Evidence for the Association between Adverse Childhood Family Environment, Child Abuse, and Caregiver Warmth and Cardiovascular Health across the lifespan: The Coronary Artery Risk Development in Young Adults (CARDIA) Study

Abstract

Background:

This study aimed to quantify the association between childhood family environment and longitudinal cardiovascular health (CVH) in adult Coronary Artery Risk Development in Young Adults (CARDIA) Study participants. We further investigated whether the association differs by adult income.

Methods:

We applied the CVH framework from the American Heart Association including metrics for smoking, cholesterol, blood pressure, glucose, body mass index, physical activity, and diet. CVH scores (range 0–14) were calculated at years 0, 7 and 20 of study. Risky Family environment (range 7–28) was assessed at year 15 retrospectively, for childhood experiences of abuse, caregiver warmth, and family or household challenges. Complete case ordinal logistic regression and mixed models associated RF (exposure) with CVH (outcome), adjusting for age, sex, race, and alcohol use.

Results:

The sample (n=2,074) had a mean age of 25.3 (±3.5) years and 56% females at baseline. Median RF was 10 with ideal CVH (≥12) met by 288 individuals at baseline (28.4%) and 165 (16.3%) at Year 20. Longitudinally, for every 1-unit greater Risky Family, the odds of attaining high CVH (≥10) decreased by 3.6% (OR=0.9645, 95%CI 0.94–0.98). Each unit greater child abuse and caregiver warmth score corresponded to 12.8% lower and 11.7% higher odds of ideal CVH (≥10), respectively, (OR=0.872, 95%CI 0.77, 0.99; OR=1.1165, 95%CI 1.01–1.24), across all 20 years of follow-up. Stratified analyses by income in adulthood demonstrated associations between risky family environment and CVH remained significant for those of highest adult income (>$74k), but not lowest (<$35k).

Conclusion:

While Risky Family environmental factors in childhood increase odds of poor longitudinal adult CVH, caregiver warmth may increase the odds of CVH, and socioeconomic attainment in adulthood may contextualize level of risk. Toward a paradigm of primordial prevention of CVD, childhood exposures and economic opportunity may play a crucial role in CVH across the lifecourse.

Introduction

Cardiovascular disease (CVD) is the leading cause of death in the United States, with mounting evidence of its beginnings (e.g., atherosclerosis and CVD risk factors) in early childhood.¹ To reduce the lifecourse morbidity and mortality caused by CVD, the American Heart Association (AHA) defined cardiovascular health (CVH) metrics for Americans to reduce the burden of CVD by 20%.² A high CVH score, indicating more ideal CVH, has been associated with lower all-cause mortality,³ and is inclusive of metrics for smoking status, body mass index (BMI), physical activity, total cholesterol, blood pressure (BP), fasting glucose, dietary intake,² and, most recently, sleep.⁴ Despite the life-saving potential in the attainment of ideal CVH, few adults are able to meet the individual components with the lowest rates in Black and Latino individuals.⁵ An early report indicated only 1.1% of a multi-ethnic cohort met all seven criteria, and just under 20% met 5–6 criteria.⁶ In another multi-ethnic cohort of just under 3,000 individuals, none met 7 criteria and only 4.4% had 5–6 criteria.⁷

With the early origins of CVD in childhood, and poor attainment of ideal CVH in adulthood, it is evident that to meet the AHA goal, risk factor mitigation must begin in childhood. This has prompted the AHA to emphasize the importance of primordial prevention, or the prevention of risk factors for CVD, with a lifecourse approach.^4,8 Specifically, Adverse childhood experiences (ACEs), which include abuse or maltreatment, and family or household challenges, such as caregiver substance use, are associated with higher odds of CVD and all-cause mortality.^9,10 ACEs are a subset of experiences in childhood hypothesized to be linked to lifecourse outcomes through their contribution to early life stress (ELS), which is particularly impactful in childhood, a period of development, learning, and growth.¹¹ Only one identified study, by Islam and colleagues, assessed the link between ELS and the AHA CVH composite score, specifically.¹² This study did show that ELS (as measured by a total of ACEs and other traumas outside of the family environment, taken together) was associated with lower CVH score in adulthood.¹² However, this was a cross-sectional study that was conducted in a single geographical location.¹²

Research supports that the most impactful early life exposures to lifecourse health outcomes are those related to interpersonal relationships termed “relational health”, by the American Academy of Pediatrics. Challenges to relational health (e.g., abuse), and safe, stable, and nurturing relationships (e.g., caregiver warmth), are both impactful on future health outcomes.¹³ Only one identified study explored and found an association between greater positive childhood experiences, inclusive of exposure to healthy relationships, and higher CVH scores in adulthood.¹⁴ Though numerous studies have linked ELS to CVD and CVD risk factors, studies exploring the association between childhood experiences and the AHA’s CVH score (a composite of protective factors against CVD taken together) have been limited to cross-sectional analyses.^12,14 Additionally, further research is needed to explore the comparison and relationship between adverse and protective childhood experiences on lifecourse CVH in the same study.

