Relationship Between Maternal Economic Vulnerability and Childhood Neurodevelopment at 2 and 5 Years of Life

Abstract

OBJECTIVE:

To assess the relationship between economic vulnerability during pregnancy and childhood neurodevelopment.

METHODS:

This is a secondary analysis of 2 parallel multicenter, randomized, controlled trials of administration of levothyroxine to pregnant individuals with subclinical hypothyroidism or hypothyroxinemia in the U.S. All participants who delivered a live, non-anomalous neonate and completed a Weschler Preschool & Primary Scale of Intelligence (WPPSI-III) at 5 years of life and a Bayley Scales of Infant Development (Bayley-III) test at 2 years were included. The primary outcome is WPPSI-III score. Secondary outcome included Bayley-III subtest scores. Multivariable analyses were used to assess the relationships between economic vulnerability during the index pregnancy — defined as a household income < 200% of the estimated federal poverty line (FPL), part-time or no employment, and use of government insurance — and the prespecified outcomes. Tests of interaction were performed to assess whether the magnitude of association differed according to whether participants were married or completed > high school (HS) education. A sensitivity analysis was performed limiting the income criteria for economic vulnerability to household income <100% of the estimated FPL.

RESULTS:

Of 955 participants who met inclusion criteria, 406 (42.5%) were considered economically vulnerable. In bivariate analysis, the WPPSI-III score and Bayley-III subtest scores were significantly lower among children of the economically vulnerable. For the WPPSI-III, Bayley-III cognitive subtest, and Bayley-III language subtest scores, the associations between economic vulnerability and lower childhood neurodevelopmental scores were primarily seen only among those who were married or completed > HS education (p for interaction < 0.05). A similar pattern was noted when restricting the income criteria for economic vulnerability to <100% FPL.

CONCLUSION:

Economic vulnerability during pregnancy is associated with an increased risk of adverse neurodevelopmental outcomes in their children at 2 and 5 years of life, particularly among those who are married or completed > HS education.

PRÉCIS:

Exposure to economic vulnerability during pregnancy is associated with increased risk of adverse childhood neurodevelopmental outcomes, particularly among participants who are married or had more education.

INTRODUCTION

In the U.S., sociodemographic markers — such as maternal race or ethnicity, English language competency, and insurance status — have been linked to adverse perinatal outcomes.^1-6 To further understand the relationship between different sociodemographic indicators and adverse health outcomes, previous investigators outside of the U.S. have utilized the exposure of a relative deprivation index, which is a metric of economic vulnerability, composed of different social aspects of access and utilization of resources (i.e., owning a car, having insurance, living in an area with a low prevalence of unemployment).^7-9 High relative deprivation indices, indicating greater socioeconomic marginalization, have been associated with inadequate prenatal care utilization.^10,11

Multiple studies have demonstrated the association between exposure to childhood adverse economic conditions, such as those captured by relative deprivation indices, and poor neurodevelopmental outcomes¹². Other social circumstances, such as parental educational status and occupation also have been associated with childhood cognitive status.^13-15 These exposures could be theorized to alter outcomes through epigenetic pathways, similar to the association seen between maternal stress (e.g., due to poverty) during pregnancy and adverse pregnancy outcomes.^16-19

Therefore, we performed an analysis to investigate whether markers of economic vulnerability (i.e., personal income, employment status, and type of insurance utilized for perinatal care) during pregnancy were associated with adverse childhood neurodevelopmental outcomes, and whether there was an interaction between economic vulnerability during pregnancy and markers of social capital (i.e., marital status and level of maternal education).²⁰

