Early intervention for inhibited young children: a randomized controlled trial comparing the Turtle Program and Cool Little Kids

Abstract

Background:

Children classified as behaviorally inhibited (BI) are at risk for social anxiety. Risk for anxiety is moderated by both parental behavior and social–emotional competence. Grounded in developmental–transactional theory, the Turtle Program involves both parent and child treatment components delivered within the peer context. Our pilot work demonstrated beneficial effects of the Turtle Program (‘Turtle’) over a waitlist control group. Herein, we report results of a rigorous randomized controlled trial (RCT) comparing Turtle to the best available treatment for young children high in BI, Cool Little Kids (CLK).

Methods:

One hundred and fifty-one parents and their 3.5- to 5-year-old children selected on the basis of BI were randomly assigned to Turtle or CLK, delivered in group format over 8 weeks. Effects on child anxiety, life interference, BI, and observed parenting were examined at post-treatment and 1-year follow-up. ClinicalTrials.gov registration: NCT02308826.

Results:

No significant main effect differences were found between Turtle and CLK on child anxiety; children in both programs evidenced significant improvements in BI, anxiety severity, family accommodation, and child impairment. However, Turtle yielded increased observed warm/engaged parenting and decreased observed negative control, compared with CLK. Parental social anxiety moderated effects; parents with higher anxiety demonstrated diminished improvements in child impairment, and parent accommodation in CLK, but not in Turtle. Children of parents with higher anxiety demonstrated more improvements in child BI in Turtle, but not in CLK.

Conclusions:

Turtle and CLK are both effective early interventions for young children with BI. Turtle is more effective in improving parenting behaviors associated with the development and maintenance of child anxiety. Turtle also proved to be more effective than CLK for parents with social anxiety. Results suggest that Turtle should be recommended when parents have social anxiety; however, in the absence of parent anxiety, CLK may offer a more efficient treatment model.

Introduction

Anxiety is among the most common and earliest-emerging disorders, with 50% of cases appearing before age 6 (Dougherty et al., 2013). Anxiety is both persistent and predictive of the developmental unfolding of serious comorbidity including depression and substance abuse (Frenkel et al., 2015), with early onset typically indicating a more chronic and severe course (Ramsawh, Weisberg, Dyck, Stout, & Keller, 2011). Early identification and intervention are, therefore, of paramount importance.

Behavioral inhibition (BI) is a dispositional trait that is reliably identified as early as 4 months which, when longitudinally observed across preschool and early childhood, predicts the development of later anxiety disorders, particularly social anxiety (Chronis-Tuscano et al., 2009). Children high in BI withdraw in the face of novelty, including in social situations across toddlerhood, preschool, early childhood, and beyond. Their well-intentioned parents become increasingly overly protective and accommodate avoidance of social situations. Such parental behavior maintains and exacerbates child avoidance and anxiety in the long-term (Hastings, Rubin, Smith, & Wagner, 2019).

Despite developmental–transactional models identifying children’s interactions with parents and peers as key factors predicting persistence of BI and moderating risk for anxiety (Hastings et al., 2019; Rubin, Barstead, Smith, & Bowker, 2018), extant early interventions for young BI children have largely targeted either parenting or peer interactions (Chronis-Tuscano, Danko, Rubin, Coplan, & Novick, 2018). To address this gap, the Turtle Program was developed as a multi-component early intervention drawing from both parent–child interaction therapy for anxiety (PCIT; Choate, Pincus, Eyberg, & Barlow, 2005) and social skills facilitated play (SSFP; Coplan, Schneider, Matheson, & Graham, 2010). Turtle involves in vivo coaching of parents within the child’s peer group context, as well as in vivo social skills training to encourage approach behaviors. In a pilot randomized controlled trial (RCT), Turtle resulted in significant effects on maternal-reported child anxiety symptoms, BI as reported by parents, teacher-rated school anxiety symptoms, and observed maternal positive affect/sensitivity relative to a waitlist condition (Chronis-Tuscano et al., 2015). Importantly, unlike other programs for children high in BI that rely largely on diagnostic and parent-report data, we demonstrated cross-context effects of Turtle on teacher reports, classroom observations, and laboratory parent–child interactions (Barstead et al., 2018).

