Maternal psychological distress during the COVID-19 pandemic and structural changes of the human fetal brain.

Background: Elevated maternal psychological distress during pregnancy is linked to adverse outcomes in offspring. The potential effects of intensified levels of maternal distress during the COVID-19 pandemic on the developing fetal brain are currently unknown.
Methods: We prospectively enrolled 202 pregnant women: 65 without known COVID-19 exposures during the pandemic who underwent 92 fetal MRI scans, and 137 pre-pandemic controls who had 182 MRI scans. Multi-plane, multi-phase single shot fast spin echo T2-weighted images were acquired on a GE 1.5 T MRI Scanner. Volumes of six brain tissue types were calculated. Cortical folding measures, including brain surface area, local gyrification index, and sulcal depth were determined. At each MRI scan, maternal distress was assessed using validated stress, anxiety, and depression scales. Generalized estimating equations were utilized to compare maternal distress measures, brain volume and cortical folding differences between pandemic and pre-pandemic cohorts.
Results: Stress and depression scores are significantly higher in the pandemic cohort, compared to the pre-pandemic cohort. Fetal white matter, hippocampal, and cerebellar volumes are decreased in the pandemic cohort. Cortical surface area and local gyrification index are also decreased in all four lobes, while sulcal depth is lower in the frontal, parietal, and occipital lobes in the pandemic cohort, indicating delayed brain gyrification. Conclusions: We report impaired fetal brain growth and delayed cerebral cortical gyrification in COVID-19 pandemic era pregnancies, in the setting of heightened maternal psychological distress. The potential long-term neurodevelopmental consequences of altered fetal brain development in COVID-era pregnancies merit further study.

Introduction

Intrauterine programming refers to early developmental responses to environmental exposures that in turn may influence an individual’s lifelong health[1]. The timing, duration, and severity of fetal exposures may adversely impact tissue and organ system development through multiple pathways, including nutrition, oxygen supply, inflammatory changes, dysregulated hormonal exposure, and epigenetic changes[2]. The fetal brain is especially sensitive to such changes, and it is increasingly recognized that developmental and neuropsychiatric conditions manifesting later in life have their origins in the fetal period[3,4]. Several studies have shown that prenatal exposure to maternal psychological distress results in structural and functional changes in brain development of young children through school age, including regional changes in surface area, gray matter and amygdala volumes along with cortical thinning[5-7]. Furthermore, emerging evidence links these structural differences in brain development with neurobehavioral function in children and adolescents[8,9]. However, this body of research also highlights the challenges in distinguishing the effects of prenatal from postnatal exposures with the potential cumulative impact of prolonged exposures across extensive periods of development. Given the impact of not only the presence, but the timing, severity, and duration of adverse prenatal exposures on the developing brain, the ability to precisely characterize fetal brain development represents an advance in the field. Recent studies have demonstrated an association between maternal psychological distress and altered structural and functional development of the fetal brain[10-14], allowing for an enhanced understanding of prenatal mental health exposures on later neuropsychological function in offspring.

The Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) responsible for COVID-19 (coronavirus disease 2019) pandemic was first identified in Wuhan, China in 2019 and continues to exact widespread morbidity and mortality across the globe[15]. During the pandemic, elevated levels of depression, anxiety, post-traumatic stress and psychological distress[16-18], have been reported. In pregnant women, concerns around potential fetal COVID-19 exposure, as well as social isolation, food and housing insecurity, unemployment, and inequitable access to health care, play an important role in elevated pregnancy-related psychological distress[19]. Specific to pregnant women, both rates of anxiety and depression have increased, along with heightened symptomology of clinical mental health conditions[20-22]. One early study suggests that prenatal maternal distress during the COVID-19 pandemic may decrease amygdala-prefrontal connectivity in infants at 3 months of age, particularly in settings of lower social support[23]. The psychosocial impact of this pandemic on fetal brain development, however, remains largely under-reported. The objective of this investigation is to determine the effects of maternal mental health on in vivo human fetal brain development during the COVID-19 pandemic. Our overarching hypothesis is that heightened maternal stress, depression, and anxiety during the COVID-19 pandemic will adversely influence fetal brain growth and development, even in the absence of confirmed COVID-19 exposure. Our results show that maternal stress and depression are significantly higher in the pandemic cohort, compared to the pre-pandemic cohort. We also demonstrate decreased regional fetal brain volumes and delayed brain gyrification in the pandemic cohort.

