Introduction

Eating disorders (EDs) are severe psychiatric disorders with high rates of mortality [1], suicide [2], and considerable global disease burden [3]. The clinical syndromes of anorexia nervosa (AN) and bulimia nervosa (BN) have been recognized in the diagnostic classification of psychiatric disorders for decades; given their complexity, high rates of comorbid psychopathology [4], and associated medical complications [5], early and effective intervention is critical [6].

Restrictive EDs, particularly AN, are notoriously difficult to treat [7, 8]. Family-based treatment (FBT) is the current first-line approach for treating adolescent AN and a recommended approach for adolescent BN [9]. Although treatment of AN in youth has better recovery rates than for adults, outcomes are not optimal [10, 11]. Understanding for whom and under what conditions a certain intervention approach works best is necessary to develop the most effective treatments, as well as to identify those for whom it is less successful [12, 13]. To that end, this review highlights research on non-specific predictors, moderators, and mediators that inform change processes in FBT.

Given that the strongest evidence base to date within the broader ED field lies in FBT for AN and BN, we focus on this treatment for these EDs. Recent reviews that have addressed factors that impact treatment for adolescent EDs have been limited either by their systematic nature that did not allow for inclusion of a broader range of studies [14], or by their more narrative nature that did not specifically home in on mapping gaps in knowledge within the current literature base [15,16,17]. In contrast, this scoping review includes studies that present a specific focus on non-specific predictors, moderators and mediators of FBT and based on our summary of the evidence, we identify areas where knowledge remains lacking, and raise questions for future research.

We focused our review on evidence from studies that have (1) examined the use of FBT for AN or BN in an adolescent population (ages 12–19); (2) used an RCT design to examine treatment effects (with exceptions to include case studies and pilot work, as warranted and noted in text); and (3) identified at least one non-specific predictor, mediator, or moderator of change in the FBT therapeutic process. We did not include studies with therapy approaches that did not include FBT; studies that did not identify variables that could be explored as factors impacting treatment; and studies that focused on multi-family approaches to FBT (as this work is related to, but not FBT perse, and introduces family-systems-level factors that are beyond the scope of this review). Searches of electronic databases (PubMed, PsychInfo, Google Scholar) were conducted using specific keywords (family-based treatment, adolescent, eating disorders, moderators, mechanisms of change). Reference lists from other recent and comprehensive reviews were also consulted, e.g., [15]. The initial search using the aforementioned key words and other sources identified 63 studies that were read in full; in total, 36 met eligibility including 11 RCTs (Table 1), with the main findings summarized below.

Table 1 Study characteristics and key findings
Full size table

Defining non-specific predictors, moderators and mediators

Non-specific predictors, moderators, and mediators of ED outcome have been of increasing interest, with several recently published reviews [55, 56], but none has comprehensively examined these associations specifically in FBT. Mechanisms of action are processes that occur within treatment that lead to therapeutic change and include treatment mediators [57]. Not all mechanisms of action are mediators, but all mediators are mechanisms of action [13]; therefore, studying mediators may narrow down the search for causal mechanisms when the objective is to enhance features of treatment that lead to improved outcomes or remove those that do not. Mediators do not precede treatment, but instead change as a result of treatment, and it is this change in the mediator which is associated with changes in the outcome [13]. In contrast, baseline variables in randomized controlled trials (RCTs) that impact outcomes are considered non-specific predictors or moderators [13]. Baseline variables that interact with treatment type to affect outcome are referred to as moderators, whereas variables that affect outcome, irrespective of treatment type, are considered non-specific predictors. In RCTs, posttreatment factors (not a moderator) that are uncorrelated with treatment type (not a mediator) that have a main effect on outcome may also be referred to as non-specific predictors.

