The Shifting Perspectives Study Protocol: Cognitive Remediation Therapy as an Adjunctive Treatment to Family Based Treatment for Adolescents with Anorexia Nervosa

Abstract

Background:

Adolescents with anorexia nervosa have set-shifting inefficiencies that can be exacerbated by starvation and that may interfere with outcomes of treatment interventions. Cognitive Remediation Therapy (CRT), an adjunctive treatment focused on improving set-shifting, can target inefficiencies and may augment treatment effectiveness. The best way to add CRT to the standard of care (Family Based Treatment, FBT) for adolescents with anorexia remains understudied.

Methods/design:

This is a randomized controlled trial designed to determine if CRT is effective in increasing flexibility in adolescents with anorexia and/or their parents. Participants are adolescents 12-18 years old with anorexia and their parents. 54 family groups will be randomized into one of three groups: FBT only, FBT plus Parent-focused CRT, or FBT plus Adolescent-focused CRT. Psychosocial, neurocognitive, and behavioral measures will be collected throughout the study.

Discussion:

This is the first study of its kind to apply CRT to parents. All forms of CRT in the context of anorexia have targeted the individual with anorexia’s thinking style. We propose that it may be impactful to target the parent of the adolescent with anorexia as parents carry the burden of treatment and re-nourishment of their child during FBT and may have similar thinking styles.

Conclusion:

This study takes an experimental therapeutics approach to further our understanding of the mechanisms of treatment for adolescents with anorexia. It focuses on increasing cognitive flexibility in patients or their parents and determining the appropriate dose of CRT needed to achieve positive change.

Trial registration:

ClinicalTrials.gov Identifier NCT03928028

1.0. Introduction

Anorexia Nervosa (AN) is a restrictive eating disorder characterized by significantly low body weight for one’s developmental history, age, and sex.¹ The typical age of onset is during adolescence.² It is accompanied by significant medical and psychological complications³ and has one of the highest mortality rates of all psychiatric illnesses.^4–6 The immediate goal in the treatment of AN is nutritional repletion and weight restoration to reverse physical and cognitive complications associated with malnutrition. Parents take the lead in this process and the treatment team acts as supporting consultants.^7–9 During the first phase of treatment, parents temporarily take charge over all food choices as they increase calories and food exposures. As treatment progresses, age-appropriate responsibilities involved in eating are gradually returned to the child in a developmentally appropriate manner, making sure that AN behaviors do not worsen as this occurs. Over time, symptoms of the eating disorder reduce and the adolescent returns to a normal developmental trajectory. Though effective, a significant portion of adolescents do not reach full remission,¹⁰ highlighting the need to improve the effectiveness of FBT.

1.1. Targeting Cognitive Flexibility.

Results from studies in adults with AN indicate that they have inefficiencies in executive functioning, specifically set-shifting.^11–15 Set-shifting is the ability to switch tasks, shift attention, rapidly adapt to new situations, and respond to new environmental contingencies.¹⁶ Set-shifting inefficiencies in adults with AN do not fully improve with nutritional rehabilitation¹⁷. Similar inefficiencies are found in siblings and parents of individuals with AN. ^11,18–20 Executive functioning is heritable²¹ and there is preliminary evidence of co-segregation of cognitive inefficiencies in the context of AN.¹⁹ We hypothesize that behaviors associated with inefficiencies in set-shifting in either parent or adolescent can maintain AN or interfere with treatment. Addressing cognitive flexibility in the treatment of AN may have a positive impact on outcome by facilitating early changes in flexibility that increase approach behaviors (e.g., eating a wider variety of macronutrients, eating with friends) or decrease avoidant behaviors (e.g., parental accommodation symptoms) leading to a faster rate of weight gain, which has been shown to improve outcomes.²²

Cognitive flexibility can be increased by Cognitive Remediation Therapy (CRT), an adjunctive treatment focusing on cognitive skills, particularly skills associated with executive functioning. Previous studies show that CRT positively impacts cognitive flexibility in adults with AN,^23–30 yet research on the use of CRT as an adjunctive treatment for adolescents with AN is limited and conflicting.^31–37 There is significant gap in our knowledge regarding how best to apply CRT in the treatment of adolescents. Because parents are the implementors of treatment in FBT, it may be that supplementing FBT with CRT for parents has the largest impact on outcome. In this randomized clinical trial, the primary goal is to examine whether or not CRT improves flexibility in the adolescent or their parent. The end goal of this research line is to determine whether or not providing CRT to adolescents with AN or their parents effective in enhancing outcomes in FBT.

