Perceived stress, subjective, and objective symptoms of disturbed sleep in men and women with stable heart failure

Abstract

Objective/Background:

Sleep disturbance is prevalent among patients with heart failure (HF) and is associated with increased morbidity and mortality. Stress also affects health and quality of life among patients with cardiovascular disease and likely plays a prominent role in HF. However, little is known about the associations between stress and sleep among HF patients.

Participants:

One hundred fifty-three stable New York Heart Association (NYHA) Classification I-IV HF patients with at least low symptoms of insomnia (M_age:63.0±12.8, 42% Women).

Methods:

We examined baseline stress, sleep disturbance, and sleep-related characteristics from a randomized controlled trial of cognitive behavioral therapy for insomnia including the Perceived Stress Scale, Insomnia Severity Index, Pittsburgh Sleep Quality Index, Epworth Sleepiness Scale, Sleep Disturbance Questionnaire, Dysfunctional Beliefs about Sleep Scale, PROMIS Cognitive Ability, SF-36 Mental Health, and wrist actigraphy. We used Pearson correlations and general linear models to assess stress-sleep associations, including potential moderating effects of sex and symptom severity (NYHA).

Results:

There were moderate-to-large correlations between stress and self-reported sleep disturbance, dysfunctional beliefs about sleep, cognitive ability, and mental health (p’s<0.01). High stress was associated with more objectively-measured (i.e., actigraph-assessed) awakenings and sleep fragmentation among women than men (β = −0.04, p<0.01; β = −0.71, p = 0.04). Relationships between stress and objectively-measured sleep did not vary by symptom severity.

Conclusions:

Perceived stress is related to sleep disturbance among HF patients, and effects may be sex-dependent. Subsequent research should determine the temporal links between sleep and stress and optimal opportunities for intervention among HF patients.

Introduction

Over 6 million adults in the United States have heart failure (HF), one of many conditions categorized as cardiovascular disease, which is a primary cause of hospitalization and death (Benjamin, Muntner, & Bittencourt, 2019). HF is a progressive, neurohormonal condition in which the heart fails to adequately pump or fill with blood, and consequently deprives the body of oxygen (American Heart Association, 2017). The condition is associated with sleep disturbance including insomnia, hypersomnia, poor sleep quality (i.e., wakefulness and fragmentation of nighttime sleep), waking too early, and excessive daytime sleepiness (Broström, Strömberg, Dahlström, & Fridlund, 2004; Erickson, Westlake, Dracup, Woo, & Hage, 2003; Moon, Phelan, Lauver, & Bratzke, 2015; Moon, Yoon, & Bratzke, 2017; Redeker, 2008; Redeker et al., 2010; Redeker & Stein, 2006; Wang, Lee, Tsay, & Tung, 2010). Among those with HF, disturbed sleep is associated with a high symptom burden, worse physical and mental health, cognitive impairment, poor quality of life, and mortality (Broström et al., 2004; Garcia et al., 2012; Javadi, Darvishpour, Mehrdad, & Lakeh, 2015; Javaheri, 2006; Redeker & Hilkert, 2005; Redeker & Stein, 2006; Reinhard et al., 2013).

Psychosocial stress is a well-known risk factor for cardiovascular disease and is associated with poor clinical status and mental health symptoms among patients with HF (Alhurani et al., 2018; Brummett et al., 2004; Chambers & Reiser, 1953; Das & O’Keefe, 2006; Koizumi et al., 2009). For example, a mental stress task performed in the laboratory significantly increased ischemic burden in HF patients (Wawrzyniak et al., 2015), and patients’ self-reported perceived stress predicted greater HF symptom severity over time (Endrighi et al., 2019). Stressors may include relationship problems, difficulties managing HF symptoms, and low perceived control over health (Coyne et al., 2001; Dickens, Dickson, & Piano, 2019; Dracup et al., 2003), among others.

Stress also interacts with sleep (Åkerstedt, Hume, Minors, & Waterhouse, 1994; Kim & Dimsdale, 2007); stress is associated with insomnia, greater symptoms of sleep disturbance, shorter sleep duration, and precipitates frequent episodes of nighttime awakening, both in healthy adults and in those with HF (Hall et al., 2015; Haynes, Adams, & Franzen, 1981; Mezick et al., 2009; Redeker et al., 2019; Zoccola, Dickerson, & Lam, 2009). More specifically, in one study, individuals with high perceived stress exhibited worse nighttime sleep and daytime problems (i.e., shorter sleep duration, worse sleep quality, a greater likelihood of sleep apnea, sleepiness and fatigue) and an increased cardiovascular risk profile (i.e., higher body mass index and inflammation) compared to those with low stress, outlining a potential mechanistic pathway from perceived stress to cardiovascular disease (Kashani, Eliasson, & Vernalis, 2012). Among patients with HF, a lower day-night ratio of the stress hormone cortisol is associated with greater symptoms of insomnia and negative sleep-related cognitions, two indicators of physiological hyperarousal (Redeker et al., 2019). However, it is unknown if there are different associations between stress and subjective (i.e., self-report) compared to objective (i.e., actigraph-assessed) sleep disturbance among those with HF.

