Examining Associations Between the Timing of Maternal Depressive Symptoms, Parenting Behaviors, and Child Outcomes

Abstract

Objective:

Perinatal depression is associated with poor psychosocial functioning and impaired parent–child relationship quality and confers risk for offspring psychopathology and altered brain structure. This study investigated the relative associations between the timing of perinatal depressive symptoms (prenatal, postnatal, and concurrent) and toddler psychopathology symptoms, toddler brain structure, and parenting behaviors. The study then explored whether parenting behaviors mediated relations between perinatal depressive symptoms and child psychopathology outcomes.

Method:

Using longitudinal data from 220 pregnant women and singleton offspring in the Early Life Adversity and Biological Embedding Study, associations from the first trimester of pregnancy to 3 years postpartum were examined. Linear mixed effect analyses examined associations between the timing of maternal depressive symptoms and toddler brain and behavioral development. The study also investigated whether maternal depressive symptoms were indirectly associated with psychopathology symptoms of children through parenting behaviors. The data included multiple time points of maternal depressive symptoms (prenatal, early postnatal, late postnatal, and concurrent) and children’s outcomes (psychopathology symptoms and brain structure at ages 2 and 3 years), which allowed examination of dynamic associations within dyads.

Results:

Higher levels of early postnatal maternal depressive symptoms, exceeding prenatal and concurrent depressive symptoms, were associated with greater externalizing symptoms in toddlers (ß = .15, p = .02, q = .05). Nonsupportive parenting behaviors partially mediated these relations (p = .02) and were directly associated with greater externalizing symptoms in children (ß = .14, p = .03).

Conclusion:

Findings suggest temporal differences in the relation of perinatal depressive symptoms to children’s psychopathology. Specifically, postnatal maternal depressive symptoms were associated with toddlers’ externalizing symptoms. Nonsupportive parenting behaviors partially explained these relations. These results emphasize the importance of early and accessible identification, intervention, and treatment of postnatal depressive symptoms.

Plain language summary:

This study aimed to better understand associations between maternal depressive symptoms across the perinatal period and child socioemotional outcomes. Using longitudinal data from 220 pairs of pregnant women and their children, the authors examined associations between the timing of maternal depressive symptoms and toddler brain development and behavior. The authors found that postnatal maternal depressive symptoms were associated with toddlers’ externalizing symptoms, and nonsupportive parenting behaviors partially mediated this relationship. Findings highlight the importance of accessible perinatal mental healthcare, including for mothers with subclinical depressive symptoms.

Study registration information:

Investigating the Path From Prenatal Depression Through Parenting Behaviors to Offspring Internalizing Problems and Amygdala and Hippocampal Volume; https://osf.io/s3vma.

Perinatal depression, characterized by depression during pregnancy or within the first year after birth, is associated with poor psychosocial functioning, adverse birth outcomes, and impaired parent–child relationship quality.¹ The prevalence of perinatal depression ranges from 12% to 26%,² disproportionately affecting mothers with lower socioeconomic status and chronic stress.³ Maternal depression in the perinatal period is among the earliest environmental adversities to which children can be exposed; may confer risk for greater overall psychopathology, poor self-control, and altered brain structure and functioning; and may affect the parent–child relationship.^4–6 Perinatal depression has epigenetic-related psychophysiological associations: in utero, depression may change DNA methylation and gene expression,⁷ circadian rhythms,⁸ stress reactivity,^9,10 and immune responses¹¹; following birth, early caregiving behaviors may be associated with altered neurophysiological and socioemotional development in infancy and beyond.¹²

In addition to behavioral and biophysiological effects, perinatal depression may negatively influence a mother’s emerging identity as a parent. Mothers with prenatal depression tend to experience more negative expectations and cognitions about their children that persist through the first 2 years of life¹³ and report lower perceived parenting competency, which may negatively impact parenting behaviors and exacerbate depressive symptoms.¹⁴ Further, mothers with greater prenatal depressive symptoms exhibit lower levels of responsiveness and sensitivity¹⁵ and higher levels of harsh parenting¹⁶ during interactions with their toddlers. When looking at parenting effects in dyads with maternal depression, increased harsh parenting mediates the relation between prenatal depression and preschool internalizing symptoms.¹⁷

Whereas research has historically focused on prenatal depression, postnatal maternal depression, defined by depression in the child’s first year of life, is also associated with an increased risk of psychopathology in children. Maternal depression in the first year of life is associated with higher rates of internalizing and externalizing problems in school-age children.¹⁸ Findings on the increased risk for developing psychopathology in offspring of mothers with postnatal depression are widely replicated and have been associated with the effects of early parenting behaviors¹⁹ as well as epigenetic and biological factors. Similar to prenatal depression, postnatal depression is associated with more withdrawn and harsh parenting behavior,²⁰ fewer displays of affection and sensitivity toward children,²¹ and less perceived parenting competency and emotional attachment to their children.²² Nonsupportive parenting behaviors (eg, ignoring or not engaging with the child, harsh tone, excessive criticism of the child) are associated with difficult temperament, emotional dysregulation, and withdrawn behavior in infancy and toddlerhood.²³

Given the critical, lasting effects of maternal depression on offspring, a primary goal of the current study was to investigate the effects of maternal depressive symptoms as a function of developmental timing (prenatally, postnatally, and in early childhood) and to examine whether parenting may contribute to the intergenerational transmission of depression and psychopathology more broadly. There is evidence that exposure to both prenatal and postnatal depression is negatively associated with children’s outcomes and that prior depressive episodes confer risk for later episodes²⁴; however, the literature is less clear about the relative effects of the specific timing of maternal depressive symptoms within the perinatal period on children’s outcomes. For instance, previous longitudinal studies found that prenatal depression had the strongest association with externalizing symptoms and total problems in early to middle childhood (ages 2–8 years old) compared with postnatal depression²⁵ and concurrent depression, as defined by maternal depressive symptoms measured at the same time as the child’s symptoms.²⁶ Another study found that concurrent maternal anxiety and depression had the strongest association with internalizing and externalizing symptoms of children compared with maternal symptoms during the prenatal or postpartum periods.¹⁷ Thus, further research is needed to elucidate the temporal relations between maternal depressive symptoms in the perinatal period and psychopathology symptoms in toddlers to inform the timing of prevention and early intervention.

