Allostatic Load Mediates Associations Between Race, Ethnicity, and Hypertensive Disorders of Pregnancy

Abstract

OBJECTIVE:

To evaluate whether chronic stress exposure, measured by allostatic load (a biological measure of chronic stress embodiment, including stressors exacerbated by structural inequities [e.g., structural racism]) and patient-reported perceived stress in the first trimester of pregnancy, mediates the association between self-identified race/ethnicity and hypertensive disorders of pregnancy (HDP).

METHODS:

This was a secondary analysis of data from the large prospective cohort study, Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-be. We evaluated self-identified race/ethnicity as an independent variable (non-Hispanic [NH]-Black, Hispanic, Asian, NH-White), while our outcome of interest was HDP (i.e., gestational hypertension, preeclampsia/eclampsia). Allostatic load was operationalized using regression- and count-based approaches. Perceived stress was collected using Cohen’s perceived stress scale. We investigated allostatic load and perceived stress and used causal mediation analyses with a counterfactual approach to evaluate whether they mediated the association between self-identified race/ethnicity and HDP, adjusting for age and tobacco use. Mediation analyses were conducted for each minoritized racial/ethnic group compared to NH-White participants.

RESULTS:

The sample included 645 participants who developed HDP and 2,438 participants without HDP or other adverse pregnancy outcome. Allostatic load and perceived stress varied by race/ethnicity, while HDP varied by allostatic load but not perceived stress. Allostatic load was a partial mediator exclusively in the comparison of NH-Black versus NH-White participants (0.027, 95% CI: 0.013 to 0.040, p<0.001; 28.9%). Perceived stress was not a significant mediator.

CONCLUSION:

First-trimester allostatic load mediated the association between self-identified race/ethnicity and HDP for NH-Black and NH-White participants. This mediation effect was not observed in other racial/ethnic comparisons. These results demonstrate a physiological pathway through which racism may contribute to adverse pregnancy outcomes and suggest that interventions targeting allostatic load reduction could help address racial and ethnic disparities in HDP.

Precis:

First-trimester allostatic load mediates the association between self-identified race, ethnicity, and hypertensive disorders of pregnancy in a nulliparous cohort.

Introduction

Chronic stress has long been posited as a potential driver of health inequities, including pregnancy outcomes.^1–12 Hypertensive disorders of pregnancy (HDP; gestational hypertension, preeclampsia, and eclampsia) are significant contributors to pregnancy-related mortality and morbidity, as well as accelerated cardiovascular disease among birthing individuals.^{13,14,1,15–18} Among individuals who had pregnancy-related deaths in U.S. hospitals from 2017 to 2019, 31.6% were diagnosed with HDP.¹⁹ Substantial racial and ethnic disparities persist in pregnancy outcomes and cardiovascular disease, and these disparities have worsened in recent years.^19–28 Importantly, racism has long functioned as a chronic stressor for racially and ethnically minoritized populations—particularly for Black and Indigenous communities^29–32—though its role in driving physiological dysregulation and health inequities only recently began to receive sustained attention in biomedical research.³³

The mechanism by which the body maintains homeostasis in response to stress is termed “allostasis.”³⁴ Chronic stress can lead to systemic dysregulation and cumulative physiological burden, known as allostatic load (AL).^34,35 AL represents the biological embodiment of chronic stress and is associated with increased susceptibility to chronic diseases, including cardiovascular disease and metabolic disorders.^35–45,12 In contrast, perceived stress reflects an individual’s subjective evaluation of life stressors as overwhelming, uncontrollable, or unpredictable.⁴⁶ Despite evidence linking chronic stress to health disparities, the role of AL and perceived stress in racial/ethnic disparities in HDP remains underexplored.^47–49 Understanding whether biological (AL) or patient-reported (perceived stress) measures of chronic stress mediate these disparities may help inform targeted interventions. The purpose of the present study was to (1) evaluate first-trimester AL and perceived stress by self-identified race/ethnicity and HDP status and (2) determine whether first-trimester AL or perceived stress mediates the association between self-identified race/ethnicity and HDP in a large, nulliparous population.

