Emotion coaching skills as an augmentation to family-based therapy for adolescents with anorexia nervosa: A pilot effectiveness study with families with high expressed emotion

Abstract

Objective:

To examine the feasibility and acceptability of augmenting family-based treatment (FBT) for adolescents with anorexia nervosa (AN) or atypical anorexia nervosa (AAN) with a parent emotion coaching intervention (EC) focused on reducing parent expressed emotion.

Method:

In this pilot effectiveness trial, families of adolescents with AN/AAN exhibiting high expressed emotion received standard FBT with either 1) EC group or 2) support group (an attention control condition focused on psychoeducation).

Results:

Forty-one adolescents with AN or AAN were recruited (88% female, M_age=14.9±1.6 years, 95% White: Non-Hispanic, 1% White: Hispanic, 1% Bi-racial: Asian). Most study adolescents were diagnosed with AN (59%) while 41% were diagnosed with AAN. Participating parents were predominantly mothers (95%). Recruitment and retention rates were moderately high (76% and 71%, respectively). High acceptability and feasibility ratings were obtained from parents and interventionists with 100% reporting the EC intervention was “beneficial”- “very beneficial.” The FBT+EC group demonstrated higher parental warmth scores at post-treatment compared to the control group (standardized effect size difference, d = 1.58), which was maintained at 3-month follow-up. Finally, at post-treatment, the FBT+EC group demonstrated higher rates of full remission from AN/AAN (40%) compared to FBT+support (27%), and were nine times more likely to be weight restored by 3-month follow-up.

Discussion:

Augmenting FBT with emotion coaching for parents with high expressed emotion is acceptable, feasible, and demonstrates preliminary effectiveness.

Introduction

Family-based treatment (FBT) is considered the first-line, evidence-based treatment for adolescents with anorexia nervosa (AN), yet outcomes remain suboptimal, with an average of 40% of patients achieving full remission by end of treatment (Le Grange et al., 2016; Lock & Le Grange, 2019; Lock et al., 2010; Madden et al., 2015). Furthermore, individuals with atypical anorexia nervosa (AAN) experience comparable physical and psychological impairment to their AN counterparts (Harrop et al., 2021; Sawyer et al., 2016) yet have not been included in prior FBT randomized controlled trials (RCTs), underscoring the need for treatment outcome research including this population. Enhancing FBT through targeting predictors of early non-response to FBT is critical to improve outcomes in adolescent AN/AAN.

In FBT, parents manage their adolescents’ eating to ensure weight restoration to a healthy body weight. Treatment processes can elicit negative emotional interactions between the parent and adolescent and increase caregiver burden and distress (Matthews et al., 2018). In this context, a parent’s attitudes and behaviors toward the patient -- or expressed emotion (EE) -- is a robust predictor of FBT response. Specifically, critical comments and parental warmth, two domains of EE, are associated with treatment outcomes (Bohon et al., 2023; Le Grange et al., 2011; Rienecke et al., 2016; Szmukler et al., 1985). High critical comments predict FBT drop-out and non-response, explaining up to 34% of the variance in weight restoration (Allan et al., 2018; Darcy et al., 2013; Kyriacou et al., 2008; Lock et al., 2006; Rienecke et al., 2016; Rienecke et al., 2017; Szmukler et al., 1985; van Furth et al., 1996). Conversely, high parental warmth predicts improved psychological functioning and weight restoration (Le Grange et al., 2011; Rienecke, 2017). As such, EE shows significant promise as a mechanism to enhance FBT outcomes (Allan et al., 2018; Le Grange et al., 2011).

Parent emotion coaching (EC) interventions target EE by teaching parents strategies to supportively respond to their child’s emotions and to reduce unhelpful behaviors, like criticism, building on developmental theory of emotion socialization (Eisenberg et al., 1998). Skills promoted in EC include active listening (e.g., reflection), emotional support (e.g., validating emotions), and emotion coaching (e.g., identifying coping strategies). In diverse clinical and community samples, parent EC led to increased use of positive skills (e.g., validation, parental warmth) and reductions in negative skills (e.g., criticism) relative to treatment as usual (Havighurst et al., 2020). Parent EC interventions have been implemented successfully as augmentations to evidence-based behavioral interventions in families of youth with internalizing and externalizing disorders when targeting processes of parent-child emotion communication, and are warranted to optimize treatment outcomes (Chronis-Tuscano et al., 2016; Lenze et al., 2011; Shaffer et al., 2023).

EC interventions are not entirely novel to eating disorder treatment. For example, emotion focused family therapy (EFFT), developed as an adjunct to FBT and piloted in a 2-day parent intervention, was associated with improved parent self-efficacy, and reduced fears about treatment (Lafrance Robinson et al., 2016). However, follow-up data was not collected for EFFT, and treatment generalizability was limited due to a lack of randomization. Parent-focused treatment (PFT; a separated parent-focused format of FBT), is associated with decreased maternal criticism compared to standard FBT (Allan et al., 2018). Notably, observed changes may have occurred spuriously since EE was not specifically targeted in PFT any more than in FBT.