Here, we present results of a large-scale RCT comparing Turtle to Cool Little Kids (CLK), a parent-only group treatment that provides psychoeducation about anxiety development and strategies targeting parental anxiety and accommodation. CLK has demonstrated long-term effects on young children classified as BI, including fewer anxiety symptoms and disorders at 3-year follow-up compared with a no-treatment control group (Rapee, Kennedy, Ingram, Edwards, & Sweeney, 2010), making it a stringent comparison condition.

A primary aim was to compare Turtle and CLK on parent questionnaires, diagnostic interviews, and observed parenting, with the primary outcome measure being child anxiety symptoms (NCT02308826). Second, we aimed to identify for whom the more intensive Turtle Program was indicated, to guide treatment personalization in real-world practice. We examined parent social anxiety as the putative moderator given research showing a high prevalence of anxiety disorders in parents of children with BI and the greater tendency among anxious parents to exhibit overcontrol and accommodate child avoidance of social situations (Murray, Creswell, & Cooper, 2009). We hypothesized that Turtle would be superior to CLK for parents with social anxiety.

Methods

Participants

Participants were recruited from preschool/pre-kindergarten programs, childcare centers, pediatricians, and community organizations in the Washington, DC metropolitan area. For inclusion, children were required to: (a) be between ages 45–64 months and attending school, (b) score within the top 15% on the parent-rated Behavioral Inhibition Questionnaire (BIQ; Bishop, Spence, & McDonald, 2003), based on conceptual models of BI (Kagan et al., 2007), (c) never have been diagnosed with an autism spectrum disorder (ASD), or be below the clinical cutoff (≤15) on the Social Communication Questionnaire (SCQ; Eaves, Wingert, Ho, & Mickelson, 2006) screener for ASD, (d) not be receiving anxiety treatment, and (e) not have selective mutism.

Procedures

Interested parents completed a telephone screen assessing basic inclusion criteria. At baseline and post-treatment, qualifying parent–child dyads attended a laboratory visit, where a diagnostic interview and observations of parent–child interaction were completed. At 1-year follow-up, the child diagnostic interview was completed via phone. Questionnaires were completed online at all three time points. Families were paid $50 for post-treatment and $75 for 1-year follow-up assessments. Study materials and procedures were approved by the University of Maryland institutional review board. Written informed consent was obtained from all parents. One hundred fifty-one families were block randomized by study staff based on child sex and anxiety disorder status (Table 1). This sample size is grounded in our previous experience with this population, assumption of a moderate effect size, and expected dropout during longitudinal follow-up. The parent who reported spending more time in the caregiver role and/or who agreed to attend all treatment sessions and laboratory assessments in each family was designated as ‘primary’. Co-parents were invited to attend treatment, but their attendance was not required. Data were collected between 2015 and 2020.

Table 1.

Primary parent and child characteristics at baseline assessment

	Turtle	CLK	t
Primary parent (N = 151)
Age in years, M (SD)	38 (4.4)	39.4 (5.7)
Sex (% female) Parent race, (%)	88	83
White	69	61
Asian	21	16
Black	7	20
Other	3	3
Hispanic or Latinx, (%)	7	7
Parent education, (%)
3 years of college or less	9	12
4 years of college (bachelor’s)	24	24
Master’s degree or equivalent	48	36
Doctoral degree or equivalent	19	28
Median household income		$150,000+
Social anxiety disorder, current (%)	4	10
Child (N = 151)
Age in months, M (SD)	53.2 (5.5)	52.7 (5.9)
Sex (% female)	56	46
Child race, (%)
White	58	43
Asian	19	9
Black	7	18
Other	16	30
Hispanic or Latinx, (%)	7	11
Any anxiety disorder (%)	64	64
Social anxiety disorder (%)	60	56
Outcome measures
Total Anxiety Symptom Severity, M (SD)	6.52 (2.89)	7.08 (3.88)	1.00
Social Anxiety Symptom Severity, M (SD)	3.40 (1.32)	3.49 (1.44)	0.41
Behavioral Inhibition Questionnaire (BIQ), M (SD)	5.17 (0.64)	5.10 (0.69)	−0.57
PP Current Social Anxiety Symptom Severity, M (SD)	0.92 (1.14)	1.06 (1.54)	0.62
Family Accommodation Scale (FAS), M (SD)	2.12 (0.68)	2.19 (0.76)	0.65
Child Anxiety Life Interference Scale (CALIS)	2.34 (0.61)	2.40 (0.71)	0.50
Engagement (FP)	3.06 (0.96)	2.87 (0.94)	−1.15
Negative control (FP)	2.42 (0.88)	2.69 (0.95)	1.78
Positive affect (FP)	2.50 (0.90)	2.23 (0.95)	−1.78
Engagement (Lego)	3.14 (0.89)	3.14 (1.08)	0.02
Negative control (Lego)	2.67 (1.07)	2.83 (1.14)	0.87
Positive affect (Lego)	2.38 (0.91)	2.27 (0.88)	−0.69

