Who Responds to an Adaptive Intervention for Adolescents With Anorexia Nervosa Being Treated With Family-Based Treatment? Outcomes From a Randomized Clinical Trial

Abstract

Objective:

Anorexia Nervosa (AN) is a serious psychiatric disorder with high morbidity and mortality. The current evidence supports the use of Family-Based Treatment (FBT), but recovery rates are about 40%. Improving treatment outcomes among adolescents will save lives, improve health, and prevent chronicity. This study examined the relative efficacy of adding Intensive Parental Coaching (IPC) to standard FBT for those who do not respond by weight gain of 2.4 kg by session 4, a clinical predictor of outcome at the end of treatment (EOT), to improve remission rates (>94% of expected mean body mass index (mBMI)).

Method:

107 adolescents and their families were recruited, of whom 69 failed to respond early and were randomized to continue with standard FBT or to receive 3 sessions of IPC. Participants were adolescents with DSM-5 AN between the ages of 12–18 years recruited from across the US; 6.5% were male participants and 11% were Hispanic.

Results:

Main outcomes were mBMI >94% expected for age, height, and sex. Secondary outcomes include change in eating-related cognitions. Adding IPC to early non-responders did not improve outcomes except for those whose parents demonstrated low parental self-efficacy at baseline (moderator). Early weight gain of 2.4kg by session 4 was replicated as a predictor of end of treatment remission.

Conclusion:

Parental self-efficacy (PSE) moderates IPC as an adaptive treatment for adolescent AN. Baseline PSE assessment can be used to identify those families most likely to benefit from adding IPC to FBT. Future studies should consider earlier interventions to improve early response rates.

Clinical trial registration information:

Adaptive Treatment for Adolescent Anorexia Nervosa; https://clinicaltrials.gov/; NCT03097874.

INTRODUCTION

Family-based treatment (FBT) has the largest evidence base supporting effectiveness for adolescent anorexia nervosa (AN). For those receiving FBT, 33–42% fully recover^{1, 2} and remain so 3–4 years after treatment.^3–5 There is substantial room for improvement. Studies of FBT find that weight gain of 2.4 kg by session 4 predicts good outcomes in 60–70% of cases^6–8 thus, one way to improve outcomes in line with precision medicine is to offer an adaptive treatment to patients depending on weight gain at session 4 to address the specific needs of early non-responders (~44%). Pilot studies demonstrated the feasibility of using a randomized adaptive design employing intensive parental coaching (IPC) combined with FBT.⁹ The aim of this study is to confirm whether adding IPC is more effective than standard FBT in achieving weight remission in early non-responders. This study examined the following hypothesis: Adolescent participants with DSM-5 AN who do not respond with sufficient early weight gain by session 4 (early non-responders) randomized to FBT + IPC will have higher remission rates (weight >94% expected body weight) at EOT than those who continue with standard FBT. We also examine potential moderators of treatment effects. We also aimed to examine the early response marker (2.4kg weight gain by session 4) as an end of treatment predictor of remission and recovery.^10–12

METHOD

A detailed description of the study protocol has been published and the study registered on Clinical Trials (NCT03097874).¹³ Participants for this trial were recruited nationally through eating disorder informational websites, Facebook advertisements, and from specialty treatment programs across the US. After informed consent and assent, according to approved protocols by Institutional Review Boards at two treatment sites, 107 participants with DSM-5 AN who were medically stable for outpatient treatment and their families were treated by therapists at Stanford University and University of California San Francisco (UCSF) either in person (n=61) or via teleconferencing (n=33), or a combination of methods due to the immediate COVID-19 transition to teleconferencing (n=13). Participants on medications met entry criteria while on a stable dose of medications for two months prior to starting treatment. Potential participants with medical conditions or treatments that could affect weight gain (e.g., diabetes mellitus, thyroid disease) or were experiencing active psychosis were excluded from the study. Of those who entered the study, 36% (N=38) responded early with weight gain of 2.4 kg by session 4. Those who did not respond early were randomized using a random number generator to either continue in standard FBT or to the adaptive treatment (i.e., FBT+IPC). The study was sufficiently powered (power=0.87, alpha=0.05, two-tailed) to test the primary hypothesis.

