Mid‐pregnancy uterine artery Dopplers and child neurodevelopment

Abstract

Objective

Resistance to blood flow in the maternal uterine artery (UtA) has been associated with both preeclampsia (PE) and small‐for‐gestational age (SGA). Our objective was to assess the association of increased maternal UtA resistance with child neurodevelopmental and behavioral outcomes in low‐risk pregnancies.

Study Design

Serial ultrasound exams were prospectively performed in 2334 racial and ethnically diverse, healthy patients with low‐risk singleton pregnancies (July 2009 to Jan 2013) as part of the NICHD‐Fetal Growth Studies. Maternal UtA Dopplers were performed between 18w0d and 25w6d and pregnancy outcomes were collected at delivery including the occurrence of PE and SGA birthweight. Mother–child pairs from the original study were re‐recruited to the NIH‐Environmental Influences on Child Health Outcomes (ECHO) Study (May 2017 to April 2019). Analysis of neurodevelopment was performed using age‐appropriate versions of the NIH Toolbox for children including executive (inhibitory control/attention, cognitive flexibility, and working memory) and motor domains (dexterity, grip strength, and standing balance). Assessment of behavioral problems was obtained by parent report using the Child Behavioral Checklist. Multivariable adjusted general linear and logistic regression models were used to determine the mean difference in standardized test scores of observed child executive and motor functions and parent‐reported child behaviors associated with a 1‐standard deviation (SD) increment in UtA pulsatility index (PI).

Results

A total of 675 patients had both interrogation of mid‐pregnancy UtA Dopplers and at least one measure of executive or motor domain function or parent‐reported behavior between the ages of 4 and 8 years. Each 1‐SD increase in UtA PI was associated with a significant reduction in cognitive flexibility in children aged 4–7 years (n = 457; adjusted mean difference[aMD] −0.97; 95% CI −1.8, −0.14). There were no associations between motor function and increased UtA resistance. A 1‐SD increase in UtA PI was not associated with any increased parent report of child behavioral problems. A 1‐SD increase in UtA PI was associated with a 1.5‐fold increased risk of PE. In mediation analyses, differences in executive function domains associated with increased UtA PI Doppler findings were not explained by either maternal PE, SGA, or birthweight z‐scores.

Conclusion

Increased maternal mid‐pregnancy UtA resistance appears to be associated with small decreases in executive domain function (cognitive flexibility) but not motor domain function. In this exploratory study, the observed effect sizes were modest and the clinical implications are uncertain, but they suggest the possibility of long‐term neurodevelopmental effects of subclinical placental dysfunction in low‐risk pregnancies. The findings, however, are biologically plausible and contribute to a growing body of literature exploring antenatal predictors of child neurodevelopment. Larger confirmatory studies will be needed to determine if maternal UtA resistance can contribute to the early identification of at‐risk children.

Trial Registration

Clinicaltrials.gov identifiers: NCT00912132, NCT03266198

1. INTRODUCTION

In the United States, approximately 1 in 6 children demonstrate some degree of developmental disability or delay with many of these conditions having their origins in the antenatal period [1, 2]. In healthy pregnancies, placental trophoblastic tissue invades the maternal myometrium, remodeling the vasculature, and enhancing oxygen delivery to the fetus [3]. Remodeling requires obliteration of the musculoelastic layer of the maternal spiral arteries, resulting in a low‐resistance placental circulation [4]. Inadequate trophoblastic invasion results in increased impedance to utero‐placental blood flow [5]. These pathologic findings have been identified in pregnancies complicated by fetal growth restriction (FGR) and preeclampsia (PE) and linked to poorer neurodevelopmental outcomes [3, 6, 7, 8, 9].

Increased resistance to utero‐placental blood flow can be detected antenatally by evaluating Doppler flow patterns in the maternal uterine artery (UtA). Normally, there is a continuous forward Doppler waveform throughout the cardiac cycle beginning as early as 8–10 weeks’ gestation [3]. This constant forward waveform is reflected by a low resistance Doppler pattern and quantified as a low pulsatility index (PI). With increased vascular resistance, there is reduced forward flow during late systole and early diastole, resulting in an elevated UtA PI [3]. There is a strong association between increased UtA PI measurements and evidence of early‐onset dysfunctional trophoblastic and placental development [10, 11, 12].