Turtle, Turtle Program; CLK, Cool Little Kids; PP, primary parent; FP, free play; Lego, Lego task.

^**p < .001;

^*p < .05.

Measures

Child BI/Anxiety.

Behavioral inhibition:

The Behavioral Inhibition Questionnaire (BIQ; Bishop et al., 2003) was administered for inclusion, and at each timepoint. The BIQ is a 30-item measure comprising six subscales measuring BI in response to adults, peers, performance demands, novel settings, physical challenges, and parental separation. The BIQ possesses strong internal consistency, convergent validity with traits related to BI, and divergent validity with conceptually distinct traits. Cronbach alpha (a) in this study ranged from 0.87 to 0.94.

Child diagnostic assessment:

The Anxiety Disorders Interview Schedule for Children for DSM-5 – Child and Parent Version (ADIS-V-CP; Silverman, Albano, & Barlow, 1996; Silverman& Albano, 2020) is a semi-structured interview administered by independent evaluators masked to treatment randomization at pre-treatment, post-treatment, and 1-year follow-up to assess total anxiety and social anxiety severity based on DSM-5 criteria. Impairment/distress was measured using the clinician-generated Clinician Severity Rating (CSR), which ranges from 0 to 8 (≥4 indicating clinically significant impairment). Total anxiety severity was calculated as the sum of separation, specific, social, and generalized anxiety module CSRs (Ginsburg, Drake, Tein, Teetsel, & Riddle, 2015). The ADIS-IV-CP has demonstrated good reliability in preschoolers (Kennedy, Rapee, & Edwards, 2009). Intraclass correlations (ICCs) ranged from .78 to .87 for total anxiety and from .77 to .93 for social anxiety.

Child impairment:

The Child Anxiety Life Interference Scale – Preschool Version (CALIS-PV; Kennedy et al., 2009) is a 24-item parent-report measure of the impact of a child’s anxiety on the child and parents’ lives. The CALIS-PV comprises three factors: interference at home, outside the home, and on parent life. The CALIS-PV has demonstrated strong internal consistency. Cronbach alpha (a) for this study ranged from 0.90 to 0.91.

Parent anxiety.

Parent diagnostic assessment:

Current parent social anxiety severity was assessed at baseline only using the ADIS-V – Adult & Lifetime Version (ADIS-V-L; Brown & Barlow, 2014). The interview was administered via telephone by independent evaluators blind to treatment randomization. As in the ADIS-V-CP, the ADIS-V-L assesses the degree of distress and impairment associated with each disorder using the clinician-generated CSR. The ADIS-IV-L has shown strong psychometrics. Twenty percent of interviews were coded for inter-rater reliability; ICC = .93.

Parenting.

Family accommodation of child anxiety:

The Family Accommodation Scale (FAS; Lebowitz et al., 2013) is a 13-item parent report that measures the way family members behave to relieve child distress caused by symptoms of anxiety over the past month. The FAS includes four subscales: parent participation in anxiety symptoms, modification of family routines, distress resulting from accommodation, and undesirable consequences of not accommodating. A high degree of internal consistency among subscales has been reported (Lebowitz et al., 2013). Cronbach alpha (a) in this study ranged from 0.88 to.90.

