Introduction

Establishing a robust maternal-infant bond in early life is central to infant survival and the development of social-emotional regulation (Ainsworth et al. 1978; Brockington et al. 2006). This bond generates maternal affective experiences which motivate maternal caregiving (e.g. emotional warmth) and contingent responsiveness (e.g. feeding and soothing) designed to promote infant survival and healthy development (Belsky et al. 1984; Feldman 2012). Despite this central role in healthy infant and child development, bonding problems are not uncommon. Between 10 and 25% of postpartum mothers referred to psychiatrists report bonding problems ranging from infant-focused anxiety, hostility, and aggression to rejection (Brockington 2004; De Falco et al. 2014; Lovejoy et al. 2000; Robson and Moss 1970; Seng et al. 2013). Bonding problems escalate to neglect and abuse in severe cases (Brockington 2004; Brockington 1996). Such infants are at risk of lifelong disadvantage, including on-going attachment problems, cognitive delay and emotional and behavioural dysregulation.(De Wolff and van Ijzendoorn 1997; Murray et al. 1996; Tarantola 2018).

Maternal perinatal depression is one of the most well studied risk factors for maternal-infant bonding problems and the usual focus of intervention in the perinatal period (Fisher et al. 2016; Hiscock et al. 2008; Powell et al. 2013; Rossen et al. 2016; Rossen et al. 2019; Rowe and Fisher 2010; Tsivos et al. 2015). However, reported effect sizes have been consistently low-to-modest (Alderdice 2013; Bakermans-Kranenburg 2005; Fontein-Kuipers 2014), suggesting a need to look beyond intervening in the perinatal period alone. More recently, it has become clear that maternal postpartum depression mostly does not arise at the time of pregnancy, but is commonly preceded by similar problems well prior to pregnancy, in adolescence and young adulthood (Patton et al. 2015). This more expansive understanding of aetiology raises a question about whether preconception mental health might predict postpartum maternal-infant bonding problems, opening the possibility of very different strategies to tackle root causes of poor bonding well prior to becoming a parent (Hall et al. 2015; Muzik et al. 2013; Seng et al. 2013), with the aim of disrupting intergenerational cycles of risk (Patton et al. 2015).

The purpose of this study was to examine the extent to which elevated depression and anxiety symptoms prior to becoming a parent predict maternal-infant bonding problems in the first year postpartum, using rare Australian prospective intergenerational data collected over two generations and 22 years of observation in the Victorian Intergenerational Health Cohort Study. Specifically, we examined the extent to which elevated depression and anxiety symptoms from adolescence to young adulthood before pregnancy predict maternal-infant bonding problems at 2 and/or 12 months postpartum, net of preconception socio-demographic confounders and continuity of symptoms into the perinatal period.

Methods

Study design and participants

Data were drawn from the Victorian Intergenerational Health Cohort Study (VIHCS), a prospective study of preconception predictors of infant and child health which is described in detail elsewhere (Patton et al. 2015). It arose from a cohort study commencing in 1992 in Victoria, Australia (The Victorian Adolescent Health Cohort Study; VAHCS) (Patton et al. 2014). Briefly, a close-to-representative sample of 1943 mid-secondary school students (1000 female) were assessed using a broad range of age-appropriate mental health and psychosocial factors bi-annually in adolescence (VAHCS waves 1–6: mean age 14.9–17.4 years) and three times in young adulthood (VAHCS waves 7–9: mean ages 20.7, 24.1, and 29.1 years). VIHCS began in 2006 during the ninth wave of VAHCS with a focus on parental mental health, substance use, partner relationship quality, parent-infant bonding, and early patterns of infant temperament and behaviour.

Between 2006 and 2013 (participant ages 29–35 years, encompassing median maternal and paternal age for Australian births (Australian Bureau of Statistics 2013), female VAHCS participants were screened bi-annually for pregnancies. Participants were invited to complete telephone interviews in trimester three (VIHCS wave 1), 2 months’ postpartum (VIHCS wave 2), and 1 year postpartum (VIHCS wave 3) for every infant born during VIHCS screening, and thus, many women participated with more than one infant. Parents or guardians of VAHCS participants provided informed written consent at recruitment, and participants provided informed verbal consent at subsequent VAHCS and VIHCS waves. Protocols were approved by the human research ethics committee at the Royal Children’s Hospital, Melbourne.

