Premature Termination of Treatment among Adults Attending Treatment for Eating Disorders: A Critical Review

The article presents a critical review regarding the premature termination of eating disorder’s treatment among inpatients and outpatients, with the aim of identifying—emotional, psychodynamic and family-related aspects involved. The search strategy used the following MeSH terms combined by Boolean operators: “eating disorders” AND “treatment” OR “patient dropouts” OR “drop-out/dropouts” OR “attrition” OR “premature termination” AND “empirical study” OR “qualitative research”, for on Med-line/PUBMED, PsycINFO and EMBASE databases. This article follows the PRISMA Guidelines. A total of 26 studies composed this review, of which 24 were original research articles, 1 was a review and 1 a theoretical article. Only two articles applied qualitative methods analyzing categories of content obtained by in-depth interviews, three combine quantitative and qualitative methods and other three present qualitative analyses while discuss-ing quantitative studies. Further qualitative studies should be carried out to clarify meanings of dropout, premature termination of treatment, and attrition. Different expectations held by patients and by therapeutic teams, and the interpersonal difficulties of these types of patients, stand out as difficulties in constructing “therapeutic alliances”, with impacts on dropout, premature termination and attrition rates.


Introduction
The landmark research study on dropout (DO) for eating disorders (ED) was published in 1983 [1]. This phenomenon is described as a complex and multifaceted reality that challenges clinicians and researchers. During the following decades, research on the subject has evolved, especially regarding methodology, benefiting from both theoretical and review studies.
This review discusses some concepts in Eating Disorders field, methodological advances and changes that have occurred in the use of terms over the last decade. Our search included all types of ED in adults undergoing inpatient or outpatient treatment. One immediate question was the term "dropout" itself and the biases regarding its use [2] [3] [4]. In order to simplify the question of vocabulary, this review examines advances in research that focus on patient dropout, more technically termed "Premature Termination of Treatment" (PTT), to avoid the biases around the term DO noted by Sly [2], Campbell [3] and Wallier [4].
Beginning with the reviews published in the first decade of this century concerning inpatients [4] and outpatients [5], respectively, very important issues called for further research and are considered here. One pertinent question, at this point, is therefore: What should be considered new since the issues of bias over the term "dropout" and over methods to be used were first highlighted regarding each modality of treatment.
In a viewpoint article in 2009, Campbell [3] expressed concern over the rise in ED-related DO rates, which more than doubled over a period of 16 years. To retain participants in ED research trials was proven to be a major challenge, and probably an even greater issue in standard clinical settings. In this regard, we might recall the remark "We are just getting worse at the job" [3]. In view of this scenario, a new review seemed pertinent.

Method
The authors reviewed studies focused on processes of premature termination of treatment (PTT), DO and attrition by inpatient and outpatient adults in treatment for ED. The guiding question was formulated on the bases of conclusions drawn from an earlier literature review on the subject [6]: "What factors are involved in the complex phenomenon of PPT-taking into account emotional, psychodynamic and family-related aspects-among patients engaged in treatment programs?". We define "premature termination of treatment"-and its synonyms DO and attrition-as discontinuation of treatment by a patient without the team's prior consent, as opposed to unilateral discharge by the treatment team. A search strategy was then carried out on the EMBASE, PsycInfo/APA and Medline/PUBMED databases, using the following MeSH terms combined by Boolean operators: "eating disorders" AND "treatment" OR "patient dropouts" OR "drop-out/dropouts" OR "attrition" OR "premature termination" AND "empirical study" OR "qualitative research", for on Medline/PUBMED, EMBASE and PsycINFO databases. The search was conducted in February, 2016, and no specification of a timeframe neither a language filter was determined, in order to allow a broad search corresponding to the critical nature of this review. The aim was to go beyond evaluating objective results, and reveal contradictions, thus enabling a discussion of issues and progress in the area, including both research and clinical practice. The search resulted in 76 articles-seven of which showed up in more than one database, giving a total of 83 finds.
All 76 articles were then screened by reading the abstracts and judged according to the following inclusion criteria: studies which 1) focused on the topic of PTT, DO or attrition from treatment for ED; 2) dealt with samples comprised solely of adults (16 y.o. and over); 3) indicated the application of a consistent and clearly described method that would enable replication; 4) contained a consistent discussion on the topic and possibly served as a pertinent contribution to answering the guiding question cited above. Forty articles were then excluded.
Of the 36 remaining, only five were written in languages other than English (two in German, one in Italian, one in Polish and one in Icelandic, and all these had been translated into English). These 36 studies were read in full, summarized and evaluated by the authors on the bases of the same inclusion criteria listed above. During this stage, ten further articles were excluded for different reasons: one for showing an inconsistent method, three for failing to focus on the central topic, two for failing to focus on ED, two for including adolescents in the sample and two for unreliable translation (those in Polish and Icelandic).
The articles were summarized and theme units were identified and selected according to qualitative saturation criteria and their relevance to the guiding question. The flowchart ( Figure 1) below describes the search.