The CARDIA Study includes longitudinal follow-up of a large sample of individuals with demographic diversity (e.g., diverse SES, geographic diversity across the United States), and inclusivity of individuals who identify as White and individuals who identify as Black. The characteristics of this cohort offer an opportunity to further characterize the association between childhood environment and CVH. The CARDIA Study collected retrospective data using a questionnaire, the Risky Family Environment questionnaire (RF), that broadly captures experiences associated with ELS, and is closely aligned with measurements of ACEs as defined in prior studies. In this study, the primary aim was to examine the association between childhood RF environment and CVH longitudinally, across multiple time points from early to mid-adulthood, over 20 years of follow-up in the CARDIA cohort. It was hypothesized that RF would be associated with decreased attainment of ideal CVH, sustained over time. To explore the impact of exposures of RF that are directly related to relational health, sub-scales of the RF questionnaire for child abuse and caregiver warmth were also used. We hypothesized that greater exposure to child abuse would be associated with poorer longitudinal CVH across adulthood, but that experiences of caregiver warmth would buffer against this relationship when testing the interaction between child abuse and caregiver warmth. Lastly. the current study considered the important contextualizing role of structural drivers of health that may influence lifecourse CVH trajectory, as found in the prior study by Islam and colleagues. These investigators observed that the relationship between childhood trauma and poor adult CVH was significant only in adults in the lowest tier of annual income (<$50,000 vs. ≥$50,000) in a population of Black adults.¹² Therefore, the secondary aim of this study was to examine whether the relationship between RF and CVH differs by income in adulthood.

Methods

The CARDIA Study is a prospective epidemiologic study initiated between 1985 and 1986 with ongoing data collection at multiple sites across the United States (Minneapolis, Minnesota; Birmingham, Alabama; Chicago, Illinois; and Oakland, California). It aims to explore the development of risk factors and CVD in Black and white adults initially enrolled at 18–30 years of age. Details related to the CARDIA study design have described previously,¹⁵ and the data is publicly available upon request to interested researchers.

For the current study, the cohort for analysis included 2,074 individuals. Participants who had complete Risky Family (RF) Environment Questionnaire data at year 15, as part of an ancillary CARDIA study cohort (n=3,650), were included. Individuals were excluded if missing CVH component data (years 0, 7, or 20 of follow-up; n=1,383), or covariate data (n=193) (Figure 1). The timeline of data collection in The CARDIA Study for variables applied to the current study can be found in Figure S1. The CARDIA study was approved by institutional review committees at each respective location and all participants gave informed consent.

Figure 1.

Flow chart demonstrating complete data for final analytic sample (n=2074) in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

Participants were excluded for having incomplete Risk Family (RF) data (at year 15), incomplete Cardiovascular Health (CVH) score component data (at any of the CARDIA study years), or covariate data yielding a complete case sample for analysis (n=2,074).

Exposure: Risky Family Environment

At year 15 of CARDIA study follow-up, the Risky Family (RF) Environment Questionnaire (also called “Risky Families” questionnaire) was collected. The questionnaire asks about experiences before the age of 18 years of age (retrospectively). It has been tested for reliability and validity and includes 7 items assessing experiences in the childhood environment (i.e., emotional abuse, physical abuse, substance use in the home, and adult affection, and support), as well as family awareness of a child’s whereabouts, and household disorganization, each scores with a 4-point Likert scale related to the frequency of exposure (“rarely or none of the time”, “some of little of the time”, “occasionally or moderate amount of time”, or “most or all of the time”).¹⁶ The RF is scored by summing all items after reverse coding the items related to adult affection and support. For the total RF score, higher score indicates more adverse RF environment (range 7–28). The questionnaire has been used prior to calculate sub-scales including a child abuse sub-scale and caregiver warmth sub-scale.¹⁷

Exposures: Early Childhood Relational Health

The child abuse and caregiver warmth sub-scales of the RF Questionnaire were calculated as in prior studies,¹⁷ to assess specific aspects of the family environment related to early childhood relational health captured by the RF Questionnaire. The child abuse subscale was a mean of the values for the RF Questionnaire items: “How often did a parent or other adult in the household…” 1) “…push, grab, shove, or hit you so hard you had marks or were injured?” and 2) “…swear at you, insult you, put you down, or act in a way that made you feel threatened?” (range 1–4). The caregiver warmth sub-scale was assessed adult affection and support as: (“How often did a parent or other adult in the household…” 1) “…express physical affection for you, such as hugging or other physical gesture of warmth and affection?” and 2) “…make you feel that you were loved, supported, and cared for?”), (range 1–4).

Outcome: Cardiovascular Health

Consistent with the American Heart Association (AHA) Strategic Planning Task Force and Statistical Committee 2020 Guidelines,² the CVH score included the original seven components, using data collected in CARDIA on tobacco use, in a physical activity questionnaire, from the diet history, and through metrics for height, weight, lipids, blood pressure (BP), and fasting glucose all at baseline and follow-up years 7 and 20. The AHA CVH index was computed from the 7 health metrics used to define ideal CVH, as defined in the 2020 Impact Goals (Table S1), scored as 1 for presence and 0 for absence. The values for each metric were used to derive a summary score. For each participant, this index ranges from 0 (poor CV health) to 7 (ideal CV health). We then calculated a CVH score for each of the 7 components to further categorize participants based on whether their health components fell in the poor (0 points), intermediate (1 point), and ideal (2 points). For analysis purposes, we used a continuous variable based on number of CVH index components and categorized as 0–7 points for low, 8–9 points for intermediate, and ≥10 points as ideal CVH score. Dichotomous categorization of CVH score (<10 and ≥10) was also used in longitudinal analyses for greater sample size.

Covariates

Consistent with prior analyses in a study assessing the relationship between RF and adult outcomes, we considered the following covariates: age, sex, and race.¹⁷ Given CARDIA methodology is intentionally inclusive of individuals identifying as both White and Black, we also tested for effect modification (with a test for interaction) between RF and race. While a prior study assessing the association between RF and outcomes in adulthood included additional models with covariates related to lifestyle including alcohol consumption, CVH in the current study included these measures as the current outcome of interest, except for alcohol use, which is not a component of CVH score. Given alcohol consumption is associated with CVH, participant alcohol use was included as an additional covariate in the current analyses.¹⁸ Socioeconomic status (SES) has been associated with CVH.¹⁹ Given the effect of adult income on the association observed between RF and CVH in a prior study, adult income was included as a covariate in the current analyses,¹² which in CARDIA, was measured as total combined family income over the last 12 months.