Measures

Preconception exposure measures

Depression and anxiety symptoms in adolescence were assessed in waves 1–6 (age 14–17 years) using the computerized version of the revised Clinical Interview Schedule (CIS-R) (Lewis et al. 1992). Participants self-administered the questionnaire using laptop computers in the classroom, with telephone follow-up of those absent from school on assessment days. At each wave, total score was dichotomised at ≥ 12 to identify mixed depression-anxiety symptoms at a level lower than major disorder, but which a general practitioner would view as clinically significant (Lewis et al. 1992). The scale has demonstrated good reliability (α = 0.88) (Lewis et al. 1992). Presence of elevated depression and anxiety symptoms during adolescence was identified (0 = no waves, 1 = ≥ 1 wave) for each participant.

Depression and anxiety symptoms in young adulthood were assessed using the CIS-R at wave 7 (age 20). At waves 8 and 9 (ages 24 and 29, respectively), the General Health Questionnaire (GHQ-12) (Goldberg et al. 1997; Patton et al. 2014) was used with a cut-off score ≥ 3 to denote a validated threshold that has been found to indicate psychological distress that a health practitioner would be concerned about with sensitivity 76% and specificity 83% (Donath 2001), and corresponds to a CIS-R threshold ≥ 12 (Lewis et al. 1992). All young adult assessments were administered via computer-assisted telephone interview, by trained interviewers following a standardized assessment protocol. The scale has demonstrated good reliability (α = 0.75–.90) (Hankins 2008). Presence of elevated depression and anxiety symptoms during young adulthood was derived as per the adolescent waves (0 = no waves, 1 = ≥ 1 wave).

Persistence of elevated depression and anxiety symptoms from adolescence to young adulthood was classified for each participant using a four-level variable based on the presence or absence of elevated depression and anxiety symptoms in the adolescent and young adulthood waves (0 = no elevated symptoms, 1 = adolescent elevated symptoms only, 2 = young adult elevated symptoms only, and 3 = both adolescent and young adult elevated symptoms).

Perinatal outcome measure

Maternal-infant bonding was assessed at 2 and 12 months postpartum using the Postpartum Bonding Questionnaire (PBQ) (Brockington et al. 2001; Brockington et al. 2006). The PBQ is designed to screen for maternal-infant bonding problems as perceived by the mother and consists of 25 items representing four factors: ‘general impaired bonding’ (e.g. I feel happy when my baby smiles or laughs—reversed); ‘rejection and anger’ (e.g. I feel angry with my baby); ‘infant-focused anxiety’ (e.g. My baby makes me feel anxious); and ‘risk of abuse’ (e.g. I feel like hurting my baby). Each item is scored on a 6-point scale from 0 (‘always’) to 5 (‘never’). Previous research has shown that a total summed score has greater specificity and predictive value than the ‘general impaired bonding’ subscale score for bonding problems diagnosed through expert consensus using interview and case record data (Brockington et al. 2006). Based on this, responses for analyses reported in this study were summed to create a single continuous distribution representing ‘total bonding problems’ (range 0 to 60). Summed continuous measures were also created for ‘rejection and anger’ (range 0–35) and ‘infant-focused anxiety’ (range 0–20). Low numbers in this community sample reporting risk of abuse (n = 3) prohibited analysis. Furthermore, this factor has not been replicated in community-based factorial studies (Arian et al. 2019). The scale and its subscales have demonstrated adequate reliability (α = 0.63–0.79) (Mathews et al. 2019). The total scale and subscale scores were each standardized to obtain z-scores, such that mean associations can be interpreted in units of standard deviations.

Covariates

Potential baseline confounding variables

All analyses were controlled for theoretically relevant preconception baseline demographic variables derived using information collected via self-report during mothers’ adolescence, including mothers’ parents’ high school completion and divorce/separation, mother’s high school completion and mother’s ethnicity (Caucasian versus non-Caucasian) (Bhopal and Donaldson 1998).

