As can be found in Table 1, paired-samples t-tests indicated that adoptive mothers scored significantly higher than adoptive fathers in terms of pre-adoptive reflective functioning (t(47) = 2.75, p =.009). The Levene’s test showed no significant differences between mothers and fathers in variance for pre-adoptive reflective functioning (F(1,94) = 0.627, p =.430). Adoptive mothers and fathers did not differ significantly in their perceptions of socioemotional child difficulties or dimensions of child temperament (all ps > 0.05). Additionally, independent-samples t-tests showed that adopted boys and girls did not differ significantly in their age at placement, anthropometric measures, socioemotional difficulties, or temperament (all ps > 0.05). However, parents of adopted girls scored significantly higher on pre-adoptive reflective functioning compared with parents of adopted boys (t(94) =–0.53, p =.03). Finally, the Little’s MCAR test indicated that missing data were missing completely at random, χ2(22, N = 96) = 22.98, p =.40. Consequently, missing data were handled by using the expectation-maximization algorithm [61].

Regarding the anthropometric measures, only a few children (8.9%; three boys and one girl) had a BMI indicative of being underweight (Z-score <–2). Nonetheless, there was substantial variation in BMI scores, indicating a spectrum of healthy and unhealthy weight. Moreover, 31.1% of the children had WFA or LFA Z-scores below −2. For each of these ratios, a score of −2 indicated a growth delay at the time of child placement, reflecting a ratio deviating by two or more standard deviations from the mean for children of the same age.

Regarding socioemotional difficulties, the US norms of the CBCL were used [59, 62], indicating that 13.5% of adoptive parents reported elevated externalizing difficulties, with 8.3% falling in the borderline range (T-scores between 60 and 63) and 7.2% in the clinical range (T-scores ≥64). For internalizing difficulties, 6.2% reported elevated difficulties, with 5.2% in the borderline range (T-scores between 60 and 63) and 1% in the clinical range (T-scores ≥64). A total of 8.3% of children exhibited elevated total socioemotional difficulties, with 5.1% in the borderline range (T-scores between 60 and 63) and 3% in the clinical range (T-scores ≥64). Levene’s tests showed no significant differences between mothers and fathers in variance for internalizing (F(1,75) = 0.008, p =.930), externalizing (F(1,75) = 0.147, p =.702), or total (F(1,75) = 0.060, p =.807) difficulties.

Pearson bivariate correlations were calculated separately for adoptive mothers and fathers and are presented in Table 2. Correlations between maternal and paternal reports of internalizing, externalizing, and total difficulties were r =.71 (p <.01), r =.78 (p <.01), and r =.76 (p <.01), respectively. To avoid multicollinearity due to the high correlation between BMI and WFA (r =.79), BMI was used in all subsequent models because it provides a comprehensive measure of both weight and height, making it a more holistic indicator of growth status compared with WFA. Outcomes for analyses using WFA as predictor are reported in the Supplementary Materials. Pre-adoptive reflective functioning was not significantly correlated with any other variables. However, it should be noted that correlation with other variables is not a requirement for a variable to serve as a moderator. With regard to potential covariates, maternal self-reported level of education was negatively associated with child age at placement. Therefore, while only maternal education showed a correlation, both maternal and paternal education levels were included as covariates to account for potential shared family-level influences in the analyses.

Results of the moderation analyses are presented in Tables 3, 4 and 5. Each model controls for the effects of other proxies used to assess early adversity, in addition to child temperament, child sex, and parent education.

Among adoptive fathers, the analyses indicated that the total model investigating child age at placement as main predictor explained a significant proportion of the variance in total socioemotional child difficulties as perceived by adoptive fathers (R2 = 0.54, F(10, 37) = 4.33, p <.001). Specifically, child age at placement significantly predicted total socioemotional child difficulties (β = 0.51, SE = 0.16, p =.003). Furthermore, the direct association between paternal pre-adoptive reflective functioning and total socioemotional child difficulties was not statistically significant (β =–0.20, SE = 0.12, p =.121). However, the interaction term between paternal pre-adoptive reflective functioning and child age at placement was statistically significant (β =–0.37, SE = 0.14, p =.013) and significantly moderated the association between child age at placement and total socioemotional child difficulties (R2 change = 0.08, F(1, 37) = 6.78, p =.013) (see Fig. 1). As Fig. 1 indicates, the association between child age at placement and child difficulties was most pronounced at the lowest level of pre-adoptive reflective functioning, less pronounced at intermediate levels, and non-significant at the highest level of pre-adoptive reflective functioning.

