nonhypertensive adults having concurrent sleep disorders with poor or short sleep.13 The current study looks more closely at this database in regard to hypertension control with the aim of assessing whether these same self-reported sleep symptoms predict RHTN in those identified with severe hypertension.
Data were obtained from the 2005–2008 NHANES, conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention (CDC). NHANES is a cross-sectional stratified multistage probability sample of the civilian noninstitutionalized US adults aged 18 years and older. Complete details on recruitment, design, and surveys used are described on the NHANES Web site (http://www.cdc.gov/nchs/nhanes.htm).14
Sleep variables were obtained by merging the sleep-related data collected by questionnaire with files containing information concerning demographics, health insurance, health status, medical history, depression screening, body measurements, physical activity, alcohol use, food frequency and drug use questionnaire, cholesterol screening, and current medications by respondent sequence number. The NHANES study was approved by a CDC human subjects committee. Since all data in the NHANES database is de-identified, the study obtained exemption by University Hospitals Case Medical Center Institutional Review Board.
A total of 3306 adults in the sample were deemed as having hypertension from a cohort of 10,526 patients with complete sleep symptoms data available. HTN was defined based on either self-report (HTN previously diagnosed by a physician) or if the NHANES-measured BP readings were >140/90 mm Hg, or 130/80 mm Hg if patients were diabetic or had chronic kidney disease. Diabetes and chronic kidney disease were defined by self-report. BP measurements were collected at the time of the interview using a mobile examination center where two physicians obtained 3 consecutive readings of BP for each participant, with 5 minutes of rest in a sitting position between each reading. A total of 379 patients were identified as having severe hypertension based on treatment with a minimum of 3 antihypertensive medications including a diuretic at the time of the data collection. Initially self-reported, this information was validated by direct medication checkup by NHANES personnel at the time of data collection.
Outcome Variable: RHTN
The sample population was dichotomized into controlled severe hypertension (CSHTN) and RHTN according to collected BP measurements in the NHANES database. CSHTN was defined as having a BP ≤140/90 mm Hg and ≤130/80 mm Hg in patients with diabetes and chronic kidney disease on treatment with ≥3 antihypertensive medications including a diuretic. Patients were considered to have RHTN if ≥1 of the 3 collected BPs were ≥140/90 mm Hg or ≥130/80 mm Hg in patients with diabetes or chronic kidney disease despite therapy with ≥3 BP medications, of which one was a diuretic.
Predictor Variables: Sleep Characteristics
Symptoms of OSA included (1) self-report of having physician-diagnosed OSA, (2) self-report of symptoms of snoring ≥3 times per week in the past 12 months, (3) self-report of symptoms of snorting ≥3 times per week in the past 12 months, (4) combined symptoms of sleep apnea of snoring or snorting ≥3 times per week in the past 12 months, (5) severe sleep apnea based on symptoms defined as symptoms of snoring or snorting ≥5 times per week in the past 12 months, and (6) severe combined sleep apnea defined as symptoms of snoring and snorting ≥5 times per week in the past 12 months.
Sleep duration was collected as continuous variables (in hours) based on response to the question: “How much sleep do you usually get at night on weekdays or workdays?” Two different cut points were used for defining short habitual sleep time (SHST) (dichotomized yes/no): <7 hours per weeknight and <6 hours per weeknight.15
Insomnia was defined as one self-reported sleep complaint plus ≥1 self-reported daytime functional impairment due to lack of sleep.16 The following sleep complaints were inquired about: “trouble falling asleep,”“waking up during the night and had trouble getting back to sleep,”“waking up too early in the morning and unable to get back to sleep,” and “feeling unrested during the day, no matter how many hours of sleep had.” Daytime functional impairments related to sleepiness measures included difficulties in carrying out specific regular daily activities in the last month in the following 9 areas: “concentrating on the things,”“remembering things,”“getting things done because too sleepy or tired to drive or take public transportation?,”“performing employed or volunteer work or attending school,”“working on a hobby, eg, sewing, collecting, gardening,” and “taking care of financial affairs and doing paperwork (eg, paying bills or keeping financial records).” Patients who reported ≥1 daytime functional impairment and any one sleep complaint occurring 5 to 15 times per month were defined as having “mild/moderate insomnia.” Patients with ≥1 daytime functional impairment and any sleep complaint occurring >15 times per month were considered to have “severe insomnia.” Insomnia with short sleep duration was defined as having any one sleep complaint occurring ≥5 times per month and any 1 daytime functional impairment and sleep duration <7 hours per weeknight.
