Insomnia Did Not Predict Incident Hypertension in Older Adults in the Cardiovascular Health Study

Abstract

Study Objective:

We hypothesized that the sleep complaints of insomnia predict incident hypertension, particularly in African Americans. The purpose of this study was to analyze insomnia complaints as predictors of incident hypertension in the Cardiovascular Health Study (CHS), stratifying by gender and allowing for race and sleep variable interaction.

Design:

This is a prospective cohort study over a 6-year period of follow-up.

Setting:

This is a community-based study of participants in Forsyth County, North Carolina; Pittsburgh, Pennsylvania; Sacramento County, California; and Washington County, Maryland.

Participants:

The study analyzed data from 1419 older individuals (baseline mean age 73.4 ± 4.4 years) from the Cardiovascular Health Study who were not hypertensive at baseline.

Interventions:

none

Measurements:

We constructed relative risks of incident hypertension over a 6-year period for insomnia complaints singly and in combination.

Results:

Difficulty falling asleep, singly or in combination with other sleep complaints, predicted a statistically significant reduction of risk for incident hypertension for non-African American men in 6 years of follow-up. Insomnia complaints did not predict incident hypertension in 6 years of follow-up in women or in African Americans, although there may not have been enough power to show a significant association for African Americans.

Conclusions:

Insomnia did not predict hypertension in this older cohort which was free of hypertension at baseline. Difficulty falling asleep was associated with reduced risk of hypertension in non-African American men.

Citation:

Phillips B; Bůžková P; Enright P. Insomnia did not predict incident hypertension in older adults in the cardiovascular health study. SLEEP 2009;32(1):65-72.

INSOMNIA IS THE MOST COMMON SLEEP COMPLAINT IN THE UNITED STATES, AND IS ESTIMATED TO AFFECT 10% TO 30% OF AMERICANS. Insomnia is associated with reduced quality of life and may precede depression^1,2 but has not clearly been shown to be causally associated with adverse outcomes. In fact, Kripke and colleagues found a reduced mortality rate for those individuals complaining of sleeping difficulty in a 6-year follow-up study.³ Efforts to establish the prevalence, potential consequences and efficacy of treatment of insomnia have been complicated by the subjective nature and variety of its symptoms. There are several characterizations of insomnia in the literature. In addition, the DSM-III-R⁴ and DSM-IV⁵ classifications, the Research Diagnostic Criteria for Insomnia Disorder,⁶ and the International Classification of Sleep Disorders⁷ include varying criteria for making the diagnosis of insomnia. A recent National Institutes of Health (NIH) Consensus Conference on insomnia concluded, “Insomnia has historically been defined by complaints of disturbed sleep in the presence of adequate opportunity and circumstance for sleep. The disturbance may consist of one or more of three features: (1) difficulty in initiating sleep; (2) difficulty in maintaining sleep; or (3) waking up too early. A fourth characteristic, nonrestorative or poor-quality sleep, has frequently been included in the definition, although there is controversy as to whether individuals with this complaint share similar pathophysiologic mechanisms with the others.”⁸ The symptom of nonrestorative sleep, as noted by the NIH consensus conference, is particularly nonspecific and prevalent, and is strongly linked to depression.⁹ The definition used to define insomnia affects its impact; in one study, some definitions of insomnia predicted incident hypertension but others did not.¹⁰ While insomnia has not been causally associated with mortality over 6- or 7-year periods,^3,11 longitudinal studies of this duration may not be long enough to demonstrate such effects, if they exist. It is possible that investigation of “upstream” events such as hypertension or other prevalent cardiovascular conditions could yield important information about the natural history of insomnia. Indeed, difficulty falling asleep was associated with an increased risk of angina in a cross-sectional study.¹² In summary, the relationship between sleep complaints, medical illness, mood disorders, and hypertension is complex, and more work is needed in this area. The NIH has urged specific research efforts to “Apply well-defined diagnostic criteria to assess the prevalence, incidence, and natural history of insomnia and sleep complaints” and to “Assess the effects of insomnia on public health burdens such as premature death and health care costs.”⁸

An intriguing relationship exists between sleep duration and hypertension. Gangswisch¹³ demonstrated that short sleep duration predicted hypertension in an 8-year longitudinal analysis of the National Health and Nutrition Examination Survey. In addition, Lauderdale¹⁴ found that African Americans (AA) had shorter objectively measured sleep durations than did Caucasians in the Coronary Artery Risk Development in Young Adults study. This association between short sleep duration and hypertension could be one of the mechanisms accounting for the increased prevalence of hypertension and African Americans.

