Abstract
Background
The relationship between psychological stress and heart failure (HF) has not been well studied. We sought to assess the relationship between perceived stress and incident HF.
Methods
We used data from the national REasons for Geographic And Racial Differences in Stroke (REGARDS) study, a large prospective biracial cohort study that enrolled community-dwellers aged 45 years and older between 2003 and 2007, with follow-up. We included participants free of suspected prevalent HF who completed the Cohen 4-item Perceived Stress Scale (PSS-4). Our outcome variables were incident HF event, HF with reduced ejection fraction events, and HF with preserved ejection fraction events. We estimated Cox proportional hazard models to determine if PSS-4 quartiles were independently associated with incident HF events, adjusting for sociodemographics, social support, unhealthy behaviors, comorbid conditions, and physiologic parameters. We also tested interactions by baseline statin use, given its anti-inflammatory properties.
Results
Among 25,785 participants with a mean age of 64 ± 9.3 years, 55% were female and 40% were Black. Over a median follow-up of 10.1 years, 1109 ± 4.3% experienced an incident HF event. In fully adjusted models, the PSS-4 was not associated with HF or HF with reduced ejection fraction. However, PSS-4 quartiles 2–4 (compared with the lowest quartile) were associated with incident HF with preserved ejection fraction (Q2 hazard ratio 1.37, 95% confidence interval 1.00–1.88; Q3 hazard ratio 1.42, 95% confidence interval 1.03–1.95; Q4 hazard ratio 1.41, 95% confidence interval 1.04–1.92). Notably, this association was attenuated among participants who took a statin at baseline (P for interaction = .07).
Conclusions
Elevated perceived stress was associated with incident HF with preserved ejection fraction but not HF with reduced ejection fraction.
Lay Summary
New strategies for preventing heart failure (HF) are needed. We studied whether psychological stress is a risk factor for developing HF. We found that those with higher levels of stress were at increased risk of developing a specific type of HF called HF with preserved ejection fraction, or HFpEF. This finding suggests that psychological stress may be important for preventing HF.
Graphical Abstract
Heart failure (HF) affects more than 6 million people across the United States, and is expected to increase over the coming years. Despite novel therapeutic options and significant advances in HF care over the past 2 decades, outcomes for adults with HF remain poor with a 5-year mortality rate of approximately 50%.1 This finding underscores the importance of developing strategies for HF primary prevention.
Traditional risk factors for HF such as age and comorbid conditions do not fully explain incident HF. Several prediction models for HF already exist and have predominantly included traditional risk factors like age, hypertension, smoking history, coronary artery disease, and diabetes.2,3 Although these traditional risk factors account for much of the population-attributable risk for developing HF, approximately 30% are not explained by these factors.4 This factor supports the need to identify additional nontraditional risk factors that may be targeted as part of a comprehensive strategy to prevent incident HF.
Psychological stress is a nontraditional risk factor that may merit additional investigation. Psychological stress has been implicated in the development of multiple disease processes, including HF.5 The proposed mechanism for stress leading to HF relates to the activation of inflammatory and neurohormonal pathways involving the hypothalamic–pituitary axis and the sympathetic nervous system.6 With stress-induced activation of these pathways, HF can result from increased atherosclerosis, myocardial fibrosis and apoptosis, and contractile dysfunction.7 Prior work has shown increased rates of HF in patients with stress-related disorders including post-traumatic stress disorder, anxiety, depression, and adjustment disorders.8,9 However, there are limited data on whether high levels of perceived psychological stress are associated with incident HF. To address this important gap in the literature, we examined the association between stress, as measured by Cohen 4-item Perceived Stress Scale (PSS-4) at the study baseline, and incident HF using data from the geographically diverse REasons for Geographic and Racial Differences in Stroke (REGARDS) cohort. Given distinct pathophysiologies and thus the biological plausibility that stress can have a different impact on HF with reduced ejection fraction (HFrEF) compared with HF with preserved ejection fraction (HFpEF),7 we examined the association of perceived stress with any HF, as well as HFrEF and HFpEF separately. Given a HFpEF pathophysiology very closely linked to stress and inflammation,10,11 we hypothesized that perceived stress would be associated with HFpEF. Finally, given its anti-inflammatory properties and pleiotropic effects, we examined whether associations differed according to the baseline use of statins.
METHODS
Data Source
Our study sample included participants from the REGARDS cohort, which has been described previously.12 Briefly, the REGARDS cohort was recruited to examine geographic and racial differences in stroke mortality and therefore oversampled populations with higher incident and fatal stroke (persons from the Southeastern United States and Black persons). REGARDS is composed of 30,239 community-dwelling White and Black adults, aged at least 45 years (42% Black participants and 55% female) residing in the continental United States. Participants were recruited from 2003 to 2007 with ongoing follow-up.
Baseline data collection for REGARDS was completed using computer-assisted telephone interviews to collect medical history, functional status, health behaviors, and psychosocial measures, including the PSS-4. In-home examinations were conducted by trained health care professionals using standardized, quality-controlled protocols to collect physiologic measures (blood pressure, height and weight, and waist circumference), blood and urine samples, electrocardiograms, and medication use by pill bottle review. Every 6 months thereafter, participants were contacted to assess for new cardiovascular events, including for HF admissions or mortality.