UtA Doppler findings consistent with increased utero‐placental resistance have been used to screen for pregnancy complications, such as FGR and PE, and have been associated with adverse perinatal outcomes [13, 14, 15, 16]. Notably, most of these studies were performed in cohorts at increased risk for these complications and subsequent childhood neurodevelopment has not been well characterized [2, 6, 7, 17]. The objective of this study was to assess the association between mid‐pregnancy UtA PI in low‐risk gestations with child neurodevelopmental and behavioral outcomes. Our hypothesis is that small increases in mid‐trimester UtA resistance may result in reductions of oxygenated blood flow to the anterior regions of cerebral development, primarily impacting child executive function domains and secondarily, motor domains and childhood behavior.

2. METHODS

The study population was drawn from a cohort originally recruited to the National Institute of Child Health and Human Development Fetal Growth Study (NICHD‐FGS). The NICHD‐FGS enrolled 2802 patients aged 18–40 years with uncomplicated pregnancies to evaluate fetal growth and associated perinatal outcomes between July 2009 and January 2013 at 12 sites in the United States [18].

Patients were recruited to the NICHD‐FGS between 8 and 13 weeks’ gestation with singleton, non‐anomalous, low‐risk pregnancies. Pregnancies complicated by substance use, assisted reproductive technologies, chronic medical conditions, or previous obstetric complications were excluded. A total of 2585 patients were followed through delivery. Data collection included maternal and antenatal variables, pregnancy complications including PE, and birth outcomes. Serial sonographic measures were obtained, including fetal biometry, umbilical and UtA Doppler findings [19]. The stratification of birthweight as small, appropriate, or large for gestational age is based on the gestational age at delivery and the birthweight centiles (< 10th; 10–90th; and > 90th) provided by Duryea et al. [20].

Children delivered in the NICHD‐FGS were reassessed between the ages of 4 and 8 years as part of the Environmental Influences on Child Health Outcomes (ECHO) Study. The ECHO cohort consisted of 1116 mother–child dyads recruited between May 2017 and April 2019 from 10 of the original NICHD‐FGS sites. Each site attempted to contact parents by phone, e‐mail, or text. Mother–child dyads were invited to return to the original study site for assessment. Those unable to return were offered participation via telephone/computer survey or home visits. A total of 730 children had in‐person visits with collection of interval history, physical examination, standardized observations of executive and motor domains, and parental reports of behavior. Of these mother–child pairs, 675 had Doppler measures of UtA PI and at least one follow‐up neurodevelopmental assessment; comprising our analytic cohort (Figure 1). A comparison of sociodemographic and medical variables for this analytic cohort with the originally recruited NICHD‐FGS cohort is provided (Table S1).

FIGURE 1.

Study population flowchart.

Written informed consent was obtained from the parent or legal guardian and assent of the child depending on age and state regulations. The study was approved by the Medical University of South Carolina Institutional Review Board (IRB) and a central IRB at Columbia University Medical Center.

Neurodevelopmental outcomes were assessed using the National Institute of Health (NIH) Toolbox which included standardized observations of executive motor domains [21]. The NIH Toolbox includes three reliable tests for executive domain function in prepubertal children and these were considered the primary outcome. These tests included measurements of inhibitory control and sustained attention (Flanker Test), cognitive flexibility (Dimensional Change Card Sorting Test), and working memory (List Sorting Memory Test). These three tests are the most common measures used to evaluate executive domain function in research, especially for pediatric populations [22, 23, 24]. Each test of executive domain function consisted of three variations: the first administered to all children and the other two administered to age‐specific subgroups. Each test was thoroughly explained by a member of the study team. For motor domain function, the study team conducted three standardized tests administered to all children, including the Nine Hole Pegboard Dexterity Test (dominant hand), the Grip Strength Test (non‐dominant hand) for forearm muscle strength, and the Standing Balance Test for postural control [25, 26, 27]. Visual tests were administered on an iPad with assistance from a trained member of the study team. Both written and verbal instructions were given to each child. Raw scores were converted to normalized t scores. Overall scores were analyzed as continuous variables with higher scores indicating greater functioning. Each NIH Toolbox Battery test administered to the child during the study visit is briefly described in Table S2.