2. Materials and Methods

2.1. Trial Design and Purpose

The Shifting Perspectives Study is a first of a series of two randomized controlled trials that take an experimental therapeutics approach as outlined by the NIMH Strategic Plan for Research.³⁸ In this approach, the focus is on malleable etiological or maintenance targets for intervention. If the treatment impacts the hypothesized target, the next step is to evaluate whether or not there is improved outcome in the experimental condition and if this outcome is linked to changes in the target.³⁹

In the Shifting Perspectives Study, cognitive flexibility is the target. It is a putative endophenotype (or intermediate phenotype) for anorexia nervosa¹¹ (thus, trait inefficiencies would be expected in a parent and child) and can be impacted by stress⁴⁰ and hunger⁴¹ (state-based inefficiencies). We hypothesize that CRT will improve cognitive flexibility and that that improvements in flexibility can lead to reductions in avoidance and increases in approach-based motivation. For adolescents, this leads to more frequent social interactions and a greater salience of social reward, which can motivate treatment and increase the likelihood of consuming sufficient and varied nutrients (leading to faster weight gain). In parents, flexibility increases approach-based motivation and decreases avoidance, which in turn increases willingness to conduct exposures with their child and reduces the likelihood of parents accommodating symptoms of the eating disorder – leading to provision of sufficient and varied nutrition. The end result is a greater rate of weight gain for the adolescent accompanied by more behavioral gains in terms of food variety. An added benefit of increasing parental flexibility is that parents model flexibility for the adolescent, who may learn to be more flexible by observing parental behavior. Greater flexibility can lead to reductions in relapse (See Figure 1).

Figure 1. Theoretical Model Depicting the Role of Flexibility in Influencing Outcome.

Note. This figure demonstrates hypothesized mechanisms of Cognitive Remediation Therapy (CRT) leading to increases in food variety and weight gain.

Families are randomized to either Family Based Treatment alone (FBT), FBT + parent focused CRT (CRT_p), or FBT + adolescent-focused CRT (CRT_A). In line with the experimental therapeutics approach, specific a-priori thresholds are in place for evaluating if flexibility improves (in parents or the adolescent) sufficiently to continue investigating the impact that CRT could have on enhancing outcomes (see Table 1). The aim is to increase flexibility above and beyond what may be observed in FBT alone. The rationale for this is that in adolescents, renourishment alone may lead to improvements in flexibility. For parents, a reduction in stress and caregiver burden as their child recovers may result in improvements in flexibility. Thus, we seek to determine if CRT is truly effective in increasing flexibility above and beyond standard treatment.

Table 1.

Milestone Criteria for Go / No-Go Decision

Adolescent		Parent
Criteria	Effect Size	Criteria	Effect Size
FBT + CRT group demonstrates greater improvement than FBT alone group on at least one of the following sub-tests of the Delis Kaplan:		FBT + CRT group demonstrates greater improvement than FBT alone group on at least one of the following subtests of the Delis Kaplan:
Number-Letter Sequencing	f = 0.40	Number-Letter Sequencing	f = 0.40
Inhibition		Inhibition
Inhibition/Switching		Inhibition/Switching
Correct Response Shifting		Correct Response Shifting Category
Category Switching		Switching
Description 1 or 2		Description 1 or 2
FBT + CRT group demonstrates greater improvement than FBT alone group on at least two of the following subtests of the Delis-Kaplan:		FBT + CRT group demonstrates greater improvement than FBT alone group on at least two of the following subtests of the Delis-Kaplan:
Number-Letter Sequencing		Number-Letter Sequencing
Inhibition	f = 0.24-0.26	Inhibition	f =0.24-0.26
Inhibition/Switching		Inhibition/Switching
Correct Response Shifting		Correct Response Shifting
Category Switching		Category Switching
Description 1 or 2		Description 1 or 2
BRIEF Inhibit or Shift		BRIEF Inhibition or Shift

Note. A “Go” decision need only satisfy one of either criteria for both parent and adolescent CRT.

2.2. Aims

The primary aim is to identify whether or not we will see greater improvements in flexibility in adolescents or parents who receive CRT compared to their counterparts who receive FBT alone. A second aim is to determine the dose of CRT needed to increase cognitive flexibility. If we are able to demonstrate significant increases in cognitive flexibility, a second phase of this work will focus on replicating target engagement and dose and establish that there is a better outcome when CRT is added to FBT. The methods for this second phase are identical to the protocol described herein.

2.3. Study Design

The study is a randomized, three group design, single-center clinical trial that commenced in May of 2019, with recruitment starting in August of 2019. Figure 2 depicts the study design layout for each arm. Families receive 15 sessions of manualized FBT and/or CRT_P/A over six months. For those in a CRT condition, CRT occurs prior to FBT. Assessment takes place at five different time points (as outlined in Figure 2 and Table 2).

Figure 2.

Study Diagram Depicting Flow of Treatment and Research Activities

Table 2.

