Promoting mother‐infant relationships and underlying neural correlates: Results from a randomized controlled trial of a home‐visiting program for adolescent mothers in Brazil

Abstract Poverty and teenage pregnancy are common in low‐and‐middle‐income countries and can impede the development of healthy parent‐child relationships. This study aimed to test whether a home‐visiting intervention could improve early attachment relationships between adolescent mothers and their infants living in poverty in Brazil. Analyses were conducted on secondary outcomes from a randomized controlled trial (NCT0280718) testing the efficacy of a home‐visiting program, Primeiros Laços, on adolescent mothers’ health and parenting skills and their infants’ development. Pregnant youth were randomized to intervention (n = 40) or care‐as‐usual (CAU, n = 40) from the first trimester of pregnancy until infants were aged 24 months. Mother‐infant attachment was coded during a mother‐infant interaction when the infants were aged 12 months. Electrophysiological correlates of social processing (mean amplitude of the Nc component) were measured while infants viewed facial images of the mother and a stranger at age 6 months. Infants in the intervention group were more securely attached and more involved with their mothers than those receiving CAU at 12 months. Smaller Nc amplitudes to the mother's face at 6 months were associated with better social behavior at 12 months. Our findings indicate that the Primeiros Laços Program is effective in enhancing the development of mother‐infant attachment.

other risk-factors for impaired child development, including adolescent parenthood (Lund et al., 2018). Adolescent parenthood remains a common social problem, particularly in LMICs (World Health Organisation, 2020). Brazil has one of the highest rates of adolescent pregnancy: one in five babies is born to a teenage mother, and three in five of those mothers do not work or study (UNFPA, 2017). Parenthood disrupts typical emotional and cognitive development in adolescence, which can impair the emergence of reflective and responsive parenting skills necessary to foster secure parent-infant relationships (Flaherty & Sadler, 2011). Being born to an adolescent mother is associated with lower cognitive ability and elevated behavioral problems in childhood (Lee et al., 2020;Morinis et al., 2013), which persist into adolescence and adulthood resulting in increased systemic and mental health disorders, welfare dependency, lower income, and higher criminality (Jaffee et al., 2001;Shaw et al., 2006). These outcomes may occur because of the lower quality of maternal care that adolescent mothers provide, including lower tendency to touch, call, smile at, and accept their child's emotions (Firk et al., 2018). Adolescent mothers are less sensitive to their child's needs, provide less verbal stimulation, and play less with their babies (Crugnola et al., 2014). These responsive behaviors from the mother are essential because they play a causal role in shaping attachment (Zeegers et al., 2017). Secure parent-child attachments help children to regulate emotions, develop autonomy and resilience, improve their cognitive development, and engage in positive and cooperative relationships (Groh et al., 2017). However, insecure and disorganized attachment places children at risk of problematic behaviors and psychopathology (Ainsworth et al., 1978;Groh et al., 2017) and is a significant contributor to public costs in adolescents with antisocial behavior (Bachmann et al., 2019). The estimated rate of secure attachment in the general population is 59%-67% (Bakerman-Kranenburg & van IJzendoorn, 2009;Lewis-Morrarty et al., 2014), while rates of secure attachment in infants of adolescent mothers are ∼33% (Moran et al., 2008).
Home-visiting programs (HPVs) designed to meet families' needs with young children lead to positive effects on maternal and child health, child development, responsive parenting, secure attachment, and maltreatment rates (Filene et al., 2016). The most well-established HVPs (e.g., Early Head Start, Nurse-Family Partnership) support infantparent relationships and are grounded in attachment theory, given the crucial role of early parent-infant attachment in development (Mountain et al, 2017). HVPs have been used with adolescent mothers in high-income countries and showed improved parenting knowledge and behaviors, including increased maternal sensitivity and stimulation of the baby as well as improved socio-emotional development and reduced externalizing behavior problems in early-childhood in offspring (Barlow et al., 2015;Guttentag et al., 2014). HVPs have also shown positive effects on maternal parenting behaviors and offspring development in vulnerable adult mothers in LMICs, including India (Andrew et al., 2020), South Africa (Christodoulou et al., 2019), Rwanda (Betancourt et al., 2020) and Brazil (Gonçalves et al., 2019).
Yet, despite the potential to improve maternal parenting behavior and infant development, to our knowledge no published work has assessed the efficacy of HVPs developed specifically for adolescent

