Bringing doctors together

or

A free, private network for verified physicians

Leading voices

Dr. Panagis Galiatsatos

Critical Care · Johns Hopkins

View

Dr. Kevin Chung

Critical Care · SeaStar Medical

View

Dr. Amos E. Dodi

Critical Care · MMC

View

Dr. Robert Hyzy

Critical Care · UMHS

View

Dr. Vin Gupta

Critical Care · Manatt Health

View

Dr. John Kellum

Critical Care · University of Pittsburgh

View

Dr. Christopher Worsham

Critical Care · Harvard Medical School

View

Dr. Kristin Walter

Critical Care · JAMA

View

Dr. Christopher Seymour

Critical Care · University of Pittsburgh

View

Dr. Hassan Bencheqroun, MD

Critical Care · TeleICU

View
  • Discover your community

  • Stay current

  • A profile that represents you

  • Find your people

About RoonPrivacyTerms

    Overview

    Associations Between Social Determinants of Health, Allostatic Load, and Anthracycline Cardiotoxicity in a Diverse Patient Population.

    Mar 27, 2025DOI: 10.1161/jaha.124.036649Journal ArticleAHA

    Written by: Lili ZhangJustin SongWaqas HanifRachel ClarkMagued HarounMrunalini DandamudiPatricia Guia SimozaLeandro SlipczukMario J GarciaMin PuCarlos A GongoraTomas G NeilanDella F MakowerEarle C ChambersCarlos J Rodriguez

