Fairburn's cognitive-behavioral model of bulimia nervosa proposes that the disorder arises from excessive reliance on weight and shape as bases for self-evaluation (Fairburn, 1981, 1997a; Fairburn et al., 1986, 1993b). Extreme concerns about the size and shape of the body, in combination with low self-esteem, lead to increasingly determined attempts to limit the quantity and type of foods consumed. These persistent efforts create physiological and psychological vulnerability to episodes of binge eating. Individuals try to undo these lapses in restraint by vomiting, taking laxatives, and imposing still more stringent exercise regimens and dietary rules; however, resort to these behaviors reinforces the bulimic cycle by triggering distress, diminishing self-esteem, renewing concern about weight and shape, and increasing deprivation. CBT is designed to address each of the principal elements in the model.

Support for the cognitive-behavioral analysis of bulimia nervosa comes from a number of sources, including risk factor research, correlational studies, and some experimental investigations (Vitousek, 1996; Cooper, 1997; Fairburn, 1997a; Byrne and McLean, 2002; Fairburn et al., 2003). In addition, patterns of treatment response are consistent with this model of symptom maintenance. Across therapeutic modalities, the reduction of dietary restraint mediates decreases in bingeing and purging (Wilson et al., 2002). The central role assigned to cognitive factors is affirmed by two findings: dismantled versions of CBT that retain its behavioral components but omit direct work on beliefs are less effective than the full treatment package (Fairburn et al., 1991, 1993a; Thackwray et al., 1993; Cooper and Steere, 1995), and the persistence of distorted attitudes at posttreatment predicts relapse (Fairburn et al., 1993a).

The standard intervention is a structured, manual-based approach that includes 19 individual sessions spanning 5 months (Fairburn et al., 1993b; Wilson et al., 1997). Treatment is divided into three stages, which are characterized by distinct therapeutic goals.

The first phase begins by establishing a therapeutic relationship and presenting the CBT model and treatment rationale. The clinician emphasizes that therapy will address all facets of the eating disorder, with particular stress in the beginning on the importance of reducing dietary restraint. Many individuals with bulimia nervosa hope that therapy will help them excise the unwanted behaviors of bingeing and purging so that they can diet more effectively and achieve a lower preferred weight. The message delivered at the inception of CBT contains both good news and bad news from these patients’ perspective. They have not become trapped in their current pattern of behavior because they are greedy or crazy or lack self-control, but because they are attempting to impose unreasonable and counterproductive standards of dietary restraint. Overeating is the normal response to food deprivation in humans and animals alike; indeed, it should not be construed as ‘overeating’ at all, but as a lawful reaction to conditions of deficit or irregular supply. The unwelcome corollary is that the two goals of eliminating bulimic behavior and achieving a higher level of dietary restraint are incompatible. In order to gain freedom from binge-eating, bulimic individuals must adopt a pattern of regular eating.

Accordingly, the intervention begins with a strong emphasis on consuming (and retaining) the regular, spaced meals and snacks that reduce susceptibility to bulimic episodes. A number of behavioral techniques are introduced during the first stage, including self-monitoring and the scheduling of alternative activities to replace binge-eating and purging (see section on Attention to eating and weight). Patients are also provided with psychoeducational material about dietary restraint, nutrition, weight regulation, and the consequences of bulimia. Some of this information is intended to correct erroneous beliefs about specific bulimic behaviors. For example, laxative abuse is usually based on the assumption that cathartics prevent weight gain by shooting food so rapidly through the intestinal tract that calories cannot be absorbed. In fact, even massive doses of laxatives eliminate only a small fraction of the calories consumed during binges (Bo-Linn et al., 1983). Other psychoeducational material is helpful in underscoring the CBT model or decreasing concern about the consequences of giving up bulimic behavior. For example, while most patients fear that the lessening of dietary restraint will cause substantial weight gain, the evidence shows that the great majority of patients gain little or no weight after a successful course of CBT (Fairburn, 1993, 1995; Fairburn et al., 1993a).

In the second stage, the emphasis on regular eating patterns continues; in addition, patients are asked to start reintroducing excluded foods into their diets and to resume eating in settings (such as restaurants) and social situations that they may have been avoiding. Cognitive restructuring techniques are used to analyze thinking patterns that help sustain symptoms, including dichotomous judgments about eating, weight, and personal performance. Patients are encouraged to review the evidence for and against their beliefs in order to reach reasoned conclusions that can be used to guide their behavior.

The final stage focuses on relapse prevention strategies. The patient reviews the tactics that she has found especially helpful during treatment, anticipates high-risk situations, and outlines an individualized ‘maintenance plan.’

Consistent with general CBT principles, therapists combine a directive, problem-solving focus with a collaborative style throughout therapy. Clients must take an active role in achieving symptom control through collecting data, generating solutions, and practicing new behaviors. An important goal is for patients to develop the skills and self-confidence that allow them to ‘become their own therapists’ during and after the time-limited course of CBT (Wilson et al., 1997).

The empirical examination of CBT for bulimia nervosa has followed a thoughtful and systematic sequence of investigation. The approach was developed through clinical experimentation with some of the first bulimic cases reported in the literature, and was guided by a clear, concise model of symptom maintenance. It was translated into a manualized intervention and tested in more than 25 controlled trials in a variety of settings against a number of well-chosen alternative modalities. Within 15 years of the time the approach was proposed (Fairburn, 1981), research had begun to examine therapeutic mechanisms, combined and sequential treatment approaches, generality of effects across different patient populations and providers, and dissemination strategies. On the basis of this impressive body of evidence, CBT has earned the status of treatment of choice for bulimia nervosa (Wilson, 1996; Agras, 1997; Compas et al., 1998; Wilson and Fairburn, 1998; American Psychiatric Association, 2000; Cochrane Depression Anxiety and Neurosis Group, 2000; Fairburn and Harrison, 2003).

In the reduction of both core and associated symptoms, CBT is clearly superior to wait-list control conditions, and matches or exceeds all other examined psychological interventions, including psychodynamic therapy, supportive treatment, IPT, BT, stress management, exposure and response prevention, and nutritional counseling. Reviews indicate that CBT results in mean reductions of 73–93% for binge eating and 77–94% for purging; total remission of symptoms is attained by one-third to one-half of CBT-treated patients when results are analyzed on an intent-to-treat basis (Craighead and Agras, 1991; Wilson et al., 1997; Wilson and Fairburn, 1998; Fairburn and Harrison, 2003; Thompson-Brenner et al., 2003). Consistent with the theoretical model that informs the treatment approach, CBT has also been shown to reduce dietary restraint, decrease depression, enhance self-esteem, and produce positive changes on global measures of adjustment and social functioning (Fairburn et al., 1991; Garner et al., 1993; Wilson et al., 1997; Wilson and Fairburn, 1998).

One clear advantage is that CBT works quickly in comparison with other psychotherapies (Wilson and Fairburn, 1998). For example, Wilson et al. (1999) determined that CBT had already produced most of the improvement evident at posttest by the third week of treatment. The rapid gains associated with CBT are also enduring. Reductions in binge eating and purging are characteristically maintained at 6–12-month follow-up assessments (Wilson et al., 1997). In the longest follow-up reported to date, 71% of the participants who had achieved full symptom remission by the end of the active treatment period remained symptom-free an average of 5.8 years later (Fairburn et al., 1995).

