The Modifying Role of Resilience on Allostatic Load and Cardiovascular Disease Risk in the Jackson Heart Study

Abstract

We examined whether resilience modified associations between allostatic load (AL), a physiological indicator of coping with repeated stressors, and cardiovascular disease (CVD) among 2758 African Americans in the Jackson Heart Study. Baseline AL was quantified using biological measures of metabolic, cardiovascular, and immune markers. We constructed a multidimensional resilience measure using validated questionnaires for social support, social networks, religious experiences, and optimism. Participants were followed until 2016 for stroke, coronary heart disease (CHD), and heart failure (HF). We used multivariable-adjusted, sex-stratified Cox regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between dichotomous AL and CVD. High AL was associated with CHD among women (HR = 1.73, 95% CI = 1.00, 2.99) and HF among women (HR = 1.52, 95% CI = 0.98, 2.37) and men (HR = 2.17, 95% CI = 1.28, 3.68). Among women, resilience did not modify the AL-CVD relationship. Among men, we observed higher stroke risk among men with low resilience (HR = 2.21, 95% CI = 0.94, 5.22) and no association among those with high resilience. Counterintuitively, high AL was associated with greater HF (HR = 5.80, 95% CI = 2.32, 14.47) in the subgroup of men with high resilience. Future studies addressing different facets of resilience are needed to elucidate underlying mechanisms for CVD prevention among African Americans.

Introduction

Reductions in the burden of cardiovascular disease (CVD) in the US have not been equitably experienced across the US population. For example, while CVD death rates were similar for African Americans (1071.6 deaths per 100,000) and White adults (1032.3 deaths per 100,000) in the 1970s, the deceleration in mortality has been more pronounced among Whites than African Americans, resulting in a CVD mortality ratio of 1.21 [1]. While the reasons for racial disparities in CVD have not been fully elucidated, a recent area of focus has centered on the role of psychosocial risk factors, including chronic stress, as a potential contributor to the persistent gap in CVD between African Americans and Whites [2, 3].

Experiences of chronic stress, which among African Americans may uniquely result from socioeconomic disadvantage, low neighborhood safety, and discrimination, are thought to disrupt the ability of physiological systems to adapt, resulting in wear and tear on the body, also known as allostatic load (AL) [4, 5]. AL has been conceptualized as the chronic imbalance and possible degradation of the proper function of multiple body systems, including the sympathetic nervous system, the hypothalamic-pituitary-adrenal (HPA) axis, and the immune system, which can be measured through a range of serum-based biomarkers. Those who experience social stigma and discrimination due to race, ethnicity, and sexual orientation, among other factors, may exhibit increased physiological stress responses and allostatic imbalance, which are detrimental to health. Despite various parameterizations of AL in the literature, AL has been linked with poorer health outcomes, including mood and anxiety disorders, cancer, and CVD [6]. Among African Americans specifically, higher AL is associated with worse CVD mortality [7], but it is important to note that not all African Americans facing these stressors develop CVD. As Taylor and colleagues [8] note, “African Americans as a population show the sustained ability to survive an evolving array of social, economic and environmental adversities that date back to more than a century before the founding of the United States.” The mitigating role of unique, positive traits (e.g., resilience) should therefore be explored to inform asset-based interventions that might reduce the disproportionate CVD burden among African Americans.

Although the conceptualization of resilience has evolved over time, fundamentally, resilience reflects positive adaptation, or the ability to overcome stress or adversity [9]. This capacity to adapt has also been framed as the utilization of “resilience resources” to overcome adversity and reduce the impact of negative health-related stressors [10]. Recent operationalizations of resilience have focused on the multidimensional nature of resilience, incorporating individual (e.g., hope and optimism), interpersonal (e.g., close social ties), and community-level (e.g., religious service attendance) resources instead of focusing on these features independently [11, 12]. Mechanistically, greater resilience resources should reduce the physiological stress response through indirect pathways, such as the promotion of adaptive coping and behavioral mechanisms (e.g., physical activity and diet) [13].

In a recent review of 13 studies, Park et al. [10] summarized associations between multilevel resilience resources and CVD and showed that possessing individual-, interpersonal-, and community-level resilience resources was inversely associated with CVD outcomes. In addition to the scant number of studies identified, other limitations of the existing literature were acknowledged, including study populations that were overwhelmingly White, a lack of studies reporting estimates by race, sex, or socioeconomic position, a focus on examining one resilience resource at a time, and the need for studies to specifically explore adversity (i.e., stress) in relation to resilience resources and subsequent health effects.

