Introduction
Postpartum psychosis (PP) is the most severe perinatal mental health disorder and occurs in one-two per 1000 deliveries. However, women with a previous diagnosis of bipolar disorder (BD), schizoaffective disorder or PP are at greater risk of suffering from postpartum psychosis than the general population and also have a significant high risk of developing a postpartum depressive relapse (Jones & Craddock, Reference Jones and Craddock2001; Viguera et al., Reference Viguera, Tondo, Koukopoulos, Reginaldi, Lepri and Baldessarini2011; Wesseloo et al., Reference Wesseloo, Kamperman, Munk-Olsen, Pop, Kushner and Bergink2016). Considering the severity of PP, it is surprising that, to date, little research has been conducted on mother–infant interaction and infant development in this clinical population, in contrast to the considerable amount of research in perinatal depression (Bind et al., Reference Bind, Biaggi, Bairead, Du Preez, Hazelgrove, Waites and Pariante2021; Liu et al., Reference Liu, Kaaya, Chai, McCoy, Surkan, Black and Smith-Fawzi2017; Murray, Fiori-Cowley, Hooper, & Cooper, Reference Murray, Fiori-Cowley, Hooper and Cooper1996; Stein et al., Reference Stein, Pearson, Goodman, Rapa, Rahman, McCallum and Pariante2014). Difficulties in mother–infant interaction may be evident early on after delivery, providing an opportunity to intervene already in pregnancy, to promote the well-being of the dyad.
Nowadays it is still not clear whether an episode of postpartum psychosis could represent a risk for the mother interaction and the infant development. Evidence to date suggests that women with postpartum psychosis (PP) and women with postnatal depression (PND) may have similar difficulties in mother–infant interaction (Hornstein et al., Reference Hornstein, Trautmann-Villalba, Hohm, Rave, Wortmann-Fleischer and Schwarz2006; Noorlander, Bergink, & van den Berg, Reference Noorlander, Bergink and van den Berg2008). Furthermore, research has shown that impairments in mother–infant interaction following a severe postnatal episode may remain even when psychiatric symptoms resolve (Hipwell, Goossens, Melhuish, & Kumar, Reference Hipwell, Goossens, Melhuish and Kumar2000). More recently, research on maternal bonding towards the infant (Biaggi et al., Reference Biaggi, Hazelgrove, Waites, Fuste, Conroy, Howard and Dazzan2021; Gilden et al., Reference Gilden, Molenaar, Smit, Hoogendijk, Rommel, Kamperman and Bergink2020), has reported low evidence for a considerable negative impact of postpartum psychosis on maternal bonding, confirming previous reports (Hornstein et al., Reference Hornstein, Trautmann-Villalba, Hohm, Rave, Wortmann-Fleischer and Schwarz2006; Noorlander et al., Reference Noorlander, Bergink and van den Berg2008). However, there is still limited research on mother–infant interaction in this clinical population. With regards to infant development, a recent study (Chen et al., Reference Chen, Pan, Bai, Huang, Tsai, Su and Hsu2021) has reported an increased risk for Attention-Deficit/Hyperactivity Disorder (ADHD) in the offspring born to women who had experienced PP, while previous research reported no difference in the development of infants born to women with and without PP and healthy controls (McNeil, Persson-Blennow, Binett, Harty, & Karyd, Reference McNeil, Persson-Blennow, Binett, Harty and Karyd1988). In summary, only few studies have been conducted in this clinical population, with mixed findings and a variety of methodological limitations, for example in the selection of the healthy controls and in the evaluation of mother–infant interaction not blind to maternal diagnosis.
There is also limited research in women at-risk-of-PP as a group, i.e., regardless of whether they develop a postpartum episode, and therefore, more studies need to be conducted. In fact, while women with BD and affective psychosis are at increased risk of obstetric complications and their infants are at increased risk of less optimal outcomes at birth (Frayne et al., Reference Frayne, Nguyen, Allen, Hauck, Liira and Vickery2019; Judd et al., Reference Judd, Komiti, Sheehan, Newman, Castle and Everall2014; Solé et al., Reference Solé, Torres, Roca, Hernández, Roda, Sureda and Garcia-Esteve2019; Zhong et al., Reference Zhong, Gelaye, Fricchione, Avillach, Karlson and Williams2018), less is known about offspring later outcomes, occurring during infancy and childhood. Studies conducted to date have shown that, compared to controls, women at-risk-of-PP report a more negative perceived bonding towards their infants in the first 12 months postpartum (Biaggi et al., Reference Biaggi, Hazelgrove, Waites, Fuste, Conroy, Howard and Dazzan2021; Boekhorst, Beerthuizen, Hillegers, Pop, & Bergink, Reference Boekhorst, Beerthuizen, Hillegers, Pop and Bergink2021). Furthermore, there is evidence that women at-risk-of-PP display more difficulties in the interaction with their infants during the first year postpartum (e.g. lower maternal sensitivity, involvement, contact, contingent responsiveness, more tension as well as less infant positive affect, communication and dyadic coordination) and have infants with less optimal developmental outcomes (e.g. more early risk characteristics within 4 years, such as delayed walking and less optimal global functioning at 6 years) compared to controls (Anke et al., Reference Anke, Slinning, Moe, Brunborg, Siqveland and Skjelstad2019, Reference Anke, Slinning, Moe, Brunborg, Siqveland and Skjelstad2020; Henriksson & McNeil, Reference Henriksson and McNeil2004; McNeil & Kaij, Reference McNeil and Kaij1987; Naslund, Persson-Blennow, McNeil, & Kaij, Reference Naslund, Persson-Blennow, McNeil and Kaij1985).