Assessments were conducted at baseline, 3 months within treatment, EOT, and 6 months and 12 months post-treatment by trained assessors masked to treatment allocation. Assessments included weight, height, semi-structured interviews, and questionnaires. Adolescents completed the Eating Disorder Examination (EDE) version 16,¹⁴ a semi-structured interview that assesses eating disorder cognitions and behaviors; for this study, the global score was used. Adolescents also completed measures to assess psychiatric comorbidities, including the Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS; baseline only),¹⁵ the Yale-Brown-Cornell Eating Disorder Scale (YBC-EDS),^{17, 18} the Beck Depression Inventory (BDI),¹⁹ the Beck Anxiety Inventory (BAI),²⁰ and the Children’s Yale-Brown Obsessive-Compulsive Scale (CY-BOCS).²¹ The Compulsive Exercise Test (CET),²² and the Commitment to Exercise Scale (CES)²³ were administered to assess for compulsive and overexercise behaviors. Parents completed the Parents vs Anorexia Nervosa Scale (PVAN),¹⁶ a measure of parental self-efficacy, at all major assessment points as well as within treatment. The Helping Alliance Questionnaire (HAQ)²⁴ was administered to parents and collected after session one and at 3 months within treatment.

There were two treatments used in the study. Standard manualized FBT,²⁵ an 18-session treatment delivered over an average of 9 months designed to promote parental management and disruption of the behaviors maintaining AN (e.g. severe dieting, overexercise, purging, food weighing, weight checking, etc.), or FBT+IPC, which adds three sessions to standard FBT to identify key issues that are interfering with parental success at disrupting the maintaining behaviors of AN, starting after session 4.^{9, 26} The IPC sessions involved the following interventions: 1) reiteration of the crisis that poor early weight gain portents for future recovery; 2) a parent-only session allowing the parents and the therapist to identify why the parents think they are unsuccessful and find new strategies they might employ; 3) a second family meal focused on helping the parents try new strategies to help their child eat. Following these three IPC sessions, treatment resumes according to standard manualized FBT. Therapists conducted treatments in both arms of the randomized study. All therapists received training in FBT and in the additional sessions of IPC, and were supervised weekly at both sites (JL and DLG).

Statistical Approach

This study employed an independent Data Collection Center (DCC) coordinated by investigators not part of the intervention team (BJ, BM, CB) who entered data and conducted the analyses. The primary outcome of the study is weight remission (weight >94% expected body weight [EBW]) by EOT. The core analysis strategy is mixed effects modeling,²⁷ which fully utilizes the repeatedly measured primary and secondary outcomes (baseline, 3-month mid-treatment, 9-month (EOT), 6- and 12-month follow-up). For maximum likelihood (ML) estimation of mixed effects models, the Mplus program version 8.6 was used.²⁸ Data points that are missing due to attrition or missing assessments were handled assuming that data are missing at random conditional on observed information.²⁹ The results of these analyses were converted to group differences at EOT and follow-up assessments.

To assess differences in remission rate between the FBT-FBT and the FBT+IPC groups, longitudinal mixed-effects modeling was conducted utilizing 4 repeated measures of remission (3-month mid-treatment, 9-month EOT, and 6- and 12-month follow-up), as there is no variation in the data at baseline (i.e., no one experiences remission yet by design). A piecewise growth model was used, allowing for random intercepts and treating remission status as a binary outcome. All individuals with remission data at one or more assessment points were included in the analysis following the intent-to-treat principle. EBW was also analyzed as a secondary outcome utilizing the same analysis strategy but treating it as a continuous outcome. We examined potential moderators of intervention effects on remission, our primary outcome. For this investigation, we used mixed effects modeling applying the analytical criteria for detecting moderators in the MacArthur framework.³⁰ Data for early responders (non-randomized) is provided as cohort data for descriptive and completeness purposes.