The aim of this analysis was to assess the effect of specific components of insomnia on incident hypertension in the Cardiovascular Health Study. We hypothesized that insomnia complaints predict incident hypertension. As hypertension in older adults has been shown to be a risk factor for stroke, heart failure, and death, identifying easily assessable predisposing risk factors remains important. We were particularly interested in relationships between sleep complaints and hypertension in African Americans, since little data is available about this group, and the CHS cohort is enriched with African Americans. Excess cardiovascular morbidity among African Americans likely has many causes, but we hypothesized that sleep complaints could be among the factors contributing to this phenomenon.

METHODS

Study Population

Participants in the Cardiovascular Health Study were selected using a Medicare eligibility list provided by the U.S. Health Care Financing Administration (HCFA) for the 4 participating communities: Forsyth County, North Carolina; Pittsburgh, Pennsylvania; Sacramento County, California; and Washington County, Maryland. These communities are diverse in proportion of minorities, education and income level, degree of urbanization, death rates, and availability of medical care. The initial study cohort of 5201 participants was recruited and examined in 1989–90. To enhance the representation of African American people in the study, an additional cohort of 687 African Americans was enrolled in 1992–93, bringing the total number of African Americans in the cohort to 924. Thus, 17% of the females and 14% of males in the cohort were African American.

The following were exclusion factors for study entry: institutionalization; terminal illness; inability to walk, communicate, or give informed consent; or being likely to move from the area during the next 3 years. Enrolled CHS participants were younger, more educated, and more likely to be married and Caucasian than those who refused or were ineligible. The CHS design and recruitment are described in detail elsewhere.^15,16 The research protocol was reviewed and approved by the institutional review board for human studies at the 4 clinical centers, and informed consent was obtained.

Examinations

Study participants completed standardized interviews in both the home and the field center, as well as extensive examinations at the field centers at baseline and during the annual examinations. Not all components were repeated at each annual visit. Standing height was measured in stocking feet to the nearest centimeter using a stadiometer, and weight was measured using a balance beam scale, recalibrated monthly. A medication inventory, including antihypertensive medications, was obtained at each examination.¹⁷ Participants assessed their general health by answering the question: “Would you say, in general, your health is (excellent, very good, good, fair, or poor)?” Symptoms of depression were assessed using the modified Center for Epidemiologic Studies Depression (CES-D) scale of 0–30.¹⁸ The categorical variable “Depression” (a mood, not a diagnostic category) was defined as a score ≥ 10. Diabetes was staged as normal, impaired fasting glucose {110–125 mg/dL} and diabetic {taking insulin or oral hypoglycemic or fasting glucose ≥ 126 mg/dL}. We assessed alcohol intake as reported previously.¹⁹ At the baseline visit and annually until 1999, participants separately reported their usual frequency of consumption of beer, wine, and liquor, and the usual number of 12-ounce cans or bottles of beer, 6-ounce glasses of wine, and shots of liquor that they drank on each occasion. At baseline, participants reported whether they had changed their pattern of consumption during the previous 5 years and whether they ever regularly consumed ≥ 5 drinks daily. Participants who reported abstention at baseline but responded yes to either of these questions were classified as former drinkers. Participants were categorized into categories according to weekly ethanol consumption as follows: none, former, < 1 drink weekly, 1 to 6 drinks weekly, ≥ 7 drinks weekly.

