The effect of maternal childhood maltreatment on postpartum mother–child bonding and maternal hair glucocorticoids

ABSTRACT

Background: Maternal experiences of childhood maltreatment (CM) constitute a risk factor for impairments in the mother–child relationship. One mechanism underlying this intergenerational transmission may be maternal hypothalamic–pituitary–adrenal axis dysregulation. Yet, few studies have examined different maltreatment subtypes, including emotional neglect, considered concurrent depressive symptoms, and used long-term integrated glucocorticoid measures.

Objective: This study aimed to investigate associations between maternal CM history, postpartum glucocorticoids in hair, and mother–child bonding. In exploratory analyses, we tested whether specific subtypes of CM had differential implications for glucocorticoid secretion and bonding.

Methods: During pregnancy, N = 269 mothers from the prospective cohort study DREAM_HAIR provided retrospective information on CM and current information regarding psychological and hair-related variables. Hair samples were collected 8 weeks after delivery for quantification of maternal long-term hair cortisol and cortisone concentrations in 2-cm scalp-near hair samples. Mother–child bonding was measured 8 weeks and 14 months after birth using the Postpartum Bonding Questionnaire.

Results: While bivariate correlations revealed significant associations of CM with bonding and hair cortisol, regression findings showed CM was associated with impaired bonding 8 weeks (overall CM trend-level; emotional neglect p = .038) and 14 months (emotional neglect trend level p = .041) after birth, however not after controlling for depressive symptoms at the time point of the outcome. In regression analyses, CM was not associated with maternal hair glucocorticoids 8 weeks postpartum. Maternal hair glucocorticoid concentrations were not related to mother–child bonding and did not mediate associations between CM and mother–child bonding.

Conclusion: Data tentatively suggest that mothers with CM experiences, in particular emotional neglect, may be at risk for suboptimal bonding to their child, however current depressive symptoms seem to be more important. Our data provide no evidence for a crucial role of glucocorticoid secretion, yet aetiological processes of long-term glucocorticoid secretion and bonding are complex and more severely affected samples should be examined.

HIGHLIGHTS

• Maternal childhood maltreatment experiences, in particular emotional neglect, were associated with impaired mother–child bonding postpartum, however not after considering current depressive symptoms.

• While maternal childhood maltreatment correlated with maternal hair cortisol 8 weeks postpartum, this was not confirmed in regression analyses controlling for relevant confounders.

• Maternal hair glucocorticoid concentrations were not associated with impaired mother–child bonding.

Antecedentes: Las experiencias maternas de maltrato infantil (MI) constituyen un factor de riesgo para deterioro en la relación madre-hijo. Un mecanismo subyacente a esta transmisión intergeneracional puede ser la desregulación del eje hipotálamo-hipófisis-adrenal. Sin embargo, pocos estudios han examinado los diferentes subtipos de maltrato, incluido negligencia emocional, han considerado síntomas depresivos concurrentes y han utilizado mediciones integradas de glucocorticoides a largo plazo.

Objetivo: Este estudio tuvo como objetivo investigar las asociaciones entre los antecedentes maternos de MI, glucocorticoides de cabello en el postparto y el vinculo madre-hijo. En el análisis exploratorio, probamos si subtipos específicos de MI tenían implicancias diferenciales para la secreción de glucocorticoides y el vínculo.

Métodos: Durante el embarazo, N = 269 madres del estudio de cohorte prospectivo DREAMHAR proporcionaron información retrospectiva en MI e información actual sobre variables psicológicas y capilares. Se recolectaron muestras de cabello 8 semanas después del parto para cuantificar las concentraciones de cortisol y cortisona de cabello a largo plazo en muestras de 2 cms de cabello cercano al cuero cabelludo. El vínculo madre-hijo se midió a las 8 semanas y 14 meses después del nacimiento utilizando el Cuestionario de Vinculo del Postparto.

Resultados: Si bien las correlaciones bivariadas revelaron asociaciones significativas del MI con el vínculo y cortisol capilar, los hallazgos de regresión mostraron que el MI se asoció con una alteración del vínculo a las 8 semanas (nivel de tendencia general del MI; negligencia emocional p = .038) y a los 14 meses (negligencia emocional p = .041) después del nacimiento, sin embargo, no después de controlar los síntomas depresivos en el momento del resultado. En los análisis de regresión, el MI no se asoció con glucocorticoides del cabello materno a las 8 semanas postparto. Las concentraciones de glucocorticoides en el cabello materno no se relacionaron con el vínculo madre-hijo y no mediaron asociaciones entre el MI y el vinculo madre-hijo.

Conclusión: Estos datos sugieren tentativamente que madres con experiencias de MI, en particular negligencia emocional, pueden estar en riesgo de tener un vínculo subóptimo con su hijo; sin embargo, los síntomas depresivos actuales parecen ser mas importantes. Nuestros datos no proporcionan evidencia de un papel crucial de la secreción de glucocorticoides, sin embargo, los procesos etiológicos de la secreción de glucocorticoides a largo plazo y el vinculo son complejos y se deben examinar en muestras más gravemente afectadas.

KEYWORDS:

• Child maltreatment
• childhood trauma
• emotional neglect
• DREAM study
• hair cortisol
• mother–child bonding

PALABRAS CLAVE:

• Maltrato infantil
• trauma infantil
• negligencia emocional
• estudio DREAM
• cortisol capilar
• vinculo madre-hijo

1. Introduction

Mother–child bonding is considered an important component of the mother–child relationship and an early predictor of infant developmental outcomes (Le Bas et al., 2022). The term ‘bonding’ describes the mother’s self-reported affectional tie to her child and should be conceptually distinguished from ‘attachment’ or ‘mother–child interactions’, as these concepts reflect the child’s tie to their mother (Bicking Kinsey & Hupcey, 2013). An early stable bond enables positive parenting behaviours (Muzik et al., 2013) and fosters more optimal child development (Blum et al., 2022; Le Bas et al., 2020; 2022). In contrast, bonding impairments are characterized by maternal anger, irritability, and rejection towards the child (Brockington et al., 2006). Considering the significant impact bonding can have on the offspring’s development, investigating both biological and psychological factors that affect the formation of a stable bond is pivotal.