RESULTS

Participants were recruited from October 2017 to February 2021 (see Consort Figure 1). One hundred and seven participants were recruited to the study, of which 38 were early responders and not randomized, while 69 early non-responders were randomized to one of the two treatment conditions (FBT-FBT or FBT+IPC) across the two treatment sites (see Table 1 for demographics and baseline characteristics). The randomized participants were on average 14.7 (1.6) years of age, 93% female participants, and 64% white. Seventy-two percent had been previously hospitalized for AN, duration of AN was 7.76 (6.25) months, and 43% had a diagnosed comorbid psychiatric disorder. (Table 1). Study retention was good, with 91% of the main outcome data collected at EOT, 77% at 6-month follow-up, and 74% at 12-month follow-up. Treatment dropout (defined as not attending 3 or more consecutive sessions) was also low, with 3 dropping out in standard FBT and 8 dropping out in the FBT+IPC arm. There were no serious adverse events.

Figure 1. Consort Diagram.

Note: AN = anorexia nervosa; EBW = expected body weight; FBT-AN = family-based treatment for anorexia nervosa; PI = principal investigator SU = Stanford University; UC = University of California San Francisco

Table 1.

Demographic Information for Sample by Treatment Group and Site

Site	UCSF			Stanford			Total
Group	FBT	IPC	NR	FBT	IPC	NR	FBT	IPC	NR	Total
Ethnicity, n (%)
Hispanic or Latino	1 (5.56)	2 (10.53)	1 (5.0)	3 (18.75)	2 (12.5)	3 (16.67)	4 (11.76)	4 (11.43)	4 (10.53)	12 (11.21)
Not Hispanic or Latino	17 (94.44)	17 (89.47)	19 (95.0)	13 (81.25)	14 (87.5)	15 (83.33)	30 (88.24)	31 (88.57)	34 (89.47)	95 (88.79)
Sex, n (%)
Male	3 (16.67)	1 (5.26)	1 (5.0)	0 (0.0)	1 (6.25)	1 (5.56)	3 (8.82)	2 (5.71)	2 (5.26)	7 (6.54)
Female	15 (83.33)	18 (94.74)	19 (95.0)	16 (100.0)	15 (93.75)	17 (94.44)	31 (91.18)	33 (94.29)	36 (94.74)	100 (93.46)
Race, n (%)
Asian	6 (33.33)	4 (21.05)	7 (35.0)	0 (0.0)	1 (6.25)	3 (16.67)	6 (17.65)	5 (14.29)	10 (26.32)	21 (19.63)
Black or African American	0 (0.0)	1 (5.26)	0 (0.0)	1 (6.25)	1 (6.25)	0 (0.0)	1 (2.94)	2 (5.71)	0 (0.0)	3 (2.8)
More than one race	2 (11.11)	2 (10.53)	4 (20.0)	5 (31.25)	2 (12.5)	0 (0.0)	7 (20.59)	4 (11.43)	4 (10.53)	15 (14.02)
Other	1 (5.56)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (2.94)	0 (0.0)	0 (0.0)	1 (0.93)
White	9 (50.0)	12 (63.16)	9 (45.0)	10 (62.5)	12 (75.0)	15 (83.33)	19 (55.88)	24 (68.57)	24 (63.16)	67 (62.62)
Prior hospital, n (%)
No	2 (11.11)	4 (21.05)	5 (25.0)	5 (31.25)	8 (50.0)	3 (16.67)	7 (20.59)	12 (34.29)	8 (21.05)	27 (25.23)
Yes	16 (88.89)	15 (78.95)	15 (75.0)	11 (68.75)	8 (50.0)	15 (83.33)	27 (79.41)	23 (65.71)	30 (78.95)	80 (74.77)
Has comorbidity, n (%)
No	12 (66.67)	13 (68.42)	17 (85.0)	7 (43.75)	7 (43.75)	7 (38.89)	19 (55.88)	20 (57.14)	24 (63.16)	63 (58.88)
Yes	6 (33.33)	6 (31.58)	3 (15.0)	9 (56.25)	9 (56.25)	11 (61.11)	15 (44.12)	15 (42.86)	14 (36.84)	44 (41.12)
Age, yrs, mean (SD)	14.72 (1.74)	14.37 (1.77)	14.8 (1.24)	15.0 (1.55)	14.75 (1.24)	15.39 (1.2)	14.85 (1.64)	14.54 (1.54)	15.08 (1.24)	14.83 (1.48)
%EBW, mean (SD)	83.04 (4.69)	82.91 (6.26)	83.52 (6.44)	83.47 (4.47)	84.3 (3.78)	82.92 (4.07)	83.24 (4.52)	83.54 (5.25)	83.24 (5.39)	83.34 (5.04)
EDE Global, mean (SD)	2.20 (1.45)	2.52 (1.63)	2.42 (1.61)	2.61 (1.59)	2.35 (1.64)	2.37 (1.57)	2.42 (1.52)	2.42 (1.61)	2.39 (1.57)	2.41 (1.55)
Duration of illness in months, mean (SD)	5.96 (5.09)	7.55 (5.7)	6.08 (2.74)	10.38 (8.4)	6.64 (5.15)	7.53 (5.28)	8.04 (7.1)	7.14 (5.4)	6.74 (4.11)	7.29 (5.6)