Observed parenting:

Parent–child dyads were observed in two situations at pre- and post-treatment: (a) Free Play and (b) Lego Task (parent asked to guide child to create a Lego structure). Parental behavior was indexed by blinded raters with an adaptation of the Maternal Warmth and Control Scale (Rubin, Cheah, Smith, & Wagner, 2016). For each task, parent behavior was assigned a global score on a 1 (low) to 5 (high) scale for each of three categories: (a) Engagement, (b) Positive affect, and (c) Negative control. Coders established reliability on 22% of Free Play and 25% of Lego cases, attaining ICCs of .75 or better with a master coder for each category. Additional detail about these tasks and codes is provided in Appendix S1.

Interventions.

Turtle program:

Turtle consists of eight weekly, concurrent parent, and child group treatment sessions, each lasting 1.5 hr (Danko, O’Brien, Rubin, & Chronis-Tuscano, 2018). In the parent component, sessions were divided into three phases: child-directed interaction (CDI), bravery-directed interaction (BDI; Choate, Pincus, Eyberg, & Barlow, 2005), and Parent-Directed Interaction (PDI). In CDI, parents learned skills to strengthen the parent–child relationship (e.g., following the child’s lead, positive reinforcement) and received live, in vivo coaching. In the BDI phase, parents learned to model approach behavior, prompt and reinforce targeted social behavior, and conduct graduated exposures to aid children’s adjustment to social situations. Skills were promoted through direct instruction, role-play, at-home practice, and in vivo coaching sessions within the peer group context. Parents practiced skills with their children while the parent group observed live through a television in the parent group room. Depending on each child’s individualized bravery goal, in vivo coaching with the parent and child may have involved interactions with another child (from the SSFP group) or adult. Parents were also coached in the peer group context at drop-off and during the graduation session. In PDI, parents were taught to differentiate anxious versus oppositional behavior and child behavior management strategies. Treatment integrity checks demonstrated 98.99% adherence.

Concurrently, children received modified SSFP (Coplan et al., 2010). SSFP employs training and modeling in social skills, social problem-solving, emotion regulation, and relaxation. The SSFP didactic portion lasted 10 min, with sensitivity to the attention span of young children, followed by free play and structured activities, during which leaders systematically encouraged children to approach social interactions. Turtle Program therapists were trained by licensed clinical psychologists with experience implementing the Turtle Program and PCIT; weekly supervision was provided. Treatment integrity checks for SSFP demonstrated 91.40% adherence.

Cool Little Kids:

Cool Little Kids (Rapee, Kennedy, Ingram, Edwards, & Sweeney, 2005) is a 6-week parent psychoeducation group, delivered by the same clinicians as the Turtle parent component to control for therapist effects. As in Turtle, five to seven families participated in each cohort, and each session lasted 120 min, to control for therapist contact time. Sessions 1–4 were held weekly, with a 1-month break between Sessions 5 and 6. Session 1 included a discussion of the nature and development of anxiety. Session 2 included basic parent-management techniques and the role of parent overprotection in maintaining anxiety. Sessions 3 to 5 involved creating and implementing exposure hierarchies, and the application of cognitive restructuring to parents’ own worries. Session 6 included continued application of techniques, with an emphasis on developmental transitions. Parent group therapists received didactic training and weekly supervision on CLK. The developer of CLK provided fidelity ratings and feedback on a subset of sessions. Treatment integrity checks demonstrated 97.57% adherence.

Analytic plan

Patterns of missingness were examined prior to running analyses. Missingness in parent-report measures ranged from 2.65% to 5.30% at baseline and 15.89% to 21.85% at post-treatment and follow-up. Missingness on the ADIS-V-P ranged from 0.66% to 1.99% at baseline, and 11.92% to 16.56% at post-treatment and follow-up. Missingness in observed parenting ranged from 5.30% to 13.25% at pre- and post-treatment.

Paired samples t-tests were used to examine change from baseline to post-treatment and post-treatment to follow-up.

Main hypotheses were tested using structural equation modeling (SEM) through the lavaan package in R. The direct effect of condition on post-treatment and follow-up measures was examined. For each model, pre-treatment values were included as covariates.