Based on the clinical need among adolescents with AA/AAN, risks associated with high parent EE during FBT, and growing evidence for EC interventions, we designed a strategic augmentation of FBT with EC for high EE parents. This intervention was designed to teach parents EC skills to better support their adolescent with difficult emotions that frequently arise during FBT (Lafrance Robinson et al., 2016; Matthews et al., 2018; Treasure & Nazar, 2016). Initially, an open pilot was conducted to gather parent and interventionist feasibility and acceptability data, which were subsequently used to modify the FBT+EC manual for the pilot RCT (Aarnio-Peterson et al., 2023; Shaffer & Aarnio-Peterson, 2020). The primary aim of this study was to assess feasibility and acceptability of augmenting FBT with a parent EC group (FBT+EC) in high EE families of adolescents with AN/AAN. Secondly, we examined whether FBT+EC reduced criticism, increased warmth, and improved weight restoration and remission rates compared to the control condition.

We predicted high ratings of acceptability and feasibility of FBT+EC by families and interventionists. Further, we anticipated that parents in FBT+EC would have lower criticism and higher warmth scores than parents of adolescents in the control condition at post-treatment and at follow-up. Finally, adolescents in FBT+EC would demonstrate greater weight restoration and higher remission rates compared to participants in the control condition at post-treatment and follow-up.

Method

Participants

Consecutive patients presenting to the eating disorders program during the recruitment period (September 2020-September 2022) were assessed for eligibility (see Figure 1 CONSORT diagram). Inclusion criteria were DSM-5 criteria for AN/AAN, age 12–17, and a parent who lived with the adolescent and screened positive on the Five Minute Speech Sample (FMSS) (Magaña et al., 1986), and adolescent inability to read or speak English fluently. Exclusion criteria were diagnosis of a severe developmental disability (e.g., autism) or medical condition affecting weight (e.g., cystic fibrosis), active suicidality, active psychosis, or previous FBT.

Figure 1.

CONSORT Diagram

Study Design

Methods for this clinical trial (ClinicalTrials.gov Identifier: NCT04421989) have been previously published (Aarnio-Peterson et al., 2023). The study intervention was conducted from June 2020-March 2023 during the COVID-19 pandemic. Thus, all assessments and interventions were conducted via telehealth. Participants included high EE parents and their adolescents with AN/AAN, presenting for FBT in a multidisciplinary eating disorders program within Cincinnati Children’s Hospital Medical Center in the United States. While both parents participated in FBT, only one eligible parent was required to participate in the group components. To qualify as “high EE,” parents completed the FMSS (Magaña et al., 1986) during which they spoke for 5 uninterrupted minutes about their adolescent with AN/AAN. This speech sample was then coded using an empirically valid, standardized coding schema for EE.

Participants were randomized to FBT+EC (treatment) or FBT+support (control). As an effectiveness pilot study, all participants received institutional standard of care treatment (i.e., FBT and medical monitoring by an adolescent medicine physician). FBT is comprised of three phases conducted in weekly 50-minute sessions. Phase I supports parents to manage weight restoration, the second phase transitions management of eating back to the adolescent in a developmentally appropriate manner, and the third phase focuses on typical adolescent development (Lock & Le Grange, 2015). During COVID-19 (2020–2022), many participants received virtual FBT, which has demonstrated feasibility and similar clinical outcomes to standard face-to-face FBT (Anderson et al., 2017; Matheson et al., 2020). Standard care was augmented with additional participation in the parent EC group (FBT+EC condition) or the parent support group (FBT+ support condition). EC and support groups were conducted virtually and led by master’s level clinicians. FBT interventionists received weekly supervision by author DLG to ensure treatment fidelity.

Study assessments, including adolescent diagnostic interviews, the FMSS, and self-report questionnaires, were completed virtually with adolescents and parents at baseline, one month (i.e., halfway through EC or support interventions), end of treatment, and 3 months post-treatment. All assessments were administered by independent, trained assessors who were not involved in treatment delivery.

Treatment Conditions

FBT + EC Condition.

Participants randomized to FBT+EC received standard FBT (Lock & Le Grange, 2015). Parents received 10 weekly, 1-hour group interventions occurring separately, yet concurrently with FBT sessions. FBT clinicians differed from the EC interventionists. Most study participants began EC during phase I of FBT and a minority started EC during FBT Phase II given the timing of enrollment (e.g., were waiting for group to begin). The timing of group start with FBT phase did not vary significantly between EC or support conditions (χ² = 0.37; p = .55). The structure of EC parent group sessions included: review of homework, skills-based didactic, and role plays and live coaching with interventionist feedback. Between sessions, parents were assigned self-reflective homework intended to apply session content to their own goals. Parents were also instructed to practice skills via daily 5- to 10-minute debriefing with their adolescent following challenging or emotion-provoking meals/snacks.