CVD Assessment

Health status was assessed at baseline through self-report of physician diagnosis of diseases. Self-report of cardiovascular diseases were validated according to standardized criteria by the medications used and by examination data such as blood pressure, electrocardiograms,²¹ echocardiograms,²² and carotid ultrasound.²³ Cardiovascular events were validated by a review of medical records and adjudicated according to standardized criteria.²⁴

Pulmonary Assessments

Spirometry was performed according to American Thoracic Society criteria,²⁵ with reference values previously obtained from healthy members of our cohort.²⁶ The smoking status of each participant at each exam was categorized as never smoker, former smoker, or current smoker, using responses to the standardized ATS DLD-78 respiratory questionnaire.²⁷

Hypertension Assessments

We defined hypertension as seated blood pressure average systolic ≥ 140 mm Hg, or seated blood pressure average diastolic ≥ 90 mm Hg, or history of hypertension and taking antihypertensive medication. Antihypertensive medications included β-blockers, calcium-channel blockers, diuretics, vasodilators, β-blockers with diuretics, angiotensin converting enzyme inhibitors, angiotensin converting enzyme with diuretics, vasodilators with diuretics, angiotensin type 2 antagonists, or angiotensin type 2 antagonists with diuretics. Most CHS blood pressures were taken in the morning.

Assessment of Sleep Complaints

Three questions that are included in most definitions of insomnia were included in the evaluation at year 5 (1992–1993), which is the year that the African American cohort was added. They were: Do you usually have trouble falling asleep? (Difficulty Falling Asleep, DFA); Do you usually wake up several times at night? (Sleep Continuity Disturbance, SCD); Do you usually wake up far too early? (Early Morning Awakening, EMA). The possible responses were limited to “yes,” “no,” or “don't know.” Only 1% to 3% of the questions were answered “don't know” and were coded as missing.

Analyses

To address incident hypertension, we excluded from analysis individuals who were hypertensive at year 5, the study's baseline. We analyzed the following sleep complaints as predictors of incident hypertension: difficulty falling asleep (DFA), early morning awakening (EMA), waking up repeatedly (sleep continuity disturbance, SCD), and DFA combinations (DFA and EMA, DFA and SCD, and all three complaints combined [DFA and SCD and EMA]). Race contrasts were done for African Americans versus non-African Americans; and men and women were analyzed separately. We allowed for interaction of race with sleep complaints to assess possible differentiable risk of incident hypertension associated with sleep complaints among African Americans compared to the risk among non-African Americans. In assessing the possible relationship between sleep complaints and hypertension, we assumed that the risk, if any, would be small, and opted for a long exposure period of 6 years. Thus, the outcome variable was incident hypertension at year 11, which was 6 years after the exposure to the sleep complaint predictors.

Because CHS is a cohort study, it enabled us to estimate directly relative risks as opposed to odds ratios. To construct relative risks we used Poisson log-link model with a model-robust sandwich estimator for standard errors.²⁸ Statistical analyses were carried out in R version 2.5.1.²⁹ We did not adjust P-values for multiple testing. Model I was adjusted for year 5 age and education (< 12 grades or ≥ 12 grades). In addition to the Model I covariates, the fully adjusted Model II was adjusted for income (< $25,000 US or ≥ $25,000 US), body mass index, smoking status, alcohol intake, FEV1, coronary heart disease, diabetes stage, height, and estrogen use for women.

RESULTS

Demographics

Of the total 5888 CHS participants, 353 (6.0%) died by year 5, and 497 (8.4%) were lost to follow-up or did not have their hypertension status assessed at year 5. On year 5 assessment, 2920 (49.6%) were hypertensive, leaving 2118 participants (36.0%) available for inclusion in this analysis of incident hypertension. Of the 2118 participants, 431 (20.3%) died by year 11, and 268 (12.7%) were lost to follow-up or did not have their hypertension status assessed at year 11 and could not be included in the analysis. After exclusion of those individuals, our study population consisted of 1419 CHS participants. There were 150 African Americans, 1264 Caucasians, and 5 individuals who were American Indian/Alaskan, Asian/Pacific Islander, or other. Among the 579 men there were 70 African Americans and among the 840 women 80 were African Americans.

Baseline characteristics are summarized in Table 1 for the subpopulation of 1419 people who were not hypertensive at year 5 and who were available to have their hypertension status assessed at year 11. Sixty-four percent remained free of hypertension at year 11. The baseline mean age was 73.5 years, ranging from 64 to 91 years. Forty-one percent were males, and 11% were African Americans. Those with hypertension at year 11 were more likely to be women, less educated, have a lower income and FEV1, and have a higher BMI and worse glucose status than normotensive people. Hypertensive women were more likely to use estrogen than normotensive women.