Methods

Study participants

This study involved two sequential enrollments: (1) 137 healthy pregnant women from March 2014 to February 2020 (“pre-pandemic”); (2) 65 women without known COVID-19 exposures from June 2020 to April 2021 (“pandemic”) for a fetal brain magnetic resonance imaging (MRI) study from low-risk obstetrical community hospitals in Washington, DC (Supplementary Fig. 1). The first enrollment period was part of a longitudinal study of normative fetal brain development in low-risk obstetric patients, and the second was a natural history observational study of fetal brain development during the COVID-19 pandemic. Study procedures were identical across both enrollment periods. Healthy pregnant volunteers had a normal prenatal history that included normal screening, laboratory, and ultrasound studies. Exclusion criteria were multiple gestation pregnancy, known or suspected congenital infection, syndromic or dysgenetic features in the fetus, documented chromosomal abnormalities, or any maternal contraindication to MRI. Similarly, subjects reporting the use of medications or substances other than prenatal vitamins or supplements were excluded (e.g., prescribed medications, tobacco, marijuana, or alcohol use). Enrolled fetuses found to have structural (encephaloclastic or dysgenetic) brain abnormalities on fetal MRI, or postnatal confirmation of a genetic syndrome were subsequently excluded from the study. Parental education and employment data were collected from each participant during study visits. Following approval by the Institutional Review Board at Children’s National Hospital (Protocol 1373 for the pre-pandemic cohort, approved on January 9, 2011, and Protocol 14257 for the pandemic cohort, approved on May 1, 2020), written informed consent was obtained from all participants.

Maternal distress

Four well-validated maternal distress measures were completed by each pregnant woman on the day of the MRI, including the Spielberger State Anxiety Inventory (SSAI, range: 20 to 80)[24], Spielberger Trait Anxiety Inventory (STAI, range: 20 to 80)[24], Perceived Stress Scale (PSS, range: 0 to 40)[25], and Edinburgh Postnatal Depression Scale (EPDS, range: 0 to 30)[26]. Values higher than the following thresholds were considered elevated: maternal state anxiety >40, trait anxiety >40, stress > 15 and depression >10[25,27-29].

MRI data acquisition

Multi-plane multi-phase single shot fast spin echo (SSFSE) T2-weighted images for fetal brain were acquired on a 1.5 Tesla GE Discovery MR450 scanner (GE Healthcare, Milwaukee, WI, USA) using an eight-channel surface receiver coil (USAI, Aurora, OH). The following acquisition parameters were used: echo time = 160 ms; repetition time = 1100 ms; field of view = 320 × 320 mm2; matrix = 256 × 256; 2 mm slice thickness and 50 to 70 consecutive slices for full fetal brain coverage in the axial, coronal, and sagittal planes for a final in-plane resolution of 1.25 × 1.25 mm2. Each subject was scanned up to two time points in the fetal period.

Image processing

Motion correction was first conducted on fetal brain T2-weighted multi-plane images using the slice-to-volume registration method[30]. This procedure reduced interslice motion artifacts and provided images with enhanced contrast and resolution, and coherent anatomic boundaries in 3D space. 3D brain images with severe motion artifacts that affected the ability to distinguish brain tissues such as cortical gray matter (CGM), white matter (WM), lateral ventricles, brainstem and cerebellum were excluded from the analysis. Automatic segmentation of the brain tissues was then implemented using the Developing Brain Region Annotation with Expectation-Maximization (Draw-EM) algorithm[31,32]. Draw-EM utilizes Expectation-Maximization (EM) algorithm to segment a brain into different tissue types as well as detailed structures of the brain[31]. Two sets of tissue labels were generated from Draw-EM: the segmentation file with 9 labels[31] and the parcellation file with 50 labels[32]. Manual correction of tissue labels of the segmentation and parcellation files was performed by a trained research team member (K.K.), who had more than 5 years of experience using ITK-SNAP in fetal brain segmentation during the time of this work[33]. Fifty-five scans (20%) were randomly chosen and segmented by a second trained examiner (N.R.A.) to evaluate the inter-rater reliability. The intraclass correlation coefficients for all measured regions between the two examiners were higher than 0.95.

