Testing a cascade model linking prenatal inflammation to child executive function

Abstract

Inflammation during pregnancy is beginning to be understood as a risk factor predicting poor infant health and neurodevelopmental outcomes. The long-term sequelae associated with exposure to prenatal inflammation are less well established. The current study examined associations between maternal inflammation during pregnancy, markers of infant neurodevelopment (general cognitive ability, negative affect, and sleep quality), and preschool executive function (EF) in a longitudinal sample of 40 African American mother-infant dyads. Mothers completed a blood draw in the third trimester of pregnancy to measure plasma levels of C-reactive protein (CRP) and pro-inflammatory cytokines (e.g., interleukin 6 [IL-6], tumor necrosis factor-alpha [TNF-α]). When infants were 6 months of age, we assessed general cognitive ability via the Bayley-III, negative affect via the Still-Face Paradigm, and sleep quality via actigraphy monitoring. When children were 4 years of age, we assessed their EF ability using four tasks from the EF Touch battery. Elevated levels of maternal CRP, IL-6, and TNF-α were associated with poorer infant general cognitive ability. Although there were no direct effects of prenatal inflammation on preschool EF, we observed an indirect relationship between IL-6 and preschool EF ability via infant general cognitive ability. Our findings suggest that prenatal inflammation may have long-lasting, cascading implications for child neurodevelopment. Implications of these findings for health disparities in women and children of color are discussed.

1. Introduction

The Developmental Origins of Health and Disease [1] and fetal origins [2] hypotheses predict that prenatal conditions may have lasting effects on child health and developmental outcomes. Researchers testing these hypotheses have found that specific at-risk groups of children, such as children born low birthweight (LBW) and/or preterm (PT), are at increased risk for a host of neurodevelopmental difficulties, including low IQ and poor executive function as well as increased risk for ADHD, internalizing and externalizing behavior problems, and autism spectrum disorders [3–6]. Beyond broad proxies for adverse prenatal experiences, researchers have started to examine specific risk factors within the prenatal environment that are related to individual differences in child neurodevelopment. One factor that has begun to receive attention is the role of prenatal inflammation, the primary risk factor assessed in the current study.

Inflammation is a critical process by which the body defends itself against infection, illness, and injury [7]. Exposure to an external (e.g., physical damage) or internal (e.g., pathogenic intrusion) insult initiates a release of pro-inflammatory cytokines, signaling molecules that recruit white blood cells and T cells to mobilize the body’s response to fight pathogens and repair cellular damage. Whereas acute inflammation is an adaptive bodily response to perceived or actual infection, chronic, low-grade levels of inflammation that arise in the absence of a specific trigger (i.e., systemic inflammation) may confer increased risk for chronic health problems, including cardiovascular disease, cancer, metabolic disorders, as well as overall mortality [8,9].

Maternal inflammation during pregnancy has received increasing attention in recent years, as elevated levels of C-reactive protein (CRP) and pro-inflammatory cytokines (e.g., interleukin 6 [IL-6], tumor necrosis factor-alpha [TNF-α]) have been associated with poorer infant health and neurodevelopment. Results from one prospective study indicate that maternal inflammation during the antenatal period predicts a composite of adverse neonatal outcomes, including respiratory compromise, sepsis, and intraventricular hemorrhage [10]. Maternal infection, one cause of elevated inflammation, has been associated with elevated risk of schizophrenia and autism in offspring [11]. Whereas findings from pre-clinical studies suggest these associations are causal [12], the applicability of this research is limited since it relies on animal (rather than human) models of pregnancy. As a result, the association between prenatal inflammation and neurodevelopment in humans remains unclear.

The few human studies that have investigated prenatal inflammation and infant outcomes are consistent with the animal literature (for a recent review, see [13]). Elevated inflammation during pregnancy has been shown to predict differences in infant brain structure and function, including variations in amygdala volume, as well as connectivity between various functional brain networks [14–16]. In turn, variations in brain volume and connectivity predicted differences in cognition and behavior at age two [14,15]. Moreover, inflammation during pregnancy has been shown to mediate the effects of maternal distress and socioeconomic hardship on infant outcomes, including prematurity [17] and neurological abnormalities [18]. Thus, elevated levels of inflammation during pregnancy predict abnormal neurodevelopment in children, which may manifest in various alterations of central nervous system functioning.

A critical next step in the study of human prenatal inflammation on offspring neurodevelopment is the need for research outside of primarily White/European-American [15,19] and Hispanic [17] populations. Previous research in this area has yet to include a representative proportion of African American women despite findings that women of color experience greater inflammation during pregnancy than their White counterparts [20]. This may be due to differential experiences of systemic stressors such as chronic psychological stress [21], experiences of racial discrimination [20], and the accessibility of healthcare [22]. Perhaps as a consequence of systemic inflammation, African American women experience higher proportions of adverse health and birth outcomes, such as preterm birth [23]. Research is needed to determine whether associations between inflammation and neurodevelopment primarily reported in White and Hispanic women and children are similar in African American samples. Additionally, much of the research described thus far has examined links between inflammation and clinically relevant child outcomes (e.g., prematurity, medical complications) rather than more subtle differences in behavior or long-term developmental sequelae. Therefore, we know little about how widespread and enduring the associations are between prenatal inflammation and child outcomes.