For assessment of behavioral problems, parents were asked to complete a Child Behavioral Checklist (CBCL) evaluating internalizing and externalizing problems, attention deficits, oppositional defiance, depression, and anxiety [28]. Internalizing problems (symptoms of depression, anxiety, and social withdrawal) are associated with substance use, school drop‐out, and suicide in adolescents [29]. Externalizing problems (aggressive and delinquent social behaviors, hyperactivity, and attention deficits) are associated with poor self‐regulation and maladjustment [30, 31]. For the CBCL, outcome scores were averaged as an overall score and assessed continuously; higher scores indicating greater behavioral problems [32].

Maternal UtA systolic and diastolic flow and the PI measurement were obtained between 18 0/7 and 25 6/7 weeks of gestation for each patient [18]. Bilateral UtA PI measurements were obtained three times and measurements were averaged. All sonographers were trained and credentialed at a single center (Columbia University), with measurements subject to post hoc quality assurance [33]. The UtA PI is calculated as the systolic flow, minus the diastolic flow, divided by the mean. While the assessment of the association between UtA PI exposure and outcomes was continuous, we selected a greater than 1‐SD increase in UtA PI as our primary exposure. We selected a 1‐SD increment in UtA PI because it was a relatively small increase, and our objective was to assess the impact of a more subtle rise in UtA PI that would generally fall within the normal range in a pre‐selected low‐risk population. Covariates included maternal age at NICHD‐FGS enrollment, maternal self‐reported race and ethnicity, pre‐gravid BMI, gestational age at delivery, parity (0 vs. 1+), education, marital status, child's age at ECHO study visit, gender, and birthweight z‐score. Prenatal stress and depressive symptoms were also covariates given previously published evidence of an association with childhood neurodevelopmental outcomes in this cohort [34]. Prenatal stress and depression were assessed by maternal report at six time points during pregnancy using the Perceived Stress Scale (PSS) and the Edinburgh Postpartum Depression Scale (EPDS), respectively [18, 33, 35, 36]. For both, higher scores indicated greater perceived stress or depressive symptoms.

Co‐variates were summarized using medians (ranges) for continuous variables and counts (percentages) for categorical variables. The association between UtA PI Doppler parameters and neurodevelopmental and behavioral outcomes was assessed as a continuous exposure with general linear and logistic regression models. Regression analyses were conducted on each dependent neurodevelopmental or behavioral standardized score against the UtA PI Doppler findings. We reported the adjusted mean differences (aMD) and 95% confidence intervals (95% CI) for each standard deviation increase in the continuous UtA PI measure in both the crude and multivariable adjusted models. Stratification by child age is also reported for executive domain functioning where age‐specific testing was used. Logistic regression was performed to calculate odds ratios (OR) with 95% CI to evaluate the association between a 1‐SD increase in UtA PI with PE risk.

The potential mediation of our findings by PE, SGA, or birthweight z‐score was considered, and those results are reported in a supplemental mediation analysis due to their potential effects on neurodevelopmental outcomes. Birthweight z‐scores were calculated according to the birthweight percentiles provided by Villar and colleagues using Intergrowth‐21 data according to gestational age and sex [37]. The impact of breastfeeding for at least six months was also considered in a sensitivity analysis [38]. Breastfeeding status was self‐reported and coded as either “yes” or “no,” and duration (in weeks) was assessed as a continuous variable. A second sensitivity analysis was performed repeating all analyses with log‐transformed UtA PI Doppler parameters to ensure that the findings are robust and not driven by skewness of the distribution of the predictor.

All analyses were performed using R version 4.4.0. A two‐sided p value ≤ 0.05 was considered statistically significant. The sample sizes ranged between 423 and 540 children for the full sample. In performing approximately 40 tests, our minimum detectable effect size ranges from Cohen's f ² = 0.31 to 0.04 to have a power of 80% at a 0.05 significance level. Traditionally, Cohen's f ² = 0.02 and 0.15 are a small and a medium effect size, respectively, indicating that we are likely to have sufficient power to detect a medium effect size association.

3. RESULTS

The sample characteristics are presented in Table 1 (n = 675), with broad racial and ethnic diversity. Most patients were married or living with their partners (73%), with 73.8% having at least some college education. Approximately 45.9% of patients were primigravid, with an average gestational age at delivery of 39.29 weeks and birthweight of 3348 g (1240–4890 g). PE occurred in 17 patients (2.5%). The average maternal age at enrollment was 29 (18–40) years. Of children evaluated, 51.7% were biologically male with an average age of 6.9 (4.73–8.87) years at follow‐up.

TABLE 1.

Sample characteristics (n = 675).