Maternal antenatal and postpartum depressive symptoms

Analyses estimating controlled direct effects (see below) were in addition adjusted for maternal antenatal and concurrent postpartum depressive symptoms to examine preconception effects independent of continuity of symptoms into the perinatal period, and account for possible effects of maternal mood on reporting bonding quality. Maternal antenatal and postpartum depressive symptoms were measured in trimester three of pregnancy, at 2 months and at 1 year postpartum using the Edinburgh Postnatal Depression Scale (EPDS) (Cox et al. 1987). For each assessment, presence of elevated depressive symptoms was defined as a total score ≥ 10, a threshold that is appropriate for use in community samples and when administered via telephone (Gibson et al. 2009). This cut-off has been recommended for use in detecting mild-to-severe postnatal depression, with a recent meta-analytic estimates of 95% sensitivity and 82% specificity in detecting major depressive disorder (National-Collaborating-Centre-for-Mental-Health 2018). The scale has demonstrated good reliability (α = 0.88) (Cox et al. 1987).

Statistical analysis

We estimated associations between continuity of maternal preconception depression and anxiety symptoms and subsequent maternal-infant bonding at each postpartum timepoint separately, using linear regression. Models were fitted within a generalized estimating equation (GEE) framework to account for correlation between observations due to clustering of children within families. For each outcome, we conducted a series of progressively adjusted analyses, sequenced by timing of covariate. We first adjusted for baseline demographic characteristics. We then repeated analyses further adjusting for maternal antenatal and concurrent postnatal depressive symptoms to examine the extent to which preconception associations were independent of maternal depressive symptoms that persist into the perinatal period. This provides an estimate of the controlled direct effect of preconception depressive symptoms under certain assumptions (VanderWeele 2014).

Analyses included all mothers who responded at least once in each phase (adolescent, young adult, and postnatally). There were low levels of missing data on most variables (< 10%). We addressed potential bias due to missing data using multiple imputation by chained equations, including all analysis variables and three auxiliary variables in imputation models (maternal age at birth of study child, maternal parental history of divorce or separation, and parity) (van Buuren et al. 1999). Thirty imputed datasets were produced, and parameter estimates were obtained by pooling results across imputed datasets using Rubin’s rules (Rubin 1987). In supplementary analyses, we repeated analyses using available case data for comparison. Additionally, in supplementary work, analyses were repeated with a binary version of the outcome (higher versus lower levels of bonding problems) within a logistic regression GEE framework. Confirmatory supplementary analyses were conducted using the continuous exposures and an interaction term used to capture stage-specific and persistent symptoms across adolescence and young adulthood. To do this, adolescent and young adulthood continuous depression and anxiety scores were derived by standardizing (z-score) continuous scores at each wave, and selecting the maximum standardized score within each developmental period. Interpretation of these analyses was based on inspection of the contour plots. All analyses were conducted in Stata 15 (StataCorp 2015).

Results

Figure 1 presents the flow of participants through the study. Of the 465 VAHCS women reporting 748 live births during the screening period for VIHCS, 398 (86%) VAHCS women participated in VIHCS with 609 (81%) infants. Of these, 3 women did not participate postnatally, resulting in a final analysis sample of 395 women with 606 infants for this study. There were no notable differences between women retained in VAHCS at VIHCS commencement and therefore screened for pregnancies and the total VAHCS study sample on measured demographic, mental health, or risky behaviours variables at VAHCS study entry. Similarly, women who participated in VIHCS were broadly representative of those with live births during screening on measured baseline characteristics (Spry et al. 2020).

Fig. 1
figure 1

Sampling and ascertainment of the Victorian Intergenerational Health Cohort Study (VIHCS), from 2006 to 2014. *Eligibility defined as all live births occurring during VIHCS screening (1 September 2006 to 30 June 2013). †Of the VAHCS women who did not participate for one or more eligible VIHCS children, 67 were completely excluded and the remaining women were recruited to participate in the study with at least one other child. ‡The cohort remained open during waves 1–3, so that parents could enter the study at wave 2 or 3 if they had missed earlier wave/s. Many parents participated with more than one child born during the VIHCS recruitment phase. **The analysis sample comprises all infants with preconception data from both adolescence and young adulthood, and at one or both of the 2 months and 1 year postpartum assessments

Full size image

Table 1 presents the estimated preconception and perinatal characteristics of women who participated in VIHCS. For 1 in 5 infants, mothers reported that their parents had either divorced or separated. Over a third of the sample were from homes where neither parent completed high school. Women reported elevated depression and anxiety symptoms at some point prior to conception for the majority of infants (387; 64%). Women reported symptoms of postnatal depression for 7% of infants at 2 months postpartum, and 8% of infants at 12 months postpartum. Scores on maternal-reported mother-infant bonding problems were low, with a mean score of 8 at each of 2 and 12 months postpartum.