    METHODS

    The data that support the findings of this study are available from the corresponding author upon reasonable request. In this study, we included consecutive adults who received anthracycline‐based chemotherapy between January 2016 and December 2019 for any type of cancer at a single academic health system that serves a large multiracial‐ethnic community in the Bronx, New York. Individuals with preexisting HF, defined by the International Classification of Diseases, Ninth Revision (ICD‐9; 428.9) or Tenth Revision (ICD‐10) code (I50.9), were excluded. Only individuals who underwent a left ventricular ejection fraction (LVEF) assessment at baseline were included in the analysis. Patients lacking AL measures were excluded from the analysis. Of 1115 individuals treated with anthracyclines,116 were excluded due to a prior history of HF, 74 were excluded due to inadequate baseline LVEF assessment, and 5 were excluded due to missing AL measures (Figure 1). We categorized self‐reported race and ethnicity into 5 nonoverlapping groups: NH White, NH Black, Hispanic, Asian, and other/unknown/declined. We first identified self‐reported Hispanics, and then only among NH individuals, we specified self‐reported race. Being Hispanic is an ethnicity not a race; but self‐reported race among Hispanics is less reliable.30 Most Hispanic individuals (88%) in our cohort self‐reported race as “Unknown or Other” race. In the final analysis, we included NH White, NH Black, and Hispanic patients because the sample sizes of other racial and ethnic groups (ie, Asian) were small. Therefore, the final analysis included 718 individuals, including 206 NH White, 226 NH Black, and 286 Hispanic individuals (Figure 1).LVEF indicates left ventricular ejection fraction.The index dates of the cohort were the dates when the subjects received the first dose of anthracycline therapy. The end of follow‐up was March 1, 2023. This study was approved by the institutional review board of Albert Einstein College of Medicine and a waiver of informed consent was granted.A composite AL score was calculated as described previously,25, 31, 32 using 9 biomarkers reflective of 4 physiologic systems collected before the administration of anthracyclines. The physiologic systems were as follows: (1) metabolic, including body mass index (calculated as weight in kilograms divided by height in meters squared) and creatinine, albumin, and glucose levels; (2) immune, consisting of white blood cell count; (3) cardiac, including systolic blood pressure, diastolic blood pressure, and heart rate; and (4) hematologic, consisting of serum hemoglobin level. Cutoffs were established using those of prior studies evaluating AL and established laboratory thresholds for abnormal values. Variable cutoffs were described in Table 1. A point was assigned if the biomarker value was above normal cutoffs for all biomarkers except albumin and hemoglobin levels for which a point was assigned if the values were below the normal threshold. For body mass index, a point was assigned if the value was greater than 25.0 kg/m2 or less than 18.0 kg/m2. All the assigned points were summed, with a possible range from 0 to 9.
    Composition of Allostatic Load Score
    The cumulative dose of anthracyclines was extracted from pharmacy database and verified by nursing notes during infusion. The dose of anthracycline agents other than doxorubicin was converted to doxorubicin isoequivalent dose based on established guidelines (Table S1).33 The conversion of liposomal doxorubicin to doxorubicin is based on toxicity dose ratio of 0.33.34 Demographics, tumor characteristics at the index date, and other cancer treatments were collected by chart review.35 Preexisting atrial fibrillation, coronary artery disease, chronic kidney disease, cerebrovascular accident, myocardial infarction, and cardiovascular risk factors (hypertension, diabetes, hyperlipidemia, obesity, current or former smoking status) were identified by ICD codes.35 Obesity was defined by ICD codes or by body mass index ≥30 kg/m2 before initiation of anthracyclines. Laboratory data and outpatient prescriptions 1 year before the index dates were reviewed. If multiple laboratory results were available, the most recent ones before the index dates were chosen for the analysis. LVEF was extracted from echocardiogram reports. LVEF was assessed using the biplane Simpson method according to the echocardiography laboratory's standard protocol.36
