Sustained Change in Adoptive Families: A Follow-up Mediation Analysis of Adoption-Specific Therapy (ADAPT) on Behavioral Problems and Attachment in Adopted Children and Parental Stress in Adoptive Parents

2.1 Study design and participants

This study presents the secondary analysis of three-month follow-up data from a parent randomized controlled trial (RCT) that evaluated the efficacy of Adoption-Specific Therapy (ADAPT). The original study was a randomized controlled trial registered in the Iranian Registry of Clinical Trials (IRCT20250107064305N1), and the present manuscript reports secondary and follow-up analyses of that trial. The full methodology, including recruitment, randomization procedure, intervention details, and immediate post-test results, have been published previously. The present analysis focuses exclusively on the follow-up data collected three months after the completion of the post-test assessments and introduces advanced statistical modeling not reported in the initial efficacy paper.

The original sample consisted of 30 adoptive families from Tehran and Alborz provinces in Iran, randomly assigned to either the ADAPT intervention group (n = 15) or a waitlist control group (n = 15). The final sample for the current follow-up and mediation analysis comprised 25 families who provided complete data at all three time points: pre-test (T1), post-test (T2), and three-month follow-up (T3). This included 12 families in the ADAPT group and 13 families in the control group. All analyses reported in this paper (both RM-MANCOVA and mediation models) were conducted on this complete-case sample of 25 families. Attrition from the original sample was due to logistical reasons (relocation, scheduling conflicts) and not to adverse effects of the intervention, with no significant demographic differences between completers and non-completers. All participating children were adopted after the age of three and were between 7 and 12 years old at the time of the study.

2.2 Ethical considerations and informed consent

The study protocol was reviewed and approved by the Ethics Committee of Iran University of Medical Sciences (IR.IUMS.REC.1403.113) and was registered in the Iranian Registry of Clinical Trials (ID Number: IRCT20250107064305N1). Written informed consent was obtained from all adoptive parents prior to participation after providing a full explanation of the study aims, procedures, and confidentiality of the data. As the study involved children aged 7–12 years, age-appropriate verbal assent was obtained from all child participants. During the first session, children were asked whether they understood the reason for attending therapy, and the study procedures and therapeutic activities were explained to them in simple, developmentally appropriate language in accordance with the treatment protocol. Participation was voluntary, and families were informed that they could withdraw from the study at any time without any consequences.

2.3 Intervention: Adoption-specific Therapy (ADAPT)

As detailed in the primary outcome paper, ADAPT is a manualized, multi-modular intervention comprising 34 sessions delivered over 20 weeks: 14 individual child sessions, 14 individual parent sessions, and 6 conjoint parent-child sessions. The therapy addresses core adoption-related themes, including managing behavior, developing a coherent adoption narrative, processing grief and loss, building attachment, exploring identity, and managing public discourse about adoption. The intervention was delivered by a licensed clinical psychologist (the first author) under regular supervision to ensure treatment fidelity. The control group was placed on a waitlist and received no structured intervention during the active study period but was offered a compensatory educational program after all study assessments were complete.

2.4 Measures

The same standardized, parent-report measures used at T1 and T2 were administered at the three-month follow-up (T3).

2.5 Analytical strategy for the current study

The analysis proceeded in two sequential phases to address the two main aims of sustainability and mechanisms of change.

All analyses were performed using IBM SPSS Statistics version 27, with an alpha level of.05. We acknowledge that the sample size, while sufficient for detecting main effects, may limit the statistical power to detect smaller or more complex indirect effects in the mediation models.

3.1 Preliminary analyses and descriptive statistics

The final analytical sample consisted of 25 families who provided complete data at all three assessment points: pre-test (T1), post-test (T2), and the three-month follow-up (T3). This included 12 families in the ADAPT group and 13 in the waitlist control group. Independent samples *t*-tests and chi-square analyses confirmed no significant baseline differences between the two groups on key demographic variables, including child’s age, sex, age at adoption, and length of prior institutional stay, or on pre-test scores of the primary outcome measures (all *p* > .05), supporting successful randomization.

Descriptive statistics (means and standard deviations) for the Child Behavior Checklist (CBCL), Kinship Center Attachment Questionnaire (KCAQ), and Parenting Stress Index (PSI) across the three time points for both groups are presented in Table 1. Visual inspection reveals a clear pattern of improvement from T1 to T3 in the ADAPT group across all measures, while scores in the control group remained stable (see Figures 1, 2 and 3).