Parents sometimes found EC goals (i.e., emotional support) and FBT goals to be at odds. For example, in FBT+EC parents were coached to normalize and validate their child’s anxiety about consuming a feared food, while also discouraged from accommodating or enabling maladaptive eating behavior (e.g., food refusal). Thus, EC interventionists were aware of the therapeutic goals and modality of the FBT re-nourishment process and provided tailored guidance to maintain effective EC skills while staying on track with FBT. As part of the FBT+EC pilot study, we developed additional content in the EC treatment manual that addressed common emotion-related challenges experienced by families in FBT and provided guidance to EC therapists in navigating those challenges while maintaining fidelity to both FBT and EC.

FBT+ support condition.

FBT+support was the attention control condition in which parents attended 10 weekly, 1-hour support group sessions as opposed to waitlist assignment or usual care. Sessions consisted of psychoeducation on the causes and treatment of AN/AAN and included topics like medical issues associated with AN/AAN, the various levels of care in eating disorder treatment, taking time off from work to support an ill adolescent, self-care, and balancing time for other household demands.

Randomization Method for Pilot Trial

Participants were assigned 1:1 into the FBT+EC or the FBT+support conditions using block randomization using R and the ‘blockrand’ package (Snow, 2013; Team, 2010). Condition assignment occurred according to a 2×2×2 randomization table, stratified on three factors: 1) recruitment upon discharge from a medical hospitalization secondary to AN/AAN versus outpatient setting, 2) adolescent sex, and 3) adolescent diagnosis (AN or AAN). Block sizes of 2 or 4 were randomly chosen to ensure equal allocation to the two conditions across each of the stratum. The randomization list was held by a research staff member independent of the study to reduce any potential biases related to group assignments and was not assigned until all baseline measures were obtained.

Safety.

All participants met regularly with their adolescent medicine physician to ensure medical stability, in line with the institutional standard of care. If a participant became medically unstable (hypothermic [body temperature <36.3°C], bradycardic [heart rate <50 beats/min or QT interval corrected for heart rate >0.45], orthostatic [pulse increase >35, systolic blood pressure decrease > 10 mm Hg], or experienced acute weight loss <75% expected body weight [EBW]), medical hospitalization was required [32, 33]. Further, a data safety monitoring board (DSMB) consisting of three eating disorder experts independent of the study team met twice yearly throughout the RCT to review any adverse events and ensure participant safety.

Study Measures

Feasibility and Acceptability Questionnaire.

The Feasibility and Acceptability Questionnaire was created by the study team and comprises 18 items completed by parents. Questions assess the usefulness of the EC intervention, what the parent learned as a result of the intervention (e.g., “I have become more aware of my own emotions”), and the perceived effects of the intervention on the adolescent’s symptoms and treatment (e.g., “My teen better understands her emotions”).

Acceptability of the Intervention Measure (AIM); Intervention Appropriateness Measure (IAP); Feasibility of the Intervention Measure (FIM)

(Weiner et al., 2017). EC interventionists completed the AIM, IAM, and FIM to assess acceptability, appropriateness, and feasibility, respectively. Higher scores are indicative of higher ratings of intervention acceptability, appropriateness, and feasibility. Acceptable discriminant content validity has been demonstrated for each of the three measures with alpha values ranging from 0.87–0.89 (Weiner et al., 2017).

Eating Disorder Examination (EDE)

(Fairburn & Cooper, 1993). The EDE, one of the most widely used measures of eating disorder assessment, is a structured clinical interview designed to assess eating disorder psychopathology and comprises four subscales: Weight Concerns, Shape Concerns, Eating Concerns, and Restraint. These subscales are averaged to obtain the Global EDE Scale. Reliability of the EDE Global Scale was good at α = 0.84. Higher scores are associated with higher levels of eating disorder symptoms.

Five Minute Speech Sample (FMSS)

(Magaña et al., 1986). The FMSS is an empirically validated approach to assess parents’ attitudes toward their child (Magaña et al., 1986). An adapted version of the FMSS assessed baseline parent EE directed toward the adolescent with AN/AAN. The FMSS prompts a parent to speak for five minutes about their child and their relationship with their child, and these responses are recorded and coded for EE. The FMSS has demonstrated 6–8 week stability and concurrent validity with a semi-structured interview assessing affective attitudes (i.e., Camberwell Family Interview, (Magaña et al., 1986; Vaughn & Leff, 1976)). We used an adapted version of the Family Affective Attitude Rating Scale (FAARS) (Bullock & Dishion, 2004), a validated coding scheme for the FMSS designed to provide global ratings on dimensions of EE that are developmentally relevant for families with adolescents, criticism and warmth (Bullock & Dishion, 2004; McKenna et al., 2020).