With respect to this latter limitation, few have examined the extent to which resilience modifies the association between AL and CVD among African American individuals. In a cross-sectional analysis, those living in areas with lower-than-expected rates of CVD had higher odds of select resilience-related characteristics (optimism and purpose of life) [14]. Moreover, in a prior study including more than 10,000 African American women enrolled in the Women’s Health Initiative (WHI), we did not observe effect modification of the association between stressful life events and incident CVD by resilience [15]. Our null findings may be the result of the unidimensional resilience measure (3-item Brief Resilience Scale) or the absence of a biological representation of stress. To overcome prior limitations, we examined (1) whether resilience modified the association between AL and CVD and (2) whether the association differed by sex in a large study of African American adults, with available biological measures to define AL (e.g., hemoglobin A1C [HbA1C] and c-reactive protein [CRP]) and multiple self-report measures to characterize resilience (e.g., social support, social networks, religious experiences, and optimism). We hypothesized that the detrimental impact of AL on CVD risk would be mitigated among women and men with greater self-reported resilience.

Methods

Study Population

For this analysis, we used the JHS, a large, prospective study of risk factors for CVD development and progression in a cohort of 5306 African American adults aged 21–94 years at baseline from the tri-county area (Hinds, Madison, Rankin counties) of metropolitan Jackson, Mississippi. Details about the study have been described [16–18]. In brief, 5306 JHS participants were enrolled between 2000 and 2004 (examination 1) and followed in 2005–2008 (examination 2) and 2009–2013 (examination 3). Study participants completed home interviews, self-administered questionnaires, and inclinic examinations to obtain demographic, psychosocial, behavioral, anthropometric, and clinical data. In the current analyses, participants were excluded if they had baseline CVD (n = 569), if baseline components needed to calculate AL were missing (n = 492), or if measures needed to calculate resilience were missing (n = 1487), resulting in a sample of 2758 adults (mean age 52.3 ± 12.3 years; 65% female). The JHS was approved by the institutional review boards of the University of Mississippi Medical Center, Jackson State University, and Tougaloo College. All participants provided written informed consent.

Measures

Allostatic Load

Based on prior studies [7, 19–21], we selected six physiological indicators (c-reactive protein [CRP], systolic blood pressure [SBP], diastolic blood pressure [DBP], sex-specific waist circumference, the ratio of high-density lipoprotein [HDL] cholesterol to low-density lipoprotein [LDL] cholesterol, hemoglobin A1C [HbA1C]) that were collected at examination 1 to construct AL. For each physiological indicator, the continuous score was dichotomized as 0 or 1, based on cutoffs commonly used in the literature (Supplementary Table 1). The dichotomous variables were then summed to calculate a total AL score (range: 0–6, median = 2). Participants who scored <2 were considered low AL, and those with scores ≥2 were considered to have high AL.

Resilience

In line with recommendations from Park and colleagues [10], our main goal was to understand how concurrent possession of multiple resilience resources impacts the AL-CVD relationship. Therefore, we conceptualized resilience as a multidimensional framework inclusive of four domains: social support, social networks, religious experiences, and optimism [22]. Each domain was measured at examination 1, except optimism which was measured at the first annual follow-up call. Social support was measured using the Interpersonal Social Support Evaluation List (range: 18–62), a scale that includes 16 items that describe emotional support (appraisal), others with whom one can interact (belonging), material aid (tangible), and others with whom one believes she/he compares favorably (self-esteem). Higher scores indicate greater social support. The social network domain was adapted from the Berkman Social Network Index (SNI, range: 0–5), which consists of 5 items that ask participants about the following types of social connections: marital status, number of friends, number of relatives, membership in community groups, and frequency of social contact. Higher scores indicate larger social networks [23]. Religiosity was measured using (1) Daily Spiritual Experiences Scale which includes six items rated on a 6-point scale that assess the frequency with which one finds strength and comfort in religion, is spiritually touched by the beauty of creation, and/or feels God’s presence; (2) Religious Practices, which includes two items that assess how often participants attended worship service and how often they prayed somewhere other than a church/place of worship; and (3) Religious Coping, which included one item rated on a 4-point scale that assessed the extent to which religion was involved with dealing with stress. Combining these three scales produced a religiosity measure that ranged from 9 to 54, with higher scores indicating greater religiosity [24]. Optimism was assessed using the 6-item Life Orientation Test-Revised (LOT-R) Scale (which consists of three positively worded items and three negatively worded items), which were reverse coded to indicate greater optimism, resulting in a total optimism score ranging from 6 to 24 [25].

All four domains were summed to generate a total resilience score (range: 31–147) with higher scores indicating greater resilience. Supplementary Table 2 shows descriptive characteristics of the four individual resilience domains and the overall resilience score in the total study population and according to the CVD subtype. We also assessed correlations between the total resilience score with social support (rho = 0.81), social networks (rho = 0.29), religiosity (rho = 0.73), and optimism (rho = 0.47). Median resilience in this study population was 125; individuals with scores less than 125 were categorized as having low resilience, and those with scores of 125 or greater were categorized as having high resilience. In sensitivity analyses, we also categorized the total resilience score according to tertiles and observed no qualitative differences in our results.