Covariates included self-reported demographics, health-related variables, and substance use (use of alcohol and illegal drug usage). Age was reported in years at the time of screening. Race was dichotomized as Caucasian vs other. Financial strain was measured as a continuous variable by poverty income ratio (PIR), a variable obtained by dividing the family income by the poverty threshold. Education was dichotomized with a cutoff set at high school graduation vs no high school graduate. Insurance status was dichotomized as covered by any type of health insurance vs not insured. Marital status was defined based on living with a partner/married or other. Study-collected serum nicotine was used to determine current smoking status, with patients smoking ≤10 ng/mL considered nonsmokers.17 Self-reported use of any alcohol, available for participants’ 20 years or older was defined as having ≥1 drink per month. Illegal narcotics or stimulant drug use was coded as positive in patients self-reporting current or past use of marijuana, hashish, cocaine, heroin, or methamphetamine.
Weight, height, and waist circumference at the time of the interview were collected by trained health technicians. Body mass index (BMI)18 was computed as the ratio of weight in kg to height in cm2 and overweight/obesity was defined as BMI ≥25 kg/m2. Waist circumference was used to define central obesity as >88 cm in women or >102 cm in men. Direct high-density lipoprotein (HDL) levels were assessed from blood drawn at the time of data collection in mg/dL. Participants were considered to have diabetes based on self-reported physician-diagnosed diabetes. Patients with hemoglobin A1c≥7% at the time of NHANES data collection were considered to have poorly controlled diabetes.19 Self-reported “overall general health” was also dichotomized by collapsing “fair” and “poor health” groups into the newly created variable “unsatisfactory health” vs “satisfactory health.” Depression was assessed using the 9-Item Patient Health Questionnaire (PHQ), and a PHQ score >10 was considered an indication of major depression.20
General and sleep characteristics of the analytic sample were analyzed according to RHTN status by using sample weights analyses in SAS 9.2 (Proc Survey; SAS Institute, Inc, Cary, NC) and the Taylor Series Linearization approach.21 Univariate and multivariate nested hierarchical logistical regression modeling for this sample was obtained after the weights were normalized (standardized) to the size of the subsamples.22 Models were adjusted for age, sex, race, family income, education, health insurance, marital status, smoking, alcohol use, illegal drug use, sedentary leisure time, unsatisfactory health status, overweight/obesity, direct HDL, poorly controlled diabetes, and depression. The final model was additionally adjusted for physician-diagnosed sleep apnea, insomnia, and SHST <6 hours per weeknight. The consistency of the presented weighted results was tested in unweighted analyses.23 Two-tailed P values of <.05 were considered significant.
Of 3306 patients with HTN, there were 379 with severe HTN: 269 with CSHTN and 110 with RHTN. Among RHTN patients, 94 had diabetes or CKD. The descriptive characteristics for the final cohorts are described in Table I. Age, sex, ethnic distribution, PIR, sedentary leisure time, illicit drug usage, obesity, and alcohol use and depression scores did not differ significantly between the two groups. Patients with RHTN were more likely to be married, have a lower education level, be current smokers, have unsatisfactory health, and have poorly controlled diabetes as compared with controls. In addition, they were more likely to have diabetes and poorly controlled diabetes.