These audio-recorded responses were securely uploaded on a shared drive and coded by trained members of the research team who were blind to treatment condition. All FMSS recordings were double-coded and discrepant scores were conferenced to 100% agreement. On the FAARS, criticism scores range from 1 (“no evidence”) to 9 (“two or more concrete, unambiguous examples”). High criticism was defined as scores of 5 or higher, which corresponds to “one concrete, unambiguous example.” Warmth scores ranged from 1 (“no evidence”) to 9 (“two or more concrete, unambiguous examples”). Low warmth was defined as scores of 3 (“1 or more weak examples”) or lower. The FAARS rating scale was adapted throughout this pilot study to include examples relevant to families of adolescents with AN/AAN (See Appendix A for the coding manual).

Percent EBW (%EBW).

Adolescent weight was obtained by an assessor blind to treatment condition. Percent EBW is the current weight divided by the adolescent’s EBW based on their premorbid body mass index percentile for age-and-sex (BMI%ile) (Centers for Disease Control, 2000). EBWs were calculated by research team members to ensure a standardized calculation methodology. One research team member extracted the highest and lowest premorbid BMI%ile from the participant’s medical chart and calculated the mean premorbid BMI%ile, and %EBW (i.e., current weight divided by EBW). A second coder completed the same process for double data coding. Any discrepancies in weights entered or EBW calculations were double checked for accuracy and entered. Traditional use of percent of median BMI (%mBMI) as a goal weight can be problematic because it tends to underestimate malnutrition for those with high premorbid weights (e.g., >85^th percentile) and overestimate malnutrition for those with low premorbid weights (<10^th percentile). Recent studies and the APA Practice Guidelines have emphasized the importance of obtaining historical height and weight percentiles through growth curves in determining more accurate goal weights (Association, 2023; Jary et al., 2023; Lin et al., 2023; Norris et al., 2018).

Weight Restoration.

Weight restoration was defined as ≥95% EBW. This variable was computed at post-assessment and 3-month follow-up.

Remission.

Full remission from AN/AAN was defined as achieving weight restoration (as defined above) and an Eating Disorder Examination (EDE) score within one standard deviation of community norms, consistent with prior studies [1–4].

Analyses

Primary Aim.

Feasibility and acceptability were evaluated using descriptive statistics on the Acceptability and Feasibility Questionnaire. Fidelity was examined by using descriptive statistics on the fidelity rating scale for the EC manual. Analyses were conducted in SPSS Version 27.

Secondary Aim.

Descriptive statistics for each outcome (parental warmth, criticism, and % EBW) at time point and by treatment group was calculated, as well as a standardized effect size difference in the observed means between groups (Cohen’s d), with the goal of describing the magnitude of the observed between-group differences. Thus, we provide the absolute value of Cohen’s d, given that the metric for two of our outcomes are not inherently meaningful (Kelley & Preacher, 2012), and the following thresholds were used to interpret the effect sizes: d<0.20 was considered trivial, d=0.20 was considered small, d=0.50 was considered a medium effect, and > d=0.80 was considered a large effect (Cohen, 1988). Three separate regression-based analysis of covariance (ANCOVA) models were used estimate the marginal treatment group differences after baseline adjustment for each of our three continuous outcomes at both post-treatment and at follow-up (i.e., six models in total – one for each of the three outcomes at two timepoints). To assess weight restoration, we calculated the proportion of participants within each treatment group at each time point who met criteria for weight restoration and conducted logistic regression analyses to estimate the odds of weight restoration by treatment group at each timepoint. To examine remission rates, we calculated the proportion of participants within each treatment group at each time point who met criteria for full remission. All analyses for our secondary aim were conducted in Stata version 18. Missing data in our regression-based analyses was handled via full-information maximum likelihood estimation.

Results

Study Participants

Adolescent participants primarily identified as white non-Hispanic (95%) and female (88%), with a mean age of 14.9 years (SD = 1.6). Most adolescents (59%) were diagnosed with AN and 41% were diagnosed with AAN (See Supplemental Table 1 for descriptive statistics by diagnosis group). Mean %EBW at baseline was 89% (SD = 9.5) and 46% of adolescents required medical stabilization for their AN/AAN prior to beginning FBT, consistent with other RCTs of this population (e.g., 34–58%) (Lock et al., 2005; Lock et al., 2021; Lock et al., 2010). Parent participants were mostly mothers (95%; see Table 1). There were no significant group differences on any of these variables.

Table 1.