    To assess SDOH, we acquired zip code‐level median household income, high school graduation rate, employment rate, and percentage of people in poverty from the US Census Bureau's 2015 American Community Survey 5‐Year Data.37 Subjects' residential addresses were used to derive residence‐specific metrics. Although zip code‐level measures may be a less precise proxy for individual‐level SDOH, they can reflect the impact on individual health outcome by the community of residence and have been consistently shown as reliable predictors of health outcomes in many studies.38, 39, 40, 41, 42 Patients were categorized into quartiles for the entire cohort in ascending order from the lowest (Q1) to the highest (Q4).
    The primary end point, development of cardiotoxicity, was defined as clinical HF or an asymptomatic decline in LVEF (Table S2). All cardiotoxicity events were adjudicated by a cardiologist based on medical records. Clinical HF was determined by a diagnosis made by a board‐certified physician in the chart or based on chart‐review findings of (1) symptoms (dyspnea, decreased exercise tolerance, orthopnea, paroxysmal nocturnal dyspnea); (2) physical examination findings (peripheral edema, crackles, neck‐vein distension, rales, S3 gallop, or fluid retention not explained by renal or liver failure); and (3) diagnostic tests (increased B‐type natriuretic peptide, pulmonary congestion on chest x‐ray, or abnormal LVEF). HF cases needed to meet at least 1 of the criteria in the symptoms or physical examination AND one of the criteria in diagnostic tests. An asymptomatic decline in LVEF was defined as an absolute decrease of ≥10% from baseline to <50% without evidence of clinical HF.43 All‐cause death was ascertained from the Social Security Death Index, the SEER (Surveillance, Epidemiology, and End Results) registry, and hospital records.
    Individuals were categorized into AL high (>2) and AL low groups (≤2) using the median AL score of the cohort. Continuous data were summarized as mean±SD and categorical variables as frequency and percentage. Differences in characteristics between individuals with high and low AL score were assessed using the Student's t test for continuous variables and the Pearson's chi‐square test for categorical variables. The time‐to‐event analysis was used to evaluate the cumulative incidence of cardiotoxicity. The date when the cardiotoxicity criteria were met for the first time was the event date. Death was considered a censoring event. For individuals free of cardiotoxicity, the last contact date or death date was the censoring date. The cardiotoxicity‐free probabilities by high and low AL score and by AL quartiles were presented by the Kaplan–Meier method and compared using the log‐rank test. Univariable and multivariable (adjusting for race and ethnicity, age, sex, SDOH, hypertension, diabetes, hyperlipidemia, obesity, current or former smoker, cumulative anthracycline dose, baseline LVEF, cancer type, and cancer metastasis) Cox proportional‐hazards models were performed to assess the association between AL score and incident cardiotoxicity. As a small percentage of the cohort (45/718, 6%) had missing values in covariables, they were dropped in the multivariable model. Linear regression was performed to assess the association of demographics and SDOH variables with AL score. Additionally, given the plausibility that race and ethnicity associate with chronic stress, which can further affect the risk of anthracycline cardiotoxicity, mediation analysis was performed to examine the prespecified hypothesized mediator path: NH Black or Hispanic race and ethnicity→ ↑ AL → ↑ the risk of anthracycline cardiotoxicity, using the causal mediation analysis implemented in “mediate” command using Stata 18 (StataCorp, College Station, TX).44 We performed the following sensitivity analysis in assessing the association between AL and cardiotoxicity: (1) Fine and Gray's proportional subhazards model (considering death as the competing risk); and (2) limiting the analysis to individuals who underwent doxorubicin treatment. We additionally performed the sensitivity analysis using the individual biomarkers to replace AL score in the multivariable Cox proportional Hazard model. A 2‐sided P value <0.05 was considered significant. Analyses were performed with Stata 18 (StataCorp, College Station, TX).