Ten brain regions of both the right and left hemispheres were extracted from segmentation and parcellation files (Supplementary Fig. 2): the frontal, parietal, temporal, and occipital lobes, anterior and posterior cingulate gyrus, insula, and corpus callosum were extracted from the parcellation file, and the deep gray matter (DGM) and ventricles from the segmentation file. These brain regions were imported to BrainSuite version 18a to generate 3D surface mesh models[34]. Each mesh model contained a set of 3D coordinates of the surface vertices and a set of triangular mesh. Every surface vertex was associated with one of these 10 brain regions.

Fetal brain volumes and cortical folding

Brain tissue volumes from the segmentation file were determined based on the voxel sizes of the images, including CGM, WM, DGM, cerebellum, brainstem, and hippocampus (Supplementary Fig. 2d–f)[35].

To characterize 3D fetal brain morphology, three cortical features, including the surface area, local gyrification index, and sulcal depth, were measured on the brain surface of the four brain lobes (frontal, parietal, temporal, and occipital lobes) (Supplementary Fig. 2g–l)[36-39]. Areas of the four lobes of WM surface were calculated as the summation of the areas formed by the triangular mesh[40]. To calculate local gyrification index and sulcal depth, convex hull surface of the 10 brain regions was first created[41]. Local gyrification index quantifies the amount of cortex buried within the sulcal folds, representing the extent of cortical folding. For each vertex on the surface, the local gyrification index is defined as the ratio between the area of a circular region of the vertex on the surface and the corresponding area on the convex hull for the vertex[42-44]. The sulcal depth was computed as the distance from each vertex on the brain surface to the nearest point on a convex hull for each hemisphere[45]. The surface area, local gyrification index and sulcal depth were calculated on the inner surface of the CGM (i.e., the gray and white junction)[36,37,46,47].

Statistics and reproducibility

Demographic data are presented as frequency and percent or median and quartile (25th, 75th), where appropriate. Data were explored for departures from normality using the Shapiro-Wilks test. Non-normally distributed parameters included gestational age (GA), maternal age, birth weight, birth head circumference, Apgar score, and maternal distress measures (i.e., stress, anxiety, depression). The fetal and maternal demographics were therefore compared between pre-pandemic and pandemic cohorts using non-parametric tests such as Wilcoxon-Mann-Whitney tests for continuous data and using Chi-square tests for categorical data.

Given that some mothers had repeated scans and thereby presented correlated data, we chose to use separate generalized estimating equations (GEEs) to examine fetal brain tissue volumes and cortical features in association with cohort status. GEE is a robust statistical method employed to study population-averaged patterns or trends over time for longitudinal data, allowing for multiple measurements per subject[48]. If an individual fetus was scanned at two time points in the fetal period, then both scans (if successful) were included for data analysis. Our modeling strategy was as follows. First (“Step 1”), we examined with separate models the associations between cohort status (pre-pandemic: 0 [referent]; pandemic: 1) and maternal distress measures (SSAI, STAI, PSS, and EPDS), adjusted for GA (weeks) at MRI and fetal sex[48,49]. In addition, the distress measures were further compared between pre-pandemic and pandemic cohorts in the low and high distress groups based on their corresponding threshold (SSAI: 40; STAI: 40; PSS: 15; EPDS: 10). Therefore, a total of 12 GEE models were implemented. Second (“Step 2”), separate GEE models were utilized to assess the associations between cohort status and fetal brain tissue volumes (i.e., brain tissue volumes for the six brain tissues) and cortical features (i.e., surface area, local gyrification index, and sulcal depth), controlled for GA at MRI (weeks), fetal sex, and each maternal distress measure (as a continuous variable) within each GEE model to determine whether cohort status was associated with fetal brain tissue volumes and cortical features. Specifically, 18 GEE models were adjusted for cohort status, GA at MRI, and fetal sex to determine the differences in each brain region between pre-pandemic and pandemic cohorts, with an additional 72 models that were further adjusted for each maternal distress measure, for a total of 90 GEE models implemented. Lastly (“Step 3”), the entire cohort was separated into high distress and low distress groups for each maternal distress measure based on published cut points (i.e., 40 for anxiety[24], 15 for stress[25], or 10 for depression)[29] for those significant maternal distress measures for all subjects found in Step 1, and separate GEE models were conducted to investigate the association between cohort status and fetal brain tissue volumes and global cortical features (i.e., combined measures of the four lobes) in each group following adjustment for GA at MRI and fetal sex. Therefore, a total of 36 GEE models were implemented, where 24 models were for brain tissue volumes and 12 models were for cortical features.