Mother–infant interaction is an important predictor for infant cognitive and socio-emotional development (Leclère et al., Reference Leclère, Viaux, Avril, Achard, Chetouani, Missonnier and Cohen2014; Murray et al., Reference Murray, Fiori-Cowley, Hooper and Cooper1996). It potentially represents the most important mediator in the association between perinatal mental illness and child development (Stein et al., Reference Stein, Pearson, Goodman, Rapa, Rahman, McCallum and Pariante2014), and can play a crucial role and even buffer the negative effects of antenatal stress on the child (Herba, Glover, Ramchandani, & Rondon, Reference Herba, Glover, Ramchandani and Rondon2016). In addition, some of the effects of maternal mental illness on mother–infant interaction and infant development may be moderated by the sex of the child, as it has been shown in several studies in depression. In fact, boys are more susceptible to the effects of PND on cognitive and behavioural development and may also be at greater risk of suffering from a less optimal mother–infant interaction, while girls are more at risk of internalizing problems (Goodman et al., Reference Goodman, Rouse, Connell, Broth, Hall and Heyward2011; Murray et al., Reference Murray, Fiori-Cowley, Hooper and Cooper1996; Murray, Kempton, Woolgar, & Hooper, Reference Murray, Kempton, Woolgar and Hooper1993; Stein et al., Reference Stein, Pearson, Goodman, Rapa, Rahman, McCallum and Pariante2014). However, this has never been investigated in women at risk of/with PP.
The current study aims to fill these gaps in the literature and to investigate mother–infant interaction and infant development in women at-risk-of-PP (AR), both in those who do and do not develop a psychiatric relapse within 4 weeks after delivery (the AR-unwell and AR-well, respectively). We also examined whether mother–infant interaction is a predictor of infant development and the role of infant sex on these dimensions in this clinical population. We hypothesised that (1) AR-unwell women would show less optimal interactions with their infants and have infants with less optimal developmental outcomes than AR-well women; and that women at-risk-of-PP as a group would show less optimal mother–infant interactions and infant development than healthy controls; that (2) mother–infant interaction would predict infant development and would explain the association between mental illness and infant development; and finally that, (3) differences in mother–infant interaction and in infant cognitive, language and motor development and adaptive behaviour would be more evident in boys, while differences in emotional development would be more evident in girls.
Methods
Design
This study is part of the Psychiatry Research and Motherhood study – psychosis cohort (PRAM-P), a prospective longitudinal study that recruited and followed-up a group of women at-risk-of-postpartum psychosis and a group of healthy control women (and their offspring) from 25 weeks' gestation (baseline) to 12 months' post partum (Biaggi et al., Reference Biaggi, Hazelgrove, Waites, Fuste, Conroy, Howard and Dazzan2021; Hazelgrove et al., Reference Hazelgrove, Biaggi, Waites, Fuste, Osborne, Conroy and Dazzan2021). Symptom severity was assessed at 25 weeks' gestation and postnatally – at 8 weeks'(8w) and 12 months'(12 m) post partum. Mother–infant interaction was assessed at 8w and 12 m post partum, and infant development at 12 m post partum.
Sample
We included 103 women, 43 considered at-risk-of-PP (because of a diagnosis of bipolar disorder, schizoaffective disorder or previous PP) (AR), and 60 healthy controls (HC). The AR women had the following diagnoses: 33 (76.7%) bipolar disorder, 6 (14%) schizoaffective disorder, and 4 (9.3%) previous PP. Healthy controls had to be free from any current or previous psychiatric disorder and family history of PP in their first-degree relatives. AR women were identified from Perinatal Psychiatry Services and HC from King's College Hospital or GP surgeries. Inclusion criteria were: late second or third trimester of a singleton pregnancy, age ≥18 and fluency in English. Exclusion criteria were: uterine anomalies, pregnancy complications, unlikelihood to keep the baby after delivery or contraindication to MRI scan (data on MRI scan are reported separately). AR women were also excluded if their diagnosis was unclear or if they were currently too unwell to participate. The study was approved by the local ethics committee (REC: 10/H0807/14) and participants provided written informed consent for themselves and their offspring.
Measures
Socio-demographic, obstetric and infant related characteristics
We collected socio-demographic, health, obstetric, current pregnancy and infant related information at baseline, 8w and 12 m using a semi-structured interview; delivery and neonate characteristics at 6 days' post partum using the maternal discharge summary.