In addition to understanding the role of timing in maternal depressive symptoms, examining variation in types of parenting behaviors (eg, supportive, nonsupportive) may provide insight into how these behaviors may differentially relate to children’s outcomes. Supportive parental behaviors (eg, using a warm tone, displaying physical affection, praising the child, and facilitating play) are associated with fewer behavior problems and less emotion dysregulation, whereas less sensitive and responsive parenting is associated with increased risk for depressive symptoms and conduct problems in preschool and early childhood.^27–29 Nonsupportive (eg, negative and intrusive) parental behaviors have been associated with noncompliance, aggression, and other early manifestations of self-control and externalizing problems in toddlers.³⁰ Given the lasting effects of postnatal maternal depression on mental health in children, investigating the role of parenting in the association between the timing of maternal depression and children’s outcomes may delineate mechanisms underlying the intergenerational transmission of psychopathology and inform intervention targets. Additionally, much of the past work on maternal depression and parenting behaviors has focused on internalizing symptoms of children. However, when internalizing symptoms are broadly measured through parent-report surveys, they are difficult to capture precisely before 3 years of age,³¹ and more complex symptoms, such as guilt or shame, may not emerge until age 3.³² Thus, further research is needed to identify potential parenting mediators of the relation between maternal depressive symptoms and children’s externalizing symptoms, which are more common and observable in toddlers.³³

Another mechanism by which parenting may influence children’s behavioral development is through the associations of parenting with children’s brain development. In addition to strong associations between social disadvantage and early cognitive development,³⁴ the psychosocial environment relates to brain development, with parenting from the primary caregiver serving as an essential driver of this relation. Supportive early caregiving is strongly associated with hippocampal volume in school-age children,^35,36 and harsh parenting is associated with smaller total gray matter, white matter, and amygdala volumes over time in school-age children.^37–39 In examining associations between amygdala volume and parenting, variation in harsh parenting seems particularly salient, likely due to the role of the amygdala in sensitivity to stress and social input.⁴⁰ Smaller amygdala and hippocampal volumes are associated with pediatric depression,⁴¹ emotion dysregulation,⁴² and altered stress reactivity.⁴³ Although most studies observe smaller amygdala and hippocampal volumes related to exposure to prenatal maternal depression, some evidence suggests that maternal depressive symptoms may be specifically associated with larger right amygdala volume, particularly in girls.⁴⁴ Understanding potential environmental pathways, such as the timing of perinatal depressive symptoms and relations to parenting, that are associated with early brain development is crucial in identifying risk for psychopathology in young children.

To address the above-defined gaps in the literature, we aimed to understand the relation between perinatal depressive symptoms (prenatal, postnatal, and concurrent at ages 2 and 3) and subsequent psychopathology symptoms in toddlers (aim 1), brain development in toddlers (aim 2), and parenting behaviors (aim 3). Further, we aimed to investigate whether parenting behaviors were associated with psychopathology symptoms in toddlers (aim 4) and whether parenting behaviors would mediate the relations between maternal depressive symptoms and toddlers’ psychopathology symptoms (aim 5).

For aim 1, we examined the relation between maternal depressive symptoms (prenatal, postnatal, and concurrent) and psychopathology outcomes in toddlers (internalizing, externalizing, and total psychopathology symptoms). Based on prior literature, we hypothesized that greater maternal depressive symptoms, specifically during the prenatal period, would be associated with greater psychopathology symptoms in children. For aim 2, we examined the relation between maternal depressive symptoms (prenatal, postnatal, and concurrent) and toddlers’ brain structure (amygdala, hippocampal, white matter, and total brain volumes). We expected that greater maternal depressive symptoms, specifically during the prenatal period, would be associated with smaller amygdala, hippocampal, and total brain volumes at ages 2 and 3 years. For aim 3, we examined the relation between maternal depressive symptoms (prenatal and postnatal) and parenting behaviors (supportive and nonsupportive) at age 1 year to determine if there were specific time periods when maternal depressive symptoms had the strongest associations with parenting behaviors. We hypothesized that prenatal maternal depressive symptoms would be positively associated with nonsupportive parenting behaviors and negatively associated with supportive parenting behaviors at year 1. For aim 4, we examined the relation between parenting behaviors at age 1 year and children’s psychopathology symptoms at ages 2 and 3 years. We expected that greater nonsupportive parenting behaviors and fewer supportive parenting behaviors at year 1 would be associated with greater psychopathology of children at ages 2 and 3 years. Finally, for aim 5, we investigated whether parenting behaviors at age 1 year mediated the relation between the timing of maternal depressive symptoms and children’s psychopathology symptoms. Here, we hypothesized that greater nonsupportive and fewer supportive parenting behaviors at year 1 would mediate the relation between maternal depressive symptoms and children’s psychopathology symptoms. We used a longitudinal prospective study that began in utero to address these aims and examine the relations of the timing of perinatal depressive symptoms to child brain and behavioral outcomes. We preregistered these hypotheses on the Open Science Framework (https://osf.io/s3vma/) (see “Deviations From the Pre-Registration” in the Supplement, available online).