The REGARDS study was previously approved by the institutional review boards of participating centers. This ancillary study was approved by the institutional review board. All participants provided written informed consent.
Study Population
We included participants from the REGARDS HF-free cohort13 based on medication use and completed a baseline PSS-4. The exclusion cascade is shown in Figure 1.
Figure 1.
Exclusion cascade. REGARDS, REasons for Geographic And Racial Differences in Stroke.
Measures
The primary exposure variable was baseline PSS-4 (obtained at the time of REGARDS study enrollment), a validated four item version of the PSS,14 which assessed perceived stress experienced over the prior month. Questions inquired about the degree to which the respondent felt (1) unable to control important things in your life, (2) confident in your ability to handle personal problems, (3) that things were going your way, and (4) that difficulties were piling up in life. Each category was scored using a 5-point scale (0 = never, 1 = almost never, 2 = sometimes, 3 = fairly often, 4 = very often). Final composite scores ranged from 0 to 16, where a higher score indicated a higher level of perceived stress. The PSS-4 was validated as an abridged version of the Cohen PSS to be used in telephone interviews. Because there are no validated cut points to distinguish between high or low stress, we chose to examine PSS-4 quartiles similar to prior work examining the association of PSS-4 and other cardiovascular end points.15
The primary outcome was incident adjudicated HF events—this included any HF, HFrEF, and HFpEF, which were examined separately. These 3 outcomes were determined a priori. HF events were defined as a hospitalization or death where HF was the underlying cause through 2016. Adjudication of an incident HF hospitalization was performed independently by 2 clinician investigators based on signs and symptoms, laboratory studies, electrocardiograms, and assessments of left ventricular function documented in the medical records.16,17 Adjudication of a death with HF as the underlying cause was based on interviews with participant proxies, medical records for hospitalizations around the time of death, and death certificates. Subtypes of HF were determined based on diagnostic testing performed at the time of an incident event similar to prior studies—HFrEF was defined as a left ventricular ejection fraction (LVEF) of less than 50% or a qualitative report of reduced LVEF (this included those with an LVEF of 40%–50%, which is commonly described as mid-range or mildly reduced ejection fraction given shared characteristics including responsiveness to treatment)18; HFpEF was defined as an LVEF of 50% or greater or a qualitative report of preserved LVEF.
We chose covariates for our statistical models a priori based on prior literature19,20 and clinical expertise. These variables included sociodemographic characteristics (age, sex, race, annual household income, educational achievement, geographic region of residence), social support variables (marital status, social isolation as defined by 1 visit or fewer by friends/family in prior month, and number of other adults who live in the household), health behaviors (cigarette smoking status, alcohol use defined according to the National Institute on Alcohol Abuse and Alcoholism, and self-reported exercise frequency), cardiovascular comorbid conditions (atrial fibrillation, history of coronary heart disease, dyslipidemia, and diabetes mellitus), physiologic parameters (body mass index, systolic blood pressure, diastolic blood pressure, left ventricular hypertrophy, estimated glomerular filtration rate, urinary albumin to creatinine ratio, depressive symptoms [based on the Center for Epidemiologic Studies Depression Scale score of ≥4], physical functioning [based on physical component score of the Short-Form 12], and baseline statin use). These variables were routinely collected at baseline when participants were initially enrolled in REGARDS. Similar variables have been used in prior studies examining incident HF.20
Effect Modification
To understand potential effect modification by specific factors, we planned several interaction tests a priori. First, because of the anti-inflammatory effect of statins and the proposed relationship between perceived stress, systemic inflammation, and incident HF, we examined the association between PSS-4 score and incident HF stratified by baseline statin use. Given known sex-related21 and race-related22 differences in left ventricular remodeling to stressors, we also examined whether associations between the PSS-4 score and incident HF differed according to sex and race. Given the possibility that the negative impact of increased psychologic stress might attenuate with increasing age, we examined whether associations between the PSS-4 score and incident HF differed according to advanced age (≥75 years old vs <75 years old)—we used 75 years given prior work in REGARDS that has shown effect modification of other risk factors for incident HF at this threshold.17 Prior work has shown that the PSS was associated with cardiovascular disease only in the setting of an income of less than $35,000.15 We accordingly also examined whether associations between PSS-4 score and incident HF differed according to an income of less than 35,000 compared with an income of $35,000 or more.
Statistical Analyses
We calculated descriptive data for baseline participant characteristics (mean with standard deviation for continuous variables or percentages for categorical variables). We then calculated the incident rates of HF events (inclusive of hospitalizations and deaths), HFpEF, and HFrEF for each PSS-4 score quartile as number of cases per 10,000 person-years.