Childhood maltreatment (CM), which includes all types of abuse and neglect of a child under 18 years by a parent or caregiver, is a global issue with high prevalence rates ranging from 6.5%–54.8% depending on the region and type of CM (Stoltenborgh et al., 2015). It has been shown to not only impact the victims’ physical and mental health, but may even continue to affect their own offspring, among others through intergenerational transmission of disrupted mother–child relationships (Vaillancourt et al., 2017). As such, studies have emphasized an impact of maternal history of CM on poorer mother–child relationship outcomes, e.g. poorer mother–child interaction or reduced sensitivity (Hughes and Cossar, 2016; Savage et al., 2019; Souch et al., 2022). Regarding bonding, initial research indicates increased risk for bonding impairments associated with CM (Lehnig et al., 2019; Muzik et al., 2013) and suggests that current psychopathological, including depressive symptoms, may partly explain this relationship (Frohberg et al., 2022; Lehnig et al., 2019; Muzik et al., 2013). Interestingly, differential effects of specific subtypes of CM (i.e. sexual abuse, physical abuse, emotional abuse, physical neglect, and emotional neglect) on bonding were reported, suggesting that emotional neglect was associated with higher bonding impairment (Lehnig et al., 2019), whereas physical abuse predicted less bonding impairment (Frohberg et al., 2022; Lehnig et al., 2019).

Existing literature indicates that alterations of the mother’s endocrine stress response system, the hypothalamic–pituitary–adrenal (HPA) axis with the main output being the steroid hormone cortisol, could link maternal CM with poorer bonding (Nyström-Hansen et al., 2019). Outside the perinatal context, findings linking CM to altered cortisol levels have been mixed. A recent meta-analysis found blunted cortisol stress reactivity in adults exposed to CM, but no effect on basal cortisol secretion using traditional cortisol assessment methods (e.g. salivary cortisol) or long-term assessment methods using hair cortisol analysis (Schär et al., 2022). Yet, another meta-analysis focusing specifically on hair cortisol (HairF), as a more robust and reliable measure (Stalder & Kirschbaum, 2012), and adversity, including CM, indicated an effect of CM on both attenuated and increased HairF (Khoury et al., 2019). This highlights the need to account for moderating variables (e.g. type and timing of adversity, clinical status) in determining the pattern of cortisol alterations following adversity (Khoury et al., 2019). A recent study for instance found blunted hair cortisol only in participants who reported childhood sexual abuse (Burnett et al., 2023). Furthermore, Pittner et al. (2020) recently examined hair cortisone (HairE) concentrations, finding elevations due to abuse, whereas no previous study in this context has examined the ratio between HairF and HairE. Thus, considering the enzyme 11β-hydroxysteroid-dehydrogenase type 2 (11β-HSD2) converts cortisol to its inactive form cortisone, the additional assessment of HairE concentrations (Stalder et al., 2013) and the ratio of HairF/HairE (Scharlau et al., 2018) may represent a promising approach to obtain further long-term markers of cortisol metabolism, also in the perinatal period (Bergunde et al., 2022; Jaramillo et al., 2023; Scharlau et al., 2018).

Considering the perinatal period represents a unique time with great endocrine changes (King et al., 2022) and high psychosocial demands, the relationship between CM and cortisol levels might differ in this time. Studies employing short-term measures of cortisol found both directions of cortisol dysregulation in CM-exposed women during pregnancy (elevated diurnal, blunted phasic cortisol; Epstein et al., 2021; elevated diurnal cortisol; Rinne et al., 2023) and postpartum (elevated salivary cortisol; Gonzalez et al., 2012; reduced salivary cortisol; Juul et al., 2016). Studies examining longer-term cortisol measured in hair during pregnancy and CM indicated childhood neglect and abuse to be associated with reduced and elevated HairF (and HairE) respectively (Kelsall et al., 2023; Penner et al., 2023). To add to that, recent research in our cohort revealed that maternal experience of at least one lifetime traumatic event, but not specifically CM, was associated with elevated prenatal maternal HairF and HairE, respectively (Steudte-Schmiedgen et al., 2023). Regarding the postpartum period, CM showed a positive linear (Nyström-Hansen et al., 2019) and curvilinear (Broeks et al., 2023) association with HairF at 3 and 4 months postpartum respectively.

Cortisol is known to exert diverse effects on neurobiological systems involved in behavioural and affective processes and initial evidence has demonstrated it may affect the mother–child relationship (Almanza-Sepulveda et al., 2020). During the first week after birth, mothers who had higher salivary cortisol levels were found to show more sensitive mothering behaviours (Barrett & Fleming, 2011). As the postpartum period progresses in time, evidence suggests that alterations in HPA axis function may be indicative of impaired mothering (Finegood et al., 2016). In studies that did not explicitly address mothers who had experienced CM and that employed shorter-term cortisol measures, both directions of cortisol dysregulation were found to be related to less optimal parenting (Crockett et al., 2013; Finegood et al., 2016). In contrast, while Bos et al. (2018) found no association between prenatal and postpartum maternal cortisol levels and parenting behaviour at 6 weeks postpartum, Beijers et al. (2022) found salivary cortisol levels at 6 weeks postpartum to be positively related to caregiving quality when offspring were 3 years old. Initial findings employing hair cortisol analysis indicate that increased maternal HairF may be related to poorer parenting at 4–6 months postpartum (Nyström-Hansen et al., 2019; Tarullo AR et al., 2017) and might act as a mediator of the association of CM and child abuse potential when children were aged 5–12 years (Kluczniok et al., 2020). While promising evidence indicates that postpartum altered HPA axis activity might link maternal CM to impaired parenting (Gonzalez et al., 2012; Juul et al., 2016; Nyström-Hansen et al., 2019), research on bonding and long-term endocrine secretion is lacking to date, in particular also with regard to subtypes of CM.