Measures and Assessments Administered for Each Outcome

Outcome Category	Measure	Time Point
		T1	T2	T3	T4	T5
Target: Cognitive Flexibility	Behavior Rating Inventory of Executive Functioning (BRIEF)	x		x	x	x
	Delis-Kaplan Executive Functioning System (DKEFS)	x		x	x	x
	Wechsler Abbreviated Scale of Intelligence	x				x#
Outcome: Weight & ED Symptoms	Accommodation and Enabling Scale for Eating Disorders (AESED)*	x	x	x	x	x
	Anorectic Behavior Observation Scale (ABOS)*	x	x	x	x	x
	Eating Disorder Examination Questionnaire (EDE-Q) #	x	x	x	x	x
Outcome: Behavioral Flexibility	Buffet Challenge#	x			x	x
	Meal Planning / OSOG*	x			x	x
Treatment Credibility & Therapeutic Alliance	Credibility and Expectancy Questionnaire (CEQ)	x
	CRT Treatment Evaluation Questionnaire (TEQ)					x
	Working Alliance Inventory (WAI)		x	x	x	x
Comorbid Symptoms & Potential Moderators	Autism Spectrum Quotient*	x	x	x	x
	Behavior Assessment System for Children (BASC)	x	x	x	x	x
	BIS/BAS	x	x	x	x	x
	Depression Anxiety and Stress Scale (DASS)	x	x	x	x	x
	Detail and Flexibility Questionnaire	x	x	x	x	x
	Eating Disorder Flexibility Index#	x	x	x	x	x
	Experience of Caregiving Inventory*	x	x	x	x	x
	Family Questionnaire*	x	x	x	x	x
	Intolerance of Uncertainty Scale	x	x	x	x	x
	Line Bisection Task	x		x	x	x
	Obsessive-Compulsive Inventory	x	x	x	x	x
	Parent Versus Anorexia Scale (PVA)*	x	x	x	x	x
	Rey-Osterrieth Complex Figure Task (Rey-O)	x		x	x	x
	Social Reward Questionnaire#	x	x	x	x	x

^*= Parent only

^#= Adolescent Only

2.4. Ethical Considerations

Approval of the study protocol was by Children’s Hospital of Philadelphia Institutional Review Board. Written informed consent for each adolescent study participant is obtained prior to any data collection; parents provide written informed consent for their child and themselves. This study is registered on ClinicalTrials.gov (NCT03928028).

2.5. Setting

This randomized controlled trial takes place in the research arm of the Eating Disorder Assessment at Treatment Program (EDATP) at Children’s Hospital of Philadelphia (CHOP). The EDATP is an interdisciplinary team of core medical, behavioral health, and nutrition staff. The program provides family-based treatment on an outpatient basis. An inpatient medical stabilization unit is available when adolescents have medical complications due to malnutrition.

2.6. Participants and Eligibility Criteria

Participants adolescents aged 12–18 who meet DSM-5 criteria for Anorexia Nervosa and are medically stable for outpatient care. Both biological parents of the adolescent must be willing to participate. Siblings currently living with the participant are asked to attend at least the first four sessions of treatment.

Epidemiological data indicate anxiety, depression, and obsessive compulsive disorder are common among individuals with AN;^42–46 in order to increase external validity, individuals with these comorbidities will be included. Likewise, we will not exclude participants for the use of medication, with the exception of atypical antipsychotics (e.g., olanzapine). Atypical antipsychotics are sometimes prescribed to assist in weight gain or to reduce inflexibility and preservative behavior in adolescents with anorexia nervosa.^47,48 Any prescribed medications or dosage changes will be tracked throughout the study. Concurrent psychotherapy will not be permitted during the study. Full exclusion and inclusion criteria are in Table 3.

Table 3.

Inclusion and Exclusion Criteria for Parents and Adolescent

Adolescent		Parent
Inclusion Criteria	Exclusion Criteria	Inclusion Criteria	Exclusion Criteria
Age = 12-18 Meets DSM-V criteria for Anorexia Nervosa Medically stable for outpatient treatment Fluent in English No co-morbid condition that would exclude participation Medical clearance from primary care physician Biological parents and caregivers who live with adolescent willing to engage in treatment	Outside of age range Adolescent is adopted Adolescent is pregnant Presence of suicidal ideation, psychosis, bipolar disorder, substance use disorder, autism spectrum disorder, or developmental disability Presence of a brain disorder or injury (such as TBI) that could impact the ability to engage in treatment. Use of anti-psychotic medication	Age > 18 Child with diagnosis of Anorexia Nervosa Both biological parents willing to participate in treatment Fluent in English Parents will be included if pregnant as neither FBT nor CRT should have iatrogenic effects No co-morbid condition that would exclude participation	Presence of suicidal ideation, psychosis, bipolar disorder, substance use disorder, autism spectrum disorder, or developmental disability Presence of a brain disorder or injury (such as TBI) that could impact the ability to engage in treatment. Use of anti-psychotic medication

2.7. Recruitment and Initial Screening

Potential participants are identified during presentation for treatment at CHOP for an eating disorder. This includes during initial hospitalization for medical stabilization, during outpatient intake assessment, or during the initial medical visit and evaluation. Potentially eligible adolescents are flagged by the clinical team and reviewed for eligibility. Those who may be eligible are approached and, if willing, screened for final eligibility. If interested, families are scheduled for a baseline assessment.