We estimated the Cox proportional hazards regression models to determine whether baseline PSS-4 score quartiles were associated with incident HF events. The fully adjusted model included all covariates listed previously. For analyses of effect modification, we added cross-product terms to the models and performed a Wald test. If the cross-product terms were statistically significant, we calculated stratum-specific estimates of the association of PSS-4 with HF from these models. To account for missing covariate values in the regression analysis, we used multiple imputation by chained equations. Estimates from 30 separate imputations were combined and summarized using Rubin’s rules.23 Variables that were imputed include education, social isolation, number of adults living in the household (n = 13), physical component score, smoking, alcohol use, exercise, body mass index, systolic blood pressure, diastolic blood pressure, estimated glomerular filtration rate, urinary albumin to creatinine ratio, and depressive symptoms.
Given the well-described differences in the demographics, pathophysiology, treatment, and disease trajectory of HFrEF compared with HFpEF, we repeated the Cox proportional hazards regression models to determine whether the PSS-4 score was differentially associated with incident HFrEF and incident HFpEF using a Lunn–McNeil extension to the Cox proportional hazards model. We also repeated the interactions analyses for both HFrEF and HFpEF.
To account for competing risk, we conducted a sensitivity analysis where we repeated the primary analyses using the subdistribution proportional hazards model to incorporate the competing risk of non-HF mortality (for analyses of any HF, HFrEF, and HFpEF), HFpEF (for analyses of HFrEF), and HFrEF (for analyses of HFpEF).
To better understand the association between PSS-4 as a continuous variable and each outcome, we examined the association between PSS-4 and incident HF using the best fitting second-degree fractional polynomial.
We used 2-sided hypothesis testing and P value of less than .05 to determine statistical significance for main effects and a P value of less than 0.10 to determine statistical significance of interaction terms. We managed the data in SAS version 9.4 (SAS Institute, Cary, NC) and performed statistical analysis using STATA version14 (StataCorp, College Station, TX).
RESULTS
Study Population
In the study population of 25,785 participants, the mean age was 64.0 ± 9.3 years, 55.0% were female, 40.0% were Black, and 47.4% had a household income of less than $35,000. (Table 1). The PSS-4 quartiles were 0 (n = 6585), 1–2 (n = 6136), 3–4 (n = 5834), and 5 or higher (n = 7830). When compared with participants in quartile 1 those in quartiles 2–4 were more likely to be female, Black, have lower household incomes, lower educational achievement, social isolation, and lower PCS. Participants in the highest stress quartile (PSS of ≥5) were more likely to smoke currently and less likely to exercise, have a higher body mass index, and had higher prevalence of atrial fibrillation, diabetes mellitus, left ventricular hypertrophy, and depressive symptoms compared with other groups. There were no differences in baseline statin use across the PSS-4 quartiles.
Table 1.
Characteristics by PSS Quartiles
| Characteristic | Total | Q1 (PSS = 0) | Q2 (PSS = 1–2) | Q3 (PSS = 3–4) | Q4 (PSS ≥ 5) | P value |
|---|---|---|---|---|---|---|
|
| ||||||
| N | 25,785 | 6585 | 6136 | 5834 | 7230 | |
| Age, mean (SD) | 64 (9.3) | 66 (8.8) | 64 (9.2) | 64 (9.4) | 63 (9.8) | <.001 |
| Female sex | 14,177 (55.0%) | 2967 (45.1%) | 3186 (51.9%) | 3301 (56.6%) | 4723 (65.3%) | <.001 |
| Black race | 10,309 (40.0%) | 2597 (39.4%) | 2044 (33.3%) | 2221 (38.1%) | 3447 (47.7%) | <.001 |
| Annual household income | <.001 | |||||
| ≥$35,000 | 1,2227 (47.4%) | 3361 (51.0%) | 3337 (54.4%) | 2866 (49.1%) | 2663 (36.8%) | |
| <$35,000 | 10,476 (40.6%) | 2462 (37.4%) | 2085 (34.0%) | 2264 (38.8%) | 3665 (50.7%) | |