Variables (N = 675)	Median (range)	Frequency (%)
Maternal age at enrollment, years	29.00 (18.00–40.00)
Maternal race and ethnicity
Non‐Hispanic White		221 (32.7%)
Non‐Hispanic Black		216 (32.0%)
Hispanic		155 (23.0%)
Asian & Pacific Islander		83 (12.3%)
Pre‐gravid BMI (kg/m2)	24.20 (19.00, 44.90)
Marital status
married/not married but living with a partner		493 (73.0%)
widowed/divorced/separated		23 (3.4%)
never been married		159 (23.6%)
Maternal education
Less than high school		71 (10.5%)
High school diploma or GED or equivalent		106 (15.7%)
Some college or associate degree		193 (28.6%)
Bachelor's degree		173 (25.6%)
Master's degree or Advanced degree		132 (19.6%)
Maternal Parity
0		310 (45.9%)
1		231 (34.2%)
2		94 (13.9%)
3		32 (4.7%)
4		6 (0.9%)
5		2 (0.3%)
Preeclampsia		17 (2.5%)
PSS averaged	9.20 (0.00, 26.25)
EPDS averaged	3.83 (0.00, 19.80)
Gestational age at delivery, weeks	39.29 (30.43, 42.71)
Birth weight, g	3348 (1240, 4890)
Child biological sex
Female		326 (48.3%)
Male		349 (51.7%)
Child age at ECHO visit, years	6.94 (4.73, 8.87)

The mean UtA PI between 18 and 25 weeks’ gestation was 0.94 (SD = 0.4). The trend in mean PI reduction between 18 and 25 weeks is shown in Figure 2.

FIGURE 2.

Mean uterine artery PI by gestational age. Mean uterine artery PI between 18 and 25 weeks’ gestation was 0.94 ± SD 0.4.

We found no association between mid‐pregnancy UtA PI measurements and child executive domain functioning among all tested children in the cohort (Table 2). However, a 1‐SD increase in UtA PI was associated with a significant reduction in cognitive flexibility in children aged 4–7 years (aMD −0.97;95% CI −1.8, −0.14). A small subset of children at age 8 (n = 78) demonstrated a significantly higher mean score for inhibitory control/sustained attention (aMD 1.95; 95% CI 0.35, 3.55) with a 1‐SD increase in UtA PI. There was no effect of an elevated UtA PI on inhibitory control/sustained attention among the 462 children aged 4–7 years or all children overall.

TABLE 2.

Associations between UtA PI and child executive functions (NIH Toolbox). All models were adjusted for maternal age at NICHD‐FGS enrollment, maternal self‐reported race and ethnicity, pre‐gravid BMI, parity, education, marital status, child's age at ECHO study visit, gender, birth weight z‐score, and summated prenatal Edinburgh Depression Scale and Prenatal Stress Score.

			Uterine artery PI changes
Executive functions domain	n	Mean score (SD)	* 1 SD increase in PI mean difference (95% CI)
Inhibitory control/sustained attention (all ages)	540	48.18 (8.77)	0.05 (−0.69, 0.8)
Inhibitory control/sustained attention (ages 4–7)	462	48.51 (8.89)	−0.29 (−1.14, 0.56)
Inhibitory control/sustained Attention (age 8)	78	46.18 (7.68)	1.95 (0.35, 3.55) *
Cognitive flexibility (all ages)	535	48.12 (8.42)	−0.46 (−1.19, 0.28)
Cognitive flexibility (ages 4–7)	457	48.24 (8.45)	−0.97 (−1.8, −0.14) *
Cognitive flexibility (age 8)	78	47.39 (8.23)	1.31 (−0.27, 2.9)
Working memory (all ages)	482	48.55 (10.17)	−0.58 (−1.49, 0.34)
Working memory (ages 4–6)	230	46.78 (10.05)	−1.37 (−2.77, 0.04)
Working memory (ages 7–8)	252	50.24 (10.01)	−0.02 (−1.28, 1.24)

^* p < 0.05.

Due to concern for selection bias associated with differing sociodemographic profiles or other important covariates between the included analytic cohort and the original NICHD‐FGS cohort from which they arose, we calculated propensity weights and applied inverse propensity weights to adjust for this potential bias (Table 3). Results remained similar to the primary outcomes.

We found no association between increased UtA resistance (1‐SD increase in UtA PI) and motor domain functioning (Table 3).

TABLE 3.