Although the positive effects of CBT are robust and stable, it is also well-established that no more than 50% of patients recover completely, while a substantial minority obtain minimal symptom relief from participation in this mode of therapy. Across studies, the half or more of patients who do not attain full recovery through CBT continue to binge an average of 2.6 times per week and to purge 3.3 times per week at treatment termination (Thompson-Brenner et al., 2003). Little is known about the factors that influence response to CBT, as the few variables that appear to be associated with outcome in individual studies are seldom replicated across them (Wilson and Fairburn, 1998). The most consistent predictors of poor response are comorbid personality disorder and high baseline frequency of bingeing and purging. In addition, the strong relationship between symptom reduction during the first few weeks of CBT and eventual outcome provides some rational basis for continuing the standard approach or considering modified, supplementary, or alternative interventions (Wilson et al., 1999; Agras et al., 2000). Unfortunately, there is scant empirical basis for anticipating that patients who fail to respond to CBT will derive greater benefit from a different treatment approach (Wilson et al., 2000).

Clinical lore holds that CBT is appropriate only for relatively ‘simple’ cases of bulimia nervosa such as that represented by Sharon, but contraindicated for complex, severe, and/or comorbid symptom pictures, exemplified by the description of Emily. Certainly, Sharon is far more likely than Emily to be symptom-free after 19 sessions of CBT; however, it is a fallacy to conclude that CBT is therefore the wrong treatment for a patient with Emily's symptom profile. Such reasoning holds only if an alternative approach is known to support superior outcomes in comparable patients (Hollon and Kriss, 1984; Wilson, 1995, 1996). No such evidence exists in the treatment of bulimia. A reasonable course for Emily's case might be clinical experimentation with modifying CBT by increasing its intensity and/or duration, and by including additional components to address difficulties with affect regulation (e.g., Segal et al., 2002), self-harm (e.g., Linehan, 1993), and substance abuse (e.g., A. T. Beck et al., 2001; Parks et al., 2001) (see discussions in Wilson, 1996; Wilson et al., 1997 and Fairburn et al., 2003).

More generally, commentators from both within and outside the CBT orientation have suggested that modifications to the basic approach might provide greater benefit to a broader range of patients. Critiques of CBT for bulimia nervosa usually highlight three overlapping limitations. First, the standard intervention is narrowly focused on specific eating disorder symptoms, paying minimal attention to interpersonal issues or generic concerns about self-worth (Hollon and Beck, 1994; Vitousek, 1996; Garner et al., 1997). CBT for other disorders characteristically extends to a wider range of topics as relevant to individual cases, and there is no obvious clinical justification for restricting the scope of CBT for bulimic patients. Second, the manual-based approach relies predominantly on behavioral tactics, with fairly cursory attention paid to the exploration of beliefs and less to the role of affect (Hollon and Beck, 1994; Meyer et al., 1998; Ainsworth et al., 2002). A third and related concern is that CBT appears less effective in reducing patients’ focus on weight and shape than in eliminating the behavioral symptoms of bingeing and purging (Wilson, 1999). Greater change might be obtained through more emphasis on cognitive work; in addition, closer focus on body image issues through therapist-assisted exposure and other targeted CBT techniques could be beneficial (Tuschen and Bent, 1995; Rosen, 1996; Fairburn, 1997a; Wilson et al., 1997; Wilson, 1999; Fairburn et al., 2003).

On the basis of these observations, Fairburn et al. (2003) recently proposed a revised model of the maintenance of bulimia nervosa and outlined a broader approach to its treatment. The new formulation is intended to supplement rather than replace the original model, principally through the inclusion of four additional foci, if indicated for individual patients: perfectionism, low self-esteem, mood intolerance, and interpersonal difficulties.

If the standard course of manual-based CBT is not sufficient for all patients, the full treatment may be unnecessary for some (Wilson, 1995; Wilson et al., 1997). The pressures of cost containment and the scarcity of trained specialists have stimulated efforts to find economical, readily disseminable treatments for bulimic patients. Several streamlined interventions consistent with the CBT approach have been evaluated. One of these involves self-help manuals (Schmidt and Treasure, 1993; Cooper, 1995; Fairburn, 1995) designed for direct use by bulimic individuals with or without guidance by a professional or paraprofessional (Fairburn and Carter, 1997; Birchall and Palmer, 2002; Carter, 2002). Another possibility is an abbreviated CBT intervention that can be applied in primary care settings (Waller et al., 1996). Additional alternatives are group CBT that includes all components of the standard model but can be delivered economically to multiple patients (Agras, 2003; Chen et al., 2003) or a shorter group series that presents the psychoeducational content covered in the full approach (Olmsted and Kaplan, 1995).

Each of these approaches has been examined, and the same general conclusion appears to apply across all: truncated and/or group-administered variants of CBT provide substantial benefit to a subgroup of patients, but typically yield lower rates of improvement and remission than the complete individual approach (e.g., Olmsted et al., 1991; Treasure et al., 1994, 1996; Thiels et al., 1998; Mitchell et al., 2001; Palmer et al., 2002; Chen et al., 2003). The appropriate use of self-help and/or group psychoeducation may be as initial interventions in a stepped-care model, with individuals who fail to respond offered a subsequent course of the full treatment; conclusions about brief CBT and a group version of standard CBT are more tentative pending the accumulation of additional data.

In contrast to CBT, the use of IPT in the treatment of bulimia nervosa is not predicated on an elegant, disorder-specific model of symptom maintenance. The approach was first applied to this population because it suited the purposes of clinical researchers who needed a short-term, well-specified modality that had minimal conceptual or procedural overlap with CBT. IPT fulfilled these specifications admirably, and was initially selected for comparison with CBT and BT in a trial conducted by Fairburn et al. (1991).

Although IPT was in some senses chosen as a foil, it would be unjust to both the researchers and IPT itself to assume that it was meant to be a ‘straw treatment’ that would make the results of CBT appear more impressive by contrast. IPT was already established as an effective therapy for depressed outpatients (Weissman et al., 1979; Elkin et al., 1989), and would have been a poor bet for investigators seeking an attention-placebo condition. Moreover, if there is no elaborated ‘interpersonal theory’ of bulimia nervosa, there is substantial evidence that interpersonal issues are implicated in the disorder. Family problems, sensitivity to criticism, conflict avoidance, and concern about social presentation are all prominent in bulimic patients, and binge episodes are often precipitated by interpersonal stress. Therefore, IPT offered a credible alternative treatment that was manual based and approximately matched to CBT in format, yet focused on different issues, employed different techniques, and presumably worked through different mechanisms.

In its original formulation for depressed patients, IPT was also designed as a ‘research treatment’ that gave structure to the emphasis many clinicians place on their clients’ relationships (Klerman et al., 1984). Drawing on Sullivan's (1953) interpersonal approach, IPT makes few assumptions about the variables that produce specific symptom patterns. The rationale for its use across diagnostic categories and clients is that all psychiatric disorders develop and persist in a social context, and are often ameliorated by resolving interpersonal problems. IPT focuses on patients’ current social relationships rather than attempting to address childhood issues or enduring personality characteristics (Weissman and Markowitz, 1994).

