Cortical Thinning and Neuropsychiatric Outcomes in Children Exposed to Prenatal Adversity: A Role for Placental CRH?

Ninety-seven mother-child dyads participated in a longitudinal study of the effect of fetal exposure to maternal stress hormones on child MRI measures (see the CONSORT diagram in Figure S1 in the data supplement that accompanies the online edition of this article; see also Tables S1A–C in the data supplement). Women provided informed consent to provide blood samples at five intervals during gestation: 13.5–16.6 weeks (mean=15.3 weeks), 17.8–20.5 weeks (mean=19.2 weeks), 23.7–26.5 weeks (mean=24.9 weeks), 29.9–32.3 weeks (mean=30.9 weeks), and 34.6–38.1 weeks (mean=35.9 weeks). All women were English-speaking healthy adult (>18 years of age) pregnant women with singleton intrauterine pregnancies. Women were excluded if they had multiple births; used tobacco, alcohol, or others drugs during pregnancy; had uterine or cervical abnormalities; or had any conditions associated with dysregulated neuroendocrine function. Their children (48 boys, 49 girls) were enrolled in the study at 6–9 years of age (mean=7.3 years [SD=0.91]).

Gestational age at testing was determined by last menstrual period and was confirmed by obstetric ultrasonographic biometry before 20 weeks. Maternal blood samples (20 mL) were collected serially at five intervals throughout pregnancy by antecubital venipuncture into siliconized EDTA (purple top) Vacutainers and then immediately chilled to 60°C. Samples were centrifuged at 2,000 g for 15 minutes, decanted into polypropylene tubes prepared with aprotinin (Sigma Chemical, St. Louis; 500 KIU/mL blood), and stored at −80°C until assayed.

Following previously reported methods (40), the concentration of total maternal pCRH was determined by radioimmunoassay (Bachem Peninsula Laboratories, San Carlos, Calif.). The CRH assay had less than 0.01% cross-reactivity with ovine CRH, 36% cross-reactivity with bovine CRH, and nondetectable reactivity with human ACTH. The intra- and interassay coefficients of variance ranged from 5% to 15%. The minimum detectable dose of the assay is 2.04 pg/mL (see the online data supplement for additional assay details).

Data reduction for the radioimmunoassay was conducted with a computer-assisted four-parameter logistics program (41). Values exhibiting greater than 25% error (deviation from the standard curve) were not included in the analyses. A subset of samples (N=60) was sent to a clinical laboratory (Quest Diagnostics) for further validation. The correlation between the two sets of data was 0.87 (p<0.01). Shared variance between 19- and 31-week CRH samples was a modest 14% (r=0.38). Mother-child dyads were recruited from two separate studies with identical prenatal assessments. To ensure comparability between the cohorts for analysis, the pCRH values were standardized within the two groups and then combined for statistical analysis (see the data supplement for details). As expected, pCRH levels increased geometrically as gestation advanced (see Figure S4 and Table S9 in the data supplement).

All children (mean age, 87.3 months [SD=10.9]) had a stable neonatal course (median Apgar score, 9 [range, 8–10]; gestational age at birth, 39.2 weeks [SD=1.5]) and were without known neonatal illness or congenital, chromosomal, or genetic anomalies. Participants had no evidence of neurological abnormalities in the newborn period. Children’s structural MR images were assessed for normal anatomical appearance. Seven children with motion artifacts and three with abnormal scans were not included in the final sample (N=97). These 10 children did not differ significantly on any measure from those whose scans were usable (see Table S8 in the data supplement). At 6–9 years of age, no physical conditions were reported by the parents in a structured interview format of the MacArthur Health and Behavior Questionnaire (42). The majority of children (88%) were right hand dominant (Edinburgh Handedness Inventory) (43).