METHOD

Study Design and Participants

Data were collected as part of the Early Life Adversity and Biological Embedding (eLABE) Study, a longitudinal study designed to investigate the impacts of early psychosocial adversity on neural, cognitive, and emotional development. The study originally enrolled 395 healthy pregnant women and their singleton offspring. Participants were oversampled for exposure to socioeconomic adversity (eg, poverty, access to health care) and recruited in 2 obstetric clinics: 1 community clinic and 1 private insurance clinic. Exclusion criteria included multiple gestations, alcohol or substance use during pregnancy (except for marijuana or tobacco), known pregnancy complications, and fetal congenital anomalies. Prenatal assessments were conducted during each trimester of pregnancy. Mother–child dyads were assessed shortly after birth (4 months postpartum) and returned for follow-up assessments at ages 1, 2, and 3 years. Data collection at these time points is complete and is ongoing at subsequent waves. Written informed consent was obtained from mothers after describing study protocols and before participation. Current analyses included all participants with scores on parent–child interaction (PCI) tasks, with neuroimaging and psychopathology symptom data at ages 2 or 3 years, and without evidence of significant injury on neonatal magnetic resonance imaging (MRI) scans (N = 220 mother–child dyads). All protocols were approved by the Institutional Review Board at Washington University in St. Louis.

Measures

Demographics.

Demographic data, including family income and child age, sex, race, and ethnicity, were collected via maternal self-report and medical records. Income-to-needs ratio (INR) was measured by dividing a family’s total income by the federal poverty guidelines of needs based on family size. We used INR to create a threshold score for above or below the poverty line (threshold of 2) to control for economic adversity. According to medical records and self-reports, 49 participants reported taking psychotropic medications during their pregnancy, which was accounted for in analyses. Our analyses did not include race, a socially defined construct, per recommendations^45,46 that highlight the risks of perpetuating systemic racism and using race as a proxy or indirect measure of experiences such as discrimination and subsequent stress experienced by participants and socioeconomic disadvantage (which was included in our models). The current project did not include measures of discrimination or race-based trauma or stress, but these experiences should be explored in future work, given racial disparities in prenatal health care and maternal mortality.⁴⁷

Maternal Depressive Symptoms.

Maternal depressive symptoms during pregnancy were measured using the Edinburgh Postnatal Depression Scale (EPDS),⁴⁸ and mean scores were calculated across trimesters (detailed score breakdowns and diagnosis history are available in “Maternal Depression” in the Supplement, available online). Prenatal averages were used to capture potential effects of exposure to depressive symptoms across pregnancy. We had no trimester-specific hypotheses based on past literature, which found no trimester-specific associations between depressive symptoms and child outcomes.²⁶ Postnatal maternal depressive symptoms were measured at 4 and 12 months after delivery using the EPDS to better understand the effects of depressive symptoms early in the postnatal period compared with later (detailed score breakdowns and comparison of postnatal time points are available in “Maternal Depression” in the Supplement and illustrated in Figure S1, available online). The EPDS is the recommended measure of perinatal depression because it makes a distinction between depressive symptoms and normal physical pregnancy and postpartum symptoms.⁴⁹ Maternal depressive symptoms at offspring ages 2 and 3 were assessed using the Beck Depression Inventory (BDI).⁵⁰ Prenatal and 4-month postnatal maternal depressive symptoms were moderately correlated (r = 0.51). Postnatal depressive symptoms at 4 and 12 months were strongly correlated (r = 0.88). Prenatal and 4-month postnatal maternal depressive symptoms (EPDS) were moderately correlated with symptoms at years 2 and 3 (BDI) (r = 0.31 and r = 0.44, respectively).

Parenting Behaviors.

Three videotaped PCI tasks (feeding, teaching, and free play) were coded using the Parent–Child Interaction Rating Scales.⁵¹ Parental sensitivity, positive regard, intrusiveness, detachment, and negative regard were assessed and scored using a 7-point scale, with higher scores representing more of that behavior. Coders were trained and needed to reach 92% reliability (exact or within 1 point on the coding scale) with a master coder before coding interactions independently.⁵² For ongoing reliability, 20% of interaction tapes were double-coded. A small portion (approximately 14%) of the PCI tasks were conducted and videotaped remotely due to COVID-19. These dyads were included in analyses, and PCI status (remote vs in-person task completion) was controlled for in all analyses with parenting variables. We see the same pattern of results when we exclude the remote PCI, and parenting behaviors coded in the in-person and remote interactions show the same relations to child outcome variables (Figure S2, available online).

Supportive (sensitivity and positive regard) and nonsupportive (intrusiveness, detachment, and negative regard) parenting composite scores were created and averaged across the 3 interaction tasks. Results did not significantly differ when using individual behaviors or task scores, so supportive and nonsupportive composite scores across the 3 tasks were used for all analyses. Prior work has identified positive parenting behaviors to include responsivity, warmth, and displays of support and negative parenting behaviors to include hostility, intrusiveness, and detachment.⁵³ Thus, parenting behavior scores were categorized in supportive and nonsupportive composite scores based on whether they were capturing positive or negative parenting behaviors. This design also allows for comparison across other studies with similar definitions of parenting behavior.^54,55 Many studies have found meaningful differences in associations and outcomes based on the type of parenting behavior (supportive vs nonsupportive).⁵⁶ A single parenting score would obscure important context for associations with parenting behaviors and make very different parenting profiles indiscernible (ie, parents who scored high or low on both composites would fall in the middle of a single parenting dimension score). Distinguishing between supportive and nonsupportive parenting behaviors and separately investigating their relation to maternal depressive symptoms and child outcomes allows for specificity in findings and suggestions for future work. Supportive and non-supportive composite scores were negatively correlated (r = −0.79) (a complete correlation matrix between subscales is available in Table S1, available online). Internal consistencies for both parenting composites were acceptable (supportive parenting behavior: α = .82; nonsupportive parenting behavior: α = .73). Tables S2 and S3, available online, provide additional details on supportive and nonsupportive composites.

Child Socioemotional Outcomes.