Clinical assessment
We used the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) (First, Reference First1996) supplemented by medical notes, the Hamilton Depression Rating Scale (HAM-D) (Hamilton, Reference Hamilton1960), the Young Mania Rating Scale (YMRS) (Young, Biggs, Ziegler, & Meyer, Reference Young, Biggs, Ziegler and Meyer1978), the Positive and Negative Syndrome Scale (PANSS) (Kay, Fiszbein, & Opler, Reference Kay, Fiszbein and Opler1987) and the Global Assessment of Functioning (GAF) (American Psychiatric Association, 1994) to assess previous and current DSM-IV Axis I disorders, symptom severity, global functioning and medication use during pregnancy and in the first 12-month postpartum period. Participants' diagnosis and symptom severity at each time point were confirmed in consensus meetings. The AR women were classified as having a psychiatric relapse (AR-unwell) if, in the first four weeks' post partum (time frame chosen according to the DSM-IV postpartum-onset specifier), they either: (a) met DSM-IV diagnostic criteria for a psychotic, manic, hypomanic, depressive or mixed episode; or (b) had a combination of DSM-IV symptoms that, whilst not meeting diagnostic criteria, impacted on their daily functioning (e.g. their ability to care for the baby or themselves) and/or were of sufficient intensity to require a change in treatment (either pharmacological or management plan). This broader definition was used to capture all affective relapse events as AR women were closely monitored by Perinatal services and most took psychotropic medication to prevent PP, or to treat symptoms as soon as they developed. A severe relapse was considered if, in the first 4 weeks' post partum, women experienced psychotic, manic, mixed symptoms, and/or psychiatric hospitalization (Wesseloo et al., Reference Wesseloo, Kamperman, Munk-Olsen, Pop, Kushner and Bergink2016). As maternal IQ is a strong predictor of offspring IQ (Eriksen et al., Reference Eriksen, Kesmodel, Underbjerg, Kilburn, Bertrand and Mortensen2013), we additionally assessed maternal Full-Scale IQ (FSIQ) using the Wechsler Adult Intelligence Scale Revised version (WAIS-R) (Wechsler & De Lemos, Reference Wechsler and De Lemos1981) at 30 weeks' gestation, to control for group differences in infant development.
Mother–infant interaction
Mother–infant interaction was assessed at 8w and 12 m with the Child-Adult Relationship Experimental Index (Infant CARE-Index), which was used to code a three-minute video-recorded observation of a free-play interaction. The CARE-index evaluates the quality of the adult-infant interaction from 0 to 15 months (Crittenden, Reference Crittenden2010), has been extensively used in research, including with women with severe postpartum mental disorders (Kenny, Conroy, Pariante, Seneviratne, & Pawlby, Reference Kenny, Conroy, Pariante, Seneviratne and Pawlby2013) as well as in a variety of different cultures (Hautamäki, Reference Hautamäki, Farnfield and Holmes2014). The CARE-Index assesses three patterns of interaction for the adult (Sensitivity, Control and Unresponsiveness) and four for the infant (Cooperativeness, Compulsiveness, Difficultness and Passivity) as well as Dyadic Synchrony, which evaluates the degree of togetherness of the dyad and the level of risk for the infant's future development if that interaction continued. All patterns as well as Dyadic Synchrony are scored on a scale 0–14. Scores obtained are clustered in different categories: At risk (0–4); Inept (5–6); Adequate (7–10), Sensitive (11–14). The maternal scales and (separately) the child scales are linearly dependent; therefore, statistical analyses should only include up to two of the maternal scales and up to three of the infant scales (Crittenden, Reference Crittenden2010). For this paper, we selected maternal Control, infant Compulsiveness, infant Difficultness and Dyadic Synchrony, based on correlation analyses (online Supporting Tables S1 and S2). All video-recorded interactions were scored by three trained raters blind to maternal mental health. The interclass correlation coefficient (ICC) (based on an absolute-agreement, 2-way mixed-effects model) was calculated using SPSS 25 and it ranged from 0.90 to 0.95 for Dyadic Synchrony at 8w and 12 m. In terms of infant age at assessment, infants born to AR were significantly older than those of HC at the first time point; at the second time point, infants of the AR-unwell were significantly older than infants of the AR-well (Table 1).
Table 1. Socio-demographics, clinical, pregnancy and infant related characteristics
AR, women at risk of PP; HC, healthy control women; AR-unwell, women at risk of PP who developed a psychiatric episode within 4 weeks after delivery; AR-well, women at risk of PP who remained well within 4 weeks after delivery; a n = 71 (HC = 31; AR = 40); b n = 61 (HC = 29; AR = 32); c n = 102 (HC = 60; AR = 42); d n = 85 (HC-45; AR = 40); e AR n = 41. Results provided in bold are statistically significant.1Including women employed but currently on maternity or sick leave. 2Antipsychotics, mood-stabilizers, antidepressants, benzodiazepines or combination. The Index of Multiple Deprivation (IMD) score is a UK government measure of relative deprivation based on the area of living (Noble, M., W.G, Dibben, C., Gan, Smith, Mclennan, D., Anttila, C., et al., 2004. The English Indices of Deprivation. Neighbourhood Renewal Unit. Report to the Office of the Deputy Prime Minister, London).