In the current study, we examined the association between prenatal maternal inflammation and infant behavioral outcomes thought to be markers of central nervous system integrity in a sample of African American women and children. Infant general cognitive ability, negative affect, and sleep quality are thought to have neurobiological underpinnings [24,25], are impacted by prenatal conditions [26–28], and are related to later child functioning [29–31]. Therefore, we focused on these three facets of infant behavior as relevant neurodevelopmental markers that may be programmed by prenatal exposure to inflammation. We hypothesized that greater circulating CRP, IL-6, and TNF- α during pregnancy would be associated with poorer general cognitive ability, more observed negative affect, and poorer sleep quality in infants at 6 months of age.

A secondary aim of this study was to investigate whether associations between prenatal inflammation and child outcomes endure into childhood. Therefore, we tested the association between prenatal inflammation and child executive function (EF) at age 4, including whether this relation is mediated through infant neurodevelopment. One recent study found that maternal inflammation during the third trimester of pregnancy prospectively predicted child ADHD symptoms when children were 4–6 years of age [19]. Given that ADHD is a disorder characterized by deficits in EF, we expected to observe an inverse relationship between prenatal inflammation and child EF in our sample. Further, prior research suggests that the association between prenatal risk factors and childhood EF is indirect, mediated through earlier measures of general cognitive ability [32]. Therefore, we hypothesized that the association between prenatal inflammation and child EF might be indirect, mediated through markers of infant neurodevelopment.

2. Materials and methods

2.1. Participants

Data come from the Neonatal and Pediatric Sleep (NAPS) Study, a longitudinal investigation of 95 African American mother-infant dyads. For these analyses, we included a subset of women who were recruited during pregnancy and completed a third trimester blood draw. Women were recruited via electronic medical records from a large university hospital in the Southeastern United States, as well as via flyers posted in local OB/GYN clinics and other community locations. Potential dyads were excluded if mothers were younger than 18 years of age, did not identify as Black/African American, did not speak fluent English, or if infants were not singletons. Additional information about this sample has been reported elsewhere [30].

The current analyses included 40 dyads with prenatal inflammation data and data from at least one 6-month outcome. Of these, 33 had 4-year follow up EF data. Participants who were included in the analytic sample did not differ from excluded participants on the basis of prenatal inflammation levels, maternal education, child sex, prematurity status, or 4-year EF ability. In the analytic sample, six infants (15%) were born prematurely (< 37 weeks gestational age [GA]). For infants born preterm, 6-month visits were delayed until infants reached the appropriate adjusted age. On average, mothers in the analytic sample were 29 years old at childbirth (range = 20–42 years) and had an average of 15.0 years of education (SD = 2.2), with 49% having earned at least a four-year college degree. Of the 37 (92.5%) mothers who reported their use of government assistance programs, their household structure, and their health history at the 3-month post-partum visit, 18 (51.4%) had used Women, Infants, and Children (WIC), 18 (51.4%) were on Medicaid, and 22 (59.5%) lived in the same household as their child’s biological father. One (2.7%) participant had a history of diabetes, and eight (21.6%) had a history of hypertension. Of the 35 (87.5%) mothers who self-reported their height and weight at the 3-month postpartum visit, body mass index [BMI] ranged from 19.64 to 48.47, with over two-thirds (n = 24; 68.6%) of mothers having self-reported BMI in the obese range (BMI > 30). Of the 33 (82.5%) mothers who reported their birth mode, 52% had a vaginal birth whereas 48% experienced a cesarean birth.

2.2. Procedure

Women visited a campus laboratory during their third trimester of pregnancy (M_GA = 29.2 weeks; range = 26.1 – 32.7 weeks) for a non-fasting blood draw. A trained phlebotomist collected whole blood samples (12 mL) via venipuncture into two 6 mL EDTA tubes that were stored on ice until being centrifuged for 10 min at 2000 x g at 4 °C. Plasma was aliquoted into six 200 mL aliquots and stored at − 80 °C until time of assay.

When infants were 6 months of age, dyads participated in a home visit during which trained research assistants (RAs) conducted videotaped parent-child interactions, infant cognitive assessments, and parental questionnaires. Following the daytime home visit, families completed a week-long sleep assessment, which included seven days and nights of actigraphy monitoring. At the beginning of the home visit, a lightweight actogram (Actiwatch-2) was placed on the infant’s left ankle and remained there for the duration of the sleep assessment week. Families were compensated up to $130 in the form of a gift card and infants received a small gift.

