Does Treating Maternal Depression Affect Maternal and Child Hypothalamic Pituitary Adrenal Axis Activity? Six‐Year Follow‐Up of a Cluster Randomized Trial in Rural Pakistan

ABSTRACT

Perinatal depression is a risk factor for adverse long‐term outcomes in both mothers and children, including stress system dysregulation. Yet, it is unclear whether treating perinatal depression has lasting intergenerational effects on biological stress regulation. This study evaluated the effects of a psychosocial intervention programme initiated during the perinatal period and sustained throughout the child's early childhood (up to 5 years of age) on hair‐derived measures of hypothalamic pituitary adrenal (HPA) axis activity in mothers and their children. Mothers with prenatal depression were recruited during their third trimester of pregnancy and randomly assigned to a peer‐delivered psychosocial depression intervention (n = 283) or control (n = 287). At six years postpartum (one year after the conclusion of the intervention), hair samples were collected from mothers (n = 392) and children (n = 366) and analysed for cortisol, cortisone, and dehydroepiandosterone (DHEA). Additionally, 584 mothers without prenatal depression were recruited during pregnancy, of which 426 mothers and 389 children provided hair samples at six years. Mixed models showed no evidence of intervention effects on maternal HPA axis activity or child cortisone and DHEA. In contrast, children whose mothers received the intervention exhibited higher cortisol levels relative to children in the control group, bringing them closer to cortisol levels observed in the non‐depressed group. There was also an interaction between the intervention and baseline maternal depression severity, with stronger intervention effects on cortisol among children whose mothers experienced more severe depressive symptoms during pregnancy. Results suggest that treating prenatal depression can lead to durable improvements in stress system regulation in children, especially among those whose mothers had the highest symptom burden prenatally.

1. Introduction

Perinatal maternal depression has been linked to adverse outcomes throughout the life course for both mothers and children. Among mothers, perinatal depression is associated with recurrent depressive episodes and increased risk for later suicidality and functional impairment (J. H. Goodman ²⁰⁰⁴; Lindahl et al. ²⁰⁰⁵; O’Hara and McCabe ²⁰¹³). Additionally, there is a large body of observational evidence showing that children born to mothers with depression are at increased risk for a range of poor cognitive, emotional, and physical health outcomes (Morales et al. 2023; Sutherland et al. ²⁰²²). The global burden of perinatal maternal depression is high, especially in low‐ and middle‐income countries (LMIC) where it affects roughly 1 in 4 women (Roddy Mitchell et al. 2023). Despite these well‐established associations, the biological mechanisms linking maternal depression to later maternal and child outcomes remain incompletely understood. One proposed pathway involves dysregulation of the physiological stress response systems in both mother and child, particularly the hypothalamic‐pituitary‐adrenal (HPA) axis. Maternal depression is a significant stressor that may lead to long‐term activation of the HPA axis in mothers. From the child's perspective, exposure to maternal depression symptoms in utero and early in life may lead to lasting changes in the HPA axis through both biological embedding and postnatal caregiving environments. While interventions for maternal depression are known to improve maternal mental health, it remains unclear whether such interventions can produce lasting changes in intergenerational stress regulation. The goal of this paper is to examine the effects of a psychosocial intervention for prenatal depression on HPA axis activity in mothers and their children in a cohort of mother‐child dyads in rural Pakistan.

The HPA axis plays a central role in coordinating endocrine responses to stress by activation of a complex set of signalling pathways and release of hormones including glucocorticoids (cortisol, cortisone) and dehydroepiandosterone (DHEA). Cortisol and DHEA are synthesised and released in concert by the adrenals, with evidence showing that they mediate opposing biological functions and have different effects on neurological and developmental outcomes (Kamin and Kertes 2017). Cortisone is biologically inactive and derived when cortisol is metabolised by 11β‐hydroxysteroid dehydrogenase (11β‐HSD) (Chapman et al. 2013). Assessing both cortisol and cortisone allows for a more thorough examination of both active and inactive circulating glucocorticoids.

HPA axis function is important for regulating physiologic systems across the body, including immune function, metabolism, and inflammation (Chrousos and Gold 1992). Dysregulation of the HPA axis may lead to worse overall mental and physical health and increased disease risk (Adam et al. 2017). HPA axis activity can be measured non‐invasively through hormone levels in hair. Hair‐derived hormones provide a measure of chronic hormone secretion over the course of months, in contrast with other analytes, such as saliva, that capture concurrent or daily hormone levels (Russell et al. 2012).