Research highlights
• Teenage pregnancy is common in low-and-middle-income countries (LMICs) • Adolescent mothers often have less secure relationships with their babies • A home-visiting intervention improved attachment relationships in this population • Better attachment was associated with enhanced neural processing of social stimuli mothers in LMICs. We therefore conducted a randomized controlled trial (RCT) of an HVP designed for adolescent mothers and their infants living in poverty in Brazil (Primeiros Laços). We created the program based on the Nurse-Family Partnership and Minding the Baby, two programs with a solid conceptual basis and empirically-supported efficacy (Olds, 2006;Slade et al., 2005), but adapted for the socio-cultural characteristics and health system organization in Brazil Pinheiro et al., 2018). The primary aim of the trial was to investigate the efficacy of the HVP on infants' cognitive, motor and language development, and also the effects of the intervention on important secondary outcomes including maternal well-being, parenting behaviors, mother-infant attachment relationships and neurobiological mechanisms involved in the intervention (clinicaltrials.gov NCT02807818). In the current analysis, we focused on two of these secondary outcomes: (1) mother-infant attachment relationships, and (2) neural correlates of infants' social development, as well as the relationship between these outcomes.
Mother-infant attachment relationships were measured during a mother-infant interaction episode when the infants were aged 12 months. Neural correlates of social development were measured with a component of the event-related potential (ERP), the Nc, derived from electroencephalography (EEG) recorded while the infants viewed images of their mother's face and a stranger's face when they were aged 6 months. The Nc reflects attentional and memory processes and is sensitive to social information; in the first year of life, the Nc is larger (i.e. has greater amplitude) when viewing the mother than a stranger, reflecting enhanced attentional allocation to the socially-salient mother (de Haan & Nelson, 1997). This effect reverses from the second year of life, coinciding with the emergence of more exploratory social behaviors with individuals other than the mother (Carver et al., 2003). The Nc to mother/stranger faces appears to be particularly sensitive to children's social behavior and relationships with their mothers. One study reported that, compared to securely-attached preschoolers, insecurely-attached preschoolers showed larger Nc amplitudes when viewing their mother's face, suggestive of greater attentional allocation to the mother in children with less stable maternal relationships (Kungl et al., 2017). Similarly, another study found that infants who showed more social engagement behaviors with the mother following a brief separation showed smaller Nc amplitudes to the mother's face than a stranger's face, interpreted as reflecting less attentional allocation to the mother in infants who held a stable representation of the mother as being someone to whom the infants could direct their social-engagement behaviors while exploring their environment (Swingler et al., 2007).
We hypothesized that (1) mother-infant dyads who received the intervention would show stronger attachment relationships at 12 months than those receiving care-as-usual, (2) consistent with previous work investigating the Nc in the first year of life (de Haan & Nelson, 1997), at age 6 months the infants in the intervention group would show more differentiated neural responses to mother and stranger face stimuli, as indicated by larger Nc to the mother's than stranger's face, compared to infants receiving care-as-usual, and (3) more differentiated Nc at 6 months would be associated with better socialinteraction behavior with mothers at 12 months.

Study design and participants
The current report presents analysis of secondary outcome measures collected as part of an RCT (clinicaltrials.gov NCT02807818) con-  Figure 1). Randomization was stratified by primary healthcare unit type (with vs. without family health support teams) and grandmother's years of schooling.
Group allocation ratio was 1:

Intervention
Primeiros Laços is an HVP delivered by trained nurses tailored to firsttime pregnant adolescents and their infants, running from the first 16 weeks of pregnancy until the child is aged 24 months. Nurses were specialized in maternal or mental health and were supervised weekly by senior nurses and psychologists. Primeiros Laços is based on three theoretical frameworks (Attachment theory, Bowlby et al. 1979; Self-Efficacy theory, Bandura, 1977;Bioecological Development theory, Bronfenbrenner, 1998) and was structured in five axes (health care, health environment, parenting and attachment, social and family network, life project) (see eTable for further details). Our team developed the program based on existing HVPs (Olds, 2006;Slade et al., 2005). Each mother-infant dyad's care plan was carefully designed to strengthen maternal competences for warm and responsive care. A key factor was establishing positive relationships between home-visitors and the family. During the visits, nurses helped parents develop child-centered interactions, improve their bond with their infant, reflect on their attachment history and the parenting they received, consider their child as individuals with their own needs, feelings, and thoughts, and improve their parenting skills by modeling. The Visitors encouraged attuned parenting and stimulated sensitive behaviors, such as being attentive to the child's communicative signals or following their lead.
Parents were also given support to think reflectively. The frequency of visits was weekly (first/last month of pregnancy/puerperium), biweekly (gestation/2-20 months of child's age), and monthly (21-24 months of child's age). Mothers were expected to receive 40-42 visits by the time their infants were aged 12 months and 60-62 visits by the time their infants were 24 months of age. Participants who received the intervention also had access to public health services (care-as-usual).