Therapeutic change in FBT for AN and BN

Non-specific predictors

Non-specific predictors are fixed or variable factors that are not related to the treatment received and precede and have a main effect on that outcome (i.e., the predictor has the same effect on the outcome regardless of the treatment type). While commonly defined as baseline characteristics (e.g., sex; treatment site), non-specific predictors may also comprise post-treatment measures such as early weight gain if they demonstrate a main effect but do not interact or correlate with treatment type [13]. Although non-specific predictors may apply within other treatment modalities (e.g., cognitive behavioral therapy (CBT)), in the interest of providing readers with information most applicable to the study of FBT, we focus in the following summary only on the impact that non-specific predictors had within studies of the FBT modality.

Findings for studies examining AN

Baseline factors

Several studies identified factors related to severity of illness specifically at baseline (i.e., not only as an outcome). For example, Eisler et al. explored the comparative impact of two forms of outpatient family interventions (conjoint family therapy vs. separated family therapy) on both nutritional and psychological outcomes, determined by the Morgan and Russell Scales [31]. Regardless of treatment type, baseline non-specific predictors of better outcomes relative to the Morgan and Russell Scales included less previous ED treatment, shorter duration of illness, and higher percent estimated body weight (%EBW) at the beginning of treatment [31]. In an RCT comparing FBT with adolescent-focused therapy (AFT) for AN [19], remission was defined a priori as achieving a weight that is ≥ 95%EBW and a global Eating Disorder Examination (EDE) score within one SD of norms. Based on this definition, individuals who had prior hospitalization, were older, or had a longer illness duration were less likely to achieve remission status by end-of-treatment (EOT) [43].

Weight response

Five studies have investigated whether early response in terms of weight gain impacted remission status and other symptoms at EOT and at follow-up. In secondary analyses of the RCT comparing FBT v. AFT mentioned above [19], Accurso et al. explored the impact of timing of weight gain on psychological symptoms [18]. Weight gain significantly predicted improved psychological outcomes, including EDE scores [18]. However, the impact of weight restoration on ED symptoms diminished over the 2-year treatment and follow-up time period, suggesting that earlier weight restoration had the greatest impact on ED symptom improvement. In contrast, depressive symptoms but not ED symptoms tended to ease over time even in the absence of weight gain [18].

In this same sample, Le Grange et al. assessed the association of early weight gain and remission from AN (defined as above) [44]. Results indicated that the earliest predictor of EOT remission was gaining > 5.8 lbs by session 3. Madden et al. also studied whether early weight change in FBT predicted remission at EOT and 12-month follow-up in an RCT comparing brief hospitalization for medical stabilization and hospitalization for weight restoration, followed by 20 sessions of FBT [34]. Analysis of the combined group found that gaining 3.97 lbs by session 4 of FBT predicted higher %EBW and EOT remission. In a study where response to treatment was defined as having achieved ≥ 95% EBW following 20 sessions of manulized FBT, analyses showed that a gain of at least 2.88% in EBW by session 4 best predicted remission at EOT [28]. Finally, using data from an RCT comparing conjoint FBT v. parent-focused therapy (PFT, a separated FBT format) [22], colleagues investigated the timing and amount of weight gain that predicted remission [37]. Early responders for those receiving conjoint FBT had gained > 6.17 lbs by session 5, while early responders for PFT had gained > 5.02 lbs by session 5. Weight gain of these amounts were the best predictors of remission at EOT in these respective treatments. Taken together, these studies suggest early weight gain as an early marker of treatment response in FBT for adolescent AN and the potential to identify those who may need more intensive intervention to increase the likelihood of achieving remission, regardless of treatment type.

Weight response in subclinical presentations

While not directly tested in the context of the study, some factors impacting treatment outcome related to weight have been suggested by case studies that address EDs in ‘subclinical’ presentations. For example, in a case report of an individual who began FBT with low weight but denial of fear of weight gain (i.e., ‘subclinical AN’), weight gain itself became an important focus of progress in treatment, and an ultimate indicator of the success of the intervention as a whole. In this study, the patient was an identical twin; twin status is a non-specific predictor that was not empirically assessed in this case study, but was noted for its impact on the patient specifically in the effect of having a constantly visible image of oneself at a normal weight [48]. An FBT-based approach was also used in a case series (n = 7) describing an intervention for individuals with ‘mild’ ED presentations characterized by low weight [54]. In this 5-session, FBT-informed intervention, factors contributing to positive outcome in both weight status and cognitive symptom remission were noted as parents taking control of meals and making nutrition “mandatory”, with consistent praise and support.