Mothers were re-contacted and invited to participate in a lab visit when their child was nearing 4 years old. Mother-child dyads visited a campus laboratory to complete a two-hour visit led by a trained RA. Children completed a battery of assessments designed to measure cognition and EF. Mothers sat in an adjacent room fitted with a two-way mirror where they completed online and paper questionnaires that measured demographic information and information about maternal and child health and development. Mothers were compensated $50 for the visit and children received a small gift. All procedures were approved by the University of North Carolina at Chapel Hill Institutional Review Board, and participants provided informed consent.

2.3. Measures

2.3.1. Prenatal inflammation

CRP was assayed from plasma using an enzyme-linked immunosorbent assay (ELISA) from Thermo Fisher (Cat #BMS288INST). IL-6 and TNF-a were assayed using a standard multiplex array from EMD Millipore on the Luminex 200 S (Cat# HSTCMAG-28SK); the multiplex included granulocyte-macrophage colony-stimulating factor (GM-CSF), fractalkine, interferon gamma (IFN-γ), IL-1β, IL-2, IL-4, IL-6, IL-10 and TNF-α. From these, we selected inflammatory markers for analysis that had been used in prior research to investigate the relation between early life inflammation and child outcomes (e.g., [14,15]). All samples were run on 1.5 plates on the same day and a single standard curve was applied. Therefore, there is no inter-assay coefficient of variance (CV). Lower limits of detection (LLOD) from the standard curves and percentage of samples below the LLOD were as follows: CRP (0.3 mg/L, 0%), IL-6 (0.22 pg/mL, 0%), and TNF-α (2.68 pg/mL, 18%). Average intra-assay CV values were acceptable (< 5%) for all inflammatory markers.

2.3.2. Infant cognitive ability

Infant general cognitive ability was measured at 6 months using the cognitive subscale of the Bayley Scales of Infant Development (BSID-III) [33]. The BSID-III is a widely used measure of cognitive development for children in the first three years of life that measures abilities such as sensorimotor development, object manipulation, memory, and simple problem solving. Scaled scores were calculated based on infant performance and age at assessment.

2.3.3. Infant negative affect

Infant negative affect was coded from the Still-Face Paradigm (SFP) [34] which was administered at the 6-month daytime home visit. The SFP is a multi-episode parent-child interaction task that elicits a range of child affective states and behaviors. Infants were placed in a car seat on the floor and mothers sat directly in front of their infant. Two cameras were set up to videotape the face and body of both the parent and infant. The SFP consists of three 2-minute episodes, separated by a 15-second head turn. In the normal episode, mothers were instructed to interact with their infant as they normally would. In the still-face episode, mothers were told to look at their infant with an expressionless face and were instructed not to talk to or touch their infant. In the reunion episode, mothers were told to again interact with their infant as they would normally. The SFP was cut short if the infant expressed extreme distress (e.g., hard crying) for 15 s or more.

Trained RAs coded the SFP episodes in 5 s intervals. For each interval, coders marked the presence or absence of certain infant behaviors. Of interest to the current investigation were codes pertaining to infant affect. Infant negative affect (NA) was coded if the infant displayed a negative expression (e.g., sharply lowered brows, tightly closed eyes, and/or downward-turned corners of the mouth). Infant positive affect (PA) was coded if the infant displayed a positive expression (e.g., raised corners of mouth, raised cheeks, wide mouth, surprised/playful expression). If NA and PA were present simultaneously (e.g., tightly closed eyes, but raised corners of mouth), NA was coded. If distinct NA and PA expressions were present in the same interval, then the expression that was present for the majority of the interval was coded. Interrater reliability was measured using Cohen’s kappa, which exceeded 0.70 for all codes. Because the length of the SFP task could vary, infant NA was transformed into a proportion score to represent the proportion of epochs in each episode in which the behavior was present. As the still-face episode elicited the most NA, we used the proportion of NA during this episode as our marker of observed negative affect.

2.3.4. Infant sleep quality

Infant sleep quality was assessed in the week following the 6-month daytime home visit using actigraphy, which uses an accelerometer to measure limb movement in 15-second epochs. At the end of the sleep assessment week, actigraphy data were downloaded to a PC computer and edited using Phillips Actiware software (version 6.0). Actogram algorithm settings were selected as follows: immobile minutes for sleep onset were set to 5 min; minimum rest interval size was set to 20 min; multiple rest intervals per day were allowed; and automatically set minor rest intervals were allowed. Consistent with previous validation studies [35], the activity threshold for scoring the infant as awake was set to the Low setting (20 activity counts) at 6 months. Intervals of sleep/wake that were not automatically detected were manually entered, following a procedure detailed elsewhere [36].

Sleep quality variables derived from actigraphy included longest sleep period (LSP; longest period where the infant was coded as being continuously asleep), nighttime sleep ratio (NSR; ratio of nighttime [7PM-7AM] sleep to total [24-hour] sleep), and number of long night wakings (NW; periods of wakefulness ≥ 5 min). Infant LSP and NW were calculated for each night and averaged over the sleep assessment week, whereas NSR was calculated for the entire study week (e.g., weekly nighttime sleep divided by weekly total sleep). Longer LSPs, fewer NW, and larger NSRs were indicative of better sleep quality.