METHODS

This is a secondary analysis of two parallel multicenter, randomized, placebo-controlled trials conducted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development-Maternal Fetal Medicine Units Network focused on treatment of subclinical hypothyroidism or hypothyroxinemia during pregnancy and subsequent childhood neurological outcomes. Institutional review board approval was obtained at each institution prior to initiation of the trials.²¹ In brief, participants with a singleton pregnancy between 8 weeks 0 days and 20 weeks 6 days of gestation who were diagnosed with either thyroid disorder were randomized to treatment with levothyroxine or placebo for the duration of pregnancy, with adjustments made to achieve a euthyroid state. Childhood neurodevelopmental outcomes were assessed with the Weschler Preschool & Primary Scale of Intelligence (WPPSI-III) at 5 years of life or the Bayley Scales of Infant Development (Bayley-III) at 2 years of life. For the purposes of our analysis, we limited inclusion to those who delivered a live, non-anomalous neonate.

The primary exposure in this analysis is economic vulnerability during pregnancy, defined as having each of the following three components: household estimated federal poverty level (FPL) of less than 200%, part-time or no employment, and use of government insurance. As the parent trial did not collect FPL from participants a priori, we estimated the percent below FPL based on self-reported marital status, household income, parity of the participant, and year of study inclusion.²² A FPL of less than 200% was chosen as part of the combined metric, as this threshold is commonly used to demonstrate a high economic need for state-based assistance, such as Medicaid.²³ The other markers of economic vulnerability have previously been demonstrated in the literature to be associated with adverse perinatal outcomes.^24,25 The data used to compose economic vulnerability were derived from information collected during the baseline randomization visit of the index pregnancy.

The primary outcome for this analysis is the WPPSI-III score at 5 years of life; this was chosen as the primary outcome in order to limit detection bias for adverse neurodevelopmental outcomes. The WPPSI-III, developed to assess cognitive functioning, has an expected population mean of 100 with a standard deviation of 15, and is useful in detecting mild to moderate cognitive delay.^21,26 Secondary outcomes include Bayley-III scores for cognitive, language, and motor subtests at 2 years of life. Johnson et al note that Bayley-III cognitive and language scores < 85 are indicative of moderate-to-severe neurodevelopmental delay.^27,28 Both scoring systems show robust interrater reliability and validity for detection of neurodevelopmental delay in the U.S.

Confounding variables considered were maternal age, maternal race or ethnicity, substance use during pregnancy, trial center, trial arm (e.g., levothyroxine or placebo), baseline thyroid status (e.g., hypothyroxinemia or subclinical hypothyroidism), marital status, maternal educational level, and maternal housing status. Maternal race and ethnicity were self-reported and categorized as non-Hispanic White, non-Hispanic Black, Hispanic, or other. These variables were selected a priori after a literature search focusing on the relationship between relative deprivation indices and adverse perinatal outcomes.^10,11,24,25 Race and ethnicity were specifically selected as confounding variables given the association between markers of economic deprivation, such as household income²⁹, and previous data demonstrating a relationship between race and ethnicity and Bayley-III scores among preterm neonates; however, the latter finding is confounded by other sociodemographic and biomedical factors, such as receipt of early intervention.³⁰

Given our hypothesis that economic vulnerability would lead to adverse neurodevelopmental outcomes through maternal stress pathways (e.g., activation of maternal hypothalamic-pituitary-adrenal axis with increase in cortisol) and subsequent adverse obstetrical outcomes^16,17,31, we chose not to control for preterm birth or gestational age at delivery. Furthermore, given data demonstrating that maternal mental illness may lie along the causal pathway linking economic vulnerability to adverse childhood neurodevelopmental outcomes, we did not control for presence of maternal mental illness.³² Controlling for the aforementioned variables would lead to collider stratification bias in the model.³³