| Declined to report | 3082 (12.0%) | 762 (11.6%) | 714 (11.6%) | 704 (12.1%) | 902 (12.5%) | |
| Education < high school | 2934/25768 (11.4%) | 671/6580 (10.2%) | 486/6133 (7.9%) | 596/5831 (10.2%) | 1181/7224 (16.3%) | <.001 |
| Geographic region of residence | <.001 | |||||
| Belt | 8861 (34.4%) | 2250 (34.2%) | 2048 (33.4%) | 1966 (33.7%) | 2597 (35.9%) | |
| Buckle | 5367 (20.8%) | 1306 (19.8%) | 1220 (19.9%) | 1238 (21.2%) | 1603 (22.2%) | |
| Non-belt/buckle | 11,557 (44.8%) | 3029 (46.0%) | 2868 (46.7%) | 2630 (45.1%) | 3030 (41.9%) | |
| Married | 15,449 (59.9%) | 4125 (62.6%) | 3897 (63.5%) | 3563 (61.1%) | 3864 (53.4%) | <.001 |
| Social Isolation | 3104/25560 (12.1%) | 703/6474 (10.9%) | 642/6097 (10.5%) | 653/5800 (11.3%) | 1106/7189 (15.4%) | <.001 |
| No. of other adults living in household, median (IQR) | 1 (0, 1) | 1 (0, 1) | 1 (1, 1) | 1 (0, 1) | 1 (0, 1) | <.001 |
| Smoking status | <.001 | |||||
| Current | 3813/25694 (14.8%) | 866/6557 (13.2%) | 772/6113 (12.6%) | 780/5822 (13.4%) | 1395/7202 (19.4%) | |
| Past | 10,121/25,694 (39.4%) | 2826/6557 (43.1%) | 2504/6113 (41.0%) | 2302/5822 (39.5%) | 2489/7202 (34.6%) | |
| Never | 11,760/25,694 (45.8%) | 2865/6557 (43.7%) | 2837/6113 (46.4%) | 2740/5822 (47.1%) | 3318/7202 (46.1%) | |
| Alcohol use | <.001 | |||||
| Heavy | 1087/25,293 (4.3%) | 319/6448 (4.9%) | 264/6051 (4.4%) | 243/5734 (4.2%) | 261/7060 (3.7%) | |
| Moderate | 8727/25,293 (34.5%) | 2253/6448 (34.9%) | 2285/6051 (37.8%) | 2051/5734 (35.8%) | 2138/7060 (30.3%) | |
| None | 15479/25,293 (61.2%) | 3876/6448 (60.1%) | 3502/6051 (57.9%) | 3440/5734 (60.0%) | 4661/7060 (66.0%) | |
| Exercise frequency | <.001 | |||||
| ≥4×/per week | 7751/25,427 (30.5%) | 2326/6492 (35.8%) | 1962/6055 (32.4%) | 1677/5754 (29.1%) | 1786/7126 (25.1%) | |
| 1–3×/week | 9414/25,427 (37.0%) | 2279/6492 (35.1%) | 2306/6055 (38.1%) | 2246/5754 (39.0%) | 2583/7126 (36.2%) | |
| None | 8262/25,427 (32.5%) | 1887/6492 (29.1%) | 1787/6055 (29.5%) | 1831/5754 (31.8%) | 2757/7126 (38.7%) | |
| Atrial fibrillation | 1894 (7.3%) | 399 (6.1%) | 394 (6.4%) | 415 (7.1%) | 686 (9.5%) | <.001 |
| Coronary heart disease | 3751 (14.7%) | 977 (14.9%) | 855 (14.0%) | 831 (14.3%) | 1088 (15.2%) | .23 |
| Dyslipidemia | 14,379 (57.9%) | 3762 (59.3%) | 3406 (57.5%) | 3218 (57.0%) | 3993 (57.5%) | .052 |
| Diabetes mellitus | 4698 (18.9%) | 1162 (18.3%) | 1001 (16.9%) | 978 (17.4%) | 1557 (22.4%) | <.001 |
| Body mass index, kg/m2, mean (SD) | 28.9 (5.9) | 28.8 (5.5) | 28.7 (5.7) | 28.8 (5.9) | 29.5 (6.5) | <.001 |
| Systolic blood pressure, mm Hg, mean (SD) | 127 (16) | 128 (16) | 127 (16) | 127 (16) | 127 (17) | <.001 |
| Diastolic blood pressure, mm Hg, mean (SD) | 77 (9.6) | 77 (9.3) | 76 (9.3) | 76 (9.5) | 77 (10) | <.001 |
| Left ventricular hypertrophy | 2346 (9.1%) | 633 (9.7%) | 503 (8.2%) | 482 (8.3%) | 728 (10.1%) | <.001 |
| eGFR, mean (SD) | 86 (19) | 85 (19) | 86 (18) | 86 (19) | 88 (20) | <.001 |
| Urinary ACR (mg/g) | <.001 | |||||
| <10 | 15,732/24,624 (63.9%) | 4012/6308 (63.6%) | 3880/5887 (65.9%) | 3662/5589 (65.5%) | 4178/6840 (61.1%) | |
| 10–29 | 5568/24624 (22.6%) | 1427/6308 (22.6%) | 1292/5887 (21.9%) | 1222/5589 (21.9%) | 1627/6840 (23.8%) | |
| 30–300 | 2778/24,624 (11.3%) | 745/6308 (11.8%) | 598/5887 (10.2%) | 570/5589 (10.2%) | 865/6840 (12.6%) | |
| >300 | 546/24,624 (2.2%) | 124/6308 (2.0%) | 117/5887 (2.0%) | 135/5589 (2.4%) | 170/6840 (2.5%) | |
| Depressive symptoms | 2608/25,594 (10.2%) | 149/6524 (2.3%) | 221/6077 (3.6%) | 407/5791 (7.0%) | 1831/7202 (25.4%) | <.001 |
| Physical component score, mean (SD) | 47 (10) | 50 (8.2) | 49 (9) | 47 (9.9) | 44 (11) | <.001 |
| Statin use | 7548 (29.3%) | 2025 (30.8%) | 1779 (29.0%) | 1679 (28.8%) | 2065 (28.6%) | .022 |
PSS, Perceived Stress Scale Score; SD, standard deviation; IQR, interquartile range; eGFR, estimated glomerular filtration rate; ACR, urinary albumin/creatinine ratio.