Table 1.

Baseline Characteristics at Year 5 by Hypertensive Status at Year 11

	Total (n=1,419)	Normotensive at Year 11 (n=912)	Hypertensive at Year 11 (n=507)	P value
Age	73.5 ± 4.39	73.53 ± 4.47	73.46 ± 4.26	0.77
Gender (Male)	579 (40.8)	399 (43.8)	180 (35.5)	< 0.01
Race (AA)	150 (10.6)	99 (10.9)	51 (10.1)	0.71
DFA	281 (20.1)	171 (19)	110 (22)	0.21
SCD	813 (58.4)	513 (57.3)	300 (60.2)	0.32
EMA	421 (30.4)	262 (29.4)	159 (32.3)	0.3
DFA & SCD	235 (16.7)	143 (15.9)	92 (18.4)	0.26
DFA & EMA	168 (12)	101 (11.2)	67 (13.3)	0.27
DFA & EMA – SCD	158 (11.2)	95 (10.5)	63 (12.6)	0.28
Education (≥ 12 yrs)	1110 (78.4)	737 (81.1)	373 (73.7)	< 0.01
Income (≥ $25,000)	629 (47.7)	422 (49.9)	207 (43.7)	0.03
Estrogen use (women)	140 (9.9)	74 (8.1)	66 (13)	< 0.01
CESD (Depressed)	1213 (85.6)	792 (86.9)	421 (83.2)	0.07
CHD	224 (15.8)	140 (15.4)	84 (16.6)	0.6
BMI	26.26 ± 4.09	26.06 ± 3.93	26.62 ± 4.35	0.02
FEV1	2.13 ± 0.7	2.18 ± 0.69	2.06 ± 0.73	< 0.01
Smoking status				0.58
Never	626 (45.2)	395 (44.2)	231 (47)
Former	608 (43.9)	401 (44.9)	207 (42.1)
Current	152 (11)	98 (11)	54 (11)
Glucose status				0.03
Normal	1144 (83.2)	746 (84.8)	398 (80.4)
IFG	104 (7.6)	66 (7.5)	38 (7.7)
Diabetes	127 (9.2)	68 (7.7)	59 (11.9)
Alcohol use				0.07
None	572 (40.6)	346 (38.3)	226 (44.8)
Previous	105 (7.5)	65 (7.2)	40 (7.9)
< 1 drink	280 (19.9)	197 (21.8)	83 (16.5)
1 to 6 drinks	254 (18)	165 (18.3)	89 (17.7)
≥ 7 drinks	197 (14)	131 (14.5)	66 (13.1)

Numbers are means ± standard deviations or counts (percentages within columns). P-values are from t-tests or Chi-squared tests used as appropriate, comparing those who were hypertensive at year 11 to those who were normotensive at year 11.

EMA=early morning awakening; DFA=difficulty falling asleep; SCD=sleep continuity disturbance; IFG=impaired fasting glucose; BMI=body mass index; FEV1=forced expired volume in one second; CHD=coronary heart disease; AA=African American

Health Characteristics

Our subpopulation of 1419 people was generally healthier than the CHS population. Self-reported health was excellent or very good for 47%, good for 41%, and fair or poor for 12%, compared with 36%, 41%, and 23% for the general CHS population at year 5. At year 11, self-reported health was excellent or very good for 30%, good for 48%, and fair or poor for 22% of the study group, compared with 25%, 44%, and 31% for the general CHS population. Thus, our findings may not be extendable to the general older population.

Associations Between Sleep Complaints and Incident Hypertension

The cross-tabulation of hypertension status at year 11 and sleep complaints at year 5 are presented by race and gender in Table 2. Interestingly, sleep complaints tended to “cluster,” especially for women. Women who reported early morning awakening and difficulty falling asleep nearly always had sleep continuity disturbance as well. In other words, 97% of women who reported DFA and EMA also reported SCD. Similarly, 88% of men who reported DFA and EMA also reported SCD.

Table 2.