Results
The final 26 studies included one literature review [4], one viewpoint article [3] and 24 original research articles, only two of which were qualitative in nature analyzing categories of content obtained by in-depth interviews [7] [8]; and three described mixed-methods research [9] [10] [11]. The following Table 1 summarizes the characteristics and results of each of the studies included in the current review.

[10] Italy
Dropping out in a sample of anorexic patients and the evolution of those within the sample who abandoned treatment.

Mixed Methods
Variables related to the severity of the disease are not a predictive factor for dropping out, whereas clinical features such as depressive symptoms, hostility and presence of self-injurious compulsive behavior were quite important. Initial response to treatment seems to lead to a favorable medium-term prognosis among patients who abandoned therapy.

[25] Italy
Dropping out by EDed patients from outpatient cognitive-behavioral therapy.
Quantitative DO from ED treatment was affected primarily by factors which were related to patients' attitude and behavior. Therefore, these factors should be carefully addressed in patients with ED to improve participation in treatment and outcome. articles adopt quantitative methods, although some (three) combine the two types. Topics addressed: Several topics were found: 1) The relationship between narcissism and droppingout in Cognitive Behavioral Therapy (CBT) BT; 2) Different Evaluations of dropping-out by treatment teams and by patients; 3) Retrospective study searching for pre-treatment factors predicting DO in BN and atypical BN outpatients; 4) Analyses of whether differential treatment effects in the targeted mechanism of change and ED symptoms are associated with patterns of PTT during the treatment and follow-up phases in ED psychotherapy Randomized Clinical Trials (RCT); 5) A qualitative study of non-attenders in order to elicit the reasons behind non-engagement; 6) Definitions of high-risk patient profiles and the adoption of appropriate preventive strategies; 7) The roles of individual patients' characteristics and process-based factors for termination of outpatient treatment; 8) Identification of personality factors predictive of termination from hospitalization; 9) Comparisons of motivation, alliance and aspects of illness severity among groups upon admission, and comparison of outcomes upon discharge; 10) Early responses to treatment by completer groups, shifts in motivation and other aspects; 11) Attempts to identify predictors of dropping out obtained from a Web-based program in order to improve motivation in women with AN, BN or related sub-threshold eating pathologies; 12) Attempt at determining whether motivational interviewing in the form of a brief, pre-treatment intervention would be associated with higher completion rates of subsequent intensive treatment for ED.

Is There a More Adequate Term Than Dropout? Or A Better
Approach to the Topic? Bjork Open Journal of Medical Psychology notation of the term "dropout", especially in research. In the same direction, Sly [2], in a theoretical added article, points out that this "generally pejorative label" has "misdirected the research community into persisting with the examination of patient characteristics". He suggests replacing "Dropout" (DO) with "Premature Termination of Treatment" (PTT), a less pejorative and more neutral term.
The vast majority of the studies (18 out of 26) nevertheless used "dropout".
The use of other terms has tended to be more frequent in recent studies, but is still rare. Two studies use "attrition" and two use "PTT". As mentioned, then, the term "dropout" is still widely used. It should be noted that the problem and the solution are not in the name itself but in how it is used in research designs, and the term's connotations in the field, including negative, or pejorative,