Prior research indicates that stress exposure is associated with increased HPA axis activity, often indexed by elevated cortisol and cortisone levels in hair (Li et al. 2023; Stalder et al. ²⁰¹⁷; Vanaelst et al. ²⁰¹³). While fewer studies have examined DHEA, existing work points to low, rather than high, DHEA levels as a result of physiological dysregulation from chronic stress (Kamin and Kertes 2017). Depression represents a particularly prevalent, and often chronic, stressor. However, research linking depression with HPA axis activity is somewhat mixed. Some work shows a positive association between depression and cortisol (Staufenbiel et al. 2013) while other work shows negative or null associations. (Psarraki et al. 2021). Additionally, a growing body of evidence suggests that exposure to prolonged or severe stress, such as more severe depression, may lead to decreased HPA axis activity (Dedovic and Ngiam 2015; O’Keane et al. ²⁰¹²; Wardenaar et al. ²⁰¹¹). This pattern may be due to a negative feedback mechanism within the HPA axis by which excessive activation of the system triggers down‐regulation over time (Danese and McEwen ²⁰¹²; Miller et al. ²⁰⁰⁷).

The effects of stress and depression on HPA axis activity may also extend across generations (Glover et al. 2010). Maternal distress during pregnancy may lead to increased in utero exposure to glucocorticoids for the child, resulting in altered brain development in areas associated with the HPA axis (Howland et al. 2017). This may lead to HPA dysregulation that persists into childhood and beyond. There is evidence pointing to an enduring effect of prenatal stress exposure on child HPA axis activity, however the direction of the effects is somewhat mixed. Some studies find that maternal depression, stress, and psychopathology during pregnancy is associated with increased hair cortisol and cortisone among children when they are 3–6 years old (Molenaar et al. 2019; Slopen et al. ²⁰¹⁸; Thayer et al. ²⁰²⁴). Other work examining cortisol in urine and saliva has shown that prenatal stress and anxiety are associated with blunted cortisol output in childhood and adolescence (O’Donnell et al. ²⁰¹³; Send et al. ²⁰¹⁹). Overall, the association between prenatal maternal depression and child HPA axis activity is complex and may vary based on mediating and moderating factors, such as severity of maternal depressive symptoms (Bleker et al. 2020). For instance, in a sample in rural South India, Fernandes et al. ²⁰¹⁵ found a dose‐dependent and U‐shaped association between maternal depressive symptoms and infant salivary cortisol reactivity, with increased HPA axis activity only among children of mothers with very low and very high depressive symptoms (Fernandes et al. 2015). In addition to prenatal depression, postnatal maternal depression may also impact child stress system function through increased environmental stress exposure (e.g., harsh parenting, household chaos) (S. H. Goodman and Gotlib ¹⁹⁹⁹). Results with postnatal depression vary, with studies showing both increased and decreased cortisol in relation to maternal depression (Bates et al. 2022).

Psychological interventions have shown some promise in altering hair‐derived HPA axis activity associated with stress exposure (Botschek et al. 2023). Randomized trials of cognitive‐behavioural and stress reduction interventions have shown that adults who receive treatment show relative decreases in hair cortisol compared to those who do not receive treatment (La Marca et al. 2024; Puhlmann et al. ²⁰²¹; Romero‐Gonzalez et al. ²⁰²⁰). However, other trials show null results (Kische et al. 2022; Woud et al. ²⁰²¹), suggesting that more research is needed. Studies of dyadic interventions show that targeting parental behaviour may lead to lower hair cortisol among young children (Sullivan et al. 2024). Only a few studies have examined effects of psychological treatments targeting adult mental health on intergenerational HPA axis activity. Among these, two studies of prenatal stress reduction interventions conducted in Spain and the United States show decreased salivary and hair cortisol among infants of mothers who received treatment (Puertas‐Gonzalez et al. ²⁰²³; Urizar and Muñoz ²⁰¹¹). Similarly, one study of a cognitive‐behavioural compassion intervention for parents conducted in the United States found that young children whose parents received treatment showed lower hair cortisol compared to those in the control group (Poehlmann‐Tynan et al. ²⁰²⁰). In LMIC settings, there are few studies examining HPA axis activity as an outcome following psychological interventions. A stress reduction intervention for war‐affected adolescents in Jordan was associated with overall decreases in hair cortisol and a slower rise in cortisol concentration over time (Dajani et al. 2018; Panter‐Brick et al. ²⁰²⁰).