Potential associations between sleep and stress may also differ by a patient’s sex. In the general population, women are more vulnerable to some sleep disorders including restless legs syndrome and insomnia, and report more perceived stress than men (Berger, Luedemann, Trenkwalder, John, & Kessler, 2004; Cohen & Janicki-Deverts, 2012; B. Zhang & Wing, 2006). Laugsand and colleagues (2013) found that women with greater cumulative symptoms of sleep disturbance (difficulty initiating or maintaining sleep, and non-restorative sleep) had a higher risk of incident HF compared to men, but there was an inverse association between self-reported sleep disturbance and HF risk for men (Laugsand, Strand, Platou, Vatten, & Janszky, 2013). Other studies report no sex differences in the relationship between sleep and HF risk (Ingelsson, Lind, Ärnlöv, & Sundström, 2007), or in sleep quality among patients with HF (Erickson et al., 2003; Redeker & Hilkert, 2005).

The New York Heart Association (NYHA) classification is used to categorize HF patients based on symptom severity and functional status (I: without symptoms, II: symptoms with exertion, III: symptoms with any physical activity, class IV: symptoms at rest). As observed with sex differences, there is limited literature concerning NYHA class and sleep. In one study, worse symptom severity (i.e., a higher NYHA class) was associated with a greater risk of insomnia (Príncipe-Rodríguez, Strohl, Hadziefendic, & Piña, 2005). Others found that HF symptom severity predicted worse self-reported sleep quality, insufficient sleep, and daytime dysfunction (Nasir, Shahid, & Shabbir, 2015; Redeker & Hilkert, 2005). However, NYHA class was unrelated to actigraph-assessed sleep characteristics (Redeker & Hilkert, 2005). Thus, sleep disturbance and stress may each be associated with greater symptoms, suggesting that NYHA class could moderate sleep-stress associations, but there may be differential associations between objective versus subjective sleep and cardiovascular disease severity.

The goals of this study are (1) to describe cross-sectional relationships between perceived stress and subjective sleep characteristics, sleep-related cognition, cognitive abilities, mental health symptoms, and objective sleep characteristics of sleep, assessed with actigraphy among patients with stable HF and, (2) to test whether sex or HF symptom severity (i.e., NYHA class) moderate sleep-stress associations. We hypothesize that perceived stress will be directly associated with subjective and objective sleep characteristics and that these associations will be more robust for women than men.

Methods

Design

We conducted secondary analyses of data from a randomized controlled trial in which HF patients participated in group-based cognitive behavioral therapy for insomnia or received education about HF and sleep hygiene in the attention control condition. A complete description of the original study procedures is published elsewhere (Redeker et al., 2017).

Sample

A total of 166 patients with stable HF were recruited from the community, the VA Connecticut Healthcare System echocardiogram database, and the Yale-New Haven Hospital HF program during a routine visit. Study participants had received stable doses of cardiovascular medications for at least 2 weeks and had no hospitalizations within the past month, were at least 18 years of age, cognitively intact, lived at home, had a NYHA Classification of I-IV, experienced insomnia symptoms for at least 1 month and scored ≥8 on the Insomnia Severity Index. Thus, each participant had at least subthreshold insomnia (Bastien, Vallières, & Morin, 2001). Study exclusion criteria were an unstable medical or psychological disorder, untreated sleep disordered breathing, restless legs syndrome, narcolepsy, regularly engaging in night or shift work, a seizure disorder, neurological or musculoskeletal disorders that restricted the non-dominant arm (due to possible confounding effects on wrist actigraph recordings). Participants with sleep apnea were included if the condition was mild (an apnea hypopnea index of <15 (Zinchuk et al., 2018)) or if they were adherent to continuous positive airway treatment (CPAP) by self-report during a patient interview. If participants did not have a sleep study within the last year recorded in their medical record or were not currently using CPAP therapy we conducted two nights of home polysomnography to screen for sleep apnea (Apnea Risk Evaluation System, SleepMed, Braintree, MA (Redeker et al., 2017)).

Approximately 13 individuals were missing self-report data pertinent to perceived stress, leaving a sample of 153 individuals to examine subjective sleep. Of that sample, 139 participants had actigraph data, with which to examine objective sleep.

Procedure

The Institutional Review Board approved study procedures, and all participants provided written informed consent. Participants completed a battery of questionnaires at baseline to elicit demographic and clinical information, perceived stress, self-report measures of sleep, sleep-related cognitions, cognitive functioning, and mental health symptoms. Participants wore actigraphs on their non-dominant wrists continuously for 14 days while engaging in their normal routines and completed daily sleep diaries in the morning after each night of actigraph recording.