Number of participants with missing data include education (n = 17), social isolation (n = 225), number of adults living in the household (n = 13), Physical component score (n = 1073), smoking (n = 91), alcohol use (n = 492), exercise (n = 358), body mass index (n = 113), systolic blood pressure (n = 34), diastolic blood pressure (n = 35), eGFR (n = 1022), ACR (n = 1161), and depressive symptoms (n = 191).
Association Between PSS and Incident HF Events
This cohort experienced 1109 incident HF hospitalization or death events over a median of 10.1 years of follow-up, for an incidence rate of 47.7 incident HF events per 10,000 person-years. Among 821 HF hospitalizations, 356 were HFpEF events and 465 were HFrEF events. HF-related deaths could not be classified into HF subtypes because LVEF data were not available.
Participants in the highest PSS quartile had the highest crude incidence of HF at 56.3 events per 10,000 person-years compared with 44.7 events in Q1, 42.6 events in Q2, and 46.6 events in Q3. In the unadjusted models, participants in the highest quartile of PSS had increased hazard of incident HF as compared with those in the lowest quartile of PSS (Q1); but those in Q2 and Q3 did not have a higher hazard than those in Q1. In the fully adjusted model, the HR for incident HF was no longer significantly higher for those in Q4 compared with Q1 (Figure 2). These results were unchanged in a sensitivity analysis accounting for mortality as a competing risk (Supplemental Table 1). There were no significant interactions for sex (P = .84), race (P = .59), age (P = .12), income (P = .63), or baseline statin use (P = .24).
Figure 2.
Forest plot for association of PSS-4 quartile and incident HF. The adjusted model includes age, sex, race, annual household income, educational attainment, geographic region of residence, marital status, social isolation, number of other adults living in the household, smoking status, alcohol use, exercise frequency, atrial fibrillation, history of coronary heart disease, dyslipidemia, diabetes mellitus, body mass index, systolic blood pressure, diastolic blood pressure, left ventricular hypertrophy, estimated glomerular filtration rate, urinary albumin to creatinine ratio, depressive symptoms, physical functioning, and statin use. CI, confidence interval; HF, heart failure; HR, hazard ratio; PSS, Perceived Stress Scale Score.
When examining the relationship between PSS-4 quartile and incident HFrEF, the hazard for incident HFrEF was similar across the PSS-4 quartiles in unadjusted and adjusted analyses (Figure 3). These results were unchanged in a sensitivity analysis accounting for mortality and incident HFpEF as competing risks (Supplemental Table 2). There were no significant interactions for baseline statin use (P = .42), sex (P = .88), race (P = .96), age (P = .96), or income (P = .84).
Figure 3.
Forest plot for Association of PSS-4 quartile with incident HFrEF and incident HFpEF square indicates point estimate for HFrEF and triangle indicates point estimate for HFpEF. The adjusted model includes age, sex, race, annual household income, educational attainment, geographic region of residence, marital status, social isolation, number of other adults living in the household, smoking status, alcohol use, exercise frequency, atrial fibrillation, history of coronary heart disease, dyslipidemia, diabetes mellitus, body mass index, systolic blood pressure, diastolic blood pressure, left ventricular hypertrophy, estimated glomerular filtration rate, urinary albumin to creatinine ratio, depressive symptoms, physical functioning, and statin use. CI, confidence interval; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; HR, hazard ratio; PSS, Perceived Stress Scale Score.
In contrast, when examining the relationship between PSS-4 quartile and incident HFpEF, there was significantly higher hazard of incident HFpEF for those in Q2 (hazard ratio [HR] 1.37, 95% confidence interval [CI] 1.00–1.88), Q3 (HR 1.42, 95% CI 1.03–1.95), and Q4 (HR 1.41, 95% CI 1.04–1.92) compared with those in Q1 (Figure 3). These results were unchanged in a sensitivity analysis accounting for mortality and incident HFrEF as competing risks (Supplemental Table 2). There was an interaction with baseline statin use (P = .07). The HRs for incident HFpEF stratified by baseline statin use are shown in Table 2; as shown, the hazard of incident HFpEF was lower among participants with baseline statin use. There were no other significant interactions (sex P = .85; race P = .45; age P = .15; income P = .38).
Table 2.
Hazard Ratios for Association Between PSS Quartiles and Incident HFpEF Event According to Baseline Statin Use
| Model | Q1 (PSS = 0) | Q2 (PSS = 1–2) | Q3 (PSS = 3–4) | Q4 (PSS ≥ 5) |
|---|---|---|---|---|
|
| ||||
| No baseline statin use | ||||
| N (events) | 4520 (36) | 4320 (56) | 4120 (41) | 5094 (68) |
| Event rate, person-years | 9/10,000 | 14/10,000 | 11/10,000 | 15/10,000 |
| Crude | 1.00 (ref) | 1.62 (1.07,2.46) | 1.27 (0.81,1.98) | 1.80 (1.20,2.70) |
| Adjusted | 1.00 (ref) | 1.78 (1.17,2.71) | 1.32 (0.84,2.08) | 1.57 (1.03,2.40) |
| Baseline statin use | ||||
| Model | Q1 (PSS=0) | Q2 (PSS=1–2) | Q3 (PSS=3–4) | Q4 (PSS≥5) |
| N (events) | 1997 (36) | 1762 (28) | 1653 (43) | 2031 (48) |
| Event rate, person-years | 19/10,000 | 17/10,000 | 29/10,000 | 27/10,000 |
| Crude | 1.00 (ref) | 0.90 (0.55,1.47) | 1.53 (0.98,2.39) | 1.45 (0.94,2.24) |
| Adjusted | 1.00 (ref) | 0.97 (0.59,1.60) | 1.58 (1.01,2.47) | 1.26 (0.80,1.99) |
Adjusted model includes age, sex, race, annual household income, educational attainment, geographic region of residence, marital status, social isolation, number of other adults living in the household, smoking status, alcohol use, exercise frequency, atrial fibrillation, history of coronary heart disease, dyslipidemia, diabetes, body mass index, systolic blood pressure, diastolic blood pressure, left ventricular hypertrophy, estimated glomerular filtration rate, urinary albumin to creatinine ratio, depressive symptoms, physical functioning, and statin use.