The adaptation of IPT for bulimia is outlined in several descriptive articles (Fairburn, 1993, 1997b, 2002b; Apple, 1999; Wilfley et al., 2003). Therapy is delivered in 19 sessions over 18–20 weeks, scheduled twice weekly in the first month, weekly for the subsequent 2 months, then in alternate weeks. This represents a slight reduction and rearrangement of the sessions specified for work with depressed patients, in order to align the format more precisely with CBT and BT. With a few exceptions, the intervention for bulimia is otherwise identical to the approach detailed in the IPT manual for depression (Klerman et al., 1984; Weissman et al., 2000). Two changes in content are prescribed: the initial sessions involve an analysis of the chronology and context of eating disorder symptoms; thereafter, discussion of disorder-specific material is actively discouraged to maintain the focus on interpersonal issues. The first of these modifications is entirely consistent with the principles of IPT for depression; however, the second represents a departure that was intended to sharpen the distinction between IPT and CBT/BT (Palmer, 2000). In other applications, IPT does not exclude direct work on current symptoms. As reformulated for bulimia, IPT avoids any reference to eating patterns, compensatory behaviors, and weight concern between the first and last few sessions of the treatment course. If these topics are raised by patients, therapists are instructed to try to ‘limit patients’ discussion of their disordered eating behaviors to 10 seconds or less’ (Apple, 1999, p. 717).

The first phase of IPT is completed in three or four sessions, which are devoted to a thorough assessment of the interpersonal context surrounding bulimic symptoms. Therapist and patient trace the historical association between significant events, relationships, mood, self-esteem, and changes in eating patterns and weight. This review is used to create a ‘life chart’ that illustrates the connection between experiences and symptoms. The assessment also includes identification of interpersonal triggers for episodes of binge-eating (Fairburn, 2002b).

On the basis of the information collected and organized during this initial phase, therapist and patient identify one or more problem areas that will become the focus of the next stage of treatment. Paralleling IPT for depression, these are drawn from four categories: grief reactions, interpersonal role disputes, difficulties arising from role transitions (such as moving out of the parental home or starting work), and interpersonal deficits. For bulimic patients, the most common targets are role disputes (relevant for 64% of clients) and role transitions (identified in 36%); issues related to grief (12%) or interpersonal deficits (16%) are less often implicated for this population (Fairburn, 1997b).

With reference to the case examples outlined earlier, a natural focus of IPT for Sharon might be her impending transition from living alone to forming a new household with her fiancé. In view of Emily's profound social isolation, therapy might focus on her interpersonal deficits or unresolved issues in her conflicted relationships with her estranged husband and family. Unfortunately, just as the standard CBT intervention may not be effective in Emily's case, patients presenting with longstanding interpersonal deficits are difficult to help through IPT as well (Fairburn, 1997b).

In the final phase of treatment, the patient and therapist review progress to date, discuss remaining difficulties, and anticipate and plan for possible future problems. At this point, patients are encouraged to identify any changes in eating-disordered symptoms over the course of therapy, and to note their linkage to improvements in relationship patterns (Apple, 1999).

In the study that prompted the adaptation of IPT for bulimia nervosa, the approach appeared moderately effective when status was assessed at the end of the treatment period (Fairburn et al., 1991). IPT and CBT were associated with comparable reductions in binge frequency and depression, as well as equivalent improvements in social functioning; however, CBT was more effective than IPT (or BT) in modifying attitudes about weight and shape, and produced greater reductions in dietary restraint and vomiting frequency. Data collected after a 1-year closed follow-up period revealed some surprising trends (Fairburn et al., 1993a). While participants in the BT condition were doing quite poorly, those who had received IPT had caught up to the CBT-treated patients so that the groups had become statistically indistinguishable across all indices of outcome. A similar pattern of results was obtained in a subsequent multisite study (Agras et al., 2000). CBT again outperformed IPT at posttreatment assessment; once more, no differences were discernible by follow-up as a function of continuing improvement in participants previously treated with IPT.

The unanticipated efficacy of IPT at follow-up seemed to raise important questions for models of bulimia nervosa. Clearly, some individuals were able to accomplish significant (if slightly delayed) changes in their eating-disordered behavior even when therapy paid little or no attention to the specific symptoms that prompted them to seek treatment. The different temporal pattern of change also supported the view that these modalities worked through alternative mechanisms. Fairburn speculated that IPT might facilitate change by increasing patients’ feelings of self-worth, indirectly lessening their tendency to evaluate themselves on the basis of body shape and weight (Fairburn, 1988, 1997b; Fairburn et al., 1991). This hypothesis appeared consistent with the lag between the active treatment phase and the achievement of symptom control—perhaps it simply took more time for individuals to translate improvements in self-esteem into modifications of their eating behavior.

The intriguing intimation of a delayed treatment response or ‘sleeper effect,’ however, was disconfirmed by further analyses of data from the second study (Wilson et al., 2002). In fact, the same proportion of patients who remained symptomatic after IPT or CBT continued to improve during the follow-up period; thus, IPT appeared to ‘catch up’ to CBT simply because there were more symptomatic patients left at the end of IPT who were still eligible for a late shift toward recovery. The conjecture that the two treatments work through different mechanisms was contradicted as well. The mediators believed to account for improvement in IPT—improved self-esteem and interpersonal functioning—showed no relationship to symptom changes in either IPT or CBT. Instead, both treatments decreased bingeing and purging through reductions in dietary restraint, with CBT appearing more effective than IPT at posttest because it accomplished this objective more rapidly.

Although it remains unclear how IPT works, the equivalence of IPT and CBT by follow-up supports the conclusion that these modalities are comparably effective (Fairburn, 1993). On that basis, either treatment is a defensible first-choice alternative for bulimia nervosa, with the selection between them influenced by patient and therapist preference, availability of expertise, and the importance of prompt symptom control. Another possibility is that IPT might be reserved as a second-line treatment for patients who do not achieve satisfactory results through CBT. The sole study that has investigated the merits of such sequential treatment, however, was not supportive (Mitchell et al., 2002).

Dominant schools of family therapy have taken a keen interest in anorexia nervosa and had considerable impact on the field (e.g., Minuchin et al., 1978; Selvini-Palazzoli, 1978; Dare, 1985; Dare and Eisler, 1992). Minuchin's structural model identified anorexia as the prototype ‘psychosomatic’ disorder, in which family dysfunction (including enmeshment, overprotectiveness, rigidity, and conflict avoidance) is expressed by the symptom-bearing child. The treatment approach featured ‘family lunch sessions,’ during which the therapist observed family dynamics and carried out on-the-spot interventions. Parents were urged to unite and force their anorexic child to eat—in some instances by holding her down and pushing food into her mouth with the therapist's encouragement and support. The method was widely publicized through the distribution of filmed sessions, which for a time were shown routinely in abnormal psychology classes throughout the United States. Many undergraduates found these disturbing to view—as did most eating disorder specialists. The results Minuchin claimed to have achieved, however, were every bit as dramatic as the sample sessions: nearly 90% of patients were said to be doing well at follow-up (Rosman et al., 1978). Critics have questioned the rigor, representativeness, and even the veracity of these data; many experts also dispute the assumption that family dynamics are uniform or causal in anorexia nervosa (e.g., Yager, 1982; Rakoff, 1983; Vandereycken, 1987).