Associations between UtA PI on motor development (NIH Toolbox). All models were adjusted for maternal age at NICHD‐FGS enrollment, maternal self‐reported race and ethnicity, pre‐gravid BMI, parity, education, marital status, child's age at ECHO study visit, gender, birth weight z‐score, and summated prenatal Edinburgh Depression Scale and Prenatal Stress Score.

			Uterine artery changes
Motor functions domain	n	Mean score (SD)	* 1 SD increase in PI mean difference (95% CI)
Dexterity	555	53.28 (13.20)	−1.05 (−2.18, 0.07)
Grip strength	537	49.24 (9.32)	−0.35 (−1.16, 0.46)
Standing balance	483	49.55 (18.35)	−0.2 (−1.84, 1.44)

^* p < 0.05.

The associations between increased UtA PI measurement and parent report of behavioral problems are shown in Table 4. There were no significant associations between a 1‐SD increase in UtA PI and child behavioral problems (overall or individual domains). The number of included subjects varies across Tables 2, 3, 4 due to variable numbers of children who underwent the specific neurodevelopmental and behavioral tests offered. When we repeated the analysis to include the results for only those children who had all the neurodevelopmental and behavioral outcomes of interest (n = 376), the estimated regression coefficients remained similarly significant (Table S4).

TABLE 4.

Association between UtA PI and child behavioral problems (Child Behavior CheckList). All models were adjusted for maternal age at NICHD‐FGS enrollment, maternal self‐reported race and ethnicity, pre‐gravid BMI, parity, education, marital status, child's age at ECHO study visit, gender, birthweight z‐score, and summated prenatal Edinburgh Depression Scale and Prenatal Stress Score.

			Uterine artery changes
Behavioral Problems Domain	n	Mean score (SD)	* 1 SD increase in PI mean difference (95% CI)
Internalizing problems	433	45.9 (9.12)	0.26 (−0.57, 1.08)
Externalizing problems	433	45.7 (9.3)	−0.26 (−1.13, 0.6)
Attention problems	434	53.3 (5.19)	−0.13 (−0.61, 0.36)
Oppositional defiant problems	434	53.0 (4.72)	−0.08 (−0.51, 0.35)
Conduct problems	434	52.4 (4.43)	0.02 (−0.39, 0.43)
Depressive problems	434	52.7 (4.54)	−0.03 (−0.47, 0.4)
Anxiety problems	434	52.5 (4.61)	0.22 (−0.22, 0.67)
Total problems	433	45.32 (9.97)	−0.03 (−0.93, 0.88)
Subgroup of children with greater problems identified on the Child Behavior Checklist (normalized t‐score ≥ 60)
Internalizing problems	37/433		−0.48 (−2.37, 1.4)
Externalizing problems	31/433		−0.36 (−2.46, 1.74)
Attention problems	43/434		−0.44 (−4.83, 3.95)
Oppositional defiant problems	39/434		0.11 (−2.3, 2.52)
Conduct problems	38/434		−0.53 (−2.9, 1.83)
Depressive problems	57/434		0.45 (−1.01, 1.91)
Anxiety problems	51/434		0.33 (−1.04, 1.7)
Total problems	36/433		−1.19 (−3.38, 1)

*p < 0.05.

Elevation of the UtA PI was associated with an increased risk of PE (OR 1.71 per 1SD; 95% CI 1.16, 2.43). Mediation analyses were performed and revealed that neither PE (n = 17;2.5%), SGA (n = 57, 8.4%), nor birthweight z‐scores exhibited significant mediation effects on executive functioning. Despite the sample size, our power was low in some of the child age subgroups and mediation was not performed in the children 8 years of age due to the extremely low prevalence of the mediator (Tables S5–S7).

We conducted a sensitivity analysis including only those patients who breastfed their child for at least 6 months (n = 345). Due to the sample size reduction, most statistical significance became non‐significant, however, the effect sizes remained similar (Table S8). We also repeated all analyses with log‐transformed UtA PI, and all significant associations with the raw UtA PI measures remained significant (Table S9).

4. DISCUSSION

4.1. Main findings

In this large, prospective, racially and ethnically diverse cohort of low‐risk pregnancies, a 1‐SD elevation in mid‐pregnancy UtA PI was associated with small but significant reductions in age‐specific child executive domain functioning. No association was found with motor skills. While the observed effect sizes are modest and their clinical implication uncertain, the findings demonstrate a link between subtle placental vascular findings with later childhood outcomes.