An a priori decision was made to evaluate the relationship between economic vulnerability and childhood neurodevelopmental outcomes by markers of social capital — defined by marital status (i.e., single or divorced vs. married or living with partner) and maternal educational level (i.e., having completed less than a high school education vs. completed high school or more education). The aforementioned variables were chosen based on a review of the literature of relative deprivation indices, as well as social theory, rooted in Bourdieu’s theories of capital, that illustrates the relationship of social capital, defined as certain reciprocal or obligatory connections between individuals that can be used for economic or cultural benefit, to health outcomes.^7,20 We hypothesized that children born to individuals with higher social capital (e.g., married or with higher levels of education) would be less likely to demonstrate adverse neurodevelopment in the context of economic vulnerability. A post-hoc sensitivity analysis was performed for economically-vulnerable participants who had a household FPL of <100%, as this population would exhibit the highest economic needs and would be eligible for multiple forms of state-sponsored economic assistance programs.²³

Bivariate analyses were performed using chi-square, Fisher’s exact, and Wilcoxon rank-sum tests, as appropriate. Multivariable quantile regression models were used to evaluate the associations between economic vulnerability and child neurodevelopmental outcomes. To examine possible interactions between economic vulnerability and education status and between economic vulnerability and marital status, quantile regression models that included interaction terms, along with the potential confounders described above, were generated. A quantile regression model was chosen in order to estimate the adjusted median outcomes from the exposure and covariates, rather than the adjusted mean that is generated using linear regression.³⁴ Regression models were then run separately based on education status or marital status if a significant interaction was found. An interaction analysis assessed whether a given exposure’s association with an outcome varies by other characteristics (e.g., maternal education, marital status). An interaction that is significant implies that the exposure’s strength of association with the outcome differs in magnitude for different strata.

Statistical significance was defined as p<0.05, and all tests were two-tailed. No imputation for missing data was performed. All reported analyses and outcomes adhere to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.³⁵ All statistical analyses were performed with SAS (version 9.4).

RESULTS

A total of 955 participants met inclusion criteria, of whom 406 (42.5%) were classified as economically vulnerable (Figure 1). Univariable analyses revealed economically vulnerable participants to be younger (27.2±5.6 vs. 28.8±5.5 years, p<0.001), more likely to be single or divorced (28.3% vs. 20.4%, p=0.005), be of Hispanic ethnicity (70.7% vs. 35.0%, p<0.001), have completed a high school education or less (82.8% vs. 39.0%, p<0.001), have a larger household size (p<0.001), have used tobacco during pregnancy (9.9% vs. 5.3%, p=0.007), less likely to consume alcohol during pregnancy (3.0% vs 8.2%, p<0.001) and to rent or live with parents (83.0% vs. 52.5%, p<0.001) (Table 1).

Figure 1.

A flow chart demonstrating the final study population based on the initial study sample from the parent trial. Of 955 patients, 645 had an income <200% at the federal poverty level, 525 were unemployed or had part-time work, and 406 had government health insurance.

Table 1.

Demographics

	Economically vulnerable, n=406	Not economically vulnerable, n=549	p-valuea
Maternal characteristic
Maternal age, in years	27.2±5.6	28.8±5.5	<0.001
Single or divorced	115 (28.3)	112 (20.4)	0.005
Race-ethnicity			<0.001
Hispanic	287(70.7)	192 (35.0)
Non-Hispanic White	35 (8.6)	284 (51.7)
Non-Hispanic Black	76 (18.7)	60 (10.9)
Otherb	8 (2.0)	13 (2.4)
High school education or less	336 (82.8)	214 (39.0)	<0.001
Household size			<0.001
2	43 (10.6)	45 (8.2)
3	79 (19.5)	211 (38.4)
4	116 (28.6)_	141 (25.7)
5	99 (24.4)	92 (16.8)
6+	69 (17.0)	60 (10.9)
Smoking during pregnancy	40 (9.9)	29 (5.3)	0.007
Alcohol use during pregnancy	12 (3.0)	45 (8.2)	<0.001
Illicit drug use during pregnancy	4 (1.0)	6 (1.1)	1.0
Rent or live with parents	337 (83.0)	288 (52.5)	<0.001
Delivery less than 37 weeks	39 (9.6)	40 (7.3)	0.20
Trial arm
Subclinical hypothyroidism	219 (53.9)	334 (60.8)	0.03
Hypothyroxinemia	187 (46.1)	215 (39.2)