The most influential contemporary form of family therapy for anorexia nervosa is the ‘Maudsley model,’ which combines elements from both structural and strategic approaches (Dare and Eisler, 1995, 1997; Lock et al., 2001). Following Minuchin, therapists direct parents to assume control over the anorexic child's eating behavior and orchestrate crises during meal sessions to empower them in this role. The Maudsley approach is more closely aligned with strategic family therapy, however, in favoring an ‘agnostic’ view of etiology. Family members are charged with responsibility for the anorexic individual's recovery, but explicitly exonerated from blame for her disorder.

As applied to adolescent patients, the Maudsley approach involves 10–20 family sessions spaced over 6–12 months. The ‘conjoint’ format specifies that all family members—siblings as well as parents and the anorexic child—should be seen together. A recently published manual (Lock et al., 2001) describes the implementation of conjoint family therapy (CFT) in detail. As discussed below, a form of ‘separated’ family therapy has also been devised and tested.

CFT is divided into three phases, with transition from one to the next dependent on the achievement of specific objectives. The approach is highly structured—indeed, almost scripted, particularly in the early sessions. The key therapeutic maneuver in Phase I is to reestablish parental authority in the family system, with particular reference to asserting control over the anorexic child's eating and weight. Several tactics are adopted to further this goal. Using a sympathetic but authoritative style, the therapist works to heighten the parents’ level of anxiety by underscoring the severity of their daughter's condition. In an ‘almost ritualistic’ fashion (Lask, 1992), clinicians are advised to assume a ‘portentous, brooding, and grave manner’ (Lock et al., 2001, p. 208) when they greet parents; in the first session, they should ‘concentrate on the horror of this life-threatening illness,’ (p. 52) warning parents that ‘something very drastic has to happen for you to save [your child's] life’ (p. 47).

Another recommended technique is the externalization of anorexic symptoms. The disorder is construed as an alien force that has overtaken the patient so completely that she is incapable of controlling her own behavior—and therefore critically in need of her parents’ forceful intervention. This ‘benevolent dissociation’ is intended to assuage parents’ guilt about using strong measures to combat the illness, as well as to convey support for the patient as an individual distinct from her disorder (Lock et al., 2001; Russell, 2001).

Parents are asked to bring food to the second treatment session, and coached by the therapist to find ways of compelling their daughter to eat. Outside of therapy, they are advised to keep her under parental supervision 24 hours a day during the first few weeks, temporarily arranging leaves of absence from school and work to accomplish the task of refeeding. External control tactics are also applied to other eating-disordered behaviors; for example, parents are told to lock the refrigerator and cupboard doors if necessary to prevent binge-eating, and to inform neighborhood pharmacies that their daughter must not be allowed to purchase laxatives.

Phase II begins when the patient is surrendering relatively consistently to the demand to increase her intake and weight, typically after 3–5 months of weekly family sessions (Lock et al., 2001). During this stage, the therapist encourages a gradual fading of close supervision and reinforces the patient's return to age-appropriate activities and levels of autonomy. The message is that anorexia nervosa deprived her of the right to make her own choices, as she had functionally regressed to a child-like incompetence and dependency; now that she is beginning to improve, she is entitled to reclaim more control over her life in this and other domains. Parents are asked to focus their attention on strengthening the marital relationship; all members of the family are enlisted in reestablishing ‘intergenerational boundaries’ between the parental dyad and the children.

Phase III is initiated after the patient demonstrates her capacity to maintain a stable weight without high levels of external control. This stage involves several sessions spaced 4–6 weeks apart. Parents are provided with information about normal adolescent development, and the emphasis on fostering independence continues; however, the therapist also works to instill fear about the possibility that symptoms could resurface, in order to ensure continued parental vigilance to the risk of relapse.

A modified form of family therapy is recommended for adult patients. It is inappropriate (as well as impossible) for the family or partner of an adult patient to seize control of her eating behavior—clearly, the 29-year-old Amanda's parents cannot be advised to hold her down and push food into her mouth. Instead, CFT for older patients focuses on restructuring family relationships so that the eating disorder no longer dominates the picture. This application has not been described in the same detail as family therapy for adolescents and, as discussed below, appears to be much less effective. It should also be noted that in the case of adult patients, a decision to implement CFT does imply certain assumptions about the significance of family dynamics in the maintenance of symptoms. It may well be possible to take an ‘agnostic’ view of etiology when using the approach with adolescents, as the tactic of enlisting parents as treatment agents can be justified on purely pragmatic grounds. Because adults necessarily retain principal responsibility for the management of their own symptoms, however, a preference for working with such individuals through a family unit that may no longer reside together requires a theory-based explanation. Many individual therapists might schedule a few sessions with the spouse, parents, or friends of an older anorexic patient (generally because she requests it); however, a therapist who elects to see her primarily or exclusively with her family members present is making a much stronger statement about why she became or remains ill.

Family therapy is the most extensively researched treatment for anorexia nervosa, contributing at least one cell to half of all controlled trials of psychotherapy. Only one of these studies found family therapy clearly superior to a comparison treatment, and the effect was restricted to patients who carried particularly favorable prognoses by virtue of their young age and brief duration of illness. At least for this subgroup, however, no alternative treatments have been demonstrated to work better than some version of family therapy. On the basis of the accumulated evidence, family therapy is the sole intervention that currently meets the standard of an ‘empirically supported treatment’ for adolescent anorexia nervosa.

The strong association between recency of onset and the likelihood of positive response to family therapy was evident in the first trial conducted by the originators of the Maudsley approach (Russell et al., 1987). In that study, CFT was much more effective than a supportive, dynamically oriented individual therapy with a subset of patients who had become anorexic before the age of 19 and been symptomatic for less than 3 years. The effects of initial treatment were still discernible at 5-year follow-up: 90% of those who had received CFT were classified as ‘recovered,’ while 45% of the patients originally allocated to individual therapy remained anorexic or bulimic (Eisler et al., 1997). In contrast, family therapy was neither effective nor differentially effective for other subsets of patients who had a longer history or a later onset; in fact, there was a tendency for the latter group to do better in individual therapy, although few patients responded well to either treatment.

Subsequent research has confirmed the importance of short duration as a predictor of response to CFT. Indeed, the data suggest that the window for successful intervention is even narrower than the 3-year period used to form subgroups in the Russell et al. (1987) study. In a project carried out by the same investigators, all participants had been anorexic for just 2–36 months, with an average duration of 12.9 months and a mean age of 15.5 years (Eisler et al., 2000). Even within this extremely restricted range, there was a significant correlation between how very recent onset had been and treatment outcome in either of two forms of family therapy. Patients who were doing well at 1 year had been anorexic for a mean of 8 months at the inception of treatment, compared with 16 months for those with intermediate or poor outcomes. Another historical variable was also linked to treatment response. When patients who had received repeated prior treatment on an inpatient or outpatient basis were compared with those obtaining therapy for the first time, the contrast was again sharp: 73% of the treatment veterans did poorly in family therapy, while only 19% of the novices failed to improve.