Cross-Tabulation of Sleep Complaints and Hypertension by Race and Gender

	Hypertension
	Men				Women
	non-AAs		AAs		non-AAs		AAs
	no	Yes	no	yes	No	yes	no	Yes
DFA
no	303	142	37	15	347	210	42	24
yes	44	14	11	6	110	84	6	6
EMA
no	249	103	37	11	310	200	33	20
yes	95	51	12	10	138	89	17	9
SCD
no	158	60	25	9	175	118	24	11
yes	189	95	24	12	275	174	25	19
DFA & SCD
no	312	147	41	18	364	220	42	24
yes	38	10	7	3	92	73	6	6
DFA & EMA
no	323	146	42	17	392	246	45	26
yes	26	11	7	4	64	48	4	4
DFA & EMA & SCD
no	325	148	44	18	395	246	45	26
yes	25	9	5	3	61	47	4	4

AAs =African Americans

EMA=early morning awakening; DFA=difficulty falling asleep; SCD=sleep continuity disturbance

Results of the analyses are presented in Table 3 for Model I and Table 4 for the fully adjusted Model II, including estimated relative risks (RRs), 95% confidence intervals (CIs) of the RRs and the P-values. Of particular interest are the data relating to those non-African American men who endorsed the symptom of difficulty falling asleep. In the fully adjusted model in men, the relative risk of incident hypertension in 6 years associated with DFA among non-African Americans was estimated as 0.47 (95% CI 0.25–0.87); this was a statistically significant reduction of 53% in incident hypertension. DFA in combination with sleep continuity disturbance or in combination with both of the other sleep complaints also predicted a reduced risk of hypertension over 6 years of follow-up under the fully adjusted model II in non-African American men. There was an increased risk of incident hypertension (RR 2.03, P-value 0.05) over 6 years in African American men who endorsed the symptom of EMA when adjusting for age and education alone, which became smaller (RR 1.89) and was no longer statistically significant in the fully adjusted model. Adding depression to the fully-adjusted model resulted in very minor changes in the estimates of relative risks. It produced a slightly higher relative risk for incident hypertension for African American men who endorsed EMA (RR 2.14, P = 0.04), it and eliminated the significance of the combination of DFA with SCD or EMA in predicting reduced risk of hypertension for non-AA men (data not shown).

Table 3.

Relative Risks of Incident Hypertension in 6 Years Associated with Sleep Complaints, Model I

Model I	RR (P-value) 95% CI
	Men		Women
	Non-AAs	AAs	Non-AAs	AAs
DFA	0.79 (0.31)	1.15 (0.72)	1.13 (0.22)	1.32 (0.44)
	0.50-1.25	0.52-2.56	0.93-1.38	0.65-2.66
EMA	1.21 (0.18)	2.03 (0.05)	0.99 (0.89)	0.90 (0.73)
	0.92-1.61	0.99-4.16	0.81-1.2	0.48-1.66
SCD	1.25 (0.11)	1.54 (0.25)	0.98 (0.8)	1.42 (0.24)
	0.95-1.63	0.74-3.2	0.81-1.17	0.79-2.54
DFA & SCD	0.68 (0.15)	1.09 (0.86)	1.17 (0.15)	1.32 (0.44)
	0.40-1.16	0.39-3.05	0.95-1.44	0.65-2.67
DFA & EMA	0.97 (0.92)	1.33 (0.54)	1.10 (0.47)	1.26 (0.58)
	0.58-1.62	0.53-3.32	0.86-1.4	0.55-2.89
DFA & EMA & SCD	0.86 (0.61)	1.51 (0.44)	1.12 (0.39)	1.26 (0.58)
	0.49-1.52	0.54-4.23	0.87-1.43	0.55-2.9

Adjusted for age and education, using interaction of race and sleep complaint.

AAs =African Americans

EMA=early morning awakening; DFA=difficulty falling asleep; SCD=sleep continuity disturbance

Table 4.