The association between PSS-4 as a continuous variable and each outcome in an adjusted model are shown in Supplemental Figures 1–3. In general, there was a pattern of higher log HR with higher PSS-4 score, although the 95% CI were fairly wide in these models.
DISCUSSION
In this geographically diverse US cohort of 25,785 participants, perceived stress was not associated with incident HF. However, PSS-4 was associated with incident HFpEF hospitalization, but not with incident HFrEF hospitalization. Given the increasing prevalence of HF, particularly HFpEF, there is a need to identify population-based strategies to curb the incidence of HF. Prior work assessing the relationship between stress and HF has primarily focused on individuals with baseline stress-related disorders and did not assess differences in HF subtype. To our knowledge, this study is the first effort to quantify the relationship between subclinical stress and incidence of HF stratified by HF subtype.
The underlying pathophysiology of HFrEF and HFpEF differ substantially. Whereas direct myocardial injury and cardiomyocyte loss mediates the development of HFrEF, inflammation at the microvascular level mediates the development of HFpEF.7 Comorbid conditions have previously been implicated as the inciting factors for this cascade in HFpEF. Several of these comorbid conditions, including diabetes mellitus24 and coronary artery disease15 are increased in individuals with elevated stress. Our study reveals that an association between perceived stress and HFpEF remains, even after adjusting for cardiovascular comorbid conditions. It is, therefore, conceivable that psychological stress represents a separate risk factor for HFpEF. This notion is supported by prior literature showing that chronic psychological stress can increase IL-6, C-reactive protein, IL-1-beta, and tumor necrosis factor-alpha—cytokines integral to the proposed pathophysiology of HFpEF.10,11 Although inflammation has been implicated in the pathophysiology of HFrEF as well, this condition is characterized by a different cytokine and metabolomic profile than HFpEF, and thus may respond differently to chronic stress.25,26 Future study is necessary to better understand these complex relationships, identify mediators of our observed association, and determine whether stress-related mechanisms can serve as a basis for future therapeutic strategies. Given prior findings demonstrating an association between social determinants of health and incident HF,20 incorporating the social determinants of health into further examination of stress-related mechanisms is similarly warranted.
Our study additionally suggests that statins may attenuate the association between perceived stress and HFpEF, although this work requires replication. Prior work has shown that statins decrease systemic inflammation,27 providing rationale for the benefits of statins in preventing incident HFpEF. In prevalent HF populations, data have shown that statins decrease sympathetic activity, increase HR variability, and increase availability of nitric oxide28; biopsy specimens have shown that patients with HFpEF who take statins had less cardiomyocyte hypertrophy and lower cardiomyocyte resting tension compared with those who did not take statins.29 It is possible that statins have similar beneficial effects on endovascular and cardiomyocyte function that can prevent HF in those with elevated perceived stress. Importantly, we could not account for time-varying statin use. Moreover, we may not have accounted for confounding variables involved in the complex relationship between statins, inflammation, stress, and HF. Consequently, this finding regarding statins should only be considered hypothesis generating. Given the biological plausibility and potential implications on clinical care,30 future work examining the potential role for statins in this clinical context is warranted.
There were several strengths of this study, including the large sample size, geographic and racial diversity, and adjudicated HF outcomes, which were subtyped into HFrEF and HFpEF. The PSS-4 was collected for all except 5 participants; the low degree of missingness minimized related bias. We also incorporated a broad array of covariates into the statistical models that controlled for several domains, including sociodemographics, social support, unhealthy behaviors, and physiologic parameters. There were also some limitations to note. First, PSS-4 is a self-reported survey and was obtained at a single point at baseline and assessed stress during the prior month. Accordingly, biases common to self-reported surveys like recall bias and social desirability bias could have impacted findings. Also, if the incident HF event occurred after a long latency period, the PSS-4 score may not represent perceived stress at the time of the incident event, especially given the dynamic nature of perceived stress. Moreover, if stress increased over the study period before the incident event, the baseline PSS-4 could have underestimated the cumulative stress over the study period and subsequently led to a degree of misclassification bias. We also did not have data on interim life events that could have increased stress, such as spousal injury or death. Understanding whether stress-related mechanisms underlie the potential negative impact of life events on chronic medication conditions like HF represents another fruitful area for study. Second, although we adjusted for multiple confounders, there was the risk of residual confounding. Third, HF events were adjudicated based on hospitalizations and death and did not capture incident HF diagnosed during an ambulatory visit or through surveillance. Although approximately one-third of incident HF may be diagnosed in the ambulatory setting, a substantial proportion of individuals diagnosed with HF in the ambulatory setting are hospitalized within 1 year of diagnosis and would have thus been captured as an incident HF event at that time.31 Fourth, we could not further subtype HF based on etiology. For example, if patients acquired a stress-related cardiomyopathy, this factor could have had an impact on findings observed here. In contrast, because stress-related cardiomyopathies typically present with reduced ejection fraction, they would have been grouped with HFrEF; because we did not observe an association of perceived stress with HFrEF, the inclusion (or exclusion) of this phenotype would not have changed the overall findings.