Statistical Analysis

A descriptive analysis of the cohort inclusive of a comparison between low and high RF score (dichotomized at the median). For continuous variables (non-normally distributed) with median and interquartile range (IQR), Wilcoxon rank sum test were used whereas for continuous variables (normally distributed) with mean and standard deviation, two-sample t-tests were used to compare RF score groups. For categorical variables, Chi-square tests were used to assess association. For our primary aim, linear mixed models (for continuous CVH score) or generalized mixed models (for categorical CVH score), with random intercepts for each participant, were used to examine the effect of RF (independent variable) on CVH (dependent variable) over time (baseline through 20 year follow-up). We repeated the same analyses using the child abuse and caregiver warmth sub-scales as independent variables.

A separate model and graphical depiction were completed to demonstrate the interaction between child abuse and caregiver warmth. This model included child abuse, caregiver warmth, and their interaction, with the outcome CVH, and covariates. To aid in interpretation, this interaction analysis was graphed where three lines represented the slope of childhood abuse (x-axis) predicting CVH (y-axis) such that one line was plotted for caregiver warmth (mean, 3.2), a second for low caregiver warmth below one standard deviation of the mean (−2.4), and a third for high caregiver warmth or (one standard deviation above mean (2.4). Predictions were obtained using mean values of all other predictors in the model.

For our second aim, to assess the relationship between RF and CVH in the context of adult SES (adult income), a secondary analysis stratifying by adult income was completed, with CVH score as a continuous outcome variable longitudinally (across all years). All analyses were performed using SAS 9.4 (Cary, NC, USA). A p-value of < 0.05 was considered statistically significant for main results.

Sensitivity Analyses

While the main analysis applied longitudinal statistical modeling including CVH score at all years of follow-up, we conducted cross-sectional analyses between our exposures and outcome at each time point (year 0, 7, and 20). Regarding CVH scoring, calculation has not been consistent in previous studies given the use of varying cut off values for ideal CVH score (e.g., ≥10, ≥12).³ We addressed this limitation by conducting a sensitivity analyses for cuts offs of CVH categories as follows: 1) as done in the main analyses of this study, categorized as 0–7 points for low, 8–9 points for intermediate, and ≥10 points as ideal CVH, and 2) CVH score low as 0–8, intermediate as 9–11, and ideal as 12–14. Analyses were conducted using linear regression to assess the association between RF environment score and CVH score as a continuous dependent variable. After assessing the proportional assumption, ordinal logistic regression analyses were used to assess the association between RF and ideal CVH index as a 3-level categorical variable (low, intermediate, and high).

Results

The cohort (n=2,074) had an average age of 25.25 (SD 3.51) at baseline (Table 1). The sample was 55% female and 61% White. At baseline, 55% of the cohort met ≥5 categories for ideal CVH, with most (97%) meeting the ideal metric for fasting glucose (<100 mg/dL) and fewest (2%) meeting the ideal metric for diet (4–5 dietary components). By year 7, 45% of participants attained ≥5 categories for ideal CVH, with a further decline to 24% by year 20 (Table S2). Less than 1% of participants were in the ideal range for all 7 categories of ideal CVH (≥12) at all time points.

Table 1.

Baseline and Cardiovascular Health (CVH) characteristics of the study sample by low and high Risky Family (RF) childhood scores: The Coronary Artery Risk Development in Young Adults (CARDIA) Study

Variable‡		Risk Family Score		p-value	Total (n=2074) N (%)
		≤10 (n=1060) N(%)†	>10 (n=1014) N (%)
Age		25.1 (3.6)	25.4 (3.5)	0.06	25.24 (3.5)
Sex	Female	575 (54.3)	568 (56.0)	0.42	1143 (55.1)
	Male	485 (45.8)	446 (44.0)		931 (44.9)
Race	White	664 (62.6)	602 (59.4)	0.13	1266 (61.0)
	Black	396 (37.4)	412 (40.6)		808 (39.0)
Income (Year 7)	<$35k	431 (40.7)	468 (46.2)	0.001*	889 (43.4)
	≥$35k	629 (59.3)	546 (53.8)		1,175 (56.7)
Childhood socioeconomic status (mean parental education)		13 [12, 16]	12 [11.5, 15]	<0.001*	13 [12, 15]
Alcohol use	Yes	940 (88.7)	885 (87.3)	0.33	1825 (88.0)
Risky Family Score (Year 15)		8 [7, 9]	14 [12, 16]	<0.001*	10 [8, 14]
Child Abuse Score		1 [1, 1]	1.5 [1, 2]	<0.001*	1 [1, 2]
Caregiver Warmth Score		4 [3.5, 4]	2.5 [2, 3.5]	<0.001*	3.5 [2.5, 4]
Smoking (Year 0)	Current	190 (17.9)	268 (26.4)	<0.001*	458 (22.1)
	Former ≤ 12 months ago	49 (4.6)	67 (6.6)		116 (5.6)
	Never or quit >12 months ago	821 (77.5)	679 (66.7)		1500 (72.3)
Body Mass Index (kg/m2)		23.98 (4.3)	24.59 (5.0)	0.003*	24.27 (4.6)
Blood pressure (mmHg)	SBP >= 140 OR DBP >= 90	18 (1.7)	17 (1.68)	0.55	35 (1.7)
	SBP 120–140 or DBP 80–90 or treated	236 (22.3)	206 (20.3)		442 (21.3)
	SBP <120 or DBP <80	806 (76.0)	791 (78.0)		1597 (77.0)
Total cholesterol (mg/dL)	>= 240 mg/dl	46 (4.3)	39 (3.9)	0.054	85 (4.1)
	200–240 mg/dl or treated	179 (16.9)	213 (21.0)		392 (18.9)
	<200 mg/dl	835 (78.8)	762 (75.2)		1597 (77.0)
Diet	0–1 components	589 (55.6)	567 (55.9)	0.28	1156 (55.7)
	2–3 components	449 (42.4)	435 (42.9)		884 (42.6)
	4–5 components	22 (2.1)	12 (1.2)		34 (1.6)
Fasting plasma glucose (mg/dL, Year 0)	>= 126 mg/dl	6 (0.6)	6 (0.6)	0.54	12 (0.6)
	100–126 mg/dl or treated	25 (2.4)	17 (1.7)		42 (2.0)
	<100 mg/dl	1029 (97.1)	991 (97.7)		2020 (97.4)
Physical activity (Year 0)	None	97 (9.2)	101 (10.0)	0.68	198 (9.5)
	1–149 moderate, 1–74 vigorous	311 (29.3)	307 (30.3)		618 (29.8)
	≥150 moderate or ≥75 vigorous	652 (61.5)	606 (59.8)		1258 (60.7)
CVH Score at Year 0		11 [10, 12]	10 [9, 12]	<0.001*	11 [9, 12]
CVH Score at Year 7		10 [9, 12]	10 [9, 11]	<0.001*	10 [9, 12]
CVH Score at Year 20		9 [8, 11]	9 [7, 11]	0.002*	9 [8, 11]