The structural MRI scan was acquired with a 3-T Philips Achieva system. Children received earplugs and watched a movie while in the scanner to increase compliance and minimize movement. A high-resolution T₁ anatomical scan was acquired in the sagittal plane with 1 mm³ isotropic voxel dimensions. An inversion-recovery spoiled gradient recalled acquisition sequence with optimal parameters was applied: TR=11 ms, TE=3.3 ms, TI=1,100 ms, turbo field echo factor=192, number of slices: 150, no sensitivity encoding (SENSE) acceleration, flip angle=180°, shot interval (time from inversion pulse to the center of acquisition)=2,200 ms. The images were reviewed by the MRI operator (who was unaware of any of the study parameters) immediately after the scan was completed. If there were visible signs of motion artifacts, the subject was asked to stay for an additional scan. If the subject agreed, a second scan was acquired.

Cortical surface reconstruction and volumetric segmentation were performed with FreeSurfer (http://surfer.nmr.mgh.harvard.edu/). Streamlined image processing procedures included application of intensity normalization prior to segmentation to minimize errors in identifying the boundaries (44); removal of nonbrain tissues (45); and transformation of images into Talairach space. Pial and white matter surfaces were located by finding the highest intensity gradient. Surface inflation was applied to each individual brain (46), and the inflated brains were registered to a spherical atlas. Cortical thickness was the closest distance from the gray matter or white matter surface to the pial surface at each vertex on the tessellated surface (47).

Smoothing of images was done prior to regression. The command line interface was used for false discovery rate corrections and to accommodate more than one nuisance variable. Gestational age at birth, birth weight, age of the child at testing, sex, and handedness were included as covariates. A threshold of p<0.05 was used for all statistical tests, and outcomes were corrected for multiple comparisons using false discovery rate.

After false discovery rate corrections and spatial normalization, the number of vertices that were significantly associated with pCRH was determined in each region of the cortex. The numbers of significant vertices were summed for each region and divided by the total number of vertices in that area to provide the percentage of vertices that were significantly associated with pCRH concentrations. The same procedure was used for the number of significant vertices in each lobe. For hemispheric and whole brain percentages, the procedure was the same except that the total number of subcortical vertices was subtracted from the total.

We conducted a battery of tests that interrogate several brain regions and circuits to obtain a relatively broad assessment of cognitive and emotional function.

The Child Behavior Checklist (48) is one of the most widely used parental interviews for identifying affective and conduct disorder problems in children. Structured interviews (rather than a questionnaire) were administered to mothers about their children’s behavior. We focused on two inclusive scales; one that assessed internalizing problems (sum of anxious-depressed, withdrawn-depressed, and somatic-complaints scores) and one that evaluated externalizing problems (sum of rule-breaking and aggressive behavior problem scores). Responses were recorded on a Likert-type scale ranging from 0 (not true) to 2 (very true or often true). Scores were summed, and standardized scores were computed that are age and sex specific.

The flanker task is an executive function task that requires the ability to resolve conflicts when competing information is present (49). Participants view five arrows arrayed horizontally on a screen and are instructed to press a left or right response button based on the direction of the center arrow (target). They are instructed to ignore the surrounding arrows, which are either congruent (all aligned in the same direction) or incongruent with the center arrow. This task consists of 24 congruent trials and 24 incongruent trials. Each set of arrows is presented until the child responds (maximum of 5,000 ms), with a 750-ms intertrial interval. Among the scores, median reaction time to targets with incongruent distractors correlates most highly with other indexes of performance derived from this test (all r values >0.70). The association between externalizing scores on the Child Behavior Checklist and the median reaction time on the flanker task was not significant (r=0.09, p=0.41).