Table I. General Characteristics of Severe Hypertension Patientsa According to Hypertension Control Statusb (N=379)
|Mean age, y||67.6±1.2||66.4±1.2|
|Male sex, %||48.5±7.4||42.9±3.5|
|Race (Caucasian), %||75.7±4.0||74.4±4.2|
|Mean poverty income ratio (family income/poverty threshold), %||2.2±0.2||2.5±0.1|
|Low education (<high school graduation), %||39.7±4.2d||27.7±3.5|
|No health insurance, %||2.4±1.4||4.3±1.8|
|Marital status (married or living with partner), %||56.3±6.2d||55.5±4.7|
|Current smoker (serum nicotine >10 ng/mL), %||12.9±4.3d||12.1±2.2|
|Alcohol consumption (alcohol ≥1 time per mo), %||43.9±6.3||50.9±3.8|
|Illegal drugs use (current or former use of marijuana, hashish, cocaine, and/or heroin), %||1.8±1.3||4.8±1.3|
|Mean sedentary leisure time (h per d of TV and computer use for entertainment), %||4.1±0.4||3.8±0.1|
|Unsatisfactory health (self-reported fair or poor health), %||69.6±4.1d||56.0±5.0|
|Overweight/obesity (body mass index ≥25 kg/m2), %||83.6±5.0||75.0±3.4|
|Central obesity (waist circumference >88 cm in women or >102 in men), %||75.6±4.7||66.8±3.9|
|Mean direct high-density lipoprotein, mg/dL||49.8±1.4||50.8±1.2|
|Physician-diagnosed diabetes, %||86.4±4.5d||21.6±3.0|
|Poorly controlled diabetes, % (hemoglobin A1c≥7%)||23.8±5.4d||9.2±1.9|
|Depression (PHQ-9 score≥10), %||15.1±3.1||14.3±2.1|
In relation to sleep characteristics (Table II), there was no difference in any of the surrogate symptom measures of sleep apnea, sleep duration, or insomnia between RHTN and CSHTN groups. History of physician-diagnosed OSA was marginally greater in the RHTN group as compared with the CSHTN group.
Table II. Sleep Characteristicsa of Severe Hypertensionb in Patients According to Hypertension Control Status (N=379)c
|Physician-diagnosed obstructive sleep apnea, %||17.5±2.9e||17.4±3.4|
|Snoring (≥3 times per wk in the past 12 mo), %||45.5±5.3||58.3±4.8|
|Snorting (≥3 times per wk in the past 12 mo), %||15.3±3.1||22.1±3.4|
|Combined symptoms of sleep apnea (snoring or snorting ≥3 times per wk in the past 12 mo), %||49.8±4.9||59.3±4.7|
|Severe symptoms of sleep apnea (snoring or snorting ≥5 times per wk in the past 12 mo), %||36.7±6.9||39.0±4.0|
|Severe combined symptoms of sleep apnea (snoring and snorting ≥5 times per wk in the past 12 mo), %||9.4±3.3||9.9±3.0|
|SHST <7 h per weeknight, %||39.8±4.1||42.0±3.2|
|SHST <6 h per weeknight, %||17.5±4.7||16.1±2.3|
|Mild/moderate insomnia (≥1 sleep disturbance 5–15 times per mo and ≥1 daytime functional impairment), %||37.3±4.4||32.7±4.0|
|Severe insomnia (≥1 sleep disturbance ≥15 times per mo and ≥1 daytime functional impairment), %||10.2±2.8||12.3±3.0|
|Insomnia with SHST <7 h per weeknight (≥1 sleep disturbance ≥5 times per mo and ≥1 daytime functional impairment and with SHST <7 h per weeknight), %||21.7±3.3||21.9±2.9|
Univariate analysis (Table III and Table IV) for the RHTN group revealed that people not completing high school had about 1.7 higher odds (95% confidence interval [CI], 1.1–2.7) of having RHTN. Also, diabetes was associated with 12.8 higher odds (95% CI, 7.4–21.9) and poorly controlled diabetes was associated with 3.1 higher odds (95% CI, 1.3–7.2) of having RHTN compared with CSHTN. The patients who reported poor health had 1.8 times higher odds (95% CI, 1.1–2.8) of having RHTN. The patients who reported snoring more than 3 times per week in the past 12 months trended towards a reduced odds of RHTN (that was marginally significant, odds ratio [OR], 0.6; 95% CI, 0.3–1.0). No other sleep symptoms were found to be significantly associated with RHTN in this cohort.