Sample Characteristics

	Treatment Group N (%)	Control Group N (%)
Cisgender Male	2 (9.1 %)	2 (10.5 %)
Cisgender Female	20 (90.9 %)	16 (84.2 %)
Transgender	0 (0 %)	1 (5.3 %)
Age M (SD)	15.00 (1.6 %)	14.74 (1.7 %)
Ethnicity: Hispanic	1 (4.5 %)	0 (0 %)
Race: Caucasian	21 (95.5 %)	19 (100 %)
Race: Biracial	1 (4.5 %)	0 (0 %)
Duration of Illness in months M (SD)	1.4 (1 %)	1.3 (.5 %)
Psychiatric medication	11 (50 %)	7 (36.8 %)
EDE Global Score M (SD)	3.80 (1.7 %)	2.73 (1.5 %)
%EBW M(SD)*	91.80 (10.4 %)	85.99 (7.4 %)
Required medical admission	8 (36.4 %)	7 (36.8 %)
FMSS Crit score M (SD)	5.36 (2.3 %)	5.26 (1.7 %)
FMSS warmth score M (SD)	5.50 (1.7 %)	5.95 (1.8 %)
Mood disorder (Ksads)	0 (0 %)	2 (12.5 %)
OCD (Ksads)	2 (10.5 %)	3 (18.8 %)
Anxiety Disorder (Ksads)	11 (57.9 %)	9 (56.3 %)
Suicidal ideation or self-harm history (Ksads)	2 (10.5 %)	3 (18.8 %)

Note: EDE= Eating Disorder Examination; %EBW=percent expected body weight; FMSS Crit score = Five Minute Speech Sample criticism score on the Family Affective Attitude Rating Scale (FAARS); FMSS warmth score = Five Minute Speech Sample warmth score on the Family Affective Attitude Rating Scale (FAARS); Ksads = Kiddie Schedule for Affective Disorders and Schizophrenia; OCD = Obsessive Compulsive Disorder.

^*%EBW differed significantly between groups at baseline: t =2.02; p = .04.

Primary Aim

Treatment Completion.

29 of 41 recruited participants (71%) completed treatment: FBT+EC (14/22= 64%) and FBT+support (15/19=79%); these rates did not differ by group. Number of FBT sessions attended did not differ by group: FBT+EC: M = 18, SD = 9.2; FBT+support: M = 17, SD = 8.7; p = 0. 77. Finally, number of group sessions (out of 10), did not differ by group: FBT+EC: M = 7.2, SD = 2.2; FBT+support: M = 8.6, SD = 1.7; p = 0.11. Rates of attrition did not vary systematically by race, ethnicity, age, or caregiver type (mother or father), nor did we observe any associations between missing data and diagnosis or baseline outcome scores, except for baseline %EBW – missing data was associated with higher % EBW at baseline, p = .04 (see Table 1).

Feasibility and Acceptability.

Recruitment and retention rates were 76% and 71%, respectively. Satisfaction with the intervention was high for both groups, with 100% of FBT+EC participants rating the EC intervention as “beneficial” to “very beneficial.” There were no significant differences between treatment groups in perceived benefits. Across groups, 73% felt “The amount of material covered in group was just right,” 71% felt “60 minutes sessions was just the right length of time,” 86% preferred the telehealth format, 81% agreed “I used the skills that I learned in the intervention,” with no differences between groups. In addition, 77.4% of group participants endorsed, “The intervention has helped me to communicate with my child about their emotions related to their eating disorder,” with the FBT+EC group being more likely to agree or strongly agree to this item compared to the FBT+support group (t = 2.13, p < 0.05).

Fidelity to FBT+EC.

Fidelity to the FBT+EC treatment manual was satisfactory. Study therapists met for weekly supervision sessions with author AS and had consistently high attendance (>90%). All FBT+EC sessions were reviewed for fidelity by one of the treatment developers, with all sessions rating between a 6 and 7 on a 7-point scale indicating “very good” to “excellent” fidelity to treatment procedures as outlined in the FBT+EC manual, including review of homework, coverage of main didactic points, and facilitation of role plays and discussions.

Secondary Aim

Parental warmth.

We observed a large effect size difference between FBT+EC and FBT+support groups in the observed means for parental warmth at post-treatment (see Figure 2 and Table 2). When we adjusted for baseline warmth scores, the 95% confidence interval (95%CI) for the differences between groups in the marginal means of parental warmth at post-treatment did not contain zero (b = 2.55, 95%CI = 1.38, 3.71), indicating the FBT+EC group was approximately 2.55 points higher on parental warmth than the FBT+Support group at post-treatment. At follow-up, there was only a trivial effect size difference between the FBT+EC group and FBT+support group for observed parental warmth scores, and the 95%CI for the estimated marginal difference between groups, covarying for baseline parental warmth scores, contained zero, b = 0.34, 95%CI = −1.20, 1.87.

Figure 2.

Parental Warmth Across Treatment Timepoints

Table 2.