Parental expressed emotion

Several studies have investigated the impact of expressed emotion (EE; measure of caregiver attitudes toward a relative with a mental or physical illness [58]) on outcomes. EE is characterized by five components: critical comments, hostility, emotional overinvolvement, positive remarks, and warmth. On the whole, EE has been shown to be relatively low in samples with EDs (e.g., [38, 41]). However, in one early study of adolescents who received family therapy for AN, results suggested that change (i.e., a decrease) in EE over the course of treatment was related to better outcomes according to the Morgan and Russell Scales [38]. Of note, although the association between levels of critical comments at baseline and treatment outcome was not replicated in later work [41], evidence of both maternal and paternal warmth was related to good outcome.

Findings from studies of EE and treatment outcomes continue to remain more consistent with its negative impact. For example, in addition to the non-specific baseline predictors described above, and counter to the null findings of Le Grange et al. [41], Eisler et al. found a significant main effect of maternal criticism, with greater criticism associated with worse outcomes [31]. In the follow-up to this study, only two factors remained as significant non-specific predictors: those with a history of inpatient treatment or maternal criticism were less likely to do well 5 years following EOT, regardless of treatment type [32]. Moreover, in the sample from the RCT comparing FBT v. AFT described above [19], Rienecke et al. explored the role of EE and family functioning on treatment outcomes [51]. Study results showed that paternal criticism predicted less improvement in ED psychopathology, and maternal hostility (a second EE subscale) predicted less improvement in general family functioning and family communication at EOT, regardless of treatment modality [51].

Building on the earlier work of Le Grange et al. [38], a study by Allan et al. also assessed changes in EE from baseline to EOT in the RCT comparing conjoint FBT v. PFT mentioned above [22]. There was a main effect of treatment type on EE at EOT such that compared to FBT, PFT was associated with a decrease in maternal criticism. Additionally, adolescents of mothers who demonstrated an increase in EE, or remained high in EE, were less likely to remit compared with adolescents for whom parental EE decreased or remained low. While this study cannot confirm why a decrease in maternal criticism was more likely to be observed in PFT than conjoint FBT, it is possible that because the adolescent is not present in PFT sessions, parents may have more of an opportunity to discuss and resolve any criticism toward their child. This explanation remains quite speculative, though.

Parental behavior

Three studies have examined parental behavior more broadly, outside of EE. Rhodes et al. evaluated the use of parent-to-parent consultations as a potential augmentation to FBT, randomizing one group to standard treatment and the other group to recieve an additional parent-to-parent consultation between weeks three and five [52]. Findings showed that parent-to-parent consultation led to small increases in the rate of early weight restoration [52]. Darcy et al. explored whether identified parental and patient behaviors observed in the first four sessions of FBT for AN predicted early response to treatment (defined by gaining at least 1.8 kg by session 4 [28]). Video recordings of FBT sessions for this study centered only on those who received FBT within a broader comparative treatment RCT [19]. Behavioral analyses showed that adolescents who were less resistant in the first two sessions and whose parents made fewer critical comments yielded the highest rates of early response (i.e., weight gain of 1.8 kg by session 4 [27]). In a third study of parental behavior, Ellison et al. examined whether key components of manualized FBT for AN predicted weight gain (i.e., parental control over eating disorder behavior; externalization of the illness; parental unity; parental consistency; or sibling support) [33]. Results indicated that parents taking control of eating disorder behavior, not criticizing the patient, and externalizing the illness predicted greater weight gain by EOT. The authors concluded that their findings demonstrated that the degree to which parents assume control over the eating disorder behavior is the central predictor of change in FBT [33].