^*p<0.05

^†Continuous variables that were normally distributed, are described using mean and standard deviation, p-values from two-sample t-tests; Continuous variables that were not normally-distributed were described using median and interquartile range (IQR), p-values from Wilcoxon rank sum tests; Categorical variables are presented as N(%), p-values from Chi-square tests

^‡All variables collected at baseline (Year 0) unless otherwise indicated; variables at baseline are representative of similar results seen at years 7 and 20 observing only differences in BMI and smoking between high and low RF groups

The median RF score was 10 (IQR: 8, 14) (Table 1). Adults at or below the median RF score (low RF) were similar to those above the median RF score (high RF) in demographic characteristics (age, sex, and race), but differed in income during adulthood (reported at year 7) and education in adulthood (reported at baseline; Table S2). Specifically, while for the entire cohort the majority had an annual income of $35–45k, those with high RF were more likely to have an adult annual income less than $25k (30% v. 24%) and fewer years of education (median 14 vs. 15). At each year, a smaller percentage of individuals in the high RF group met ≥5 categories for ideal CVH than those with low RF (Year 0: 520/1015 (51%) vs. 628/1062 (59%), Year 7: 430/1015 (42%) vs. 504/1062 (47%), and Year 20: 230/1015 (23%) vs. 273/1062 (26%)) (Figure 2, Table S1). At each year, a smaller percentage of individuals with low income (<$25k) has ideal CVH score (≥10, Table S3).

Figure 2.

Greater Risk Family (RF) Environment in childhood is more commonly seen in those with low ideal CVH score at each follow-up year in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

These graphs depict an observed pattern that Risky Family (RF) score above the median (>10) was more common in those with Cardiovascular Health (CVH) score at or below the peak (most common CVH across the population) at baseline (Panel A), year 7 (Panel B), and year 20 (Panel C) of CARDIA follow-up. The x-axis indicates the percent of total sample and the y-axis indicates number of ideal CVH metrics categories met (CVH index, range, 0–7).

CARDIA participants with high RF were significantly more likely to have obesity by BMI (115 (11.3%) vs. 82 (8%), p=0.014) and be a current or former smoker (335 [33%] vs. 239 [23%], p<0.001) at baseline, but no baseline differences were observed in any other specific category of the CVH health score when comparing low and high RF categories at baseline. CVH score was lower at all years for those with high RF compared to those with low RF (10.51 (1.75) vs. 10.35 (1.76), p=0.0001 at baseline; 10.06 (2.01) vs. 9.9 (2.05), p=0.0003 at year 7; and 9.14 (2.30) vs. 8.99 (2.31), p=0.003 at year 20.

Longitudinal Analyses

In longitudinal analyses over all 20 years of follow-up, after adjusting for age, sex, race, and alcohol use, higher RF was significantly associated with lower CVH score (ß=−0.030, p<0.001, Table 2, model 2, linear mixed model). In a generalized mixed model, for every unit higher RF score, there was 3.5% lower odds of having ideal CVH (ideal CVH as ≥10, OR=0.97 (0.965), 95%CI 0.94–0.98). In these longitudinal analyses, there was no interaction with year of examination. In other words, the observed associations were consistent over time.

Table 2.

Longitudinal Analyses of the Association Between Risky Family (RF) Environment in Childhood and Cardiovascular Health (CVH) Over 20 years of Follow-up in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

Dependent variable	Beta coefficient Model 1*	95% CI	p-value	Beta coefficient Model 2†	95% CI	p-value
CVH score (continuous)‖	−0.04	(−0.06, −0.02)	<.001	−0.03#	(−0.05, −0.01)	<.001
	Odds ratio Model 1 *	95% CI	p-value	Odds ratio Model 2 †	95% CI	p-value
iCVH (categorical)‡	0.96	(0.94, 0.98)	<.001	0.97**	(0.95, 0.99)	<.001

^*Model 1: Unadjusted;

^†Model 2: Adjusted for age, sex, race, alcohol use, and income; For categorical CVH (iCVH) longitudinal analyses, generalized mixed models with random intercepts were used

^‡iCVH (binary): ideal CVH score ≥10 vs. ideal CVH score <10

^‖Continuous CVH longitudinal analyses used linear mixed models with random intercepts to account for subject-to-subject variations, which allowed CVH score to be higher or lower for each subject

^#Example interpretation: Longitudinally, for every unit increase in RF score there was a 0.03 unit decrease in CVH over time (Beta=−0.030, 95%CI (−0.047, −0.012), p<0.001), after adjusting for age, sex, race, alcohol use, and income.