    RESULTS

    A total of 718 patients were included in the analysis (Figure 1). The cohort was racially and ethnically diverse, with 29% NH White, 31% NH Black, and 40% Hispanic. The mean age was 56.7±15.9 years, 71.3% were women. Mean baseline LVEF was 63.9±5.6%. The average cumulative anthracycline dose was 198.5±99.0 mg/m2. The median follow‐up time was 21 months and the total at risk time was 1419 person‐years. Ninety‐four individuals developed cardiotoxicity (13.1%) at the end of follow‐up. In the entire cohort, 31.8% patients died during the follow‐up and the mortality rate was higher in NH Black patients (39.8%) compared with NH White patients (29.6%) and Hispanic patients (26.9%, P=0.006).
    The mean AL score of the cohort was 2.4±1.4 and it was higher in NH Black and Hispanic individuals than in NH White individuals (2.5±1.3 NH Black versus 2.4±1.3 in Hispanic versus 2.1±1.5 in NH White, P=0.031). Women were found to have a lower AL score compared with men in this cohort (2.2±1.3 versus 2.8±1.4, P<0.001). In patients who developed cardiotoxicity, AL score was higher as compared with patients without cardiotoxicity (2.7±1.4 versus 2.3±1.3, P=0.006). We categorized the cohort based on the median AL score (AL > 2 versus AL≤2). Compared with individuals with a AL score below median, individuals with a AL score above median were older; less likely to be female; had more baseline hypertension, diabetes, hyperlipidemia, or obesity; and more likely were on an angiotensin‐converting enzyme inhibitor or angiotensin receptor blocker, aspirin, or statin (Table 2). There were no significant differences in SDOH measures and cancer metastasis rate by the level of AL score (Table 2). Patients with an AL score above median (>2) received a lower dose of anthracyclines compared with patient with an AL score below median (≤2) (189.5 mg/m2 versus 205.5 mg/m2, P=0.032). The baseline LVEF was similar between individuals in these 2 groups (64.3% versus 63.5%, P=0.059). Compared with individuals with an AL score below median (≤2), individuals with an AL score above median (>2) had worse cardiotoxicity‐free survival (Figure 2A, log‐rank test P=0.014). We subsequently categorized the cohort according to AL quartiles (Q1–Q4). There was a significant difference in cardiotoxicity‐free survival between AL quartiles and a dose–response relationship (Figure 2B, log‐rank test P=0.008).
    Baseline Characteristics by Allostatic Load Score Levels
    Student's t test for continuous variables and the Pearson's chi‐square test for categorical variables.
    Statistical significance.
    The time‐to‐event analysis was used to evaluate the cumulative incidence of cardiotoxicity. The date when the cardiotoxicity criteria were met for the first time was the event date. Death was considered a censoring event. For individuals free of cardiotoxicity, the last contact date or death date was the censoring date. A, Cardiotoxicity free survival by the median allostatic load score (allostatic load ≤2 vs allostatic load >2). B, Cardiotoxicity free survival by allostatic load score quartiles (Q1–Q4). AL indicates allostatic load.
    AL score, as a continuous variable, was significantly associated with the development of cardiotoxicity in the unadjusted model (hazard ratio [HR], 1.33 [95% CI, 1.15–1.54], P<0.001). After adjusting for race and ethnicity, sex and age, the association was slightly attenuated, but remained significant (HR, 1.22 [95% CI, 1.04–1.43], P=0.014). Additional adjustment of SDOH variables (median household income, high school graduation rate, employment rate, and poverty rate), cardiovascular risk factors (hypertension, diabetes, hyperlipidemia, obesity, being current or former smoker), baseline LVEF, anthracycline dose, cancer type, and cancer metastasis did not alter the association (model 3–7, Table 3). Each point increase in AL was associated with 20% higher risk of incident cardiotoxicity (HR, 1.20 [95% CI, 1.02–1.43], P=0.033, model 7, Table 3). In addition, in the multivariable analysis (adjusting for sex, age, race, and ethnicity), a high AL score (Q4) was significantly associated with a higher risk of cardiotoxicity compared with individuals with a low AL score (Q1) (HR, 1.88 [95% CI, 1.01–3.51], P=0.047). SDOH variables were not associated with cardiotoxicity in the adjusted models including race and ethnicity, age, sex, and AL (model 3–6, Table 3). Female sex was associated with cardiotoxicity in the models including AL, race and ethnicity, and SDOH variables (model 2–6, Table 3); however, the association was significantly attenuated after the adjustment of cardiovascular risk factors, anthracycline dose, baseline LVEF, cancer type, and cancer metastasis (model 7, Table 3). Being NH Black or Hispanic was independently associated with incident cardiotoxicity after adjusting for demographics, SDOH, cardiovascular risk factors, baseline LVEF, anthracycline dose, AL, cancer type, and cancer metastasis (model 2–7, Table 3). The association between AL and cardiotoxicity was similar by race and ethnicity and by sex (Figure S1). Sensitivity analysis using the Fine and Gray's proportional subhazards model considering death as the competing risk and limiting the analysis to individuals who underwent doxorubicin treatment only did not alter the results significantly. The sensitivity analysis using individual biomarkers replacing AL score did not yield significant associations with cardiotoxicity (Table S3).
    Associations Between Allostatic Load and Incident Cardiotoxicity
    Model 1: Unadjusted Cox proportional‐hazards model between allostatic load and incident cardiotoxicity.
    Model 2: Multivariable Cox proportional‐hazards model between allostatic load and incident cardiotoxicity adjusting for race and ethnicity, age, and sex.
    Model 3: Model 2+median household income.
    Model 4: Model 2+high school graduation rate.
    Model 5: Model 2+employment rate.
    Model 6: Model 2+poverty percentage.
    Model 7: Model 2+ hypertension, diabetes, hyperlipidemia, obesity, current or former smoker, baseline left ventricular ejection fraction, anthracycline dose, cancer type, and cancer metastasis.
    HR indicates hazard ratio.
    Correlations between baseline characteristics (demographics and SDOH) and AL are shown in the Table 4. In the unadjusted model, sex, age, race and ethnicity, median household income, and high school graduation rate were associated with AL score separately. However, after adjusting for sex, age, race, and ethnicity, SDOH variables were no longer associated with AL score. In the mediation analysis testing whether AL served as a mediator between race and ethnicity and anthracycline cardiotoxicity, the indirect effect indicating what extent race and ethnicity affected cardiotoxicity through AL was not significant (NH Black versus NH White: coefficient − 0.002 [95% CI −0.01 to 0.01], P=0.74; Hispanic versus NH White: coefficient 0.01 [95% CI −0.002 to 0.02], P=0.091).
    Association Between Characteristics and Allostatic Load
    Univariable model: linear regression between the characteristics and allostatic load score.
    Model 1: Linear regression of allostatic load adjusting for age, sex, and race and ethnicity.
    Model 2: linear regression of allostatic load adjusting for age, sex, race and ethnicity, and social determinants of health (median household income, high school graduation rate, employment rate, poverty percentage, separately).