2.3.5. Child EF

Child executive function was measured at the 4-year assessment using the Executive Function Touch (EF Touch) battery [37]. EF Touch is a computerized battery of tasks that provide performance-based measures of children’s inhibitory control, working memory, and attention shifting abilities. The battery was initially created, administered, and studied in paper and pencil format but has been computerized to standardize task administration and streamline data capture, processing, and scoring. Extensive psychometric evaluation has supported the use of the EF Touch battery with preschool populations [38,39]. These studies have demonstrated that individual tasks measure the full range of children’s ability and that tasks can be combined into overall EF composite scores with acceptable test-retest reliability and expected patterns of age-related improvements. Due to time constraints that precluded the use of the full battery and in line with previous studies that used only a subset of tasks [40], we included a subset of two inhibitory control tasks, one working memory task, and one cognitive flexibility task. Children’s performance (e.g., accuracy) was averaged across tasks to create a single measure of their EF ability.

2.3.6. Covariates

Covariates were selected based on prior evidence of a potential confounding relation between maternal prenatal inflammation and child cognitive abilities and to increase the precision of the outcome measures. We selected maternal education as an indicator of maternal socioeconomic status (SES), as maternal SES has been found to be associated with both maternal inflammation and child neurodevelopmental outcomes and therefore may act as a confounder [41, 42]. Child age at outcome assessment (both 6-month and 4-year outcomes) was included as a covariate to account for variability in timing of child visits, particularly as families at higher risk for poor neurodevelopmental outcomes may be more difficult to schedule. We additionally adjusted for prematurity (GA < 37 weeks) in order to increase the precision of our outcome measures, given that infants born premature are at increased risk of neurodevelopmental impairment [43,44].

2.4. Analytic plan

Descriptive analyses were conducted using SAS 9.4 software (SAS Institute Inc., Cary, NC, USA) and substantive models were conducted using path analysis in MPlus 8.1 [45]. To address our first aim, we estimated path models in which prenatal inflammatory biomarkers were used as predictors of infant outcomes (i.e., cognition, negative affect, sleep). The association between each inflammatory marker and each infant outcome was tested independently (i.e., in separate models). All models statistically controlled for maternal education, child prematurity, and infant age at the 6-month visit. As these path models were fully saturated, fit statistics are not presented.

We addressed our second aim using path analysis to examine the direct and indirect effects of prenatal maternal inflammation on child EF via infant neurodevelopment. Only infant outcomes that were significantly associated with inflammation in the first aim were included in these secondary models. An exemplar model is shown in Fig. 1. As described earlier, we included maternal education, child prematurity, and infant age at the 6-month visit as predictors of infant cognition and child’s age at the 4-year follow-up visit as an additional predictor of child EF. Model fit was evaluated using the root mean square error of approximation (RMSEA), comparative fit index (CFI) and Tucker-Lewis index (TLI), in which RMSEA values less than 0.08 and 0.06 indicate reasonable and good model fit, respectively, and CFI and TFI values greater than 0.90 and 0.95 indicate reasonable and good model fit, respectively. Missing data were handled using full-information maximum likelihood with robust standard errors.

Fig. 1.

Exemplar cascade model linking maternal prenatal inflammation to child EF via infant outcomes.

3. Results

3.1. Descriptive statistics

Means, standard deviations, and ranges for continuous study variables and frequencies for categorical study variables are displayed in Table 1. Correlations among all study variables are displayed in Table 2. Maternal prenatal inflammatory markers were generally in expected ranges compared to previous studies of pregnant women [46–48]. Bivariate correlations showed that inflammatory markers were marginally to highly correlated with one another (rs = 0.17 – 0.81). IL-6 (r = − .51, p = .001) and TNF-α (r = − .41, p = .02) were negatively associated with infant cognitive ability. A similar trend was observed for CRP, although this relationship was only marginally significant (r = − .31, p = .06). TNF-α was positively associated with child EF (r = .39, p = .04) and marginally positively associated with infants’ NSR (r = .37, p = .06).

Table 1.

Descriptive statistics for all study variables.

	N	Mean (SD)	Range
Maternal-level variables
Education (years)	37	15.03 (2.18)	10.00–18.00
Third trimester CRP (mg/L)	40	4.23 (2.79)	0.31–10.97
Third trimester IL-6 (pg/mL)	40	5.65 (5.55)	1.56–35.34
Third trimester TNF-α (pg/mL)	33	8.22 (3.27)	5.61–22.18
Third trimester mass index at 3 months postpartum	35	33.95 (8.14)	19.64–48.87
Infant-level variables
Age at 6-mo visit (days)	40	197.33 (15.89)	168.00–245.00
Cognitive ability (scaled score)	38	11.05 (2.77)	3.00–17.00
Negative affect (proportion score)	37	0.52 (0.43)	0.00–1.00
Night wakings (total wakings)	33	1.78 (0.77)	0.00–3.86
Longest sleep period (min)	33	290.63 (76.59)	184.21–525.18
Nighttime sleep ratio (nighttime sleep to total sleep)	33	0.79 (0.06)	0.66–0.98
Child-level variables
Age at 4-yr visit (years)	33	4.00 (0.16)	3.74–4.38
EF (proportion correct)	33	0.63 (0.15)	0.26–0.95
	N	n (%)
Prematurity status (GA < 37 weeks)	40	6 (15)
Delivery mode	31
Vaginal birth		16 (52)
Cesarean birth		15 (48)
Maternal-level variables
History of high blood pressure	37	8 (22)
History of diabetes	37	1 (3)
Use of WIC in past 3 months	37	18 (51)
Use Medicaid in past 3 months	37	18 (51)
Breastfeeding at 3 months	35
Not breastfeeding		14 (40)
Breastfeeding		21 (60)
Biological father lives in household	37	22 (59)