3. Procedures

3.1. Randomization and blinding

After consent is obtained and baseline assessment is completed, participants are randomized to one of the three groups: FBT, CRT_P, or CRT_A. Randomization employs a covariate-adaptive randomization method (using open source free software; OxMaR)⁴⁹ that ensures an equal distribution of males across the three conditions. Minimization works by re-calculating the probability of allocation to the treatment and control groups after each new participant enters the cohort. The probability of assignment depends on stratification by sex and age (defined by year in school: middle school vs high school) to assure that males have an equal probability of assignment to the different conditions, something that is not guaranteed with simple random assignment. The algorithms for doing this are programmed ahead of time; study team members enter the sex and school level (middle or high) of the participant into the program in order randomize them to groups. After baseline assessment, the study coordinator informs the family and their therapist of their treatment.

Because members of the study team are involved in coding for fidelity of treatment and post-baseline neuropsychological assessments, certain members are blinded to the conditions of the family to ensure bias does not influence the fidelity and/or assessment scores. Two teams are created in which one team knows the conditions of half of the incoming participants, but is responsible for assessing, coding, and scoring for the other half of participants for whom they are blind to their conditions. We have used this method with success in prior research.

3.2. Study measures

Primary outcomes for this study are specifically related to cognitive flexibility and will be assessed via standard neuropsychological assessment and self-report of executive function. However, we also assess credibility of treatment, therapeutic alliance, behavioral outcomes, and symptom change. Table 2 provides a summary of all measures and the timepoints they are administered. Assessments occur at the following times: Baseline (T₁), four weeks (T₂), 8-10 weeks (T₃), 13-15 weeks (T₄), and end of treatment (T₅).

3.2.1. Credibility of treatment and therapeutic alliance

3.2.1.1. Credibility and Expectancy Questionnaire.⁵⁰

The CEQ is a six item measure that assess cognitive and affective elements of how much participants believe that their treatment condition will be effective and that they will improve. Three versions of the CEQ are used, each one tailored to the specific treatment being offered (FBT, CRT_P, or CRT_A). Parents and adolescents complete the CEQ after the first session.

3.2.1.2. Working Alliance Inventory Short Revised (WAI-SR).⁵¹

The WAI-SR assesses patient and therapist perceptions of the collaborative helping relationship during treatment. The link between the quality of the working alliance and positive treatment outcomes is well established.^52,53 Patients and their parents each complete the twelve-item measure. The therapist completes a ten-item variation, the Working Alliance Inventory Short Revised – Therapist (WAI-SRT), for the family as a whole. The WAI-SR is administered at each time point after baseline.

3.2.2. Target Outcome: Cognitive Flexibility

Cognitive flexibility is assessed via standard neuropsychological assessment and self-report of executive function.

3.2.2.1. The Delis-Kaplan Executive Functioning Scale (DKEFS).⁵⁴

The DKEFS measures executive functioning valid for individuals ages 8–89 years. It has been validated for use with adolescents with AN and is part of a standardized battery developed to assess neuropsychological functioning in individuals with eating disorders.^55,56 For this study, we are using Verbal Fluency, Color-Word Interference, Card Sort, and Trail Making Test. Our primary areas of interest are in the executive functions of inhibition and shifting. Inhibition refers to the ability to stop a response or urge; whereas shifting refers to the ability to stop one response and engage in another, that is, shifting gears. Generally speaking, one needs to inhibit a learned response in order to engage in a different behavioral response. Thus, we are interested in both these elements of flexibility.

3.2.2.2. The Behavior Rating Inventory of Executive Function-2 (BRIEF-2).⁵⁷

The BRIEF-2 is an ecologically valid measure of executive functioning in everyday life across multiple contexts. It yields scores for behavioral and cognitive flexibility as well as metacognition. It is a strong and reliable predictor behavioral inflexibility, perseveration, and adaptive behavior in a variety of populations.^58–60 The BRIEF-2 is available in self-report form for adolescents and adults and a parent report version. In this study, parents will complete the BRIEF-2 about themselves and their child. Adolescents will complete the BRIEF about themselves.

3.2.3. Behavioral Outcome

In order to assess the impact of improved cognitive on behavioral flexibility in a real world and in a disease relevant way, adolescents and parents will be asked to complete menu/food-related tasks at three time points: T1, T4, and T5. Parents will be asked to complete three days’ worth of meal plans for their child with AN and visit an on-line grocery store where they can engage in mock grocery shopping. Adolescents will be asked to serve themselves and eat a meal from a buffet.

3.2.3.1. Buffet Challenge.

The Buffet Challenge is described in detail elsewhere.⁶¹ Adolescents are presented with a buffet of food items that represent both “safe” (e.g., salad, no fat yogurt) and “challenge” (e.g., pizza, cookies) foods for adolescents with eating disorders. Food items for the buffet were generated from items in the Food Choice Task.⁶² and in collaboration with the study dietitian. We chose to present adolescents with actual food items instead of a computerized task as there may be difference between being able to choose more varied food on a computer and actually choosing and consuming foods.