Covariates

Additional data from examination 1, including age (continuous), sex (male or female), educational attainment (less than high school, high school graduate/general equivalency diploma, attended vocational school, trade school, or college, missing), income based on family income, family size and calendar-year-specific poverty level [poor (<federal poverty level), lower middle (1–1.5 times the federal poverty level), upper middle (more than 1.5 but less than 3.5 times the federal poverty level), and affluent (more than 3.5 times the federal poverty level), missing], insurance status (uninsured, private only, public only, or private and public, missing), body mass index kg/m² (BMI, calculated from in-clinic measurements of standing weight and height), menopausal status among women (premenopausal, postmenopausal, missing), current alcohol use (no, yes, missing), physical activity, smoking status, use of statins, antiarrhythmic medications, blood pressure medications (no, yes, missing), and dietary intake, were examined as covariates.

The following cardiovascular health behaviors were categorized using the Life’s Simple 7 metrics defined by the American Heart Association (AHA) [26]: ideal smoking status included those participants who reported never smoking or quit smoking more than a year prior to examination; ideal BMI included participants with BMI <25 kg/m²; ideal physical activity included participant self-reports of 150 min/week or greater of moderate-intensity or 75 min per week or greater of vigorous-intensity physical activity; ideal diet included participants meeting four to five of five of the following recommendations: fruits and vegetables of 4.5 cups/day or more; fish of two 3.5-oz servings per week or more (preferably oily fish); fiber-rich whole grains of three 1-oz-equivalent servings per day or more; sodium less than 1500 mg/d or more; and sugar-sweetened beverages of 450 kcal (36 oz)/week or less; ideal total cholesterol included <200 mg/dL; ideal blood pressure included SBP <120 and DBP <80; and ideal glucose included HbA1c <5.7%, fasting plasma glucose <100 mg/dL; and no report of diabetes medications.

To examine psychosocial confounding, we also included discrimination, stress, and depressive symptoms. We examined self-reported experiences of discrimination using the JHS Discrimination (JHSDIS) Instrument [27]. Everyday discrimination (median = 1.89, range = 1.00–6.89) assessed the occurrence and frequency of everyday discrimination with nine questions such as, “How often on a day-to-day basis do you have the following experiences ‘treated with less courtesy and…less respect…people act as if…you are dishonest…you are threatened.’” Response options for each question ranged from 1 (never) to 7 (several times a day), and the score was calculated as the mean of the nine items. Lifetime discrimination (median = 3.00, range = 0.00–9.00) captured the lifetime occurrence (yes/no) of unfair treatment across 9 domains (school/training, getting a job, at work, getting housing, getting money or resources, getting medical care, on the street or public place, getting services, other reasons). The continuous everyday and lifetime discrimination scores were categorized into tertiles (low, medium, and high).

Chronic stress was measured using the Global Perceived Stress Scale (GPSS), an 8-item scale that assesses the following domain-specific stressors experienced over the previous 12 months: job, relationships, neighborhood, caregiving, legal, medical, racism and discrimination, and meeting basic needs [18]. The GPSS score was categorized into tertiles (low, medium, and high). Baseline reports of depressive symptoms over the last week were quantified using the 20-item Center for Epidemiologic Studies Depression (CES-D) Scale [28]. Participants were asked to indicate whether each item (e.g., “I had trouble keeping my mind on what I was doing”) was experienced on a scale of 0 to 3 (0 = “Rarely”; 3 = “Most or all of the time”) during the past week (some items reverse coded). The CES-D score ranges from 0 to 60, and we categorized individuals with scores ≥16 as having depressive symptoms [29].

Outcomes

Trained and certified personnel performed surveillance for incident stroke, coronary heart disease (CHD), and heart failure (HF) after baseline examination (surveillance for HF began in 2005) through December 31, 2016. Participants were contacted by telephone to identify health events (diagnostic tests, hospitalizations, or death), and then subsequently, the medical record abstraction team obtained discharge lists and death certificates from hospitals and state offices for verification [30, 31]. Eligible events were classified as first definite or probable fatal or nonfatal CHD, HF, and stroke events by a computer algorithm, and follow-up review and adjudication by two independent physician reviewers was conducted. Any disagreements in diagnoses were adjudicated by another reviewer. Details on the quality assurance for ascertainment and classification of CVD events in JHS have been previously published [31]. Incident stroke events were defined as definite or probable stroke from neuroimaging studies and autopsy based on classification from the National Survey of Stroke [32]. The minimum criterion for a definite or probable stroke was a sudden or rapid onset of neurological symptoms lasting for >24 h or leading to death. Neurologic symptoms that did not last >24 h or symptoms seen before or during admission to the hospital were not considered to be a definite or probable stroke event. Stroke events that occurred outside of the hospital or without medical diagnosis were also not considered definite or probable. Incident CHD events were characterized as a definite or probable hospitalized myocardial infarction or fatal CHD (or cardiac procedure). Classification of definite or probable myocardial infarction was based on combinations of chest pain symptoms, electrocar-diogram (ECG) changes, and cardiac enzyme levels. Fatal CHD was based on chest pain symptoms, underlying cause of death from the death certificate, and other associated hospital information or medical history. The criterion for the cardiac procedure was based on receipt of angiography and any revascularization procedures, as indicated in the medical records. Incident HF was defined as the first occurrence of either inpatient or outpatient diagnosis of unspecified failure of the heart, according to the International Classification of Diseases, Ninth Revision (ICD‐9) code 428, and/or an underlying cause of death code of 150. The definition of HF also included, but was not limited to, radiographic findings that were similar with congestive HF, increased venous pressure >16 mmHg, dilated ventricle/left ventricular function <40% by echocardiography/multiple gated acquisition, or autopsy finding of pulmonary edema.