Note. Means are provided for all continuous variables whereas percentages are provided for dichotomous variables. CRP = C-reactive protein; IL- 6 = interleukin 6; TNF- α = tumor necrosis factor-alpha; Infant cognitive ability was indexed using the BSID-III Cognitive Subscale Scaled Score; Negative affect was indexed as the proportion of 5-second intervals with coded negative affect behavior during still-face episode; Min=minutes; EF = Executive Function, which was indexed as the proportion of correct items across all EF administered EF tasks; Prematurity status was defined as gestational age less than 37 weeks; Use of WIC and Medicaid in past 3 months was ascertained at the 3-month postpartum visit; Biological father living in the household was ascertained at the 3-month postpartum visit.

Table 2.

Correlations among all study variables.

	1	2	3	4	5	6	7	8	9	10	11	12
1. Maternal education (years)	–
2. Infant age at 6-mo visit (days)	.29±	–
3. Prematurity status	−.14	.20	–
4. Third trimester CRP (mg/L)	.03	.14	.14	–
5. Third trimester IL-6 (pg/mL)	.15	.18	−.10	.28±	–
6. Third trimester TNF-α (pg/mL)	−.07	.07	−.17	.17	.81 * **	–
7. Infant cognitive ability (scaled score)	.08	−.02	−.30±	−.31±	−.51 * *	−.41 *	–
8. Infant negative affect (proportion score)	−.21	−.37 *	.00	.01	.11	.25	−.09	–
9. Infant night wakings (total wakings)	−.23	−.21	−.21	−.16	−.10	−.16	−.12	−.19	–
10. Infant longest sleep period (min)	.31±	.19	.05	.10	.01	.12	.26	−.03	−.80 * **	–
11. Infant nighttime sleep ratio (nighttime sleep to total sleep)	.30	−.02	−.16	−.04	.16	.37±	.04	.13	−.37 *	.49 * *	–
12. Child age at 4-yr visit (years)	.04	−.45 * *	−.19	−.22	.02	.43 *	−.16	.22	−.02	.14	.28	–
13. Child EF (proportion score)	.15	−.15	−.21	−.27	.12	.39 *	.22	.05	−.02	.08	.11	.43 *

Note. Prematurity status was defined as gestational age less than 37 weeks; CRP = C-reactive protein; IL-6 = interleukin 6; TNF- α = tumor necrosis factor-alpha; Infant cognitive ability was indexed using the BSID-III Cognitive Subscale Scaled Score; Negative affect indexed as the proportion of 5-second intervals with coded negative affect behavior during still-face episode; Min=minutes; EF = Executive Function

^±p < .10

^*p < .05

^**p < .01

^***p < .001.

3.2. Path models

3.2.1. Cognitive ability

After adjustment for covariates, we observed significant, negative associations between IL-6, CRP, and TNF- α and infant cognitive ability (|β| =.28 − .57, p < .05; Table 3). These associations were of moderate effect size. In path models extending to child EF at age four (Table 5), IL-6, CRP, and TNF- α continued to be negatively associated with infant cognitive ability (|β| =.28 − .57, p < .05).

Table 3.

Summary of standardized estimates for infant cognitive ability and negative affect.

Path		Cognitive Ability β (SE)			Negative Affect β (SE)
A	Prematurity (yes vs. no)	−.27± (.14)	−.38 ** (.14)	−.38 ** (.14)	.06 (.14)	.10 (.15)	.13 (.14)
B	Maternal education (years)	.00 (.21)	.01 (.15)	−.02 (.18)	−.10 (.15)	−.10 (.15)	−.06 (.15)
C	Infant age at 6-mo visit (days)	.08 (.13)	.16 (.12)	.08 (.13)	−.35 * (.15)	−.38 * (.15)	−.41 ** (.16)
D	Third trimester CRP (mg/L)	−.28 * (.12)			.06 (.15)
D	Third trimester IL-6 (pg/mL)		−.57 ** (.18)			.20 (.15)
D	Third trimester TNF-α (pg/mL)			−.45 * (.21)			.34 * (.15)
	Model R2	.16	.39	.29	.15	.18	.26

Note. CRP = C-reactive protein; IL-6 = interleukin 6; TNF- α = tumor necrosis factor-alpha; Min=minutes.