4.2. Results in the context of what is known

It is well established that poorer neurodevelopmental outcomes are linked to prematurity and low birthweight, and pregnancy complications such as FGR and PE [2, 7, 39, 40, 41, 42]. These conditions are associated with increased utero‐placental resistance, reflected by an elevated UtA PI [5, 13]. We assessed the relationship between utero‐placental resistance and neurodevelopmental outcomes using a longitudinal study design bridging mid‐pregnancy UtA Doppler measures with child neurodevelopmental and behavioral outcomes between ages of 4 and 8.

As previously noted, increased UtA resistance has been repetitively associated with abnormalities in trophoblastic proliferation and early placental development [10, 11, 12]. A recently published systematic review demonstrated higher odds of abnormal infant cognitive development in cases with pathologically identified vascular perfusion lesions [9]. The authors found a strong association between placental lesions typical of vascular malperfusion and increased risk of brain abnormalities in term neonates and neurodevelopmental impairment in both term and preterm infants.

Notably, neurodevelopmental abnormalities most commonly identified in association with impaired utero‐placental perfusion have been found in the domain of executive rather than motor functioning [8, 9]. This literature is consistent with our findings, demonstrating age‐related reductions in cognitive flexibility associated with increased UtA resistance.

While the clinical implications of these small differences are unclear, there is potential biological plausibility. In the case of FGR, it is hypothesized that the greater oxygen demands of anterior cortical regions may create a greater susceptibility to subtle hypoxia. Alternatively, the anterior cortical region may possibly be at greater risk as a consequence of compensatory redistribution of fetal blood flow. Normal cerebral blood flow favors the anterior cerebral artery distribution to the frontal lobes, which have the greatest oxygen demand and direct executive functioning. In the presence of fetal hypoxia, there is a redistribution of blood flow favoring energy conservation and preservation of critical mid‐brain function. This redistribution involves vasodilation of the middle cerebral artery with compensatory shunting of cerebral blood flow to parietal and mid‐brain regions at the expense of the anterior cerebral circulation [7]. However, whether increased maternal UtA resistance in otherwise uncomplicated pregnancies translates to such changes in the fetal circulation is unknown.

Although significant literature suggests associations between adverse antenatal conditions such as PE or FGR, and childhood neurodevelopment, there is a paucity of research evaluating the association of increased UtA resistance and adverse developmental outcomes [2, 43, 44]. Reassuringly, Okido et al. examined patients with an elevated second trimester UtA PI > 90th %, observing that outcomes in the second or third year of life did not differ between groups [45]. Our cohort included children aged 4–8 years and considered multiple neurodevelopmental and behavioral outcomes. Importantly, we evaluated neurodevelopment in association with incremental increases in UtA PI as opposed to a dichotomous finding of UtA PI above or below the 90th percentile. Beyond the Okido study, there is a scarcity of research evaluating mid‐pregnancy UtA resistance and neurobehavioral outcomes.

Notably, in the analysis of all children, we found no significant associations between increasing UtA PI and executive domain function. However, associations were identified in age‐specific subgroups. Child age is the major determinate of executive function, and the NIH Toolbox Battery includes separate tests for inhibitory control/sustained attention, cognitive flexibility, and working memory with age‐specific norms reflecting the dynamic progression in executive domain functioning between younger and older children [46, 47]. Working memory storage capacity increases dramatically between the ages of 3 and 8 years and approaches adult levels between the ages of 6 and 8 years [48, 49]. This developmental trajectory of improvement in memory, especially recollection of previously encountered stimuli, as well as the ability to recall contextual details (List Sorting Memory Test), results in age‐related improvements in memory that are relatively linear between the ages of 4 and 11 years [50, 51]. Similar age‐related improvement in performance is also the norm using the Dimensional Change Card Sorting Test for cognitive flexibility [23, 52]. These improvements are believed to reflect enhanced functional development of networks in the medial temporal lobe specialized for recollection and are identifiable in children aged 8 years using fMRI [53]. These rapid advances in executive domain functioning during early childhood inform and justify age‐specific assessment of younger and older children. Age‐specific testing may reveal either the commonly seen improvement in neurodevelopmental testing as children get older (cognitive flexibility in this cohort) or potentially reveal deficits exposed by the more challenging testing versions administered to older children [54].