As part of the buffet, adolescents are asked to approach the buffet, serve food, and consume the meal without coaching or support from parents. The meal is videotaped and the following data collected: amount of time to approach buffet, amount of time spent serving food, amount of time eating food, amount of food chosen (weight and calories), macronutrient profile of food chosen, and amount (calories and macronutrients) eaten. We also code for mealtime behaviors reported by Gianini and colleagues⁶³ (e.g., tearing, staring, dissecting).

3.2.3.2. Open Science Online Grocery (OSOG).⁶⁴

OSOG is a website that mimics an online grocery store. The store contains approximately 11,000 products across various categories (e.g., breads, produce, beverages, snack foods). Each product is an actual product stocked in American grocery stores with the exception of store brand “Howe’s” products. Product nutrition information and price for the version employed in this study were accurate as of February 2018. By clicking “check-out” consumers indicate they have finished their shopping trip. Consumers are not charged for and they do not receive any groceries. The OSOG has a logging framework that allows researchers have a granularity of understanding where users click, spend the most time, or the sequence of additions and removals from the user’s cart. In this study, parents’ food choices in their cart can be analyzed for macronutrient and caloric content. Parents will be asked to do three days of meal planning and then shop for those menus on-line. Although OSOG allows researchers to make a variety of changes to the store, participants see the “default” version which mimics the product categorization and within-page placement of standard American online grocery stores. This version did not contain any front-of-package labels, budget limits, or other custom changes.

3.2.4. Symptom outcome

For eating disorder symptom outcome, we assess weight change over the course of treatment as well as changes in eating disorder cognitions. Weight and height are measured each week. Goal weights represent expected body weight (EBW). EBW is calculated by the study dietician or medical safety officer and is based on historical growth curves, stage of pubertal development, menstrual history for females (including age of menarche, frequency of periods, and weight at last normal period), hormone levels (estradiol for females and testosterone for males) and linear growth potential and/or presence of height stunting. Comparison to prior growth patterns is essential, especially prior to the onset of AN. Therefore, EBW reflects where an individual should plot on a growth curve compared to his/her personal lifetime of growth.

3.2.4.2. Eating Disorder Examination Questionnaire (EDE-Q).⁹¹

The EDE-Q is self-report questionnaire based on the Eating Disorder Examination, a semi-structured clinical interview.⁶⁵ It assesses core symptoms of eating disorders including: dietary restriction, weight, shape, and eating concern. The EDE-Q has four scales assessing each core symptoms area and a global score that is the mean of all four scales.

3.2.4.3. Anorectic Behavior Observation Scale (ABOS).⁶⁶

The ABOS is a 30-item measure assessing eating disorder behaviors and designed to be completed by caregivers. Items are primarily behavioral in nature. It has a total score and three sub-scales: eating behavior, bulimic-like behaviors; and hyperactivity. A total score can also be generated; scores over 19 are considered clinically significant.

3.2.4. Definition of Remission

Although the aims of this trial do not focus on symptoms, the a priori criteria for remission are adolescents reaching 95% of EBW and EDE-Q scores within one standard deviation of population norms.

3.3. Investigational Intervention

3.3.1. Family Based Treatment.

FBT is a manualized family treatment for adolescents with AN.⁶⁷ It is effective for children and adolescents aged 12-18. It is a conjoint family treatment with all family members attending all sessions. Both parents (and any additional caregivers such as step-parents and grandparents) are asked to attend. The goal of treatment is return adolescents back to normal physical and psychosocial development via weight restoration and reduction of eating disorder symptoms. The treatment takes place over 3 phases: 1) refeeding and weight restoration; 2) returning control of eating to the adolescent and 3) addressing issues of adolescent development and treatment completion. The primary focus of treatment sessions is to activate and empower the parents of the affected child to restore weight. Secondary goals of FBT include aligning the parental and sibling subsystems and reestablishing healthy adolescent developmental trajectories. We use a standardized 15 session FBT protocol used in prior clinical trials.⁶⁸

3.3.2. Cognitive Remediation Therapy (CRT)

CRT is an adjunctive intervention focused on improving, remediating, or preventing worsening of executive functioning along with fostering meta-cognition (thinking about how one thinks).^69,70 Treatment involves presenting the individual with an assortment of tasks such as geometric figures, illusions, reversing sequences of numbers and letters, completing sorting tasks, and finding various routes on a map. In each task, participants are asked to solve the puzzle or task, reflect on their process in solving it, attempt new ways to approach it, and make connections to real life activities. The focus is on process instead of outcome; a focus that is especially salient for individuals with AN and their parents as CRT guides them away from a perfection-based or task-oriented perspective.^71,72 Learning to strengthen meta-cognitive skills and flexibility in a context that is not illness specific allows participants to strengthen new skills that can later be generalized to more emotionally laden situations (such as meal-times). CRT sessions are administered every week to the treatment groups prior to the standard FBT treatment. Unless touched upon by adolescent/parents, the eating disorder is not discussed. Between therapy sessions, homework is assigned so that CRT related tasks can be practiced.