Figure 1. Bar charts showing pre-, post- and follow up test of CBCL Scores.

Legend: Mean Child Behavior Checklist (CBCL) Total Problems scores for the ADAPT intervention group and the waitlist control group across pre-test (T1), post-test (T2), and three-month follow-up (T3) assessments. Lower scores indicate fewer behavioral problems.

Figure 2. Bar charts showing pre-, post- and follow up test of KCAQ Scores.

Legend: Mean Kinship Center Attachment Questionnaire (KCAQ) scores for the ADAPT intervention group and the waitlist control group across pre-test (T1), post-test (T2), and three-month follow-up (T3) assessments. Higher scores indicate more secure attachment behaviors.

Figure 3. Bar charts showing pre-, post- and follow up test of PSI Scores.

Legend: Mean Parenting Stress Index (PSI) Total Stress scores for the ADAPT intervention group and the waitlist control group across pre-test (T1), post-test (T2), and three-month follow-up (T3) assessments. Lower scores indicate lower parenting stress.

3.2 Phase 1 analysis: Testing the sustainability of intervention effects

To examine whether treatment gains were sustained at follow-up, a 2 (Group: ADAPT, Control) × 3 (Time: T1, T2, T3) mixed-design Repeated-Measures Multivariate Analysis of Covariance (RM-MANCOVA) was performed, controlling for the child’s age at adoption. Preliminary checks were conducted. Box’s M test (M = 88.325) was non-significant (*p* = .291), supporting the homogeneity of covariance matrices. Mauchly’s test indicated a violation of the sphericity assumption for all univariate tests: CBCL (W = .401, χ²(2) = 16.436, *p* < .001), KCAQ (W = .460, χ²(2) = 13.977, *p* = .001), and PSI (W = .152, χ²(2) = 33.959, *p* < .001). Therefore, all reported univariate within-subjects effects use the conservative Greenhouse-Geisser (G-G) correction.

The RM-MANCOVA revealed a significant and large multivariate Group × Time interaction effect, Pillai’s Trace = .777, F(6, 14) = 8.12, *p* = .001, partial η² = .777, observed power = .995. This indicates distinct trajectories of change over time between the groups across the combined outcome variables.

Follow-up univariate RM-ANCOVAs, using G-G correction, confirmed significant Group × Time interactions for each outcome. Detailed results, including corrected degrees of freedom, effect sizes, and observed power, are presented in Table 2.

To decompose these interactions and directly test sustainability, simple effects analyses were conducted. First, the effect of Time was examined separately within each group. In the ADAPT group, there was a significant multivariate effect of Time (Pillai’s Trace = .512, F(6, 22) = 11.55, *p* < .001), with significant linear improvements from T1 to T3 on all measures: CBCL (*p* = .027), KCAQ (*p* = .006), and PSI (*p* < .001). Critically, paired comparisons between T2 and T3 within the ADAPT group were non-significant for all outcomes (all *p* > .05), indicating maintenance of post-intervention gains without decay. In the control group, the effect of Time was non-significant (Pillai’s Trace = .145, F (6, 22) = 1.86, *p* = .179), with no significant changes on any measure across time points.

Second, between-group comparisons at each time point were conducted. While groups were equivalent at T1 (all *p* > .05), the ADAPT group showed a strong trend towards better attachment at T2 (*p* = .067) and significantly higher attachment scores at T3 (*p* = .017). Differences in CBCL and PSI scores at T3 favored the ADAPT group but did not reach statistical significance in these pairwise comparisons (*p* = .302 and *p* = .108, respectively).

3.3 Phase 2 analysis: Investigating mechanisms of sustained change

To identify the mechanisms underlying sustained changes at follow-up and to determine which key domain improvement (child behavioral problems, attachment quality, or parenting stress) acted as a mediator for change in the other domains, a multiple mediation analysis was performed using the PROCESS macro in SPSS (Models 4 and 6). In all models, the group variable (intervention = 1, control = 0) was the independent variable (X). Follow-up (T3) scores on the CBCL (behavioral problems), KCAQ (attachment), and PSI (parenting stress) served as potential mediating (M) and outcome (Y) variables. To isolate the net intervention effect from the baseline levels, pre-test (T1) scores for each of these variables were entered as covariates. To examine a wide range of plausible theoretical mechanisms, two mediation structures were tested: (1) Parallel mediation models (PROCESS Model 4), where the intervention effect is transmitted independently and simultaneously through two mediators to the third outcome, and (2) Serial mediation models (PROCESS Model 6), where the intervention effect flows through a two-stage sequence of mediators (X → M1 → M2 → Y) to the final outcome. In total, nine theoretical models (three parallel and six serial) were examined. Bootstrap resampling with 5,000 samples and a 95% bias-corrected percentile confidence interval (CI) was used to test the significance of indirect effects. A summary of the results for all models is presented in Table 3.