Our prospective study based on a large community sample aimed to fill these gaps by investigating associations of maternal CM, postpartum hair glucocorticoids (GCs), and maternal self-report mother–child bonding. First, we aimed to investigate the influence of maternal CM on bonding. We assumed that CM has an impact on poorer bonding 8 weeks and 14 months postpartum, respectively (Savage et al., 2019), and examine whether this and all subsequent associations remain significant after including current depressive symptoms as suggested by some previous studies (Frohberg et al., 2022). Second, we examined the effect of maternal CM on hair GCs at 8 weeks postpartum. Considering previous findings in the perinatal period indicating elevated HairF in traumatized (expectant) mothers (Broeks et al., 2023; Nyström-Hansen et al., 2019; Steudte-Schmiedgen et al., 2023), a positive direction of GC dysregulation was postulated. Third, we investigated whether altered hair GC levels predicted poorer mother–child bonding 8 weeks and 14 months postpartum, respectively (Almanza-Sepulveda et al., 2020). Fourth, we tested whether altered hair GCs can explain the association between maternal CM and poorer mother–child bonding. Because of previous heterogenous findings concerning hair GCs and bonding, no direction of association was postulated. Finally, we aimed to explore whether different types of CM contribute to differential implications for long-term integrated maternal GC levels and bonding.

2. Methods

2.1. Study design and participants

The present study is part of the prospective cohort study ‘DREAM – Dresden Study on Parenting, Work, and Mental Health’ and the biological sub-study DREAM_HAIR (Kress et al., 2019). The latter aims to shed light on the influence of long-term hair GCs on perinatal mental health and child outcomes, with several measurement points from birth to middle childhood (see Figure 1).

Figure 1. Measurement points of DREAM and the endocrine sub-study DREAM_HAIR. Note: Relevant measurement points for the current study were T1 DREAM, T2 DREAM, T3 DREAM, T1 DREAMhair, and T2 DREAMhair. Mothers completed ^aT1 DREAM questionnaires during pregnancy at gestational week M = 27.0 (SD = 5.5), ^bT2 DREAM questionnaires M = 8.3 weeks (SD = 1.4) after the anticipated birth date, and ^cT3 DREAM questionnaires M = 13.8 months (SD = 0.6) after birth. For the sub-study DREAMhair, mothers participated at ^dT1 DREAMhair M = 21.9 days (SD = 10.9) before birth and at ^eT2 DREAMhair M = 8.4 weeks (SD = 1.1) after the anticipated birth date.

From June 2017 to the end of 2020, expectant parents, of whom n = 2,227 expectant mothers had returned the first questionnaire, were recruited for the main DREAM study in hospital information evenings or birth preparation courses in Dresden, Germany. The DREAM study included participants with a current pregnancy, who lived in Dresden or surroundings, and whose German language skills were sufficient to complete the questionnaires. Additionally, for DREAM_HAIR, expectant parents who participated in the main study and whose due dates were at least 4 weeks away, were recruited via telephone screening. Participants were included in DREAM_HAIR if they reported no hair loss or baldness, a hair length of at least 2 cm, no serious physical illness within the past five years (e.g. cancer), and if they had not taken any GC-containing medication within the last 3 months. For hair sample collection after birth, families were mailed collection materials that included detailed instructions and a self-generated hair protocol (Stalder et al., 2014) to check for potential confounders, such as hair washes per week or hair treatments.

For the present study, only data of (expectant) mothers were used. Exclusion criteria for this study comprised multiples and preterm delivery, too early maternal hair sampling (within 7 weeks after delivery), factors compromising hair analysis (e.g. too low hair mass), and factors impacting GC secretion (e.g. smoking; Stalder et al., 2017). After exclusion of mothers who did not complete the instrument assessing the outcome bonding, i.e. the Postpartum Bonding Questionnaire (PBQ; Brockington et al., 2006) at T2 DREAM or T3 DREAM, sample 1 consisted of N = 258 and sample 2 of N = 241, respectively (see Figure 2).

Figure 2. Flowchart of exclusion criteria resulting in final samples. Note: Data used in this study were collected until January 31^st, 2022 (data collection still ongoing, ninth version of the quality-assured dataset). CTQ = Childhood Trauma Questionnaire. PBQ = Postpartum Bonding Questionnaire. bonding = Mother–child bonding.

Attrition analyses (see supplementary materials) indicated that participants who did not complete T3 (but have completed T1 and T2) were more likely to have no academic degree (15.4% versus 36.1%; χ²(1) = 4.48, p = .048) than participants that completed all relevant time points. Regarding all other study variables, no significant differences between completers and non-completers emerged.

The DREAM study was approved by the Ethics committee of the Technical University Dresden (No: EK 278062015). Prior to the first assessment, participants gave their written informed consent according to the Declaration of Helsinki.

2.2. Measures

2.2.1. Sociodemographic and hair-related characteristics

Sociodemographic information (e.g. age and education) was measured at T1 DREAM. Via a self-generated hair protocol (Stalder et al., 2014) we assessed hair-related (e.g. weekly washes, hair treatment) and health-related characteristics (e.g. medication intake, alcohol consumption) at T2 DREAM_HAIR.

2.2.2. Psychological measures

To measure the extent of maternal CM, participants regularly completed the German version of the Childhood Trauma Questionnaire (CTQ; Bernstein & Fink, 1998) at T1 DREAM_HAIR. The CTQ is a retrospective self-report instrument that aims to measure experiences of overall CM and of five subtypes in childhood and adolescence. It consists of 28 items ranging from 1 ( =  ‘not at all’) to 5 ( =  ‘very often’), with five items each assessing the respective subtype of CM and three items measuring the tendency to downplay or deny. Higher values therefore indicate a higher degree of CM (Range 25–125). The CTQ yielded good internal consistency in this study (α = .88). For descriptive purposes subtypes of CM were categorized as present in at least moderate severity according to cut-offs (Bernstein et al., 2003): emotional abuse > 12; emotional neglect > 14; physical abuse > 9; physical neglect > 9; sexual abuse > 7.

To assess bonding, the German version of the widespread Postpartum Bonding Questionnaire (PBQ; Brockington et al., 2001; Weigl & Garthus-Niegel, 2021a) was used at T2 DREAM and T3 DREAM. The PBQ comprises 25 items in which mothers are asked to think of the most challenging time with their child and to rate specific experiences on a scale ranging from 0 (‘never’) to 5 (‘always’). Higher values thus indicate a higher extent of bonding impairment (Range 0–125). Internal consistency of the PBQ was α = .87 at T2 and α = .91 at T3.