Infant development
Infant development was assessed at 12 m using the Bayley Scales of Infant and Toddler Development (Bayley-III) (Bayley, Reference Bayley2006a), which has been extensively used for both clinical and research purposes (Weiss, Oakland, & Aylward, Reference Weiss, Oakland and Aylward2010). The Bayley can be used from 0 to 42 months and consists of a series of standardized play tasks assessing infant Cognitive, Language and Motor development. A questionnaire completed by the caregiver also assesses infant Socio-Emotional development and Adaptive behaviour. All developmental domains are independent and have a separate score (adjusted for infant's age). Composite scores range from 40 to 160 (M = 100, s.d. = 15) (Bayley, Reference Bayley2006b) and scores of <85, <70 and <55 are cut-off points indicative of a mild, moderate or severe developmental delay. However, the Bayley-III may underestimate developmental delays and, therefore, cut-off threshold values may need to be higher (Bos, Reference Bos2013; Johnson, Moore, & Marlow, Reference Johnson, Moore and Marlow2014). Infants of the AR-unwell women were significantly older than those of the AR-well (Table 1).
Statistical analyses
Statistical analyses were performed in IBM SPSS Statistics Version 25 (IBM Ltd, UK). Data were first examined for normality of distribution and homoscedasticity. In univariate analyses we first compared AR to HC women, then AR-unwell to AR-well. Continuous data were analysed with Independent samples t test or Mann-Whitney U, as appropriate. Categorical data were analysed using Pearson's Chi-square test for independence (χ2). In multivariate analyses we winsorized data that did not meet assumptions for parametric analyses or used bootstrapping (with 1000 samples). Specifically, we conducted factorial ANOVA and ANCOVA to control for the effects of potential confounders, identified in correlation analyses, using Pearson's (r) or Spearman's (r s), correlation coefficients, as appropriate. When more than two potential confounders were identified, we retained two if they were highly associated (r ⩾ 0.50), to ensure adequate power. We also conducted a mixed ANOVA to investigate changes in mother–infant interaction (from 8w to 12 m). Furthermore, we conducted an exploratory analysis, using planned contrasts to test the hypothesis that boys of the AR women would have lower scores in mother–infant interaction and in all infant developmental dimensions than all other groups (AR-females, HC-males, HC- females), apart from socio-emotional development where we hypothesized that girls of AR women would have the lowest scores of all groups. Cohen's d was used for effect sizes in univariate comparisons, while partial eta squared (η p2) in ANOVAs and ANCOVAs. Finally, we conducted a mediation analysis to examine the direct effect of maternal Group (the exposure, X) on infant development at 12 m (the outcome, Y) and the indirect effect of X on Y via mother–infant interaction (the mediator, M), controlling for the effect of potential confounders. We tested this in two separate models, one with mother–infant interaction at 8w and one at 12 m, using the ‘PROCESS for SPSS and SAS’ macro (version 3.1). The number of bootstrapped samples was set to 5000 as recommended by Hayes (Hayes, Reference Hayes2013).
Results
Socio-demographic and clinical characteristics
Compared to HC, significantly more AR women were unemployed, had an unplanned pregnancy, did not breastfeed their infants and had lower FSIQ (Table 1). AR women had also significantly higher PANSS, HAM-D, YMRS scores and lower GAF scores at baseline, 8w and 12 m post partum than HC (Table 2). Of the 43 AR women, 18 (41.9%) developed a psychiatric relapse within four weeks of delivery (AR-unwell): 8 (44.4%) had symptoms of depression or depression and anxiety, 6 (33.3%) had manic or hypomanic symptoms, 3 (16.7%) had psychotic symptoms and 1 (5.6%) had mixed symptoms. Of these, 10 (55.6%) had symptoms that met DSM-IV diagnostic criteria and 5 (27.8%) had a severe relapse. Twenty-five of the AR women (58.1%) remained symptom-free within four weeks of delivery (AR-well). The AR-unwell women were more likely to be unemployed, to have an unplanned pregnancy, to have their child born at lower gestational age and to have lower FSIQ than the AR-well women. The AR-unwell women were also more likely to have experienced a psychiatric relapse in pregnancy and had higher PANSS and HAM-D and lower GAF scores at baseline, while there was no significant difference in lifetime diagnosis, previous PP, age at illness onset, duration of illness and use of psychotropic medication at baseline, 8w and 12 m (Table 1). However, in the 12 months' post partum, the AR-well also progressively developed psychiatric symptoms (46.5% by 8w and 74.4% by 12 m) and by 12 m, there was no significant difference between the two AR groups in any of the clinical symptoms (Tables 1 and 2).
Table 2. Symptom severity and global functioning in the AR and HC groups and in the AR-unwell and AR-well groups
AR, women at risk of PP; HC, healthy control women; AR-unwell, women at risk of PP who developed a psychiatric episode within 4 weeks after delivery; AR-well, women at risk of PP who remained well within 4 weeks after delivery; (AR) – PANSS missing: baseline = 2, PN1 = 5, PN2 = 6; HAM-D missing: baseline = 2, PN1 = 3, PN2 = 6; YMRS missing: baseline = 3, PN1 = 6, PN2 = 7; GAF missing: baseline = 1, PN1 = 2, PN2 = 7; (HC) – PANSS missing: PN1 = 3, PN2 = 4; HAM-D missing: PN1 = 1, PN2 = 5; YMRS missing: PN1 = 1, PN2 = 4; GAF missing: PN1 = 1, PN2 = 5. Results provided in bold are statistically significant.
