METHODS:
The Nulliparous Pregnancy Outcomes Study: Monitoring mothers-to-be
(nuMoM2b) is a geographically diverse, prospective, observational cohort
study in which 10,038 nulliparous individuals with singleton pregnancies
were enrolled between October 2010 and September 2013. Individuals were
eligible for enrollment if they were nulliparous (no prior delivery at
20 weeks or later gestational age), had a viable singleton gestation,
had an estimated gestational age of pregnancy between
60–136 weeks, and intended to
deliver at a participating clinical site. The study protocol included
three study visits during pregnancy and a final visit at the time of
delivery. Maternal characteristics and other covariates were ascertained
from baseline clinical assessments, medical record abstraction, and
standardized questionnaires by trained personnel. Details of the study
procedures have been described elsewhere. 11 Each
institutional review board approved the study, and all participants gave
written informed consent.
A follow-up study, nuMoM2b-Heart Health Study (HHS), included nuMoM2b
participants 2 to 7 years after the nuMoM2b index pregnancy. 4508 HHS
participants completed an in-person cardiovascular risk factor
evaluation. HHS participants had not withdrawn from the primary parent
study, had pregnancy outcome data available, and agreed to follow-up
contact at 6 month intervals beginning at least 6 months after delivery
of the index pregnancy. After the in-person study CVD visit,
participants agreed to be contacted at an interval of every 12 months.
Biometric data, measurements, questionnaires, and bio-specimens were
obtained at the in-person nuMoM2b-HHS visit. Further details on the
methods of the HHS have been described elsewhere. 12
This study was a secondary analysis of the nuMoM2b-HHS cohort. We
excluded participants for whom the index pregnancy ended in fetal demise
< 20 weeks or termination. Participants with missing
1st trimester index pregnancy biomarker measurements,
details of APOs, preexisting diabetes, chronic diabetes and other
missing delivery details at index pregnancy were excluded from the
primary analysis. Chronic hypertension and preexisting diabetes at the
index pregnancy, were exclusion criteria from the primary analysis, but
were included in analyses of secondary outcomes MD and incident HTN,
respectively.
Allostatic load biomarkers were processed from stored urine or serum
samples, collected in the first trimester during the nuMoM2b index
pregnancy. Samples were stored at -80oC at a central
core biorepository. Assays were completed at the HHS core laboratory
(Lundquist Institute, Torrance, CA) using standard protocols on a
Beckman AU480.
This study’s definition of allostatic load modifies the
NHANES13, 14 commonly used risk biomarkers by adding
triglycerides, insulin and glucose. Based on available data and assays,
allostatic load was defined using: clinically-measured (systolic blood
pressure (SBP) diastolic blood pressure (DBP), and body mass index (BMI)
(kg/m2)), serum-measured (cholesterol (mg/dL), low-density lipoprotein
(LDL) (mg/dL), high-density lipoprotein (HDL) (mg/dL), high sensitivity
C-reactive protein (hsCRP) (mg/dL), triglycerides (mg/dL), insulin
(ulU/mL), and glucose (mg/dL)) and urine-measured (creatinine (mg/dL)
and albumin (mg/dL)). Numerous variations and definitions of allostatic
load that have been used, and one is not clearly superior to another
[14]. We decided to use biomarkers that were available in our
dataset and had been used in other definitions of allostatic load
commonly utilized in studies of health disparities. These biomarkers
exemplify organ and tissue damage within the following physiological
systems: cardiovascular, inflammation, metabolic, and immune.14-16
A high allostatic load score was defined as four or more out of 12
biomarkers in the worst quartile; the ’worst’ quartile was lowest for
HDL and albumin and highest for the rest. 17 For each
biomarker, if values were at or above the worst quartile (high risk),
that biomarker received a value score of ”1.” Values not in the worst
quartile were characterized as ”low risk” and given a value score of
”0”. 13 The total allostatic load score was summed for
an allostatic load index ranging from 0 to 12. Low allostatic load was
reported as an allostatic load index of 4 or less, and high allostatic
load was an allostatic load index of more than 4 since this threshold
has been discriminatory. 17
The study’s primary outcome, a composite CVD-related outcome, consisted
of hypertension (HTN), and metabolic disorder (MD) newly diagnosed in
the 2 to 7 years after the index pregnancy. The diagnostic threshold for
HTN was based on confirmed elevated or high clinical measurements of
blood pressure (SBP ≥ 120 mm Hg, DBP ≥ 80 mm Hg) or antihypertensive
medication use. The diagnostic threshold for MD consisted of diabetes as
diagnosed by a health care provider, fasting glucose levels
>= 100 mg/dL, or medication use for glucose control.