Addressing Innovative Aspects-What about Experiences of Treatment?
Campbell [3] emphasizes that "this is an area where clinicians can have a substantial impact" (p. 242). She calls attention to the need to standardize research and thus enable better comparison and replication, and she invites to evaluate the treatment models. In 2009, the same author [3] stated that retention rates worsened between 1991 and 2006 (p. 240) and expressed concern over this gloomy scenario. Other authors also stress the need for new approaches to treatment.
Several research experiences took into account the complexity of the phenomenon, studying termination based on conclusions in the literature, such as those by von Brachel et al. [30] in Germany; Weiss, Mills, Westra & Carter [16] in Canada, and the very original experience described by Clinton, Almlöf, Lindström, Manneberg & Vestin [7] in Sweden. Studies with "internet based" interventions [29] [30], and motivational interventions as special approaches to these patients [7] [16] [29], should also be noted.
Campbell, Waller & Pistrang [23] noted that patients with higher scores on the Narcissistic Abuse Scale were significantly more likely to drop out of treatment, whereas no relationship was found between DO rates and higher levels of eating pathology. They concluded that such patients had "defensive personality style, seeing others as hostile and over-demanding, and presenting themselves as 'martyred'" (p. 279). This predictor for PTT should be analyzed and understood by teams because there is more evidence that termination is related to the tendency to give up, than to the eating pathology. Difficulties in interpersonal relationships analyzed by Campbell [3] suggest that this should be taken into account in research on individual treatments that seek to strengthen perseverance.
It is important to stress that this information indicates the importance of build- tients to their treatment and the difficulties involving these personality traits. We also call attention to the importance of studying the priority of therapeutic alliances as a goal at the beginning of treatment and that without positive alliance, other goals such as gaining weight, for example, tend to be more difficult and lead to termination by the patients.
Pham-Scottez et al. [14] also aim at identifying personality factors predictive of termination, but for inpatient anorexia treatment, concluding that:

Do More Comprehensive Studies Lead Us to a Process-Oriented Perspective?
The present review perceived a gradual trend in the research toward more process-oriented approaches to termination. Earlier studies tended to focus solely on finding predictors related to patient profiles. This procedure, combined with the implicit bias in the term "drop-out", may have led to the interpretation that some patients themselves, notably those with interpersonal difficulties and/or co-morbid personality disorders might be seen as responsible for the DO rates. There is consensus in the literature [32] [33] [34] [35] as to the high comorbidity between ED and personality disorders, and the frequent presence of these personality traits in cases without diagnosed personality disorders [32] [33] [35] [36]. It therefore seems reasonable to suppose that personality traits affecting interpersonal relationships should be considered an inherent feature of the challenge in ED treatments. Some treatment protocols take personality issues into account and consider it their responsibility to offer more effective treatment for specific ED patient profiles. This is done through programs that are processoriented and promote therapeutic alliance approaches. Such programs are probably more likely to address the current high DO rates among adults in ED treatment.
Our review detected that such approaches have been increasingly considered in more recent studies, suggesting a paradigm shift. Even when recent studies, e.g. [17] [24], focus on finding DO predictors, they tend to adopt process-based approaches that target the phenomenon of PTT in a broader way. The same is true of articles that use the term "attrition" [29], a word that tends to include other variables besides individual "resistance to treatment". Considering inherent features of ED patients as "resistance" seems to lead to less efficient treatment approaches. At the same time, this stance could excuse treatment teams from taking responsibility for considering such traits as an essential part of the

Overcoming the Dropout Paradigm
The paradigm shift allowed for different interpretations of DO in more recent studies, in alignment with the view that the earliest use of the term "dropout" in research, may have been a misuse [2] [3] [5]. Mahon [5] criticizes the tendency to identify "dropout" with a patient-characteristic profile. The follow-up study by Bjork et al. [26], was a landmark in signaling that termination of treatment need not always be taken in a negative sense, nor as a pathologic reaction.
For instance, these authors surprisingly observed that at intake, so-called "dropouts" had a less negative self-image and scored significantly higher than other patients on the self-emancipation cluster of the structural analysis of social behavior (SASB), used to evaluate self-image. These patients also showed lower intake levels of self-blame and self-hatred than the completers.
In the same vein, Vandereycken & Vansteenkiste [18] and Zeeck & Herzog [12] question "dropout" as a synonym for poor compliance or for failures in the treatment. These latter authors found that depressive patients are unlikely to drop out and that most patients who do drop out seek other services within three months.
We consider that a paradigm shift has been taking place over the last ten years, and it represents a welcome evolution in the area. This shift can be defined as a change in focus from the search for predictors of DO to factors most frequently involved in PTT. This has been critically discussed in several viewpoint articles by Sly [2] (not included in the present review, but added as an additional reference relevant to the discussion), Campbell [3], and in the reviews of Mahon [5] and Wallier et al. [4]. In addition, overcoming this paradigm means that the narrow and reductionist view that sees DO as resistance to treatment by patients is also shifting. We can now ask what challenges remain, or in what direction can advances in research and clinical practice for eating disorders be made, considering the findings regard to PTT?

What Is the Role of Qualitative Research?
We propose this question in order to inquire into the contribution of qualitative We found only two studies [7] [8] that take a purely in-depth qualitative approach, despite the numerous recommendations to employ qualitative methods.
Mahon [5]  to the fore unforeseen issues involved in PTT for ED and the formulation of better hypotheses to be investigated through quantitative research that will hopefully contribute to advances in this area.

