Acceptance and commitment therapy in parents of children with cancer at psychosocial risk: A randomized multiple baseline evaluation

Full reference: Bautista, A. B., Ruiz, F. J., & Suárez-Falcón, J. C. (2023). Acceptance and commitment therapy in parents of children with cancer at psychosocial risk: A randomized multiple baseline evaluation. Journal of Contextual Behavioral Science, 29, 109–121. https://doi.org/10.1016/j.jcbs.2023.06.004

Study type: Randomized multiple-baseline design (single-case experimental design)

Background and objectives

Childhood cancer is associated with significantly higher levels of emotional distress in parents and primary caregivers compared to normative samples. The Pediatric Psychosocial Preventative Health Model (PPPHM; Kazak, 2006) classifies families into three risk levels: universal (low risk), targeted (moderate risk), and clinical (high risk). Although evidence indicates that primary caregivers at higher psychosocial risk demonstrate higher treatment dropout rates—potentially due to experiential avoidance—most previous studies evaluating psychological interventions failed to stratify families by their psychosocial risk level. Recently, Repetitive Negative Thinking (RNT)-focused Acceptance and Commitment Therapy (ACT) has demonstrated large effects for depression and anxiety through single-case experimental designs and randomized controlled trials.

The objectives of this study were to evaluate the feasibility, acceptability, and preliminary efficacy of a brief (two-session), online, RNT-focused ACT protocol in parents of children with cancer at moderate and high psychosocial risk. RNT-focused ACT is grounded in Relational Frame Theory (RFT) and proposes that RNT operates as a transdiagnostic process that maintains emotional difficulties. The intervention aims to help parents develop a more flexible relationship with their repetitive negative thoughts and redirect their behavior toward valued actions, even in the presence of such thoughts.

Method

Participants

Twenty-seven potential participants were recruited from three Spanish hospitals: Hospital Vall d'Hebron (Barcelona), Hospital Miguel Servet (Zaragoza), and Hospital Son Espases (Palma de Mallorca). Inclusion criteria were: (a) being the primary caregiver of a child undergoing active cancer treatment, (b) at least 1 month post-diagnosis, (c) moderate or high psychosocial risk as determined by the Psychosocial Assessment Tool (PAT), and (d) internet access. Participants were excluded if they were currently receiving psychological or psychiatric treatment or if their child was receiving palliative care.

Of the 27 assessed, 16 met inclusion criteria, 15 consented to participate, but 3 discontinued during the baseline phase, resulting in a final sample of 12 participants. Age ranged from 30 to 62 years (M = 41.42, SD = 8.71). The majority were female (9 of 12), married, and had primary or secondary school education. Seven participants were Spanish nationals and five were from Latin American countries. Regarding risk level, 5 participants were classified as high risk (clinical) and 7 as moderate risk (targeted). The most frequent diagnosis was childhood leukemia. The mean age of children was 8.58 years (SD = 4.14), and time since diagnosis ranged from 1 to 29 months (M = 8.75, SD = 7.15).

Design

A randomized, nonconcurrent, multiple-baseline design across participants was employed. Participants were randomly assigned to 2- or 4-week baseline phases using Research Randomizer. One participant (P10) received the intervention after an 8-week baseline due to scheduling difficulties. The study included daily ecological measures throughout the protocol via the m-path app. The SCRIBE statement was followed for reporting. The primary outcome variable was emotional symptoms (measured by PHQ-4), while process variables included RNT (RNTQ-3) and valued living (VQ-3). An 8-week follow-up was conducted after completion of the second session.

Intervention

The RNT-focused ACT protocol consisted of two individual 60-minute sessions delivered online via Microsoft Teams, grounded in principles of Relational Frame Theory (RFT). The protocol was adapted from previous work by Ruiz, Peña-Vargas, and colleagues (2020) evaluated in an RCT.

In the first session, participants developed skills to focus on valued actions and disrupt RNT. RNT was introduced as the central process leading away from values. Therapists explored the main triggers for each participant's RNT and their valued actions, as well as experiential avoidance strategies linked to RNT. A physical metaphor was used to amplify the negative consequences of the inflexible pattern, and an alternative flexible pattern was presented that emphasized distancing from RNT and commitment to valued actions.

In the second session, multiple experiential exercises promoted discrimination of the RNT process and its disruption through distancing from triggers and engaging in valued actions. Four audio files were provided for home practice. All sessions were conducted by the first author (with 2 years of postgraduate ACT training) under supervision by the second author. The complete protocol is available at https://osf.io/c6q2h/.

Treatment integrity

Treatment integrity was assessed by analyzing 8 randomly selected sessions (approximately one-third of all sessions) using two independent raters who evaluated therapist adherence and competence. Mean adherence was 97.22% (Kappa = 0.86, interpreted as "almost perfect"), and mean competence was 95.83% (Kappa = 0.82, "almost perfect").

Outcome measures

The Psychosocial Assessment Tool (PAT) was used to classify participants by psychosocial risk level (scale 0–7 with three risk categories). The Personal Health Questionnaire-4 (PHQ-4) is a brief 4-item screening instrument for anxiety and depression. A daily adapted version was used that asked about symptoms in the past 24 hours (scale 0–4), demonstrating good internal consistency (α = .85) and hierarchical factor structure. Daily mean was 3.30 (SD = 3.16) with a correlation of 0.81 with the original version.