AR women have less synchronous interactions with their infants than HC
As a group, the AR women had significantly less synchronous interactions with their infants than HC at both 8w and 12 m post partum (Table 3). The difference in Dyadic Synchrony at 8w remained significant even after controlling for maternal employment (F (1,90) = 3.95, h p2 = 0.04, p = 0.050) (online Supporting Table S4). None of the maternal and infant- related factors were associated with Dyadic Synchrony at 12 m, therefore these were not included as potential confounders (online Supporting Table S5). There was also no association between medication assumption and Dyadic Synchrony at either 8w or 12 m (rs = −0.31, p = 0.06; rs = −0.29; p = 0.08). The AR group had less optimal scores than HC also in maternal Control, infant Compulsiveness and Difficultness, although these were not statistically significant. There were no significant differences in mother–infant interaction between the AR-unwell and the AR-well at either 8w or 12 m (Table 3). Interestingly, Dyadic Synchrony improved significantly from 8w to 12 m in both groups of AR and HC women (F (1,81) = 29.31, p < 0.001, h p2 = 0.27), with no significant maternal Group × Time interaction (F (1,81) = 0.00, p = 0.976, h p2 = 0.00).
Table 3. Mother–infant interaction at 8 weeks' and 12 months' post partum and infant development at 12 months' post partum
AR, women at risk of PP; HC, healthy control women; AR-unwell, women at risk of PP who developed a psychiatric episode within 4 weeks after delivery; AR-well, women at risk of PP who remained well within 4 weeks after delivery; CARE-Index 8w – AR N = 38 (AR-unwell = 17; AR-well = 21), HC – N = 55; CARE-Index 12 m – AR N = 37 (AR-unwell = 15; AR-well = 22), HC – N = 51; Bayley (Cognitive, Language and Motor)- AR- N = 40 (AR-unwell = 15; AR-well = 25)- HC = 56; Bayley (Socio-Emotional and Adaptive behaviour) – AR – N = 35 (Socio-Emotional: 23 AR-well and 12 AR-unwell; Adaptive Behaviour: 22 AR-well and 13 AR-unwell) – HC – N = 53. Results provided in bold are statistically significant.
Infants born to AR women have less optimal developmental outcomes than HC
Compared to infants born to HC women, infants born to AR women had significantly lower scores in Cognitive, Language, Motor and Socio-Emotional development at 12 m, while there was no significant difference between the groups in Adaptive behaviour (Table 3). After controlling for maternal FSIQ, the difference in Cognitive development remained significant (F (1,77) = 4.18, p = 0.044, h p2 = 0.05). After controlling for maternal employment and FSIQ, the difference in Language development remained significant (F (1,76) = 6.94, p = 0.010, h p2 = 0.08), while the difference in Socio-Emotional development became non-significant (F (1,68) = 0.46, p = 0.501, h p2 = 0.01). As none of the maternal and infant- related factors were associated with Motor development, these were not included as potential confounders (online Supporting Table S3). There was no significant difference in any developmental domain nor in infant behaviour between infants born to AR-unwell and infants born to AR-well women (Table 3).
Mother–infant interaction is associated with infant development
Dyadic Synchrony at 8w was positively associated with infant Cognitive (rs = 0.25, p = 0.019) and Language (rs = 0.32, p = 0.003) development at 12 m, and maternal Control and infant Compulsiveness at 8w were negatively associated with infant Language development at 12 m (respectively, rs = −0.23, p = 0.030; rs = −0.37, p < 0.001). Similarly, Dyadic Synchrony at 12 m was positively associated with infant Cognitive, Language and Socio-Emotional development (respectively, r = 0.29, p = 0.006; r = 0.49, p < 0.001; rs = 0.26, p = 0.019), and infant Compulsiveness and Difficultness at 12 m were negatively associated with infant Language development (respectively, rs = −0.32, p = 0.002; r = −0.22, p = 0.043).
Mother–infant interaction during the first postnatal year predicts infant development at 12 months
We then tested a mediation relationship: whether mother–infant interaction (Dyadic Synchrony), which differed between AR and HC women, predicted and mediated the relationship between maternal Group and infant development after controlling for potential confounders. As findings of the models including Dyadic Synchrony at 8w and 12 m were identical, we report here only those at 8w (Figs 1 and 2), and report those at 12 m in the online Supporting materials.
Figure 1. Mediation model – Infant Language development. Figure of the hypothesized mediation model with maternal Group (AR v. HC) as the predictor variable (X), infant Language development at 12 months (Bayley-III) as the outcome variable (Y) and mother–infant interaction (Dyadic Synchrony- CARE-Index at 8 weeks) as the mediating variable (M). A, b, c and c': path coefficients representing unstandardized regression weights and standard errors. *p < 0.05, **p < 0.01. Maternal FSIQ and employment were inserted in a second step analysis as potential confounders.
Figure 2. Mediation model – Infant Cognitive development. Figure of the hypothesized mediation model with maternal Group (AR v. HC) as the predictor variable (X), infant Cognitive development at 12 months (Bayley-III) as the outcome variable (Y) and mother–infant interaction (Dyadic Synchrony- CARE-Index at 8 weeks) as the mediating variable (M). *p < 0.05. A, b, c and c': path coefficients representing unstandardized regression weights and standard errors. Maternal FSIQ was inserted in a second step analysis as potential confounder.