Findings for studies examining BN

Compared to the studies on FBT for adolescent AN, considerably less work has examined FBT for BN [16]. In the first RCT testing the relative efficacy of FBT for BN v. individual supportive psychotherapy (SPT) [26], remission was defined as the absence of binge eating and compensatory behaviors in the previous four weeks, and partial remission was defined as no longer meeting diagnostic inclusion criteria. While there were several significant univariate non-specific predictors of treatment outcome, in multi-variate analyses, only higher eating concerns and higher depression scores at baseline were associated with decreased odds of meeting criteria for remission and partial remission at EOT, respectively [39]. At 6-month follow-up, higher EDE eating concerns remained associated with decreased odds of remission, and higher binge eating/purging frequency was associated with decreased odds of partial remission [39]. With an eye to the potential importance of early symptom change, across both the FBT and SPT treatment arms, a reduction of binge eating and purging frequency by 85% or more by session 6 (6 weeks) was predictive of remission (defined as the absence of binge eating and purge symptoms in the previous 28 days) at EOT [40]. Within this same sample, Ciao et al. explored factors that contributed to psychological change at EOT [25]. The authors found that individuals taking psychotropic medications prior to treatment had faster reductions in EDE eating concerns and lower overall EDE eating concerns at EOT [25]. Age did not predict changes in ED symptoms, but it did significantly predict change in self-esteem such that older adolescents had faster change in self-esteem than younger adolescents and older adolescents had better overall self-esteem at EOT.

A few studies have investigated non-specific predictors of outcome in the most recent BN RCT comparing FBT v. CBT [36]. In the main outcome report, those with male gender, or lower scores on a measure of eating-related obsessionality, the Yale–Brown–Cornell Eating Disorder Scale (YBC-EDS), or higher scores on certain subscales of the Family Environment Scale (FES; Cohesion; Intellectual–cultural orientation; Active–recreational orientation; Organization) at baseline all showed higher abstinence rates at EOT [36].

In a follow-up investigation of this RCT sample, Gorrell et al. examined motivation for change around eating-related obsessionality as measured by the Motivation for Change subscale of the YBC-EDS [59]. Results indicated a direct effect of lower motivation for change in obsessionality at baseline on higher EDE global scores at EOT, with no interaction between motivation and treatment type [59]. Of note, this study supported an association between motivation and decrease in ED cognitions but did not find that motivation contributed to the likelihood of binge eating/purging abstinence (i.e., behavior change) at EOT, across treatment types.

Finally, the latest secondary analyses of this RCT data examined the timing of response in treatment and its impact on remission status [49]. In line with earlier work by Le Grange et al. that highlighted the importance of early symptom change in adolescents with BN [40], results indicated that reduction in purging at session two and binge eating at session four were independently related to abstinence at EOT, regardless of treatment type (i.e., FBT, CBT, or SPT). An interaction effect was not tested in this study; however, slight differences emerged based on treatment group, such that reductions in symptoms most predictive of abstinence at EOT occurred slightly sooner for FBT.

Summary

Drawing conclusions about non-specific predictors across studies, factors that might positively impact outcomes regardless of the type of treatment for AN include younger age and shorter duration of illness, less prior ED treatment (specifically, inpatient), higher weight, less parental criticism, and greater parental warmth. In addition, there is strong support for the association between early response in weight gain and positive outcome for AN at EOT and at follow-up. For BN, better outcome in reducing binge eating and purging behavior appears to be uniquely related to less baseline EDE eating concern, compulsivity, and depression; an improved family environment; psychotropic medication use; higher motivation to change compulsive ED symptoms, and early symptom reduction in the course of treatment.

Mediators and moderators

As a reminder, mediators are variables that act after treatment has begun and indicate the mechanisms through which a treatment might achieve its objectives [13]. In contrast, moderators are factors that modify the effect of treatment on outcome (i.e., interaction effect) and can identify for whom treatments may work. To date, only three studies have examined mediation effects in AN, and one study in BN, detailed just below.