Table III. Univariate ORs for RHTNa in Patients With Severe Hypertensionb
|Mean age, y||1.04 (0.9–1.2)||.523|
|Male sex||0.8 (0.4–1.6)||.512|
|Race (Caucasian)||1.1 (0.6–1.9)||.802|
|Mean poverty income ratio (family income/poverty threshold)||1.4 (0.7–2.9)||.3986|
|Low education (<high school graduation)||1.7 (1.1–2.7)c||.019|
|No health insurance||0.5 (0.1–2.3)||.404|
|Marital status (married or living with partner)||1.0 (0.5–1.7)||.9111|
|Current smoker (serum nicotine >10 ng/mL)||1.0 (0.5–2./2)||.993|
|Alcohol consumption (alcohol ≥1 time per mo)||0.8 (0.4–1.5)||.423|
|Illegal drug use (current or former use of marijuana, hashish, cocaine, and/or heroin)||0.3 (0.1–1.4)||.126|
|Mean sedentary leisure time (h per d of TV and computer use for entertainment)||1.1 (0.5–2.1)||.830|
|Unsatisfactory health (self-reported fair or poor health)||1.8 (1.1–2.8)c||.010|
|Overweight/obesity (body mass index ≥25 kg/m2)||1.7 (0.8–3.6)||.167|
|Central obesity (waist circumference >88 cm in women or >102 in men)||1.5 (0.9–2.7)||.131|
|Mean direct high-density lipoprotein, mg/dL||0.9 (0.6–1.4)||.635|
|Physician diagnosed diabetes||12.8 (7.4–21.9)c||<.0001|
|Poorly controlled diabetes (hemoglobin A1c≥7%)||3.1 (1.3–7.2)c||.008|
|Depression (PHQ-9 score≥10)||1.1 (0.7–1.7)||.786|
Table IV. Univariate ORs of Sleep Symptoms for Resistant Hypertensiona in Patients With Severe Hypertensionb
|Physician-diagnosed obstructive sleep apnea||1.0 (0.5–2.0)||.9687|
|Snoring (≥3 times per wk in the past 12 mo)||0.6 (0.3–1.0)||.0657|
|Snorting (≥3 times per wk in the past 12 mo)||0.6 (0.3–01.3)||.1905|
|Combined symptoms of sleep apnea (snoring or snorting ≥3 times per wk in the past 12 mo)||0.7 (0.5–1.1)||.0918|
|Severe symptoms of sleep apnea (snoring or snorting ≥5 times per wk in the past 12 mo)||0.9 (0.5–1.8)||.7744|
|Severe combined symptoms of sleep apnea (snoring and snorting ≥5 times per wk in the past 12 mo)||1.0 (0.3–3.0)||.9304|
|SHST <7 h per weeknight||0.9 (0.6–1.3)||.6045|
|SHST <6 h per weeknight||1.1 (0.5–2.5)||.8060|
|Mild/moderate insomnia (≥1 sleep disturbance 5–15 times mo and ≥1 daytime functional impairment)||1.2 (0.8–1.8)||.3249|
|Severe insomnia (≥1 sleep disturbance ≥15 times mo and ≥1 daytime functional impairment)||0.8 (0.3–1.9)||.6239|
|Insomnia with SHST <7 h per weeknight (≥1 sleep disturbance ≥5 times per mo and ≥1 daytime functional impairment and with SHST<7 h per weeknight)||1.0 (0.6–1.6)||.9565|
Multivariate analyses showed consistent association between poorly controlled diabetes and RHTN (OR, 3.0; 95% CI, 1.2–7.9) (Table V). Self-reporting fair or poor health was also a significant predictor of RHTN (OR, 1.7; 95% CI, 1.1–2.7). Snoring at least 3 times per week trended towards a reduced risk of RHTN (OR, 0.5; 95% CI, 0.3–1.1).
Table V. Multivariate ORs for Resistant Hypertensiona in Patients With Severe Hypertensionb
|Low education (< high school graduation)||1.5 (0.9–2.4)||.8122|
|Illegal drugs use (current or former use of marijuana, hashish, cocaine, and/or heroin)||0.3 (0.1–1.4)||.1093|
|Unsatisfactory health (self-reported fair or poor health)||1.7 (1.1–2.7)||.0289|
|Overweight/obesity (body mass index ≥25 kg/m2)||1.7 (0.8–3.8)||.1964|
|Poorly controlled diabetes (hemoglobin A1c≥7%)||3.0 (1.2–7.9)||.0245|
|Snoringc (≥3 times per night in the past 12 mo)||0.5 (0.3–1.1)||.0742|
To the best of our knowledge, this is the first study that assessed the role of self-reported sleep complaints in severe hypertension as it exists in the community and distinguishing RHTN from medically matched patients with CSHTN. From this analysis, sleep symptoms did not predict RHTN, but other common health problems did. This analysis has the advantage of sampling from a nonclinic sample in a nationally representative US population cohort.