Note: EBW = expected body weight; FBT = Family-based treatment; IPC = intensive parental coaching; NR = non-randomized (early responder to treatment); SD = standard deviation; UCSF = University of California San Francisco

This study was initiated prior to the COVID-19 pandemic. Because of the pandemic, study assessment and treatment transitioned from in-person visits to videoconferencing. Both treatment groups received online FBT-FBT or FBT+IPC. Evidence suggests that online FBT for AN is similarly effective as in-person FBT.³¹ Thirteen participants were transitioned mid-treatment (4 in FBT-FBT, 2 in FBT+IPC, and 7 non-randomized early responders). Overall, 61 participants were treated in person, and 33 were treated via videoconferencing.

Replication/validation of early treatment response marker.

Early responders had a robust and rapid improvement in weight, with 75% (29/38) of adolescents reaching weight remission after 3 months of treatment, compared to 46% (32/69) in the early non-responder group. End of treatment weight remission rate for early responders was 63% (24/38), and for non-early responders it was 49% (31/69). End of treatment recovery rates for early responders was 60% (23/38) and for non early responders it was 38% (26/69).

Treatment differentiation adherence check.

To ensure adherence in the FBT+IPC arm and no use of IPC in the FBT-FBT arm, 20% of participant sessions 4, 5, and 6 were randomly selected and reviewed for adherence to ensure no overlap of differentiating intervention elements. Adherence to the core intervention elements of Sessions 4, 5, and 6 was high (98%). The IPC adaptations (e.g., parent-only session, second family meal) were only implemented in that arm with no crossover detected.

Weight remission rates at EOT (Table 2 and Figure 2).

Table 2.

Intent to treat (ITT) Effects on Key Outcomes Based on Longitudinal Mixed Effects Modeling

	FBT-FBT (n=34)		FBT+IPC (n=35)		Group Difference (ITT effect)
Remission	Rate	95% CI	Rate	95% CI	Rate difference (RD)		95% CI of RD	p-value
3-months	0.441	0.28, 0.62	0.475	0.29, 0.66	−0.034		−0.27, 0.22	0.801
EOT	0.502	0.33, 0.67	0.378	0.22, 0.56	0.124		−0.12, 0.33	0.326
6-month FU	0.604	0.42, 0.76	0.406	0.23, 0.60	0.192		−0.07, 0.38	0.142
12-month FU	0.656	0.46, 0.82	0.403	0.23, 0.60	0.241		−0.03, 0.41	0.078
Recovery
3-months	0.380	0.23, 0.56	0.347	0.19, 0.54	0.035		−0.22, 0.27	0.796
EOT	0.408	0.25, 0.58	0.230	0.11, 0.40	0.193		−0.06, 0.38	0.132
6-month FU	0.551	0.37, 0.72	0.188	0.07, 0.38	0.345		0.10, 0.46	0.008
12-month FU	0.582	0.39, 0.75	0.267	0.13, 0.46	0.296		0.04, 0.44	0.024
EDE Global	Mean	95% CI	Mean	95% CI	Mean	Cohen’s d	95% CI of d	p-value
3-months	1.479	0.92, 2.04	1.904	1.32, 2.48	−0.424	−.027	−0.80, 0.25	0.302
EOT	1.472	0.88, 2.07	1.756	1.09, 2.42	−0.284	−0.18	−0.76, 0.39	0.533
6-month FU	1.004	0.54, 1.47	1.812	1.22, 2.41	−0.809	−0.52	−1.01, −0.03	0.036
12-month FU	1.040	0.55, 1.53	1.651	1.06, 2.25	−0.611	−0.40	−0.90, 0.10	0.121
%EBW
3-months	93.132	90.23, 96.03	92.830	89.73, 95.94	0.302	0.06	−0.82, 0.94	0.899
EOT	93.277	90.96, 95.60	92.562	90.20, 94.92	0.716	0.15	−0.54, 0.83	0.672
6-month FU	94.214	91.78, 96.65	92.710	89.10, 96.32	1.504	0.31	−0.59, 1.21	0.499
12-month FU	95.370	93.14, 97.60	92.615	89.30, 95.93	2.754	0.57	−0.26, 1.40	0.177