Variables and Measures

Demographic and Clinical Variables

Participants reported their age, sex, race, marital status, and Veteran status. We conducted interviews and medical record review to obtain clinical data including body mass index (BMI), NYHA class, Charlson Comorbidity Index (CCI) and health history, left ventricular ejection fraction (EF), and use of cardiovascular medications (i.e., angiotensin-converting-enzyme [ACE] inhibitors, angiotensin II receptor blockers [ARB]; hydrochlorothiazide [HCTZ], loop diuretics, or a statin).

Subjective Sleep Characteristics and Sleep-Related Cognition

The Insomnia Severity Index (ISI) (Morin, 1993) is a 7-item measure, which is used to assess insomnia severity. Scores are 0-28. A cutoff of ≥8 indicates subthreshold symptoms of insomnia and ≥15 is used to indicate a diagnosis of clinical insomnia (moderate severity). The ISI has well-documented reliability, validity, and internal consistency (0.74–0.88) (Bastien et al., 2001; Chiu, Chang, Hsieh, & Tsai, 2016). We used categorical scores to describe the sample and continuous scores for analyses. Cronbach’s α was 0.73 in this sample and 0.83 in the study overall (Redeker et al., 2019).

The Pittsburgh Sleep Quality Index (PSQI) (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989) is a 19-item measure that is used to assess sleep over the previous month including subjective sleep efficiency, sleep latency, sleep duration, sleep quality, sleep disturbance, use of hypnotic medications; and daytime dysfunction due to sleepiness. The PSQI is reliable, valid, and internally consistent (Buysse et al., 1989; Buysse et al., 1991). The seven component scores yield a global score of 0-21 (21 = poorest sleep quality). We used the global score as well as the sleep duration and sleep latency component scores. The PSQI has a diagnostic sensitivity of 89.6% and specificity of 86.5%. Cronbach’s α was 0.83.

The Epworth Sleepiness Scale (ESS) (Johns, 1991) consists of 8 items, which are used to measure the general level of self-reported daytime sleepiness in common soporific situations. Items are scored to create a total of 0-24. The ESS has well-documented reliability and validity (Johns, 1992). Internal consistency was 0.83 in this study.

The Sleep Disturbance Questionnaire (SDQ) (Espie, Brooks, & Lindsay, 1989; Violani, Devoto, Lucidi, Lombardo, & Russo, 2004) is a 12-item measure of one’s beliefs about the causes of insomnia including restlessness or agitation, mental overactivity, consequences of insomnia, and lack of sleep readiness. Total scores are 12-60. The SDQ is reliable and valid (Espie et al., 1989). Cronbach’s alpha was 0.81.

The Dysfunctional Beliefs about Sleep Scale (DBAS) (Morin, Vallières, & Ivers, 2007) consists of 16 items that measure sleep disruptive cognitions on a scale of 0-10. The DBAS shows good internal consistency and reliability (Morin et al., 2007). Cronbach’s α was 0.83 in the current study.

Objective Sleep Variables

Objective sleep characteristics were measured with the Actiwatch 2 (Phillips Respironics, Inc.), a wrist-worn electronic accelerometer, continuously over 14 days in 1-min epochs. The device has a sensitivity of .01 g or greater, sampling data 32 times per second. The activity count/epoch of recording time is the sum of peak activity. Participants depressed event markers at “lights out/on” and removal times. Actigraph-measured wrist motor activity was used to compute sleep characteristics included sleep duration (amount of time between the estimated sleep onset and final awakening), sleep efficiency ([total sleep duration/time in bed] X 100), wake after sleep onset (WASO; the average number of minutes awake between sleep onset [at least 20 continuous minutes of sleep after getting into bed] and final awakening), sleep latency, sleep midpoint, and the fragmentation index ([the number of epochs with movement for at least 4 min + the number of epochs without movement < 1 min duration/number of epochs without movement that lasted > 1 min] × 100; (Mezick et al., 2009)). All actigraph data was scored by trained research assistants and was reviewed by a trained actigraph scorer to ensure accuracy and consistency of scoring. The rest period was defined as lights off to lights on and was determined (in order of consideration) using the actigraph light meter (lux = 0), event markers, and participant diaries.

While insomnia is diagnosed via clinical interview, for adults with sleep disorders, actigraphy is a more accurate method for measuring sleep duration than a sleep diary (Smith et al., 2018). Among those with chronic conditions, actigraphy may overestimate sleep duration and efficiency, and underestimate sleep latency and WASO compared to polysomnography, although agreement between the measures is generally acceptable (Conley et al., 2019). While actigraph-assessed sleep latency is not always a reliable variable, sleep latency was included because it was sensitive to differences in NYHA class among HF patients in other work (Jeon, Conley, & Redeker, 2019).

A daily sleep diary was completed each morning to assist with interpreting the wrist actigraph data (e.g., lights out and lights on time) and to document use of hypnotic medications. Diaries are reliable, valid, and often used to corroborate objective sleep data (Martin & Hakim, 2011; Matthews et al., 2018).