Paternal reflective functioning, child age at placement, and child total socioemotional difficulties 4 years after placement. The values of A pre-adoptive reflective functioning correspond with low (16th percentile), middle (50th percentile), and high (85th percentile) levels

These findings were robust for internalizing child difficulties but not for externalizing difficulties. Specifically, child age at placement significantly predicted internalizing difficulties (β = 0.40, SE = 0.17, p =.025) and externalizing difficulties (β = 0.58, SE = 0.17, p =.001) as reported by adopted fathers. Furthermore, paternal pre-adoptive reflective functioning did not significantly predict internalizing difficulties (β =–0.23, SE = 0.14, p =.121) or externalizing difficulties (β =–0.11, SE = 0.14, p =.447). Finally, the interaction term between child age at placement and paternal pre-adoptive reflective functioning significantly predicted internalizing child difficulties (β =–0.34, SE = 0.154, p =.035), but did not significantly predict externalizing child difficulties (β =–0.22, SE = 0.15, p =.151).

Among adoptive mothers, the total model explained a significant proportion of the variance in total socioemotional child difficulties as perceived by adoptive mothers as well (R2 = 0.46, F (10, 37) = 3.19, p =.005). While child age at placement significantly predicted total socioemotional child difficulties as perceived by mothers (β = 0.42, SE = 0.16, p =.012), neither the direct effect of maternal pre-adoptive reflective functioning (β =–0.14, SE = 0.15, p = .325) nor the interaction effect of maternal pre-adoptive reflective functioning and child age at placement (β =–0.28, SE = 0.16, p =.092) significantly predicted total socioemotional child difficulties. The interaction of child age at placement and maternal pre-adoptive reflective functioning was not significant, but showed a trend toward statistical significance (R2 change = 0.04, F(1, 37) = 3, p =.092).

These findings were robust regarding externalizing child difficulties, as child age at placement significantly predicted externalizing difficulties (β = 0.47, SE = 0.18, p =.012). In contrast, the direct effect of child age at placement on internalizing difficulties was not statistically significant but showed a trend toward significance (β = 0.28, SE = 0.17, p =.113). Furthermore, maternal pre-adoptive reflective functioning did not significantly predict internalizing (β =–0.15, SE = 0.17, p =.371) or externalizing difficulties (β =–0.00, SE = 0.17, p =.978). Finally, the interaction between child age at placement and maternal pre-adoptive reflective functioning did not significantly predict internalizing difficulties (β =–0.21, SE = 0.17, p =.234) or externalizing difficulties (β =–0.28, SE = 0.18, p =.112), although there was a trend toward statistical significance in the prediction of externalizing child difficulties.

Among adoptive fathers, the total model explained a significant proportion of the variance in total socioemotional child difficulties as perceived by adoptive fathers (R2 = 0.58, F(10, 37) = 5.11, p <.001). Neither child BMI at placement (β =–0.09, SE = 0.08, p =.275) nor paternal pre-adoptive reflective functioning (β =–0.07, SE = 0.12, p =.596) directly predicted total socioemotional difficulties. However, the interaction between child BMI at placement and paternal pre-adoptive reflective functioning did significantly predict total socioemotional child difficulties(β = 0.26, SE = 0.08, p =.002), and paternal pre-adoptive reflective functioning significantly moderated the association between child BMI at placement and total socioemotional child difficulties (R2 change = 0.13, F(1, 37) = 11.01, p =.002). As can be seen in Fig. 2, the associations with child BMI were most pronounced at the lowest level of pre-adoptive reflective functioning, less pronounced at intermediate levels, and non-significant at high levels.

Paternal reflective functioning, child BMI, and child total socioemotional difficulties 4 years after placement. The values of pre-adoptive reflective functioning correspond to low (16th percentile), middle (50th percentile), and high (85th percentile) levels

These results were generally consistent for both internalizing and externalizing child difficulties, as child BMI at placement did not predict internalizing difficulties (β =–0.00, SE = 0.08, p =.953) or externalizing difficulties (β =–0.05, SE = 0.08, p =.540) as perceived by adoptive fathers. Furthermore, paternal pre-adoptive reflective functioning did not significantly predict internalizing difficulties (β =–0.06, SE = 0.13, p =.659) or externalizing difficulties (β = 0.02, SE = 0.13, p =.880). Similarly, consistent with the findings for total socioemotional child difficulties, the interaction between child BMI at placement and paternal pre-adoptive reflective functioning significantly predicted internalizing difficulties (β = 0.32, SE = 0.08, p <.001) and externalizing difficulties (β = 0.24, SE = 0.08, p =.003). However, a notable exception was observed for internalizing difficulties, where the significant moderation effect was found at both the lowest and highest levels of the moderator. In contrast, the moderation effect for total socioemotional difficulties was significant at the lowest and intermediate levels of the moderator.