Data are mean ± SD or n (%) unless otherwise specified. Bold indicates a p-value of less than 0.05

^aChi-squared or Fisher’s exact test for categorical variables, Wilcoxon rank-sum test for continuous variables

^b“Other” race-ethnicity was a prespecified formal category used during the parent trials. This included 13 individuals who self-identified as Asian, 1 individual who self-identified as American Indian/Alaskan Native, and 7 who identified as “Other”

Bivariate analyses demonstrated that median WPPSI-III score and all Bayley-III subtest scores were significantly lower for the children of economically vulnerable participants (Table 2). Median scores for the WPPSI-III were 91 (interquartile range (IQR): 82-98) and 99 (IQR: 89-111) for economically vulnerable versus non-economically vulnerable groups, respectively (p<0.001). Median and IQR scores for Bayley-III tests were 85 (IQR: 80-90) and 95 (IQR: 85-100) for the cognitive test, 94 (IQR: 88-100) and 100 (IQR: 94-107) for the motor test, and 86 (IQR: 79-91) and 94 (IQR: 86-106) for the language test for economically vulnerable versus non-economically vulnerable groups, respectively (all p<0.001).

Table 2.

Association between economic vulnerability and childhood neurodevelopmental outcomes

Name of test	Test score		Difference in scores(95% CI)	p-value
	Economically vulnerable, n=406	Not economically vulnerable, n=549
WPPSI-III	91 (82–98)	99 (89–111)	−8.0 (−9.9 to −6.1)	<0.001
Bayley-III Cognitive	85 (80–90)	95 (85–100)	−10.0 (−12.7 to −7.3)	<0.001
Bayley-III Motor	94 (88–100)	100 (94–107)	−6.0 (−8.6 to −3.4)	<0.001
Bayley-III Language	86 (79–91)	94 (86–106)	−8.0 (−8.7 to −7.3)	<0.001

Data are shown as median (interquartile range). Bold indicates a p-value of less than 0.05

We found significant interactions between economic vulnerability and educational attainment, and well as between economic vulnerability and marital status, with regard to WPPSI-III scores (p for interaction = 0.001 and 0.008, respectively). Specifically, economic vulnerability was associated with a significantly lower WPPSI-III score only among those whose mother had at least a high school education (96 vs. 105, adjusted difference: −4.4, 95% CI: −8.2 to −0.59, p = 0.024). Also, economic vulnerability was associated with a significantly lower score only among those whose mother was married (90 vs. 102, adjusted difference in median scores: −4.0, 95% CI: −6.9 to −1.1, p = 0.007). Among those who either did not complete high school or those who were unmarried or unpartnered, WPPSI-III scores were similar between economically vulnerable and non-economically vulnerable groups (Table 3).

Table 3.

Subgroup analysis of economic vulnerability and childhood neurodevelopmental outcomes by markers of social capital

	Test score		Adjusted difference in scores (95% CI)a
	Economically vulnerable, n=406	Not economically vulnerable, n=549
WPPSI-III a
Greater than HS educationb
Yes (n=405)	96 (85–104)	105 (98–115)	-4.4 (−8.2 to −0.59)
No (n=550)	90 (82–97)	90 (81–99)	-0.24 (−2.9 to 2.4)
Married or partneredc
Yes (n=728)	90 (82–99)	102 (93–113)	-4.0 (−6.9 to −1.1)
No (n=227)	91 (82–96)	90.5 (81–99)	1.5 (−1.7 to 4.8)
Bayley-III Cognitive d
Greater than HS educationb
Yes (n=405)	90 (85–95)	100 (90–105)	-5.0 (−9.2 to −0.7)
No (n=550)	85 (80–90)	85 (80–90)	0 (−1.1 to 1.1)
Married or partneredc
Yes (n=728)	85 (80–90)	95 (85–105)	0 (−1.0 to 1.0)
No (n=227)	85 (80–95)	90 (80–95)	0.71 (−2.16 to 3.59)
Bayley-III Language e
Greater than HS educationb
Yes (396)	90 (83–97)	100 (94–112)	-4.6 (−8.5 to −0.67)
No (540)	86 (77–91)	86 (79–94)	-3.0 (−5.1 to −0.85)
Married or partneredc
Yes (n=714)	86 (77–91)	97 (89–109)	-3.4 (−5.8 to −1.09)
No (n=222)	89 (79–94)	89 (79–97)	0.37 (−3.4 to 4.1)