Moderation of the association between parent anxiety and child outcomes was examined using multigroup analysis. Chi-square difference (Δχ²) tests were used to compare models with regression coefficients free to vary across groups with models with regression coefficients constrained to be equal across groups. A significant chi-square difference test indicated that constraining regression coefficients to be equal across groups would significantly worsen model fit, suggesting moderation by treatment condition. Robust full-information maximum likelihood was used to accommodate missing data (Enders, 2001). The comparative fit index (CFI), root mean square error of approximation (RMSEA), and chi-square statistic were used to assess model fit. A CFI value of 0.95 and RMSEA cutoff of 0.06 indicate good fit, respectively (Hu & Bentler, 1999).

Results

Preliminary analyses

Results of preliminary analyses demonstrated no significant differences in clinical and demographic variables between conditions (Table 1). Given the intent-to-treat approach, all randomized families were included in analyses (Figure 1 CONSORT). Of 151 families randomized, 16 (10.6% of total sample; seven from Turtle and nine from CLK) dropped out. Primary parents who dropped out were more likely to have a household income of less than $150,000/year (χ² = 5.05, p = .02), be Black/African American (χ² = 5.05, p = .02), and have less than a bachelor’s degree (χ² = 24.63, p < .001). Co-parents attended 32% of CLK sessions and 51% of Turtle sessions (see Novick et al., 2020). Tests of mean change are presented in Table S1.

Figure 1.

Consort diagram. CLK, Cool Little Kids; Turtle, The Turtle Program; BIQ, Behavioral Inhibition Questionnaire; SCQ, Social Communication Questionnaire; ASD, autism spectrum disorder; Tx, treatment; BI, behaviorally inhibited. Five families who did not complete post-assessment completed one-year follow-up assessment in CLK. Two families who did not complete post-assessment completed one-year follow-up assessments in Turtle

Main effects of treatment

A summary of saturated path models examining treatment effects on child anxiety, parent accommodation, and associated impairment can be found in Table 2. Main effects of treatment condition on post-treatment and follow-up outcome diagnostic interviews and questionnaires were not detected; however, treatment group differences were found on observed parenting (Table 3). Parents in Turtle displayed significantly more engagement and positive affect, and less negative control during Free Play at post-treatment, relative to parents in CLK. During the Lego Task, parents in Turtle displayed significantly more engagement and positive affect directed to their children at post-treatment, relative to parents in CLK.

Table 2.

Findings from separate path models regressing study outcomes on treatment group and covariates

	Total anxiety symptom severity		Social anxiety symptom severity		Behavioral Inhibition Questionnaire (BIQ)		Family Accommodation Scale (FAS)		Child Anxiety Life Interference Scale (CALIS)
	Post-treatment, B (β)	Follow-up, B (β)	Post-treatment, B (β)	Follow-up, B (β)	Post-treatment, B (β)	Follow-up, B (β)	Post-treatment, B (β)	Follow-up, B (β)	Post-treatment, B (β)	Follow-up, B (β)
PP Black	−0.18 (−0.02)	0.87 (0.12)	−0.8 (−0.02)	0.39 (0.11)	−0.22 (−0.11)	0.07 (0.03)	−0.09 (−0.05)	−0.14 (−0.08)	−0.01 (−0.003)	0.11 (0.07)
PP Asian and other	−0.52 (−0.08)	0.09 (0.02)	−0.08 (−0.03)	−0.29 (0.09)	−0.1 (−0.06)	−0.21 (−0.10)	−0.15 (−0.10)	−0.23 (−0.16)*	−0.05 (−0.04)	−0.08 (−0.06)
PP less than bachelor	−0.68 (−0.07)	1.68 (0.19)	−0.09 (−0.02)	0.76 (0.17)	0.46 (0.19)*	−0.014 (−0.004)	0.31 (0.14)	−0.06 (−0.03)	0.03 (0.02)	0.08 (0.04)
PP graduate degree	0.50 (0.09)	−0.41 (−0.08)	0.33 (0.12)	−0.11 (−0.04)	0.11 (0.08)	0.16 (0.08)	−0.001 (−0.001)	0.09 (0.07)	0.03 (0.03)	0.13 (0.11)
PP income	−0.09 (−0.02)	−0.58 (−0.12)	0.31 (0.12)	−0.15 (0.06)	−0.05 (−0.04)	−0.14 (−0.07)	0.03 (0.03)	−0.13 (−0.11)	−0.09 (−0.08)	−0.18 (−0.15)
T1 variable	0.54 (0.68)**	0.25 (0.34)**	0.62 (0.65)**	0.23 (0.25)**	0.59 (0.56)**	0.62 (0.45)**	0.46 (0.52)**	0.45 (0.54)**	0.59 (0.70)**	0.37 (0.46)**
Treatment group	0.05 (0.01)	0.02 (0.04)	−0.01 (0.002)	0.03 (0.01)	0.03 (0.03)	0.18 (0.10)	0.01 (0.01)	−0.01 (−0.01)	−0.06 (−0.05)	0.04 (0.03)