Depressive symptoms were included as a confounder and measured at T2 DREAM and T3 DREAM using the German version of the Edinburgh Postnatal Depression Scale (EPDS; Bergant et al., 1998; Weigl & Garthus-Niegel, 2021b). The EPDS is a self-report instrument, consisting of ten items with a scale ranging from 0 to 3, higher values indicate higher severity of depressive symptoms over the past week (Range 0–30). In this study, internal consistency was α = .76 at T2 and α = .82 at T3.

2.2.3. GC analyses

Hair strands of each participant with a minimum length of 2 cm and a diameter of 3 mm were taken scalp-near from a posterior vertex position at T2 DREAM_HAIR and stored in aluminium foil. For analyses of HairF and HairE, samples were sent in four batches to the Kirschbaum Laboratory at the Technische Universität Dresden. GCs were assessed in the scalp-near 2 cm segment via liquid chromatography-tandem mass spectrometry (Gao et al., 2013), reflecting integrated cortisol and cortisone levels over the last two months (i.e. in this study reflecting the first eight weeks after birth). Quality control procedures confirmed that the coefficient of variability (CV%) for cortisol (7.4%) and cortisone (9.6%) averaged across the four batches was below the recommended 10%.

2.3. Statistical analyses

Analyses were conducted with IBM SPSS Statistics 28 and 29 and R version 4.3.2 (R Core Team, 2023). Non-detectable values (n = 2 for HairF) were replaced with the lowest detectable value in the sample to reduce bias (Herbers et al., 2021). For psychological variables, missing values were replaced with the mean if at least 80% of items were completed. As HairF, HairE, and the HairF/HairE ratio lacked normality, they were log-transformed. Outliers of the logarithmized GCs (M ± 3 SD) were excluded (HairF: n = 2; HairE: n = 5; HairF/HairE ratio: n = 4). A significant effect was defined as having p < .05 and values p > .05 but p < .10 are discussed carefully as potentially indicating an effect at trend-level. We examined associations between the main study variables (i.e. overall CM, hair GCs, bonding, depressive symptoms, and the CM subtypes) via Spearman correlations and to control the false discovery rate, the Benjamini-Hochberg procedure was applied with FDR of 5% (Benjamini & Hochberg, 1995). Then we checked if potential confounders, i.e. storage time, hair treatment, hair washing, time point of hair sample collection in weeks postpartum, sun exposure, use of oral contraceptives, body mass index, parity, age, and education (Stalder et al., 2017), were associated with hair GCs by calculating Spearman correlations (supplementary materials). Significantly associated variables (hair washing frequency (HairF), storage time (HairF, HairE), hair treatment (HairF, HairE), and time point of hair collection (HairF, HairE)) were controlled for in subsequent analyses with the respective GC as outcome or mediator variable. To test effects of COVID-19 pandemic exposure (categorization regarding participation before or after 9 March 2020; Petzold et al., 2020) hierarchical regression analyses controlling for batch were calculated. COVID-19 exposure affected neither hair GCs or bonding and was not considered further (see supplementary materials). To control for the effects of batch, we regressed batch on the HairF, HairE, and HairF/HairE ratio concentrations and saved the residuals for use in subsequent analyses (Marceau et al., 2021).

Main analyses involved hierarchical multiple regressions to test associations of overall CM and CM subtypes with (1) bonding and (2) hair GCs as well as the association between hair GCs and (3) bonding. Given studies have shown an impact of depressive symptoms on HairF, HairE, and bonding (Green et al., 2021; Psarraki et al., 2021; Tichelman et al., 2019), for each regression, the main predictor was entered first, followed by hair GC-specific confounders identified above for analyses predicting hair GCs, and finally maternal depressive symptoms at the time when the respective outcome variable was assessed. Finally, using the PROCESS macro model 4 (Hayes, 2018), we tested if the hair GCs mediated the relation between overall CM and subtypes of CM indicated in hierarchical regression analyses and bonding. As some residuals lacked normality or homoscedasticity, bootstrapping with 2,000 iterations for regression and mediation analyses was used (Field, 2013).

We detected n = 4 multivariate outliers with Mahalanobis Distance (see supplementary materials for outlier characteristics). Inspection of these outliers revealed no clearly diverging patterns and as sensitivity analyses indicated no relevant differences when excluding these outliers, results including outliers are reported.

3. Results

3.1. Sample description

For an overview of sample characteristics, see Table 1. Mothers were more highly educated than the general German population (Statistisches Bundesamt, 2020), with 66.2% having an academic degree. The majority was born in Germany and was primiparous. At T2 n = 21 (7.8%) and at T3 n = 31 (11.5%) were above the cut-off of 26 in the PBQ, indicating impaired bonding. Except for emotional abuse, prevalence rates of CM were lower than in the general German population (Witt et al., 2017). More precisely, n = 42 (15.6%) mothers reported having experienced at least one subtype of moderate to severe CM and n = 20 (7.4%) reported at least moderate emotional abuse, n = 6 (2.2%) reported physical abuse, n = 12 (4.5%) reported sexual abuse, n = 15 (5.6%) reported emotional neglect, and n = 13 (4.8%) reported physical neglect.

Table 1. Sample description (N^a= 269).