Data are shown as median (interquartile range). HS = high school

^aThere was a significant interaction between marital status and economic vulnerability (p=0.008) and between education and economic vulnerability (p=0.001) for the WPPSI-III neurodevelopmental outcome; therefore, separate regression models were performed

^bAdjusted for maternal age, maternal race/ethnicity, marital status, substance use during pregnancy, trial center, levothyroxine or placebo treatment group, baseline thyroid status, and maternal housing status

^cAdjusted for maternal age, maternal race/ethnicity, maternal educational level, substance use during pregnancy, trial center, levothyroxine or placebo treatment group, baseline thyroid status, and maternal housing status

^dThere was a significant interaction between education status and economic vulnerability (p<0.001) and between education and economic vulnerability (p<0.001) for the Bayley-III cognitive neurodevelopmental outcome, therefore, separate regression models were performed

^eThere was a significant interaction between marital status and economic vulnerability (p<0.001) and between education and economic vulnerability (p<0.001) for the Bayley-III neurodevelopmental outcome, therefore, separate regression models were performed

We also noted significant interactions between economic vulnerability and markers of social capital with both Bayley-III cognitive (p <0.001 for both marital status and educational attainment) and language scores (p < 0.001 for both marital status and educational attainment) (Table 3). There were no significant interactions between economic vulnerability and markers of social capital with Bayley-III motor scores.

For our sensitivity analysis, restricting the FPL cutoff to <100%, a total of 314 (32.9%) were classified as economically vulnerable (Table 4). Similar to the findings using the FPL <200% criteria, univariable analyses revealed economically vulnerable participants to be younger (27.3±5.6 vs. 28.5±5.6 years, p<0.001), more likely to be single or divorced (29.6% vs. 20.9%, p=0.003), less likely to be non-Hispanic White (8.3% v. 45.7%, p < 0.001), have completed a high school education or less (81.8% vs. 45.7%, p<0.001), live in a smaller household size (p<0.001), have used tobacco during pregnancy (11.1% vs. 5.3%, p=0.001), have consumed less alcohol during pregnancy (3.5% vs. 7.2%, p=0.024), and to rent or live with parents (84.1% vs. 56.3%, p<0.001). Additionally, economically-vulnerable individuals, using the FPL <100% criteria, were more likely to be in the hypothyroxinemia trial arm (48.7% vs 38.8%, p=0.004) (Table 4).

Table 4.

Demographics, when <100% crude FPL cutoff instituted in definition of economic vulnerability

	Economically vulnerable, n=314	Not economically vulnerable, n=641	p-valuea
Maternal characteristic
Maternal age	27.3±5.6	28.5±5.6	<0.001
Single or divorced	93 (29.6)	134 (20.9)	0.003
Race-ethnicity			<0.001
Hispanic	214 (68.2)	265 (41.3)
Non-Hispanic White	26 (8.3)	293 (45.7)
Non-Hispanic Black	67 (21.3)	69 (10.8)
Otherb	7 (2.2)	14 (2.2)
High school education or less	257 (81.8)	293 (45.7)	<0.001
Household size			<0.001
2	29 (9.2)	59 (9.2)
3	61 (19.4)	229 (35.7)
4	95 (30.3)_	162 (25.3)
5	71 (22.6)	120 (18.7)
6+	58 (18.5)	71 (11.1)
Smoking during pregnancy	35 (11.1)	34 (5.3)	0.001
Alcohol use during pregnancy	11 (3.5)	46 (7.2)	0.024
Illicit drug use during pregnancy	3 (1.0)	7 (1.1)	1.0
Rent or live with parents	264 (84.1)	361 (56.3)	<0.001
Delivery less than 37 weeks	28 (8.9)	51 (8.0)	0.61
Trial arm
Subclinical hypothyroidism	161 (51.3)	392 (61.2)	0.004
Hypothyroxinemia	153 (48.7)	249 (38.8)