Child psychopathology symptoms were measured using parent reports on the Infant-Toddler Social and Emotional Assessment (ITSEA) at age 2.⁵⁷ Internalizing symptoms were measured using the Depression/Withdrawal and Anxiety subscales of the ITSEA, and externalizing symptoms were measured using the Aggression/Defiance, Activity/Impulsivity, and Peer Aggression subscales. Total psychopathology was measured by combining internalizing and externalizing scores. Child psychopathology symptoms at age 3 were measured using parent reports on the Child Behavior Checklist (CBCL).⁵⁸ Internalizing symptoms were assessed using the Affective and Anxiety DSM subscales, and externalizing symptoms were measured using the Attention-Deficit/Hyperactivity Disorder and Oppositional Defiant Disorder (ODD) DSM subscales. Total psychopathology at age 3 was measured by combining internalizing and externalizing subscales. ITSEA and CBCL subscales were selected to ensure that both measures captured the same symptoms to allow for analyses across time points (eg, internalizing symptom composite scores encompassed the same types of symptoms from both measures). Child internalizing, externalizing, and total symptom scores were used to examine possible differences in associations with maternal depressive symptoms and parenting. Prior work has found some differences in risk factors for internalizing and externalizing symptoms in young children.⁵⁹ Thus, our hypotheses focused broadly on different relations between maternal characteristics (eg, depressive symptoms, parenting) and child outcomes (eg, externalizing and internalizing symptom scores) to avoid unnecessary tests (eg, comparing specific subscales). Although research shows that early symptoms of psychopathology are associated with later symptoms,^60,61 discrete diagnoses without multi-informant or multimethod assessment in such early childhood are less reliable and valid,^62–64 and we are underpowered to investigate such relations. Psychopathology symptoms in children at ages 2 and 3 were z-scored to allow for comparison and multilevel analysis across time points.

Child Brain Structure.

Toddlers underwent MRI in a Siemens MAGNETOM Prisma 3T scanner (Siemens Healthineers) with a 64-channel Siemens head coil to acquire T1-weighted images (repetition time = 2,400 ms, echo time = 2.22 ms, 0.8 mm³ isometric voxels), T2-weighted images (repetition time = 3,200/4, 500 ms, echo time = 563 ms, 0.8 mm³ isometric voxels), and spin echo field map sequences (repetition time = 8,000 ms, echo time = 66 ms, 2 mm³ isotropic voxels, multiband factor = 1). MRI scans were obtained during the first few weeks of life and at ages 2 and 3. Scans at ages 2 and 3 were affected by COVID-19, and thus we included all available neuroimaging data in analyses. For participants with data at ages 2 and 3, both time points were included in models and nested within the participant. Child age at MRI was controlled for in all analyses. The Melbourne Children’s Regional Infant Brain Atlas Surface toolkit⁶⁵ was used to generate anatomical volume segmentations. Brain volumes of interest included hippocampal, amygdala, white matter, and total brain volumes.

Data Analysis

To lessen attrition bias and include all participants who had outcome data at age 2 or 3 years, we used multiple imputation, which is considered best practice for handling missing data and improves accuracy and statistical power relative to other techniques (eg, listwise deletion).⁶⁶ Given the high proportion of low-income families in our study and the impact of disadvantage on attrition, we used multiple imputation to ensure we did not unnecessarily exclude these groups from our data. Missing data were greatest for follow-up visits at age 2 and 3 years, during the first years of the COVID-19 pandemic. No demographic, prenatal, or parenting data were missing in our sample (extent of missingness by variable is available in Table 1). All missingness was in the range for acceptable use of multiple imputation.^67,68 Predictive mean matching was used for the missing continuous variables (postnatal EPDS, BDI, ITSEA, and CBCL data) to lessen bias and retain the original variabilities and skewed distributions of the data.⁶⁹ We used all the variables of interest in the imputation under a fully conditional specification, using the default settings of the mice 3.16.0 package in R (https://www.rdocumentation.org/packages/mice/versions/3.16.0) and stabilized after 20 imputations. Each model included in our analyses was run on every imputation (m = 20) and then pooled together to get 1 estimate for each coefficient using Rubin’s rules.⁶⁸

TABLE 1.

Descriptive Statistics

		Value			Missingness
Demographics/covariates	Observed range	Mean		(SD)	n	(%)
Child age at year 2 visit, y	1.92–2.50	2.02		(0.10)	24	(11)
Child age at year 3 visit, y	2.92–3.45	3.04		(0.08)	54	(25)
			%
Child sex, female %	NA		47.3		0	(0)
Child race, parent reported	NA				0	(0)
Asian		3
Black/African American		57
Othera		1
White		39
Child ethnicity, % Hispanic/Latinx	NA	8			0	(0)
		Mean		(SD)
INR	0.40–12.50	2.93		(2.99)	0	(0)
			%
PCI remote status, % remote	NA		13.6		0	(0)
		Mean		(SD)
Prenatal maternal depression (EPDS)	0–20.50	4.63		(3.91)	0	(0)
4-month postnatal maternal depression (EPDS)	0–19	5.71		(4.10)	8	(0.04)
12-month postnatal maternal depression (EPDS)	0–17	5.54		(4.25)	8	(0.04)
Maternal depression at year 2 and 3 (BDI)	0–34.5	6.61		(6.65)	66	(28)
Supportive parenting score	2.33–7	4.48		(0.92)	0	(0)
Nonsupportive parenting score	1–3.44	1.60		(0.50)	0	(0)
Child total psychopathology at year 2 (ITSEA)	0.28–2.44	1.09		(0.41)	24	(11)
Child externalizing symptoms at year 2 (ITSEA)	0.03–1.44	0.56		(0.26)	24	(11)
Child internalizing symptoms at year 2 (ITSEA)	0.11–1.19	0.53		(0.22)	22	(10)
Child total psychopathology at year 3 (CBCL)	0–9.25	2.83		(1.80)	54	(25)
Child externalizing symptoms at year 3 (CBCL)	0–11	3.60		(2.36)	54	(25)
Child internalizing symptoms at year 3 (CBCL)	0–12.5	2.05		(1.78)	54	(25)

Note: ITSEA and CBCL scores are averages across subscales within each composite. Missingness differed on some child psychopathology scales because of item-level missingness that impacted the calculation of one score but not another. BDI = Beck Depression Inventory; CBCL = Child Behavior Checklist; EPDS = Edinburgh Postnatal Depression Scale; INR = income-to-needs ratio; ITSEA = Infant-Toddler Social and Emotional Assessment; NA = not applicable; PCI = parent–child interaction.