Methods

Protocols for the Nulliparous Pregnancy Outcome Study: Monitoring Mothers-to-be (nuMoM2b) and the subsequent Heart Health Study (HHS1) have been previously published.^50,51 nuMoM2b, a prospective observational cohort study conducted between 2010 and 2013 at eight medical centers in the United States, recruited 10,038 nulliparous individuals during their first trimester and followed them for three visits across pregnancy and a fourth visit at birth. Eligibility criteria included viable singleton pregnancy, nulliparity, gestational age between 6^0/7 and 13^6/7 weeks at enrollment, and intention to give birth at the participating clinical site. Individuals were ineligible if they had fetal abnormalities or aneuploidy, planned to terminate the pregnancy, had ≥3 previous pregnancy losses, or were <13 years of age. Trained study coordinators obtained clinical characteristics of participants from assessments, standardized questionnaires, and medical records.

A total of 4,508 nuMoM2b participants completed assessments in the follow-up HHS1. During HHS1, biomarkers were assayed from first-trimester blood and urine samples that had been previously collected and stored from 4,232 participants during nuMoM2b. Both nuMoM2b and HHS1 were IRB-approved by each site and completed with informed consent from participants. The present analysis was IRB-approved at the University of Pittsburgh.

Figure 1 displays the derivation of samples for the present analyses. We selected nuMoM2b participants who also participated in HHS1, as the biomarkers used in our investigation were measured exclusively among HHS1 participants. To define a low-risk reference group for comparison with participants who developed HDP, we excluded those with HDP and other adverse pregnancy outcomes (APO), defined as gestational diabetes or diabetes mellitus, spontaneous preterm birth, small-for-gestational age, or stillbirth. We also excluded individuals with chronic hypertension diagnosed prior to nuMoM2b enrollment or before 20 weeks of gestation, as our analysis focused on hypertension arising during pregnancy. Lastly, we excluded participants without race/ethnicity data or who reported a race/ethnicity of “Other” (i.e., American Indian, Native Hawaiian, Multiracial). The “Other” category represented a highly heterogeneous group, and aggregating these identities into a single category would limit the interpretability and validity of any group-specific conditions. For AL analyses, we excluded participants missing one or more AL component values (detailed below), resulting in a sample of 3,148 participants. Analyses evaluating perceived stress included 3,328 participants.

Figure 1.

Derivation of samples from parent study for present analyses. *A total of 113 participants had two or more other adverse pregnancy outcomes (APO). PE, preeclampsia or eclampsia (includes mild, severe, and superimposed preeclampsia and eclampsia); gHTN, gestational hypertension (includes new-onset antepartum and intrapartum or postpartum hypertension).

APOs, including HDP—defined as gestational hypertension (new-onset antepartum or intrapartum/postpartum hypertension based on SBP or DBP >140 mmHg and/or >90 mmHg without end-organ manifestations), preeclampsia or eclampsia—were verified by chart abstraction with adjudication.⁵⁰ HDP were classified according to the American College of Obstetricians and Gynecologists’ 2013 guidelines.^50,52

Non-fasting peripheral blood and urine were collected during the first trimester (6^0/7-13^6/7 weeks’ gestation), via venipuncture and clean catch, respectively, and stored at −80°C at a central core biorepository (ThermoFisher Scientific; Waltham, Massachusetts, USA). At the first-trimester study visit, systolic (SBP) and diastolic (DBP) blood pressures were measured using an aneroid sphygmomanometer, following a standardized protocol, with participants seated and rested. Body mass index (BMI) was calculated as kg/m², using height measured using a stadiometer or measuring tape and weight measured with a digital scale or calibrated balance beam.

Biospecimens from serum and urine were assayed at the HHS core lab at the Lundquist Institute (Torrance, California). Specifically, a Beckman AU480 analyzer was utilized to quantitate glucose, total cholesterol, triglycerides, and high-density lipoprotein (HDL). Low-density lipoprotein (LDL) was calculated using the Friedewald equation. High-sensitivity C-reactive protein (CRP) was measured using turbidimetric analysis and spectrophotometry. A Beckman ACCESS 2 Immunoassay System was utilized to measure insulin levels. Urinary albumin and creatinine were measured via the modified Doumas and Rodkey method and modified Jaffe procedure, respectively. Albumin and creatinine were measured from urine; all other biomarkers were measured from serum.⁵³

We calculated first-trimester AL using a 12-component measure, based on prior protocols, spanning the cardiovascular, inflammatory, and metabolic domains of physiology.^53–56 The cardiovascular domain included DBP and SBP; the inflammatory domain included urinary albumin and high-sensitivity CRP; and the metabolic domain included BMI, urinary creatinine, glucose, HDL, insulin, LDL, total cholesterol, and triglycerides.⁵³ To ensure rigor and contextualize our findings within existing literature, we operationalized AL using two validated methods: the logistic regression-based approach—recommended for reproductive-aged women when outcome data are available⁵⁷—used in our primary analyses, and the count-based approach—most commonly used in the literature^57,58—used in secondary analyses.