A more synchronous interaction at 8w predicted better infant Language development at 12 m (b = 1.22, t = 2.89, p = 0.005). There was also evidence of a partial mediation, as there was both a significant direct effect of maternal Group on infant Language development (b = −4.98, 95% CI [−9.42 to −0.54]) as well as an indirect effect via Dyadic Synchrony (b = −1.67, 95% Boot CI [−3.74 to −0.16]). The corresponding partially standardized indirect effect size estimate was −0.16 (95% Boot CI [−0.34 to −0.02]) (Fig. 1). When maternal FSIQ and employment were included in the model, Dyadic Synchrony remained a significant predictor of infant Language development (b = 1.15, t = 2.47, p = 0.016), but was no longer a partial mediator (b = −1.17, 95% Boot CI [−3.41 to 0.28]).
With regards to infant Cognitive development, a more synchronous interaction at 8w predicted better infant development at 12 m (b = 1.11, t = 2.04, p = 0.045), but was not a significant mediator in the relationship between maternal Group and infant Cognitive development (b = −1.52, 95% Boot CI [−3.77 to 0.05]) (Fig. 2). When controlling for maternal FSIQ, Dyadic Synchrony remained a significant predictor of infant Cognitive development (b = 1.71, t = 2.05, p = 0.044).
Boys of AR women present the lowest cognitive, language and motor developmental scores and have the least synchronous interactions with their mothers at 12 months, while girls of AR women have the lowest socio-emotional development scores
The planned comparisons of the effect of maternal case Group (AR) and infant male Sex v. all other groups on infant Cognitive, Language and Motor development were all statistically significant, indicating that male infants of the AR women had lower scores than all other groups in all of these dimensions (Table 4). The planned comparison of the effect of maternal case Group (AR) and infant female Sex on infant Socio-Emotional development was statistically significant, indicating that the female infants of the AR women had lower scores than all other groups. The planned comparison of the effect of maternal case Group (AR) and infant male Sex v. all other groups on Dyadic Synchrony at 12 m was statistically significant, while it only reached a trend for statistical significance at 8w, suggesting that women of the AR group with male infants had the least synchronous interactions compared to the other groups (Table 4).
Table 4. Mother–infant interaction and infant development in boys and girls
AR, women at risk of PP; HC, healthy control women; AR-unwell, women at risk of PP who developed a psychiatric episode within 4 weeks after delivery; AR-well, women at risk of PP who remained well within 4 weeks after delivery; Bayley (Cognitive, Language and Motor): 32 HC males, 24 HC females, 21 AR males, 19 AR females; Bayley Socio-Emotional: 30 HC males, 23 HC females, 18 AR males, 17 AR females; CARE-Index 8 weeks: 30 HC males, 25 HC females, 19 AR males, 19 AR females; CARE-Index 12 months: 30 HC males, 21 HC females, 20 AR males, 17 AR females. Results provided in bold are statistically significant. 1Planned comparisons of AR-male infants v. all other groups (AR-females, HC males, HC females) on Cognitive, Language, Motor development, Dyadic Synchrony at 8 weeks and Dyadic Synchrony at 12 months. Planned comparison of AR-female infants v. all other groups (AR-males, HC males, HC females) on Socio-Emotional development.
Discussion
In this prospective longitudinal study we investigated mother–infant interaction and infant development in women at risk of postpartum psychosis and in those at risk who suffered a postpartum relapse within four weeks after delivery.
We have found that women at-risk-of-PP, regardless of whether or not they develop a postpartum relapse, have less synchronous interactions with their infants at both 8 weeks' and 12 months' post partum and have infants with less optimal cognitive, language, motor and socio-emotional development at 12 months, compared to healthy controls. Our data also provide no evidence for a difference in mother–infant interaction or in infant development between women at-risk-of-PP who do and do not develop a psychiatric relapse in the early postpartum period. Our results are consistent with a previous research, which found no differences in developmental outcomes of infants born to women at-risk-of-PP who developed or did not develop the illness after delivery (McNeil et al., Reference McNeil, Persson-Blennow, Binett, Harty and Karyd1988) and with studies reporting differences in mother–infant interaction and infant development in women with a history of affective psychosis and bipolar disorder in comparison to healthy controls (Anke et al., Reference Anke, Slinning, Moe, Brunborg, Siqveland and Skjelstad2019, Reference Anke, Slinning, Moe, Brunborg, Siqveland and Skjelstad2020; Henriksson & McNeil, Reference Henriksson and McNeil2004; McNeil & Kaij, Reference McNeil and Kaij1987).
These findings suggest that for a child's development, a postpartum episode in women at-risk-of-PP may not be more important than the mother's previous diagnosis of an affective psychosis. Therefore, these results highlight the need to target all women at-risk-of-PP when considering potential preventative interventions. It is indeed possible that factors associated with the woman's lifetime diagnosis, such as a history of childhood maltreatment and of adverse life events, may have a negative effect on child's development. These factors could indeed affect child development in different ways, for example through alterations in biological factors, such as cortisol, and this should be investigated in future studies.