Care-as-usual (CAU)
Adolescent mothers in the CAU group received standard prenatal and postnatal care-as-usual. In Brazil, prenatal and postpartum care is Prenatal care is organized to meet the bio-psycho-social demands of women and their families. Upon completing 40 weeks of gestation, the woman is referred to the regional hospital where her delivery will take place. Puerperal assistance consists of home care performed from the 15th day of the baby's birth and then continues with medical consultations in primary health care units (Schirmer, 2000).

Mother-infant attachment relationships at infant age 12 months
A video-recorded 14-min mother-infant interaction was conducted when the infants were aged 12 months. The interaction was semi-structured and included five situations (playing, teaching, helping, setting boundaries, dealing with stress and separation) to elicit attachment behaviors (see eAppendix 2). Videos were coded using the Emotional Availability Scale (EAS, Biringen, 2008), a widely used instrument to assess parent-child interactions that shows good reliability and validity in varied populations (Biringen et al., 2014;Ziv et al., 2000). The EAS quantifies emotional attachment between the parent and child in six dimensions: (1) Caregiver Sensitivity (affective presence/appropriate responsiveness to child), (2) Caregiver Structuring (use of proper strategies to guide child/set preventative limits), (3) Caregiver Non-intrusiveness (ability to withhold behaviors that interfere with child's interests and support age-appropriate autonomy), (4) Caregiver Non-hostility (absence of subtle/overt negative emotions), (5)

Neural correlates of infant social development at infant age 6 months
At age 6 months, infants completed a passive viewing task during EEG recording while seated on their mother's lap ∼65 cm in front of a 45 cm × 45 cm screen in a dimly lit room. The task presented static images (11 × 18 cm) of the mother's or a stranger's (another study mother's) face. Task trials began with a 1200 ms pause (blank screen) followed by a central fixation cross (black on a white background) presented until the infant was looking at the screen (timing controlled by the experimenter) and then the mother or stranger face image for 500 ms and finally an inter-trial-interval of 1200 ms (blank screen). Seventy-five images of the mother and 75 images of the stranger were presented in randomized order (see de Haan & Nelson, 1997).
One photo of the mother and one photo of the stranger was used for each baby and the two photos were repeated multiple times. Members of the research team conducting the assessments took the photos of the mothers on the day the babies took part in the EEG assessment. All mothers stood in the same location in the laboratory against a blank white wall and wore a cream-colored cloak that covered their clothes while their photograph was taken. Mothers were asked to relax their faces and maintain a neutral expression, while looking directly at the camera. The researchers then cropped the photos to be the same size and number of pixels and to ensure only the head and shoul-ders were included in the frame before the photos were added to the experimental programming software that was used to present the task. For each baby, the mother's photo was matched with a photo from one of the other mothers in the study who had a similar appearance in terms of hairstyle, facial features, skin coloring, and use of glasses.
EEG data were recorded using a 128-channel Geodesic Sensor Net  orienting to objects, higher soothability). These temperament factors were included in analysis as a covariate since child temperament is proposed to influence mother-infant attachment (Bowlby, 1969).

Statistical analyses
Statistical analyses were conducted in STATA 15 (Stata, version

15.1; StataCorp). Group differences in baseline characteristics were assessed with independent-samples t-tests (or Mann-Whitney U tests for non-normally distributed variables) for continuous variables and
Chi-square tests for categorical variables.
To test the hypothesis that mother-infant dyads who received the intervention would show stronger attachment relationships than those who received CAU (hypothesis 1), linear and logistic regression models were conducted with intervention group predicting EAS Direct scores (six linear models) and the proportion of mothers and infants categorized as being in the EAS Emotionally Available Attachment Zone (two logistic models). Eight models were conducted in total to test Hypothesis 1. Unadjusted and adjusted odds ratios (OR), unstandardized beta coefficients (B), 95% confidence intervals, and p-values are reported.

Adjusted models included infant temperament factors (Surgency,
Negative Affect, Regulation) as continuous independent variables.