Stress, Cognition, and Mental Health Symptoms

The Perceived Stress Scale (PSS) (Cohen, Kamarck, & Mermelstein, 1983) is a 10-item instrument that measures one’s general perception of stress over the past month. Scores range from 0 to 40 with higher scores indicating greater perceived stress. Reliability and validity are well-documented (Cohen & Janicki-Deverts, 2012) and the measure is often used to assess stress in epidemiological and clinical research (Richardson et al., 2012). Cronbach’s α was 0.90.

The PROMIS Cognitive Ability Short-Form subscale is an 8-item measure that quantifies positive self-assessments of cognitive functioning over the previous 7 days including, “I have been able to concentrate” and “My memory has been as good as usual.” Higher scores indicate better cognitive ability. The measure is summed to produce a total score of 0 to 40, which is converted to a T-score metric with a mean of 50, and a standard deviation of 10. Internal consistency was 0.94 in this study.

The Short Form 36-item Health Survey (SF-36) (Brazier et al., 1992) is used to assess limitations in the type and amount of activities people can complete to meet basic needs, due to mental or physical problems. Each subscale ranges from 0 to 100, with higher scores indicating a better health-related quality of life. The SF-36 has well-established validity and reliability in chronically ill and healthy populations (Brazier et al., 1992; Léger, Scheuermaier, Philip, Paillard, & Guilleminault, 2001). These analyses focused on the normed Mental Health subscale. Cronbach’s α of the subscale was 0.90.

The Charlson Comorbidity Index (CCI) (Charlson, Szatrowski, Peterson, & Gold, 1994) is a weighted index based on 19 comorbid conditions. Comorbidities were derived from the medical record and assigned 1 to 6 based on mortality risk and disease severity. Conditions were summed to create a total score predicting 10-year mortality. The index is reliable, valid, and can be used with International Classification of Diseases (ICD)-9-CM and ICD-10 codes (Charlson et al., 1994; Sundararajan et al., 2004).

Data Analysis

We downloaded actigraph data into Actiware v.6 software (Phillips Respironics, Inc.) and calculated nocturnal sleep duration, sleep latency, WASO, sleep efficiency, sleep midpoint, and sleep fragmentation for each day using the medium threshold (i.e., 40). Mean values of sleep variables were calculated for those who had at least 7 consecutive days of data, given that some participants had less than 14 days. We used SAS v.9.4 (SAS Institute Inc., Cary, NC, USA) for all analyses, examined data for errors, and corrected for outliers as needed. Skewed variables were log transformed. PSS scores were divided into tertiles of low, moderate, and high stress (≤10, 11-29, ≥30) to describe the results, and the estimates from each analysis were summarized using the SAS MIANALYZE procedure. Pearson correlations with 95% confidence intervals were calculated for the entire sample and subsamples by sex and NYHA class using Fisher’s z-transformation. In bivariate analyses, we compared the correlations between PSS, subjective and objective sleep outcomes, cognition, and metal health by sex and NYHA class. The generalized linear model (GLM) was used to examine interaction effects between PSS and sex/NYHA class after adjusting for age, BMI, EF, and comorbidity (based on CCI scores). Missing values for self-report items were imputed using the mean of the observed values for an item. We used the Markov Chain Monte Carlo multiple imputation approach with ten imputation sets for correlations, confidence intervals, and GLM models (P. Zhang, 2003).

Results

Table 1 displays demographic and clinical characteristics of the sample (n = 153). Participants were 63 years of age on average (SD = 12.8) and 58% were men. Most participants were White (75%) and non-Veterans (86%). Mean BMI was in the obese range (31.5), with an average CCI of 2.6 (SD = 1.9). Participants were primarily classified as NYHA I or II (77%, i.e., no limitation or slight limitation of physical activity). The average total PSS score was 15, in the range of moderate stress. The mean ISI score was in the moderate severity diagnostic range for insomnia (15.3; i.e., 15-21). Thirty-eight percent of patients had a reduced EF (< 45%). The most prevalent comorbid conditions were hypertension (62%) and diabetes (33%). The most commonly used medications were loop diuretics (71%), beta blockers (67%), and statins (62%).

Table 1.

Demographics and clinical characteristics of the sample (n = 153).

Variables	Mean (SD) / N (%)
Age	63.0 (12.8)
Gender: Male	89 (58.2%)
Race
White	115 (75.2%)
African American	28 (18.3%)
Native American	1 (0.6%)
Asian	1 (0.6%)
Other	8 (5.2%)
Veterans	22 (14.4%)
Body Mass Index (BMI)	31.5 (8.2)
Charlson Comorbidity Index (CCI)	2.6 (1.9)
New York Heart Association (NYHA) Classification
I	50 (32.7%)
II	67 (43.8%)
III	32 (20.9%)
IV	4 (2.6%)
Perceived Stress Scale (PSS)	15.0 (7.6)
Insomnia Severity Index (ISI)	15.3 (4.3)
Ejection Fraction (EF) %
EF < 45	57 (37.7%)
EF < 35	31 (20.5%)
Health History
Diabetes	50 (33.1%)
Hypertension	92 (62.2%)
COPD	36 (24.0%)
Peripheral Vascular Disease	23 (16.1%)
Myocardial Infarction	40 (28.4%)
Pacemaker	44 (29.1%)
LVAD	3 (2.1%)
Medications
ACE or ARB	76 (49.7%)
Beta blocker	102 (66.7%)
HCTZ	6 (5.2%)
Loop diuretic*	99 (70.7%)
Statin	95 (62.1%)

Note.