Data are mean ± SD or n (%) unless otherwise specified. Bold indicates a p-value of less than 0.05

FPL = federal poverty line

^aChi-squared or Fisher’s exact test for categorical variables, Wilcoxon rank-sum test for continuous variables

^b“Other” race-ethnicity was a prespecified formal category used during the parent trials. This included 13 individuals who self-identified as Asian, 1 individual who self-identified as American Indian/Alaskan Native, and 7 who identified as “Other”

Bivariate analyses demonstrated that median WPPSI-III score and all Bayley-III subtest scores were significantly lower for the children of economically vulnerable participants, similar to the pattern described in the main analysis (Table 5).

Table 5.

Association between economic vulnerability and childhood neurodevelopmental outcomes, when <100% crude FPL cutoff instituted in definition of economic vulnerability

Name of test	Test score		Difference in scores (95% CI)	p-value
	Economically vulnerable, n=314	Not economically vulnerable, n=641
WPPSI-III	90 (82–97)	99 (88–110)	-9.0 (−10.8 to −7.2)	<0.001
Bayley-III Cognitive	85 (80–90)	90 (85–100)	-5.0 (−9.8 to −0.15)	0.04
Bayley-III Motor	94 (88–100)	97 (91–107)	-3.0 (−4.8 to −1.2)	0.001
Bayley-III Language	86 (79–91)	94 (83–106)	-8.0 (−10.2 to −5.8)	<0.001

Data are shown as median (interquartile range). Bold indicates a p-value of less than 0.05

We found significant interactions between economic vulnerability and educational attainment with the WPPSI-III (p = 0.014), Bayley-III cognitive (p < 0.001), and Bayley-III language scores (p <0.001). We found significant interactions between economic vulnerability and marital status with the Bayley-III motor (p < 0.001) and the Bayley-III language (p < 0.001) scores (Table 6).

Table 6.

Subgroup analysis of economic vulnerability and childhood neurodevelopmental outcomes by markers of social capital, when <100% crude FPL cutoff instituted in definition of economic vulnerability

	Test score		Adjusted difference in scores (95% CI)a
	Economically vulnerable, n=314	Not economically vulnerable, n=641
WPPSI-III a
Greater than HS educationb
Yes (n=405)	94 (84–104)	105 (98–115)	−4.3 (−8.9 to 0.35)
No (n=550)	89 (82–96)	91 (83–99)	−2.5 (−5.1 to 0.15)
Bayley-III Cognitive c
Greater than HS educationb
Yes (n=405)	90 (85–95)	100 (90–105)	−5.0 (−9.2 to −0.77)
No (n=550)	85 (80–90)	85 (80–90)	0 (−1.3 to 1.3)
Married or partneredd
Yes (n=728)	85 (80–90)	95 (85–105)	0 (−1.2 to 1.2)
No (n=227)	90 (80–95)	85 (80–95)	0.56 (−2.6 to 3.7)
Bayley-III Motor e
Married or partneredd
Yes (n=722)	94 (85–100)	100 (94–107)	−3.0 (−4.9 to −1.09)
No (n=221)	97 (91–103)	94 (91–100)	−1.1 (−5.4 to 3.2)
Bayley-III Language f
Greater than HS educationb
Yes (n=396)	89 (83–94)	100 (94–112)	−4.5 (−8.8 to −0.14)
No (n=540)	86 (77–91)	86 (79–92.5)	−2.9 (−5.3 to −0.5)
Married or partneredd
Yes (n=714)	86 (77–91)	94 (86–106)	−4.1 (−6.6 to −1.6)
No (n=222)	89 (79–94)	89 (79–94)	−0.12 (−4.0 to 3.8)