The significance of these data is underscored when we consider their implications for the sample cases of Chloe and Amanda. There is no reason to anticipate that 29-year-old Amanda would respond to family therapy; indeed, she falls into the category of adult patients for whom individual therapy appeared slightly—if rarely—more effective in the initial study. Chloe, however, seems to match all specifications for the empirically supported treatment of CFT: the onset of her disorder was squarely in the middle of adolescence, she has been anorexic for just over a year, and she is still living at home in an intact (and concerned) family. If we try to extrapolate her prognosis from the figures provided by Eisler et al. (2000), however, Chloe's outlook appears less sanguine. At a duration of 13 months, she falls right in between the group of patients for whom family therapy was found to be effective and those for whom it was not. The fact that she has already been a treatment failure elsewhere is ominous as well. Even though the nonspecific therapy, nutritional counseling, drug treatment, and brief hospitalization to which she has been exposed may not represent particularly promising interventions for her disorder, their presence on her treatment record consigns her to the category from which only one-fourth of patients will emerge as successful responders to family therapy.

This pattern could have a number of plausible explanations, and the alternative possibilities hold different implications for how we should view the results of family therapy. In the early stages of anorexia nervosa, patients may not yet have crystallized their identities around the disorder, and it is conceivable that it is easier and more efficacious to exercise external control over the expression of symptoms in such cases. It is also possible that early intervention appears to work better in part because we end up counting among our ‘treatment successes’ the subset of patients whose disorders would be self-limiting with or without professional (or parental) intervention.

Although it is clear that family therapy is effective principally for briefly ill anorexic patients, it has not been established that it differs from other forms of treatment in this regard. It may well be that alternative approaches are comparably constrained. Certainly, the general pattern of correlation between duration and outcome obtains across most of the treatment trials and uncontrolled follow-up studies reported in this field; however, few have analyzed data with sufficient precision to confirm or disconfirm the stark association between months of symptom persistence and treatment response evident in Eisler et al. (2000).

The view that family therapy is preferentially indicated for the treatment of recent-onset anorexia nervosa depends on the demonstration that it resolves such cases faster or more completely or in a higher proportion of patients than alternative therapies. The answers to those questions are less clear than the wide disparity found by Russell et al. (1987) suggested. The pronounced superiority of family over individual treatment in that initial study—evident even with samples of 10 and 11 patients per cell—certainly offered a compelling basis for further investigation. Commendably, proponents of family therapy did continue to examine the method they advocate; inexplicably, however, they stopped comparing it with anything else. The Maudsley group never tried to replicate their remarkable finding with larger samples or alternative forms of individual therapy. Instead, they embarked on a series of intramural studies comparing different formats and intensities of family therapy (le Grange et al., 1992; Lock, 1999; Eisler et al., 2000), as if the case for its superiority over other modalities were already amply documented.

Three different teams of investigators did take up some of the basic questions bypassed by the Maudsley group, with mixed results. Two studies revisited the question of family versus individual treatment. When individual therapy was operationalized in the form of an ‘ego-oriented’ approach in one trial, family therapy (combined with some CBT elements) appeared slightly but not durably more effective with a sample of adolescent patients who had been ill for less than a year (Robin et al., 1994, 1995). When the individual treatment condition was CBT, both modalities yielded equivalent and fairly positive results with adolescent and young adult patients, with no trends favoring either approach for any subgroup (Ball, 1999). A third study found no differences between eight sessions of family therapy and eight sessions of group family psychoeducation when these were provided adjunctively in connection with inpatient treatment for adolescents (Geist et al., 2000).

Interestingly, the intramural research to which the Maudsley investigators turned did identify one mode of treatment that appears superior to CFT: an alternative format for delivering the same Maudsley message. In two studies, the standard ‘conjoint’ approach was compared with ‘separated’ family therapy (SFT), in which parents and the anorexic child were counseled in different sessions (le Grange et al., 1992; Eisler et al., 2000). In parent meetings, the therapist provided advice consistent with the parental control strategies of CFT, while anorexic patients received supportive individual therapy that could include discussion of both family and eating/ weight issues. On theoretical and clinical grounds, SFT was clearly expected to prove weaker than CFT, as it offered no direct opportunities to observe and intervene in family dynamics, did not include meal sessions, and did not involve siblings.

Across both trials, however, there was a trend favoring SFT over CFT. In the second and larger project (Eisler et al., 2000), nearly twice as many patients achieved a ‘good’ outcome through SFT (48% versus 26%), while fewer than half as many patients did poorly (24% versus 53%). This effect was accounted for by the subset of families in which parents frequently directed critical remarks toward the anorexic child. When subgroups of cases high and low in expressed emotion (EE) were compared, SFT was significantly and strikingly more effective than CFT with high EE families, benefiting 80% versus 29% of the patients treated; for low EE families, no trend favored either format.

It is commendable that the Maudsley investigators put themselves in a position to learn that their assumptions about what works best for anorexia nervosa were mistaken. To date, however, their response to these unusually decisive results has been disappointing—both for this specialty area and the evidence-based treatment movement as a whole. Their own findings indicate that SFT is a slightly better treatment option overall and a dramatically better one for patients with the misfortune to come from contentious families. Yet the Maudsley group recently published a manual that strongly advocates the less effective conjoint format (Lock et al., 2001), and is using that approach rather than SFT in ongoing research (Lock, 1999). At present, then, a curious anomaly attaches to the empirical standing of family therapy for anorexia nervosa. CFT is at once the best-supported treatment for recent-onset adolescent patients—and one of the very few ‘active’ modalities in the field that has been found inferior to an alternative approach. A therapist who was committed to practicing validated treatments should indeed adopt the Maudsley model for cases matching the profile of Chloe; ironically, however, he or she should not adhere to the manual written to disseminate the approach, as it describes a version of family therapy that has been shown to disadvantage a sizeable subgroup of the patients to whom it is applied.

More broadly, it should be noted that no study has yet examined the merits of the specific type of family intervention espoused in the Maudsley model. Across orientations, most specialists advocate working with parents when treating individuals in the young-to-mid-adolescent age group, sometimes using principles and techniques quite different from those associated with the Maudsley model. Only direct comparisons can illuminate which of these should be preferred. At present, all that can be stated with some confidence is that seeing family members together does not contribute to positive outcomes, and is contraindicated for a subgroup of particularly vulnerable patients.

Psychodynamic approaches do not fit comfortably into the format used to profile other modalities in this chapter. The difficulty is that there are too many alternatives to cover, none of which is dominant in the eating disorder field. They differ so markedly that there would be multiple ‘theories’ and ‘treatment descriptions’ to summarize under the section subheadings, while the ‘empirical evidence’ that has been collected bears only on the specific variants tested.