Variable	n (%) or mean ± SD (Range)
Sociodemographics
Age in years^b (M, SD, Range)	30.15 ± 3.86 (18–42)
Marital status^b (n, %)
Married	123 (45.9)
Unmarried	138 (51.5)
Divorced	7 (2.6)
Academic degree^b (n, %)	178 (66.2)
Mother tongue German^b (n, %)	253 (94.1)
Primiparous^b (n, %)	224 (83.3)
Employed before pregnancy^b (n, %)	227 (84.4)
Psychological variables
Childhood maltreatment (CTQ)^c (M, SD, Range)	32.12 ± 7.79 (25–72)
Emotional abuse (M, SD, Range)	7.30 ± 2.88 (5–21)
Physical abuse (M, SD, Range)	5.46 ± 1.48 (5–20)
Sexual abuse (M, SD, Range)	5.31 ± 1.18 (5–14)
Emotional neglect (M, SD, Range)	8.12 ± 3.27 (5–19)
Physical neglect (M, SD, Range)	5.93 ± 1.65 (5–15)
Mother-child bonding T2 (PBQ)^d (M, SD, Range)	12.83 ± 8.86 (0–52)
Mother-child bonding T3 (PBQ)^e (M, SD, Range)	14.68 ± 10.85 (0–81)
Depressive symptoms T2 (EPDS)^d (M, SD, Range)	5.11 ± 3.44 (0–17)
Depressive symptoms T3 (EPDS)^e (M, SD, Range)	5.52 ± 4.01 (0–18)
Hair-related variables^f
Time point of hair sampling (in weeks) (M, SD, Range)	8.38 ± 1.10 (7–14)
Storage time (in weeks) (M, SD, Range)	39.53 ± 14.42 (10–72)
Hair washes per week (M, SD, Range)	2.77 ± 1.18 (0.25–7)
BMI (M, SD, Range)	24.93 ± 4.03 (16.92–41.24)
Hair treatment (n, %)	53 (19.7)
Use of oral contraceptives (n, %)	25 (9.3)
Returned to menstrual cycle (n, %)	5 (1.86)
Sunlight exposure per week (in minutes)^g (M, SD, Range)	583.46 ± 499.76 (0–3480)
Raw maternal glucocorticoid concentrations at T2^f
HairF (pg/mg) (M, SD, Range)	10.40 ± 10.87 (0.73–79.10)
HairE (pg/mg) (M, SD, Range)	22.31 ± 14.69 (2.93–113.00)
HairF/HairE ratio (pg/mg) (M, SD, Range)	0.45 ± 0.26 (0.12–2.07)

Note: CTQ = Childhood Trauma Questionnaire. Bonding = mother-child bonding. PBQ = Postpartum Bonding Questionnaire. EPDS = Edinburgh Postnatal Depression Scale. BMI = Body Mass Index. HairF = Hair cortisol. HairE = Hair cortisone.

^a n varied minimally due to missing data and exclusion of outliers.

^b Measured at T1 DREAM (during pregnancy).

^c Measured at T1 DREAM_HAIR (during pregnancy).

^d Measured at T2 DREAM (8 weeks postpartum).

^e Measured at T3 DREAM (14 months postpartum).

^f Measured at T2 DREAM_HAIR (8 weeks postpartum).

^g Sunlight exposure was calculated as an interaction between total sunlight exposure and the frequency of wearing head covering.

3.2. Correlation analyses

Spearman correlations (Table 2) showed small significant correlations of CM with impaired bonding at T2 and T3 (r_s's = .20, p’s = .003 – .006), depressive symptoms at T2 and T3 (r_s's > .21, p’s < .003), and HairF at T2 (r_s = .15, p = .028). Furthermore, emotional and physical neglect were significantly associated with HairF (r_s's = .15, p’s < .030). GCs were not significantly associated with impaired mother–child bonding (p’s < .181). Emotional (T2) and physical abuse (T2, T3), and emotional neglect (T2, T3) were significantly correlated with bonding (r's > .15, p’s < .028). Also, depressive symptoms and bonding were moderately significantly correlated across assessment points (r's > .31, p’s < .001).

Table 2. Spearman correlation matrix of main study variables (N = 269^a).

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13
1. Childhood maltreatment^b	–	.20**	.20**	.15*	.08	.13	.29**	.21**	.81**	.46**	.31**	.87**	.65**
2. Bonding T2^c		–	.59**	.10	.02	.09	.48**	.31**	.15*	.16*	.06	.21**	.07
3. Bonding T3^c			–	.00	.00	.07	.39**	.43**	.12	.16*	.03	.23**	.06
4. HairF T2				–	.77**	.57**	.13	.02	.07	.05	-.01	.15*	.15*
5. HairE T2					–	.10	.05	.00	.05	.03	-.08	.12	.02
6. HairF/HairE ratio						–	.16*	.11	.06	.08	.05	.09	.15*
7. Depressive symptoms T2^d							–	.52**	.26**	.17*	.15*	.26**	.17*
8. Depressive symptoms T3^d								–	.25**	.12	.10	.17*	.03
9. Emotional abuse^b									–	.43**	.26**	.61**	.35**
10. Physical abuse^b										–	.24**	.32**	.26**
11. Sexual abuse^b											–	.12	.07
12. Emotional neglect^b												–	.51**
13. Physical neglect^b													–

Note: Two-tailed testing. GCs were log-transformed. Significant correlations *p < .05, **p < .01 presented in bold. All p-vales were FDR corrected.

^a n varied slightly due to missing data.

^b Childhood Trauma Questionnaire.

^c Postpartum Bonding Questionnaire.

^d Edinburgh Postnatal Depression Scale.

3.3. Main analyses

3.3.1. Associations between maternal CM and bonding

In multiple regression analyses, maternal CM was associated with impaired bonding at 8 weeks postpartum at trend-level ($\text{β}$ = .17, p = .056), however this was no longer the case when controlling for current depressive symptoms ($\text{β}$ = .06, p = .427). Regarding CM subscales, only emotional neglect was significantly associated with impaired bonding at 8 weeks postpartum ($\text{β}$ = .19, p = .038), however only at trend-level after adjusting for current depressive symptoms ($\text{β}$ = .15, p = .092; Table 3). Emotional abuse, physical abuse, physical neglect, and sexual abuse were not significantly associated with bonding at 8 weeks postpartum, neither before nor after adjusting for depressive symptoms (p’s = .545–.944; Table 3).

Table 3. Hierarchical regression analysis of overall CM and the CM subtypes predicting bonding at 8 weeks postpartum.