^aMixed race and Asian Indian.

Linear mixed effects regressions, with random intercepts for time point and child (with child age as a time-varying covariate) to account for repeated measures, examined the relations between maternal depressive symptoms at various time points (prenatal, postnatal, and concurrent) and our outcomes of interest (psychopathology symptoms in children [aim 1], brain structure in children [aim 2], and parenting behaviors [aim 3]). Analyses were conducted using the lmerTest 3.1–3 and lme4 1.1–35.1 packages in R. We analyzed associations between prenatal maternal depressive symptoms and psychopathology symptoms in children (aim 1), brain structure in children (aim 2), and parenting behaviors (aim 3). Child sex, INR, and maternal psychotropic medication use were included as covariates in each model (model I). To determine temporal specificity, we added postnatal depressive symptoms (model II) and concurrent depressive symptoms (model III) to each regression for our first 3 aims. Analyses investigating associations with psychopathology symptoms in children were adjusted for the 3 symptom groups, and analyses with parenting behaviors were adjusted for the 2 composite scores using multiple comparison procedures (Bonferroni). Analyses with brain outcomes in children were adjusted for the 4 regions of interest using multiple comparison procedures (Bonferroni). We assessed relations between parenting (supportive and nonsupportive in separate analyses) and psychopathology symptoms in children with the same covariates and remote PCI status (aim 4). Then, we included maternal depressive symptoms at each time point to determine whether parenting behaviors independently predicted psychopathology symptoms in children in aim 4.

Mediation models (aim 5) included postnatal maternal depressive symptoms as the predictor and parenting behaviors (supportive and nonsupportive) as the mediators for externalizing, internalizing, and total psychopathology in children. To allow for mediation analyses with longitudinal data, we ran a mediation model (a, the effect of the predictor on the mediator), an outcome model (b, the effect of the mediator on the outcome), and a model without the mediator (c1, the direct effect of the predictor on the outcome), all controlling for child age, sex, and INR. For each mediation, we calculated the indirect effect of the mediator, parenting behavior (a × b), and the total effect of the predictor, postnatal depressive symptoms, on the outcome, child psychopathology symptoms (c1 + (a × b)). These analyses were completed with the structural equation modeling function sem from the lavaan 0.6.17 package in R.

RESULTS

The analyzed sample included 220 women and their children. The means, ranges, standard deviations, and missingness for relevant demographic variables, covariates, predictors, and outcomes are shown in Table 1.

Aim 1: Temporal Relations of Maternal Depressive Symptoms to Child’s Psychopathology

Prenatal maternal depressive symptoms were not associated with total psychopathology symptoms in children at ages 2 and 3 years (ß = .08, p = .10) (Table S4, model I, available online). However, greater postnatal maternal depressive symptoms at 4 months postpartum were associated with greater total psychopathology symptoms in children at ages 2 and 3 years, even after controlling for child sex, age, INR, psychotropic medication use, and prenatal maternal depressive symptoms (ß = .18, p = .002, q = .002) (Figure 1; Table S4, model II, available online). However, this relation did not hold after adding concurrent maternal depressive symptoms to the model and was trend level after adjusting for multiple comparisons (ß = .11, p = .05, q = .08) (Figure 1, Table 2). Concurrent symptoms were associated with psychopathology symptoms in children (ß = .19, p < .001, q < .001) (Table 2). Maternal depressive-symptoms at 12 months postpartum were not significantly associated with total psychopathology symptoms in children at ages 2 and 3 years after accounting for concurrent maternal symptoms (ß = .10, p = .06) (Table S5, available online).

FIGURE 1. Postnatal Maternal Depressive Symptoms Predicting Toddler Psychopathology Symptoms.

Note: Bold statistics indicate associations that passed Bonferroni correction. Greater postnatal maternal depressive symptoms (4 months postpartum) were associated with greater child externalizing symptoms, after accounting for concurrent maternal depression. Please note color figures are available online.

TABLE 2.

Maternal Depressive Symptoms Predicting Toddler Psychopathology Symptoms

	Outcome variable: total child psychopathology symptoms					Outcome variable: child externalizing symptoms					Outcome variable: child internalizing symptoms
Predictor	B	SE	t	p (q)	95% CI	B	SE	t	p (q)	95% CI	B	SE	t	p (q)	95% CI
Intercept	−1.05	1.18	−0.89	.38	[−3.38, 1.28]	−.61	0.94	−0.65	.52	[−2.46, 1.23]	−1.26	1.53	−0.82	.41	[−4.27, 1.76]
Prenatal maternal depressive symptoms	−.02	0.05	−0.44	.66	[−0.13, 0.08]	.001	0.06	0.02	.98	[−0.03, 0.12]	−.05	0.08	−0.65	.52	[−0.20, 0.10]
4-month postnatal maternal depressive symptoms	.11	0.06	1.91	.05 (.08)	[−0.003, 0.22]	.15	0.06	2.36	.02 (.05)	[0.01, 0.07]	.07	0.07	0.88	.38	[−0.08, 0.21]
Concurrent maternal depressive symptoms (ages 2 and 3)	.19	0.05	3.86	<.001 (<.001)	[0.09, 0.29]	.15	0.06	2.62	.01 (.01)	[0.04, 0.27]	.23	0.06	1.67	<.001 (<.001)	[0.10, 0.36]
Child age	.09	0.04	2.00	.05 (.14)	[0.001, 0.17]	.05	0.04	1.18	.24	[−0.03, 0.14]	.10	0.06	1.67	.10	[−0.02, 0.22]
Child sex (male)	−.02	0.09	−0.20	.84	[−0.19, 0.16]	.06	0.10	0.55	.59	[−0.15, 0.26]	−.09	0.12	−0.74	.46	[−0.33, 0.15]
INR	−.28	0.10	−2.89	.004 (.006)	[−0.46, −0.09]	−.34	0.11	−3.24	.001 (.004)	[−0.54, −0.13]	−.21	0.13	−1.60	.11	[−0.47, 0.05]
Maternal antidepressant use (on medication)	.03	0.11	0.24	.81	[−0.20, 0.25]	.01	0.13	0.05	.96	[−0.25, 0.26]	.05	0.15	0.31	.76	[−0.24, 0.33]

Note: Bold indicates predictors that passed Bonferroni correction. INR = income-to-needs ratio.