Sub-analyses of other potential confounders were also implemented. First, we conducted GEE analyses for the associations between fetal brain volumes/brain cortical features and each maternal distress measure (treated as a continuous variable), adjusting for fetal sex and GA at MRI for all subjects (including both pre-pandemic and pandemic cohorts). A total of 72 GEE models were implemented. Second, we conducted the analysis of the GA-cohort status interaction for the brain cortical features, by further adjusting the GA-cohort status interaction in the GEE models in “Step 2” above, without adjusting for the maternal distress measures. A total of 12 GEE models were implemented. Third, two sensitivity analyses were conducted: (1) exclusion of scans performed below 28 weeks gestation and (2) exclusion of mothers greater than 40 years of age as potential outlier. A total of 180 GEE models were implemented (36 GEE models without adjustment for maternal distress measures and 144 GEE models with adjustment for maternal distress measures). Fourth, we evaluated potential differences in laterality by fitting the GEEs by the two hemispheres to investigate the effect of the cohort status on the fetal brain volumes/brain cortical features, adjusting for GA at MRI and fetal sex. A total of 36 GEE models were implemented. Lastly, the effect of parental education and employment on the fetal brain volumes/brain cortical features was examined. A total of 72 GEE models were implemented.

For demographic pre-pandemic vs. pandemic comparisons, statistical significance was assumed for p < 0.05, two-tailed. All subsequent p values were also adjusted for multiple testing using the false discovery rate method based on the number of outcomes (6 tissues or 4 lobes)[50]. All analyses performed in this study were conducted using MATLAB R2019a (The MathWorks, Inc., Natick, MA, USA)[48].

Table 1

Demographics of 202 pregnant women who underwent 274 prenatal MRI studies.

N [%] or Median [IQR]	All Subjects	Pre-pandemic	Pandemic	p
Number of subjects	202	137	65
Female fetus	91 [48]	58 [45]	33 [54]	0.22
Male fetus	100 [52]	72 [55]	28 [46]
With 1 scan	130 [64]	92 [67]	38 [58]	0.23
With 2 scans	72 [36]	45 [33]	27 [42]
Number of MR scans	274	182	92
Time point 1	147 [54]	101 [55]	46 [50]	0.39
Time point 2	127 [46]	81 [45]	46 [50]
GA at MRI	30.4	30.2	30.8	0.15
	[26.1, 35.3]	[27.0, 35.9]	[25.3, 34.1]
Maternal age (years)*	34.1	34.0	34.6	0.63
	[31.0, 36.9]	[31.0, 37.8]	[30.5, 36.1]
Maternal parity (primiparous/multiparous)†	109/80	73/55	36/25	0.80
GA at birth (weeks)‡	39.6	39.7	39.4	0.51
	[38.7, 40.4]	[38.4, 40.6]	[38.9, 40.1]
Birth weight (kg)§	3355	3382	3285	0.91
	[3078, 3677]	[3079, 3670]	[3064, 3690]
Birth head circumference (cm)#	34.3	34.0	34.5	0.19
	[33.5, 35.5]	[33.0, 35.0]	[33.7, 35.6]
Apgar score at 1 min††	8	8	8	0.66
	[8, 9]	[8, 9]	[8, 9]
Apgar score at 5 min‡‡	9	9	9	0.32
	[9, 9]	[9, 9]	[9, 9]
Delivery Mode§§				0.10
Vaginal	101 [68]	59 [62]	42 [78]
Elective C-section	26 [17]	18 [19]	8 [15]
Emergency C-section	22 [15]	18 [19]	4 [7]
Race/Ethnicity##				0.07
White	100 [57]	68 [56]	32 [58]
Black	33 [19]	25 [21]	8 [15]
Hispanic or Latino	20 [11]	12 [10]	8 [15]
Asian or Pacific Islander	10 [6]	4 [3]	6 [11]
Others	13 [7]	12 [10]	1 [2]