Variable	ß	B	SE	BCa 95% CI	p
Model 1^c
Childhood maltreatment^a	${.17}^{†}$	.19	.10	[.01, .40]	.056^ϯ
Model 2^d
Childhood maltreatment^a	.06	.07	.08	[-.08, .24]	.427
Depressive symptoms^b	.46	1.20	.17	[.88, 1.53]	< .001
Model 1^e
Emotional abuse^a	.01	.02	.31	[-.88, .51]	.944
Physical abuse^a	-.02	-.10	.80	[-1.04, 2.52]	.901
Sexual abuse^a	.01	.07	.47	[-.65, 1.14]	.874
Emotional neglect^a	.19	.52	.25	[.06, 1.08]	.038
Physical neglect^a	.01	.04	.54	[-1.07, 1.09]	.942
Model 2^f
Emotional abuse^a	-.05	-.16	.27	[-.66, .29]	.545
Physical abuse^a	-.02	-.11	.61	[-.92, 1.95]	.831
Sexual abuse^a	-.02	-.14	.37	[-.86, .76]	.651
Emotional neglect^a	${.15}^{†}$	.40	.23	[-.04, .92]	.092^ϯ
Physical neglect^a	-.01	-.04	.46	[-.90, .75]	.938
Depressive symptoms^b	.46	1.20	.17	[.89, 1.50]	< .001

Note: ß = Standardized beta coefficient. B = Unstandardized beta. SE = Standard error for unstandardized beta based on 95% bias corrected and accelerated bootstrap confidence interval (2,000 iterations). BCa 95% CI= 95% bias corrected and accelerated bootstrap confidence interval. All values significant at p < .05. ^ϯ = trend-level significant.

^a Childhood Trauma Questionnaire.

^b Edinburgh Postnatal Depression Scale at 8 weeks after anticipated birth.

^c F(1,256) = 7.51, p = .007, R² adj. = 0.03.

^d F(2,255) = 37.71, p = < .001, R² adj. = 0.22.

^e F(5,252) = 2.02, p = .076, R² adj. = 0.02.

^f F(6,251) = 13.12, p = < .001, R² adj. = 0.22.

Regarding bonding at 14 months postpartum, maternal CM was not significantly associated, neither before ($\text{β}$ = .11, p = .116) nor after controlling for depressive symptoms ($\text{β}$ = .08, p = .260; see supplementary materials). Only emotional neglect was significantly prospectively associated with impaired bonding at 14 months postpartum at trend-level ($\text{β}$ = .15, p = .098), but not after adjustment for depressive symptoms ($\text{β}$ = .12, p = .191; see supplementary materials). Emotional abuse, physical abuse, physical neglect, and sexual abuse were not significantly associated with bonding at 14 months postpartum, neither before nor after adjusting for depressive symptoms (p’s = .191–.943; see supplements for full results).

3.3.2. Associations between maternal CM and GCs

In multiple regression analyses, overall maternal CM (all p ≥ .232) and maternal CM subscales (all p ≥ .120) were not associated with HairF, HairE, or the HairF/HairE ratio after adjusting for hair-related confounders identified through preliminary analyses described in section 2.3 and maternal depressive symptoms (see supplementary materials for full results).

3.3.3. Associations between GCs and bonding

In multiple regression analyses, hair GCs were not significantly associated with mother–child bonding 8 weeks postpartum before (all p = .127–.464) and after adjusting for depressive symptoms (all p = .360–.525). The same pattern of results emerged for bonding 14 months postpartum (all p = .513–.997; see supplementary materials for full results).

3.3.4. Mediation analyses

Mediation analyses testing whether hair GCs mediated associations between maternal CM and bonding, controlling for depressive symptoms and the described hair-specific confounders, indicated no significant indirect effects (see Table 4 and Figure 3a,c). Considering emotional neglect was related at trend-level to mother–child bonding at T2 and T3, mediation was also tested with emotional neglect as the predictor. Results revealed no significant indirect effects for bonding at T2 and bonding at T3 (see supplementary materials and Figure 3b,d).

Figure 3. Mediation model of HairF mediating the relationship between childhood maltreatment and emotional neglect with mother-child bonding at 8 weeks and 14 months postpartum. Note: Standardized regression coefficients are shown. All dotted arrows were non-significant with 95% CI including zero and p > .05. All analyses were controlled for depressive symptoms and relevant hair-related confounders.

Table 4. Mediation results testing whether hair GCs mediated an effect of maternal CM on bonding 8 weeks and 14 months postpartum using PROCESS Model 4.

	Path a			Path b			Path c’ (direct effect)			indirect effect
	B	SE	CI	B	SE	CI	B	SE	CI	B	SE	CI
CM -> Bonding T2^a
HairF^b	0.00	0.00	-0.00, 0.01	1.37	1.41	-1.27, 4.27	0.06	0.08	-0.10, .24	0.00	0.01	-0.01, 0.02
HairE^c	-0.00	0.00	-0.00, 0.00	1.26	2.18	-2.80, 5.90	0.08	0.08	-0.06, 0.26	-0.00	0.00	-0.01, 0.01
Ratio^d	0.00	0.00	-0.00, 0.00	2.06	2.65	-3.06, 7.13	0.06	0.08	-0.10, 0.24	0.00	0.01	-0.01, 0.02
CM -> Bonding T3^a
HairF^e	0.00	0.00	-.00, 0.01	-1.37	1.95	-5.40, 2.48	0.11	0.10	-0.07, 0.31	-0.00	0.01	-0.03, 0.01
HairE^f	0.00	0.00	-0.00, 0.00	0.04	2.46	-4.77, 4.84	0.14	0.10	-0.04, 0.36	0.00	0.00	-0.01, 0.01
Ratio^g	0.00	0.00	-0.00, 0.01	0.28	3.25	-6.25, 6.42	0.11	0.10	-0.07, 0.32	0.00	0.01	-0.02, 0.02

Note: Path a, Path b, the direct, and indirect effects of the mediation analyses are presented. HairF, HairE, and the HairF/HairE ratio were log-transformed, outliers were excluded and batch-corrected residuals were used. Heteroscedasticity consistent standard errors were applied. Confounders in these analyses were depressive symptoms and the relevant hair-specific variables described above. B = Unstandardized regression coefficient. SE = Standard error. CI = 95% bootstrap confidence interval based on 2,000 bootstrap samples. bonding = Mother-child bonding.

^a Assessed via the Postpartum Bonding Questionnaire at T2 DREAM (8 weeks after birth) and T3 DREAM (14 months postpartum) and CM assessed with the Childhood Trauma Questionnaire at T1 DREAM (during pregnancy).