3.3.3. Managing suicidality and medical instability

The clinical trial follows standard operating procedures of the Eating Disorder Assessment and Treatment Program’s outpatient clinic. Using the Columbia Severity Suicide Rating Scale (C-SSRC),⁷³ Lifetime/current is assessed at intake and since last visit is assessed at every clinic appointment. If needed, the adolescent will be hospitalized for monitoring and management. We will assess parents for suicidality using the C-SSRC. Should parents express suicidal ideation, appropriate steps will be taken, including: development of a safety plan, information about suicide prevention hotlines, referral to a provider, and frequent contact with the study team.

Adolescents seen in the trial are required to be followed medically and to allow communication between providers. If weight does not increase or decreases substantially, immediate risk will be assessed using standard clinic protocols. Should hospitalization be required, adolescents will be allowed to return to the study once medically stable. Weight will be measured and the patient will be assessed for eating disorder symptomatology using the short form of the Eating Disorder Examination – Questionnaire⁷⁴ prior to hospitalization and post-discharge. If at any point during the study an adolescent is no longer appropriate for outpatient treatment (as determined by the Medical Safety Officer), the adolescent will be removed from the study and referred to a higher level of care.

3.3.4. Fidelity and Definition of Treatment Completion

To ensure competent and uniform treatment delivery, treatment sessions are videotaped and a random selection of 20% audiotapes per patient are reviewed for adherence to FBT and the two CRT conditions. Fidelity is assessed using published FBT fidelity measures⁷⁵ and CRT fidelity measures received from K. Tchanturia (personal communication, May 6, 2019). Therapists who are not adherent will be retrained; if adherence is not reached, they will be removed from the trial. Treatment completion is defined as completing 13 out of 15 sessions.

For families who drop out, we will characterize the reason for drop out via the method proposed by Dejong and colleagues.⁷⁶ In addition to reporting on the percentage of sessions completed, we will categorize drop out as clinical withdrawal (e.g., recommendation of higher level of care by the treatment team), logistical withdrawal (e.g., change in parent employment that prohibits travel to treatment), progress withdrawal (withdrawal from the study when both clinician and parents agree that treatment goals have been met), patient/parent initiated withdrawal (e.g., desire for individual treatment).

3.4. Missing Data and Missing Sessions

We will make every effort to avoid missing data. Research intake days at baseline and during follow-up sessions involve a full day at our research site, ensuring that ample time is given to complete each of the research measures during the day. Participant reimbursement is given after the completion of research measures for each assessment day. Should the family need to cancel a therapy session or if the therapist is sick for a therapy session, all attempts will be made to reschedule within the week to ensure the family stays on schedule (including seeing another study therapist if needed). If an appointment cannot be rescheduled within the week, the family will have two sessions scheduled the following week. If the missed session falls during the period where sessions occur every two weeks, the missed session can be made up the following week.

For missing data, if assumptions that the pattern of missingness is occurring at random (MAR) appear tenable, then missing data will be imputed prior to analysis using data augmentation procedures.^77,78 Data are MAR if systematic differences in the outcome variable between those with missing data and those with complete data can be explained by the observed variables (e.g., participant sex or age).⁷⁹ The type of missing data⁸⁰ will be evaluated and the most appropriate method for dealing with it will be chosen (e.g., multiple imputation).

3.6. Planned Analyses

We will use a mixed subjects design to evaluate target engagement. We include 5 different assessment points in the design, though not each time point is used in a given analysis. For example, we only use two time point to establish pre-post change in the dependent variables but use 4/5 assessment points to determine dose. To examine target engagement, we will use a 2 (time) x 3 (treatment) Multivariate Analysis of Covariance (MANCOVA) to establish whether or not flexibility increased over the course of treatment. MANCOVA was chosen due to the number of dependent variables. Prior research from our group indicates that the dependent variables typically correlate around r = 0.3. As IQ is related to executive functioning, it will serve as a covariate. If significant, the MANCOVA will be followed by a series of planned ANCOVAs of FBT vs the appropriate CRT condition. We are interested in how well the CRT condition of interest (adolescent focused CRT + FBT or parent focused CRT+ FBT) performed against FBT alone. FBT alone serves as a control for any gains made in set-shifting due to re-nourishment or maturation (in the adolescent) or reduction in stress due to treatment (in the parents) as well as a control for any practice effects due to multiple administrations of the neuropsychological measures.