Table 1 Estimated preconception and perinatal characteristics of 606 infants to 395 women
Full size table

Table 2 presents the mean score for postpartum bonding problems at 2 and 12 months postpartum by history of preconception depression and anxiety symptoms. Overall, there appeared to be a pattern where those reporting persistent elevations in depression and anxiety symptoms from adolescence to young adulthood, as well as young-adulthood-limited depression and anxiety symptoms, reported higher levels of bonding problems.

Table 2 Mean unstandardized bonding problem scores at 2 and 12 months postpartum, by preconception history of elevated depression and anxiety symptoms, in 606 infants of 395 women
Full size table

Table 3 presents associations between preconception depressive and anxiety symptoms and later bonding problems at 2 months postpartum. After adjustment for preconception socio-demographic confounders, women who reported a history of persistent preconception mental health problems had standardized mean scores around one-third of a standard deviation higher than women without a history of mental health problems on the total bonding problem scale (β = 0.30 [95% CI 0.04, 0.55]) and the infant-focused anxiety subscale (β = 0.31 [0.05, 0.56]). There was weak evidence of an association with rejection-anger. After further adjusting for antenatal and concurrent postpartum maternal depression to estimate the controlled direct effect of preconception depression on bonding problems at 2 months, mean differences across all three bonding outcome attenuated. Results were similar when repeated using logistic regression (Appendices A-B), when using (non-imputed) available case data (Appendix C-D), and when using continuous exposures (Appendices E-F).

Table 3 Estimated adjusted associations between continuity of preconception elevated depression and anxiety symptoms and subsequent maternal-infant bonding problems at 2 months postpartum, in 606 infants of 395 women
Full size table

Table 4 presents associations between preconception history of depressive and anxiety symptoms and later bonding problems at 12 months postpartum. After adjustment for preconception socio-demographic confounders, there were links between persistent preconception mental health problems and all three bonding outcomes, with larger standardized mean differences than those observed for depression at 2 months postpartum. Young-adulthood-limited preconception problems also predicted bonding problems at 12 months postpartum with effect sizes similar to those for persistent preconception problems. After further adjusting for antenatal and concurrent postpartum maternal depression to estimate the controlled direct effect of preconception depression on bonding problems at 12 months, associations with infant-focussed anxiety remained. There was weaker evidence of an association with general bonding; similarly, the association with maternal rejection and anger towards the infant attenuated somewhat. Again, results were similar when repeated using logistic regression (Appendices A-B), when using (non-imputed) available case data (Appendix C-D), and when using continuous exposures (Appendices E-F).

Table 4 Estimated adjusted associations between continuity of preconception elevated depression and anxiety symptoms and subsequent maternal-infant bonding problems at 1 year postpartum, in 606 infants of 395 women
Full size table

Discussion

Elevated depression and anxiety symptoms in the decades prior to conception are associated with higher levels of maternal-infant bonding problems in the first year postpartum. This was particularly so in those reporting a history of depression and anxiety symptoms in young adulthood, regardless of whether elevated symptoms persisted from adolescence or first emerged in the young adult period. Importantly, associations strengthened from 2 to 12 months. The high proportion of young women reporting elevated depression and anxiety symptoms in the adolescent and young adult years is consistent with other prospective studies (Copeland et al. 2011; Moffitt et al. 2010) and suggests that supporting mental health of girls and women well before pregnancy not only brings contemporaneous benefits but may also yield longer term (intergenerational) benefits in reducing postpartum bonding problems (Patton et al. 2018).

Maternal history of depression and anxiety symptoms was most clearly associated with caregiver anxiety about the infant (infant-focused anxiety) across the first year of life. However, there was also evidence of heightened maternal rejection and anger towards the infant at 12 months. This suggests that a history of depression and anxiety symptoms may impact bonding in ways well beyond caregiver anxiety, with increases in anger and hostility (which are common symptoms of postnatal depression (Ou and Hall 2018; Parfitt and Ayers 2012)) being part of a larger multi-final outcome of mood regulation problems from earlier life stages. In all cases, preconception risk relationships intensified, not diminished, across the postpartum period. It is possible that the increasing complexity of care for an older infant may be an important activator of bonding problems in women with a persistent history of depression and anxiety from adolescence onwards (Rossen et al. 2017). In other words, those who are still having difficulties bonding at this stage may be more likely to be those who entered pregnancy and parenthood with pre-existing vulnerability. We suggest that this area warrants further research.