For the logistic regression-based method (rAL), the 12 AL components were standardized to a mean of 0 and standard deviation of 1. We conducted multivariable logistic regression with the 12 components as predictors and HDP status as the outcome. The resulting regression coefficients (Appendix 1) were used as weights for AL components. Each participant’s standardized component values were multiplied by the corresponding regression coefficient. Finally, a composite AL score was calculated for each participant by summing their 12 weighted component values.

For the count-based method (cAL), high-risk thresholds were defined using high-risk quartiles calculated from HHS1 participants with complete AL data (n=4,232; Appendix 2). Participants received a score of 1 for each component in high-risk range and 0 for each in normal range. A composite AL score was calculated by summing the 12 dichotomous (0/1) risk scores, yielding a total score ranging from 0 to 12.

Perceived stress was measured during the first trimester visit of nuMoM2b using Cohen’s perceived stress scale⁴⁶, which measures perceived stress over the last month. Scores range from 0 to 40 (with higher scores indicating greater stress), and the scale has been validated in diverse racial/ethnic and pregnant populations⁵⁹.

All analyses were completed using R version 4.1.2.⁶⁰ We compared sample demographics and clinical characteristics between participants with HDP and those with no APO by using unpaired student t-tests for quantitative variables (participant age, gestational age at birth, birthweight, AL components, rAL), Pearson chi-squared tests for categorical variables (race/ethnicity, education, health insurance, tobacco use), and Wilcoxon rank-sum tests for cAL.

We evaluated associations of first-trimester (1) rAL and (2) perceived stress with self-identified race/ethnicity using linear regression models adjusted for participant age and tobacco use (Y/N: self-reported use during the three months prior to pregnancy). In secondary analyses, we evaluated cAL using Poisson regression with the same covariate adjustments. Age and tobacco use are well-established HDP risk factors and may confound the relationship between race/ethnicity and HDP due to differences in age distribution and tobacco use across groups.^17,61 Tobacco use is associated with altered stress response,^62,63 which may influence the pathways examined in this study. We initially ran all models without any adjustment and results were largely consistent; we present adjusted models for clarity and to account for these known confounders. To compare AL and perceived stress by HDP status, we stratified participants by race/ethnicity and used unpaired student t-tests to compare those who developed HDP with those who developed no APO. We also examined the relationship between first-trimester AL and perceived stress using the Spearman correlation test. False discovery rate adjusted p-values were calculated to account for multiple comparisons.

We used the R ‘mediation’ package to conduct causal inference-based mediation analyses to evaluate whether first-trimester AL mediates the association between self-identified race/ethnicity (NH-Black, Hispanic, Asian, and NH-White) and HDP.⁶⁴ As a preliminary step to establish a significant relationship between the exposure and outcome, we used logistic regression to evaluate the association between self-identified race/ethnicity and HDP. We then performed model-based mediation analyses. First, we fit the mediator model by regressing AL on self-identified race/ethnicity, participant age, and tobacco use in a linear regression. Next, we fit the outcome model by regressing HDP status on AL, self-identified race/ethnicity, participant age, and tobacco use in a logistic regression. We used the ‘mediate’ function with nonparametric bootstrapping (1,000 simulations) to decompose the total effect into causal mediation and direct effects. We conducted three mediation analyses to evaluate whether AL mediated the association between race/ethnicity and HDP for (1) NH-Black, (2) Hispanic, and (3) Asian participants, using NH-White participants as the comparison group, reflecting our focus on stress exposures and HDP among minoritized individuals. These steps were repeated to evaluate first-trimester perceived stress as a potential mediator.