^b n = 256.

^c n = 253.

^d n = 254.

^e n = 231.

^f n = 228.

^g n = 230.

4. Discussion

This study aimed to increase our understanding of the sequelae of CM by examining associations of maternal CM severity, postpartum HPA axis activity, and mother–child bonding. While we found significant bivariate correlations between overall CM, physical abuse, and emotional neglect with mother–child bonding at 8 weeks and 14 months postpartum, regression analyses showed only an association of overall maternal CM with impaired bonding at 8 weeks (trend-level) and in particular an association with the subtype emotional neglect and bonding at 8 weeks and 14 months (trend-level) postpartum, respectively. However, these associations lost significance when considering concurrent depressive symptoms at each respective timepoint. Also, while HairF was significantly positively related to overall CM, emotional neglect, and physical neglect in bivariate corrrelations, hair GCs showed no significant associations with CM in regression analyses adjusting for confounders and no relations with bonding at 8 weeks and 14 months postpartum, respectively.

These findings partly align with the extensive literature describing a transmission of one’s own CM experiences to parenting behaviour among postpartum women (Hughes & Cossar, 2016). The present results also confirm previous research documenting the importance of the mother’s current mental health state for bonding outcomes, with depressive symptoms being significantly positively associated with bonding impairment (Höflich et al., 2022) and 57.1% of mothers with postpartum depression admitted to a Mother and Baby Unit reporting impaired bonding (Gilden et al., 2020). Thus, our findings highlight that while maternal CM and emotional neglect experiences might also be related to the self-reported emotional tie of a mother to her child, more variance in bonding outcomes is explained by maternal postpartum mental health, i.e. depressive symptoms. Consequently, it may be that while CM is a relevant predictor of several pregnancy outcomes (Kern et al., 2022), current depressive symptoms may be of greater importance in identifying mothers at risk regarding postpartum bonding difficulties. This fits with findings from Frohberg et al. (2022), who found no direct association between CM subtypes and mother–child bonding, but did confirm a significant mediating effect of maternal psychopathological symptoms regarding the relationship between overall CM and CM subtypes with mother–child bonding. In addition, our results align with the study by Muzik et al. (2013), who also found an important role for maternal psychopathology in the CM-bonding relationship. Together, these results highlight that supporting mothers with postpartum depressive symptoms represents an important prevention pathway (Chamberlain et al., 2019; Weidner et al., 2021) and could help reduce the impact CM experiences can have on the affective mother–child relationship. Findings can also be seen from the perspective that even if a mother experienced CM, this may not causally determine her bond to her child. Instead amongst many other factors, her current mental health, which can be effectively treated with psychotherapy in the postpartum period (Höflich et al., 2022), appears more central.

The fact that overall CM was only associated with bonding 8 weeks postpartum at trend-level, whereas emotional neglect was significantly associated may be explained by the fact that emotional neglect was frequently reported by participants in the present investigation and showed the highest variance compared to other CM subtypes. Hence, this may have increased our ability to detect effects (Bühner, 2011). It is important to consider that this study involved a community sample of mothers with an above-average health and education status (Kress et al., 2019) and below-average CM burden compared to the general German population (Witt et al., 2017). Emotional neglect, which includes inadequate nurturance or affection, is one of the most prevalent forms of CM worldwide that has been linked to negative parenting outcomes (reviewed in Hughes & Cossar, 2016) and impaired bonding (Lehnig et al., 2019). In our study, emotional neglect experiences were related to impaired bonding (before controlling for depressive symptoms) and were associated with postpartum depressive symptoms both in the early and later postpartum period, emphasizing the importance of this CM form for postpartum bonding and mental health outcomes.

This finding further confirms previous studies that suggested differential patterns of bonding depending on the particular CM subtype experienced (physical abuse: Frohberg et al., 2022; emotional neglect: Lehnig et al., 2019) and supports the promising approach to additionally assess CM subtypes. Furthermore, results pertaining to emotional neglect underscore the importance of a supportive early environment where parents are emotionally available and sensitive towards their child. If a mother lacked these experiences in her own childhood, this may interfere with developing a stable affective bond with her own offspring, potentially via reduced emotional expressiveness (e.g. Yu et al., 2023) or higher levels of alexithymia (Khan & Jaffee, 2022), the inability to express, describe, or distinguish one’s emotions.

Concerning hair GCs, contrary to prior findings, regression analyses revealed no associations between overall maternal CM and HairF. Our results are in line with previous findings in this cohort showing no association between CM and hair GCs during pregnancy (Steudte-Schmiedgen et al., 2023). However, other studies within the postpartum period have indicated a positive association (Broeks et al., 2023; Nyström-Hansen et al., 2019), while outside the postpartum period a negative effect in mixed-sex adult samples (Schalinski et al., 2019) or no effect of CM subtypes on HairF (Oresta et al., 2021) has been reported. Clearly more research within and outside the perinatal period is needed, taking account also of other variables, such as the specific age of CM occurrence (Schalinski et al., 2019), further lifetime trauma (Steudte-Schmiedgen et al., 2023), and epigenetic signatures (Varadarajan et al., 2022), which might play a role in HPA axis modifications. It should also be highlighted that we thoroughly controlled for potential confounding factors of hair GCs (e.g. Stalder et al., 2017), increasing confidence in robustness of our null findings. Furthermore, previous studies that found a significant effect of overall CM on HairF in the postpartum context examined mainly clinical samples (Nyström-Hansen et al., 2019), whereas we investigated a non-clinical community sample, which could explain diverging findings.