^**Example interpretation: Longitudinally, for every unit increase in RF score the odds of ideal CVH (CVH score ≥10) decreased by 3.5 % (OR=0.965, 95%CI (0.945, 0.985), p<0.001), after adjusting for age, sex, race, alcohol use, and income.

Relational Health: Child Abuse and Caregiver Warmth

Longitudinal analyses by sub-scales (Table 3) demonstrated that when adjusting for age, sex, race, and alcohol use, child abuse was associated with lower CVH score (ß=−0.119, p=0.03, linear mixed model), with each unit greater child abuse score corresponding to 12.8% lower odds of ideal CVH (CVH ≥10, OR=0.87 (0.872), 95%CI 0.77, 0.99, generalized mixed model) across all 20 years of follow-up. Greater caregiver warmth was associated with greater CVH score (ß=0.085, p=0.049, linear mixed model), with each unit greater caregiver warmth score corresponding to 11.7% higher odds of ideal CVH (OR=1.12 (1.1165), 95%CI 1.01–1.24, generalized mixed model) across all 20 years of follow-up.

Table 3.

Longitudinal Analyses of the Association of Child Abuse and Caregiver Warmth with Cardiovascular Health (CVH) Over 20 years of Follow-up in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

Dependent variable	Beta coefficient Model 1*	95% CI	p-value	Beta coefficient Model 2†	95% CI	p-value
Exposure (independent variable): Child Abuse (continuous)
CVH score‖	−0.17	(−0.28, −0.05)	0.005	−0.12	(−0.23, −0.01)	0.030
	Odds ratio Model 1*	95% CI	p-value	Odds ratio Model 2†	95% CI	p-value
iCVH2	0.83	(0.73, 0.95)	0.007	0.87	(0.77, 0.99)	0.038
Exposure (independent variable): Caregiver Warmth (continuous)
	Beta coefficient Model 1*	95% CI	p-value	Beta coefficient 2†	95% CI	p-value
CVH score1	0.10	(0.01, 0.19)	0.028	0.09#	(0.00, 0.17)	0.049
	Odds ratio Model 1*	95% CI	p-value	Odds ratio Model 2†	95% CI	p-value
iCVH‡	1.13	(1.02, 1.26)	0.018	1.12**	(1.01, 1.24)	0.034

^*Model 1: Unadjusted;

^†Model 2: Adjusted for age, sex, race, alcohol use, and income.

^‡Ideal Cardiovascular Health (iCVH): binary as ideal CVH score ≥10 vs. ideal CVH <10; for categorical CVH (iCVH) longitudinal analyses, generalized mixed models with random intercepts were used

^‖Continuous Cardiovascular Health (CVH) longitudinal analyses used linear mixed models with random intercepts to account for subject-to-subject variations, which allowed CVH score to be higher or lower for each subject

^#Example interpretation: Longitudinally, for every unit increase in degree of caregiver warmth there was a 0.085 unit increase in CVH over time (Beta=0.085, 95%CI (0, 0.169), p=0.049), after adjusting for age, sex, race, alcohol use, and income.

^**Example interpretation: Longitudinally, for every unit increase in caregiver warmth the odds of ideal CVH (CVH score ≥10) increased by 11.7 % (OR=1.117, 95%CI (1.009, 1.236)), after adjusting for age, sex, race, alcohol use, and income.

There was a significant interaction between child abuse and caregiver warmth (p=0.020 in an unadjusted mode, p=0.028 in a model adjusted for age, sex, and alcohol use), such that greatest caregiver warmth and greatest child abuse exposure was associated with the lowest CVH score, while greatest caregiver warmth and lowest child abuse exposure was associated with highest CVH score (Figure 3).

Figure 3.

Lowest Child Abuse Predicts Greatest Cardiovascular Health (CVH) in the Setting of High Caregiver Warmth in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

The interaction between child abuse and caregiver warmth was significant (p=0.028). Predictions were obtained using mean values of all other predictors in the model. This was graphed using three points of caregiver warmth (mean=3.2, one standard deviation below the mean (low)=2.4; and one standard deviation above the mean (high)=4). For each value of caregiver warmth, a line with the slope of childhood abuse predicting cardiovascular health score (CVH) is shown. For those with exposure to the greatest caregiver warmth (+1SD) and lowest child abuse (0 on the x-axis), CVH was highest. As child abuse exposure increased to highest (4 on the x-axis) for those with also the greatest exposure to caregiver warmth (+1SD), CVH was lowest. Predictions were obtained using mean values of all other predictors (including age, sex, and alcohol use) in the model.

Analyses Stratified by Income

Stratified analyses (Table 4) showed the associations between RF with longitudinal CVH score (continuous variable) differed across income levels when adjusting for age, sex, race, and alcohol use (interaction trend toward significance, p=0.066, Table 4, Model 2). Greater RF was associated with lower CVH across all years for income strata $35–74k and ≥$75k (ß=−0.04, p=0.005 at $35–74k; ß=−0.07, p=0.002 at ≥$75k), but there were no significant associations between RF and CVH among participants with annual income less than $25k or $25-$34k. Though there was no significant interaction between child abuse and income (p-interaction=0.280), there was between caregiver warmth and income (p-interaction=0.008, Table 4, Model 2). Greater warmth was associated with higher CVH only for those with an income of $75k or greater (ß=0.38, p<0.001). No significant interaction was observed between RF and race and therefore analyses were not stratified by race, as was the approach was used prior for assessing the relationship between RF and adult outcomes in the CARDIA study.¹⁷

Table 4.