    DISCUSSION

    This large diverse well‐characterized cohort of individuals with cancer from an underserved community offers unique opportunities to investigate the mechanism of racial and ethnic disparities in anthracycline cardiotoxicity. We have several notable findings. (1) NH Black and Hispanic individuals with cancer had a higher level of stress and physiologic dysregulation, demonstrated by a higher AL score, in comparison to NH White individuals. (2) AL score was significantly associated with the development of cardiotoxicity independent of race and ethnicity, sex, age, SDOH, cardiovascular risk factors, baseline LVEF, anthracycline dose, cancer type, and cancer metastasis. Each point increase in AL was associated with 20% higher risk of incident cardiotoxicity and there is a dose–response relationship. (3) The association between AL and cardiotoxicity was similar by race and ethnicity and sex; (4) Although SDOH variables were associated with AL score in the unadjusted model, the relationship between AL and cardiotoxicity was independent of SDOH. (5) AL did not serve as a significant mediator in the association between race and ethnicity and anthracycline cardiotoxicity. To our knowledge, the current study is the first to examine the relationship between AL and cardiotoxicity in patients with cancer.
    Current understanding of disparities in cardio‐oncology is limited by a relative paucity of data.12, 14, 45, 46 Although small retrospective studies reported NH Black individuals have a higher risk of cardiotoxicity from cancer therapies, prior studies are limited by lack of inclusion of Hispanic individuals and lack of characterization of cardiovascular risk factors and SDOH.9, 11 No studies have addressed the mechanism of disparities, which can be crucial to develop strategies to alleviate the gap between race and ethnicity. AL, as a measure of physiological burden of chronic stress, has been associated with adverse health outcomes in cardiovascular disease and cancer.17, 18, 19, 20, 21, 22, 23, 24, 25 However, it has not been investigated in a population involving cardio‐oncology to inform the risk of chemotherapy related cardiotoxicity. In the current study, we leveraged a large diverse cohort of patients with cancer treated with anthracyclines and used the routinely collected biomarkers to construct AL score and inform the risk of anthracycline cardiotoxicity. In our analysis, the AL score is a strong and independent predictor of the development of anthracycline cardiotoxicity and there is a dose–response relationship between AL score and cardiotoxicity.
    The current study highlights the key role of chronic stress and physiologic dysregulation in the pathogenesis of cardiotoxicity and underscores the importance of using stress‐related biomarkers to inform cardio‐oncology practice. Prior biomarker studies in cardio‐oncology were largely related to troponins and natriuretic peptides, reflective of myocardial injury, and other novel markers related to oxidative stress, fibrosis, and inflammation.47, 48, 49, 50 AL provides a unique framework to investigate the impact of stress on cardiotoxicity. The calculation of AL score is heterogeneous throughout literature, making the comparison between studies challenging. Despite such variability, a high AL has been consistently associated with adverse health outcomes.16 We used a method that was previously validated and lower in cost because these biomarkers are widely available in routine clinical care.25, 31, 32 The levels of AL score in the current cohort was similar to previous studies conducted in populations with cancer.24, 25, 51 We did not identify significant associations between individual biomarkers with cardiotoxicity (Table S3) in our cohort, despite some markers (ie, body mass index and hemoglobin) being associated with incident HF in general population.52, 53, 54 This is likely due to limited statistical power. However, the composite AL score synthesizing several markers reflective of multisystem stress levels offered improved prognostication and achieved statistical significance as shown by the main results.
    Future studies are warranted to investigate the predictive value of some other stress‐related biomarkers, such as dehydroepiandrosterone sulfate, urinary cortisol, urinary epinephrine, urinary norepinephrine, urinary norepinephrine, C‐reactive protein, or clinometric interview. The contribution of stress to cardiovascular disease has been proposed through heightened inflammation and oxidative stress.55 One potential pathway involves stress‐associated neurobiological mechanism, where a greater psychosocial stress leads to heightened neurobiological activity, resulting in a biological cascade that involves increased bone marrow activity and arterial inflammation.56, 57 However, mechanistic pathways are yet to be studied in the population involving cardio‐oncology.
    Our study is unique in addressing the complex relationship between race and ethnicity, SDOH, stress, and cardiotoxicity. We found that being NH Black or Hispanic was associated with a higher AL score, which further associated with an increased risk of anthracycline cardiotoxicity and these associations persisted after adjusting for SDOH. In contrast, the associations between SDOH variables and AL or incident cardiotoxicity were weakened after the adjustment of race and ethnicity.13 Our findings are consistent with the NHANES (National Health and Nutrition Examination Survey) data where Black participants had a consistently higher AL than White participants even after adjusting for socioeconomic status.27 It is plausible that there is a disassociation between the presence of psychosocial stress and physiological dysregulation resulting from race and ethnicity and SDOH. AL score, reflective of physiological burden, could capture a more direct and downstream effect of stress to health outcomes, therefore AL is a stronger predictor of cardiotoxicity. Although the mediation analysis did not demonstrate a significant mediator path: NH Black or Hispanic race and ethnicity→ ↑ AL → ↑ the risk of anthracycline cardiotoxicity, this analysis may be limited by the statistical power. How SDOH affect health outcomes can be more complex and can involve a multitude of related variables, leading to a composite effect. Variabilities in individual's neurobiological vulnerability (ie, genetics, constitutional factors), psychosocial resources, learned patterns of coping and community environment can potentially affect the impact of SDOH on cardiotoxicity,58 highlighting that AL reflects the individual physiologic response to stress, not just the presence of the stress itself. Future studies can focus on using AL to identify high‐risk individuals before chemotherapy to guide cardiac function surveillance and intensified cardiovascular risk factor modification or cardioprotective therapies. These strategies may potentially mitigate racial and ethnic disparities in cardiotoxicity in diverse patient populations. There are no studies characterizing cardiotoxicity in Asian population in the United States. Future investigations are warranted to include other unstudied racial and ethnic groups to overcome disparities in cardio‐oncology.