To test the hypothesis that infants in the intervention group would
show more differentiated neural responses to mother and stranger face stimuli compared to those in the CAU group (hypothesis 2), one mixed-model ANOVA was conducted on mean Nc amplitudes with the between-subjects factor group (intervention, CAU) and the withinsubjects factors electrode cluster (frontal, central) and condition (mother, stranger). Significant interactions between the factors were further investigated using Bonferroni-corrected planned pairwise contrasts between the levels of each factor. The ANOVA was repeated with the covariates age (given known effects of age on the Nc, Carver et al.,

2003; de Haan & Nelson, 1997) and infant temperament factors.
To test the hypothesis that more differentiated neural correlates of social development at age 6 months would be associated with better social-interaction behavior with mothers at age 12 months (hypothesis 3), Spearman correlation coefficients were computed between 6-month mean Nc amplitudes for mother/stranger conditions and dimensional 12-month mother/infant EAS Direct scores. To limit the number of tests conducted, correlation coefficients were only computed for mother/infant EAS Direct scores that differed significantly between groups. One multinomial logistic regression model was conducted to test whether 6-month mean Nc amplitudes to mother/stranger faces at frontal and electrode clusters predicted 12-month Emotionally Available Attachment Zones.

Sample characteristics
Mothers from the intervention and CAU groups did not differ significantly in baseline sociodemographic characteristics ( EAS, Emotional Availability Scale; Mean Nc amplitude, the mean amplitude measured in the time-range of the Nc (400-600 ms); μv, amplitude measured in microvolts.

Mother-infant attachment relationships at infant age 12 months
EAS Direct scores and EAS Emotional Attachment Zones are shown by group in Table 2. Results of the linear and logistic regression models testing effects of intervention group on these variables are presented in Table 3. Linear regression showed that membership of the Intervention compared to CAU group was associated with significantly higher Child Involvement scores, but was not significantly associated with the remaining EAS Direct scores. Logistic regression revealed a significant effect of intervention group on Child Emotionally Available Attachment Zone, with a significantly higher proportion of infants classified as Emotionally Available (securely attached) in the intervention than CAU group. Group did not predict maternal Emotionally Available Attachment Zone.

TA B L E 3
Results of linear and logistic regression models testing the effects of the intervention on mother-infant attachment relationships when infants were aged 12 months

Neural correlates of infant social development at infant age 6 months
Grand averages of the Nc are plotted by condition, group, and electrode cluster in Figure 2; group means are shown in Table 2. The 2 × 2 × 2 ANOVA revealed no significant effect of group on mean Nc amplitudes (F(1,26) = 1.69, p = 0.205, η 2 = 0.06) and no significant interactions between group and electrode cluster or condition (all F ≤ 1.42, p ≥ 0.244).

3.4
Associations between neural correlates of infant social development at age 6 months and mother-infant attachment relationships at age 12 months Spearman correlation coefficients were computed between mean Nc amplitudes for the mother and stranger faces at the frontal and central electrode clusters and the EAS Direct score variables that differed significantly between groups (Child Involvement Index scores; four F I G U R E 3 Associations between neural correlates of social development at 6 months and attachment behavior at 12 months. Scatterplot shows the significant positive correlation between mean amplitude of the Nc component while viewing the mother's face at frontal scalp at age 6 months and infants' attachment behavior (EAS child involvement index) at 12 months. The regression line represents the correlation computed in the intervention and CAU groups combined; for information purposes only, intervention and CAU infants are plotted in different colors (blue = intervention, yellow = control) correlation coefficients were computed in total). Mean Nc amplitude at the frontal cluster in the mother condition was significantly positively correlated with EAS Child Involvement index (rho(26) = .400, p = .046); infants with more positive (smaller) Nc amplitudes had higher involvement scores (Figure 3). There were no further significant associations (rho ≤ .300, p ≥ .126). The multinomial logistic regression model examining whether mean Nc amplitude for mother/stranger conditions at frontal and central electrodes predicted child emotional attachment zone was non-significant (χ 2 (8) = 14.65, p = 066). These findings suggest that Primeiros Laços buffers the adverse effects of risk factors such as poverty and adolescent motherhood on early attachment development. These positive effects likely have long-term benefits. Behaviors quantified by the EAS Child Involvement index (e.g., making social overtures to the mother) are early indicators of good social interaction skills, which are protective factors contributing to resilience (Domitrovich et al., 2017). Secure early attachment relationships are protective against a range of psychological, socioemotional and behavioral problems in children growing up in adversity (Groh et al., 2017). Further, early-life attachment behaviors are transmitted across generations (Verhage et al., 2016). Therefore, early HVP interventions such as Primeiros Laços represent important methods of breaking transgenerational insecure attachment cycles in highrisk families.