Primary analyses examined rAL and secondary analyses examined cAL; Poisson regression was used for cAL in place of linear regression. Finally, to address potential sampling and selection biases due to overrepresentation of HDP cases, we repeated the mediation analyses using inverse probability weighting (Appendix 1, available online at http://links.lww.com/xxx).

Results

The analytic sample included 3,148 participants, of whom 686 (21.8%) developed HDP and 2,462 (78.2%) had no APO (Table 1). The majority of participants self-identified race/ethnicity as NH-White (69.3%), with smaller proportions identifying as Asian (2.9%), Hispanic (15.8%), and NH-Black (12.0%). Participants with HDP gave birth at a significantly earlier gestational age than those without an APO. They were also significantly older and more likely to use tobacco than no APO participants.

Table 1.

Sample demographics and clinical characteristics of participants with HDP and No APO. Values represent indicated units ± SD.

Sample for Allostatic Load Analyses
Demographics/Characteristics	Total Sample	HDP	No APO	p-value
n (%)	3,148	686 (21.8%)	2,462 (78.2%)	-
Self-identified Race/Ethnicity, n (%): Asian	92	19 (20.7%)	73 (79.3%)	<0.001***
Hispanic	498	76 (15.3%)	422 (84.7%)
Non-Hispanic Black	377	110 (29.2%)	267 (70.8%)
Non-Hispanic White	2,181	481 (22.1%)	1,700 (77.9%)
Mean Participant Age at Visit, years	27.1 ± 5.4	27.6 ± 5.7	26.9 ± 5.3	0.036*
Mean Gestational Age at Birth of Infant, weeks	39.2 ± 1.5	38.9 ± 2.0	39.3 ± 1.3	<0.001***
Mean Infant Birthweight, grams	3,418 ± 423	3,378 ± 508	3,429 ± 396	0.017*
Education†, n (%): Yes	1,122 (35.6%)	242 (35.3%)	880 (35.7%)	0.822
Government Health Insurance, n (%): Yes	781 (24.8%)	175 (25.5%)	606 (24.6%)	0.624
Tobacco Use‡, n(%): Yes	448 (14.2%)	127 (18.5%)	321 (13.0%)	<0.001***
Mean First-trimester Allostatic Load Components
Diastolic Blood Pressure (mmHg)	67.1 ± 8.1	69.4 ± 8.6	66.4 ± 7.8	<0.001***
Systolic Blood Pressure (mmHg)	108.9 ± 10.4	112.4 ± 10.7	108.0 ± 10.2	<0.001***
Albumin (g/dL)	0.66 ± 1.8	0.71 ± 2.5	0.65 ± 1.6	0.543
C-Reactive Protein (mg/L)	0.70 ± 0.80	0.84 ± 0.82	0.66 ± 0.79	<0.001***
Body Mass Index (kg/m2)	26.1 ± 6.0	28.3 ± 6.7	25.5 ± 5.6	<0.001***
Creatinine (mg/dL)	99.7 ± 71.4	100.3 ± 71.3	99.5 ± 71.5	0.811
Glucose (mg/dL)	87.4 ± 14.1	88.3 ± 13.8	87.1 ± 14.1	0.052
High-density lipoprotein (mg/dL)	73.0 ± 15.1	72.8 ± 15.8	73.0 ± 14.9	0.668
Insulin (uIU/mL)	15.7 ± 21.3	19.6 ± 27.4	14.6 ± 19.1	<0.001***
Low-density lipoprotein (mg/dL)	89.2 ± 27.1	91.2 ± 26.6	88.6 ± 27.2	0.025*
Triglycerides (mg/dL)	125.2 ± 47.8	134.2 ± 50.5	122.7 ± 46.7	<0.001***
Total Cholesterol (mg/dL)	187.2 ± 35.2	190.8 ± 34.9	186.2 ± 35.2	0.002**
Median Regression-based Allostatic Load Score	−0.03 ± 0.01	0.20 ± 0.02	−0.10 ± 0.01	<0.001***
Median Count-based Allostatic Load Score [IQR]	3.0 [2, 4]	3.7 [2, 5]	2.8 [1, 4]	<0.001***
Sample for Perceived Stress Analyses
Demographics/Characteristics	Overall	HDP	No APO	p-value
n (%)	3,328	724 (21.8%)	2,604 (78.2%)	-
Self-identified Race/Ethnicity, n (%):Non-Hispanic Black	419	122 (29.1%)	297 (70.9%)	<0.001***
Hispanic	543	76 (14.0%)	467 (86.0%)	-
Asian	99	22 (22.2%)	77 (77.8%)	-
Non-Hispanic White	2,267	504 (22.2%)	1,763 (77.8%)	-
Mean Participant Age at Visit, years	27.1 ± 5.4	27.5 ± 5.7	26.9 ± 5.3	0.017*
Mean Gestational Age at Birth of Infant, weeks	39.2 ± 1.4	38.9 ± 1.9	39.3 ± 1.3	<0.001***
Mean Infant Birthweight, grams	3,419 ± 422	3,381 ± 500	3,429 ± 397	0.018*
Education†, n (%): Yes	1,196 (35.9%)	254 (35.1%)	942 (36.2%)	0.588
Government Health Insurance, n (%): Yes	845 (25.4%)	185 (25.6%)	660 (25.3%)	0.906
Tobacco Use‡, n(%): Yes	475 (14.3%)	133 (18.6%)	342 (13.0%)	<0.001***
Median Perceived Stress Score	12 ± 6.6	12 ± 6.4	13 ± 6.6	0.704