Poorly controlled diabetes and self-reported fair or poor health were significant, and probably more important, predictors for RHTN. The general association of diabetes and HTN is well-known. For instance, in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), diabetes predicted lack of BP control during the course of the study.24 Conversely, clinical trials have indicated that in order to achieve the lower BP goal recommended for patients with diabetes, an average of 2.8 to 4.2 antihypertensive medications are required.25 The degree to which insulin resistance directly contributes to the development of HTN vs simply being associated with HTN because of common underlying causes (ie, obesity) remains to be determined. Pathophysiologic effects of insulin resistance per se may also contribute to HTN; these include increased sympathetic nervous activity, vascular smooth muscle cell proliferation, and increased sodium retention. Therefore, diabetic control could be proposed as an important health priority to help control RHTN. There has been a thought that patient’s overall level of well-being and perception of functional capacity may be more sensitive to the pharmacotherapy of antihypertensive drugs. Also, compliance, which is frequently related to a patient’s sense of deterioration in quality of life secondary to medical treatment, may well be the determinate of success with any antihypertensive regimen.26 One possibility is that the RHTN cohort, with greater unsatisfactory health, may not be as compliant to the pharmacologic therapy. While all of these factors can be indirectly attributed to poor sleep, chronic snoring, or untreated sleep apnea, self-reports of such conditions appear to take a backseat when RHTN is detected in the community.
It was contrary to our expectations that self-reported sleep symptoms did not relate to RHTN. In fact, there was a trend towards a protective effect of frequent snoring in terms of RHTN. Past studies including small hospital case series have shown that among cohorts with RHTN, OSA was highly prevalent.10 Our findings of a lack of association among RHTN, sleep symptoms, and sleep apnea could have resulted from a relative insensitivity or nonspecificity of these symptoms to detect OSA, whereas past studies used the “gold standard” polysomnography to diagnose OSA using values of apnea-hypopnea index. The use of polysomnography is cost prohibitive and thus was not done in the NHANES study. Further mechanistic and interventional studies are needed to address the role of sleep apnea in RHTN to understand what features of clinical studies have led to the belief that OSA is a risk factor of RHTN.3,27,28
While many studies have shown a significantly higher prevalence of HTN in habitual snorers than among nonsnorers,12,29 other studies have reported no association after the adjustment of confounding factors such as smoking, alcohol usage, age, and obesity.30,31 These observations are based on clinic cohorts where attributes of the presenting population in both medical and social domains could differ and offer different associations. The cause of the marginally significant protective role of snoring is unknown but could result from artifact secondary to screening and treatment decisions of OSA in RHTN vs CSHTN.
Strengths and Limitations
The strengths of the study include the national community-based sample of hypertensive patients. Limitations of the NHANES data include a cross-sectional design, self-reported data on sleep, and data paucity in respect to OSA treatment for patients already diagnosed with OSA in the cohort. The prevalence of severe HTN was found to be lower compared with previous published clinical studies. Since NHANES does not include recently hospitalized or institutionalized adults, some of the members of the population might be excluded from NHANES.
This study shows a significant association between self-reported poor health and poor controlled diabetes and RHTN in a representative sample of US population with severe hypertension, independently of confounders including sleep symptoms. Longitudinal and interventional community-based studies are warranted to assess the predictive value of the diabetic control and assess the role of OSA symptoms in screening, diagnosis, and treatment in severe HTN.
Acknowledgments and disclosures: Dr Kingman Strohl serves on the medical advisory boards of SleepMed and Novasom and has grants from the VA Research Foundation and the National Institutes of Health National Heart, Lung, and Blood Institute. Supported by National Institutes of Health grants HL007913 and HS00059-14. The other authors have no conflict of interests and no financial disclosures to make.