The goal of the current study is to examine the effect of a psychosocial maternal depression intervention on hair‐derived HPA axis indicators in mothers and their 6‐year‐old children in rural Pakistan. The intervention, the Thinking Healthy + Bachpan Care Programme (TH‐BCP), was offered to mothers from the third trimester of pregnancy through 5 years postpartum. Prior evaluations of TH‐BCP have demonstrated promising results for maternal and child health outcomes, although questions remain regarding long‐term sustainability and implementation effectiveness. At 3 months postpartum, mothers in the intervention group exhibited lower depression severity compared to those in the control group (Sikander et al. 2019). By 3 years postpartum, intervention group mothers showed improved overall mental health, while children showed time‐limited improvements in socioemotional functioning that were evident at 6 and 12 months but not sustained at 3 years (Maselko et al. 2020; Sevim et al. ²⁰²⁴). At the 6–8‐year follow‐ups, results indicated that the intervention had sustained effects on maternal depression, though long‐term effects on child outcomes were inconsistent, with benefits observed only in specific subgroups, including children from more disadvantaged families and girls. (manuscript under review).

With regard to the intervention's effect on HPA axis activity, in a subsample of dyads at 12 months postpartum, we found that mothers in the intervention group had lower cortisol and cortisone and higher DHEA compared to mothers in the control group. Children in the intervention group had higher DHEA, but showed no differences in cortisol or cortisone compared to the control group (Baranov et al. 2022). In the current study, we examine HPA axis activity in the full sample one year after the conclusion of the intervention (6 years postpartum) to better understand the persistence of these impacts. We hypothesise that the intervention will reduce mean cortisol and cortisone and increase mean DHEA in mothers and children. As an additional exploratory analysis, we will also examine whether intervention effects are moderated by maternal depression severity at baseline. Of note, although child sex and socioeconomic markers were associated with other outcomes of the intervention, they were outside the scope of this analysis and were not considered as moderators.

2. Method

2.1. Study Sample and Design

Data were drawn from the Bachpan (Urdu for ‘childhood’) study, a cohort study with an embedded cluster randomized trial of a psychosocial maternal depression intervention conducted in a rural sub‐district of Rawalpindi, Pakistan. Participants were recruited during their third trimester of pregnancy and screened for depression using the Patient Health Questionnaire‐9 (PHQ‐9) (Kroenke et al. 2001). Those who screened positive for depression (PHQ‐9 score ≥ 10) resided in village clusters that were randomly assigned in a 1:1 ratio to intervention or control using stratified randomisation, with union council as the stratification factor. Within each union council, an even number of village clusters were identified to ensure balanced allocation. In total, 20 village clusters were assigned to the intervention arm (n = 283 women) and 20 village clusters were assigned to the control arm (n = 287 women). The control group received basic information on depression and referrals for services. Cluster randomisation was chosen to avoid contamination across the treatment and control groups among women living in the same community. In addition, in the same village clusters, women who did not screen positive for depression were invited to participate in the cohort portion of the study (n = 584). Mothers and their children were followed from baseline to when children were 6 years old. HPA axis measures were not included as outcomes in the pre‐registered trial of this intervention, as the main target was maternal depression (Clinical Trial registration: NCT01826903). Sample flow is shown in Figure 1. Ethical approval for human subjects research was obtained from the Institutional Review Boards of the University of North Carolina‐ Chapel Hill, Duke University, and the Human Development Research Foundation in Islamabad, Pakistan. All participating mothers provided informed consent for themselves and their children and children provided assent.

FIGURE 1.

CONSORT sample flow chart.

2.2. Intervention

The Thinking Healthy + Bachpan Care Programme (TH‐BCP) comprises three sequential interventions delivered from pregnancy through 5 years postnatal: Thinking Healthy Programme Peer‐delivered (THPP), Thinking Healthy Programme Peer‐delivered Plus (THPP+), and Thinking Healthy Programme Peer‐delivered Preschool (THPP‐PS). Details of the intervention phases are provided in Table 1.

TABLE 1.

Summary of the thinking healthy + bachpan care programme.