Effect Modification of Income in Adulthood on Longitudinal Association Between Risky Family (RF) Environment in Childhood and Cardiovascular Health (CVH) Score¹ Over 20 years of Follow-up in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

Level‡	Model 1* Interaction p-value	Beta coefficient Model 1*	95% CI	p-value	Model 2† Interaction p-value	Beta coefficient Model 2†	95% CI	p-value
Risky Family (RF) Environment ‖
1:less than $25k	0.035	0.005	(−0.03, 0.04)	0.77	0.066	0.001	(−0.03, 0.03)	0.95
2:$25k-$34k		−0.022	(−0.07, 0.03)	0.38		−0.026	(−0.07, 0.02)	0.27
3:$35k-$74k		−0.035	(−0.06, −0.01)	0.015		−0.038	(−0.07, −0.01)	0.005
4:$75k and above		−0.079	(−0.13, −0.03)	<.001		−0.070	(−0.11, −0.03)	0.002
Caregiver Warmth
1:less than $25k	0.007	−0.053	(−0.21, 0.11)	0.52	0.008	−0.039	(−0.19, 0.11)	0.61
2:$25k-$34k		−0.062	(−0.28, 0.15)	0.57		−0.059	(−0.26, 0.14)	0.57
3:$35k-$74k		0.096	(−0.05, 0.24)	0.19		0.133	(−0.00, 0.27)	0.057
4:$75k and above		0.410	(0.18, 0.64)	<.001		0.375	(0.16, 0.59)	<0.001

^*Model 1: Unadjusted model;

^†Model 2: Adjusted for age, sex, race, and alcohol use

^‡For analyses in this table, all outcomes are continuous CVH score where linear mixed models including an interaction term (risky family*income) were used

^‖Example interpretation: Longitudinally, the associations between Risky Family (RF) with Cardiovascular Health (CVH) were different across different income strata (interaction p-value=0.066, when adjusting for age, sex, alcohol use, and race) such that, for example, among participants with annual income $75k and above, for every one unit increase in RF score, there was a 0.07 (Beta=−0.070, 95% CI (−0.114, −0.025), p=0.002) unit decrease in CVH across time, but there were no significant associations between RF and CVH among participants with annual income less than $34k.

Sensitivity Analyses

Cross-sectional analyses by year demonstrated results that were consistent with the longitudinal analysis inclusive of all years of follow-up. Results also were consistent for a CVH score cut-off of ≥10 (Table S4) and ≥12 (Table S5).

Discussion

This study is the first to identify a link between adverse childhood family environments and CVH score across multiple time points in the lifecourse, with longitudinal follow-up over decades using data from the CARDIA cohort. We observed that greater RF was associated with lower odds of attaining a greater CVH category (for every unit increase in RF) over the entire 20 years of follow-up (Figure 4). Measures of relational health in childhood, by sub-scales of the RF questionnaire, were specifically associated with CVH. Though greater exposure to child abuse decreased the odds of attaining a higher category of CVH, greater exposure to caregiver warmth increased the odds of attaining ideal CVH in adulthood across 20 years of follow-up. Our findings also highlight the interaction between types of CVH risk exposures across the lifecourse, as we observed variation in impacts on CVH when the level of caregiver warmth and child abuse exposures were considered together and when accounting for income in adulthood. Relational health experienced through the stability of caregiver relationships in childhood (e.g., the opposite of the oscillation between caregiver warmth and abuse) may be associated with higher cardiovascular health in adulthood. In analyses stratified by income, associations between RF and CVH were significant only for CARDIA participants with highest income levels.

Figure 4.

Childhood Family Environment is Associated Odds of Cardiovascular Health (CVH) Over Time in the Context of Socioeconomic Status in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

This figure depicts a lifecourse timeline from childhood (<18 years of age) through early adulthood (average age of 25 years) to mid-adulthood (20 years later) as a trajectory toward the attainment of cardiovascular health (CVH), inclusive of smoking status, body mass index, physical activity, total cholesterol, blood pressure, fasting glucose, and dietary intake.² Reported experiences of adversity in a childhood family environment (RF) were associated with 4% lower odds of attaining ideal CVH across the lifecourse; an association that was statistically significant in those with annual income above $35k only. Child abuse, specifically, was associated with near 17% lower odds of ideal CVH across the lifecourse that was independent of an association with annual income. However, caregiver warmth was associated with near 12% greater odds of ideal CVH across the lifecourse for those with the greatest annual income (>$75k). All of these associations remained significant in the context of childhood socioeconomic status, but were each impacted differently by the context of annual income in adulthood as reported. These findings demonstrate an association between early life RF and CVH, but also suggest that future studies assessing the impact of early life exposures and/or interventions on later life health outcomes should consider socioeconomic contexts across the lifecourse.

Our study adds to the literature in demonstrating that childhood adversity, an outcome that has been associated with higher lifecourse morbidity and mortality prior,¹⁹ is also associated with lower odds of attaining cardiovascular health decades later in life. The literature across the last decade has reproduced findings of an association between childhood adversity and disease, specifically CVD. For example, prior studies show that adverse childhood environments are correlated with subclinical atherosclerosis, incident type 2 diabetes, hypertension, hyperlipidemia, and cardiovascular disease and mortality in CARDIA and other cohorts.^20–23 Our study is among the first to modify the conceptual models of these collective studies to instead consider an outcome variable beyond metrics of disease or individual risk factors, and instead of a modifiable risk factor composite measure. This study further considers the protective exposure of caregiver warmth, beyond solely a risk exposure (adversity). Therefore, the current conceptual model exemplifies the importance of considering the most upstream protective factors (i.e., caregiver warmth or relational health) towards the promotion behavioral and clinical factors that can mitigate disease risk (i.e., higher CVH scores).