Mediators

Although a study detailed above [52] found that the use of parent-to-parent consultations led to small increases in body weight, the effect of this intervention on the rate of weight restoration was not mediated by parental efficacy. In another study investigating a mediation effect of self-efficacy, Byrne et al. explored data from a comparative treatment RCT [19], hypothesizing that increases in parental (FBT) vs. adolescent (AFT) self-efficacy would mediate change in these respective treatments [23]. Findings revealed that increases in parental self-efficacy resulted in significantly greater weight gain for adolescents who received FBT, but increases in adolescent self-efficacy did not impact weight gain in those who received AFT [23]. Of note, in an earlier examination of potential mechanisms of change in this same study sample, early changes (from baseline to week four) on measures of depression, self-esteem, self-efficacy, parental self-efficacy, and BMI percentile were examined as potential mediators of remission at EOT, defined as ≥ 95% EBW plus within one SD of EDE norms. Of those tested, no treatment mediators of remission were identified [43]. Finally, findings from Sadeh-Sharvit et al. mirrored the effects from prior work [23], such that change in maternal self-efficacy from baseline to session 8 mediated weight gain by session 10. Only parents receiving FBT reported significantly greater self-efficacy [53].

For BN, one study investigating potential mediators of treatment effect did not find any significant treatment-by mediator effects. However, as a main effect, this study found that change in ED cognitions mid-treatment, specifically a subscale corresponding to weight concern, led to improved likelihood of abstinence from binge eating and purging behavior at EOT and 6-month follow-up [45]. These results suggest that it is possible that FBT-BN exerts its effects by changing disordered thinking, a possibility that should be explored more thoroughly as an a priori hypothesis in future work.

Moderators

Findings for studies examining AN

Treatment alliance

Forsberg et al. [35] examined the possible differential impact of treatment alliance on remission status (i.e., remission was defined as achieving ≥ 95%mBMI and EDE global scores within 1SD of community norms, with partial remission defined as > 85th %mBMI) within data from a prior RCT that compared FBT with AFT [19]. Results did not reveal a main effect or interaction effect for alliance on full remission; although a main effect was found for partial remission, there was no alliance-by-treatment moderation effect.

Parental expressed emotion

Based on earlier work that provided foundational hypotheses for the importance of EE in family therapy for EDs [35, 38], three studies have identified potential moderating effects related to EE. Eisler et al. [31] found that youth with high EE families were more likely to achieve good clinical outcomes (per the Morgan and Russell Scales) in a separated rather than conjoint format of FBT, whereas outcomes did not differ according to treatment type among those with low EE families. At 5-year follow-up, patients from families with high levels of maternal criticism had done less well at EOT if they received conjoint as opposed to separated family therapy [32]. Examining data from an RCT described above [19], Rienecke et al. found a significant interaction between maternal hostility and treatment outcome, indicating that adolescents whose mothers displayed hostility had greater increases in %EBW in an individual treatment (AFT) as opposed to FBT [51].

Eating related obsessionality

In addition to these parental factors, there may also be a unique effect of certain patient-level features on outcomes, depending on the format of FBT. Specifically, in an optimal FBT dose study, Lock et al. compared short-term FBT (10 sessions over 6 months) to long-term FBT (20 sessions over 12 months [60]. While treatment dose was not a non-specific predictor of outcome, subgroups with more severe eating-related obsessive–compulsive thinking (i.e., YBC-EDS scores) were more successful with the higher dose of FBT [60]. Single-parent or divorced families also benefitted from the longer format.