    LIMITATIONS

    Results of the present study should be interpreted in the context of the study design. The adjustment of SDOH is based on neighborhood‐level data, not individual‐level SDOH. However, neighborhood‐level SDOH can reflect the impact on individual health outcome by the community of residence and have been consistently shown as reliable predictors of health outcomes in many studies.38, 39, 40, 41, 42 The study was limited by race and ethnicity collection by hospital records, which may result in misclassification. Some SDOH measures, such as food insecurity, access to health services, and racial residential segregation, were not within the scope of the current study. In the cohort with cancer, AL score may be affected by cancer diagnosis and cancer severity; however, AL has been consistently studied in oncological research.22, 23, 24, 25 In our analysis, the association between AL and cardiotoxicity was independent of cancer type and cancer metastasis. The implementation of AL score is heterogeneous throughout the literature, limiting the comparison between studies. Inconsistencies in assessing AL are mainly due to the absence of a “gold standard,” lack of harmonization, and the availability of parameters. Although the method we used in this study was not compared with other operationalizations, our results were consistent with prior findings on the negative impact of AL on cancer and cardiovascular outcomes.17, 18, 19, 20, 21, 22, 23, 24, 25 Due to lack of validation and standardization of AL calculation, AL score is still primarily used in the research setting, however, its potential clinical implications are being recognized and actively investigated as a potential clinically actionable tool to improve health equity in cardio‐oncology.16 Additionally, there were discrepancies in the methods of determining baseline (ICD codes) and follow‐up events (adjudication), which might have led to misclassification of incident cardiotoxicity. However, the misclassification would not affect the association between AL and cardiotoxicity because it was unlikely to be differential based on the levels of AL. Finally, due to the retrospective nature of the present study, it was challenging to determine the degree of loss to follow‐up. Because the investigators' institution is the largest health care network that serves the entire Bronx, we would assume that the number of losses to follow‐up was small.

    CONCLUSIONS

    To our knowledge, the current study is the first to examine the relationship between AL and cardiotoxicity in patients with cancer. In this large diverse cohort of individuals with cancer from an underserved community, we found that NH Black and Hispanic individuals had a higher AL score than NH White individuals and AL score was a strong independent predictor of incident anthracycline cardiotoxicity. The current study highlights a key role of stress in the development of anthracycline cardiotoxicity and understories risk stratification using stress‐related biomarkers before starting chemotherapy.

    Sources of Funding

    Dr. Lili Zhang received research support from the American Heart Association (23CDA1046240) and the National Center for Advancing Translational Sciences, National Institutes of Health through CTSA award (K12TR004411). Dr. Tomas G. Neilan received research support from the National Institutes of Health (K24HL150238).

    Disclosures

    None.

    Supporting information

    Tables S1–S3
    Figure S1
    JZDr. Justin Zhang
    Leading Voice

    Dr. Justin Zhang

    Pulmonology & Critical Care · Critical Care Medicine
    UCLA Health
    justinz
    Los Angeles County, CA, USA