Note: CI = confidence interval; EBW = expected body weight; EDE = eating disorder examination; EOT = end-of-treatment; FBT = family-based treatment; FU = follow-up; IPC = intensive parental coaching

Figure 2. Remission and Recovery Rates for Study Participants Across all Time Points.

Note: BL = baseline; EOT = end-of-treatment; FBT = family-based treatment; FU = follow-up; IPC = intensive parental coaching; mo = month

The main outcome of this study was EOT weight remission defined as achieving an EBW of > 94% for age, height, and sex. The estimated remission rate was 50% for those who received FBT-FBT and 38% for those who received FBT+IPC, with no statistical difference between the groups (see Table 2). The early responders (results not reported in the table) had a remission rate of 63%.

Recovery rates at EOT (Table 2 and Figure 2).

We compared categorical recovery rates that we defined as requiring weight remission as defined above and EDE global score within one standard deviation of the community mean for adolescents (1.6; SD 1.4). The estimated recovery rate at EOT (see Table 2) for those in FBT-FBT was 41% and 23% for those in FBT+IPC. For this study, the early responder recovery rate was 60%.

Remission rates at follow-up (Table 2).

We compared 6-and 12-month follow-up weight remission outcomes. For those in FBT-FBT, the estimated 6-month weight remission rate was 60%, and 66% by 12-month follow-up. For those in FBT+IPC, the remission rate at 6-month follow-up was 41%, and 40% at 12-month follow-up. These rates were not significantly different. For those that responded early, 67% were remitted at 6-month follow-up, and 61% at 12-month follow-up.

Weight change (Table 2).

We compared weight change as a continuous outcome variable by calculating percent EBW. For those in FBT-FBT, the estimated mean EBW was 93.3% at EOT, 94.2% at 6 months, and 95.4% at 12 months. For those in FBT+IPC, the mean EBW was 92.6% at EOT, 92.7% at 6 months, and 92.6% at 12 months. For those who responded early, the mean EBW at EOT was 97.0%, 96.2% at 6 months, and 97.0% at 12 months.

Eating disorder cognitions (Table 2).

We compared improvement in eating-related cognitions via the EDE. For those who received FBT-FBT, the estimated mean EOT global EDE score was 1.47, by 6-months post-treatment it was 1.00, and by 12-months post-treatment it was 1.04. For those who were treated with FBT+IPC, the mean EOT global EDE was 1.76, by 6-months post-treatment it was 1.81, and at 12-months post-treatment it was 1.65. For early responders, the mean EOT global EDE was 0.82, by 6-months post-treatment it was 0.69, and by 12-months post-treatment it was 0.63. The mean EOT global EDE scores for all groups fell below one standard deviation of the community adolescent mean of 1.60 (SD 1.40) following treatment, and all continued to improve during follow-up.

Moderators of Outcome of Weight Remission (main outcome).

Among 16 baseline variables, baseline PSE (as measured by the PVAN) was a significant moderator of treatment effect on remission at EOT (significant treatment × PVAN interaction with p=.022) (see Figure 3). For those with PVAN scores <= 19 (i.e., randomized non-early responders with lower PSE= 26%, N= 28/107), FBT+IPC shows higher remission, although not significantly higher (SRD=0.17, p=.383). For those with PVAN scores > 19 (greater PSE), FBT-FBT shows significantly higher remission (SRD=0.41, p=.017) at EOT. In addition, AN duration was a significant moderator of treatment effect (see Figure 3) on remission at 3 months (significant treatment × AN duration interaction with p=.016). For those with duration <= 6 months, FBT-FBT shows higher remission, although not significantly higher (SRD=0.17, p=.292). For those with duration > 6 months, FBT+IPC shows a significantly higher remission (SRD=0.45, p=.050) at 3 months.