Some research indicates that ≥8 sessions of CRT are necessary for improvement in cognitive flexibility with AN.⁸¹ It is unknown how many sessions will be needed for significant change in parents of adolescents with AN. To determine optimal dose needed for change in flexibility, we will model change across the 15 sessions using latent growth curve modeling (LGCM). LGCM is a multivariate method that models individual and average trajectory (across time) and model linear and/or quadratic trends. We will visually assess for the point (dose) at which there is a significant decrease in the rate of improvement (quadratic trend). Parameter estimation will employ Maximum Likelihood (ML) estimation.^82,83 We will evaluate model fit using chi-square (χ²), Comparative Fit Index (CFI), Standardized Root Mean Squared Residual (SRMR), and Root Mean Squared Error of Approximation (RMSEA).⁸⁴ We will use the following values as indicative of acceptable fit between the model and the observed data variance/covariance matrixes: CFI, SRMR, and RMSEA, < .95, < .80, and < .05, respectively. We will assess and compare different models using BIC and novel criteria such as Scaled Unit Information Prior BIC.^85,86 Examining standardized model residuals in addition to global fit indices (GFI) permits us to identify model misspecification even if GFI values suggest adequate fit.⁸⁴ For these analyses we will model change for the participant group (adolescent, mother, father) who had significant change in a dependent variable. We will also focus on the dependent variables that increased significantly by end of treatment.

3.6.1. Power analysis.

Our sample size estimation is based on a mixed MANCOVA design with two repeated measures from baseline to end-of-treatment (EOT; within subjects component) and three groups (CRT_p, CRT_A, and FBT; between subjects component). Based on a review of the literature, we are powering for a medium effect size. Our effect size estimate is Cohen’s f statistic.⁸⁷ This effect size estimate is a ratio of variation explained to that unexplained by the model (treatment groups), and is expressed as follows: $f=\sqrt{\frac{{\eta }^2}{\left(1-{\eta }^2\right)}}$, where ${\eta }^2=\frac{{\sigma }_{model}^2}{{\sigma }_{Total}^2}$; proportion of variability due to the model.^88,89 Heuristics for small, medium, and large effect sizes are .1, .25, and .4, respectively. We selected an alpha of 0.15, given the nature of the study and the difficulty in recruiting participants from a low prevalence condition. Alpha of 0.15 is essential to provide adequate power for this study.⁹⁰ As any analysis powered for a medium effect will also be powered for a large effect, we only highlight the expected medium effect. For all analyses, we specified an α=0.15, two time points, and a correlation between measures of .3 (based on unpublished data). We focused our power analysis on between and within effects. It is not necessary to examine the interaction. Accounting for a 20% dropout rate, we need a sample of 45 treatment completers in order to conduct the analyses. Thus, we will recruit a total sample of 54. Our power for these analyses varies based on the number of groups and whether or not we are examining between or within group effects. For example, using MANCVOA with three groups, assuming f = .26, we are powered at .664 for between groups and .929 for within groups. Likewise, using ANCOVA to compare two groups and assuming f = .26, we are powered at .616; however, within-group comparisons are powered at .821. If we assume a large effect (which was observed by Harrison and colleagues³² in an open trial of CRT), power estimates range from ,84-.997 with the current sample size.

In order to determine the sample size needed to determine dose, we used Mplus 7.2 software⁸³ and conducted a LGCM Monte Carlo simulation with 1000 replications^83,91–93 to determine power to identify CRT_P/A dosage based on a half standard deviation increase (rate of change; slope) from baseline (effect size) above FBT. With 15 participants in 3 groups across 5 time points, parameter and standard error bias were ≤ 5%, confidence interval coverage ≥ 90%, and power = 0.59.

Discussion

CRT for AN is an adjunctive intervention designed to target cognitive flexibility and central coherence. It has been delivered in individual,⁶⁸,⁷²,⁸¹,⁹⁴,⁹⁵ group,^96–100 and family¹⁰¹ and self-help¹⁰² formats. The majority of studies examining the use of CRT for adolescents with AN have been pilot and feasibility studies and most have not had a control group. Those that have used a control group have either been in an inpatient setting or used a control group that may have had active components. For example, art therapy has been used in the past as a valid contact control group for adolescents; however, the art therapy manual used in a prior RCT was based on empirical evidence and may have inadvertently fostered set shifting.⁶⁸ The best way to use CRT to help adolescents is still unknown. Given that neurocognitive inefficiencies are inherited and parents are primarily responsible for treatment, we included a CRT group delivered directly to parents. We hypothesized that increases in flexibility in the parents would facilities delivery of FBT at home. In order to account for any changes that could occur due to brain maturation, we used a minimization procedure to allocate families to conditions. This increased the likelihood that we would have comparable numbers of older and younger adolescents and males and females in each condition. As reduced flexibility could be trait or state-based, we compare FBT+CRT_A/P to FBT alone, allowing for us to have more confidence that improvements in the CRT condition(s) above and beyond the FBT condition are due actual improvements in flexibility (and not brain maturation or weight gain in adolescents, or stress reductions in parents).