A set of follow-up analyses examined specific forms of psychopathology in children. Prenatal maternal depressive symptoms, again, were not associated with either externalizing or internalizing symptoms of children (Table 2). Postnatal maternal depressive symptoms at 4 months postpartum were associated with greater externalizing symptoms of children after controlling for child sex, age, INR, psychotropic medication use, and both prenatal and concurrent maternal depressive symptoms (ß = .15, p = .02, q = .05) (Figure 1, Table 2), though concurrent maternal symptoms were, again, independently associated with child externalizing symptoms. Depressive symptoms at 4 months postnatal were not associated with children’s internalizing symptoms (ß = .07, p = .38), though concurrent maternal depressive symptoms were (ß = .23, p < .001, q < .001) (Table 2). The association between early postnatal depressive symptoms (4 months postnatal) and children externalizing symptoms remained significant even when also controlling for children’s internalizing symptoms (ß = .13, p = .03) (Table S6, available online), suggesting a specific relation to children’s externalizing symptoms. Maternal depressive symptoms at 12 months postpartum were not significantly associated with children’s internalizing symptoms or externalizing symptoms at ages 2 and 3 years, after accounting for concurrent maternal symptoms (ß = .09, p = .19 and ß = .11, p = .08, respectively) (Table S5, available online).

Aim 2: Maternal Depressive Symptoms and Child’s Brain Structure

Maternal depressive symptoms (prenatal, 4 months postnatal, and concurrent) were not associated with child hippocampal, amygdala, total brain, or white matter volumes at ages 2 and 3, controlling for the volume of the region of interest at birth (Tables S7–S10, available online).

Aim 3: Maternal Depressive Symptoms and Parenting

Prenatal depressive symptoms were associated with nonsupportive parenting behaviors at age 1 (ß = 1.89, p < .001, q < .001) (Table S11, model I, available online), but not supportive parenting behaviors (Table S12, available online). In line with our preregistered hypothesis, prenatal depressive symptoms continued to be associated with nonsupportive parenting after controlling for INR, child sex, PCI remote status, psychotropic medication use, and child age and accounting for maternal depressive symptoms concurrent with parenting behaviors (at 12-month postnatal follow-up) (ß = .11, p = .02, q = .05) (Table S11, model II, available online). Postnatal depressive symptoms at 4 and 12 months postpartum were not associated with supportive or nonsupportive parenting behaviors (Tables S11–S13, available online).

Aim 4: Parenting and Child’s Psychopathology

As we hypothesized, both supportive parenting (ß = −.11, p = .04, q = .04) (Table S14, model I, available online) and nonsupportive parenting (ß = .11, p = .03, q = .04) (Table S14, model II, available online) scores at age 1 were associated with total psychopathology symptoms in children at ages 2 and 3. Supportive and nonsupportive parenting behaviors continued to be associated with total psychopathology symptoms in children with prenatal and 4-month postnatal depressive symptoms and all covariates in the model (ß = −.12, p = .03, q = .04 and ß = .11, p = .04, q = .04, respectively) (Table S14, models III and IV, available online) and were trend level after adding concurrent depressive symptoms and adjusting for multiple comparisons (ß = −.12, p = .03, q = .06 and ß = .10, p = .06, respectively) (Table S14, models V and VI, available online).

Follow-up analyses investigated the effects of parenting behaviors on specific forms of psychopathology in children. Supportive parenting behaviors were not associated with internalizing or externalizing symptoms in children, even without maternal depressive symptoms in the model (Table S14, models VII and VIII, available online). Nonsupportive parenting behaviors were associated with children’s externalizing symptoms, but not internalizing symptoms (Table S14, model VIIII, available online), without maternal depressive symptoms in the model, and were trend level after multiple comparisons correction (ß = .13, p = .04, q = .08) (Table S14, model X, available online).

Aim 5: Parenting as a Mediator of Relations Between Maternal Depressive Symptoms and Children’s Psychopathology

Supportive parenting behaviors did not mediate the relation between maternal depressive symptoms 4 months postnatal and psychopathology symptoms in children (Table S15, available online). However, nonsupportive parenting behaviors partially mediated the relation between maternal depressive symptoms 4 months postpartum and total psychopathology symptoms in children (ab = 0.007, p = .02) (Figure 2; Table S15, available online). Nonsupportive parenting behaviors also partially mediated the relation between postnatal maternal depressive symptoms 4 months postpartum and children’s externalizing symptoms (ab = 0.010, p = .02) (Figure 2; Table S15, available online). All mediation models fit well (comparative fit index >0.95, root mean square error of approximation >0.05, standardized root mean square residual >.02).

FIGURE 2. Nonsupportive Parenting Behavior Mediation Models.

Note: Nonsupportive parenting behaviors mediated the relation between (a) 4-month postnatal maternal depressive symptoms and total child psychopathology and (b) 4-month postnatal maternal depressive symptoms and child externalizing symptoms.