Individually, HTN and MD were assessed as secondary outcomes and were
chosen due to their strong associations with CVD risk and
mortality.12 Definitions for these outcomes were
standard and previously reported in detail. 18
Obstetric, medical history, clinical features of pregnancy, maternal
demographic, and health behavior characteristics, all measured during
the index pregnancy, were evaluated as risk factors for CVD. Obstetric
and medical history included: gravida, prior miscarriages, and previous
abdominal surgery. Clinical features of pregnancy included bleeding in
the first trimester. Maternal demographic and health behavior
characteristics included maternal age, education, smoking, federal
poverty level, and health insurance status. We report time between index
pregnancy and HHS visit.
Non-Hispanic black race has been associated with chronic stress and
allostatic load. 32 Thus, although race is a social
construct, we evaluated it as a proxy for social experience, systematic,
racism, and other unmeasured social determinants of health that
potentially manifest through chronic stress. Outcomes were compared
between people of self-reported non-Hispanic Black race, and
non-Hispanic White, Hispanic, Asian, Native American, Native Hawaiian,
Multiracial, and additional racial backgrounds. It was not possible to
analyze some groups separately due to small numbers.
Risk factors associated with the composite outcome were identified by
testing differences in percentages with chi-square between individuals
with and without composite outcome. Unadjusted and adjusted odds ratios
(ORs) and 95% confidence interval (CIs) were calculated from
bi-variable logistic regression models between individuals with and
without composite outcome.
Our primary analysis assessed the association between allostatic load
and composite outcome. Unadjusted odds ratios (ORs) and 95% confidence
interval (CIs) for the association of high allostatic load with
composite outcome were calculated from bi-variable logistic regression
models. As secondary outcomes, we evaluated each component of composite
outcome in a separate model using a similar methodology. For
multivariable modeling of composite outcome, maternal age, smoking
status, gravidity, the time between index pregnancy, bleeding at the
first trimester, and health insurance status were chosen either a
priori based on reported associations19, 20 or were
risk factors with an association with the outcome of P-value
<0.10. The same covariates were used in the HTN model with the
addition of preexisting diabetes and in the MD model with the addition
of chronic hypertension. As an additional exploratory analysis, we
modeled each of the twelve individual allostatic load component with
three outcomes for a total of 36 comparisons. A sensitivity analysis of
the primary outcome was performed excluding blood pressure and insulin
from the allostatic load definition, and similarly for each secondary
outcome allostatic load was redefined excluding blood pressure and
insulin separately.
To test whether allostatic load is a pathway that contributes to racial
disparities in CVDs, we conducted a four-step mediation analysis to test
whether allostatic load is a mediator of the relationship between
self-reported race and composite outcome. We first examined the
association of maternal race on composite outcome (path c, Figure 2).21-23 Second, we examined the impact of maternal race
on allostatic load (path a, Figure 2). 21-23 Third, we
report the association between allostatic load with composite outcome
(path b, Figure 2). 21-23 In the final step, we
assessed whether the race-composite outcome relationship was mediated by
allostatic load (path c’, Figure 2). We conducted a sensitivity analysis
of the mediation examining the association between allostatic load and
composite outcome, limiting the analytical population to individuals of
non-Hispanic Black and non-Hispanic White race and ethnicity. This was
repeated for secondary outcomes.
As an exploratory analysis, we tested whether there’s an effect
modification by race between allostatic load and composite outcome. In
unadjusted and adjusted models of composite outcome, we tested for an
interaction between race (non-Hispanic Black vs. “Non-Hispanic White,
Hispanic, Asian, Native American, and Native Hawaiian, multiracial and
additional racial backgrounds”.) and high allostatic load. A
significant interaction would demonstrate a difference in the
association between high allostatic load and CVD outcomes for people of
non-Hispanic Black race compared to people of “Non-Hispanic White,
Hispanic, Asian, Native American, and Native Hawaiian, multiracial and
additional racial backgrounds”. (i.e., moderation). We conducted a
sensitivity analysis of the exploratory moderation examining the
association, limiting the analytical population to individuals of
non-Hispanic Black and non-Hispanic White race and ethnicity. This was
repeated for secondary outcomes.
Data analyses were conducted using SAS 9.4 software (SAS Institute Inc.,
Cary, NC, USA). All tests were performed at a significance level of p
< 0.05, and all single degrees of freedom tests were 2-sided.