N = 274 MR scans for volumetric data, and N = 204 MR scans for cortical features. GA Gestational age. IQR Interquartile range. *: Based on 200 (99%) subjects. †: Based on 189 (94%) subjects. ‡: Based on 165 (82%) subjects. §: Based on 156 (77%) subjects. #: Based on 108 (53%) subjects. ††: Based on 146 (72%) subjects. ‡‡: Based on 146 (72%) subjects. §§: Based on 149 (74%) subjects. ##: Based on 176 (87%) subjects.

Table 2

The results of the generalized estimating equations (GEEs) for the associations between fetal brain volumes/brain cortical features and cohort status (0: pre-pandemic; 1: pandemic), adjusting for fetal sex, gestational age at MRI (weeks) and each maternal distress measure.

	Brain volumes (mm3)
	SSAI		STAI		PSS		EPDS
	β	p	β	p	β	p	β	p
CGM	660	0.64	734	0.60	817	0.55	349	0.81
WM	−5154	0.01*	−4903	0.01*	−5096	0.01*	−5479	0.01*
DGM	−92	0.71	−63	0.80	3.8	0.99	−143	0.56
Hippocampus	−76	<0.01*	−74	<0.01*	−67	<0.01*	−72	<0.01*
Cerebellum	−468	0.02*	−442	0.03	−412	0.05	−484	0.02*
Brainstem	1.9	0.97	8.1	0.86	−9.3	0.85	−6.3	0.89
	Lobe Surface Area (mm2)
	SSAI		STAI		PSS		EPDS
	β	p	β	p	β	p	β	p
Frontal	−460	<0.01*	−455	<0.01*	−478	<0.01*	−476	<0.01*
Parietal	−463	<0.01*	−460	<0.01*	−472	<0.01*	−474	<0.01*
Temporal	−428	<0.01*	−429	<0.01*	−446	<0.01*	−441	<0.01*
Occipital	−258	<0.01*	−254	<0.01*	−274	<0.01*	−268	<0.01*
	Local Gyrification Index (×10−3)
	SSAI		STAI		PSS		EPDS
	β	p	β	p	β	p	β	p
Frontal	−69	<0.01*	−68	<0.01*	−69	<0.01*	−68	<0.01*
Parietal	−117	<0.01*	−116	<0.01*	−117	<0.01*	−117	<0.01*
Temporal	−76	<0.01*	−77	<0.01*	−83	<0.01*	−80	<0.01*
Occipital	−73	<0.01*	−73	<0.01*	−76	<0.01*	−74	<0.01*
	Sulcal Depth (×10−3 mm)
	SSAI		STAI		PSS		EPDS
	β	p	β	p	β	p	β	p
Frontal	−113	0.04	−119	0.04*	−123	0.04*	−123	0.03*
Parietal	−157	0.02*	−167	0.02*	−181	0.01*	−181	0.01*
Temporal	−73	0.15	−80	0.13	−96	0.08	−87	0.09
Occipital	−113	0.01*	−115	0.01*	−123	0.01*	−125	0.01*

The β and p represent the coefficient and its significance of the cohort status in each GEE. SSAI Spielberger State Anxiety Inventory. STAI Spielberger Trait Anxiety Inventory. PSS Perceived Stress Scale. EPDS Edinburgh Postnatal Depression Scale. CGM Cortical gray matter. WM White matter. DGM Deep gray matter. Bold p: p < 0.05. *: q < 0.05.

