Low socioeconomic status early in life is related to a higher risk of child and adolescent behavior problems (1, 2). Previous research has shown that children living in low-income or single-parent households (1), or having a parent with less than a post-secondary education (3) are more likely to experience mental health challenges, including behavior problems. The reason for these associations is guided by the Family Stress Model, which describes how economic stress can create stress within the family unit, leading to emotional distress, disrupted parenting, and the potential outcomes of child and adolescent behavioral problems (4).

Breastfeeding is one behavior that has been positively linked to improved child behavior (5), partly through the pathways of improved parent–child relationships (5) and infant attachment (6). In alignment with the Family Stress Model, breastfeeding may act as a resource that disrupts the connection between socioeconomic stress and child behavior problems by facilitating positive parent–child relationships and secure attachment between the parent and child. These positive factors may reduce the negative effects of socioeconomic stress. While breastfeeding may be one factor that facilitates this pathway, it is important to note that the development of secure parent–child relationships and attachment is possible and available to families who do not breastfeed through other pathways such as responsive parenting and warm interactions (7, 8).

In high-income countries, families living with low socioeconomic status experience more barriers to breastfeeding, such as inconsistent social support (9), and psychosocial stress (10) and are less likely to breastfeed their children (11). However, previous study has shown that breastfeeding has a stronger association with positive child cognitive outcomes for those who are already experiencing risk factors (12, 13). Whether the moderating role of breastfeeding extends to other outcomes, such as child behavior, is unknown and has not been previously examined. Therefore, our exploratory research aimed to determine if breastfeeding moderates the association between family SES and child behavior scores.

The study population, stratified by higher vs. lower family SES, is described in Table 1. While 72.0% of the sample had no SES risk factors, 19.1% had one risk factor and 8.8% had two or three. As expected, families with lower SES were less likely to breastfeed; 13.9% were exclusively breastfeeding at 6 months with a mean duration of any breastfeeding of 9.9 months (standard deviation [SD] = 7.1), compared to 20.1% of higher SES families exclusively breastfeeding at 6 months with a mean duration of 11.6 months [SD = 6.6; p < 0.001 for both breastfeeding variables]. Child behavior scores were also elevated in families with lower SES, with a mean total behavior score of 42.9 [SD = 10.3], compared to a mean total score of 40.6 [SD = 8.5] in the high SES group (p < 0.001, however, Cohen’s d to measure the effect size of the difference is small at −0.26 [95% CI: = − 0.36, −0.17]). Families with lower SES were more likely to experience prenatal maternal depression (p < 0.001) and have three or more children (p < 0.001).

In adjusted regression models, lower family SES was related to higher (worse) total behavior scores (2.06 [95%CI: 1.06, 3.07] point increase, or approximately 1/5 of a standard deviation), while breastfeeding exclusivity and duration were related to lower (better) total behavior scores in a dose-dependent manner (−2.43 [95% CI: −3.74, −1.11] for exclusive breastfeeding, and −1.17 [95%CI: −2.22, −0.11] for partial breastfeeding; −0.07 [95% CI: −0.13, −0.003] for each month of breastfeeding, Table 2). The same direction and similar magnitude of effect were seen for both internalizing and externalizing behavior scores.

Significant adjusted interactions were observed between family SES, and exclusive breastfeeding at 6 months, for both internalizing and total behavior scores (p < 0.05 for internalizing scores and p < 0.01 for total scores). A marginally significant interaction was observed between family SES and exclusive breastfeeding at 6 months for externalizing behavior scores (p = 0.06). Significant adjusted interactions were observed between family SES and breastfeeding duration for both internalizing and total behavior scores (p < 0.05 for both internalizing and total scores), although this interaction was not significant for externalizing behavior scores (p = 0.11). Complete regression model statistics, including interaction beta coefficients and partial eta-squared effect sizes, can be found in Supplementary Table 1. To enhance interpretation of the interactions, Figures 1A–F show the change in internalizing, externalizing, and total CBCL scores by breastfeeding for both lower and higher family SES groups. These interaction plots show that the “socioeconomic gap” in behavior scores becomes smaller with more exclusive and longer breastfeeding. For example, among those with higher SES, there is little change in total behavior scores with more exclusive breastfeeding (i.e., no significant differences between breastfeeding levels, calculated using ANOVA, F(2, 1,237) = 1.90, p = 0.150, Figure 1E, orange line), while for those with lower SES, more exclusive breastfeeding is related to better total behavior scores (calculated using ANOVA, F(2, 444) = 8.88, p<0.001, Figure 1E, green line). The direction of effects and interpretation is the same for panels A, B, C, and F. No interactions were observed for other breastfeeding exposures (i.e., ever breastfeeding, breastfeeding mode, or breastfeeding exclusivity at 3 months).