The Repetitive Negative Thinking Questionnaire-3 (RNTQ-3) is a 3-item measure assessing engagement in RNT using a 5-point Likert scale (range 0–12), demonstrating excellent internal consistency (α = .89) and unidimensional structure. Daily mean was 2.40 (SD = 2.44) with a correlation of 0.74 with the original PTQ. The Values Questionnaire-3 (VQ-3) consists of 3 items measuring progress in valued living (scale 0–12), with α = .85 and unidimensional structure. Daily mean was 6.75 (SD = 3.70) with a correlation of 0.51 with the original Values Questionnaire Progress subscale.

Data analysis

Longitudinal factorial invariance of the daily measures was tested via dynamic factor analyses (lavaan package in R). Missing data (7.1% of total) were imputed using an ARIMA model (ImputeTS package in R). Visual analysis of daily measures was conducted along with Theil-Sen slope estimates. A Bayesian Multilevel Intercept (BMI) model was applied to estimate within-case effect sizes (δ) and Bayes factors. Interpretation of Bayes factors was: 1 = no evidence; 1–3 = anecdotal; 3–10 = substantial/moderate; 10–30 = strong; 30–100 = very strong; >100 = extreme. Gradual log response ratio (LRRg), a percentage change effect size robust to autocorrelation, was calculated. Finally, design-comparable standardized mean differences (d) were computed using the scdhlm package in R to estimate between-case effects.

Results

Factorial invariance

Hierarchical models for PHQ-4 and unidimensional models for RNTQ-3 and VQ-3 converged satisfactorily across all participants, except for PHQ-4 in P6 due to a floor effect (values clustered at the minimum). Fit indices were adequate across all measures, confirming their validity as outcome instruments over time.

Within-participant effects

Regarding emotional symptoms (PHQ-4), 7 of 11 participants (P1, P3, P4, P5, P7, P9, P10) demonstrated Bayes factors > 3, indicating substantial evidence of an intervention effect. Two participants (P2, P11) showed BF approximately 1 (no evidence), while P12 showed a negative effect. One participant (P6) was excluded from PHQ-4 analyses due to the floor effect.

For RNT (RNTQ-3), 8 of 11 participants (P1, P3, P4, P5, P7, P8, P9, P10) showed BF > 3. For valued living (VQ-3), 5 of 12 participants (P3, P6, P7, P8, P12) demonstrated BF > 3.

Between-participant effects

Using the BMI model, Bayes factors were extreme for PHQ-4 when ≥8 participants showed change (BF = 167), strong for ≥9 participants (BF = 17.5), and moderate for ≥10 participants (BF = 3.62). For RNTQ-3, they were extreme when ≥9 participants changed (BF = 167) and strong for ≥10 (BF = 17.2). For VQ-3, they were very strong for ≥7 participants (BF = 29.4) and moderate for ≥8 (BF = 5.92).

Design-comparable effect sizes

Standardized effect sizes were: PHQ-4 d = 0.83 (SE = 0.27, 95% CI [0.26, 1.40]), RNTQ-3 d = 0.81 (SE = 0.23, 95% CI [0.34, 1.28]), and VQ-3 d = 1.07 (SE = 0.42, 95% CI [0.22, 1.91]). When analyzing only participants with baseline clinical levels on PHQ-4, effect sizes were larger: PHQ-4 d = 1.44, RNTQ-3 d = 1.76, VQ-3 d = 1.74.

Follow-up

At the 4-week follow-up assessment after completion of the second session, 8 participants showed normal PHQ-4 scores, 3 showed mild symptoms, and 1 showed moderate symptoms. Four of the 5 participants with baseline clinical levels showed scores below the clinical cutoff at follow-up. Overall baseline mean was 4.60 (SD = 3.29) on PHQ-4, reduced to 2.60 (SD = 1.81) at the 4-week follow-up.

Qualitative evaluation

All participants rated the intervention as useful and indicated they would recommend it to other parents. Most participants specifically recalled the balloon exercise (metaphor), and valued the emphasis on valued actions.

Discussion and conclusions

The authors conclude that the brief RNT-focused ACT protocol demonstrated feasibility, acceptability, and produced clinically significant improvements in emotional symptoms, RNT, and valued living in parents of children with cancer at psychosocial risk. The 100% retention rate (zero dropouts after intervention initiation) suggests that the brief online format is well-accepted by this clinical population. The large effect sizes found (particularly in participants with baseline clinical levels) are notable given that these parents experience continuous and persistent stressors (ongoing oncological treatment of their children).

The authors acknowledge several limitations: (1) the absence of an active control group (a common limitation in SCEDs) prevents ruling out nonspecific effects; (2) follow-up is relatively brief; (3) a single therapist (the first author) implemented all interventions, which could confound therapist effects with protocol effects; (4) the nonconcurrent design is less robust than a concurrent design; (5) the sample is homogeneous in several respects (primary caregivers, predominantly female, predominantly leukemia); and (6) the exclusion of P6 due to floor effect reduces the analytical sample size.

The authors suggest that the ACT model appears particularly well-suited for parents experiencing emotional difficulties linked to pediatric cancer, as it contextualizes distress as a natural and realistic response to extreme stress while providing tools to live more flexibly and in line with what matters despite that distress.

Significance and contribution

This study contributes to the literature on psychological interventions for parents of children with cancer, demonstrating that a brief acceptance and commitment therapy protocol can produce clinically meaningful changes even in parents experiencing continuous and significant stress. The research provides evidence that two-session interventions delivered online can generate large effect sizes in parents at moderate and high psychosocial risk. The findings expand the literature on efficient psychological interventions for contexts of severe pediatric health crises, where parents experience both emotional difficulties and challenges in maintaining consistent support for their seriously ill child.