Phase	1	2	3
Intervention name	Thinking healthy programme peer‐delivered (THPP)	Thinking healthy programme peer‐delivered, plus (THPP+)	Thinking healthy programme peer‐delivered, preschool (THPP‐PS)
Timing	Pregnancy–6m	6 months–3 years	3–5 years
Structure	12 weekly sessions prenatally to 2m postpartum, then monthly group sessions	Monthly group booster sessions	Monthly individual sessions
Total number of sessions	14	18	21
Mean sessions attended	11.5	11.4	11.3
Median (Q1, Q3) sessions attended	13 (10, 14)	13 (10, 18)	14 (0, 18)
Areas covered	Mother's health, mother‐child interaction, mother's interaction with others		Mother's health and wellbeing, child wellbeing and development
Delivery agents	Trained peer volunteers with supervision

Mothers in the intervention group first received THPP, adapted from the WHO Thinking Healthy Program (World Health Organization ²⁰¹⁵), utilising cognitive behavioural approaches to reduce depressive symptoms and strengthen social support. THPP consisted of 12 weekly individual sessions delivered from the third trimester of pregnancy through 2 months postpartum, followed by monthly group sessions through 6 months postpartum. From 6 months to 3 years postpartum, participants received THPP+, consisting of 18 group ‘booster’ sessions. The first six booster sessions were delivered monthly, then bimonthly until 36 months postpartum. THPP + sessions were built upon the THPP intervention and were delivered by the same peer facilitators. From 3 to 5 years postpartum, the intervention extended into THPP‐PS, which encompassed individual monthly sessions emphasising child development and well‐being alongside maternal health. Session topics included maternal well‐being (mother's personal health, sleep, and diet), the mother‐child relationship (responding to infant's emotional needs, providing a stimulating environment, feeding), and the mother's relationships with others (increasing social support at home and in the community). Across the sessions, peer facilitators used culturally grounded vignettes to illustrate behavioural activation, cognitive restructuring, and problem‐solving strategies. Sessions also included psychoeducation and mood tracking.

In later phases, Learning Through Play (Rahman et al. 2009) was integrated into the intervention. Learning Through Play is an evidence‐based programme designed to promote children's cognitive, language, and social development. Mothers received education on early childhood development and guidance on age‐appropriate, enriching activities to engage in with their children. The intervention was targeted towards mothers only throughout all phases. All sessions were led by trained peer volunteers who were lay married women living in the same communities as the participating mothers with depression. In the first three years of the programme, peers were successful in gaining and sustaining satisfactory competence (Atif et al. 2019). For mothers, session attendance varied over time, with mothers attending fewer sessions on average as the intervention progressed (see Table 1).

Women in both the intervention and control groups who met criteria for depression (PHQ ≥ 10) received enhanced usual care. This included informing participants of their depression status, providing information on mental health resources, and notifying their respective Lady Health Workers (LHWs). Additionally, all 11 primary care physicians in the subdistrict were trained on the WHO Mental Health Gap Action Program guidelines for maternal depression (World Health Organization ²⁰¹⁰). Given the limited availability of mental health services in the study setting, receipt of care outside of this was unlikely.

2.3. Procedures

Mother and child hair samples were collected at the 6‐year home visit using standardized hair collection procedures (Russell et al. 2012). Approximately 200 strands of hair were clipped form the posterior vertex as close to the scalp as possible. Samples were analysed by Dresden LabService using liquid chromatography mass spectrometry (LC‐MS) (Gao et al. 2013) and are presented in picogrammes per milligramme (pg/mg). Survey‐based measures were also collected during home visits using Open Data Kit.

2.4. Measures

2.4.1. Maternal and Child HPA Axis Activity

The 2 cm of hair closest to the scalp was analysed for cortisol, cortisone, and DHEA. Each cm of hair approximates one month of hormone secretion, with more recent months closer to the scalp (Russell et al. 2012). Therefore, hormone values reported here approximate the previous 2 months of cortisol, cortisone, and DHEA secretion. There were 6 mother cortisol and 5 child cortisol outliers (> 3 SDs above the mean), which were winsorized to 3 SDs above the mean. In addition, there were 94 mother DHEA, 10 child cortisol, and 142 child DHEA values below the detectable limit, which were winsorized to half the minimum values (Walther et al. 2019). In total we have cortisol, cortisone, and DHEA values from 818 mothers and 755 children.

2.4.2. Survey‐Based Measures

Maternal depression was measured at baseline with the PHQ‐9, a 9‐item self‐report of depressive symptoms in the previous two weeks. Scores range from 0 to 27, with higher scores indicating more severe depressive symptoms. Scores of 10–14 are considered moderate depressive symptoms, 15–20 indicate moderately severe depressive symptoms, and 21–27 indicate severe depressive symptoms. This measure has been translated into Urdu and validated in this population (Gallis et al. 2018). Life events were measured at baseline with the Life Events and Difficulties Scale, which asks participants to report on whether they experienced a series of stressful events (e.g., death of a family member, loss of employment) over the previous year (Brown and Harris 1989).