Turtle Program = reference group (coded as 1); PP, primary parent.

^*p < .05;

^**p < .001.

Table 3.

Findings from separate path models regressing observed parenting outcomes on treatment group and covariates

	Free play task			Lego task
	Engagement,B (β)	Negative control, B (β)	Positive affect, B (β)	Engagement, B (β)	Negative control, B (β)	Positive affect, B (β)
PP Black	0.10 (0.03)	0.10 (0.03)	0.25 (0.07)	−0.51 (−0.15)*	−0.04 (−0.01)	−0.17 (−0.05)
PP Asian and other	0.04 (0.02)	0.35 (0.14)*	−0.01 (−0.002)	−0.14 (−0.05)	0.24 (0.09)	−0.27 (−0.09)
PP less than bachelors	−0.51 (−0.12)	0.04 (0.01)	−0.56 (−0.13)	0.05 (0.01)	0.45 (0.12)	0.18 (0.04)
PP graduate degree	−0.34 (−0.13)	0.11 (0.05)	0.07 (0.03)	−0.33 (−0.14)	0.10 (0.04)	−0.06 (−0.02)
Income	0.11 (0.05)	−0.06 (−0.03)	0.20 (0.08)	0.02 (0.01)	0.17 (0.08)	0.03 (0.01)
T1 variable	0.53 (0.43)**	0.46 (0.41)**	0.35 (0.26)**	0.51 (0.44)**	0.46 (0.48)**	0.49 (0.35)**
Treatment group	0.82 (0.35)**	−0.84 (−0.40)**	1.43 (0.57)**	0.62 (0.27)**	−0.29 (−0.13)	1.16 (0.47)**

Turtle Program = reference group (coded as 1), PP, primary parent.

^*p < .05;

^**p <.001.

Moderation of treatment effects by parent social anxiety

A summary of multigroup analyses can be found in Table 4.

Table 4.

Results of multigroup models examining moderating role of treatment group

	Post-treatment BIQ		Follow-up CALIS		Follow-up FAS
	b (SE)	β	b (SE)	B	b (SE)	β
Turtle
PP social anxiety severity	−0.12 (0.05)^,*	−0.19^,*	−0.05 (0.06)^	−0.11^	−0.02 (0.05)^	−0.04^
T1 variable	0.60 (0.09)**	0.57**	0.40 (0.08)**	0.45**	0.46 (0.07)	0.52
PP less than bachelors	0.50 (0.18)*	0.20*	0.18 (0.16)	0.09	−0.002 (0.22)	−0.001
PP graduate degree	0.10 (0.13)	0.07	0.13 (0.10)	0.12	0.05 (0.12)	0.04
Income	0.003 (0.10)	0.002	−0.13 (0.10)	−0.13	−0.07 (0.10)	−0.06
CLK
PP social anxiety severity	0.05 (0.04)^	0.11^	0.14 (0.07)^,*	0.36^,*	0.12 (0.05)^,*	0.31^,*
T1 variable	0.60 (0.09)**	0.59**	0.40 (0.08)**	0.45**	0.46 (0.07)**	0.56**
PP less than bachelors	0.50 (0.18)*	0.21*	0.18 (0.16)	0.08	−0.002 (0.22)	−0.001
PP graduate degree	0.10 (0.13)	0.07	0.13 (0.10)	0.10	0.05 (0.12)	0.04
Income	0.003 (0.10)	0.002	−0.13 (0.10)	−0.11	−0.07 (0.10)	−0.06
Model fit
χ2	χ2(4) = 2.82, p = .59		χ2(4) = 2.96, p = .57		χ2(4) = 2.17, p = .70
CFI	1.00		1.00		1.00
RMSEA	0.00		0.00		0.00

Turtle, Turtle Program; CLK, Cool Little Kids; PP, primary parent; BIQ, Behavioral Inhibition Questionnaire; FAS, Family Accommodation Scale; CALIS, Child Anxiety Life Interference Scale.