Data are shown as median (interquartile range). HS = high school

^aThere was a significant interaction between education status and economic vulnerability (p=0.014) for the WPPSI-III neurodevelopmental outcome; therefore, separate regression models were performed

^bAdjusted for maternal age, maternal race/ethnicity, marital status, substance use during pregnancy, trial center, levothyroxine or placebo treatment group, baseline thyroid status, and maternal housing status

^cThere was a significant interaction between education status and economic vulnerability (p<0.001) and between marital status and economic vulnerability (p<0.001) for the Bayley-III cognitive neurodevelopmental outcome therefore, separate regression models were performed

^dAdjusted for maternal age, maternal race/ethnicity, maternal educational level, substance use during pregnancy, trial center, levothyroxine or placebo treatment group, baseline thyroid status, and maternal housing status

^eThere was a significant interaction between marital status and economic vulnerability for the Bayley-III motor neurodevelopmental outcome (p<0.001) therefore, separate regression models were performed

^fThere was a significant interaction between education status and economic vulnerability (p<0.001) and between marital status and economic vulnerability (p<0.001) for the Bayley-III language neurodevelopmental outcome, therefore, separate regression models were performed

DISCUSSION

Those who are economically vulnerable during pregnancy have children with lower childhood neurodevelopmental scores 2 and 5 years after birth. In contrast to our initial hypothesis, for most of these measures — specifically WPPSI-III, Bayley-III cognitive subtest, and Bayley-III language subtest scores— this association was most evident among economically-vulnerable participants with markers of high social capital, such as marriage or educational attainment. This dynamic appears to be the case because people with markers of lower social capital have children with lower scores such that economic vulnerability is not multiplicative.

Relative deprivation indices were pioneered by British sociologist Peter Townsend in order to counter the prevailing theory of “absolute poverty”, which neglected the social and historical networks to which individuals belonged that afforded them differential access to resources.⁷ Since that time, researchers in the United Kingdom and France have utilized deprivation indices to assess whether there exists a relationship of sociopolitical and economic marginalization with differences in perinatal care.^9-11 However, these studies did not evaluate whether deprivation indices were associated with long-term consequences for offspring.

Relating markers of social capital, such as the completion of a high school education, to adverse pregnancy outcomes has become a research focus in understanding reproductive health disparities. Data from New York City between 2008 and 2012 demonstrated that higher educational attainment was associated with lower rates of severe maternal morbidity (283.9/10000 deliveries among those with less than a high school education vs. 164.5/10000 deliveries among those with at least a college degree).³⁶ However, our data demonstrate worse neurodevelopmental outcomes among children who are born to economically-vulnerable individuals with high social capital, such as those with high school education or higher, when compared to their non-economically-vulnerable counterparts. Therefore, our analysis suggests a unique interaction between economic vulnerability and social capital that should be considered in further analyses of sociodemographic markers associated with adverse pregnancy outcomes.