Figure 3. Parental Self-Efficacy and Duration of Illness By Treatment Group as Moderators of Remission Outcome.

Note: AN = anorexia nervosa; BL = baseline; EOT = end-of-treatment; FBT = family-based treatment; FU = follow-up; IPC = intensive parental coaching; mo = month; PVAN = parents versus anorexia nervosa scale; SRD = success rate difference

Hospital Use.

Seven participants randomized to FBT-FBT (20.6%) and 7 participants randomized to FBT+IPC (20%) were hospitalized during the course of treatment for eating-disorder-related medical reasons. An additional three participants in FBT+IPC were hospitalized for psychiatric reasons only (e.g., suicidality). The total number of hospitalization days was 146 for each condition. Examining hospitalizations for medical instability only, the FBT-FBT group utilized 139 days (15 hospitalizations; range: 4–35 days; mean: 9.3 days, SD: 7.82) whereas the FBT+IPC group utilized 117 days (16 hospitalizations; range: 2–14 days; mean: 7.3 days, SD: 3.14); Cohen’s d=0.34, 95% CI: −3.61, 1.88). An additional two hospitalizations in FBT-FBT and five hospitalizations in FBT+IPC were for non-eating disorder-related medical reasons. Among the early responders, only four participants required hospitalization for medical instability, and none for non-eating disorder-related reasons, accounting for a total of 24 days (4 hospitalizations; range: 3–8 days; mean: 6 days).

DISCUSSION

This study found that adding a 3-session intervention for early non-responders to FBT that focused on improving PSE by addressing parental concerns and behaviors related to weight restoration did not improve overall outcomes by the EOT or at any follow-up point. Nonetheless, on average, by EOT, all participants had mean weights close to an average %EBW, and all had eating-related cognitions in the non-clinical range (within 1 SD of the population mean). The overall categorical recovery rate (as defined above) of 43% is in line with previous studies of FBT for AN, despite this being a sample at risk for poorer outcomes given high hospitalization rates prior to study entry and the substantial proportion of early non-responders.^{12, 32, 33} However, parents with lower PSE at baseline benefited from receiving IPC as an adaptive intervention. Adolescents whose parents had lower PSE and received FBT+IPC had higher remission rates; while the addition of IPC did not help parents with higher PSE. The therapeutic target of the IPC sessions is to improve PSE, and to the extent that the group that needed it most benefited from IPC, supports the view that a subgroup of parents with lower PSE who fail to meet early weight gain by session 4 benefits from the addition of IPC to standard FBT. Providing IPC to this subgroup increased weight remission at EOT from 39% to 56%. Therapists can potentially use this information to better match the intervention to family needs. The PVAN, a brief 7-item parental self-report measure, identifies those families who will likely benefit from adding IPC to FBT when weight gain does not reach 2.4kg by session 4.¹⁶ In addition, duration of illness, consistent with prior research,(34) moderated treatment effect, with those who had longer duration of AN doing better in FBT+IPC by three months. However, the treatment effect of FBT+IPC appears not to be sustained after the 3-month mark, suggesting that additional intervention may be necessary to capitalize on the gains early in FBT+IPC.

The study also confirmed that early weight gain of 2.4 kg by session 4 is a prospective predictor of remission at a rate of 63% and recovery rate of 60% in line with previous reports.^10–12 This replication is important as it highlights the validity of this marker of early response and its clinical utility. Not only is this marker important as a clinical indicator of progress, but it also establishes a specific minimal weight gain goal early in FBT. Strategies to improve rates of achieving early response should target this marker.