^{^}Signifies paths significantly vary between groups based on ∆χ² difference test.

^*p < .05;

^**p < .001.

Post-treatment behavioral inhibition.

Based on Δχ² tests, we constrained paths for parent education, household income, and pre-treatment BIQ to be equal across groups (all p-values > .05). Δχ² tests revealed that the parent social anxiety severity path should remain free across groups, suggesting moderation by treatment (Δχ²(1) = 225.70, p < .001). The final model, in which only parent social anxiety severity was allowed to differ between conditions, fits the data well (χ²(4) = 2.82, p = .59; CFI = 1.00; RMSEA = 0.00). Parent anxiety severity predicted lower post-treatment BIQ only in Turtle. This association was not found in CLK. Parents with greater social anxiety severity reported lower child BI at post-treatment in Turtle, whereas there was no relation between parent social anxiety and child post-treatment BI in CLK.

Follow-up life interference.

Based on Δχ² tests, we constrained paths for parent education and pre-treatment CALIS to be equal across groups (all p-values > .05). Δχ² tests revealed that the parent social anxiety severity path should remain free across groups, suggesting moderation by treatment (Δχ²(1) = 12.836, p < .001). The final model, in which only parent social anxiety was allowed to differ between treatment groups, fits the data well (χ²(4) = 2.96, p = .57; CFI = 1.00; RMSEA = 0.00). Parent social anxiety severity predicted greater follow-up impairment in CLK. This association was not found in Turtle. That is, parents with more severe social anxiety reported higher child impairment following CLK, whereas there was no relation between parent social anxiety severity and post-treatment child impairment in Turtle.

Follow-up parent accommodation.

Based on Δχ² tests, we constrained paths for parent education and pre-treatment FAS to be equal across groups (all p-values > .05). Δχ² tests revealed that the parent anxiety path should remain free across groups, suggesting moderation by treatment (Δχ²(1) = 4.24, p = .04). The final model, in which only parent social anxiety was allowed to differ between treatment groups, fits the data well (χ²(4) = 2.17, p = .70; CFI = 1.00; RMSEA = 0.00). Parent social anxiety only predicted greater follow-up family accommodation (FAS) in CLK, but not Turtle. That is, parents with greater social anxiety severity reported continued accommodation of child anxiety at post-treatment in CLK, whereas there was no relation between parent social anxiety and parent accommodation in Turtle.

Other outcomes showed no evidence of moderation.

Discussion

This RCT compared the Turtle Program with the best-established treatment for inhibited young children, CLK. Children in both Turtle and CLK demonstrated significant improvements in anxiety severity, BI, family accommodation, and child impairment from pre- to post-treatment and post-treatment to follow-up, with no significant between-group differences. Observed parent–child interactions yielded significant main effects favoring Turtle, such that parents receiving Turtle demonstrated significant increases in both observed engagement and positive affect directed to their children, and significant decreases in negative control from pre- to post-treatment; parents in CLK demonstrated significant increases in observed positive affect from pre- to post-treatment during Free Play, but not in the other observational categories. In fact, parental engagement during the Lego Task worsened following CLK.

Cool Little Kids is a stringent comparison condition that has demonstrated lasting effects in multiple RCTs (Kennedy et al., 2009; Rapee et al., 2010); thus, it is unsurprising that children of parents receiving CLK showed improvements on many outcomes. Of note, CLK sessions were lengthened from 90 to 120 min in this study to match therapist contact hours across conditions. Although the added time did not involve the provision of additional content, it is possible that this may have affected outcomes.

Further, given the intensity of the PCIT-inspired Turtle Program parent component, we expected that Turtle would outperform CLK on objective parenting observations. It is particularly notable that effects on parental overcontrol were most pronounced during Free Play, a situation that should not elicit parental control. Given that parent overcontrol predicts the stability of BI (Rubin, Burgess, & Hastings, 2002) and moderates risk for anxiety among inhibited children (Lewis-Morrarty et al., 2012), this finding suggests that Turtle may interrupt the developmental cascade to anxiety and co-occurring psychopathology for young children characterized as high BI.