^*Counts prescription of any loop diuretic among Bumex, Demadex, and Lasix.

ACE: angiotensin-converting-enzyme inhibitors; ARB: angiotensin II receptor blockers; BMI: Body Mass Index; CCI: Charlson Comorbidity Index; COPD: chronic obstructive pulmonary disease; EF: ejection fraction; HCTZ hydrochlorothiazide; ISI: Insomnia Severity Index; LVAD: left ventricular assist device; PSS: Perceived Stress Scale

Compared to individuals with low stress (PSS ≤ 10), those with high stress (PSS ≥ 30), had 9 minutes longer WASO (69.5 vs. 78.3), but sleep efficiency and sleep fragmentation were comparable (80.4% vs. 78.8%; 38.7% vs. 40.5%). Women and those in higher NYHA classes reported greater stress, but mean group differences were not statistically significant (women: 16.26 ± 7.46 vs. men: 14.12 ± 7.57; t(151) = 1.74, p = 0.08; NYHA Class I and II: 14.40 ± 7.50, vs. III and IV: 16.90 ± 7.70, t(151) = 1.71, p = 0.09).

Cross-sectional associations between stress and sleep

Table 2 displays the correlations between perceived stress and self-reported sleep, cognition, and mental health across the entire sample, and by sex and NYHA class, respectively. Overall, higher levels of perceived stress were associated with poorer sleep quality (r = 0.29, p < 0.01), and more daytime sleepiness (r = 0.22, p < 0.01), attributions about sleep disturbance (r = 0.34, p < 0.01), and dysfunctional beliefs about sleep (r = 0.38, p < 0.01). Stress was unrelated to insomnia symptoms (r = 0.10, p = 0.23) or sleep duration reported on the PSQI (r = −0.05, p = 0.55). Finally, stress was correlated with decreased cognitive ability (r = −0.50, p < 0.01) and worse mental health (r = −0.76, p < 0.01). When comparing the correlations based on sex or NYHA class there were no significant group differences in stress and self-reported sleep.

Table 2.

Correlations between perceived stress, self-reported sleep, cognition, and mental health (n = 153).

	PSS	PSS by Sex			PSS by NYHA
	Overall r [95% CI] p-value	Male N=89 r [95% CI] p-value	Female N=64 r [95% CI] p-value	Equality Test for Correlations* p-value	I/II N=117 r [95% CI] p-value	III/IV N=36 r [95% CI] p-value	Equality Test for Correlations* p-value
Insomnia Symptoms (ISI)	0.10 [−0.06, 0.26] (.228)	0.15 [−0.06, 0.35] (.159)	0.04 [−0.22, 0.29] (.772)	.498	0.06 [−0.13, 0.24] (.528)	0.15 [−0.20, 0.46] (.411)	.669
Global PSQI Score	0.29 [0.13, 0.44] (<.001)	0.25 [0.03, 0.44] (.025)	0.33 [0.09, 0.54] (.008)	.597	0.26 [0.08, 0.43] (.006)	0.40 [0.06, 0.65] (.021)	.474
Sleep Duration (PSQI)	−0.05 [−0.21, 0.11] (.550)	−0.06 [−0.27, 0.15] (.557)	0.00 [−0.25, 0.26] (.975)	.688	−0.06 [−0.24, 0.13] (.570)	−0.09 [−0.40, 0.25] (.624)	.880
Sleep Latency (PSQI)	0.06 [−0.10, 0.22] (.441)	−0.08 [−0.30, 0.13] (.449)	0.19 [−0.06, 0.42] (.130)	.101	0.09 [−0.10, 0.27] (.372)	−0.01 [−0.34, 0.32] (.968)	.640
Daytime Sleepiness (ESS)	0.22 [0.06, 0.37] (.007)	0.30 [0.10, 0.49] (.004)	0.08 [−0.17, 0.323] (.538)	.164	0.25 [0.07, 0.41] (.008)	0.16 [−0.20, 0.47] (.381)	.650
Sleep Disturbance (SDQ)	0.34 [0.19, 0.47] (<.001)	0.30 [0.10, 0.48] (.004)	0.34 [0.10, 0.54] (.005)	.808	0.33 [0.16, 0.48] (<.001)	0.38 [0.05, 0.63] (.021)	.785
Dysfunctional Beliefs (DBAS)	0.38 [0.22, 0.50] (<.001)	0.40 [0.21, 0.57] (<.001)	0.31 [0.07, 0.52] (.005)	.488	0.36 [0.19, 0.51] (<.001)	0.31 [−0.02, 0.59] (.067)	.782
Cognitive Ability (PROMIS)	−0.50 [−0.61, −0.37] (<.001)	−0.49 [−0.63, −0.31] (<.001)	−0.54 [−0.69, −0.34] (<.001)	.973	−0.47 [−0.60, −0.31] (<.001)	−0.54 [−0.74, −0.26] (<.001)	.571
Mental Health (SF-36)	−0.76 [−0.85, −0.69] (<.001)	−0.79 [−0.85, −0.69] (<.001)	−0.73 [−0.83, −0.59] (<.001)	.744	−0.74 [−0.81, −0.65] (<.001)	−0.81 [−0.90, −0.66] (<.001)	.334

Notes.