Le Grange et al. aimed to identify treatment moderators and mediators of remission at EOT for adolescents with AN who received FBT v. AFT [43]. Of the 17 pre-randomization variables that were examined as potential moderators, only eating-related obsessionality (YBC-EDS) and ED-specific psychopathology (EDE Global scores) emerged as moderators at EOT. Participants with higher baseline scores on these measures benefited more from FBT than AFT. Perhaps a more tentative finding emerged for AN type at follow-up, such that those with binge eating/purging type were more likely to have a change in their remission status having received FBT, rather than AFT. In secondary analysis of these data, a decrease in overall eating-relating obsessionality (YBC-EDS) was significantly associated with an increase in weight for individuals who received FBT [42]. However, there was no relation between change in obsessionality scores and change in weight in the AFT group. The findings of Agras et al. diverge slightly from favoring FBT in this moderation effect [24]. Specifically, in a study that randomized participants to two family therapies, FBT v. systemic family therapy, findings suggested that those with higher baseline YBC-EDS scores gained significantly more weight with systemic family therapy than with FBT [24]. In SFT, the focus is placed on the family system rather than on behavioral features of the ED as it would be in FBT (e.g., intervening on food rituals). Therefore, although this explanation is speculative in nature, it is possible that the exploration of family patterns of beliefs and behaviors in SFT facilitates less anxiety and exacerbation of eating-related obsessionality than the more symptom-focused approach of FBT. However, in secondary analyses of this same study data, it seems as if an interaction effect occurred between comorbid symptoms and treatment type, such that those who had higher depression and obsessionality scores and received FBT were less likely to require early hospitalization [20].

Finally, a non-randomized pilot study of FBT for a sample of youth with AN, BN, and Eating Disorder Not Otherwise Specified bears mention in that the authors compared their clinical sample to a non-clinical sample at baseline in an effort to determine the meaning of any changes post-intervention and at extended follow-up [47]. In this study, patients who enrolled in the treatment study were more rigid toward themselves and had a higher impulse control and a higher drive toward perfectionism (for themselves) than the non-clinical comparison group at baseline. Further, while other aspects of general psychopathology and ED symptoms remitted 36 months after the start of FBT, perfectionism scores remained high. While no predictors of treatment effects were directly tested in this study, findings suggest that an important potential variable of interest in impacting treatment response may be trait level rigidity and/or self-oriented perfectionism.

Findings for studies examining BN

Le Grange et al. did not find any moderators of remission or partial remission at EOT when comparing outcomes for treatment with FBT v. SPT [39]. However, at 6-month follow-up, baseline EDE global scores moderated the effects of treatment on partial remission (defined as above) such that individuals with less severe baseline global EDE scores had greater partial remission when treated with FBT relative to SPT [39].

In contrast, using these same data and growth model analyses, Ciao et al. found that participants with greater baseline purging had faster change in eating concerns when receiving FBT v. SPT, whereas when baseline purging was low, participants did comparably well in both treatments [25]. Age was also a significant moderator such that younger adolescents receiving FBT v. SPT demonstrated greater change in eating concerns, whereas older adolescents showed an equal rate of change in both treatments; age did not moderate any other outcomes [25].

In another RCT for adolescents with BN, Le Grange et al. explored moderators of treatment outcomes when comparing FBT to CBT [36]. Results of this study showed that family conflict emerged as a significant moderator such that participants with lower FES Conflict subscale scores responded better to FBT than CBT, but there was no differentiation between the treatments in families with higher FES Conflict scores [36]. Similar to earlier findings in studies of AN [31], these results suggest that FBT may not be the treatment of choice for high-conflict families.

Summary

Study of moderators of therapeutic change in treatment for AN suggests that individual therapy or a separated format of FBT may be advantageous in the context of EE defined maternal hostility. Further, for those with elevated eating-related obsessionality, a longer course of treatment might be indicated, and patients with more severe ED psychopathology or eating-related obsessionality may have better weight and cognitive symptom outcomes in a behaviorally focused approach (FBT) rather than an individual approach (AFT). Of note, some work also suggests that systemic family therapy, rather than FBT, may provide superior weight gain for those with high obsessionality.

For BN, FBT encourages parental control and management of ED behaviors, while CBT or SPT for BN are principally individual therapy approaches. While the literature based on adolescent BN is limited, available findings suggest that FBT might be more suited to address certain symptoms in adolescents who are younger or with lower family conflict, relative to an individual therapy. The finding that those with baseline higher behavioral symptoms (purging) and lower cognitive symptoms (EDE global scores) fared better in FBT relative to weight gain and cognitive symptom improvement rather than in SPT is interesting, and warrants further investigation.