Statistical Analysis

Baseline characteristics according to AL and resilience were evaluated with chi-square tests or t-tests. Cumulative incidence graphs for incident stroke, CHD, and HF according to dichotomous AL were generated using Kaplan-Meier plots, and P values were calculated from log-rank tests. We used Cox proportional hazards regression to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between AL and time to incident stroke, CHD, and HF stratified by sex. The time between the date of JHS enrollment and the date of the CVD event (or date of the last contact) was the underlying time metric. In our primary analysis, we estimated cause-specific (i.e., CVD) hazard functions by censoring all deaths that were not CVD related on their date of death and censoring individuals who did not experience a CVD event/death during the follow-up at their date of loss-to-follow-up. We tested the effect measure modification of the sex-stratified AL-CVD risk relationships according to resilience (low vs. high) by including a multiplicative interaction term in the Cox proportional hazards regression model. In all models, we adjusted for the following covariates, which were significantly associated with AL (chi-square p-value from Table 1) and any of the CVD outcomes (Wald chi-square p-value from Cox univariable models, data not tabled): age, education, income, insurance status, menopausal status (among women), current alcohol status, use of blood pressure medication, and ideal AHA physical activity. We omitted the Life’s Simple 7 metrics corresponding to BMI, total cholesterol, blood pressure, and glucose to avoid over-adjustment for factors represented in the AL definition. Proportional hazards were tested with a multiplicative interaction term between AL and follow-up time (natural log scale), and we observed no significant deviations.

Table 1.

Baseline characteristics of participants in the Jackson Heart Study according to allostatic load and resilience