DBAS: Dysfunctional Beliefs and Attitudes About Sleep; ESS: Epworth Sleepiness Scale; PSQI: Pittsburgh Sleep Quality Index; PSS: Perceived Stress Scale; SF-36: 36-item Short Form Health Survey. Estimates were obtained from 10 imputation data sets.

^*Equality test for correlations was performed to reject the null hypothesis of equal correlation coefficients between two groups using Fisher’s z-transformation.

Moderating effects of sex and NYHA class

In adjusted GLMs, perceived stress was significantly associated with daytime sleepiness (β = 0.11, SE = 0.05, p = 0.03), for men only. There was also a NYHA*Stress interaction (β = −0.34, SE = 0.16, p = 0.02), whereas those with greater perceived stress and in higher NYHA classes also had worse mental health. Finally, perceived stress was directly associated with dysfunctional beliefs and attitudes about sleep as measured by the SDQ and DBAS (β = 0.03, SE = 0.01, t(146) = −2.85, 95% CI: 0.03, 0.04, p = 0.005; β = 0.06, SE = 0.02, t(146) = 3.58, 95% CI: 0.03, 0.10, p < 0.001), and worse cognitive ability (β = −0.51, SE = 0.11, t(147) = −4.48, 95% CI: −0.73, −0.28, p < 0.001).

Correlations between perceived stress and actigraph-assessed sleep variables are displayed in Table 3. Fourteen additional participants were missing actigraphy data resulting in a sample of n = 139. There were no statistically significant demographic differences between those with and without actigraphy data. Correlations between perceived stress and actigraph variables were not significant (duration, efficiency, WASO, latency, midpoint, or sleep fragmentation; all p’s > .05). When comparing correlations by sex, women displayed significantly worse sleep efficiency and greater WASO (i.e., nighttime wakefulness) and sleep fragmentation (p’s < 0.01-0.03). Based on the adjusted GLMs, there were significant interactions between Sex and Stress, whereas women’s stress was directly associated with WASO (β = −0.04, SE = 0.01, p < 0.01) and sleep fragmentation (β = −0.71, SE=0.37, p = 0.04), while the inverse association was observed for men (Figures 1a and 1b). Among individuals who reported low perceived stress (PSS ≤ 10), women had lower WASO (48.28 min. vs. 70.67 min) and fragmentation (30.75% vs. 41.65%) than men. However, among those with high perceived stress (PSS ≥ 30), women demonstrated significantly greater WASO (77.58 vs. 52.62) and fragmentation than men (36.01 vs. 32.78). Results from adjusted models indicated that NYHA-Stress interactions were not significantly associated with objective sleep disturbance (all p’s > 0.05).

Table 3.

Correlations between perceived stress and sleep assessed with actigraphy (n = 139).

	PSS	PSS by Sex			PSS by NYHA
	Overall r [95% CI] p-value	Male N=79 r [95% CI] p-value	Female N=60 r [95% CI] p-value	Equality Test for Correlations* p-value	I/II N=109 r [95% CI] p-value	III/IV N=30 r [95% CI] p-value	Equality Test for Correlations* p-value
Sleep Duration	−0.09 [−0.26, 0.07] (.270)	−0.20 [−0.40, 0.02] (.072)	0.06 [−0.20, 0.31] (.656)	.130	−0.16 [−0.33, 0.04] (.105)	0.08 [−0.27, 0.44] (.606)	.234
Sleep Efficiency	0.02 [−0.14, 0.19] (.783)	0.17 [−0.05, 0.38] (.124)	−0.19 [−0.43, 0.06] (.135)	.033	0.02 [−0.17, 0.21] (.843)	−0.02 [−0.38, 0.34] (.907)	.846
WASO**	−0.14 [−0.30, 0.03] (.0984)	−0.33 [−0.51, −0.11] (.003)	0.22 [−0.03, 0.45] (.083)	.001	−0.13 [−0.31, 0.06] (.191)	−0.18 [−0.51, 0.19] (.332	.782
Sleep Latency**	−0.12 [−0.28, 0.05] (.165)	−0.23 [−0.43, 0.01] (.038)	0.01 [−0.24, 0.27] (.914)	.150	−0.18 [−0.36, 0.00] (.055)	0.15 [−0.22, 0.48] (.435)	.118
Sleep Midpoint	0.14 [−0.02, 0.30] (.090)	0.09 [−0.13, 0.31] (.407)	0.16 [−0.09, 0.40] (.181)	.675	0.16 [−0.03, 0.33] (.104)	−0.02 [−0.38, 0.34] (.899)	.397
Sleep Fragmentation Index	−0.09 [−0.26, 0.07] (.271)	−0.23 [−0.43, −0.01] (.043)	0.21 [−0.05, 0.44] (.105)	.011	−0.07 [−0.25, 0.12] (.469)	−0.16 [−0.49, 0.21] (.400)	.671

Notes.