In this specialty area, the designation of a treatment approach as ‘psychodynamic’ conveys little information about the conceptual model that guides it or the techniques it subsumes. Drive-conflict, object relations, and self-psychological models disagree about why people become anorexic and how they should be helped to recover (Goodsitt, 1997). According to different accounts, self-starvation is a defense against oral impregnation or aggressive fantasies (Waller et al., 1940/1964; Masserman, 1941; Freud, 1958; Szyrynski, 1973), a reaction to maternal impingement and/or hostility (Masterson, 1977; Selvini-Palazzoli, 1978), or a desperate attempt to organize and empower the self (Bruch, 1973; Casper, 1982; Goodsitt, 1985, 1997; Geist, 1989; Strober, 1991). Therapists may be advised to interpret the meaning of the patient's symptoms (Thoma, 1967; Sours, 1974, 1980; Crisp, 1980, 1997) or to offer her a healing relationship with a caring adult (Goodsitt, 1997); alternatively, both of these prescriptions may be misguided and perhaps downright dangerous (Bruch, 1988). Depending on the source consulted, eating and weight issues should be addressed, ignored, or delegated to someone other than the therapist. Some treatment proposals specify a 25-session course of outpatient psychotherapy (Treasure and Ward, 1997a), while others advocate 6–24 months of residential care (Story, 1982; Strober and Yager, 1985). Clearly, any attempt to generalize across such diverse models, methods, and formats would be uninformative.

Only a few characteristics help to distinguish psychodynamic approaches from alternative methods (although none is universal across or exclusive to this group of therapies). Psychodynamic therapists are more likely to endorse the view that ‘it's not about eating and weight,’ to explore the origin of symptoms, to focus on longstanding conflicts or deficits, to encourage the expression of emotion, and to highlight the therapeutic relationship. They are, in general, less likely to emphasize the provision of facts about the disorder, to give advice about the management of eating and weight, to examine disorder-specific beliefs, to suggest extra-therapy activities, or to use an active, directive style during sessions.

Another factor complicating the review of psychodynamic models is that they are routinely combined with other approaches in the treatment of anorexia nervosa. Virtually all therapists find their accustomed modes of practice challenged by the distinctive features of this disorder, and many venture outside familiar frameworks in search of better alternatives (Garner and Bemis, 1982; Casper, 1987; Tobin and Johnson, 1991; Palmer, 2000). The identity crisis seems especially acute, however, for those who practice nondirective forms of therapy. To a greater extent than family therapists or CBT therapists, clinicians whose primary affiliation is psychodynamic tend to favor a ‘hyphenated’ approach when working with anorexic patients, borrowing elements from family systems, CBT, interpersonal, experiential, and medical models. Most are (commendably) reluctant to overlook patients’ current health, weight, eating behavior, and patently false beliefs, while tracing the origins of their difficulties to early developmental deficits. In response, some adopt a pragmatic eclectic approach, importing symptom management strategies from other orientations to put alongside the techniques they prefer. Others modify psychodynamic therapy itself to suit the distinctive features of patients with this disorder.

For example, the influential theorist and therapist Hilde Bruch (1973, 1978, 1988) outlined a causal model of anorexia nervosa consonant with her psychoanalytic training, yet cautioned that traditional psychodynamic therapy was ‘singularly ineffective’ and ‘potentially harmful, even fatal’ when applied to these patients. She recommended using a more direct ‘fact-finding treatment’ that enlisted the patient as a ‘true collaborator’ in the effort to identify and challenge specific ‘false assumptions or illogical deductions’ (Bruch, 1962, 1978, 1985). The therapeutic style that Bruch described as more effective with this population bears a striking resemblance to Aaron Beck's cognitive therapy (A. T. Beck, 1976; A. T. Beck et al., 1979; J. S. Beck, 1995; Greenberger and Padesky, 1995) and adapted versions designed for use with anorexic patients (Garner and Bemis, 1982, 1985; see subsequent section on CBT).

Only a few psychodynamic interventions for anorexia nervosa have been outlined in detail, including the hybrid approaches labeled ‘feminist psychoanalytic therapy’ (which also incorporates elements of CBT; Bloom et al., 1994) and cognitive analytical therapy (CAT; Treasure and Ward, 1997a). Still fewer have been both specified and examined in controlled trials, including CAT and focal psychoanalytic psychotherapy (FPP; Dare and Crowther, 1995).

CAT is a time-limited dynamic therapy (Ryle, 1990) that is described as ‘uniquely positioned between [the] extremes’ of symptom focus and insight orientation (Bell, 1999, p. 36). As applied to anorexic patients, the format involves 20 weekly sessions followed by 3–5 monthly follow-up visits (Treasure and Ward, 1997a; Dare et al., 2001; Tanner and Connan, 2003). Working collaboratively, the therapist and patient identify target problems and analyze the ‘traps,’ ‘snags,’ and ‘dilemmas’ through which these are maintained. Therapy also examines interpersonal patterns, termed ‘reciprocal roles,’ which are traceable to early relationships and form the background for the patient's present experience. This information is mapped on to a visual schematic called the ‘sequential diagrammatic reformulation’ that depicts connections between the individual's symptoms and her relationships and self-concept, and becomes the basis for ongoing monitoring and discussion in therapy.

FPP is a more traditional approach based on Malan's work (1976, 1979) and adapted for anorexia nervosa by Dare and Crowther (1995). In the context of research, treatment is delivered weekly for 1 year. Therapy is organized around a ‘focal hypothesis’ that links the patient's internalized representation of significant people in her past to her evolving feelings for the therapist and the function of her symptom in current personal relations. Anorexia nervosa is viewed as a means of gaining a spurious sense of control, in an effort to manage patients’ central fear of the vulnerability experienced through closeness, and as a powerful method of soliciting care from others. Analysis of transference and countertransference phenomena is prominent in this mode of treatment. Weight is monitored by a nurse and reported to the therapist, who charts and discusses weight changes with the patient in connection with other developments as they unfold in psychotherapy.

Six controlled trials have included at least one cell of psychodynamic therapy. Unfortunately, three of these describe the modalities in minimalist terms, saying little more than that one treatment condition was ‘psychodynamically oriented.’ In view of the diversity of models, such brevity means that we have no way to determine which treatment principles are being supported or invalidated by the results. One of these investigations found a psychodynamic condition inferior to dietary counseling (Hall and Crisp, 1987), one inferior to family therapy for adolescent but not late-onset cases (Russell et al., 1987), and one equivalent to inpatient care and superior to treatment as usual (Crisp et al., 1991).

The remaining three trials, two of which included CAT, provide more interpretable information. As noted in the section on family therapy, Robin et al. (1994) reported that ego-oriented individual therapy was slightly less effective than family therapy with young, recent-onset anorexic patients. A pilot study comparing CAT to an ‘educational behavioral’ treatment found no differences between conditions at 1-year follow-up (Treasure et al., 1995). Most patients in both groups gained weight but only about a third of each achieved a good outcome.