Significant associations were observed between fetal exposure to pCRH averaged across gestation and areas of cortical thinning in children, as illustrated in the pial maps in Figure 1. Children exposed to high levels of pCRH throughout fetal life exhibited significant thinning in 12% of the whole cortical mantle (see Table S2A in the data supplement). The anatomical distribution of the cortical thinning associated with CRH exposure involved the left and right hemispheres equally (12% and 11% of hemisphere reduced, respectively) (see Table S2A in the data supplement). Regional analyses indicated that cortical thinning associated with pCRH levels was primarily in the temporal (15%) and frontal (16%) regions (Table 1). Panels B–D in Figure 1 are illustrative scatterplots of the associations between total fetal exposure to pCRH across gestation and cortical thinning in the frontal and temporal areas (all significant, p<0.01–0.001). The representative scatterplots, coupled with the data in Table 1, highlight the widespread association between fetal exposures to pCRH and regional cortical thinning, which is most notable in the temporal and frontal regions. Average concentrations of pCRH across gestation or levels at any gestational interval were not associated with increased cortical thickness either globally or in any region (see Table S2B in the data supplement). Focusing on two gestational ages, 19 weeks (early, when pCRH production begins to accelerate) and 31 weeks (late, the time strongly associated with preterm birth), we found that maternal levels of pCRH were significantly associated with cortical thinning.

Fetal exposure to pCRH early in gestation (19 weeks) was associated with significant thinning of the frontal poles. The significant association between pCRH and cortical thinning was bilateral but strongest in the right (B=−0.14, p<0.001; 78% of the structure) compared with the left frontal pole (B=−6.88, p<0.05; 69% of the structure) (see Table S3 in the data supplement). In Figure 2, pial maps depict areas of significant thinning linked with fetal pCRH exposure early in gestation, and the scatterplots illustrate the regions of the frontal cortex with the strongest associations. (Findings were essentially unchanged with log transformed pCRH values; see Table S5 in the data supplement.) Nearly identical associations were observed in a smaller subsample (N=57) of children exposed to high pCRH levels measured even earlier, at 15 weeks’ gestation (see Figure S3A–D in the data supplement).

Thinning of the frontal pole has been associated with externalizing behaviors (50), defined as actions that direct energy outward and tend to harm others (51). We tested whether the effects of pCRH exposure on cortical thinning contributed to the development of these behaviors (Figure 2E). Our model supported the belief that reduced cortical volume in the frontal pole, associated with elevated fetal exposure to concentrations of pCRH at 19 weeks’ gestation, contributes to externalizing symptoms in 6- to 9-year-old children (indirect effect=0.93; 95% bias-corrected confidence interval [BCCI]=0.17, 2.05; p<0.05). Neither internalizing problems nor reaction time on the flanker task was associated with cortical thinning in this region.

Exposure to elevated levels of pCRH later in gestation (31 weeks) was associated with cortical thinning in the lateral temporal and paracentral regions (Figure 3A). The effect was localized to the left paracentral region and was bilateral in the temporal cortex. Remarkably, 98% of the lateral surface of the right temporal pole and 66% of the left temporal pole was significantly thinner in children exposed to high levels of pCRH at 31 weeks’ gestation (see Table S4 in the data supplement). The scatterplots in Figure 3 illustrate these regional associations. (Similar associations were observed at 25 weeks’ gestation in a smaller subsample [N=50].)

Structures comprising the temporal cortex subserve numerous cognitive and emotional functions. Temporal cortical thinning has been reported in individuals with attentional deficits (52), especially in the right temporal pole (53), which is active in tasks requiring attention to relevant stimulation (54). We employed a statistical model (55) to test whether the effects of exposure to pCRH late in prenatal life on cortical thinning contributed to impairment of attention in 6- to 9-year-olds. The model (Figure 3E) indicated that the reduced right temporal pole volume associated with exposure to pCRH at 31 weeks’ gestation may partially account for poorer performance on a visual processing and sustained attention test (indirect effect=42.11; 95% BCCI=4.27, 128.31; p<0.01) (56, 57).

Exploratory analyses suggest significant sex differences in the nature and degree of association between pCRH levels and global as well as regional cortical thinning. At both 19 and 31 weeks’ gestation, the associations were stronger in girls (see Figure S2 in the data supplement). The association between fetal exposure to pCRH at 19 weeks’ gestation and cortical thinning involved most cortical areas in girls (see Figure S2A) but minimal areas in boys (see Figure S2B). Fetal exposure to pCRH at 31 weeks affected cortical thinning globally in boys (see Figure S2C) but locally in the temporal pole in girls (see Figure S2D; similar to the findings for the combined sexes).