PSS: Perceived Stress Scale; WASO: Wake after sleep onset.

^*Equality test for correlations was performed to reject the null hypothesis of equal correlation coefficients between two groups using Fisher z-transformation.

^**WASO and Sleep Latency were log-transformed.

Figure 1.

Scatter plots of log-transformed wake after sleep onset (WASO), sleep fragmentation, and perceived stress according to sex.

Discussion

To our knowledge this is the first examination of the associations between perceived stress and sleep disturbance among patients with stable HF. Our inquiry merges two, largely separate, literatures concerning the risks of disturbed sleep and psychosocial stress for this patient population (Alhurani et al., 2018; Redeker et al., 2010). As hypothesized, there were moderate-to-large associations between perceived stress and multiple self-report measures of sleep, sleep-related cognition, cognitive abilities, and mental health symptoms. Contrary to our hypothesis, stress was not correlated with insomnia severity or actigraph-assessed sleep characteristics. As all participants endorsed at least mild symptoms of insomnia and reported moderate levels of stress on average, less variance in both variables may explain their nonsignificant association. Adjusted analyses revealed distinct sex differences in the association between perceived stress and sleep disturbance. As hypothesized, stress was directly related to objective wakefulness and sleep fragmentation among women. Yet, stress was indirectly associated with wakefulness and sleep fragmentation among men. NYHA class may exacerbate the effects of stress on mental health. Although NYHA did not moderate the associations between stress and sleep, most patients were Class I or II, suggesting that the sleep-stress relationship may be robust for patients with stable, mild disease and invariable across severity of illness.

Previous evidence demonstrates relationships between perceived stress, cognition, mental health, and sleep among adults with cardiovascular disease (Bidulescu et al., 2010; Huang et al., 2011; Kashani et al., 2012; Redeker et al., 2018), literature which we extend to HF. There were robust correlations between stress and cognitive impairment, mental health symptoms, and problems with sleep continuity. However, in adjusted models, stress was not related to objective sleep duration, efficiency, latency, or midpoint, or to self-reported sleep. Thus, those who perceive greater stress may have sleep disturbance of which they are unaware. For example, individuals with high perceived stress had 9 more minutes of WASO than those with low stress. Multimodal assessment of sleep may be advantageous when assessing sleep disturbance or individuals with a history of sleep problems and HF.

The role of sex has often been neglected within cardiovascular disease research (Orth-Gomér & Deter, 2015), and extends to understanding stress and sleep in HF. Yet, sex appears to be a crucial moderator of the sleep-stress relationship among people with HF (Redeker & Stein, 2006). Perceived stress was directly related to WASO and sleep fragmentation in women, while the opposite pattern was observed in men. A previous study found similar positive associations between symptoms of insomnia and risk of HF in healthy women, although negative associations were reported for men (Laugsand et al., 2013). For those with HF, sex differences in sleep and stress may relate to group differences in physiology (e.g., sex-specific neuroendocrine activity (Wirth & Gaffey, 2013)). For example, stress hormone regulation may more dependent on sleep quality in men compared to women, which would result in a greater physiological imperative for men who experience greater stress to have better sleep quality (Bassett, Lupis, Gianferante, Rohleder, & Wolf, 2015)). There could also be sex-dependent patterns in reporting about stress; men may underreport stress while women might believe it is more socially acceptable to endorse stress or other somatic symptoms (Bassett et al., 2015; O’Meara et al., 2007), Men and women also seem to cope with stress differently (e.g., greater avoidant coping by men (Dracup et al., 2003)).These effects coalesce , in the general population as women endorse more stress and difficulty initiating and maintaining sleep, (Cohen & Janicki-Deverts, 2012; Tamres, Janicki, & Helgeson, 2002; B. Zhang & Wing, 2006), and sleep worsens during and after menopause (Kravitz et al., 2003). Thus, sleep is more likely to be disrupted in women, particularly those with HF, and may be further exacerbated by stress. It is unknown if sex differences in stress and sleep lead to different health outcomes for people with HF.

Associations between stress and sleep are likely bidirectional and may occur through physiological hyperarousal. Mechanistically, chronic stress and associated sleep disturbance could separately or jointly provoke hyperarousal via increased sympathetic activity, inflammation, and maladaptive neuroendocrine functioning (Bonnet & Arand, 2010; Edmondson, Green, Ye, Halazun, & Davidson, 2014; Krantz, Sheps, Carney, & Natelson, 2000; Kupper, Denollet, Widdershoven, & Kop, 2013; Wawrzyniak et al., 2015). Although our measure of perceived stress is well-validated, using more objective, reproducible measures of the physiological mechanisms of stress (e.g., synthesis and regulation of hypothalamic-pituitary-adrenal axis or sympathetic nervous system hormones (Gaffey & Wirth, 2014; Wirth & Gaffey, 2013)) is necessary to understand the pathophysiological signatures of stress in HF.