Interestingly, our results differ from those reported in depression, which have shown that perinatal depression can represent a risk factor for both mother–infant interaction and child development (Bind et al., Reference Bind, Biaggi, Bairead, Du Preez, Hazelgrove, Waites and Pariante2021; Murray et al., Reference Murray, Fiori-Cowley, Hooper and Cooper1996; Stein et al., Reference Stein, Pearson, Goodman, Rapa, Rahman, McCallum and Pariante2014), particularly if depression is severe and persistent (Netsi et al., Reference Netsi, Pearson, Murray, Cooper, Craske and Stein2018). This may be explained by the fact that symptoms of PP (particularly psychotic symptoms) usually resolve earlier than those of perinatal depression, resulting in the child being exposed to maternal symptoms for a shorter period of time (Murray & Hipwell, Reference Murray and Hipwell1995), given that a longer symptom duration has been associated with more adverse child outcomes (Murray, Halligan, & Cooper, Reference Murray, Halligan, Cooper, Bremner and Wachs2010). It is also possible that only severe episodes or specific symptoms affect the child. Infants can, in fact, be exposed to various presentations, even in case of frank PP, as they can experience an overactive, agitated, irritable or depressed mother, or a mother with rapid mood swings or a variety of psychotic symptoms (Hipwell & Kumar, Reference Hipwell, Kumar, Murray and Cooper1997). We could not investigate this in the current study due to the limited sample size. Another possible explanation for the lack of difference between the two AR groups is the potential impact of psychiatric symptoms developed after four weeks' post partum. In fact, by eight weeks' and 12 months' after delivery many of the at-risk women who had remained well in the first four weeks, had also developed psychiatric symptoms, and these may have impacted both mother–infant interaction and infant development. This would be consistent with our previous evidence from this sample where we similarly found that women at-risk-of-PP reported a less optimal quality of perceived bonding towards the infant in the first year postpartum than healthy controls, while no difference was observed between the AR-well and the AR-unwell. This was explained by the fact that a considerable number of at-risk women developed psychiatric symptoms after four weeks, and these affected their emotional bonding towards their infants (Biaggi et al., Reference Biaggi, Hazelgrove, Waites, Fuste, Conroy, Howard and Dazzan2021). However, considering the relatively small number of women in the AR-unwell group, the finding of no difference between the two AR groups in mother–infant interaction and infant development will need to be confirmed in future studies with larger samples.
It should also be noted that, although the developmental scores of infants of women at-risk-of-PP (particularly those of boys) were significantly lower than controls, most were not indicative of significant delays. However, as mentioned earlier, the Bayley-III may underestimate developmental delays, and researchers and clinicians have warned that its cut-off threshold values may need to be increased (Bos, Reference Bos2013; Johnson et al., Reference Johnson, Moore and Marlow2014). Therefore, it is important to consider that, when using the new suggested cut-off, some of the lower developmental scores that we identified become indicative of significant delays. To this end, a new version of the Bayley (Bayley-IV) is currently undergoing further validation and standardization (Aylward & Zhu, Reference Aylward and Zhu2019).
In terms of mother–infant interaction, we should highlight that in both at-risk and control women, interactions became more synchronous over time, in line with the normative improvement that occurs from the neonatal period to 12 months, as mothers and infants get to know each other and improve their level of attunement to one another (Fuchs, Mohler, Resch, & Kaess, Reference Fuchs, Mohler, Resch and Kaess2015), as previously shown also in depressed women (Bind et al., Reference Bind, Biaggi, Bairead, Du Preez, Hazelgrove, Waites and Pariante2021). Nevertheless, dyads of the at-risk group continued to remain at higher developmental risk for the child at 12 months. In fact, they were classified as ‘at Risk’ for child development at 8 weeks and as ‘Inept’ at 12 months, indicating that the interaction was still not adequate, and an intervention might be advisable.
Therefore, these findings are clinically meaningful, also because a high Dyadic Synchrony is associated with more optimal cognitive, emotional and behavioural child development and with a secure attachment (Harrist & Waugh, Reference Harrist and Waugh2002; Leclère et al., Reference Leclère, Viaux, Avril, Achard, Chetouani, Missonnier and Cohen2014). In particular, we found that, in this clinical population, a better mother–infant interaction during the first year postpartum (higher dyadic synchrony, less maternal control, infant compulsiveness and difficultness) is associated with better infant developmental outcomes at 12 months. Specifically, the level of synchrony in the dyad as early as 8 weeks' post partum predicts both infant cognitive and language development at 12 months. These findings clearly resemble those in PND where disturbances in the mother–infant interaction at two months predicted less optimal infant cognitive development at 18 months (Murray et al., Reference Murray, Fiori-Cowley, Hooper and Cooper1996).
However, our data show that Dyadic Synchrony was only a partial mediator in the relationship between maternal risk for PP and infant language (but not cognitive) development, contrary to what observed in PND (Milgrom, Westley, & Gemmill, Reference Milgrom, Westley and Gemmill2004; Murray et al., Reference Murray, Kempton, Woolgar and Hooper1993). These results suggest that mother–infant interaction is an important contributing factor in the association between maternal Group and infant Language development, although it does not completely explain the difference observed between the AR and HC groups in this dimension. It is possible that mother–infant interaction plays more of a role in mediating the relationship between PP (rather than maternal risk for PP) and infant development, or that other or more specific characteristics of the mother–infant interaction better explain differences in infant development. To this end, a previous study also showed that, although lower maternal sensitivity and higher remoteness and less infant engagement at 2 months' post partum significantly predicted worse infant cognitive performance at 18 months, these did not explain the lower cognitive scores of the boys of women with PND (Murray et al., Reference Murray, Fiori-Cowley, Hooper and Cooper1996). On the contrary, in another study, the same authors found that speech of depressed women at 2 months post partum, which was less focused on the infant experience, largely accounted for the negative effects of PND on boys' cognitive development at 18 months (Murray et al., Reference Murray, Kempton, Woolgar and Hooper1993).