Discussion

Although it is clear that FBT for youth with AN and BN produces positive outcomes in a majority of cases, the process by which this treatment works, and for which subsets of the patient population this treatment should be the first choice, are far from clear. Given that a majority of our findings were specific to AN, much of this discussion addresses conclusions and implications for this subgroup. The lack of RCTs for adolescent BN limits our understanding of the workings of FBT for this patient population—a dilemma that we address below. Throughout, we discuss potential treatment implications based on the results of our review.

Several non-specific predictors were replicated across AN studies, particularly related to the severity of the illness when beginning treatment. Those starting treatment who were older or with longer duration of illness, more weight suppressed, or with more prior hospitalization or ED treatment, fared poorly in treatment regardless of receiving AFT or FBT and regardless of receiving a separated or conjoint FBT format [31, 43]. Although these findings are intuitive in nature, their relative robustness underscores the need for continued improvement in early intervention efforts whether an individual or family therapy approach is used [6]. Within FBT, PFT, or AFT, another particularly consistent finding for AN is the importance of early response in weight gain. It appears that weight gain of ~ 2 kg by the end of four weeks of treatment is most predictive of improved psychological outcomes (i.e., ED cognitive symptoms and depression [18]) and weight remission [18, 34, 37, 44] at EOT. Given the medical acuity of AN, providing patients and parents with appropriate recommendations for early weight goals should be a priority across both individual and family therapy treatments. Although comparatively limited, studies that have examined patient- and parent-level factors that predict clinical outcomes in BN suggest that, similarly to AN, early change in treatment is important [40, 49], and better prognosis in weight gain and cognitive symptom improvement may be related to lower eating concern [39] and obsessionality [36, 59]. The finding that baseline depression scores [39] and psychotropic medication use [25] impacted BN outcomes warrants future investigation.

A majority of studies focused on patient-level response, but there were additional parent-level factors that showed evidence of a direct effect on patient status at EOT for those with AN. In particular, several studies demonstrated that high EE, and in particular, parental criticism, was a non-specific predictor that negatively impacted progress in treatment [31, 38, 41, 50, 51]. When studying parental behavior outside of EE, two other studies found associations between fewer critical comments and improved weight gain [27, 33]. Criticism also demonstrated a moderating effect, leading to poorer outcomes for those who received conjoint v. separated FBT [31, 32] and FBT v. AFT [51], suggesting that for high EE families, individual therapy or a separated format of FBT (i.e., PFT) may be the better option for those with AN. Similarly, studies of BN suggest that higher family conflict (assessed with the FES) may indicate that better outcomes in behavioral symptom remission may result within in an individual, rather than family-based approach to treatment [36].

In addition to choosing a certain treatment format based on the presence of high EE or family conflict, another potentially helpful treatment tool may be psychoeducation at the start of treatment. Psychoeducation may be particularly helpful in guiding parents to align their behavior and communication with the best chance of good outcomes. For example, families and the treatment team may give the illness a name in an effort to separate the healthy part of the child from the ED, also called externalization [51]. By viewing the ED as an illness separate from the young person, parents may decrease criticism toward their child, and in turn, increase the likelihood of positive treatment outcomes. Clinicians may also provide psychoeducation around praise and positive reinforcement as a means of behavior change, as opposed to a critical response. Although it may be a challenge for some parents to remain compassionate toward the young person during difficult periods of resistance from the ED, providing guidance around how to minimize criticism and increase parental warmth may be useful in creating an environment that optimizes treatment progress.

In addition to helping parents communicate effectively with their child while managing the ED (e.g., with less hostility and greater warmth), preliminary evidence suggests that finding ways to increase parental self-efficacy in the context of treatment may support better outcomes in weight gain, particularly in FBT compared to AFT [23]. However, earlier work that tested associations between parental efficacy and either the rate of weight restoration [52], or full remission status (i.e., weight ≥ 95%EBE plus EDE scores within 1 SD of norms) [43] did not find significant mediation effects. A pilot attempt to boost parental efficacy with intensive parental coaching early in treatment demonstrated the acceptability and preliminary utility of this approach in increasing the likelihood of improved weight gain by EOT [21]. Even despite some mixed findings in facilitating faster weight gain or cognitive change (i.e., remission), the premise that increasing parental self-efficacy in supporting the young person specifically in bringing about improved weight gain warrants further investigation.