The well‐studied link between UtA impedance and PE was reinforced by our results [55]. We found that a 1‐SD increase in UtA PI was associated with a more than 1.5‐fold increase in maternal risk of PE. While the risk of PE was significantly increased in this study, the magnitude of that increased risk is less than has been reported by others. That difference may be explained by the fact that the NICHD‐FGS cohort was low‐risk excluding women with a history of PE or an SGA infant or any maternal vascular disease.

4.3. Clinical and research implications

In complicated pregnancies, endothelial dysfunction and hypoperfusion may result in chronic fetal hypoxia [56]. Upregulation of immune and angiogenic factors in response to hypoperfusion of the placenta has been implicated in cases of PE [57, 58]. Additionally, fetal hypoxia can result in compensatory redistribution of blood flow away from the frontal lobes, which has been associated with childhood behavioral and cognitive deficits [59, 60]. However, whether small increases in UtA resistance are associated with changes in fetal circulation that affect subsequent child neurodevelopmental or behavioral changes is unknown. More research is needed to confirm these findings and determine whether changes in UtA Dopplers might be useful for the early diagnosis of placental dysfunction or as a therapeutic target. A recently published study by Agrawal et al. evaluated UtA Dopplers, maternal angiogenic factors, and subsequent placental pathologic signs of vascular malperfusion [61]. Among women diagnosed with maternal vascular malperfusion based on placental pathology (diffuse villous changes suggestive of acceleration maturation), 71.6% had mean UtA PI > 95th percentile, and low placental growth factor (PlGF) in 80.7% [61]. These findings suggest that changes in placental pathology that represent ischemia and reperfusion injury (maternal vascular malperfusion) can be present in the setting of normal UtA Dopplers, and allude to the potential significance of subtle changes in UtA flow that influence placental perfusion.

Beyond the dichotomous diagnosis of SGA (birthweight < 10th percentile), small newborn size has also been associated with impaired neurodevelopment [62]. To test the potential mediation of our findings by newborn size, we converted each birthweight into a gestational age and sex specific z‐scores for a continuous analysis of birthweight mediation. Neither PE, SGA, nor birthweight z‐score had a significant impact on our findings of an association between increased UtA PI and child executive domain function. A similar worthwhile research question would be the continuous effect of maternal blood pressure (as opposed to the dichotomous diagnosis of PE) on neurodevelopmental outcome. Unfortunately, maternal blood pressure measures at delivery were not collected in our cohort.

The International Society of Ultrasound in Obstetrics and Gynecology recommends screening for PE in the first trimester with UtA Dopplers to identify candidates for aspirin administration [13]. Conversely, ACOG recommends a maternal risk stratification system to determine candidates for aspirin therapy but does not advise universal UtA Doppler interrogation [63]. Aspirin is hypothesized to prevent endothelial dysfunction associated with placental hypoxia [64]. Our study demonstrates that overlooked increases in the UtA PI in the mid‐gestation have potential implications for childhood executive domain functioning and may represent a reasonable target for intervention.

We believe that these exploratory findings may provide a roadmap for future larger studies to include mid‐gestation UtA Dopplers along with middle cerebral artery Dopplers in longitudinal and long‐term assessments of child neurodevelopmental and behavioral trajectories. However, until such studies are performed, changes in UtA resistance should not prompt early delivery without evidence of other evolving risks, as gestational age remains the most significant predictor of neurodevelopmental outcome. Despite the associations that have been demonstrated between UtA Doppler and a variety of pregnancy outcomes, it remains to be demonstrated if UtA Doppler can be used in clinical care to improve maternal‐fetal health and long‐term child outcomes.

4.4. Strengths and limitations

Findings may be less generalizable to patient populations with higher rates of vascular comorbidity. The NICHD‐FGS is recognized for its prospective, high‐quality data collection and ultrasound measurements, robust follow‐up, and minimal attrition bias [33]. The current study allowed for the integration of antenatal UtA Doppler findings with multiple domains of child neurodevelopment. The study is strengthened by using validated instruments and questionnaires for assessing executive and motor domains and behavioral outcomes [47]. Findings were robust in mediation analyses, inverse propensity weighted analysis, and sensitivity analyses including only those women who breastfed for more than 6 months and with log‐transformation of the data to control for non‐normal distribution. Another strength of the study is the pre‐selected low‐risk population. Patients with any history of vascular, endocrine, or rheumatological disease, prior adverse pregnancy outcomes, or exposures to noxious substances or medications were excluded. While pregnancy‐associated disorders did occur, they were infrequent. By studying low‐risk patients and smaller increases in the UtA PI Doppler, subtle associations between increased UtA resistance and neurodevelopment could be identified. Using an incremental 1‐SD increase in UtA PI as our independent variable (as opposed to extremes) is a relatively unique approach to this type of analysis and more revealing in a low‐risk population.