DISCUSSION

Given the prevalence of perinatal depression and its impact on families, the current study aimed to investigate the timing effects of depressive symptoms and the potential mediating role of parenting behaviors on brain and behavioral development. Our results demonstrated a key relation between early postnatal depressive symptoms and children’s externalizing symptoms specifically (aim 1). Our findings that greater supportive and fewer nonsupportive parenting behaviors were associated with fewer subsequent psychopathology symptoms in toddlers became trend level after correcting for multiple comparisons (aim 4). Notably, nonsupportive parenting behaviors mediated the relations between early postnatal maternal depressive symptoms and children’s externalizing symptoms and total psychopathology (aim 5). This finding partially provided preliminary evidence consistent with our hypothesized mechanism of parenting behaviors contributing to the relation between maternal depressive symptoms and psychopathology symptoms in offspring.

Although prenatal maternal depressive symptoms were not significantly associated with any psychopathology symptoms in children as we had expected, greater maternal depressive symptoms in the early postnatal period (4 months postpartum) were associated with greater externalizing symptoms in children, even after controlling for concurrent maternal depressive symptoms (aim 1). Depressive symptoms later in the postnatal period (12 months postpartum) were not significantly associated with psychopathology symptoms in children after accounting for concurrent depressive symptoms. Crucially, this finding highlights the heterogeneity of depression in the postnatal period and suggests a unique relation between maternal depressive symptoms in the early postpartum period and subsequent externalizing symptoms in children. There is some evidence in support of postnatal depression heterogeneity; specifically, early postnatal depressive symptoms tend to be more severe than, but etiologically distinct from, major depressive disorder, whereas symptoms in the later postpartum period may be similar to episodic major depressive disorder outside of the perinatal period.⁷⁰ This difference may explain why we found associations between early postnatal symptoms and psychopathology in children, but did not see the same association with later postnatal symptoms when accounting for concurrent depressive symptoms. Prior work has also found that early postnatal depressive symptoms are negatively associated with the early development of mother–child attachment,^71,72 which is shown to have lasting impacts on a child’s social and emotional development.⁷³ Our results support the importance of maternal well-being in the early postnatal period, possibly a critical time for attachment, maternal self-perception, and identity development as a parent, and psychological intervention.

In addition to the influence of the early postnatal period, these findings may identify a unique influence of postnatal maternal depressive symptoms on externalizing symptoms specifically in preschoolers. One explanation for this relation to externalizing rather than internalizing symptoms in children may be due to child age in the sample. Whereas internalizing and externalizing symptoms often co-occur, externalizing behaviors tend to be most frequent and observable during toddlerhood, and internalizing symptoms may be more difficult to detect,^33,74 which may explain the low rates of symptoms in our sample, particularly internalizing symptoms in toddlers at age 2 years. Recent work has led to improved, developmentally appropriate assessment of psychopathology in toddlers^75,76; however, in children before the age of 3, capturing internalizing symptoms may be less precise or rely on proxy measures, such as early executive control⁷⁷ and temperament.⁷⁸ It is also possible that maternal depressive symptoms may relate to early parenting behaviors in a way that confers greater risk for the development of externalizing symptoms than internalizing symptoms. Specificity in relations between caregiver depression, parenting behaviors, and internalizing and externalizing symptoms in children has been explored,⁷⁹ but further work is needed to delineate distinct mechanisms. Interestingly, concurrent maternal depressive symptoms were strongly associated with internalizing symptoms in children. Although research shows that parent and child internalizing symptoms are positively associated, reporter bias may inflate this relation, whereby mothers with depression may overendorse their child’s depressive symptoms.⁸⁰ Future research should investigate the relations between timing of maternal depressive symptoms and psychopathology symptoms in children across an extended period, such as through ages 6 to 8 years, to determine whether our findings persist beyond age 3 years.

It is possible that associations between maternal depressive symptoms and the dimension of children’s psychopathology symptoms differ based on the timing of maternal depressive symptoms. Our findings that early postnatal maternal depressive symptoms may be specifically associated with children’s externalizing symptoms contrasts with previous studies finding that prenatal depressive symptoms were most strongly related to children’s externalizing symptoms and total problems.^25,26 However, the methods used by prior studies differed from ours. For instance, some studies used convenience sampling to create population cohorts and took place outside of the United States,^37,81,82 and some studies used the Center for Epidemiologic Studies Depression Scale (CES-D) rather than the EPDS to measure prenatal and postnatal depressive symptoms.^18,26 Although prenatal depressive symptoms in our sample were largely at a subclinical level, prior work has also found associations between subclinical perinatal depressive symptoms and internalizing and externalizing symptoms in toddlers.⁸³ Such findings highlight the importance of mental health care, including thorough perinatal depression screening (emphasizing symptom quantity rather than a simple binary threshold) and enhanced support throughout pregnancy, the postpartum period, and beyond. This is particularly important for mothers who experience depressive symptoms prenatally, as they are at greater risk for subsequent depressive symptoms.^24,84 Our findings on subclinical depressive symptoms are consistent with some work in clinical populations that find maternal depression history is associated with greater nonsupportive parenting behaviors and greater externalizing symptoms in children.^4,85

We did not replicate associations between prenatal maternal depressive symptoms and smaller subcortical volumes, which was found in older children (aim 2).⁸⁶ However, prior findings were relatively mixed regarding the directionality of volume differences (smaller or larger) and which specific brain regions were related to maternal depressive symptoms.⁸⁷ Further, it is possible that the relation between maternal depressive symptoms and brain development may not be observable in 2- and 3-year-old children but might become more apparent in older children. It is also possible that more severe maternal depressive symptoms may show a stronger relation to offspring brain volumes.⁸² In addition, given that our study oversampled for socioeconomic adversity, it is possible that prior work did not account for environmental confounds with maternal depression, which thus may have driven differences in brain structure. For instance, maternal stress and socioeconomic disadvantage, which have been correlated with maternal depressive symptoms,⁸⁸ have contributed to differences in brain volume, as reported by previous work, including in this sample, that did account for perinatal maternal depressive symptoms.⁸⁹ Future research should continue to investigate possible relations between maternal depressive symptoms and structural development in early childhood to clarify some of the inconclusive past work.