^±p < .10

^*p < .05

^**p < .01

^***p < .001.

Table 5.

Summary of path models linking prenatal inflammation, infant neurodevelopment, and child EF with standardized estimates.

			Child EF β (SE)
Path	Inflammatory Marker		Third trimester CRP (mg/L)	Third trimester IL-6 (pg/mL)	Third trimester TNF-α (pg/mL)
A	Prematurity (yes vs. no)	→ Infant cognitive ability	−.27± (.14)	−.38 ** (.14)	−.38 ** (.14)
B	Maternal education (years)	→ Infant cognitive ability	.00 (.21)	.01 (.15)	−.03 (0.18)
C	Infant age at 6- mo visit (days)	→ Infant cognitive ability	.08 (.13)	.16 (.12)	.07 (.13)
D	Third trimester inflammatory marker	→ Infant cognitive ability	−.28 * (.12)	−.57 ** (.18)	−.48 * (.21)
E	Infant cognitive ability	→ Child EF	.30± (.18)	.38 * (.17)	.35± (.19)
F	Third trimester inflammatory marker	→ Child EF	−.09 (.17)	.41 (.29)	.41± (.21)
G	Child age at 4-yr visit (years)	→ Child EF	.44 ** (.14)	.46 *** (.13)	.36 * (.16)
	Indirect Effect of Inflammatory		−.08 (.07)	−.09± (.05)	−.17 (.12)
	Marker on Child EF
	Model R2 (Infant Cognition)		.16	.39	.31
	Model R2 (Child EF)		.35	.39	.41

Note. CRP = C-reactive protein; IL-6 = interleukin 6; TNF- α = tumor necrosis factor-alpha; EF = Executive Function.

^±p < .10

^*p < .05

^**p < .01

^***p < .001.

In the model testing the effects of IL-6, infant cognitive ability was positively associated with child EF (β = .38, p = .03). Whereas IL-6 was not significantly directly associated with child EF, there was a marginally significant indirect effect of prenatal IL-6 on child EF through infant cognition (β = − .09, p = .07). This model fit the data well (RMSEA =.00, CFI = 1.00, TLI = 1.00). In the extended path model testing the effects of TNF-α, infant cognition was marginally positively associated with child EF (β = .35, p = .07). Elevated prenatal TNF-α was also marginally associated with better performance on the EF battery at 4- years of age (β = .41, p = .06), but we found no indirect effects between TNF- α and child EF through infant cognitive ability. The model fit the data well (RMSEA =.00, CFI = 1.00, TLI = 1.00). Finally, in the extended path model testing the effects of CRP, infant cognitive ability was marginally positively associated with child EF (β = .30, p = .10). Prenatal CRP was neither directly nor indirectly associated with child EF and the model demonstrated poor fit to the data (RMSEA =.12; CFI =.85, TLI =.58).

3.2.2. Negative affect

As shown in Table 3, prenatal TNF-α was positively associated with infant negative affect (β = .34, p = .02) with a moderate effect size. Levels of prenatal CRP and IL-6 were not related to infant negative affect at 6 months. In path models extending to child EF, there were no significant relationships between infant negative affect and child EF, nor were there significant direct or indirect effects between any inflammatory marker and child EF.

3.2.3. Sleep quality

As shown in Table 4, elevated levels of TNF-α were marginally negatively associated with infant night wakings (β = − .33, p = .05) and positively associated with infant NSR (β = .41, p = .004) but were not associated with LSP. Levels of prenatal CRP and IL-6 were not related to sleep at 6 months. In path models extending to child EF, there were no significant associations between infant sleep measures and child EF, nor were there significant direct or indirect effects between any inflammatory marker and child EF.

Table 4.

Summary of standardized estimates for the models of infant sleep.

Path		Night Wakings β (SE)			Longest Sleep Period β (SE)			Nighttime Sleep Ratio β (SE)
A	Prematurity (yes vs. no)	−.20 (.14)	−.23 (.12)	−.25 * (.12)	.01 (.26)	.01 (.24)	.03 (.24)	−.21 (.25)	−.20 (.24)	−.17 (.25)
B	Maternal education (years)	−.16 (.20)	−.16 (.21)	−.28 (.24)	.26± (.15)	.27± (.15)	.37 * (.18)	.34 * (.15)	.33 * (.15)	.42 ** (.16)
C	Infant age at 6-mo visit (days)	−.11 (.13)	−.12 (.14)	−.05 (.15)	.09 (.17)	.10 (.18)	.04 (.18)	−.10 (.15)	−.12 (.15)	−.18 (.15)
D	Third trimester CRP (mg/L)	−.11 (.19)			.05 (.26)			−.05 (.15)
D	Third trimester IL-6 (pg/mL)		−.08 (.09)			−.02 (.16)			.10 (.10)
D	Third trimester TNF-α (pg/mL)			−.33± (.17)			.24 (.16)			.41 ** (.14)
	Model R2	.11	.11	.19	.10	.09	.16	.19	.19	.34

Note. CRP = C-reactive protein; IL-6 = interleukin 6; TNF- α = tumor necrosis factor-alpha; Min=minutes.