While there have been many studies that have associated adverse childhood experiences (ACEs) with poor health outcomes including cardiovascular risk and disease,^23–26 the current study calls attention toward health trajectories and the potential to identify risks that can be mitigated by primordial prevention.⁸ Only one prior study, by Islam and colleagues, has explored the relationship between experiences of adversity in childhood and later life CVH.¹² The findings of our current study and that of Islam et al., both identified that greater exposure to adversity in childhood was associated with lower levels of CVH in adulthood. However, the study by Islam et al. was a cross-sectional analysis in a cohort of Black individuals exclusively from Atlanta, Georgia, that used the Early Trauma Inventory (ETI) to survey exposure to a broad range of dichotomous exposures in childhood that were not specific to the family or household environment (e.g., included death or illness of a friend, witnessing someone murdered, or being exposed to a natural disaster).¹² This inventory did not include experiences of affection or support. The current study builds on the findings of Islam and colleagues in using data from a bi-racial, longitudinal cohort with a survey specific to early family environments inclusive of caregiver warmth (i.e., relational health). Further, the duration of exposure to a stressor can dictate its impact on associated health outcomes,²⁷ as may especially be the case for exposures to adversity in early life.¹¹ The RF Questionnaire and its incremental scoring offers the opportunity to include frequency of exposure by Likert scale. Therefore, our study is able to demonstrate that experiences in childhood, specific to the family environment and inclusive of metrics of relational health in a cohort of both black and white individuals, is associated with CVH longitudinally.

The topic of relational health has been the focus of calls to action for research aimed at understanding and ultimately mitigating adversity in childhood.¹³ While the mounting evidence of the impacts of child maltreatment on CVD risk aligns with the current expected finding of an association between RF and CVH,^24,28 only one prior cross-sectional study,¹⁴ and the current longitudinal study assessed the relationship between positive relational health (herein, caregiver warmth in childhood), and cardiovascular health in adulthood. The prior cross-sectional study demonstrated that those with greater positive childhood experiences, inclusive of parental warmth, had higher CVH score in adulthood (n=1255 diverse cohort of age 34–84).¹⁴ In the current study, it is demonstrated that caregiver warmth may have a role in increasing the potential to attain improved CVH by over 10% across a 20-year longitudinal period, for each unit of exposure to affection or support in childhood. Future studies should consider harnessing healthy caregiver relationships with children as a means of attaining intergenerational CVH. However, such studies should include assessment of both positive (e.g., caregiver warmth) and negative (e.g., child abuse) components of relational health. For example, our analysis suggests there may be an interaction between exposure to child abuse and caregiver warmth when taken together. Future work assessing early life risk factors, should require analyses that consider both experiences. CARDIA participants with low caregiver warmth, and also low child abuse exposures, had lower CVH score than those with high caregiver warmth and low child abuse exposures. However, for those with low caregiver warmth, there was little difference in CVH for those with simultaneous low or high child abuse exposure. Taken together these results suggest that both constant deprivation of caregiver warmth (i.e., low caregiver warmth) and potential fluctuations between experiences of warmth and abuse (i.e., experiencing both caregiver abuse and warmth in childhood), may have profound long-term influence on the decreased likelihood of higher CVH attainment in adulthood. This work supports the prior described framework of relational health that identifies how healthy relationships in childhood are defined not only as nurturing and affectionate (e.g., warm), but stable (e.g., without simultaneous childhood experiences of abuse), and that such warm and stable relationships contribute significantly to optimal health outcomes throughout the lifecourse.¹³ Literature has also demonstrated that an authoritative caregiving or parenting style, one that consistently, “balances high levels of responsiveness with high levels of demandingness,”²⁹ is associated with low early life cardiometabolic risk (e.g., non-obese status, higher fruit and vegetable consumption).^29,30 Therefore, future studies may consider the impact of caregiving styles and contexts on lifecourse cardiovascular health for future generations.

The study of both individual and systemic impacts on the relationship between childhood experiences and CVH will enable future work to consider multi-tiered interventions. We demonstrate that one such contextualizing factor may be socioeconomic inequity across the lifecourse. It was herein found that the relationship between RF and poor lifecourse CVH was seen for those with higher income in adulthood ($35-$74k and ≥$75k), but not lower (<$35k annual income). It is possible that the effect of childhood RF on later life CVH is statistically observed only in higher income strata due to lower levels of exposure to other forms of structural adversity unmeasured in the current study (e.g., limited resources, decreased access to healthcare). This may be contrary to lower income strata, which may have an unobserved impact on CVH across the lifecourse.^31,32 Importantly, prior findings by Islam and colleagues in a cohort of Black participants demonstrated that exposure to a greater number of traumatic experiences in childhood was associated with poorer CVH only for those in lower income strata (<$50k).¹² Other sociodemographic characteristic differences in the cohorts may have accounted for the different results found including that the current study compared to the prior had larger sample size (2074 vs. 457), and broader geographic (multi-site cohort vs. single urban location) and income ($35–74k vs. <$25k) composition. Most notably the studies differed in measured exposures. The prior study, surveyed a broad range of traumatic experiences using the Early Trauma Inventory (ETI), which may have captured exposures that are particularly susceptible to systemic discrimination. For example, racial and ethnic disparities exist for morbidity and mortality, as younger life expectancy with homicide as a leading cause of death is seen in Black populations (and both relate to the ETI items of experiencing death or illness of a friend or family member, or witnessing someone murdered).^33,34 Further, evacuation challenges during natural disasters have disproportionately impacted Black populations, as observed during Hurricanes Katrina and Harvey, for example, (which may relate to the ETI item being exposed to a natural disaster).³⁵ The experiences of adverse family environments, specifically, such as violence or substance use in the home were surveyed in both studies. Such household family experiences were specified in the current study without compounding in the same questionnaire the measurement of other general trauma exposures. General trauma exposures (such as natural disasters or loss of a loved one) may be seen across all income strata and, while still impacted by disparities, may be more universally traumatic and less obscured by other adverse experiences disproportionately experienced by marginalized populations (e.g., exposure to poverty and racism).^16,36 Importantly, both the study by Islam and colleagues and the current observed an association between child abuse and lower CVH, though the prior found those of low income had a relationship between emotional and sexual, but not physical, abuse experiences and lower CVH, while the current study measured emotional and physical abuse together as an exposure which was found to be associated with lower CVH but only prior to stratifying by income status.¹² The findings taken together inform a hypothesis to be explored in future work. Building equitable systems and structures for economic opportunity as well as resources for traumatic experience responsiveness surrounding families and therein early childhood environments,³⁷ could improve trajectories of CVH. Future work should explore the relationship between childhood exposures, including better phenotyping traumatic and abusive experiences, and adult health trajectories.