The direction of the relations of maternal depressive symptoms and nonsupportive parenting behaviors aligned with our hypotheses and highlighted a unique association between prenatal depressive symptoms and nonsupportive parenting behaviors (aim 3). Even after accounting for maternal depressive symptoms at 12 months postpartum (concurrent with the parenting behavior observations), prenatal depressive symptoms continued to be associated with nonsupportive parenting behaviors. This extends past work that has found associations between prenatal depression and nonsupportive parenting¹⁶ by establishing this relation even in a sample with less severe depressive symptoms and nonsupportive parenting scores. In our sample, nonsupportive parenting behaviors were relatively low (average score of 1.60/7), which prevented our ability to study associations with more extreme forms of nonsupportive parenting. However, importantly, this finding suggests that relations between prenatal depressive symptoms and nonsupportive parenting behaviors exist even at lower, milder levels. The direction of relations between prenatal depressive symptoms and supportive parenting behaviors also aligned with our hypotheses, but were not significant after accounting for INR. This finding highlights the importance of accounting for socioeconomic disadvantage when investigating associations between maternal well-being and parenting behavior given the strong relations between socioeconomic status and parenting found here as well as in other work.^34,90 Although we had more variability in supportive parenting behaviors in our sample (average score of 4.48/7), it is possible that we would have found relations between depressive symptoms and supportive parenting, in addition to relations with INR, if we had a clinical sample with more severe maternal depressive symptoms or had more extreme parenting scores represented.

Past studies have found that supportive parenting behaviors are associated with fewer internalizing and externalizing symptoms.⁵⁶ Whereas we observed relations of the same direction in our study, we did not find significant relations between parenting behaviors and total child psychopathology symptoms after correcting for multiple comparisons (aim 4). The association between greater nonsupportive parenting behaviors and greater child externalizing symptoms was also not significant after accounting for maternal depressive symptoms and adjusting for multiple comparisons (aim 4). These findings suggest that early postnatal and concurrent maternal depressive symptoms may be most relevant for studying influences on total and externalizing psychopathology symptoms in toddlers, in addition to early parenting behaviors. Clinically, these findings underscore the importance of treating and addressing maternal depressive symptoms when promoting early childhood well-being.

Nonsupportive parenting behaviors partially mediated the relation between postnatal maternal depressive symptoms and total psychopathology and externalizing symptoms in children (aim 5). Supportive parenting behaviors did not mediate the relation between early postnatal maternal depressive symptoms and total child psychopathology. Nonsupportive parenting behaviors may relate to other processes, such as emotion regulation, that underlie the development of psychopathology.⁹¹ For example, parents whose behavior was coded as nonsupportive may be modeling ineffective emotion regulation or inhibitory control. These findings provide some preliminary evidence in support of our proposed mechanism that parenting mediates the relation between maternal depressive symptoms and child psychopathology symptoms. This finding may provide additional insight into the mechanisms that underlie the intergenerational transmission of psychopathology and offer a target for further investigation and future intervention.

Our study has several strengths, including the longitudinal design, which allowed us to investigate the temporal influences of maternal depressive symptoms within the same mother–child dyads. Further, our sample was relatively racially diverse and representative of the St. Louis community, allowing for findings to be generalizable within this context.

There were also limitations. Attrition is a concern in all longitudinal designs, and our study was impacted by the COVID-19 pandemic. The pandemic occurred when the children were ages 2 and 3 years, which most heavily affected our study visits with MRI data collection. This was a nonclinical sample, and our dimensional measures of depression allowed us to identify associations between the timing of depressive symptoms and child outcomes, even when depressive symptoms did not reach clinical thresholds. However, our sample limited the conclusions we could make about clinically significant depression compared with the presence of depressive symptoms below the clinical threshold. Thus, future research may benefit from oversampling for prenatal depression to ensure greater variability. Further, the lower rates of clinical depression in our sample may be related to a high proportion of the sample reportedly taking antidepressant medications during pregnancy compared with prior studies.^92,93 Whereas we controlled for medication use, future work should further consider treatment effects on depressive symptoms across the perinatal period given the observed relations between postnatal depressive symptoms and psychopathology in toddlers, specifically externalizing symptoms. Additionally, using 2 different measures of depressive symptoms, the EPDS for prenatal and postnatal periods and the BDI for age 2- and 3-year follow-up visits, may be seen as a limitation. However, we used the EPDS per the US Preventive Services Task Force guidelines and recommendations for measuring perinatal depression,⁴⁹ which ensured that normative pregnancy symptoms, particularly physical symptoms, were not conflated with depressive symptoms.

Finally, future work may benefit from assessing parenting with a measure that clearly and uniformly separates discrete parenting behaviors rather than one with qualitative composites (eg, the low end of detachment defined by warmth or intrusiveness). However, it remains an empirical question whether such discrete behaviors are the most ecologically valid. Further, our measure was able to capture the nuance of the scores and anchors through use of all supportive and nonsupportive subscales (ie, capturing warmth and intrusiveness through their respective subscales), and each subscale score represented the degree to which parents exhibited that specific behavior. Improving measures of parenting behaviors to better understand these relations is an important area for future study.

Our findings indicated critical temporal differences in the relation of maternal depressive symptoms and children’s socioemotional outcomes, most notably the relation between early postnatal maternal depressive symptoms and externalizing symptoms in toddlers. This finding highlights the value of early identification, intervention, and treatment of postnatal depressive symptoms, even if below clinical thresholds, and the necessity for mental health education throughout prenatal and postpartum care. Our study also identified nonsupportive parenting behaviors as a significant mediator of the relation of depressive symptoms and some forms of psychopathology in children. These mediation findings reinforce prior research on parenting behaviors as a potential buffer for children’s mental health,⁹⁴ support a family-based approach to treating externalizing symptoms in children, and may be useful for clinicians working with children and families.

Data Sharing:

The complete dataset is in the process of being transferred and uploaded to National Institute of Mental Health Data Archive (NDA) by the study investigators.