Understanding the way in which preconception experiences may upset healthy developmental pathways prior to becoming a parent is important, because intervention at the earliest opportunity prior to conception is preferable to intervening post-birth when the demands on maternal resources are high and where effect sizes for existing interventions have been modest at best (Tsivos et al. 2015). Our findings suggest that continuity of symptoms into the ante- and postnatal periods may partially mediate risk for poor bonding in those with a preconception history. Indeed, it is possible that postpartum bonding problems may be a temporally specific behavioural expression of depressive symptoms. We examined these possibilities by adjusting for depressive symptoms during pregnancy and at the time of bonding assessment at 1 year postpartum. Despite some attenuation, evidence remained of preconception associations with postpartum bonding, with adjusted associations strongest for infant-focused anxiety and for bonding assessed at 12 months postpartum. These findings implicate additional mediating processes beyond continuity of symptoms alone.

Another possible explanation is that women with pre-existing depression and anxiety symptoms may already have some difficulties in nurturing relationships and may carry forward these difficulties into pregnancy and parenthood (Chamberlain et al. 2019). For example, being in a caring and supportive relationship with a partner after the birth of a child may play an important role. Preconception depression and anxiety symptoms may also disrupt a myriad of other social pathways, from education to partner selection, all of which could impact on the quality of the maternal-offspring bond. So too qualities of the child that are associated with women’s history of preconception and antenatal mental health problems, such as infant reactivity (Spry et al. 2019), may affect the bond by influencing the mother’s reaction to her baby (Nolvi et al. 2016). This is particularly likely if there is a mismatch with maternal personality or expectations. Such mechanisms of risk transmission, if demonstrated in future research, would place further weight on the importance of intervening in adolescence and young adulthood, and suggest new targets for preventive intervention that extend beyond mental health, to relationship building interventions in adolescence and young adulthood as well as educational interventions to complete secondary school, and engage further skills training post-secondary school (Merry et al. 2012).

The major strength of this study is in its prospective intergenerational design with data collected over 20 years from adolescence to young adulthood, and into the postpartum period for subsequent pregnancies. This design limits recall bias, including reporting bias for individuals with symptoms at the time of reporting. Furthermore, retention in VAHCS and participation of eligible women in VIHCS was high, with 88% of eligible VAHCS women, and 81% of infants, participating in VIHCS. Strategies to retain participants included flexible timing of assessments, option of completing a short form with core measures included only, and the opportunity to reengage at each wave even if missing at prior waves. Women participating in VIHCS remained representative of the original VAHCS sample on measured mental health, substance use, and demographic characteristics at baseline in parental adolescence. Nonetheless, as with all cohort studies, these groups may differ on unmeasured characteristics.

Levels of missing data at most waves were low. Potential bias due to missing data was addressed using multiple imputation. Other limitations included only assessing infants born to women aged 29–35 years, leaving open a possibility that the risk profiles of older and younger mothers differ. We also combined anxiety and depression symptoms in the exposure variable. While comorbidity and sequential morbidity rates are quite high (Angold et al. 1999; Costello et al. 2003), they may have different effects on bonding problems. While a mother’s perception of her relationship with her baby is an important area of study/clinical concern in its own right, future studies would also benefit from assessment of bonding by observation or other informants. Similarly, though our measure of postpartum bonding is widely used, replication of these findings using other measures of postnatal bonding would further strengthen confidence in results (Mathews et al. 2019). Finally, this paper focussed on the question of whether preconception elevations in depression and anxiety symptoms is a risk marker and whether they are independent of continuing symptoms. Future research examining the extent to which preconception risk is transmitted through other perinatal variables (e.g. parity, maternal age, intimate partner conflict, infant birth weight, and temperament) is warranted.

Conclusion

Early maternal-infant bonding problems commonly precede later childhood emotional and behavioural problems and mental disorders in adult life. Current programs targeting bonding difficulties generally focus on the perinatal period have shown somewhat limited efficacy (Tsivos et al. 2015). Findings from this study show that risks for later bonding problems are evident well prior to pregnancy, and support calls for preconception health care to extend beyond contraception and nutrition to include the mental health of young women (Catalao et al. 2019). Strong continuities in depression and anxiety symptoms across different phases of life point to the importance of designing mental health promotion interventions within a life-course framework that tend to the needs of women at all stages of reproductive life, from adolescence, to young adulthood, to parenthood and raising the next generation.