Descriptive Statistics for Primary and Secondary Outcomes by Group and Time point

	Baseline	1-Month	Post-Treatment		Follow-up
Outcome	Observed	Observed	Observed	Marginal Estimates	Observed	Marginal Estimates
Criticism, M (SD)
FBT+EC	5.4 (2.3)	5.5 (2.8)	3.7 (2.0)	3.8 (.6)	2.6 (1.6)	2.6 (.6)
FBT+support	5.3 (1.7)	3.7 (2.1)	4.3 (2.1)	4.3 (.6)	3.1 (2.0)	3.1 (.6)
Effect Size	n/a	d = .78	d = .28	d = .24	d = .25	d = .26
Warmth, M (SD)
FBT+EC	5.5 (1.7)	7.0 (1.3)	7.8 (.6)	7.9 (.3)	7.4 (1.7)	7.7 (.5)
FBT+support	5.9 (1.8)	6.3 (1.8)	5.5 (1.9)	5.4 (.6)	7.5 (1.6)	7.3 (.6)
Effect Size	n/a	d = .46	d = 1.58	d = 1.44	d = .08	d = .22
%EBW, M (SD)
FBT+EC	91.8(10.4)	95.1 (7.1)	96.6 (6.3)	94.6 (1.3)	98.0 (5.6)	96.3 (1.6)
FBT+support	86.0 (7.4)	90.8 (7.8)	92.5 (8.3)	94.4 (1.8)	92.6 (10.4)	94.1 (2.6)
Effect Size	n/a	d = .57	d = .54	d = .03	d = .65	d = .27
Weight restoration, %(n)
FBT+EC	n/a	n/a	54(7/13)	n/a	83(10/12)	n/a
FBT+support	n/a	n/a	50(7/14)	n/a	36(5/14)	n/a
Odds Ratio	n/a	n/a	1.17	n/a	8.68	n/a
EDE remission, %(n)
FBT+EC	n/a	n/a	50(5/10)	n/a	63(5/8)	n/a
FBT+support	n/a	n/a	58(7/12)	n/a	82(9/11)	n/a
Odds Ratio	n/a	n/a	.72	n/a	.37	n/a
Full remission, %(n)
FBT+EC	n/a	n/a	40(4/10)	n/a	43(3/7)	n/a
FBT+support	n/a	n/a	27(3/11)	n/a	40(4/10)	n/a
Odds Ratio	n/a	n/a	1.80	n/a	1.13

Parental criticism.

We observed a small effect size difference between the observed group means of criticism at post-treatment, with the FBT+EC group having slightly lower scores than the FBT+support group (see Figure 3 and Table 2). After adjusting for baseline criticism scores the 95%CI for the difference in the estimated marginal means contained zero, b = −0.52, 95%CI = −2.10, 1.07. This pattern of results was consistent at follow-up (b = −0.47, 95%CI = −2.05, 1.10).

Figure 3.

Parent Critical Comments Across Treatment Timepoints

%EBW.

We observed a medium effect size difference between the FBT+EC and FBT+support groups in %EBW at post-treatment (see Figure 4 and Table 2), but after adjusting for baseline %EBW, the 95%CI for the estimated marginal mean difference between groups contained zero, b = 0.17, 95%CI = −3.81, 4.16. At follow-up, we observed a similar pattern of results, b = 2.20, 95%CI = −3.38, 7.77.

Figure 4.

Adolescent Percent Expected Body Weight (EBW) Across Treatment Timepoints

Weight restoration.

The proportion of participants who met criteria for weight restoration at post-treatment in the FBT+EC group was 54% compared to 50% in the FBT+support group. However, at follow-up, only 36% of the FBT+support group was weight restored compared to 83% of the FBT+EC group (see Table 2). Logistic regression analyses revealed the FBT+EC group was 17% more likely to be weight restored at post-treatment, OR=1.17, 95%CI = 0.18, 3.85. At follow-up, the FBT+EC group was 9 times more likely to be weight restored compared to FBT+support, OR=9.0, 95%CI = 1.34, 60.65.

Remission.

Full remission rate at post-treatment for the FBT+EC group was 40%, while 27% of the FBT+support group fully remitted. At follow-up, full remission rates were comparable between groups: 40% of the FBT+support group and 43% of the FBT+EC group met criteria for full remission (see Figure 5 and Table 2).

Figure 5.

Full Remission from AN/AAN at Post-Treatment and Follow-up

Discussion

This pilot effectiveness trial suggests augmenting FBT with a virtual parent EC group is feasible and highly acceptable to adolescents with AN/AAN and their families. This is important considering time demands associated with standard care (i.e., weekly FBT + medical visits) are already high. The onset of COVID-19 necessitated a pivot to telehealth, unexpectedly providing an opportunity to test and demonstrate feasibility of virtual intervention, with recruitment and retention rates exceeding 70% among families who are more likely to drop out of treatment (Rienecke et al., 2016). Finally, this pilot study demonstrated high parent and interventionist satisfaction, with all parents in FBT+EC and all EC interventionists describing the intervention as beneficial and feasible.