Relative Risks of Incident Hypertension in 6 Years Associated with Sleep Complaints, Model II

Model II	RR (P-value) 95% CI
	Men		Women
	Non-AAs	AAs	Non-AAs	AAs
DFA	0.47 (0.02)	0.79 (0.63)	1.23 (0.08)	1.73 (0.18)
	0.25-0.87	0.30-2.08	0.98-1.54	0.77-3.91
EMA	1.13 (0.44)	1.89 (0.13)	0.99 (0.96)	1.32 (0.45)
	0.82-1.56	0.83-4.32	0.79-1.25	0.65-2.7
SCD	1.19 (0.26)	1.25 (0.61)	1.00 (0.99)	1.11 (0.76)
	0.88-1.62	0.53-2.97	0.81-1.24	0.56-2.23
DFA & SCD	0.36 (0.01)	0.42 (0.32)	1.19 (0.16)	1.73 (0.18)
	0.17-0.76	0.08-2.31	0.94-1.52	0.77-3.91
DFA & EMA	0.49 (0.07)	0.69 (0.56)	1.08 (0.6)	1.82 (0.19)
	0.23-1.05	0.19-2.43	0.81-1.46	0.75-4.4
DFA & EMA & SCD	0.41 (0.04)	0.60 (0.56)	1.09 (0.59)	1.81 (0.19)
	0.17-0.95	0.11-3.37	0.81-1.46	0.75-4.4

Adjusted for age, education, income, BMI, smoking status, alcohol intake, FEV1, CHD, diabetes stage, height and estrogen, using interaction of race and sleep complaint.

AAs =African Americans

EMA=early morning awakening; DFA=difficulty falling asleep; SCD=sleep continuity disturbance

Among women, there were no statistically significant differences in risk of incident hypertension in association with any sleep complaint for any race. Only a trend toward elevated risk of incident hypertension associated with DFA was suggested in non-African American women (RR 1.23, P = 0.08).

There were no significant associations between sleep continuity disturbance and incident hypertension at Year 11.

We hypothesized that the association of sleep complaints and incident hypertension over 6 years would be different for African Americans than for non-African Americans. However, in none of the 6 sleep complaint analysis was the interaction term of sleep complaint and race statistically significant at a 5% level or even suggestive at a 10% level.

DISCUSSION

The main finding of this analysis is that the sleep complaints of insomnia did not predict incident hypertension in this relatively healthy older cohort from the CHS. In fact, the sleep complaint of difficulty falling sleep, singly or in combination with other components of insomnia, predicted a lower risk of incident hypertension over 6 years of follow-up for non African American men. In women, there was no association between incident hypertension and DFA and its combinations with other components of insomnia. It is possible that the finding of a risk reduction for incident hypertension among non-African American men with DFA (and its combinations) is a spurious result of this post hoc subanalysis of a very healthy, select cohort. It is possible that difficulty falling asleep serves or served some survival function (e.g., increased vigilance helps to avoid predators) that somehow relates to hypertension or to a risk for hypertension.

To our knowledge this is the first prospective study to specifically address this issue in a cohort enriched with African Americans. These results are somewhat surprising, given earlier work in this area. An earlier analysis of the CHS cohort³⁰ found that health conditions were associated with disturbed sleep in a cross-sectional analysis; these authors noted that the relationship between trouble falling asleep and cardiovascular disease, including hypertension, was stronger for men than for women. However, that analysis revealed that depression was the by far the strongest correlate of sleep complaints in the cohort, confirming earlier work by Newman³¹ who noted an association between sleep complaints and cardiovascular disease but not hypertension in a cross-sectional analysis of this cohort. In a 4-year follow-up of 4,794 male Japanese workers, Suka et al found that persistent difficulty initiating sleep was associated with an increased risk of incident hypertension.³² These authors controlled for several important confounders, including job stress, but did not control for depression. The Japanese cohort was younger than the CHS cohort and was less diverse. As noted, the definition of insomnia affects not only its prevalence but also its likely impact. ¹⁰

It is possible that sleep complaints could lead to myocardial infarction or other cardiac disease through a causal pathway that does not involve hypertension. A large study that included a significant number of African Americans³³ investigated the relationship between sleep complaints and myocardial infarction, and noted that restless sleep and difficulty falling asleep predicted incident MI after adjusting for age, gender, and race. However, these relationships lost statistical significance after controlling for education, number of prescription medicines, self-rated health, and depression score.

Taken together, these data do not demonstrate a convincing relationship between insomnia complaints and hypertension. Given earlier studies showing that sleep complaints are not associated with increased incident mortality,^3,10 these results are not completely surprising. The lack of convincing correlation between insomnia complaints and incident hypertension contrasts with the consistent association between short sleep and hypertension^13,34 and highlights the important distinction between insomnia and sleep restriction.