	Allostatic load			Resilience
	Low (<2) (n = 1340)	High (≥2) (n = 1418)	p a	Low (<125) (n = 1338)	High (≥125) (n = 1420)	p a
N (%)b	1340 (48.6)	1418 (51.4)		1338 (48.5)	1420 (51.5)
Mean ± SD
Age (n = 2758)	51.7±12.5	53.8±11.9	<0.0001	52.3±12.5	53.3±12.0	0.04
Sex			<0.0001			0.0002
Women (n = 1799)	750 (41.7)	1049 (58.3)		826 (45.9)	973 (54.1)
Men (n = 959)	590 (61.5)	369 (38.5)		512 (53.4)	447 (46.6)
Education			<0.0001			<0.0001
<High school (n = 332)	135 (40.7)	197 (59.3)		212 (63.9)	120 (36.1)
High school or GED (n = 513)	219 (42.7)	294 (57.3)		274 (53.4)	239 (46.6)
Vocational, trade, or college (n = 1911)	985 (51.5)	926 (48.5)		851 (44.5)	1060 (55.5)
Income			<0.0001			<0.0001
Poor (n = 280)	118 (42.1)	162 (57.9)		190 (67.9)	90 (32.1)
Lower middle (n = 517)	231 (44.7)	286 (55.3)		309 (59.8)	208 (40.2)
Upper middle (n = 762)	356 (46.7)	406 (53.3)		342 (44.9)	420 (55.1)
Affluent (n = 823)	458 (55.7)	365 (44.4)		334 (40.6)	489 (59.4)
Insurance status			<0.0001			<0.0001
Uninsured (n = 357)	155 (43.4)	202 (56.6)		219 (61.3)	138 (38.7)
Private (n = 1751)	917 (52.4)	834 (47.6)		790 (45.1)	961 (54.9)
Public (n = 335)	123 (36.7)	212 (63.3)		192 (57.3)	143 (42.7)
Private and public (n = 309)	143 (46.3)	166 (53.7)		134 (43.4)	175 (56.6)
Current alcohol use			0.008			<0.0001
No (n = 1420)	650 (45.8)	770 (54.2)		627 (44.2)	793 (55.9)
Yes(n = 1326)	685 (51.7)	641 (48.3)		707 (53.3)	619 (46.7)
Menopausal statusc			0.001			0.27
Premenopausal (n = 236)	105 (44.5)	131 (55.5)		115 (48.7)	121 (51.3)
Postmenopausal (n = 1123)	432 (38.5)	691 (61.5)		499 (44.4)	624 (55.6)
Life’s Simple 7 Metrics
Ideal AHA smoking			0.53			<0.0001
No (n = 329)	154 (46.8)	175 (53.2)		210 (63.8)	119 (36.2)
Yes (n = 2401)	1170 (48.7)	1231 (51.3)		1115 (46.4)	1286 (53.6)
Ideal AHA body mass index			<0.0001			0.57
No (n = 2371)	1014 (42.8)	1357 (57.2)		1152 (48.6)	1219 (51.4)
Yes (n = 386)	326 (84.5)	60 (15.5)		185 (47.9)	201 (52.1)
Ideal AHA physical activity			<0.0001			0.0003
No (n = 2157)	989 (45.9)	1168 (54.2)		1086 (50.4)	1071 (49.7)
Yes (n = 601)	351 (58.4)	250 (41.6)		252 (41.9)	349 (58.1)
Ideal AHA nutrition			0.76			0.007
No (n = 2502)	1218 (48.7)	1284 (51.3)		1196 (47.8)	1306 (52.2)
Yes (n = 38)	20 (52.6)	18 (47.4)		15 (39.5)	23 (60.5)
Ideal AHA total cholesterol			<0.0001			0.64
No (n = 1477)	643 (43.5)	834 (56.5)		707 (47.9)	770 (52.1)
Yes (n = 1267)	689 (54.4)	578 (45.6)		623 (49.2)	644 (50.8)
Ideal AHA blood pressure			<0.0001			0.56
No (n = 2127)	921 (43.3)	1206 (56.7)		1039 (48.9)	1088 (51.2)
Yes (n = 623)	415 (66.6)	208 (33.4)		294 (47.2)	329 (52.8)
Ideal AHA glucose			<0.0001			0.81
No (n = 1415)	523 (37.0)	892 (63.0)		695 (49.1)	720 (50.9)
Yes (n = 1328)	809 (60.9)	519 (39.1)		636 (47.9)	692 (52.1)
Use of blood pressure medications			<0.0001			0.17
No (n = 1508)	851 (56.4)	657 (43.6)		748 (49.6)	760 (50.4)
Yes (n = 1204)	465 (38.6)	739 (61.4)		573 (47.6)	631 (52.4)
Use of antiarrhythmic medications			0.005			0.04
No (n = 2640)	1293 (49.0)	1347 (51.0)		1276 (48.3)	1364 (51.7)
Yes (n = 70)	21 (30.0)	49 (70.0)		43 (61.4)	27 (38.6)
Use of statin medications			0.36			0.06
No (n = 2419)	1186 (49.0)	1233 (51.0)		1193 (49.3)	1226 (50.7)
Yes (n = 289)	129 (44.6)	160 (55.4)		126 (43.6)	163 (56.4)
Lifetime discrimination			0.10			0.59
Low (n = 766)	361 (47.1)	405 (52.9)		360 (47.0)	406 (53.0)
Medium (n = 921)	431 (46.8)	490 (53.2)		449 (48.8)	472 (51.3)
High (n = 1071)	548 (51.2)	523 (48.8)		529 (49.4)	542 (50.6)
Everyday discrimination			0.46			<0.0001
Low (n = 917)	432 (47.1)	485 (53.9)		380 (41.4)	537 (58.6)
Medium (n = 816)	396 (48.5)	420 (51.5)		395 (48.4)	421 (51.6)
High (n = 1025)	512 (50.0)	513 (50.1)		563 (54.9)	462 (45.1)
Global Perceived Stress			0.98			<0.0001
Low (n = 859)	416 (48.4)	443 (51.6)		367 (42.7)	492 (57.3)
Medium (n = 956)	467 (48.9)	489 (51.2)		441 (46.1)	515 (53.9)
High (n = 943)	457 (48.5)	486 (51.5)		530 (56.2)	413 (43.8)
Depression			0.004			<0.0001
No (n = 2345)	1166 (49.7)	1179 (50.3)		1036 (44.2)	1309 (55.8)
Yes (n = 413)	174 (42.1)	239 (57.9)		302 (73.2)	111 (26.9)
Allostatic load			—			0.32
Low (<2)	—	—		637 (47.5)	703 (52.5)
High (≥2)	—	—		701 (49.4)	717 (50.6)

^aChi-square p-value

^bRow percentage

^cAmong women

We conducted several sensitivity analyses. First, we modeled non-CVD deaths as competing risks using the Fine-Gray model [33]. Second, we ran models with additional adjustments for variables that did not meet our above definition for confounding (ideal AHA smoking status, ideal AHA nutrition, depressive symptoms, everyday discrimination, lifetime discrimination, and GPSS) and observed no difference from our main models. Third, we excluded the 13.6% of individuals with missing income and did not observe a qualitative difference from our main results. Fourth, we examined the effect modification of the AL-CVD relationship by each individual domain (i.e., social support, social networks, religious experiences, and optimism). The median of each domain was used as the cutoff to denote low vs. high. Finally, we modeled the overall resilience score as a continuous measure instead of dichotomizing at the median. Statistical analyses were performed using SAS (version 9.4, SAS Institute, Cary, NC) and Stata software (version 14, STATA Corp., College Station, TX). All P values were two-sided with the probability of a type I error set at <5%.