Therapeutic Alliance and Relationships-Accessory Management or Treatment Itself?
This review has also identified the importance of the role of the therapeutic alliance in PTT and more specifically, the flexibility as a part of treatment and its management strategies. This may promote the management of therapeutic alliances, as suggested by Vandereycken and Vansteenkiste [18], who found that enhancing patient's autonomy and volition by giving them a personal choice at the beginning of hospitalization enhanced patients' autonomy and willingness to accept treatment, thus significantly reducing DO during the first weeks of inpatient treatment (p. 812). This experience also shows that a lenient approach that expands the patients' degree of choice in treatment can aid in reducing DO rates.
To quote Vandereycken and Vansteenkiste [18] (p. 182): "Although drop-out is generally considered as a signal of poor compliance In the same direction, Mahon [5] notes that softer and more flexible approaches can increase adherence and reduce DO. The conclusion is that flexibility should be considered an essential dimension of treatment for ED.
Furthermore, Clinton [20] has noted that "dropouts were characterised by high degrees of patient-therapist dissimilarity of expectations concerning the efficacy of psychodynamic interventions" (p. 34). Pointing out that dropout may increase when patients and therapists have substantially different notions of the value of insight-related interventions, Clinton [20] also recommends to discuss expectations related to treatment with the patients from the outset, focusing on particular areas of discrepancy to limit potential drop-out, supporting the relevance of greater flexibility in treatment settings.
Bandini et al. [25] suggest "the need for effective therapeutic guidelines, which take DO into account. It is important to facilitate the cooperation of the patient from the first session and during the whole treatment, mediated by a strong motivation" (p. 182), highlighting the role of therapist in treatment management in these interventions. It might be important to stress that this type of approach includes questions related to the clinical management of the interpersonal difficulties of ED patients.
Almost ten years ago, after highlighting personality factors such as impulsivity, Wallier et al. [4] concluded that future research should consider the thera-  Therefore, a therapeutic alliance should be considered treatment and be defined as a goal from the start. The present review indicates that the issue of ruptures implied in PTT should be part of the psychopathology involved, in order to be taken as a goal of the treatment. Therefore, rupture is a relevant psychological issue. If DO is to be addressed, patients must be offered proper support by both treatment protocols and treatment teams, and the therapeutic alliance must be adequately managed.
The intense fear of gaining weight is a core symptom in ED. The present review suggests it also plays an important role in difficulties in the therapeutic alliance in contexts where interpersonal relationship difficulties have already taken on a major dimension of the psychopathology involved. Perhaps it is time to consider treatment approaches that can offer better options in assisting patients to overcome their fears and interpersonal difficulties. It is possible that medically-centered approaches, focused on restoration of physical health parameters such as weight, can be felt as coercive by such patients, due to their emotional frailties, personality traits and difficulties in the field of interpersonal relationship. PTT is recognized as a multifaceted and complex phenomenon as it should be faced as a clinical problem in the area of interpersonal relationships. PTT remains a probable outcome, as we know, and we still seem unable to prevent it in many cases.

Closing Remarks
Research on premature termination of treatment for eating disorders has changed in several ways, as overcoming methodological problems, going deeper and becoming more comprehensive. In addition, further qualitative studies should be carried out to investigate the meanings of terminations. This concern is not new, but neither has it evolved very far. As part of the solution, we suggest using the COREQ reporting [41].
We also suggest analyzing the financial costs of high rates of treatment termination, improved and novel treatment approaches. In this respect, the fundamental role of therapeutic alliance and treatment teams must be stressed for managing a central feature of psychological issues related to ED, namely, difficulties in personal relationships. Tailored approaches are needed for potential terminators that will consider the likelihood of their dropping-out due to issues related to their therapeutic alliance. Therefore, it might be worthwhile to invest in the development of knowledge and practices about how to manage therapeutic alliances. Possible paths to recovery must be studied more deeply in order to develop motivational strategies for these patients.
We also propose the following questions: 1) Are multidisciplinary treatment teams ready to change their practices once they become aware of the present re- This study shows some limitations, including: 1) The absence of a standardized definition of "dropout", which has prevented the presentation of a systematic review that could add more evidence to the critical analysis of the findings; 2) The failure to include search terms such as family factors, personality factors, impulsivity, interpersonal relationship, interpersonal aspects and therapeutic alliance. We strongly recommend further research based on these keywords.