Our findings should be considered exploratory. Study limitations included a lack of MCA Doppler findings in the original NICHD‐FGS. Consequently, we were unable to test our hypothesis that increased UtA resistance is associated with redistribution of intracerebral blood flow away from the anterior cerebral circulation. Meher et al. have proposed that fetal neurologic injury can occur prior to detection of MCA Doppler abnormalities as a response to altered fetal hemodynamic adaptation to mild hypoxia [65]. While the lack of MCA Doppler information is a limitation, subtle increases in UtA PI may represent a marker of impaired utero‐placental perfusion and a potential antenatal predictor of childhood executive domain functioning.

The time frame for UtA Doppler evaluation is relatively wide (18–25 weeks of gestation) and may fluctuate between trimesters, but the time frame used is consistent with much of the published literature investigating mid‐trimester UtA Doppler interrogation. Identification of elevated UtA resistance in the mid‐trimester is important not only to allow for any potential interventions but also because a proportion of “abnormal” UtA Dopplers measured in the mid‐trimester may improve by the third trimester. Each UtA PI was normalized to the gestational age mean when collected for determination of a 1‐SD difference, and if any participant was assessed twice during the interval, only the later assessment was used.

The study is also limited by the absence of placental pathology. We are unable to correlate our measures of increased UtA PI with any direct identification of reduced placental size or histopathologic abnormalities. As with other studies of placental origins of disease, direct measures of fetal hypoxia are difficult to obtain. Consequently, sonographic markers of increased utero‐placental resistance are used as a surrogate for hypoxia. The threshold of placental resistance that contributes to the clinical findings of FGR and PE is unknown, but they appear to be higher than those identified in this low‐risk cohort. These data suggest that lesser degrees of utero‐placental resistance may have long‐term implications.

There is also potential bias when findings are based on parental report using the CBCL. Criticism of parent report as primary outcome data is due to potential ascertainment and recall bias and shared method variance. Nevertheless, the CBCL is widely used to assess child internalizing and externalizing behaviors with demonstrated high reliability and validity [27, 31]. In the current study, we found no significant associations between a 1‐SD increase in UtA PI and any child behavior problems identified by the CBCL. Results were also controlled for measures of both maternal perceived stress and depression, which have been associated with child neurodevelopment [34, 66].

While several covariates were included, we could not control other potentially relevant covariates such as social support, maternal anxiety, partner relationships, parenting competencies, or other unmeasured covariates. Lastly, we conducted multiple statistical tests without correction to minimize the risk of decreasing our effective statistical power or increasing the false negative rate for detecting modest associations. As in any study, some findings may have occurred by chance such as our unexpected finding of enhanced inhibitory control/sustained attention in the small subgroup of children aged 8 (n = 78). This risk represents the potential trade‐off between type 1 error and discovery in an exploratory study such as this. We concur with epidemiologists who believe that if the usual concerns regarding biases (selection, confounding, information validity, and adequate power) can be satisfied, then correction for the number of associations is unnecessary [67, 68]. In fact, it may be harmful if multiple corrections obscure modest differences between groups that should be confirmed or strengthened by repetitive or further study [69]. Reassuringly, comprehensive multivariable adjustments and inclusion of multiple mediation analyses, sensitivity analyses, inverse propensity analyses, and analysis of children who underwent all testing all demonstrated a consistency of the findings, enhancing their credibility. Nonetheless, our findings should be considered exploratory and cautious interpretation is appropriate.

5. CONCLUSIONS

Our results suggest that changes in mid‐pregnancy UtA resistance were associated with small reductions in age‐specific executive domain functioning in a low‐risk obstetric population. Specifically, reductions in cognitive flexibility were found in children aged 4–7 years with a 1‐SD increases in UtA PI. These findings were achieved through the application of validated and developmentally appropriate tools used to assess multiple domains of child neurodevelopment. The clinical implications of these findings are unknown but should be of interest to both researchers and clinicians interested in the link between placental vascular findings and later childhood outcomes. These possible associations should be explored in future studies.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supporting Information