Results

Participant Characteristics

Table 1 shows baseline characteristics of JHS participants according to AL and resilience. High AL was associated with older age and was more common among females, those with less than high school education, low income, those with public insurance, non-drinkers, postmenopausal women, those with non-ideal BMI, non-ideal physical activity, non-ideal total cholesterol, non-ideal blood pressure, or non-ideal glucose, users of blood pressure or antiarrhythmic medications, and those with clinically relevant depression scores. We mostly observed the opposite pattern of associations for factors associated with resilience, apart from age, sex, and current alcohol use: Older age, non-drinkers, and women had higher resilience compared with those of younger age, drinkers, and men, respectively. On the other hand, higher resilience was observed among those with higher education and income, those with private and public health insurance, those with ideal smoking, physical activity or nutrition, non-users of antiarrhythmic medications, and those with lower reports of everyday discrimination, lower perceived stress, and lower depression scores. AL and resilience were not associated in this study population.

The Modifying Role of Resilience on AL and Incident CVD in Sex-stratified Models

Between baseline (2005 for HF) and 2018, 91 strokes, 129 CHD, and 166 HF events occurred. Sex-stratified Kaplan-Meier cumulative incidence plots are shown in Fig. 1 – AL was unrelated to stroke among both sexes, while high AL was associated with increased risk of CHD among women only (log-rank p = 0.005) and HF risk among both women (log-rank p = 0.001) and men (log-rank p = 0.002). Among women, high AL was associated with higher risk of CHD (HR = 1.73, 95% CI = 1.00–2.99) and HF (HR = 1.52, 95% CI = 0.98–2.37, Table 2) in multivariable-adjusted models. Resilience did not significantly modify the association between AL and any CVD outcome among women.

Fig. 1.

Kaplan-Meier cumulative incidence plots of (a) stroke (women), (b) stroke (men), (c) CHD (women), (d) CHD (men), (e) HF (women), and (f) HF (men) according to low vs. high AL

Table 2.

Hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between high (vs. low) allostatic load and incident stroke, coronary heart disease, and heart failure according to resilience in sex-stratified models

	Women Overall (n = 1799)	Low resilience (n = 826)	High resilience (n = 973)	p int b
Stroke (n = 55)
Unadjusted	1.02 (0.60, 1.74)	1.43 (0.62, 3.32)	0.78 (0.38, 1.60)	0.28
Multivariable-adjusteda	0.78 (0.45, 1.36)	1.28 (0.53, 3.07)	0.57 (0.27, 1.21)	0.24
Coronary heart disease (n = 70)
Unadjusted	2.14 (1.25, 3.65)	2.17 (1.03, 4.57)	2.04 (0.95, 4.42)	0.91
Multivariable-adjusteda	1.73 (1.00, 2.99)	1.92 (0.88, 4.23)	1.61 (0.73, 3.54)	0.99
Heart failure (n = 103)
Unadjusted	1.62 (1.04, 2.51)	1.71 (0.97, 3.00)	2.47 (1.25, 4.89)	0.41
Multivariable-adjusteda	1.52 (0.98, 2.37)	1.23 (0.69, 2.19)	1.87 (0.93, 3.76)	0.33
	Men Overall (n = 959)	Low resilience (n = 512)	High resilience (n = 447)	p int b
Stroke (n = 36)
Unadjusted	1.44 (0.75, 2.77)	2.40 (1.04, 5.55)	0.50 (0.14, 1.83)	0.05
Multivariable-adjusteda	1.23 (0.62, 2.43)	2.21 (0.94, 5.22)	0.30 (0.06, 1.55)	0.02
Coronary heart disease (n = 59)
Unadjusted	1.46 (0.88, 2.44)	1.17 (0.57, 2.41)	1.86 (0.90, 3.84)	0.38
Multivariable-adjusteda	1.26 (0.74, 2.15)	1.12 (0.52, 2.41)	1.78 (0.80, 3.98)	0.50
Heart failure (n = 63)
Unadjusted	2.19 (1.33, 3.61)	1.38 (0.69, 2.75)	3.63 (1.71, 7.71)	0.06
Multivariable-adjusteda	2.17 (1.28, 3.68)	1.33 (0.65, 2.72)	5.80 (2.32, 14.47)	0.09