IQR- interquartile ratio.

^†Some college or less;

^‡Smoked tobacco three months prior to pregnancy; Quantitative variables were compared using unpaired Student t-tests; categorical variables were compared using Pearson chi-squared test; count-based allostatic load scores were compared using Wilcoxon test.

^*p-value: <0.05,

^**p-value <0.005,

^***p-value <0.001.

Median logistic regression-based allostatic load (rAL) and count-based allostatic load (cAL) were significantly higher among participants who developed HDP compared to those with no APO. Several individual AL components also differed between groups.

Participants who developed HDP had higher rAL than those with no APO across all racial/ethnic groups. Among NH-White participants, mean±SD rAL was 0.201±0.504 in the HDP group compared to −0.100±0.453 in the no APO group (false discovery rate-adjusted p-value [p_adj]<0.001). Similar differences were observed among NH-Black (0.211±0.584 vs. −0.040±0.544, p_adj=0.0003), Hispanic (0.159±0.558 vs. −0.088±0.482, p_adj=0.0005), and Asian (0.149±0.506 vs. −0.269±0.353, p_adj=0.003) participants (Figure 2A). A similar pattern was observed for cAL (Appendix 4).

Figure 2.

First-trimester regression-based allostatic load (A) and perceived stress (B) by hypertensive disorders of pregnancy (HDP) status, stratified by self-identified race and ethnicity. Raincloud plots display individual data points, box plots, and half-density curves. Lower and upper hinges correspond to the first (25^th) and third (75^th) quartiles; horizontal line indicates mean. Unpaired t-tests with false discovery rate (FDR) adjustment for multiple comparisons accounting for four tests. *FDR-adjusted p-value <.05, **FDR-adjusted p-value <.005, ***FDR-adjusted p-value <.001. APO, adverse pregnancy outcome.

In the overall sample, NH-Black participants had higher rAL than NH-White participants (Estimate [95% CI]: 0.12 [0.07 to 0.18], p<0.001; Table 2). rAL was significantly lower among Asian participants compared to NH-White participants (−0.18 [−0.28 to −0.08], p=0.0005). There was no significant difference in rAL between Hispanic and NH-White participants. Evaluation of cAL revealed similar results for Asian and Hispanic compared with NH-White participants. However, there was no significant difference between NH-Black and NH-White participants (Appendix 5). Self-identified race/ethnicity was significantly associated with HDP among NH-Black and Hispanic participants (Appendix 6).

Table 2.

Mean first-trimester stress exposures and linear regression results for the associations between self-identified race/ethnicity and first-trimester stress exposures (allostatic load, perceived stress).