This exploratory Canadian study shows that the association between family SES and child internalizing and total behavior scores is moderated by breastfeeding exclusivity and duration, suggesting that breastfeeding may be one factor that can help “close the gap” in behavior problems experienced disproportionately by children living with lower SES. These associations remained after adjustment for important confounders including prenatal maternal depression and number of older siblings.

Notably, we did not observe statistically significant interactions with externalizing behavior problems; however, the direction and magnitude of the main effects, interaction coefficients, and visualization of change in externalizing scores by breastfeeding for lower and higher SES groups were similar to internalizing and total behavior problems. Previous studies have shown that the link between breastfeeding and internalizing behavior is partially due to improved mother–child relationships (5, 22), which is one factor that may disrupt the connection between socioeconomic stress and child behavior problems. This mechanism may not apply as strongly to the outcome of externalizing behavior which would explain the lack of statistically significant findings here.

While the CBCL effect sizes in the main effects and interaction models are small (all less than a one SD or 10-point change) and may not represent individual clinical relevance, small effect sizes can still have population-level relevance. This is explained by Roses Theorem which states that small effect sizes can be meaningful at the population-level because a large number of people at a small risk can result in more incidence of the outcome than a small number of people at a high risk (23).

Similar to our findings suggesting differential breastfeeding effects depending on SES risk, another study in the United States showed that breastfeeding exclusivity and duration were related to lower prevalence of conduct problems in kindergarten for those with high genetic risk, with a much smaller association for those with low genetic risk (13). These findings suggest that those in high-risk environments (i.e., socioeconomic or genetic) will benefit more from health-promoting early life exposures (such as breastfeeding) than those in lower risk environments.

Our results are particularly important because those with lower SES are typically the populations that experience more barriers to breastfeeding and have lower breastfeeding rates (observed in other studies (24) and confirmed here in the CHILD cohort). The reasons for this are multifactorial and span the social determinants of health. For example, families living with lower SES are more likely to experience food insecurity, unstable housing or homelessness, inconsistent employment, mental health concerns, and less consistent social support, all of which add extra family stress and can impact breast milk production, the ability to access breastfeeding support, or limit the time available to breastfeed (9, 25, 26). For breastfeeding to be a possible choice for all families, policies and programs that reduce barriers to breastfeeding among those living in low SES environments are necessary (27). The results of this study support programs like the Canadian Prenatal Nutrition Program that provide targeted support to those living in more disadvantaged settings to help close the health equity gap between those living in advantaged settings (28). This study also supports policies and programs to increase SES, such as basic annual income and increasing the Canada Child Benefit to promote equity and reduce child poverty (29). Based on the current findings, breastfeeding may be one strategy that could help narrow the gap in child behavior scores between those with lower vs. higher SES.

The CHILD cohort comprises relatively higher SES families with fewer child behavior problems compared to the general Canadian population and the CBCL normative population (11, 17). Therefore, our results may underrepresent the true associations of family SES and breastfeeding with child behavior. Furthermore, the family SES variable has not been validated; the use of a composite score to define SES has been used previously, but with low education defined as less than a high school degree, instead of less than college/university which we used here due to very few mothers with less than a high school degree in our sample (18). Thirdly, unmeasured residual confounding from variables such as parenting practices, early child temperament, and infant health problems may account for some or all of the observed associations. Finally, this study was underpowered to examine nuanced interactions with a three-level SES variable, requiring further categorization which limited the resolution of this analysis.

This study provides new evidence that breastfeeding may be one factor that can help “close the SES gap” in child behavior scores. This evidence highlights the importance of breastfeeding support services, particularly in populations with lower SES.