^aMultivariable model adjusted for age (continuous), education (<high school, high school or GED, vocational, trade, or college, missing), income (poor, lower middle, upper middle, affluent, missing), insurance status (uninsured, private, public, private and public, missing), current alcohol status (no, yes, missing), ideal AHA physical activity (no, yes), menopausal status among women (premenopausal, postmenopausal, missing), and use of blood pressure medication (no, yes, missing)

^bp_int = p-value for one degree of freedom interaction between AL and resilience

Among men, high AL was associated with a higher risk of HF (HR = 2.17, 95% CI = 1.28–3.68, Table 2). Resilience was a significant effect modifier of the relationship of AL with stroke (p-interaction = 0.02). High AL was associated with two-fold higher stroke risk in the low resilience group (HR = 2.21, 95% CI = 0.94–5.22) and an almost six-fold higher HF risk (HR = 5.80, 95% CI = 2.32–14.47) in the high resilience subgroup.

Sensitivity Analyses

The subdistribution hazard ratios in the unadjusted and adjusted models did not differ after accounting for competing risks of non-CVD death (Supplementary Table 3). Effect modification by the individual domains of social support, social networks, religiosity, and optimism demonstrated significant modification of the association between AL and HF risk by social networks among women (p-interaction = 0.01) and between AL and stroke risk by religiosity among men (p-interaction = 0.02, Supplementary Table 4). Similar to the main models among women, we did not observe significant interactions between AL and resilience when the latter was treated as a continuous variable. In contrast to the main male-specific models, we did not detect significant interactions between AL and resilience for any CVD outcome when resilience was treated as a continuous variable (Supplementary Table 5).

Discussion

In this large study of community-dwelling African American men and women, we observed positive associations between AL and CVD risk, with variations according to CVD subtype, sex, and self-reported resilience. Among women, we observed that those with high AL had an increased risk of CHD and HF, while among men, an increased risk of HF was observed. In line with our hypotheses, resilience was an important effect modifier such that the association between high AL and stroke risk was significant only among men with low versus high resilience. Counter to our hypothesis, AL was significantly associated with a higher risk of HF among men who reported high resilience. Our results may reflect that strong social support, more dense social networks, higher religiosity, and greater optimism – constructs reflective of resilience resources [34] – are not enough to dampen the harmful effects of AL in contributing to HF among African American men.

First coined by McEwen and Stellar [5], AL is a framework that describes the physiological consequences of repeated efforts to cope with environmental challenges that are perceived by the individual to be stressful. Once the individual capacity to manage chronic environmental challenges occurs and buffering factors are insufficient, activation of neuroendocrine, cardiovascular, and emotional responses ensues, leading to disruptions in the coronary and cerebral artery blood flow, high blood pressure, and atherogenesis [35, 36]. Recurrent activation of the stress response, over time, results in “wear and tear,” also known as AL. Since its introduction, investigators have used the AL framework to examine racial disparities in health outcomes, with the rationale that repeated experiences of discrimination and racism among African Americans manifest in poor health outcomes through AL [7, 37]. Our main effects models demonstrating increased risks of certain CVD subtypes associated with AL, which support McEwen and Stellar’s framework, are also in line with the scant literature on AL and CVD among African American adults [7, 38, 39]. These observations motivated our evaluation of how individual assets of African Americans, specifically resilience, might alter the association between AL and greater CVD risk. Indeed, African Americans are considered to have developed resilience as a counter to the pernicious historical trauma of slavery and the transgenerational stressors that persist as a consequence of living in a race-conscious culture that under-values African Americans [8].

Theoretical models and empirical research support the plausibility of resilience, commonly conceptualized as thriving in the face of adversity, as a moderator of the AL-CVD relationship. Our hypothesis that the AL-CVD relationship would be blunted among those with high resilience was driven by Lazarus and Folkman’s transactional model of stress and coping [40], which posits that in light of stressful stimuli, individuals engage in conscious, purposeful coping actions to either directly manage stress/stressors or regulate emotions that arise as a consequence of stress/stressors. Social resources, including strong social networks and engagement in religious activities, are considered intrinsically valuable resources that should buffer stress. In addition, personality traits, such as optimism, are predicted to influence human behavior and promote health [41].

Our analysis was further motivated by research examining aspects of resilience in relation to AL with reports of better AL profiles among those with resilient qualities. In a study of 167,729 persons living in the North East regions of the Netherlands, there was an inverse association between positive effect (e.g., enthusiasm, excitement, and inspiration) and AL [42]. Moreover, better social relationships such as having higher spousal support were inversely associated with AL among adults from the Midlife in the United States (MIDUS) study [43]. In addition, a lower CVD risk factor burden among individuals with higher levels of psychosocial resilience or resilience constructs (e.g., optimism and social support/networks) has been observed, providing an indirect pathway to lower CVD risk [25, 44–46].