A subsequent study compared CAT, FPP, family therapy, and low-contact routine treatment (Dare et al., 2001). All interventions were provided on an outpatient basis to a sample of adult patients who would be expected to carry a poor prognosis as a function of their average age (26 years), extended duration of illness (6 years), low weight (BMI of 15.4), and history of prior treatment (79% overall, 43% inpatient). When reassessed a year after treatment initiation, most patients in all conditions remained underweight, with an average BMI of 16.5. A few group differences attained significance: patients assigned to routine treatment were more likely to require hospitalization than those in any of the three specialized therapies, and gained less weight than patients in focal psychoanalytic or family therapy (but not CAT). On an intent-to-treat basis, 28–37% of those allocated to the specialized treatments had recovered or significantly improved, versus 5% of those getting routine care; however, half to two-thirds of the former and three-fourths of the latter were still doing quite poorly. Moreover, because of design inequities, even the obtained differences between the specialized therapies and routine treatment are not clearly interpretable as support for any one (or all three) of these specific modalities. Widely varying lengths and densities of therapy were used across conditions; routine treatment was delivered by psychiatric trainees with less overall and disorder-specific experience, who transferred their anorexic patients/subjects to another therapist when rotated to a new service every 6 months. As a result of these discrepancies, differences in outcome cannot be attributed to particular models versus the general advantages of treatment amount, continuity, and expertise.

While the available evidence is not strongly supportive of any form of psychodynamic treatment, it should be stressed again that no intervention of any kind has so far yielded satisfactory results with the patient group most often included in these trials: older adolescents or adults with established anorexia nervosa. Psychodynamic models have been highly influential, however, in shaping the ways therapists conceptualize and treat eating disorders. Although few clinicians practice therapy precisely as specified by CAT or FPP, psychodynamic principles inform the work of most. Bruch's characterization of symptoms as a desperate struggle for a self-respecting identity, Crisp's depiction of flight from psychobiological maturity, and Goodsitt's emphasis on the importance of a healing relationship have all been incorporated into dominant clinical views of the disorder. Above all, psychodynamic approaches underscore that the eating and weight symptoms that command our attention do not define the scope of problems faced in the treatment of anorexia nervosa.

A cognitive-behavioral framework for understanding and treating anorexia nervosa was initially described by Garner and Vitousek (Garner and Bemis, 1982, 1985; Garner et al., 1997). The model overlaps substantially with Fairburn's (1981, 1985) analysis of bulimia nervosa, reflecting the CBT perspective that these disorders share many core features and maintaining variables. At the center of both is the premise that personal worth is dependent on the size and shape of the body. This dominant idea spins off a host of specific irrational beliefs, conditions a characteristic set of fears, and prompts stereotyped avoidance behaviors. Over time, anorexic and bulimic individuals begin to process information in accordance with predictable cognitive biases and respond to increasingly eccentric reinforcement contingencies. In addition, anorexic individuals incur the hard-wired consequences of semistarvation, which also contribute to the entrenchment of the pattern (Garner and Bemis, 1982).

To a greater extent than other models, cognitive accounts stress the positively reinforced and ‘organizing’ functions of anorexia nervosa, postulating that these explain its most unusual features better than the avoidance-based functions that are also present and influential (Garner and Bemis, 1982; Slade, 1982; Bemis, 1983; Vitousek and Hollon, 1990; Vitousek and Ewald, 1993; Wolff and Serpell, 1998; Fairburn et al., 1999b; Vitousek, 2005). These distinctive features include a sense of ‘specialness,’ moral certitude, competitiveness, and positive identification with the disorder (Vitousek, 2003). Because the sum of these distinctive elements seems to explain the distinctive resistance to change in this disorder, CBT for anorexia nervosa is organized around efforts to address them.

The CBT approach has been outlined in a series of papers that describe different components of the treatment package (Garner and Bemis, 1982, 1985; Garner, 1986, 1997; Orimoto and Vitousek, 1992; Pike et al., 1996; Garner et al., 1997; Vitousek et al., 1998; Wilson and Vitousek, 1999); however, no treatment manual combining this material is available. The format specifies an extended course (1–2 years) of individual therapy, supplemented with family sessions if indicated. One-to-one therapy is considered essential to the delicate work of addressing motivational issues and idiosyncratic beliefs. The long duration reflects the time required to engage reluctant patients as active participants in the change process, as well as to help them reach normal weight and recover from the persistent after-effects of semistarvation.

The CBT model has a strong bias toward accomplishing weight restoration on an outpatient basis whenever possible, to maximize patients’ sense of responsibility for decision making and minimize the risk of reactance to external control. In some cases, inpatient or partial hospital treatment may be essential; guidelines are available for using CBT principles in such settings (Bowers et al., 1997).

Treatment phases parallel the three-stage sequence common to CBT and IPT for bulimia (and to some extent, family therapy for anorexia): an initial phase for engagement, provision of rationale, and beginning steps toward change; a middle phase for focused work on identified maintaining variables; and a concluding phase for consolidation and relapse prevention (Garner et al., 1997). In practice, however, the progression of CBT for anorexia nervosa is less fixed than these divisions imply. The general movement is from developing motivation to modifying eating and weight to examining disorder-specific beliefs to focusing on broader aspects of self-concept; however, all of these issues are necessarily addressed throughout therapy, and many CBT techniques target multiple areas simultaneously.

In descriptions of CBT for anorexia nervosa, considerable space is devoted to suggested strategies for engaging patients’ interest in the prospect of change and then translating that interest into action. Four emphases are identified as crucial to the promotion of change (Vitousek et al., 1998). The first involves the nonconfrontational use of psychoeducational information to help the patient reassess the perceived risks and benefits of her symptoms and reconstrue their meaning. The second is an affirmation of the experimental method of CBT, which casts each proposed step in therapy as an opportunity to gather information rather than an irrevocable commitment to change. A third key theme is an emphasis on exploring the functional effects of patients’ choices, rather than challenging their rationality or validity. A substantial portion of the first few sessions of therapy is typically devoted to helping the patient develop a list of the advantages and disadvantages of her eating disorder, phrased in her own terms (Vitousek and Orimoto, 1993). Samples of the kinds of material provided by patients such as Chloe and Amanda are included in Table 15.1 (see discussion in the subsequent section on Attention to motivational issues). The serious attention given to the perceived advantages of symptoms can be disarming to patients who are accustomed to being warned about the dangers of their behavior. The therapist is advised to acknowledge these benefits without minimizing or disputing them. At the same time, she or he introduces a theme that will recur throughout treatment: the disadvantages the patient is experiencing are inextricably linked to subnormal weight; however, it may be possible to secure the positive effects at lower cost through alternative means.

The fourth theme is an exploration of philosophical issues that bear on patients’ attachment to symptoms and fear of change (Vitousek et al., 1998; Vitousek, 2005). CBT advocates working through each patient's personal values to convince her that her anorexic way of life violates key principles that are even more fundamental to her sense of identity. For example, Amanda echoed the common anorexic view that denying herself desired foods and working out several hours a day were testimony to her strength and self-discipline—characteristics she valued highly in herself and others. In fact, after years of anorexia nervosa, adherence to these rigid, fear-driven patterns of behavior represented the path of least resistance. For Amanda, the truly brave and difficult choice would be to violate her anorexic system of rules by eating forbidden foods or defying the impulse to exercise.