^±p < .10

^*p < .05

^**p < .01

^***p < .001.

3.3. Sensitivity analysis

To test for undue influence, we examined levels of inflammation that were three or more standard deviations from the mean. We found one participant who had high levels of both IL-6 and TNF-α. Because all assays were run in duplicate, it is unlikely that the values were due to measurement error. Nevertheless, we removed this case from the analytic sample and re-estimated the models. Substantive inferences surrounding prenatal IL-6 and infant and child outcomes did not change. However, the associations between prenatal CRP and TNF-α and infant outcomes attenuated, such that neither CRP nor TNF-α was significantly related to child cognitive ability (CRP: β = − .23, p = .10; TNF-α: β = − .12, p > .10) and TNF-α was no longer significantly related to negative affect (β = .23, p > .10). Results of sensitivity analyses are tabled in supplemental material.

4. Discussion

The purpose of this study was to examine prospective associations among maternal prenatal inflammation and infant neurodevelopmental outcomes. A secondary aim was to investigate whether the association between prenatal inflammation and child outcomes persists into the preschool years. The findings of this investigation point to a robust association between prenatal inflammation and infant cognitive ability, with elevated levels of maternal CRP, IL-6, and TNF-α associated with poorer infant cognition. We found less robust associations between prenatal inflammation and infant negative affect and sleep quality, as only TNF-α predicted these outcomes. Higher levels of TNF-α predicted greater infant negative affect and greater nighttime sleep ratio, although these associations did not hold up following the exclusion of one highly influential case. Neither CRP nor IL-6 were predictive of infant negative affect or sleep quality. Although there were no direct effects of prenatal inflammation on preschool EF, we found preliminary evidence of an indirect relationship between IL-6 and preschool EF ability via infant cognition, suggesting that prenatal inflammation may have long-lasting, cascading implications for child neurodevelopment.

Our findings of a consistent association between prenatal inflammation and infant cognitive ability are in line with previous human and animal research examining these relations, although the association between CRP and infant cognitive ability and between TNF-α and infant cognitive ability did not hold after excluding a highly influential case. Early epidemiological research has documented associations between prenatal infection, a putative cause of elevated inflammation, and various disorders of the central nervous system in offspring, including increased risk of schizophrenia and autism spectrum disorder [49]. Only recently have researchers established links between levels of inflammation and more subtle neurodevelopmental outcomes including aspects of infant and toddler cognition and behavior [14–16]. However, no study to date has documented impacts of inflammation on infant cognition in the first 6 months of life. The findings of the current investigation extend this line of work by showing consistent inverse associations between three prenatal inflammatory markers (CRP, IL-6, and TNF- α) and infant cognitive ability as measured by a common, standardized developmental assessment. The effect size of these relationships ranged from small to moderate, with the largest effect size observed for IL-6. Research has shown that maternal inflammatory cytokines, particularly IL-6, cross the placenta [50] and are subsequently related to perturbations in neonatal brain structure and circuitry [14–16]. Therefore, there is a putative biological pathway linking prenatal inflammation to child neurodevelopment. Interestingly, IL-6 and TNF- α are part of the acute-phase inflammatory response and subsequently trigger hepatic production of CRP—an acute phase protein that is considered a measure of low-grade systemic inflammation [51]. Different biomarkers may therefore be reflective of different maternal experiences and characteristics (e.g., obesity, stress, infection) and may come to impact child outcomes via different mechanistic pathways. It is noteworthy that despite these differences, all three inflammatory markers had similar associations with child cognition in the current study, potentially reflecting the coordinated effort of the maternal immune system during pregnancy [52].

After covariate adjustment, we did not observe direct associations between prenatal inflammation and child EF at age 4. However, we did observe evidence for a developmental cascade in which elevated maternal IL-6 predicted poorer infant cognitive ability, which in turn predicted poorer EF at age 4. The full indirect pathway was only marginally significant, which is not altogether surprising given the modest sample size of the current study. However, these findings do suggest that prenatal inflammation may have long-lasting implications for child neurodevelopment via their impact on foundational cognitive abilities. Previous research has similarly documented an indirect association between adverse prenatal experiences and later EF via infant cognitive ability [32]. These domains may therefore serve as early identification and intervention points for children exposed to elevated prenatal inflammation, who may be at risk for long-term neurodevelopmental difficulties. It is unknown whether the relationships we observed for EF would extend to other cognitive outcomes or to related neuropsychiatric diagnoses, although prior evidence suggests that elevated maternal inflammation may predict risk for ADHD during preschool and early childhood [19]. Thus, further research is needed to understand the full range of infant and child outcomes that are impacted by exposure to prenatal inflammation.