The finding of a stronger association between AL and stroke risk among men with low resilience, but not among those with high resilience, is in line with the transactional model of stress and coping, whereby resilience resources mitigate the impact of stress. Likewise, our CHD results among women also support the possibility of resilience reducing poor outcomes related to stress. However, the non-significant interaction in the CHD model warrants a tempered interpretation. In opposition to our hypotheses and theoretical frameworks are the observations that high AL was associated with increased HF risk among men with high resilience. A similar association was noted for women, but again, the interaction was non-significant. While we lack a definitive mechanistic explanation for these unexpected observations, and therefore urge caution in the interpretation, others have observed similar unanticipated associations when examining the effect moderation of other health outcomes by psychosocial resources. For example, Perry and colleagues [47] observed no association between gendered racism experiences and suicidal ideation among African American women with moderate levels of psychosocial resources, while stronger adverse associations were noted among women with high or low levels of psychosocial resources. It is possible that the effort needed to sustain greater social support/networks and engagement in religious activities presents a burden at higher levels of this construct with implications for HF risk.

Although some of our findings portray resilience as an accelerator of poor CVD outcomes among African American adults, we believe there is still a need to consider the potentially beneficial effects of harnessing resilience for lower CVD risk. Cognitive-behavioral and mindfulness interventions have been developed and show promise to reduce ineffective coping strategies and factors related to CVD risk (e.g., inflammation) [48]. Further study is warranted to characterize mechanisms that underlie the associations revealed in this study, with a particular focus on sex differences among African American adults. From the onset, we examined sex-specific associations as we theorized that men and women likely differ in their use of resilience resources. For example, a previous study in JHS reported sex differences in religious coping and active coping, where men were less religious and used more active coping when dealing with discrimination compared with women [49]. African American women are commonly described as portraying a strong woman persona that influences coping with stressful exposures, known as the Superwoman Schema [50]. Two key characteristics of the Superwoman Schema may be particularly relevant in explaining our findings among women: resistance to being vulnerable or dependent and prioritization of caregiving over self-care. These particular facets may prevent African American women from reaping the benefits of strong social networks/social support and engagement in religious activities in the same way that men benefit. Perhaps the need to resist vulnerability and prioritize the needs of others coupled with greater social responsibilities represent additional stressors for African American women that impact CHD and HF risk.

Our parameterization of resilience was informed by a recent review by Park and colleagues, which stressed the need for multidimensional resilience measures in the study of CVD [10]. As such, we created a novel multidimensional resilience construct that incorporated multiple psychosocial domains, including internal assets (i.e., optimism) along with external resources (i.e., social support) available to cope with stress. An important caveat is that the total resilience score we used in this study gives more weight to domains with more items. For example, the social support domain includes 16 items compared to 6 items in the optimism domain – variation that is relevant in generating the total resilience score. In addition, studies that examine associations between individual resilience resources and CVD have demonstrated stronger effects with social support and social networks than optimism or religiosity [10]. While our goal was to consider the combination of available resilience resources, there is a clear need for methodological research to produce standard multidimensional resilience measures that can be integrated within the existing large CVD cohorts. Furthermore, certain aspects that may contribute to resilience (e.g., purpose in life and acceptance of/adaptation to change) were unexplored in this analysis.

This study has several strengths aside from the inclusion of a large cohort of African American adults with a wide range of income and education levels. The rigorous ascertainment of physiologic and laboratory measures, use of validated questionnaires for the resilience constructs, and over a decade of follow-up for adjudicated CVD events are key features that limit exposure and outcome misclassification. Moreover, our use of a multidimensional resilience framework adds a significant contribution to the literature by addressing the roles of support systems, spirituality, and positive orientation in moderating CVD risk. Several limitations exist. These findings may not generalize to the US African American population, as our study population is regional and may have unique situational challenges. Moreover, there is potential for selection bias, as we excluded 28% of participants with missing information to characterize the total resilience score. We examined distributions of sex, AL, and cardiac outcomes according to whether or not participants were missing baseline resilience measures, and we noted slightly higher proportions of high AL and stroke among those with missing resilience measures. Another limitation is that all psychosocial measures were self-reported and therefore potentially affected by reporting bias. Finally, we lacked a traditional assessment of resilience, defined as the ability to thrive when faced with adversity.

In conclusion, this is the first manuscript to report on the potential moderation of resilience in the association of AL and CVD among African American adults. We observed a higher risk of developing certain CVD subtypes associated with higher AL. With respect to effect modification by resilience, some of the observed findings (e.g., stroke) were in line with our hypotheses; however, for other CVD subtypes, our findings that higher levels of resilience did not attenuate the impact of AL were not expected. Future studies, with additional constructs of resilience and repeated assessments of resilience and AL might shed light on the pathways leading to CVD. Without a clear mechanism by which the AL-CVD association would be stronger among those with high resilience resources, our findings require a cautious interpretation.

Data Availability

Requests to access the data set from qualified researchers may be sent to the Jackson Heart Study (JHS) data coordinating center.