A major aim was to examine the question, ‘For whom is the more intensive Turtle Program warranted?’ Here, we found that treatment condition moderated effects of parent anxiety on BI, life interference, and family accommodation. Parent anxiety was related to poorer treatment outcomes in terms of child anxiety life interference and parental accommodation in CLK but not in Turtle, suggesting that parent anxiety did not hinder treatment effectiveness in Turtle. It may be that Turtle’s in vivo parent coaching component, or the child group, or both reduced negative treatment outcomes that may have resulted from parent anxiety. More specifically, having the opportunity to practice skills in session with therapist support and/or the direct provision of skills to children themselves may have been more necessary for families in which parents have social anxiety.

It was interesting that children of parents with higher anxiety demonstrated greater improvements in child BI in Turtle, but not in CLK. Here again, it may be that anxious parents and their children especially thrived with the in vivo coaching and direct child component.

This study is not without limitations. Both interventions were guided by a similar theoretical model regarding the role of parents in accommodating child inhibited/anxious behavior. The study design did not allow us to disentangle effects that could be attributed to Turtle’s child group versus in vivo parent coaching. For instance, it might be the case that anxious parents required in vivo coaching in addition to psychoeducation, or it may be that direct child intervention was required in addition to the parent component when parents were anxious, or both.

We did not include a no-treatment control group in this trial. According to Chambless and Hollon (1998), comparison to the ‘already-established treatment’ is the most rigorous approach when evaluating the efficacy of psychological interventions. Prior studies of CLK (Rapee et al., 2010) and Turtle (Chronis-Tuscano et al., 2015) established clear superiority over a no-treatment condition. Given that an effective treatment for BI exists, we did not include a no-treatment control group for ethical reasons.

Future research should dismantle components of Turtle to elucidate effects attributable to the child group or PCIT coaching. Additional moderators beyond parent anxiety should be examined, including child and parent biopsychosocial markers (e.g., RSA). And, examination of mediators should be conducted to determine how these treatments work. Generalization of treatment effects to school should be examined using both observational and teacher report data. Finally, implementation of Turtle in community settings should be a priority, given our sample characteristics and dropout among socially disadvantaged parents.

Despite noted limitations, this study was conducted consistent with rigorous standards for child anxiety treatment trials (Creswell et al., 2021), such as use of multiple informants and methods, developmentally sensitive assessment, assessment of broad and specific anxiety symptoms, emphasis on impairment/life interference, and presentation of participant demographic characteristics, treatment dosage, and parent involvement. Together, these findings indicate that for many children classified as BI, Turtle and CLK both yield beneficial effects, with the exception of observed parenting for which Turtle was superior. Parent anxiety has been implicated in the development and maintenance of child BI and anxiety, as well as the tendency to accommodate child anxiety (Murray et al., 2009). The results of this study are prescriptive in that they suggest that Turtle should be recommended over CLK when parents currently struggle with social anxiety. In the absence of parent anxiety, CLK may offer a more efficient treatment model that can operate with fewer resources such as clinician time, space, and specialized equipment.

Supplementary Material

1

Appendix S1. Parent–child interactions and preliminary analyses.

Table S1. Improvements from pre- to post-treatment and post-treatment to follow-up (t-tests).

Key points.

• Parenting and peer interactions maintain and exacerbate early childhood behavioral inhibition (BI) and anxiety; thus, multi-component treatments like the Turtle Program (‘Turtle’) may interrupt progression to anxiety in at-risk children.

• This randomized controlled trial compared Turtle and Cool Little Kids (CLK) at post-treatment and 1-year follow-up.

• Turtle and CLK had comparable effects on child anxiety, accommodation, and anxiety-related impairment; Turtle had superior effects on observed parental behavior associated with the maintenance of child BI.

• Parental anxiety moderated treatment effects; Turtle outperformed CLK on child anxiety life interference, BI, and parental accommodation when parents were anxious.

• Findings inform personalization of early intervention for young children high in BI.