Findings from Parallel Mediation Models (Models 1 to 3): The results from the three parallel models indicated that while the ADAPT intervention had significantly and independently led to reduced child behavioral problems (B = -4.86, p = .032; Models 1 & 2), improved attachment (B = 8.64, p = .008; Models 1 & 3), and reduced parenting stress (B = -24.12, p = .005; Models 2 & 3) at follow-up, none of these changes served as independent and parallel mediators for influencing the third domain. In none of these models did the bootstrap confidence intervals for the specific indirect effects cross zero (all included zero). However, in Model 3 (where reduced parenting stress and improved attachment were tested as parallel mediators for reduced behavioral problems), the direct effect of the intervention on reducing behavioral problems remained significant and strong even after controlling for these two mediators (B = -8.76, p = .015). This suggests that a substantial portion of the intervention’s effect on reducing behavioral problems was transmitted through pathways other than stress reduction and attachment improvement within this parallel framework, likely stemming from direct behavioral management skill training.

Findings from Serial Mediation Models (Models 4 to 9): Given the lack of support for parallel models, six serial models were tested to examine potential causal sequences. Crucially, in none of the six serial models was the full three-path indirect effect (X → M1 → M2 → Y) statistically significant (all confidence intervals included zero). This finding indicates that the data do not strongly support a single, fixed, and exclusive linear sequence of change (e.g., “only stress reduces first, then only attachment improves, and only then do behavioral problems decrease”).

Despite this, the pattern of significant direct effects within the serial models and the notable (though non-significant) magnitudes of indirect effects provided important insights into the dynamics of change (see Table 3):

Summary and Synthesis of Findings: Overall, the mediation analyses revealed that the ADAPT intervention exerted simultaneous and direct significant effects on all three core domains within adoptive families at follow-up. However, the data do not support a single, linear, fixed mechanism (whether parallel or serial). Instead, the pattern of results is more consistent with a reciprocal and multi-pathway model of change. In this model, a strong initial reduction in parenting stress (likely resulting from the supportive and psychoeducational components of the intervention) appears to be a key lever, contributing both directly and indirectly (by fostering a more positive emotional climate for relationship improvement) to the reduction of behavioral problems. Concurrently, reduced behavioral problems (likely stemming from behavioral management components) also directly alleviate parental stress burden. Improved attachment likely benefits both from the improved emotional climate following reduced parental stress and from more positive interactions resulting from decreased behavioral conflicts. These findings suggest that the sustained effects of the ADAPT intervention are maintained not through a simple linear chain of causation, but through an interwoven and mutually reinforcing network of reciprocal influences among parental functioning, parent-child relationship, and child behavior. The lack of statistical significance for the indirect effects in these analyses can be largely attributed to the limited sample size and the consequent low statistical power for detecting such complex effects.

4.1 The sustainability of change: Evidence for lasting impact

Our first key finding is that the significant post-intervention improvements in child behavioral problems, attachment security, and parental stress were not only maintained but continued to show a positive, non-decaying trajectory three months after treatment cessation. The significant and large Group × Time interaction, coupled with the non-significant change between post-test and follow-up within the ADAPT group, robustly confirms the durability of therapeutic gains. This stability is clinically vital, as it suggests that ADAPT equips families not merely with temporary coping strategies but with fundamental shifts in relational patterns, emotional regulation, and interpretive frameworks that persist beyond the therapeutic scaffold.

The maintenance of reduced behavioral problems aligns with the trauma-informed and skill-based components of ADAPT (Waterman et al., 2018). By reframing challenging behaviors as survival strategies rooted in past adversity and by teaching concrete emotion-regulation and problem-solving skills (Modules 1 & 2), children likely internalized more adaptive ways of managing distress. The ongoing reduction in CBCL scores at follow-up indicates that these new competencies became integrated into the child’s behavioral repertoire. Similarly, the continued improvement in attachment security (KCAQ) speaks to the intervention’s success in fostering a gradual reorganization of internal working models. Activities like constructing the “invisible thread” of connection (Module 4) and processing grief through the “loss box” (Module 3) appear to have initiated a reparative relational process that continued to unfold post-treatment, facilitated by a now more sensitive and attuned caregiving environment (Hillman et al., 2020).