One of the most significant impacts of these rapid improvements in this study and in previous studies is decreased use of hospitalization, thereby decreasing suffering and lowering burden and cost, allowing for more rapid overall recovery and resumption of normal developmental processes. These observations support the continued importance of early response on outcomes and the need for rapid identification and provision of effective treatment for adolescents with AN.³⁵ Further, it suggests that the specific early sessions as manualized in FBT are effective, and other models of family therapy for AN that do not replicate these sessions may not lead to rapid early response.³⁶

Putting these results in an overall context of attempts to improve FBT is important. Various strategies to improve recovery rates have been attempted for FBT, including adding more sessions,(37) increasing treatment duration,^{37, 38} seeing parents separately from their children to address negative expressed emotion,³⁹ and adding focused individual therapy (e.g., cognitive remediation therapy (CRT) or art therapy).⁴⁰ While these adaptations demonstrated some evidence of modest improvements, the effects overall were not substantive or enduring. Adding treatments to an effective intervention often does not appear to lead to better overall outcomes for FBT.⁴¹ One reason for this may be that the interventions added may be too little too late. In this instance, we may have added IPC too late for some families that already had set a course of poor response and inadequate PSE that could not be changed, at least in the short term. This raises the question if in addition to providing IPC for those who had low PSE at baseline and who failed to gain 2.4kg by session 4, an intervention could be developed to be used during the first few sessions of FBT that would improve early response for those at risk for poor early response based on lower PSE scores at baseline. It is interesting to speculate about whether duration of AN might impact PSE and in turn treatment response; however, the data available in the current study is too small to allow for reliable statistical testing of this relationship.

There are strengths and limitations to the current study. The study employed an innovative adaptive design allowing systematic study of a treatment aimed at addressing a specific marker of early treatment response. Such studies are unusual because early markers of treatment response are often not specific enough to target; thus, truly targeted adaptive interventions to address identified targets are seldom developed and tested. Hence other researchers might find the approach used in the current study in designing future adaptive randomized clinical trials useful. The study was powered to answer the primary hypotheses, there was low study attrition, and data analytic techniques were powerful and effective approaches to calculating statistical estimates of treatment effects. The sample itself was representative of typical adolescent cases but also had 36% non-white participants and 7% male participant. Assessors were masked to randomization, and an independent data center collected the data set and conducted the analyses without PI involvement. Because treatments were provided by doctoral-level clinicians trained and supervised by experts in FBT, the generalizability of our findings might be compromised in non-specialty programs. There were comparatively larger missing data related to eating-related cognitions (EDE) than weight data. The early responders group is not a randomized group, and comparative data related to their outcomes are purely descriptive and provided for illustrative and completeness purposes.

The sample sizes precluded examination of possible differences of impact on treatment type on hospitalization rates in the randomized groups. While these data suggest early response to FBT lowers hospitalization rates compared to those who were not early responders, future studies should examine the impact of FBT on overall hospitalization use. Additionally, the PVAN, while the only specific measure available to evaluate parental self-efficacy in the context of re-feeding a child with AN, is a short measure with face validity but limited psychometric evaluation.

Another consideration in interpreting the findings of this study is that the remission and recovery categorical outcomes, while consistent with published literature and recommendations, are based on expert opinion and consensus rather than evidence.⁴² As noted in earlier reports, slight changes in the cut-points for defining recovery dramatically impact the rates of remission and recovery.^{1, 2} For example, in the current study, lowering the %EBW remission rate to >92% would have increased the overall remission rate by 20% in the non-early responder group and increased the recovery rate by 20%. Still, studies suggest that reaching EOT EBW of >94% is a strong predictor of long-term outcomes.^{12, 33} Further study of weight and EDE cut-points defining categorical outcomes are needed.

This study suggests that adding an adaptive treatment (IPC) that demonstrated promise at improving outcomes in non-early responders⁹ did not ultimately lead to greater remission or recovery on average. Instead, continuing with standard FBT appears to be reasonable and supports not giving up too soon and instead staying the course with manualized FBT as currently practiced for most patients. However, moderator analyses identified a subgroup of patients whose parents experience lower PSE at baseline experience substantial benefit from adding IPC if their child fails to gain 2.4kg by session 4.¹⁶ Clinicians may use this finding to identify at risk families and provide IPC for those (approximately 1 in 4 patients--26%) who are most likely to benefit from it, in line with precision medicine approaches. Future studies might focus on improving early response rather than trying to remediate early non-response.

Data sharing:

Deidentified data will be made available upon request.