Researchers have previously derived unique deprivation indices, which have focused primarily on type of health insurance, housing status, partner status, and household income, that correlate with psychosocial vulnerability of pregnant individuals outside of the U.S.²⁵ The presence of this association among those in the U.S. as well, along with the interaction between markers of social capital and economic vulnerability in regard to neurodevelopmental test scores, points to the importance of elucidating potential biosocial pathways in which economic vulnerability and social capital interact to affect postnatal neurodevelopment.³⁷ While a major focus of pediatric research has been on exposure to childhood poverty and adverse neurodevelopmental outcomes, our data demonstrates the need to attend to the antepartum period as a key site that may influence adverse childhood neurodevelopmental outcomes.^12,13 Hypothesized pathophysiological mechanisms that could describe our findings could include, but are not limited to, in utero epigenetic changes in fetal neuronal development pathways, preterm birth, presence of neonatal neurological damage (e.g., neonatal encephalopathy)^38,39, unequal access to early intervention resources, and postnatal social and environmental factors that could influence development and behavior.^12,18,19 Understanding these relationships is particularly important amidst the growing economic and sociopolitical inequities in the U.S.⁴⁰ Not only would such analyses contribute to the ongoing focus on disparities in reproductive healthcare quality and outcomes, but they would also add a novel epidemiological focus to the economic and social underpinnings of inequitable perinatal and childhood health outcomes.

This study has important strengths, including a large study population derived from multicenter, randomized controlled trials. The outcomes were rigorously ascertained though the use of validated neurodevelopmental tests administered by trained personnel. Also, this analysis demonstrates the applicability of economic vulnerability and social capital to the U.S. perinatal context to further understand the association of socioeconomic inequity with adverse childhood outcomes. Furthermore, economic vulnerability has been demonstrated to be a relatively stable exposure throughout the antenatal and postpartum period. Household income has been noted to decrease immediately after a birth of a child, with subsequent stabilization to antenatal values over the corresponding year.⁴¹ Sociological and economic data from the last fifty years demonstrates that for individuals who are socioeconomically and historically marginalized, the prospect of meaningful improvement in economic security (e.g., rise in household income) is limited, particularly given the enactment of retrenchment policies throughout the 1980s and 1990s.^42,43

Potential limitations of this study include our definition of economic vulnerability, which was derived from a review of the literature with data derived from a non-U.S. context. Certain aspects of our exposure definition, such as utilizing an estimated FPL, may reduce the accuracy of identifying those at risk of economic vulnerability. Furthermore, we did not incorporate renting or living with parents as a marker of social capital given the inability to stratify whether a given individual was renting or living with parents. Nevertheless, any bias introduced into the analysis from these issues should bias toward the null, and thus our finding of positive association in both our primary and sensitivity analysis is, if anything, likely to be underestimated in magnitude. The study population was derived from pregnant people diagnosed with thyroid dysfunction who chose to participate in a randomized controlled trial; therefore, the generalizability of the study is limited. We also experienced incomplete follow-up for almost 17% of the initial trial cohort due to loss to follow up, lack of neurodevelopmental outcome data, or lack of maternal income data. While the interaction analysis demonstrated that the strength of the association of child neurodevelopmental scores with economic vulnerability and with marital status differs depending upon the degree of social capital, the 95% confidence interval for some scores, which crosses 0 at the level of the substrata, may be related to the smaller sample size within the substrata. Nevertheless, our conclusion for these findings demonstrates a differential effect of economic vulnerability among individuals with markers for high and low social capital as it relates to childhood neurodevelopmental scores. This study is unable to assess causality, as it is only an observational study. Despite our rigorous statistical methodology, there is always the possibility of residual confounding that may explain the relationship between economic vulnerability and adverse childhood neurodevelopmental outcomes, as has previously been demonstrated in other literature focused on breastfeeding and childhood neurodevelopment.^44,45 Finally, we assume that the components of the composite exposures of economic vulnerability and social capital contribute equally to the exposure itself (e.g., FPL, employment, and insurance all equally contribute to economic vulnerability); our analysis does not specifically interrogate this assumption. Further research should focus on these individual attributes and which of these variables best predicts adverse childhood neurodevelopmental outcomes.

Among offspring of economically-vulnerable pregnant people who were willing to participate in a clinical trial, certain childhood neurodevelopmental outcomes are worse at 2 and 5 years of life when compared with the non-economically vulnerable, particularly among those with markers of high social capital.