2.5. Analysis

Due to a positive skew, hormone values were log transformed. All intervention effect models were estimated among women who screened positive for prenatal depression; non‐depressed participants were excluded from these models and used only for descriptive comparison. Linear mixed effects models were used to compare outcomes between the intervention and control groups. Because the dependent variables are log‐transformed, regression coefficients refer to changes in log points, which approximate percentage changes. A random intercept was specified for village cluster and fixed effects were specified for intervention group (intervention vs. control), union council (the stratification variable), interviewer, participant age, gender (for children), currently pregnant (for mothers) and variables that were unbalanced between the intervention and control groups at baseline for those in the 6‐year sample (including paternal years of education and stressful life events). We also controlled for variables relevant to hair‐derived HPA axis hormones, including weight of the hair sample, average temperature and humidity in the region on the date of sample collection, the number of days since the participant last washed their hair, and whether the participant had sweat since they last washed their hair (Feneberg et al. 2025; Staufenbiel et al. ²⁰¹⁵). For mothers, we additionally controlled for whether the mother used hair products or any steroid creams. In order to increase the precision of the intervention effect estimate, we interacted all covariates with the intervention arm indicator. This method accounts for the potential heterogeneity of covariates on the outcomes by intervention group (Lin 2013; Sevim et al. ²⁰²⁴). As a sensitivity analysis, we also ran models without hair‐relevant covariates.

Marginal effects of the intervention were derived as average predicted differences between intervention and control arms using post‐estimation margins from the fitted mixed‐effects models, averaging over the observed covariate distribution. To account for potential bias due to differential attrition and missing biomarker data, inverse probability of censoring weights were constructed using logistic regression models predicting availability of a valid 6‐year hair sample. The censoring model included baseline variables that predicted missingness at the 6‐year follow‐up: PHQ‐9 score, maternal education, household crowding (number of people per room), disability, family structure (nuclear vs. non‐nuclear), parity (first child vs. not first child), women's empowerment (decision‐making autonomy), and faith (woman's response to whether she believes faith gives her strength). Stabilised weights were applied in all outcome models so that participants with characteristics associated with higher probability of missingness were upweighted in the analytic sample (Seaman and White 2013). To test for moderation by maternal depression severity during pregnancy, we added baseline PHQ‐9 score to the model and interacted it with study arm. All analyses were conducted in Stata 18.

3. Results

Characteristics of the 818 mothers included in the 6‐year hair sample analysis are presented in Table 2. All sample characteristics were collected at baseline, except child sex, which was collected at 3 months postpartum. On average, mothers were 26.7 years old at baseline, had 1.5 children beyond the target child, and had 7.7 years of education. About 50% of the children were male. Characteristics for the full sample recruited at baseline (n = 1154) and the sample participating in the 6‐year follow‐up, including those who did not provide a hair sample (n = 900), are presented in Supporting Information S1: Tables S1 and S2, respectively. Mean observed (untransformed) cortisol, cortisone, and DHEA values by study group are presented in Figure 2.

TABLE 2.

Descriptive characteristics by study group for mothers in 6‐year hair sample analysis (n = 818).