Finally, we report here that boys of women at-risk-of-PP show the lowest scores in cognitive, language and motor development, while girls of these women show the lowest scores in socio-emotional development. Although this was an exploratory analysis, given the relatively small numbers, the results are consistent with findings in offspring of women with PND (Goodman et al., Reference Goodman, Rouse, Connell, Broth, Hall and Heyward2011; Milgrom et al., Reference Milgrom, Westley and Gemmill2004; Stein et al., Reference Stein, Pearson, Goodman, Rapa, Rahman, McCallum and Pariante2014). In fact, they suggest that even in women at risk of PP, boys are more susceptible to maternal mental health in their cognitive development while girls are more at risk in their emotional development. Furthermore, we found that, at 12 months, and at trend level also at 8 weeks, women at-risk-of-PP with boys had the least synchronous interactions compared to all other groups. Previous studies on infant sex and mother–infant interaction in depression reported mixed findings, with some research reporting that mothers had less optimal interactions with boys than girls (Murray et al., Reference Murray, Kempton, Woolgar and Hooper1993) and others not finding any sex difference (Sidor, Kunz, Schweyer, Eickhorst, & Cierpka, Reference Sidor, Kunz, Schweyer, Eickhorst and Cierpka2011). Results of this study in women at-risk-of-PP confirm those of a study previously conducted in women with a history, or recent experience of a severe mental illness, which found that mothers were more sensitive towards girls than boys (Rigby, Conroy, Miele-Norton, Pawlby, & Happé, Reference Rigby, Conroy, Miele-Norton, Pawlby and Happé2016).
Strengths and limitations
This study has numerous strengths, including the longitudinal design and the evaluation of mother–infant interaction and infant development in women at-risk-of, or with PP during the first year postpartum, a time of increased challenges for mothers (and fathers), even more for those experiencing a severe mental illness. Furthermore, the participant selection of cases and controls followed strict inclusion/exclusion criteria, clinical notes were obtained and consensus meetings were held to confirm diagnoses and symptoms. Mother–infant interaction was scored blind to maternal mental health.
There are also some important limitations to consider. We were unable to study separately the women who developed frank PP due to the small number of women with these episodes. Women at-risk-of-PP were recruited from specialist perinatal mental health services and were closely monitored throughout the perinatal period, which probably contributed to a relatively low rate of frank PP. Furthermore, we used a broader definition of PP to capture relapse of affective symptoms in the early postpartum period; nevertheless, we are confident that we are not evaluating PND, which most of the time has a later onset (Biaggi et al., Reference Biaggi, Hazelgrove, Waites, Fuste, Conroy, Howard and Dazzan2021; Doucet, Dennis, Letourneau, & Blackmore, Reference Doucet, Dennis, Letourneau and Blackmore2009; Hazelgrove et al., Reference Hazelgrove, Biaggi, Waites, Fuste, Osborne, Conroy and Dazzan2021). However, considering the relatively small number of women in our groups, these results will need to be replicated in future studies.
Furthermore, we did not evaluate father-infant interaction, which may also play an important role in infant development (Sethna et al., Reference Sethna, Perry, Domoney, Iles, Psychogiou, Rowbotham and Ramchandani2017), together with other factors, including the availability of social support. The role of these factors on infant development in this clinical population will need to be investigated in future studies.
In summary, this study provides novel findings on mother–infant interaction and infant development in the offspring of women at-risk-of-PP with and without a postpartum relapse. Although these findings will need to be confirmed in larger studies, they provide important information and suggest that all women at-risk-of-PP may benefit from support in interacting with their infants and promoting their development during the perinatal period, as this time offers a unique opportunity for interventions to promote the long-term well-being of the dyad.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291723002568
Acknowledgements
We thank the Perinatal Research Team and most of all the women who shared their early lives as mothers with us.
Author contributions
All authors have contributed to the manuscript and have approved the final article.
Funding statement
This work was supported by the Medical Research Foundation [grant number C0439] and the National Institute for Health Research (NIHR) Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London. CMP is a NIHR Senior Investigator. The funding sources played no role in the study design, collection, analysis or interpretation of findings, in writing the report or in the decision to submit the article for publication.
Competing interests
PD has received speaker's fees from Lundbeck and Janssen. CMP has received research funding from Johnson & Johnson for research on depression and inflammation and is funded by the Wellcome Trust strategy award to the Neuroimmunology of Mood Disorders and Alzheimer's Disease (NIMA) Consortium (104025), which is also funded by Janssen, GlaxoSmithKline, Lundbeck and Pfizer. LMH has received funding from the NIHR and the Nuffield Foundation for Research Programmes on maternal mental disorders. MAM has received research funding from Takeda Pharmaceuticals, Lundbeck, Johnson & Johnson and support in kind from AstraZeneca and has acted as a consultant for Lundbeck and Takeda. However, this paper is independent from this funding; there are no further declarations of interest.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Open access