FBT+EC had higher rates of remission at post-treatment (40%) compared to FBT+support (27%) and adolescents were 9 times more likely to be weight restored at follow-up. These findings mirror and, in some cases, surpass those previously published among adolescents with AN only (Le Grange et al., 2016; Lock & Le Grange, 2019; Lock et al., 2010; Madden et al., 2015). This is remarkable given that all families in our sample demonstrated high EE at baseline, thus inherently at risk for poorer FBT outcomes (Allan et al., 2018; Bohon et al., 2023; Rienecke et al., 2016).

Among FBT+EC participants, high parental warmth scores were demonstrated starting at the one-month timepoint when the adolescent’s illness is critical and, arguably, parental warmth is most needed. Early improvements in parental warmth may be facilitating more effective communication (e.g., use of active listening and validation) between parents and adolescents (while maintaining FBT behavioral principles), leading to improved adolescent eating disorder symptoms and weight restoration. High EE families are less likely to achieve early response (Bohon et al., 2023) and EC shows promise in moving warmth early in FBT. By follow-up, FBT+support “caught up” with FBT+EC on parental warmth. However, by the three-month follow-up, adolescents are no longer in the acute phase of treatment likely leading to decreased family distress secondary to the illness, promoting higher parental warmth. In contrast, small decreases in criticism were found for FBT+EC compared to FBT+support. Nonetheless, at follow-up the EC parent group decreased to a low level of criticism (e.g., 2–3 range) while the support group decreased to a slightly higher level of criticism as a group (e.g., 3–4 range).

While pilot effect sizes require replication in larger samples, they provide preliminary evidence for the effectiveness of FBT+EC compared to an attention control condition. Notably PFT compared to standard FBT is less likely to increase criticism and increases parental warmth (Allan et al., 2018). However, PFT does not specifically target these mechanisms like EC, and thus, these benefits to PFT may be more by chance. Future research is needed to explore the relative advantages of FBT+EC to PFT.

Given our study’s modest sample size, effect sizes for treatment effects should be interpreted with caution (Leon et al., 2011). With a larger sample size in a full trial, data analysis would ideally make full use of longitudinal data and use mixed effect models or other repeated measures procedures. However, given the small sample sizes and the goal of providing only a preliminary description of group differences in the outcomes, we chose to compare the means at discrete timepoints. Further, enrollment rates were impacted by: a) recruitment and intervention occurred entirely during the height of the COVID-19 pandemic, b) exclusive recruitment of high EE parents (60% of otherwise eligible participants), and c) logistical challenges of recruiting enough participants to conduct groups concurrent with each participant’s unique FBT timeline, resulting in some families losing interest/dropping out due to wait times. Finally, adolescents in FBT+EC had higher baseline %EBW, indicating a potential need to stratify for %EBW in future trials.

The group intervention format appeared to be a strength, with parents in both conditions finding the structure and associated peer support as positive. Our use of an attention control condition was an additional study strength, in which parents received weekly, 1-hour group session as opposed to waitlist assignment or usual care. This design increased confidence that the EC group content, versus group format, drove changes in treatment targets. Another strength of this study was inclusion of individuals with AAN who are typically excluded from FBT RCTs. Finally, measurement of %EBW using pre-morbid growth history was a strength of the current study. This method appears more accurate, particularly for individuals with AAN, because it accounts for premorbid growth trajectory and does not underestimate goal weights for individuals with higher premorbid weights (Norris et al., 2018).

Our promising pilot data addresses a significant gap in the pediatric AN/AAN treatment outcome literature by targeting mechanisms of non-response in FBT: high critical comments and low parental warmth. A large multi-site fully powered RCT is needed to determine treatment effectiveness.

Public significance:

Family based treatment for AN/AAN is the recommended treatment for youth but families with high criticism/low warmth are less likely to respond to this treatment. Adding a parent emotion coaching group (EC) where parents learn to talk to their adolescents about tough emotions is feasible and well-liked by families.

Anne Shaffer, Ph.D. & Claire M. Aarnio-Peterson, Ph.D.

For the purposes of this study, we are most concerned with overall impressions of criticism and warmth. This addendum is meant as a supplement to the original coding scheme, not a stand-alone replacement.

The same numerical system of rating will be used, but just for the broad Criticism and Warmth categories.

• “1” no evidence for the duration of the speech sample

• “2–3” some indication of item being coded, but no concrete evidence

• “3–4” one or more weak examples

• “5” is a borderline score in which Criticism or Warmth is present, but does not globally characterize the statements. The speech sample must one concrete, unambiguous, unqualified example, or 3 or more weak examples of the same behavior or attribute (such that the speech sample as a whole provides evidence, albeit weak for the item being rated).

• “6–8” indicates that Criticism or Warmth is clearly present and characterizes the statement as a whole. The speech sample must include at least two concrete examples and one or more weak examples of particular behaviors or attributes.