	Control group (n = 201)		Intervention group (n = 191)		Non‐depressed group (n = 426)
	Mean (SD)/n (%)	Median (IQR)	Mean (SD)/n (%)	Median (IQR)	Mean (SD)/n (%)	Median (IQR)
Baseline sample characteristics
Maternal age	27.4 (4.8)	27 (24–30)	27.1 (4.6)	27 (24–30)	26.4 (4.2)	26 (23–29)
Maternal years education	6.6 (4.5)	7 (4–10)	6.7 (4.1)	8 (5–10)	8.43 (4.5)	10 (5–12)
Paternal years education	8.7 (3.0)*	10 (8–10)	7.9 (3.6)*	9 (6–10)	8.98 (3.3)	10 (8–10)
Asset‐based SES (n)
Lowest quintile	51 (25.4)		52 (27.2)		52 (12.2)
Lower middle quintile	47 (23.4)		48 (25.1)		71 (16.7)
Middle quintile	41 (20.4)		33 (17.3)		96 (22.6)
Upper middle quintile	31 (15.4)		29 (15.2)		101 (23.7)
Upper quintile	31 (15.4)		29 (15.2)		106 (24.9)
Nuclear family (n)	31 (15.4)		32 (16.8)		46 (10.8)
Number of children a	1.7 (1.4)	1 (1–3)	1.8 (1.5)	2 (1–3)	1.21 (1.3)	1 (0–2)
Life events	4.1 (2.3)*	4 (3–6)	4.8 (2.3)*	4 (3–7)	2.87 (2.2)	3 (1–4)
PHQ‐9 total score	14.5 (3.5)	14 (12–17)	14.8 (3.7)	14 (12–17)	2.8 (2.5)	3 (0–4)
PHQ‐9 score category (n) b
Minimal	N/A		N/A		321 (75.4)
Mild	N/A		N/A		105 (24.6)
Moderate	118 (58.7)		102 (53.4)		N/A
Moderately severe	62 (30.9)		63 (33.0)		N/A
Severe	21 (10.5)		26 (13.6)		N/A
Child female (n) c	110 (54.7)		96 (50.26)		206 (48.36)
Year 6 covariates
Current pregnancy (n)	11 (5.47)		12 (6.28)		42 (9.86)
Weather on day of hair sample collection
Average temperature	24.2 (7.3)	26.9 (18.6–30.1)	24.8 (6.9)	27.4 (18.8–29.9)	23.97 (7.07)	26.0 (18.5–30.0)
Average humidity	0.57 (0.16)	0.59 (0.43–0.67)	0.56 (0.16)	0.59 (0.43–0.66)	0.55 (0.15)	0.56 (0.43–0.64)
Hair sample covariates: Mother
Number of days since last hair wash	2.3 (2.03)	2 (1–3)	2.38 (1.95)	2 (1–3)	2.35 (1.89)	2 (1–3)
Sweat since last hair wash (n)	102 (50.8)		103 (53.93)		194 (45.54)
Hair products (n)	184 (91.5)		174 (91.10)		391 (91.78)
Medical creams (n)	6 (3.0)		5 (2.62)		9 (2.11)
Weight of hair sample in mg	7.5 (0.3)	7.5 (7.5–7.5)	7.43 (0.24)	7.5 (7.5–7.5)	7.45 (0.31)	7.5 (7.5–7.5)
Hair sample covariates: Child (n = 755) d
Number of days since last hair wash	2.3 (2.0)	1 (1–2)	2.01 (1.83)	1 (1–2)	2.28 (1.91)	1 (1–3)
Sweat since last hair wash (n)	97 (52.2)		101 (55.8)		195 (48.99)
Weight of hair sample in mg	7.2 (0.9)	7.5 (7.5–7.5)	7.18 (1.07)	7.5 (7.5–7.5)	7.22 (1.00)	7.5 (7.5–7.5)

Abbreviations: PHQ‐9 = Patient Health Questionnaire‐9; SES = socioeconomic status.

^a Number of children was assessed at baseline and does not include the index child.

^b PHQ‐9 scores fall into the following categories: Minimal (score of 0–4), Mild (5–10), Moderate (10–14), Moderately Severe (15–19) and Severe (20–27).

^c Child gender was assessed at the 3‐month visit.

^d Sample sizes for child hair samples: n = 186 Control, n = 181 Intervention, n = 398 Non‐depressed.

^* Indicates statistically significant (p < 0.05) difference between intervention and control groups.

FIGURE 2.

Mean observed HPA hormone levels by study group. Error bars represent cluster‐robust standard errors. Hormone levels are untransformed. DHEA = dehydroepiandosterone.

Marginal effects on log‐transformed hormones are presented in Table 3. There was no evidence of an intervention effect on maternal cortisol (B = 0.04, 95% CI: −0.07, 0.15), with mothers in the intervention group showing cortisol levels approximately 4% higher than mothers in the control group. Similarly, there was no evidence of intervention effects on maternal cortisone (B = 0.04, 95% CI: −0.04, 0.12) or DHEA (B = 0.06, 95% CI: −0.17, 0.30). Sensitivity analyses removing hair‐relevant covariates showed similar results (see Supporting Information S1: Table S3). There was also no evidence of moderation by baseline maternal PHQ‐9 score on cortisol (B = 0.001, 95% CI: −0.04, 0.04), cortisone (B = 0.002, 95% CI: −0.03, 0.03), or DHEA (B = 0.01, 95% CI: −0.08, 0.10), suggesting that the effect of the intervention did not vary by prenatal depression severity. Supporting Information S1: Figure S1 shows the intervention effects on maternal HPA hormones by baseline maternal PHQ‐9 score.

TABLE 3.

Marginal effects of the intervention on log‐transformed maternal and child HPA axis outcomes at 6 years.