Because many anorexic individuals are passionately committed to their beliefs, work with these patients can seem more analogous to the conversion of a member of the National Rifle Association to a gun control advocate, or a religious fundamentalist to Unitarianism, than it does to psychotherapy with a depressed or anxious patient. It follows that attempts to dispute anorexic beliefs logically are seldom successful—and are not advised by CBT experts in this area. The treatment approach does make use of the conventional CBT techniques of cognitive restructuring and prospective hypothesis testing; for example, the therapist and patient might collaborate in designing an experiment to check out the patient's belief that other people respond to her more favorably when she loses weight. Contrary to stereotype, however, CBT does not assume that anorexic patients will give up their symptoms once the therapist points out their ‘errors in thinking’ and challenges their irrational beliefs. Instead, therapists are encouraged to draw on a blend of factual, functional, and value-related material to enlist patients in reexamining the relationship between anorexic symptoms and their own goals and ideals.

The processes of dietary rehabilitation and weight restoration are not carried out in isolation from the rest of CBT but integrated with the ongoing examination of patients’ beliefs. Initially, patients are encouraged to follow individualized, structured meal plans, gradually introducing larger amounts and avoided food types. Meal planning is usually incorporated into regular sessions and conducted by the primary therapist. In vivo therapy sessions may be used to assist patients with particularly challenging situations, such as grocery shopping, eating in restaurants, or trying on new clothes (see section on Attention to eating and weight).

During the course of therapy, attention gradually shifts from the focal symptoms of anorexia nervosa to more general aspects of self-concept and interpersonal relationships. It is not necessary to switch paradigms in order to address these issues, as CBT offers mode-consistent principles for work on this level. In the later stages, patients are encouraged to experiment with new strategies for achieving their goals, new sources of positive reinforcement, and new standards for gauging personal worth.

Paralleling the general state of treatment research across the eating disorders, there is a vast gap between the empirical standing of CBT for anorexia nervosa and for bulimia nervosa. During the first 7 years after the approach was proposed, several case studies (Cooper and Fairburn, 1984; Garner, 1988; Peveler and Fairburn, 1989) and one small controlled trial (Channon et al., 1989) were reported. Very recently, four comparative studies have been completed and others are in progress. Only a few tentative observations are warranted pending publication of these data.

Two recent controlled trials compared CBT with nutritional counseling. The first of these broke down after 100% of the 10 patients assigned to nutritional counseling dropped out of treatment and refused further contact; in contrast, 92% of the 25 patients allocated to CBT persisted to completion of the 20-session series (Serfaty et al., 1999). Although statistical analyses of outcome were precluded by the mass defection of patients from nutritional counseling, it was reported that those who had received CBT showed significant changes in BMI and on measures of eating disorder and general symptoms.

Findings from a second study confirmed the pattern of differential attrition from these treatment conditions (Pike et al., 2003). This investigation was designed as a ‘relapse prevention’ trial, with patients randomly assigned to 50 sessions of individual CBT or nutritional counseling and medical management after completing inpatient weight restoration. The study included a severe sample of 33 adult patients (average age 25 years) with longstanding anorexia nervosa (7.5 years). Fewer patients in the CBT condition terminated prematurely (27% for CBT versus 53% for nutritional counseling) and more met criteria for ‘good’ outcome at the end of treatment (44% versus 7%).

Another trial compared CBT, fluoxetine, and combined treatment conditions with a sample of 108 partially weight-restored anorexic patients (Halmi, 2000). Final results have not been reported, but the interim data suggested that CBT, alone or in combination with medication, also conferred some protection against premature termination compared with medication alone. The extremely high dropout rates already evident for all conditions at mid-treatment are troubling, however, and inconsistent with the patterns seen in other investigations.

Collectively, these findings support the tentative conclusion that CBT does further at least two of the goals it was expressly designed to fulfill: higher rates of initial engagement and treatment persistence in these notoriously ‘resistant’ patients. While it is tempting to attribute these desired effects to the collaborative style and motivational emphasis of CBT, such inferences are premature. Because each of these studies compared CBT with a nonpsychological intervention, they suggest only that psychotherapy—perhaps, but not necessarily, in the specific form of CBT—produces better outcomes than nutritional counseling and/or drug treatment alone (Vitousek, 2002). Ironically, the choice of weak comparison treatments has made it difficult to gauge the efficacy of CBT. In order to determine whether CBT confers any benefits beyond its apparently greater capacity to retain patients, researchers must identify comparison treatments that serve the same objective at least equally well, so that sufficient numbers of patients are willing to remain in the alternative condition.

To date, only two investigations have examined the effects of CBT relative to other forms of psychotherapy. Both used abbreviated versions of CBT (20–25 sessions) with patients who were underweight or partially weight restored. One of these found CBT equal to BT and treatment-as-usual with adult patients (Channon et al., 1989); a second obtained positive and equivalent results using CBT or family therapy with adolescents and young adults (Ball, 1999).

In view of the fact that no form of psychotherapy has been shown to work better than any other form of psychotherapy for adult anorexic patients, how should clinicians evaluate the merits of CBT relative to other treatment options, pending the availability of more instructive data? Several considerations recommend the approach as a defensible interim choice (Vitousek, 2002; Fairburn and Harrison, 2003). One argument is the documented success of a related approach for bulimia nervosa. Because anorexic and bulimic patients share many beliefs and behaviors, it seems likely that at least some of the same strategies will prove effective in treating them. Another putative benefit of CBT is its integration of direct work on eating and weight with attention to motivational issues—a characteristic that may become increasingly appealing as outpatient services are forced to take on more responsibility for the weight restoration phase of treatment.

A disadvantage of CBT is the fact that this complex intervention requires considerable training to deliver. In addition, all CBT experts concur that the treatment must be lengthy—not the 18–25 sessions used in most trials to date, but something on the order of 40–60 sessions in the usual case of established anorexia nervosa (Garner et al., 1997; Fairburn et al., 2003; Pike et al., 2003). Such extended outpatient treatment clearly requires justification through data showing that it reduces the need for expensive inpatient care and subsequent services to a greater extent than brief interventions.

Recently, this area has been invigorated by new proposals for shifts in emphasis in the basic CBT model, offered both by its originators and other CBT experts in the eating disorder field. Like the initial approach, however, these suggested revisions are based on clinical experience rather than accumulated evidence about the strengths or weaknesses of existing models. As we know very little about how well the ‘traditional’ CBT approach to anorexia nervosa works, it remains a matter of conjecture how it might be improved (Vitousek, 2002). Some specialists have suggested that the ‘traditional’ model is too narrow, paying insufficient attention to interpersonal issues and ‘deep’ aspects of the self (Leung et al., 1999); others have speculated that the model is too broad, allocating unnecessary attention to interpersonal issues and ‘deep’ aspects of the self (Fairburn et al., 1999b). Specific aspects of the approach have also been recommended for closer focus, notably work on the extreme need for self-control (Fairburn et al., 1999) and the connection between symptoms and values (Vitousek, 2005). Because the few data available already indicate that many anorexic patients will fail to achieve full recovery with CBT (or any other tested modalities), thoughtful modifications to the existing approach should be examined through systematic case study or small group designs.