Regression-based Allostatic Load
Overall Sample (n=3,148)
Race/Ethnicity (n)	Mean Allostatic Load Score (SD)	Estimate (95% CI)	p-value
Asian (n=92)	−0.183 ± 0.422	−0.18 (−0.28 to −0.08)	<0.001***
Hispanic (n=498)	−0.051 ± 0.502	0.03 (−0.02 to 0.08)	0.290
NH-Black (n=377)	0.033 ± 0.567	0.12 (0.07 to 0.18)	<0.001***
NH-White (n=2,181)	−0.034 ± 0.481	Reference
Participant Age	--	0.01 (0.01 to 0.02)	<0.001***
Tobacco Use	--	−0.08 (−0.13 to −0.03)	0.002**
Perceived Stress
Overall Sample (n=3,328)
Race/Ethnicity (n)	Mean Perceived Stress Score (SD)	Estimate (95% CI)	p-value
Asian (n=99)	12.3 ± 5.9	1.42 (0.15 to 2.70)	0.029*
Hispanic (n=543)	13.6 ± 7.0	1.32 (0.71 to 1.92)	<0.001***
NH-Black (n=419)	15.0 ± 7.2	2.08 (1.39 to 2.77)	<0.001***
NH-White (n=2,267)	11.7 ± 6.2	Reference
Participant Age	--	−0.19 (−0.24 to −0.15)	<0.001***
Tobacco Use	--	−2.87 (−3.49 to −2.25)	<0.001***

Regression models adjusted for participant age and tobacco use.

SD – Standard Deviation; NH – Non-Hispanic; CI – Confidence Interval.

^*p-value: <0.05,

^**p-value <0.005,

^***p-value <0.001.

Figure 3 shows results from causal mediation analyses evaluating first-trimester rAL as a mediator of the association between self-identified race/ethnicity and HDP. Among NH-Black compared to NH-White participants, we observed a significant average causal mediation effect (Estimate [95% CI]: 0.027 [0.013 to 0.040], p<0.001) and a significant direct effect (0.064 [0.014 to 0.120], p=0.010). These results suggest that rAL partially mediates the disparity in HDP between NH-Black and NH-White participants. The proportion mediated was 28.9% (i.e., 0.027/(0.027+0.064)*100). In the comparison of Asian and NH-White participants, we observed a significant average causal mediation effect (−0.034 [−0.051 to −0.017], p<0.001). However, the total effect was not significant (p=0.564); as such, we cannot conclude that odds of developing HDP differs overall between these groups. We did not observe evidence for mediation between Hispanic and NH-White participants. Results remained consistent after applying inverse probability weighting to the mediation analyses (Appendix 7). Results evaluating cAL (Appendix 8) were consistent with rAL for Asian and Hispanic participants. However, cAL did not show significant mediation effects for NH-Black compared to NH-White participants.

Figure 3.

Representation of first-trimester regression-based allostatic load mediating the association between race and ethnicity and hypertensive disorders of pregnancy for Non-Hispanic (NH)-Black versus NH-White (A), Hispanic versus NH-White (B), and Asian versus NH-White (C) participants. Estimates, 95% CIs (in parentheses), and p-values are provided for average causal mediated, average direct, and total effects, as well as proportions mediated (%Med) for significant average causal mediation effects. Proportion mediated can be outside 0-100% range when the mediation effect is complete (no statistically significant direct effect) or nonexistent (no statistically significant mediation effect). *p-value: <.05, ^†p-value <.005, ^‡p-value <.001.

Median perceived stress scores did not significantly differ by HDP status overall (Table 1) or in any racial/ethnic group (Figure 2B). Perceived stress was weakly but significantly correlated with rAL (Spearman coefficient: 0.032, p=0.039), but not with cAL (Spearman coefficient: 0.004, p=0.840). First-trimester perceived stress was significantly higher among NH-Black, Hispanic, and Asian participants compared to NH-White participants (Table 2). NH-Black participants showed the largest difference from NH-White (2.08 [1.39 to 2.77], p<0.001), followed by Hispanic (1.32 [0.71 to 1.92], p<0.001), and Asian (1.42 [0.15 to 2.70], p=0.029) participants.

First-trimester perceived stress did not significantly mediate the association between self-identified race/ethnicity and HDP for any racial/ethnic group (Appendix 9). While application of inverse probability weighting resulted in attenuation of the association between race/ethnicity and perceived stress among Hispanic participants, mediation effects remained similar (Appendix 3; Appendix 10).