	Predictive means (log‐transformed)		Marginal effect	95% CI
	Control	Intervention
Maternal outcomes
Cortisol	1.34	1.38	0.04	−0.07, 0.15
Cortisone	2.43	2.46	0.04	−0.04, 0.12
DHEA	0.53	0.59	0.06	−0.18, 0.29
Child outcomes
Cortisol	0.61	0.75	0.14*	0.01, 0.28
Cortisone	2.22	2.29	0.07	−0.03, 0.17
DHEA	−1.03	−1.25	−0.23	−0.69, 0.24

Note: All HPA axis outcomes are log transformed. Reported marginal effects were derived from the fitted mixed‐effects models using post‐estimation margins and averaged over the observed covariate distribution.

Abbreviations: CI = Confidence Interval; HPA = Hypothalamic Pituitary Adrenal.

^* p < 0.05.

Turning our attention to the child results, there was evidence of an intervention effect on child cortisol (B = 0.14, 95% CI: 0.01, 0.28) such that children in the intervention group showed higher cortisol compared to children in the control group, bringing them closer to children in the non‐depressed group (see Figure 2). Specifically, cortisol was approximately 15% higher in the intervention group compared to the control group. There was no evidence of intervention effects on child cortisone (B = 0.07, 95% CI: −0.03, 0.17) or DHEA (B = −0.23, 95% CI: −0.69, 0.24). Sensitivity analyses removing hair‐relevant covariates showed similar results (see Supporting Information S1: Table S3). For the moderation results, we found a significant interaction between baseline maternal PHQ‐9 score and intervention arm in predicting child cortisol (B = 0.07, 95% CI: 0.02, 0.12). The intervention effect increased as baseline maternal PHQ‐9 score increased, suggesting that the intervention had a stronger effect among children of mothers with more depressive symptoms during pregnancy (see Figure 3). Specifically, the intervention had a significant effect on child cortisol when the baseline maternal PHQ‐9 score was 15 and above (indicating moderately severe and severe depressive symptoms) but not below 15 (indicating moderate depressive symptoms). There were no significant interaction effects for child cortisone (B = 0.02, 95% CI: −0.01, 0.06) and or DHEA (B = 0.13, 95% CI: −0.02, 0.29) (See Supporting Information S1: Figure S2). Although the intervention effect on child cortisone and DHEA increased as maternal baseline PHQ‐score increased, there were no significant intervention effects on these outcomes at any level of baseline maternal depression severity.

FIGURE 3.

Intervention effect on child hair‐derived cortisol by baseline maternal depression severity. PHQ‐9 = Patient Health Questionnaire‐9.

4. Discussion

The goal of this study was to examine the effect of a psychosocial maternal depression intervention, delivered from pregnancy through 5 years postpartum, on mother and child HPA axis activity at 6 years postpartum. We did not find intervention effects on maternal cortisol, cortisone, or DHEA. In children, we found that those whose mothers received the intervention had higher cortisol compared to those in the control group, resulting in cortisol levels that were closer to the non‐depressed group. We also found that treatment for prenatal depression had stronger effects on cortisol among children of mothers with more severe depressive symptoms at baseline. There were no main intervention effects on child cortisone or DHEA. Results suggest that programs targeting perinatal maternal depression may have long‐term benefits for child HPA axis functioning, especially among those with more severe depressive symptoms.

Given that we found intervention effects on maternal cortisol, cortisone, and DHEA with a subsample of participants at 12 months postpartum, our estimates using the 6‐year wave indicate that intervention effects on mothers faded over time. The 6‐year wave was one year after the conclusion of the intervention, during which time other factors, such as stressful life experiences, may have altered HPA axis activity and obscured any intervention impacts on the mothers. In addition, many mothers in the sample had DHEA levels below the detectable limit (11% of the sample), which may have limited our ability to detect intervention effects with this outcome.

For children, recall that we found effects on child DHEA, but not cortisol or cortisone at 12 months. In this study, we saw null results for child DHEA at 6 years. This may be due to the high proportion of samples in which DHEA was below the detectable limit (19%). Low DHEA is normative in children prior to puberty onset (Guran et al. 2015). It is possible that intervention effects will be re‐emerge in this hormone when children reach adolescence and DHEA increases. Our null results with cortisone contrast previous work showing that stress and prenatal maternal psychopathology are associated with increased hair cortisone in youth (Molenaar et al. 2019; Vanaelst et al. ²⁰¹³). It may be that the intervention is primarily associated cortisol secretion, and not its conversion into cortisone via 11β‐HSD activity, although more research is needed to investigate these dynamics (O’Donnell et al. ²⁰¹²).