Perhaps most notable is the sustained reduction in parental stress (PSI). Parenting stress in adoptive contexts is often chronic, fueled by complex child needs, feelings of inadequacy, and a lack of specialized support (Canzi et al., 2019; Palacios & Sánchez-Sandoval, 2006). ADAPT’s direct, multi-pronged approach to reducing this burden—through psychoeducation, normalization of challenges, enhancement of parental self-efficacy, and the provision of a supportive therapeutic alliance—seems to have created a lasting buffer. Parents in the intervention group reported lower stress not just at the peak of therapeutic support, but months later, suggesting a fundamental change in their appraisal of and capacity to manage parenting challenges.

4.2 Unpacking the mechanisms: From linear pathways to a systemic model of change

The second, and more theoretically significant, aim was to elucidate the mechanisms of these sustained changes. Our mediation analyses yielded a clear and important conclusion: the data did not support a single, dominant, linear causal pathway (e.g., a strict behavior-focused or parent-focused sequence). None of the nine tested parallel or serial mediation models produced a statistically significant full indirect effect. This null finding, however, is not indicative of a lack of process but rather illuminates the complex, reciprocal nature of change within family systems (Cicchetti & Toth, 2009).

The pattern of significant direct effects of the intervention on all three follow-up outcomes, even after controlling for their interrelationships and baseline levels, is highly informative. It demonstrates that ADAPT exerted a simultaneous and independent influence on the child’s behavior, the quality of the parent-child bond, and the parent’s emotional state. This is consistent with the intervention’s modular, multi-target design, which concurrently addresses behavioral management, attachment narratives, and parental coping.

Rather than a simple chain of causation, the most parsimonious interpretation of our results is a reciprocal, systemic model of change. In this model, improvement in one domain catalyses and reinforces improvement in the others, creating a virtuous cycle that sustains gains over time. For instance:

Reduced Parental Stress as a Key Lever: The strong direct effect of ADAPT on reducing PSI suggests that alleviating the parental burden may be a primary entry point. As parental stress decreases, resources for sensitive, reflective caregiving are freed (Malcorps et al., 2024). A less-stressed parent is more emotionally available, patient, and capable of providing the consistent, responsive care that is the bedrock of secure attachment (Bakermans-Kranenburg, van, & Juffer, 2003). This improved caregiving climate, in turn, reduces child anxiety and the need for attention-seeking or dysregulated behaviors.

Improved Attachment as a Relational Foundation: Concurrently, ADAPT’s direct work on attachment (Module 4) helps build a stronger relational foundation. As the child feels safer and more understood, their trust in the caregiver increases, making them more receptive to parental guidance and limits. This receptivity reduces power struggles and non-compliance, thereby decreasing a major source of parental stress (Schoemaker et al., 2020).

Reduced Behavioral Problems as a Relational Relief: Directly reducing behavioral problems through skill-building (Module 1) provides immediate relief to the parent-child dyad. Fewer daily conflicts lower the overall tension in the home, making positive interactions more likely. This more harmonious environment is conducive to the growth of attachment and further reduces parental feelings of being overwhelmed or ineffective.

Thus, ADAPT appears to work by strategically intervening at multiple points within this interconnected system—child behavior, parental psychology, and the dyadic relationship—simultaneously. The sustained effects are maintained not because change follows a predetermined sequence, but because improvements in each domain continuously support and stabilize one another, creating a new, more functional family homeostasis.

4.3 Clinical and theoretical implications

These findings have profound implications for both practice and theory. Clinically, they argue against a rigid, stepwise approach to treating adoptive families. Therapists need not wait for a reduction in parental stress before addressing attachment, or for perfect behavioral control before exploring narrative work. Instead, an integrated, systemic approach like ADAPT, which targets multiple levels of the family system concurrently, appears most effective in creating robust and lasting change. The sustained reduction in stress also highlights the critical need to directly support the adoptive parent, not just as a change agent for the child, but as a client in their own right who is navigating a uniquely challenging parenting journey.

Theoretically, our results underscore the utility of a family systems perspective in understanding post-adoption adjustment. They challenge unidirectional models of influence (e.g., child problems → parent stress) and support dynamic, transactional models where child characteristics, parental functioning, and relationship quality are in constant, reciprocal interaction (Sameroff, 2009). This aligns with contemporary developmental psychopathology frameworks that emphasize the multidetermined and systemic nature of adaptive and maladaptive outcomes (Cicchetti & Toth, 2009).