Table 3

Comparisons of brain tissue volumes (least squares mean ± standard error) between pre-pandemic and pandemic cohorts from the generalized estimating equations for the associations between fetal brain volumes (cm3) and cohort status (0: pre-pandemic; 1: pandemic), adjusting for gestational age at MRI (weeks) and fetal sex.

	Pre-pandemic	Pandemic	p
Low PSS
CGM	59.6 ± 4.0	58.3 ± 4.2	0.41
WM	98.5 ± 6.6	91.1 ± 7.0	<0.01*
DGM	15.4 ± 1.0	15.4 ± 1.0	0.99
Hippocampus	1.16 ± 0.08	1.06 ± 0.08	<0.01*
Cerebellum	8.76 ± 0.92	8.23 ± 0.95	0.04
Brainstem	3.94 ± 0.13	3.91 ± 0.14	0.61
High PSS
CGM	62.5 ± 6.7	64.7±7.0	0.30
WM	99.9 ± 9.7	98.2 ± 10.3	0.60
DGM	15.6 ± 1.1	15.9±1.1	0.25
Hippocampus	1.05 ± 0.07	1.01 ± 0.08	0.22
Cerebellum	8.87 ± 1.55	8.38 ± 1.59	0.13
Brainstem	4.05 ± 0.20	4.06 ± 0.21	0.86
Low EPDS
CGM	60.5 ± 3.6	61.0 ± 3.9	0.74
WM	100.0 ± 5.9	95.0 ± 6.3	0.02
DGM	15.6 ± 1.0	15.7 ± 1.0	0.63
Hippocampus	1.14 ± 0.06	1.06 ± 0.06	<0.01*
Cerebellum	8.88 ± 0.83	8.44 ± 0.87	0.06
Brainstem	3.99 ± 0.11	4.01 ± 0.12	0.65
High EPDS
CGM	58.7 ± 13.7	55.6 ± 14.0	0.31
WM	89.9 ± 20.9	84.2 ± 21.6	0.29
DGM	14.7 ± 3.2	14.3 ± 3.2	0.51
Hippocampus	0.94 ± 0.11	0.90 ± 0.12	0.34
Cerebellum	7.59 ± 2.50	6.91 ± 2.54	0.16
Brainstem	3.78 ± 0.35	3.60 ± 0.37	0.12

CGM Cortical gray matter. WM White matter. DGM Deep gray matter. PSS Perceived Stress Scale. EPDS Edinburgh Postnatal Depression Scale. Bold p: p < 0.05. *: q < 0.05.

Table 4

Comparisons of brain lobe surface area (cm2)/LGI/sulcal depth (mm) (least squares mean ± standard error) between pre-pandemic and pandemic cohorts from the generalized estimating equations for the associations between cohort status (0: pre-pandemic; 1: pandemic) and brain lobe surface area/LGI/sulcal depth of combined four brain lobes, adjusting for gestational age at MRI (weeks) and fetal sex.

	Pre-pandemic	Pandemic	p
Low PSS
Surface Area	162.0 ± 10.2	145.4 ± 10.9	<0.01*
Local Gyrification Index	1.28 ± 0.06	1.20 ± 0.06	<0.01*
Sulcal Depth	1.60 ± 0.23	1.49 ± 0.24	0.07
High PSS
Surface Area	160.6 ± 16.8	143.3 ± 17.1	<0.01*
Local Gyrification Index	1.30±0.12	1.20 ± 0.12	<0.01*
Sulcal Depth	1.66 ± 0.38	1.50 ± 0.38	0.02*
Low EPDS
Surface Area	161.1 ± 9.1	146.2 ± 9.5	<0.01*
Local Gyrification Index	1.28 ± 0.06	1.20 ± 0.06	<0.01*
Sulcal Depth	1.61 ± 0.20	1.51 ± 0.21	0.045*
High EPDS
Surface Area	154.5 ± 32.3	134.4 ± 32.9	<0.01*
Local Gyrification Index	1.27 ± 0.24	1.16 ± 0.25	0.02*
Sulcal Depth	1.58 ± 0.70	1.34 ± 0.72	0.12

PSS Perceived Stress Scale. EPDS Edinburgh Postnatal Depression Scale. Bold p: p < 0.05. *: q < 0.05.