Our significant results with cortisol at age 6 may reflect increasing sensitivity of child cortisol to maternal depression as children age (Laurent 2017; Slopen et al. ²⁰¹⁸). We found that children in the intervention group had higher cortisol levels than those in the control group. While this is contrary to previous research showing decreasing cortisol levels in children following intervention (Poehlmann‐Tynan et al. ²⁰²⁰; Puertas‐Gonzalez et al. ²⁰²³), comparison with the prenatally non‐depressed group suggests that increased cortisol may be adaptive in this sample. The intervention may have ‘normalised’ HPA axis activity by leading to an increase in cortisol secretion. In the control group, results may reflect blunted cortisol output as a result of chronic stress. Research has shown that the direction of intervention effect on HPA axis activity may vary based on initial HPA axis functioning. For instance, Dajani et al. (2018) found that a stress reduction intervention was associated with increased cortisol among adolescents with a pattern of cortisol hyposecretion and decreased cortisol among those with a pattern of hypersecretion.

The intervention may impact child cortisol by modulating prenatal maternal HPA axis activity. By reducing stress and depression during pregnancy (Sikander et al. 2019), the intervention may have led to lower in utero exposure to glucocorticoids, and thus less HPA axis dysregulation in children (Glover et al. 2010; Howland et al. ²⁰¹⁷). The intervention may have also impacted child exposure to stress postnatally, including lowering exposure to maternal depressive symptoms and related stressors (e.g., negative parenting, household stress) (Bates et al. 2022; Sevim et al. ²⁰²⁴). We found that the intervention effect grew stronger as baseline maternal depression severity increased, suggesting that children of mothers with moderately severe and severe depressive symptoms benefitted most from the intervention.

This study adds to a growing literature showing that treating maternal depression may benefit physiological stress system function in children (Puertas‐Gonzalez et al. ²⁰²³; Urizar and Muñoz ²⁰¹¹). Importantly, our study strengthens this finding by also establishing it in a South Asian LMIC setting. Practically speaking, families in Pakistan are more likely to experience chronic stressors, such as food insecurity, interpersonal violence, and low access to healthcare, than families contributing data to other datasets in high income contexts. Such chronic stressors can impact HPA axis activity (Koss and Gunnar 2017). Simultaneously, there are potential protective factors common in this culture that may buffer mothers and children from chronic stress, such as supportive postpartum practices (e.g., chilla (LeMasters et al. ²⁰²⁰)) and extended family structures that include multiple caregivers in the home (Chung et al. 2020, 2022). The addition of diversity of contexts contributes to our understanding of the generalisability of this literature.

Results suggest that evaluating HPA axis activity as an outcome of psychological interventions may be a promising approach for discovering the full effectiveness of programs. Future research may explore whether changes in HPA axis activity as a result of this intervention translate into changes in child physical and mental health over time. This will be especially important to examine as children age into adolescence, a period of increased risk for the onset of mental health disorders. It will also be important to examine the mechanisms linking maternal depression, and its treatment, with child HPA axis activity. Such mechanisms may include parenting practices (i.e., warm or harsh parenting), mother and child coping strategies, and mother‐child relationship quality (S. H. Goodman and Gotlib ¹⁹⁹⁹). Furthermore, results of this study provide guidance on how to target interventions in the future, with findings suggesting that children of mothers with more severe depression may benefit most from treatment.

This study benefitted from many strengths, including a 6‐year longitudinal follow‐up and the inclusion of a non‐depressed comparison sample. However, results should be interpreted in light of some limitations. First, we did not measure HPA axis activity in mothers during pregnancy, so we do not have a measure of hormones before the start of the intervention. This limits our ability to study HPA axis change in the mothers pre‐to post‐intervention. We also cannot study potential in utero glucocorticoid exposure for the children. Similarly, we only use one data collection point for our outcomes at 6 years postpartum, which means we cannot study change in HPA axis activity over time, nor can we examine bi‐directional associations between maternal HPA axis activity and depressive symptoms. We also do not have information on maternal depressive symptoms before the third trimester of pregnancy. Chronicity of maternal depression before and during pregnancy may be an important moderator of the intervention's impact on HPA axis activity. Finally, although hair‐derived HPA hormones offer a measure of chronic hormone secretion, it does not allow for more fine‐grained measures of HPA axis reactivity or diurnal rhythm.

In conclusion, this study demonstrated that a psychosocial maternal depression intervention delivered in a LMIC setting had effects on child HPA axis activity at 6 years postpartum. This is promising evidence for the potential intergenerational effects of maternal depression treatment on child physiological stress system regulation. As these children age, it will be important to examine how early interventions may shape their stress system functioning, and how this may impact their physical and mental health trajectories.

Funding

This project was funded by grant R21MH131866 from the National Institute of Mental Health and grant R01HD075875 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.