While we believe that this study fills a key gap in the literature and will help to answer the question of how best to use CRT to help adolescents with AN, this was not the only approach we considered. We explored the possibility of using a family-based CRT or CRT simultaneously delivered to parents and adolescents. In the first scenario, the family would receive CRT as a unit. This has been piloted by another research group,¹⁰¹ and the focus of CRT was primarily the adolescent’s thinking style. We felt that this would not provide the specific and intentional targeting of parental set-shifting. Likewise, a parent assisted CRT (as was piloted in a self-help format) would not target parental thinking style directly – though individuals who received a caregiver assisted self-help version of CRT did note that doing CRT with their parents impacted the caregiver’s thinking style.¹⁰² In the second scenario, parents would receive CRT from one therapist while adolescents would simultaneously receive CRT from another therapist, and then the family would join together for FBT (ideally with a different therapist). While this approach to augmenting FBT would allow us to examine the synergistic interaction of parent and child on increases in flexibility, it does not seem feasible for dissemination and scaling up to the clinic. It would require three therapists to deliver the treatment. We feel the design chosen is the most parsimonious way to examine the impact of adding parental CRT to FBT.

Our comparison group is FBT alone, thus the two CRT conditions will have more contact with their therapist than those in the FBT alone condition. We chose not to create a contact control for a number of reasons, 1) it would require two additional groups (contact control for parents and contact control for adolescents) and 2) there is no established inactive contact control group that has face validity, although one group has created a control group specifically for a CRT trial.¹⁰³ As noted, art therapy has been used in the past; however, whether or not parents would view art therapy as a credible adjunctive treatment for themselves is unknown. Finally, there is no reason to expect that contact with the therapist alone would increase flexibility. Since CRT is being added in order to enhance treatment, we felt that having an FBT alone control group was the most ecologically valid.

This project is the first phase of a two-phase project. We use an experimental therapeutics approach to determine whether or not CRT positively impacts cognitive flexibility in either adolescents with AN or their parents. Given the sample size (N =54), our power is lower than conventional levels for some analyses and at or higher than the standard of .80 for others assuming a medium effect size. Assuming a large effect size, we are sufficiently powered to observe the effect. If we observe a signal (i.e., meet “go” criteria as outlined in Table 1), we will enter the second phase of the project. In this second phase, we will replicate both the increase in cognitive flexibility and the dose needed to achieve this increase using a larger sample size.

COVID-19 Related Changes to Study Protocol

After study recruitment began, a global pandemic interrupted recruitment and prevented our ability to assess and treat participants in person. In order to protect the safety of research participants, all in-person research activities stopped in March of 2020. If procedures could be completed remotely, research activities were allowed to continue. We were able to rapidly move to a telehealth format for both FBT and CRT. Study team members were trained in best practices in telehealth via resources available at CHOP and all treatment was delivered via a secure, HIPPA compliant tele-health platform. FBT has been delivered via telehealth with success in the past,¹⁰⁴ and modifications to the protocol were in-line with suggestions regarding remote delivery of FBT.¹⁰⁵ To our knowledge, CRT for AN had never been delivered via telehealth prior to this study. In order to complete CRT sessions, materials requiring manipulation were mailed to families at their homes; other materials were made available for download via a study website within one week of cessation of in-person activities. A full description of the transition of CRT to a telehealth format is available.¹⁰⁶ Review of videotaped remote sessions indicates that both CRT and FBT continue to be delivered to fidelity remotely.

Assessments were also moved to remote administration. Given the nature of some of the assessments, a number of key measures were not able to be administered remotely and participants could not complete the Buffet Challenge remotely. When permitted, assessments will resume in person, however, all treatment will continue to be delivered remotely in order to reduce heterogeneity in treatment delivery. Upon completion of the study, a full evaluation of missing data will be made and the analysis plan will be updated accordingly.

Summary

The protocol described takes an experimental therapeutics approach to evaluating the impact of adding CRT to FBT in the treatment of adolescents with AN. Working under the assumption that cognitive flexibility will be trait based, impacted by state factors, or a combination of both - we hypothesized that it may be possible that we can have a greater impact on eating disorder outcomes by adding a parent-focused version of CRT to FBT. CRT was developed over 100 years ago to improve functional outcomes in survivors of traumatic brain injuries.¹⁰⁷ It has been successfully applied to a number of psychiatric illnesses (e.g., depression,^108,109 schizophrenia^110–112), and outcomes are typically focused on improvements in daily living and executive functioning. In the context of eating disorders, many researchers focus on the impact that CRT can have on eating disorder outcomes (e.g., weight, eating disorder cognitions).¹¹³ We contend that there is likely not a direct effect of CRT and subsequent improvements on eating disorder symptoms, but rather that positively impacting flexibly will lead to changes in other areas (such as approach and avoidance) which can impact eating disorder behavior. This study will allow for us to examine whether or not we see significant improvements in flexibility in either the parent or child. Future work will replicate these findings and begin to examine a serial mediation model of the impact of CRT on eating disorder behavior.