Finally, a robust finding that spans both AN and BN is the negative impact of eating-related obsessionality. Among those with AN, adolescents with greater obsessionality required a longer course of treatment [60]. Of note, Le Grange et al. [43] showed that participants with higher levels of eating-related obsessionality had better outcomes in FBT than in individual treatment whereas Agras et al. showed that participants with higher levels of eating-related obsessionality did better in systemic family therapy than FBT [24]. In the Agras et al. [20] study, both therapies (systemic family therapy and FBT) involved members of the family in treatment in addition to the individual with AN, which suggests that family involvement overall is more efficacious than individual treatment for individuals with high levels of eating-related obsessionality. Among those with BN, there was a direct effect of lower scores on eating-related obsessionality and improved abstinence of binge eating and purging [36], as well as a direct effect of motivation to change this behavior on improved EDE scores at EOT [59]. To date, no work has tested whether certain targeted treatments (e.g., exposure-based approaches) to address eating-related obsessionality may be suitable within the context of FBT for EDs but findings suggest that this may be a promising avenue for future research.

Summary and conclusion

Simply comparing two treatments can yield a certain amount of information, but an investment in identifying mediators and moderators in clinical trial research can tell us how treatment works and for whom treatment works [46]. The delineation of non-specific predictors can also guide clinical decision-making, particularly if a known factor that impacts outcomes across treatment type can be readily manipulated (either enhanced or prevented). Caring for someone with an ED is invariably challenging, and family members often experience high levels of burden and distress [61]. Given that results from this review emphasize the important role parental behavior can play in facilitating good treatment outcome, it would be advantageous to explore clinical modifactions (e.g., increased attention to parental self-efficacy) that can improve early response to FBT for AN and BN. Boosting parental empowerment to elicit early change may be particularly key, as multiple studies have identified early response (either in weight or in bulimic symptom reduction) as an important indicator of remission. Finally, across AN and BN, a promising avenue for future research will be testing methods (e.g., exposure-based approaches) to directly target eating-related obsessionality in the context of FBT. Recent work describing successes and challenges in using a family-based approach to interoceptive exposure among adolescents with low-weight EDs suggests that disgust may be an additional variable of interest in future exposure-based adaptations to FBT [62, 63]. The preponderance of study in AN evidenced in this review demonstrates the need for increased examination of factors that directly impact and moderate FBT for BN. We also acknowledge that we focus our summary of non-specific predictors solely on those that were evidenced within studies of FBT. A more developed discussion of how these factors may impact adolescents in different treatment modalities is of interest, and could be considered as a topic of future review. Further, few studies in the extant literature are able to provide specific information on individuals who discontinued treatment, and are thus lost to follow-up. An important avenue for future work will be to ensure that as a field, we attend to learning the causes of attrition, and for whom and why a treatment led to discontinuation. Finally, many of these studies include secondary analyses of RCT data, limiting the overall number of individuals that these findings represent, as well as the potential replicability of this work in novel, non-RCT settings. Despite these limitations, results of this review underscore that focusing future research on exploring moderators and mediators of FBT outcomes will improve the overall efficacy of treatment of adolescents with restrictive EDs.

Strength and limits

A strength of this review is its focus on RCTs and rigorous evidence across the literature base. The review, while comprehensive in its inclusion of all available work, is limited by the lack of meta-analytic review, which is not indicated given the modest amount of literature that currently investigates FBT for BN.

What is already known on this subject?

To date, FBT is the first-line treatment for adolescents with AN and recommended for those with BN. Accordingly, it is known that for many adolescents, FBT may be an effective treatment approach, regardless of patient- or family-level factors.

What does this study add?

In synthesizing the evidence base for non-specific predictors, moderators, and mediators of FBT across AN and BN, this review delineates important factors that should be attended to in making treatment selections.