One unexpected finding was that mothers with higher prenatal circulating TNF-α had infants with more negative affect and poorer cognitive ability, but a greater nighttime sleep ratio. This suite of outcomes is potentially reflective of hallmark sickness behaviors associated with the production of pro-inflammatory cytokines in response to infection [53]. Specifically, pro-inflammatory cytokines like TNF-α and IL-1β have found to mediate signals between the immune and central nervous systems, conveying the need for non-specific physiological adaptations promoting survival from the immune system to the brain [54]. Impairments to cognitive functioning and behavior and increases in sleep would therefore reflect adaptive motivational changes meant to facilitate recovery [53]. Prior evidence in both animals and humans supports the hypothesis that TNF-α induces changes in sleep patterns. Treatment with TNF-α has been shown to increase lethargy in cancer patients [55] and non-rapid eye movement (NREM) sleep across several non-human species [56]. As some maternal prenatal cytokines are found to stimulate placental pro-inflammatory cytokine production [57], increased maternal prenatal pro-inflammatory cytokines like TNF- α may signal changes to the developing fetal brain which manifest as increased sickness behaviors in infancy. However, as these relationships were only found for TNF-α, additional research is needed to further probe the relationship between prenatal inflammation and infant sleep patterns.

Our study provides evidence for a link between prenatal inflammation and infant neurodevelopment in African American women and children, a group understudied in this research area. Incorporating sufficient diversity into studies of inflammation is particularly important as racial and ethnic minority groups may be doubly at risk for increased prenatal inflammation. Lifelong experiences of disproportionate socioeconomic disadvantage alongside elevated chronic stress from experiences such as discrimination put African American women at higher risk for allostatic overload and consequentially, elevated systemic inflammation, compared to White women [20]. Pregnant African American women, compared to other groups, have been shown to be at greater risk for poorer physical health and adverse birth outcomes and to have more limited access to healthcare and healthy foods [20,58]. As such, African American infants may be at higher risk for adverse neurodevelopmental outcomes that are prenatally programmed via elevated maternal stress and inflammation during pregnancy. Although our study does not compare prenatal inflammation and infant outcomes across racial and ethnic groups, we can compare our mean levels of inflammation in this study to those reported in other pregnancy samples. For example, in comparison to a prior study of a majority-White sample, participants in our study had considerably higher mean levels of IL-6 (5.65 pg/mL vs 1.11 pg/mL) and TNF-α (8.22 pg/mL vs 0.57 pg/mL [46]). Future studies could explicitly compare levels of inflammation across racial and ethnic groups as well as examine predictors of individual differences in inflammation among African American women in order to identify potentially modifiable risk factors. It is also important to note that the pernicious impact of socioeconomic disadvantage and discrimination on African American mothers and children continue to operate postnatally, potentially contributing to child neurodevelopment via additional mechanisms that we did not explore here.

While our novel findings extend the inflammation literature in several ways, it is important to note several limitations. Most notably, a primary aim of this study was to determine whether the link between prenatal inflammation, child cognition, and EF found in White/European-American and Hispanic mothers and children would generalize to African American dyads. However, we were unable to directly compare results across race and ethnicity because our study only included an African American sample. Additionally, our small sample size limited our power to detect associations with small effect sizes as well as our ability to look at additive or interactive effects of multiple inflammatory markers. Future research might investigate whether systemic and acute inflammation are uniquely associated with child neurodevelopment, or whether elevated systemic inflammation alters the impact of acute inflammation on child outcomes. Our sensitivity analyses examining potential outliers or influential cases give us confidence in our findings despite the small sample size, as does the fact that our mean levels of inflammation were in line with those reported previously in the literature [46,47].

We also did not comprehensively assess maternal health at the time that we collected blood samples, nor many other factors that could impact inflammation levels (e.g., use of non-steroidal anti-inflammatory medications, sickness, injury, nutrition). Therefore, we cannot rule out the impact of these confounders and our findings should be replicated in a larger, more controlled investigation. Though we included maternal education as an indicator of SES as a covariate in all models, future studies may benefit from additionally including a measure of maternal income or poverty. Other factors indicative of maternal and child nutrition that were not able to be considered here may also be included in future studies as covariates due to their associations with child cognitive and behavioral outcomes, including aspects of the mother and child’s diet (e.g., fatty acids, micronutrients, duration of breastfeeding). The pilot nature of our study and the limited sample size precluded our ability to test these and other potentially relevant covariates.

5. Conclusions

In sum, our findings show a consistent association between elevated prenatal inflammation and poorer infant cognitive ability, with implications for downstream EF. We demonstrated these associations in a sample of African American women and their children, an important but understudied group in the prenatal inflammation literature. These findings inform research focusing on underlying causes of health and achievement disparities in children of color, while also advancing mechanistic research that seeks to explain how prenatal inflammation comes to impact lifelong health and developmental outcomes. Early markers of central nervous system integrity may prove an important point of identification and intervention for children exposed to elevated prenatal inflammation, potentially preventing long-term adverse outcomes.

Data Availability

Data are available by request to the corresponding author.