  • To score in the range of 6 or above, consider not just what is said, but the tone with which it is expressed.

• “9” Numerous (i.e., three or more) concrete, unambiguous examples of particular behaviors or attribute.

For the purposes of the FEED study, a 5 or higher on Criticism, or a 5 or lower on Warmth, indicates study eligibility.

Double-coding is not an issue since the broad categories are utilized. Criticism and Warmth categories are explained further by sub-groupings, but this is just for the sake of defining what a critical statement is. The categorization into the smaller sub groups is not relevant at this time.

IF something is a literal strong/concrete comment but it doesn’t sound authentic or the intent is questionable, then it might actually be a weak example. Similarly, if a strong comment is followed up with a statement that mitigates or contextualizes it that might weaken the overall evaluation.

Concrete examples that are scored at the level of 6 or higher should be sufficiently elaborated (i.e., more than one concrete example of the same behavior or attribute; more than a single sentence or phrased used to describe a concrete example).

How to address comments about how teen used to be, as a child or before treatment? If they sound like they still form the basis of opinion of the teen or the relationship, can consider coding (e.g., she has been difficult ever since she was born; she started being so rigid about a year ago, and it’s still a big problem). If they are comparing past behavior or relationship quality to how things are now (e.g., she had become really withdrawn as the disorder got worse), consider not coding as current examples of warmth or criticism.

CRITICISM DOMAINS

Criticism of child

• critical of traits, persona3lity, intentions, or behaviors

• CONCRETE EXAMPLES:

  • “She’s nit-picky.”
  • “She’s so selfish.”
  • “She says she doesn’t like the meal I’ve prepared because she just wants to control everything and make it difficult.”
  • “She’s such an obsessive runner now and it’s just to burn up all her calories.”
• WEAK EXAMPLES:

  • “Sometimes if she could just relax, it would just make it a little easier on all of us.”
  • “I don’t understand why she can’t get it through her head.”
  • “She doesn’t treat her siblings well.”
  • “She is too hard on herself.” or “She’s critical of herself.”

Criticism of the relationship

• conflict, anger, resentment

• implied violence or physical confrontation

• fighting, not getting along

• CONCRETE EXAMPLES:

  • “I just want to shake her sometimes so she’ll snap out of it!”
  • “She fights with me and says I’m giving her too much food.”
  • “She’s not fun to be around.”
  • “She yells at me.”
  • “She is angry with me.”
  • “She is frustrating.” “She is frustrated with me.”
• WEAK EXAMPLES:

  • “She’s difficult for me. She wants quinoa, not what I’m making.”
  • “It’s hard and I’m just kind of sick of it.”
  • “She’s irritated with me.” “She’s somewhat irritating.”

Sarcasm/humor

• Using humor in a negative way to be critical of the child

• Joking

• Using sarcastic tone

• Can be double coded with other criticism items if the sarcasm is directly critical of those aspects (relationship or traits & behaviors)

• Needs to have a sarcastic tone or negative joking aspect to the statement

• CONCRETE EXAMPLES:

  • “It’s not that hard, you just eat. I can do it! I eat plenty. Probably eat too much.” (With sarcastic tone)
• WEAK EXAMPLES:

  • “She thinks she can fix everything and get better by herself.” (with sarcastic tone)

WARMTH DOMAINS

Warmth is comprised of the following types of statements:

Positive about child behavior

• Generally positive about the child’s personality, traits, behaviors, and intentions

• CONCRETE EXAMPLES:

  • “She’s a gifted athlete.”
  • “She’s such a good student and wants to please us.”
  • “My daughter is really competitive and wants to succeed when it comes to sports which is something I admire.”
  • “She’s so smart.”
  • “She is creative.”
  • “She is thoughtful and kind.”
• WEAK EXAMPLES:

  • “She tries to help with chores.”
  • “Kelsey tries her best.”

Love/caring

• Direct statements of love (concrete)

• Speaking about caring for child (weak examples)

• CONCRETE EXAMPLES:

  • “I love her so much.”
• WEAK EXAMPLES:

  • “I just care about her so much.”
  • “I worry about her and want what is best for her.”
  • “I don’t want her hurting.”

Positive relationship

• Speaking about doing joint activities

• Speaking positively about the dyadic relationship

• CONCRETE EXAMPLES:

  • “She’s like my best friend.”
  • “We are really close.”
  • “We really enjoy finding healthy recipes and making them.”
• WEAK EXAMPLES:

  • “We get along.” or “We get along well.”

Coders should record the number of weak and concrete statements for the criticism and warmth categories. A global rating (1 through 9) of criticism and warmth will be assigned based on the overall impressions of the speech sample and statements.

Data availability Statement:

The data that support the findings of this study are openly available in https://nda.nih.gov/ [49]. The data are also available from the authors on request.