Conclusions
Perceived stress was associated higher incidence of HFpEF hospitalization but not HFrEF hospitalization. This association may be attenuated by statins. Taken together, these data support further investigation into targeting stress and inflammation to prevent HFpEF.
Supplementary Material
Highlights.
Persons with higher perceived psychological stress are at higher risk for developing heart failure with preserved ejection fraction, but not heart failure with reduced ejection fraction.
Statins may attenuate the association between psychological stress and heart failure with preserved ejection fraction.
Psychological stress could be a novel target for preventing heart failure with preserved ejection fraction.
Acknowledgments
Supported by cooperative agreement U01 NS041588 co-funded by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), National Institutes of Health, Department of Health and Human Service. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIA. Representatives of the NINDS were involved in the review of the manuscript, but were not directly involved in the collection, management, or analysis or interpretation of the data. The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at: https://www-uab-edu.ezproxy3.lhl.uab.edu/soph/regardsstudy/.
Funding
Additionally supported by R01HL8077 (PI: Safford) from the National Heart, Lung, and Blood Institute (Bethesda, MD).
Footnotes
Disclosures
Dr. Goyal is supported by National Institute on Aging grant K76AG064428; Dr. Goyal has received personal fees for medicolegal consulting on heart failure; and has received honoraria from Akcea Inc and Bionest Inc. Dr. Levitan receives research support from Amgen, has served on Amgen advisory boards, and as a scientific consultant for a research project funded by Novartis. Dr. Sterling is supported by the National Heart Lung and Blood Institute, K23HL150160. Dr. Kronish is supported by the National Institute on Aging, P30AG164098. Dr. Safford has received research support from Amgen.
Tweet
Psychological stress may be a novel risk factor for heart failure with preserved ejection fraction.
REFERENCES
- 1.Shah KS, Xu H, Matsouaka RA, Bhatt DL, Heidenreich PA, Hernandez AF, et al. Heart failure with preserved, borderline, and reduced ejection fraction: 5-year outcomes J Am Coll Cardiol, 70 (2017), pp. 2476–2486 [DOI] [PubMed] [Google Scholar]
- 2.Jacobs L, Efremov L, Ferreira J, Thijs L, Yang W, Zhang Z, et al. Risk for incident heart failure: a subject-level meta-analysis from the heart “OMics” in AGEing (HOMAGE) Study J Am Heart Assoc, 6 (2017), Article e005231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Agarwal S, Chambless L, Ballantyne C, Astor B, Bertoni A, Chang P, et al. Prediction of incident heart failure in general practice: the Atherosclerosis Risk in Communities (ARIC) Study Circ Heart Fail, 5 (2012), pp. 422–429 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Folsom AR, Yamagishi K, Hozawa A, Chambless LE Absolute and attributable risks of heart failure incidence in relation to optimal risk factors Circ Heart Fail, 2 (2009), pp. 11–17 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Cohen S, Janicki-Deverts D, Miller G Psychological stress and disease JAMA, 298 (2007), pp. 1685–1687 [DOI] [PubMed] [Google Scholar]
- 6.Brotman DJ, Golden SH, Wittstein IS The cardiovascular toll of stress Lancet, 370 (2007), pp. 1089–1100 [DOI] [PubMed] [Google Scholar]
- 7.Paulus W, Tschöpe C A novel paradigm for heart failure with preserved ejection fraction: comorbidities drive myocardial dysfunction and remodeling through coronary microvascular endothelial inflammation J Am Coll Cardiol, 62 (2013), pp. 263–271 [DOI] [PubMed] [Google Scholar]
- 8.Roy S, Foraker R, Girton R, Mansfield A Posttraumatic stress disorder and incident heart failure among a community-based sample of US veterans Am J Public Health, 105 (2015), pp. 757–763 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ogilvie RP, Everson-Rose SA, Longstreth WT Jr, Rodriguez CJ, Diez-Roux AV, Lutsey PL Psychosocial factors and risk of incident heart failure Circ Heart Fail, 9 (2016), Article e002243. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Clark S, Song C, Li X, Keegan A, Tonelli L CD8 + T cells promote cytokine responses to stress Cytokine, 113 (2019), pp. 256–264 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Steptoe A, Hamer M, Chida Y The effects of acute psychological stress on circulating inflammatory factors in humans: a review and meta-analysis Brain Behav Immun, 21 (2007), pp. 901–912 [DOI] [PubMed] [Google Scholar]