Contrary to hypotheses, hair GCs were not significantly associated with impaired bonding across the postpartum period. Results contradict findings of traditional cortisol and HairF studies, which mainly delineated an association between cortisol and impaired parenting (Beijers et al., 2022; Crockett et al., 2013; Finegood et al., 2016; but see Bos et al., 2018). Diverging findings could be due to differences in sample characteristics, as some studies that found a significant effect examined samples suffering from chronic poverty (Crockett et al., 2013; Finegood et al., 2016) or clinical samples (Juul et al., 2016; Nyström-Hansen et al., 2019). Moreover, in contrast to our approach, studies examining HairF mostly did not control for current maternal depressive symptoms (Nyström-Hansen et al., 2019; Tarullo AR et al., 2017; except for Broeks et al., 2023). Furthermore, as postulated by Bos et al. (2018), the timing of cortisol assessment might play a crucial role. Barrett and Fleming (2011) emphasized cortisol directly after delivery as a marker of higher maternal sensitivity, whilst other studies (Crockett et al., 2013; Finegood et al., 2016) reported a negative association between cortisol and maternal sensitivity in the later postpartum period (between 4–24 months). Hence, the study of Bos et al. (2018) at 6 weeks postpartum and our study, depicting the first 8 weeks of cortisol secretion after birth, may reflect a transition period wherein no clear relationship between cortisol and the mother–child relationship can be quantified. Finally, to our knowledge, this is the first study to examine cortisol with respect to self-report mother–child bonding. While paying attention to affective components is important, the lack of additional behavioural assessment of mother–child interactions is a limitation of the present study. Future studies examining both components are necessary to draw conclusions.

Finally, mediation analyses did not yield significant results, suggesting the negative effect maternal CM experiences may have on bonding at 8 weeks and 14 months postpartum are not explained by alterations in long-term GC secretion. Instead, our findings give an indication that it might be current (i.e. postpartum depressive symptoms) rather than past adversity (i.e. maternal CM in our study) that affects bonding currently. Thus, our findings support the notion that CM is not inextricably linked to bonding impairments and prevention efforts are likely to have a positive impact if current maternal mental health is targeted.

This study has several strengths. Firstly, we focused specifically on affective aspects of mother–child bonding, and additionally examined HairE and the HairF/HairE ratio, and through hair analysis as a reliable and valid marker of long-term GC secretion. Moreover, we had a large sample, assessed CM subtypes, and investigated associations between hair GCs and the mother–child relationship not only cross-sectionally, but also longitudinally. Nevertheless, several limitations remain. Our study was conducted in a community sample with an above-average health and educational status (Kress et al., 2019) and with lower CM prevalence rates than the general German population (Witt et al., 2017), potentially compromising external validity. Moreover, 11.5% of women had the maximum score on the denial scale of the CTQ, which is comparable to previous investigations (Häuser et al., 2011), suggesting some CM experiences were not captured due to the tendency to trivialize among these women. Also, five women in our sample reported having returned to their menstrual cycle at time of T2 assessment. While sensitivity analyses excluding these women revealed no changes in main findings, oxytocin concentrations have been shown to fluctuate across the menstrual cycle (Engel et al., 2019) and oxytocin plays an important role in early contact and interactions between mother and child (Scatliffe et al., 2019). Thus, future studies should aim to disentangle the complex hormonal changes taking place during the postpartum period and how they interact with hair cortisol to affect mother–child bonding. Finally, attrition analyses also revealed that participants who dropped out at T3 more often had no academic degree, which might reduce generalizability of findings, but provides a valuable perspective on highly educated samples who may have partly overcome possible CM in the past.

5. Conclusion

This study tentatively indicates that overall maternal CM history and in particular emotional neglect experiences may constitute a risk factor for poorer bonding in a community sample of postpartum mothers. However, current symptoms of depression experienced in the postpartum period appear to be more important for mother–child bonding. Specifically, our data suggest the experience of CM alone does not have an extensive effect on mother–child bonding in a highly educated sample when taking account of depressive symptoms. To add to that, hair GCs 8 weeks postpartum might not be affected by maternal CM experiences and in turn might not affect mother–child bonding in a healthy community sample. Further research is needed to examine whether maternal CM and hair GCs might have differential implications for different aspects of the mother–child relationship and should consistently consider the current mental health status to differentiate effects. It would also be valuable to add the fathers’ perspective and explore dyadic or triadic approaches within this area of research (Seefeld et al., 2022). Moreover, the implications of maternal CM for the child’s endocrine and epigenetic profile and possible long-term development need to be addressed. However, our study provided promising preliminary evidence that in addition to overall CM, emotional neglect may be particularly relevant for mother–child bonding in the early postpartum period. Consequently, assessing maternal emotional neglect history in perinatal care could help identify mother–child dyads at risk and thus implement targeted prevention strategies (Weidner et al., 2021).

Acknowledgements

We want to thank all (expectant) mothers for supporting our project. Furthermore, we want to thank all cooperating clinics and midwives for providing access to potential participants as well as all students who performed the recruitment. Data were collected and managed using Research Electronic Data Capture (Harris et al., 2009, 2019). REDCap is a secure, web-based application designed to support data capture for research studies, hosted at the ‘Koordinierungszentrum für Klinische Studien’ at the Faculty of Medicine of the Technische Universität¨ at Dresden, Germany. LB, MK, MB, SSc, and SGN conceived the research question. MB performed the initial statistical analyses and drafted the initial manuscript under supervision of LB, MK, SSc, and SGN. LB and MK prepared the GC-data for statistical analyses, supported the initial statistical analyses, revised the statistical analyses, and prepared the manuscript for publication. LB, MK, IJ, VW, and JTM supported the conduction of the study. JTM and VW prepared the main study data for statistical analyses and IJ provided assistance with statistical procedures. WG and KW provided resources for the acquisition of data in the DREAM_HAIR study and contributed with their expertise. SGN acquired the funding, was responsible for conception and design of the DREAM study as well as the coordination and supervision of the (ongoing) data collection, and she contributed with her expertise in the research field and the interpretation of the data. All authors critically reviewed the manuscript, approved the final version as submitted, and agreed to be accountable for all aspects of the work.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The dataset analyzed during the current study is not publicly available due to legal and ethical constraints. Public sharing of participant data was not included in the informed consent of the study. The dataset is available from the corresponding author on reasonable request.

Additional information

Funding

This work was supported by the German Research Foundation (‘Deutsche Forschungsgemeinschaft’) under grant numbers GA 2287/4-1, and GA 2287/4-2. The Article Processing Charges (APC) were funded by the joint publication funds of the TU Dresden, including Carl Gustav Carus Faculty of Medicine, and the SLUB Dresden as well as the Open Access Publication Funding of the DFG.