While the current study cannot support causal mechanisms, the findings shed light on the need for future work to elucidate mechanisms underpinning the relationship between childhood exposures and poor CVH will provide direction for targeted interventions to enhance lifecourse CVH. Trauma and adversity as sources of early life stress (ELS) in childhood are associated with some potential pathways that may converge to portend risk of poor CVH including, 1) unhealthy lifestyle behaviors (e.g., smoking, diet high in processed foods), 2) dysregulation of stress physiology, and 3) associated environmental exposure risk (e.g., pollutants and endocrine disrupting chemicals). Evidence to the first comes from studies demonstrating a relationship linking both trauma and RF with smoking and substance use including finding a link between RF and smoking later in life,¹² and a link between ELS and both substance use and unhealthy diet in adulthood.^38–40 It has been hypothesized that the association may be due to the use of substances as a coping mechanism to deal with exposure to adversity.³⁹ It is also possible that household challenges (e.g., disorganization) and family environment (e.g., poor physical and mental health of caregivers) may be more likely to impact habits experienced and learned by children.⁴⁰ To the second pathway, ELS plays a key role in dysregulating the hypothalamic-pituitary-adrenal axis and cortisol curve, as well as other systems (cardiac, inflammatory, metabolic), which accumulate over time (allostatic load) to impact cardiometabolic risk into adulthood.^11,17,41,42 One of the primary mechanisms for these pathways to converge may be through adiposity. In the current study, greater RF was associated with obesity and smoking at baseline, but no other CVH categories differed by low and high RF. Prior literature also has demonstrated that ELS is associated with obesity, and many other studies have linked ELS with adiposity in children and adults including through cortisol dysregulation.^12,43–45 Additionally, chemical exposures from smoking, pollutants and processed foods, which disrupt neuroendocrine and metabolic physiology, often co-occur with non-chemical environmental exposures to stress like poverty and structural racism throughout the lifecourse.⁴⁶ Future research should aim to elucidate contributing mechanistic pathways and the way they inter-relate as potential mediators of the link between childhood exposures and adult CVH.

This study has limitations. Though CARDIA has the strength of a longitudinal cohort, it is an observational study limiting the ability to assess causality. Further, childhood exposures were assessed at year 15, a year that did not assess CVH, by retrospective analysis which is subject to recall bias and were not comprehensive to include all possible exposures in the early childhood environment. Though a strength of the RF questionnaire is in assessing frequency of occurrence and exposure to caregiver warmth, it is still limited in specifications of duration of exposure or comprehensive assessment of resilience or coping strategies. As studied by Islam and colleagues, income was observed to be the SES variable that was an effect modifier of the association between RF and CVH (i.e., not education) and so adult income was included in the current study as a covariate.¹² However, future studies should assess the impact of other variables considered to contribute to SES (i.e., education) on the relationship between childhood RF and adult CVH. Importantly, the study analysis was completed prior to the AHA updating the construct of CVH to include sleep,⁴ so future studies should aim to include 8 components of CVH. However, it may be hypothesized that given childhood adversity is associated with poor sleep in adulthood,⁴⁷ the exclusion of sleep in our metric for CVH may actually have resulted in an underestimation of the true association between RF and CVH across the lifecourse. Still, our findings highlight the importance of studying childhood exposures in addressing later life health with a need for prospective cohort design and randomized controlled trials assessing early life and intergenerational interventions.

Conclusion

Our findings demonstrated that while adverse childhood family environments may increase odds of poor CVH in adulthood, positive experiences of caregiver warmth are associated with higher odds of ideal CVH across decades of follow-up. However, caregiver warmth may not moderate the impact that high levels of child abuse exposure have on CVH. Socioeconomic status may contextualize exposure type and level of risk at the family level. Toward a paradigm of primordial prevention of CVD, healthy and stable childhood relationships and life-long economic equity may play a crucial role in the equitable opportunity for all to attain CVH across the lifecourse as future work should further explore.

What is Known

• Cardiovascular health (CVH), a composite measure of protective behavioral and clinical factors is associated with decreased cardiovascular and chronic disease morbidity and mortality.

• Childhood experiences such as childhood adversity and caregiver warmth have been associated, respectively, with lower and higher CVH in adulthood through cross-sectional studies at single timepoints.

What the Study Adds

• The current study finds childhood adversity and caregiver warmth are, respectively, associated with lower and higher CVH over multiple time points across a 20-year longitudinal period into adulthood.

• Stability of caregiver relationships (i.e., relational health) experienced in childhood may be associated with higher cardiovascular health in adulthood.

• Adult income later in life (adulthood) may modify the relationship between childhood adversity and CVH across the lifecourse into adulthood.

Abbreviations:

ACEs

Adverse Childhood Experiences

AHA

American Heart Association

BP

Blood Pressure

BMI

Body Mass Index

CARDIA

Coronary Artery Risk Development in Young Adults study

CVD

Cardiovascular Disease

CVH

Cardiovascular Health

DBP

Diastolic Blood Pressure

ELS

Early Life Stress

iCVH

Ideal Cardiovascular Health

IQR

Interquartile Range

RF

Risky Family

SBP

Systolic Blood Pressure

SD

Standard Deviation