Among adoptive mothers, the total model explained a significant proportion of the variance in total socioemotional child difficulties as perceived by adoptive mothers (R2 = 0.43, F(10, 37) = 2.76, p =.012). Furthermore, neither child BMI at placement (β =–0.11, SE = 0.09, p =.248) nor maternal pre-adoptive reflective functioning (β =–0.11, SE = 0.15, p =.493) significantly predicted total socioemotional child difficulties as perceived by mothers. Likewise, the interaction term between child BMI at placement and maternal pre-adoptive reflective functioning (β =–0.10, SE = 0.15, p =.493) was not statistically significant. Consequently, maternal pre-adoptive reflective functioning did not significantly moderate the association between child BMI at placement and total socioemotional child difficulties (R2 change = 0.01, F(1, 37) = 0.46, p =.504).

These results were robust for both internalizing and externalizing child difficulties, as child BMI at placement did not predict internalizing difficulties (β =–0.05, SE = 0.09, p =.618) or externalizing difficulties (β =–0.12, SE = 0.10, p =.254). Furthermore, maternal pre-adoptive reflective functioning did not directly predict internalizing (β =–0.11, SE = 0.17, p =.533) or externalizing difficulties (β = 0.03, SE = 0.18, p =.848). Finally, consistent with the findings for total socioemotional child difficulties, the interaction between child BMI at placement and maternal pre-adoptive reflective functioning did not significantly predict externalizing difficulties (β = 0.06, SE = 0.10, p =.576). However, the effect showed a trend toward significance for internalizing difficulties (β = 0.14, SE = 0.09, p =.135).

Among adoptive fathers, the total model explained a significant proportion of the variance in total socioemotional child difficulties as perceived by adoptive fathers (R2 = 0.46, F(10, 37) = 3.18, p =.005). Neither child LFA at placement (β = 0.00, SE = 0.10, p =.978) nor paternal pre-adoptive reflective functioning (β =–0.31, SE = 0.23, p =.197) directly predicted total socioemotional difficulties. Furthermore, the interaction between child LFA at placement and paternal pre-adoptive reflective functioning was not statistically significant (β =–0.07, SE = 0.11, p =.488). Consequently, paternal pre-adoptive reflective functioning did not significantly moderate the association of child LFA at placement and total socioemotional child difficulties (R2 change = 0.01, F(1, 37) = 0.49, p =.488).

When investigating internalizing and externalizing child difficulties as perceived by adoptive fathers separately, the findings remained unchanged. LFA at placement did not significantly predict internalizing difficulties (β = 0.07, SE = 0.11, p =.512) or externalizing difficulties (β =–0.05, SE = 0.10, p =.647). Furthermore, paternal pre-adoptive reflective functioning did not directly predict internalizing difficulties (β = 0.07, SE = 0.11, p =.525) or externalizing difficulties (β =–0.09, SE = 0.24, p =.703). Similarly, the interaction term between child LFA at placement and paternal pre-adoptive reflective functioning did not significantly predict internalizing difficulties (β =–0.09, SE = 0.11, p =.456) or externalizing difficulties (β = 0.00, SE = 0.10, p =.976).

Among adoptive mothers, the total model explained a significant proportion of the variance in total socioemotional child difficulties as perceived by adoptive mothers (R2 = 0.42, F(10, 37) = 2.68, p =.014). Neither child LFA at placement (β = 0.01, SE = 0.12, p =.912) nor maternal pre-adoptive reflective functioning (β =–0.16, SE = 0.33, p =.622) significantly predicted total socioemotional child difficulties as perceived by mothers. Likewise, the interaction term between child LFA at placement and maternal pre-adoptive reflective functioning (β =–0.02, SE = 0.16, p =.887) was not statistically significant. Consequently, maternal pre-adoptive reflective functioning did not significantly moderate the association between child LFA at placement and total socioemotional child difficulties (R2 change = 0.00, F(1, 37) = 0.02, p =.887).

When examining internalizing and externalizing child difficulties as perceived by adoptive mothers separately, the results showed that LFA at placement was not a significant predictor of internalizing difficulties (β = 0.06, SE = 0.13, p =.610) or externalizing difficulties (β =–0.04, SE = 0.13, p =.762). Furthermore, maternal pre-adoptive reflective functioning did not directly predict internalizing difficulties (β =–0.11, SE = 0.35, p =.759) or externalizing difficulties (β =–0.06, SE = 0.36, p =.863). Finally, the interaction between child LFA at placement and maternal pre-adoptive reflective functioning did not significantly predict internalizing difficulties (β = 0.02, SE = 0.17, p =.898) or externalizing difficulties (β =–0.03, SE = 0.18, p =.867).