DISCUSSION
However, in contrast to our hypothesis, there were no effects of the intervention on maternal attachment behaviors. Previous work has suggested that attachment relationships depend on the sensitivity of the caregiver's responsiveness to the needs of the child (Zeegers et al., 2017) and that maternal sensitivity is more amenable to intervention than infant attachment security (Bakermans-Kranenburg et al., 2003). Contrary to this work, Primeiros Laços did not affect maternal attachment dimensions, including maternal sensitivity. Still, previous intervention studies targeted at attachment in maltreated infants have similarly indicated that maternal sensitivity did not underlie positive effects of interventions on infant attachment (Cicchetti et al., 2006). The positive effect of Primeiros Laços on infant attachment behaviors at 12 months was supported by the pattern of findings in neural indices of infant social development at age 6 months. In contrast to our hypotheses and previous work (de Haan & Nelson, 1997), the intervention group did not show more differentiated Nc responses to mother and stranger faces compared to CAU infants. However, our dimensional analysis revealed that smaller Nc amplitudes to mother stimuli were correlated with better child involvement scores 6 months later.
This finding is consistent with previous reports of significant correlations between lower Nc amplitude to mother face stimuli and increased interaction-seeking behaviors with the mother in 6-month-old infants of adult mothers (Swingler et al., 2007) as well as lower Nc amplitudes to the mother's face in securely vs. insecurely attached preschoolers (Kungl et al., 2017).
Larger Nc amplitudes are indicative of greater attentional allocation to salient stimuli (de Haan & Nelson, 1997). Reduced Nc amplitudes in association with increased infant social behavior has been interpreted in terms of reduced attentional allocation to the mother because the infant has already established that the mother is the person to whom they should direct behavior, i.e. the foundation of the attachment bond has already been formed, allowing infants to consolidate that relationship via interaction-seeking behaviors while employing less effortful attentional resources to the mother (Swingler et al., 2007). To our knowledge we are the first to report that Nc amplitude reductions predict infant social behavior longitudinally and that this association is present in infants of impoverished adolescent mothers in LMICs who are vulnerable to adverse developmental outcomes.
Indeed, while many authors have emphasized the importance of early social relationships in guiding infant brain development and protecting against negative effects of adverse early environments on child development (Newman et al., 2015), how human infant attachment and neurobiology are related in the first months of life has rarely been investigated (for one exception see Tharner et al., 2011).
Our Nc findings may be informative as to the mechanisms by which Primeiros Laços improves infant attachment development. Fewer attentional resources devoted principally to the mother (smaller Nc amplitudes) may lead to better social-interaction skills in the first year of life and underpin the development of secure attachment in vulnerable infants. It will be important for future work to identify which aspects of Primeiros Laços, such as the relationship with the homevisitor or additional social stimulation for infants, drive these changes in infants' neurobehavioral development.
Our findings should be interpreted in the context of several limitations. Our sample size was modest, particularly for the EEG measures, and attachment data were collected at only one time-point. Future work in larger samples should include longitudinal measures of both neural correlates and behavioral measures of mother-infant attachment to better understand the effects of HVPs on the developmental trajectory of attachment in vulnerable mother-infant dyads. Future studies should also examine factors mediating the impact of HVPs on mother-infant attachment, including maternal psychopathology, which has been negatively associated with secure mother-infant attachment relationships (Verhage et al., 2016). We did not measure the abilities of the home-visitors or their relationships with the family; these factors should be investigated in further research to assess the extent to which they determine the effectiveness of HVPs on early attachment development.

CONCLUSIONS
This study is the first to demonstrate that an HVP grounded in attachment theory can enhance the early development of attachment in infants of adolescent mothers living in poverty in Brazil. Furthermore, our findings suggest that earlier-life enhancements in neural circuitry involved in social processing may contribute to improvements in infant attachment development. Our findings indicate that HVPs could be important tools for improving early socio-emotional development of vulnerable infants in low-resource countries.

ACKNOWLEDGMENTS
We thank the parents and their babies who participated in the study.

CONFLICTS OF INTEREST
Guilherme V. Polanczyk has been in the past 3 years a consultant member of the advisory board and/or speaker for Shire/Takeda and Medice.
He received travel expenses for continuing education support from Shire/Takeda and royalties from Editora Manole. The other authors report no conflicts of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings from this study are not publicly available due to confidentiality and ethical reasons. The data can be requested from the corresponding author, Professor Guilherme V Polanczyk (gvp@usp.br).