The strengths of this study are the large, well-described CHS cohort, which now includes a substantial component of African Americans. With this work, we are able to extend earlier findings demonstrating lack of increased mortality^3,10 or cardiovascular morbidity¹² in individuals with sleep complaints. Unlike the American Cancer Society Study,³ the CHS study includes objective cardiovascular and diabetes measures. Unlike the ARIC¹⁰ study, the CHS study includes subject older than 69 years of age. We were able to control for many confounders known to be associated with cardiovascular disease and hypertension. For example, blood pressure varies in relation to meal consumption and time of day³⁵; consistency in timing of blood pressure measurement is a strength of this analysis. In addition, because the questions about sleep included in the evaluation of the CHS cohort fulfill the classic criteria for insomnia complaints, we were able to analyze the risk of incident hypertension for specific components of the criteria for insomnia, separately and in combination. This analysis suggests that different components of insomnia could have different outcomes. We believe that this analysis helps to address the research need identified by the NIH⁸ to apply well-defined diagnostic criteria to assess the incidence and natural history of insomnia and sleep complaints, including public health and health care costs.

Hypertension is a prevalent and significant problem in the aging population. Among 65- to 74-year-olds, 60.9% of men and 74% of women are hypertensive. Among those 75 years and older, 69.2% of men and 83.4% of women have hypertension.³⁶ African Americans are disproportionately affected; the Third National Health and Nutrition Examination Survey found that for people over 70 years of age, the prevalence of blood pressures of ≥ 149/90 mm Hg for black women, black men, white women, and white men were 82.9%, 67.1%, 66.2%, and 59.2%, respectively.³⁷ Both the Framingham Heart Study³⁸ and the Cardiovascular Health Study³⁹ have demonstrated associations between hypertension and cardiovascular disease in older people. Systolic hypertension is associated with cognitive decline^40,41 and mortality⁴² in older persons. Thus, identification and treatment of modifiable risk factors for hypertension in older persons could be expected to improve and lengthen life in this population. Unfortunately, we are unable to demonstrate that sleep complaints contribute to hypertension in older Americans.

Potential limitations of this report include the fact that this specific analysis was not an a priori analysis for CHS, although overall determination of the predictors of hypertension, cardiovascular disease, and mortality were. Another limitation was that the participants were asked about the classic symptoms of insomnia (difficulty going to sleep, difficulty staying asleep, and early morning awakening), but not about how disturbed they were by these problems, or whether they perceived them as insomnia. It is possible that some of the participants in this study were taking hypnotics or other sleep-promoting agents, and may not have reported symptoms of insomnia at the time of evaluation because they were being treated. Unfortunately, we are not able to evaluate that possibility. Another potential cause of misclassification is that women are more likely than are men to report sleep complaints at all ages. With aging, underlying medical disease is associated with increased sleep dissatisfaction, and men with illness, including cardiac disease, appear less likely to report poor sleep quality than their female counterparts. It is possible that the men in this cohort underreported sleep complaints, leading to misclassification of participants.

As is often the case, statistical power, while adequate for the entire sample, was insufficient when interaction terms were included. We demonstrated, using simulation experiments that closely mimicked our model and data, that power statistical significance at 5% level in subgroup analysis was low. For example, for DFA + SCD + EMA, power to detect important RRs of 0.67 and 0.5 (i.e., reciprocals of 1.5 and 2) in AA men was 14% and 22% respectively.

To try to boost the power, and because none of the interaction terms between race and sleep complaints were significant at 10% level, we performed an additional analysis omitting the interaction terms (results not shown). However, significance of the relative risk estimates has not considerably improved, and as we originally hypothesized presence of the interaction terms, we decided to report the estimates based on models with interaction terms.

The implication of this study is that it is difficult to demonstrate that insomnia is causally associated with hypertension. The increased risk of incident hypertension for African American men with early morning awakening was of borderline significance, but there were three very strong reductions in risk of incident hypertension in non-African American men who had difficulty falling asleep, singly or in combination with other symptoms. There remains little strong evidence that insomnia is associated with any adverse outcome except for depression.¹

DISCLOSURE STATEMENT

This was not an industry supported study. The authors have indicated no financial conflicts of interest.