- 12.Howard VJ, Cushman M, Pulley L, Gomez CR, Go RC, Prineas RJ, et al. The REasons for Geographic and Racial Differences in Stroke study: objectives and design Neuroepidemiology, 25 (2005), pp. 135–143 [DOI] [PubMed] [Google Scholar]
- 13.Goyal P, Mefford MT, Chen L, Sterling MR, Durant RW, et al. Assembling and validating a heart failure-free cohort from the REasons for Geographic and Racial Differences in Stroke (REGARDS) study BMC Med Res Methodol, 20 (2020), pp. 1–9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Cohen S, Kamarck T, Mermelstein R A global measure of perceived stress J Health Soc Behav, 24 (1983), pp. 385–396 [PubMed] [Google Scholar]
- 15.Redmond N, Richman J, Gamboa C, Albert M, Sims M, Durant R, et al. Perceived stress is associated with incident coronary heart disease and all-cause mortality in low- but not high-income participants in the REasons for Geographic and Racial Differences in Stroke study J Am Heart Assoc, 2 (2013), Article e000447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Bailey LN, Levitan EB, Judd SE, Sterling MR, Goyal P, Cushman M, et al. Association of urine albumin excretion with incident heart failure hospitalization in community-dwelling adults JACC Heart Fail, 7 (2019), pp. 394–401 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Goyal P, Balkan L, Ringel JB, Hummel SL, Sterling MR, Kim S, et al. The Dietary Approaches to Stop Hypertension (DASH) diet pattern and incident heart failure J Card Fail, 27 (2021), pp. 512–521 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Butler J, Anker SD, Packer M Redefining heart failure with a reduced ejection fraction JAMA, 322 (2019), pp. 1761–1762 [DOI] [PubMed] [Google Scholar]
- 19.Calvillo-King L, Arnold D, Eubank K, Lo M, Yunyongying P, Stieglitz H, et al. Impact of social factors on risk of readmission or mortality in pneumonia and heart failure: systematic review J Gen Intern Med, 28 (2013), pp. 269–282 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Pinheiro L, Reshetnyak E, Sterling M, Levitan E, Safford M, Goyal P Multiple vulnerabilities to health disparities and incident heart failure hospitalization in the REGARDS Study Circ Cardiovasc Qual Outcomes, 13 (2020), Article e006438. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Scantlebury D, Borlaug B Why are women more likely than men to develop heart failure with preserved ejection fraction? Curr Opin Cardiol, 26 (2011), pp. 562–568 [DOI] [PubMed] [Google Scholar]
- 22.Okin P, Kjeldsen S, Dahlöf B, Devereux R Racial differences in incident heart failure during antihypertensive therapy Circ Cardiovasc Qual Outcomes, 4 (2011), pp. 157–164 [DOI] [PubMed] [Google Scholar]
- 23.Rubin DB, Schenker N Multiple imputation in health-care databases: an overview and some applications Stat Med, 10 (1991), pp. 585–598 [DOI] [PubMed] [Google Scholar]
- 24.Siddiqui A, Desai N, Sharma S, Aslam M, Sinha U, Madhu S Association of oxidative stress and inflammatory markers with chronic stress in patients with newly diagnosed type 2 diabetes Diabetes Metab Res Rev, 35 (2019), p. e3147. [DOI] [PubMed] [Google Scholar]
- 25.Hage C, Michopoulos F, Nilsson R, Davidsson P, Kumar C, Ekström M, et al. Metabolomic profile in HFpEF vs HFrEF patients J Cardiac Fail, 26 (2020), pp. 1050–1059 [DOI] [PubMed] [Google Scholar]
- 26.Kanagala P, Arnold J, Singh A, Chan D, Cheng A, Khan J, et al. Characterizing heart failure with preserved and reduced ejection fraction: an imaging and plasma biomarker approach PloS One, 15 (2020), Article e0232280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.A Oesterle, Laufs U, Liao J Pleiotropic effects of statins on the cardiovascular system Circ Res, 120 (2017), pp. 229–243 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.T Horwich, MacLellan W, Fonarow G Statin therapy is associated with improved survival in ischemic and non-ischemic heart failure J Am Coll Cardiol, 43 (2004), pp. 642–648 [DOI] [PubMed] [Google Scholar]
- 29.Zakeri R, Chamberlain A, Roger V, Redfield M Temporal relationship and prognostic significance of atrial fibrillation in heart failure patients with preserved ejection fraction: a community-based study Circulation, 128 (2013), pp. 1085–1093 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Daubert MA, Douglas PS Primary prevention of heart failure in women JACC Heart Fail, 7 (2019), pp. 181–191 [DOI] [PubMed] [Google Scholar]
- 31.Chamberlain A, Dunlay S, Gerber Y, Manemann S, Jiang R, Weston S, et al. Burden and timing of hospitalizations in heart failure: a community study Mayo Clin Proc, 92 (2017), pp. 184–192 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.