High perceived stress was also associated with more disordered beliefs and negative cognitions about sleep, indicated by elevated scores on the DBAS (problematic beliefs about sleep) and the SDQ (attributions about insomnia). Yet it is surprising that perceived stress was not associated with greater sleep latency, one of the cardinal symptoms of insomnia, as people who are stressed often have dysfunctional thoughts at bedtime. Based on this pattern it appears that sleep-related beliefs and cognitions are associated with, but distinct from, perceived stress. Sleep-related beliefs and cognitions may also be associated with other aspects of stress-related pathophysiology. For example, Redeker and colleagues previously showed that dysfunctional cognitions are correlated with stress-related biomarkers among individuals with stable HF (Redeker et al., 2018), and improving those cognitions a key mechanism of behavioral treatment for insomnia in this patient population (Redeker et al., 2019). Thus, dysfunctional cognitions are mutually associated with stress, an influential mediator of stress-related clinical outcomes, and part of a broader style of problematic coping that parallels a high symptom burden and prevents people with HF from improving sleep. Perhaps targeting dysfunctional cognitions could also help mitigate the effects of stress.

Clinical and Research Considerations

There may be multiple opportunities and potential benefits from understanding the role of perceived stress in HF. Routine screening for sleep and stress is a clear first step to clarify these associations in HF, particularly if self-reported stress is associated with objective sleep disturbance rather than self-reported sleep disturbance. Cardiovascular prevention guidelines have only recently recommended assessing sleep disturbance and psychosocial stressors (Arnett et al., 2019), among people with cardiovascular disorders, including those with HF (Yancy et al., 2017; Yancy et al., 2013). Continuing to measure stress throughout the healthcare trajectory could help clinicians anticipate which psychosocial and medical resources patients need to improve their management of HF (e.g., stress management at diagnosis, support groups for the people with HF or their caregivers, sleep assessment and treatment). Based on the sex differences described in our findings, interpreting information from those assessments, behavioral education and intervention may require an individualized, patient-centered approach for men compared to women.

In this sample, patients with more advanced HF reported greater stress, but sources of stress likely vary over the disease course. Understanding these changes is necessary to optimize the timing and delivery of behavioral interventions for stress or sleep. While sleep is increasingly a target in HF patients (Redeker et al., 2015; Redeker et al., 2017; Redeker & Stein, 2006), the efficacy of addressing sleep, stress, or approaching both factors simultaneously in HF should be compared. Routinely screening sleep and stress in people with HF may offer the best opportunity to mutually inform clinicians, caregivers, patients, and researchers how these factors interact and, thus, when to modify sleep or stress.

Limitations

This sample included patients with stable HF and those who met a threshold of insomnia symptoms, limiting the generalizability of these findings. Unfortunately, we did not collect information on the duration of participants’ insomnia symptoms beyond the ≥1-month inclusion criteria so it is unclear if these symptoms are chronic or more recent. Several factors may explain why stress was not related to symptoms of insomnia. It is likely that there is a ceiling effect on both the insomnia and stress measures in this study. Participants endorsed at least minimal insomnia and stress may be less of an issue for those with existing symptoms of insomnia. These analyses should be replicated in patients of HF with and without insomnia. This sample also endorsed moderate stress levels. High levels of both stress and insomnia could relate to hyperarousal. Hyperarousal is a core feature of insomnia and hyperarousal and related autonomic nervous system dysfunction are common among patients with HF (Kanno et al., 2016).

The cross-sectional nature of these analyses and potential self-report bias on the PSS also limit our ability to draw conclusions regarding causation and temporal relationships. It is unclear if an individual’s perceived stress or sleep disturbance predated their HF diagnosis or was enhanced by their condition. Given the chronic nature of HF, sleep-stress associations may persist longitudinally. For example, in a study of older women with HF, moderate positive correlations between perceived stress and fatigue were maintained over time despite increases in fatigue (Friedman & King, 1995). Assessing stress and sleep among newly diagnosed patients with HF and tracking symptoms across treatment will offer a richer evidence base concerning prospective associations, and opportunities to improve sleep, stress, and health-related quality of life. Finally, as actigraphy may underestimate WASO compared to polysomnography (Conley et al., 2019), the association between stress and WASO may be more robust than in our results.

There are clear, albeit complex, associations between perceived stress and sleep in HF. Findings from these secondary analyses largely supported our hypotheses that sleep and perceived stress interact in patients with stable HF, and high stress may be most deleterious for women. It remains unclear whether stress precipitates poor sleep or vice versa but the answer likely depends on one’s sex, and other individual differences, and may therefore, be specific to each patient. Routinely screening sleep and stress in people with HF may be the best approach to generate such data and to understand opportunities for patient-centered intervention.