Discussion

First-trimester rAL was a partial mediator, accounting for 28.9% of the Black-White disparity in development of HDP. These findings suggest that physiological dysregulation linked to chronic stress embodiment may contribute to racial disparities in HDP, potentially reflecting the influence of structural racism and adverse social drivers of health.^3,65–70 These findings align with a previous study suggesting that social-environmental stress is a key driver of HDP disparities.⁷¹ The significant direct effect between race/ethnicity and HDP demonstrates that AL is only one manifestation of social-environmental differences, with other mechanisms also contributing to this disparity. These results, combined with the lack of mediation observed among Hispanic and Asian participants, highlight the complex, multifactorial nature of HDP etiology, suggesting that factors beyond AL, including alternative protective or risk-modifying factors, contribute to racial/ethnic disparities in HDP risk.^{6,65–67,72–74}

While few studies have evaluated AL and HDP,^56,75 our findings complement those of Lueth et al., who reported partial mediation by cAL in another subset of this cohort.⁵³ However, our study had key methodological differences. We compared three minoritized racial/ethnic groups to NH-White participants, while Lueth et al. compared NH-Black participants to all other races/ethnicities combined. In contrast to Lueth et al.’s four-step approach, we used causal-inference-based mediation analyses, offering a more nuanced understanding of AL as a causal mediator. Lastly, we focused on rAL and coded cAL as integer counts in secondary analyses, while Lueth et al. dichotomized cAL as “high” versus “low”. These differences in exposure definition, approach, and cAL operationalization likely contributed to variations in findings.

Nulliparous Asian and Hispanic individuals have been reported to have lower odds of developing HDP than NH-White individuals,⁷⁶ yet few studies have evaluated first-trimester AL as a causal factor in this relationship. Protective factors, such as lower acculturation or greater family support, may buffer AL by promoting homeostasis or facilitating homeostatic recovery, with these effects potentially varying across racial/ethnic groups.^77–81 Hispanic participants had the lowest HDP rates, possibly due to multi-level protective factors. In contrast, Black individuals face institutionalized racism and systematic oppression not experienced to the same extent by other minoritized groups.^{29–32,82–84} Intergenerational effects from chronic stress (e.g., epigenetics) may contribute to variability observed in pregnancy outcomes in minoritized groups.^{31,32,83,85,86} AL was significantly higher among those who developed HDP regardless of race/ethnicity, further supporting the large body of literature establishing AL as a marker of suboptimal health, associated with disease susceptibility.^{34–36,38,42,43,87,88}

Perceived stress was higher in all minoritized groups compared to NH-White participants, consistent with previous reports.^6,89–91 However, perceived stress was not a significant mediator of the race/ethnicity-HDP association. This may reflect uniformly high first-trimester stress levels and the scale’s limited sensitivity to structural or cumulative stressors, as it captures only the past month. This finding is consistent with a previous report from the nuMoM2b cohort, where perceived stress did not affect the relationships between race/ethnicity and APOs.⁶ We found a weak correlation between perceived stress and rAL; while related, they may represent distinct pathways influencing pregnancy outcomes, though modeling interactions was beyond the scope of this analysis.

This study had many strengths including the design of the parent study, which recruited a large, diverse sample with rigorous APO phenotyping. Including both rAL and cAL enables comparability with prior studies while demonstrating how rAL may offer greater sensitivity in detecting meaningful differences, particularly in the context of health disparities. The finding that mediation occurred with rAL but not cAL likely reflects differences in how these measures are operationalized, an issue frequently critiqued in AL literature.^57,92,93 Our results support emerging recommendations for more nuanced and statistically rigorous AL measurements to better detect biologically meaningful variation, particularly in relation to social stressors and racial disparities. We intentionally avoided adjusting for variables that may lie on the causal pathway between race/ethnicity and AL or HDP, but residual confounding may persist due to unmeasured factors such as neighborhood context, structural inequities, and/or cumulative socioeconomic stressors. Collider bias is also possible, as AL components may be associated with both race/ethnicity and HDP. Finally, mediation analyses may have been underpowered to detect small effects. Despite these limitations, examining these relationships remains essential to understanding the complex pathways underlying disparities. Future studies should move beyond race/ethnicity and include direct measures of structural and contextual factors (e.g., redlining, neighborhood deprivation).⁹⁴

Our results demonstrate AL is a partially explanatory factor for the Black-White disparity of HDP and highlight the variability in how different communities of color experience the biological effects of injustice.^95–97