Facing the Challenge of Persistent Anorexia Nervosa

At a workshop session at the International Conference on Eating Disorders in Austin, TX, last May, four eating disorders specialists from the Eating Disorders Research Unit at Columbia University-New York State Psychiatric Institute tackled the challenge of understanding why anorexia nervosa (AN) can persist despite initially successful treatment.

Clues Lie in Habitual, Learned Behavior

Dr. B. Timothy Walsh, Founding Director of the Research Unit, said that despite the progress being made toward understanding AN, it is still not known why it persists despite initially successful treatment. Many factors have been theorized as predisposing factors, including female gender, genetic tendencies, environment, and our culture’s emphasis on being thin. Precipitating factors include stresses during adolescence, and perpetuating factors include habitual dieting and weight loss.

Dr. Walsh told the audience that the remarkable persistence of dieting behavior is a major roadblock to successful treatment of AN. He added that laboratory studies have shown that persistent behavior is not innate but is learned through two related but distinct processes: action-outcome learning, which is goal directed, and stimulus-response learning, or habit formation. With the action-outcome form of learning, also termed instrumental conditioning or operant conditioning, an individual learns that a certain action is likely to lead to a reward. The likelihood of repeating the action is sensitive to the reward value of the outcome, and is critical to acquisition of some new behaviors. Key neural substrates for action-outcome learning are the amygdala, the ventral striatum, and the orbitofrontal complex.

In contrast, with stimulus-response learning, or habit formation, the behavior becomes less sensitive to the value of the outcome, he said. These habits are learned behaviors, not innate, occur repeatedly, and become fixed. Once acquired, they occur automatically, almost unconsciously, and involve a structured behavioral sequence prone to be elicited by a particular context or stimulus. An example of this is the training that a Little League player undergoes to get just the right swing, often called “muscle memory.” In laboratory rats, this might be pulling a lever over and over again, until it becomes a habit. In this instance, the key neural substrates involved are the dorsolateral striatum (caudate/putamen) and the dorsolateral prefrontal cortex, he said.

Dr. Walsh noted that a major reason for the persistence of AN may be that dieting has become habitual, learned behavior. It also involves a structured behavioral sequence prone to be elicited by a particular context or stimulus. Eating behavior often involves rituals and meals, and a range of other stimuli, including negative emotion, may be enough to perpetuate the behavior. “Thus,” he added, “One hypothesis is that the eating behaviors characteristic of persons with AN begin as goal-directed learning, but then become habitual.” And, because of their habitual nature, they become highly resistant to change and therefore serve to perpetuate the disorder.

How dieting and exercise become habitual

Initially, dieting and exercise are rewarding and support action-outcome learning, according to Dr. Walsh. The reward is intermittent and behaviors are repeated and become “over-trained,” he said. And, much of this occurs during adolescence, a time that is biased toward reward, not adverse outcomes, and also a time of multiple stresses. And, finally, the pattern is enhanced by starvation.

Weight loss is rewarding in our society, due to our cultural beliefs,” Dr. Walsh told the audience. “Weight loss is particularly effective for enhancing self-esteem, especially for females,” he said, and noted that even Hilde Bruch in The Golden Cage reported that girls felt that being thin gave them power and confidence. Weight loss also becomes a way to deal with negative affect and anxiety, and is rewarding because it becomes recognized (learned) as an action that produces a favorable outcome—weight loss (action-outcome learning), he said. One more factor is that, compared to children and to adults, adolescents are more sensitive to the reward value of their actions (J Cogn Neurosci2011; 23:2123).

To produce weight loss, dieting behavior must be persistent and repeated—thus, the behavior becomes highly practiced (like overtraining by an athlete) and the habit becomes established. As dieting continues, the “rewards,” including weight loss and social praise, become intermittent. As has been shown in laboratory animals, intermittent reinforcement leads to habit formation. Then, conditioned reinforcement, or the phenomenon by which stimuli associated with rewards themselves, gains rewarding properties. Thus, the idea of dieting itself becomes rewarding.

Dr. Walsh also emphasized the factor of stress, noting that the dieting behavior characteristic of AN typically begins during a period of stress in adolescence, such as leaving home to go to camp, or to college. As Ulrike Schmidt et al. have reported (Psychol Med 1997; 27:523), individuals who developed AN reported more major difficulties with life even during the year before the onset of AN than did controls. Stress promotes habit formation at the expense of goal-directed learning, Dr. Walsh said.

Starvation also plays a role in persistent behavior. As Dr. Kenneth Carr at New York University has suggested, the neural substrate is sensitized so that rewards are more rewarding and learning during starvation appears to be more persistent.

All these behavioral disturbances may play a role in the persistence of AN and may help explain why earlier intervention—before the habit becomes established–works better, he said. The implication is that treatment must address behaviors, and accounts for the notable similarities between AN and substance use disorders. Although the AN patient is not addicted to a substance, some of the same underlying neural circuits are involved.

Targeting Persistent Problematic Behaviors

Joanna Steinglass, MD, a staff member at the Research Unit, told the audience that the technique of exposure and response prevention for AN, or AN-EXRP, may offer some help for the problem of entrenched behavior that limits food intake even after weight is restored. EXRP is a method of approaching eating-related fears systematically and without anxiety-inducing behaviors. Dr. Steinglass noted that cognitive strategies have been only of limited benefit, and studies have shown a 30% to 50% relapse rate after cognitive treatment (Eckert, 1995; Pike 2003). In contrast, behavioral strategies have been of great benefit and are the cornerstone of weight restoration treatment, she said.

According to Dr. Steinglass, AN-EXRP was developed from treatments for fear-based anxiety. (Dr. Steinglass and colleagues have shown that intake is related to pre-meal anxiety.) AN-EXRP helps patients confront rather than avoid their eating-related fears, and to work their way up from dealing with minor to major fears surrounding eating. The approach capitalizes on behavioral, experiential learning. Changing a habit requires active learning, she said, and use of repetition to learn a new set of (adaptive) eating behaviors.

Dr. Steinglass feels AN-EXRP has the potential to improve eating behavior among AN patients. The treatment is acceptable and feels relevant to patients, she said. The current manual being used by Dr. Steinglass and her coworkers is aimed at work in a structured setting; they have seen that behavioral change is associated with psychological change, consistent with the purported mechanism. The treatment success highlights the importance of—as well as the major challenges inherent in–changing habits around food. “We can make a difference even if it is very hard to change habits,” she said.

The Impact of Eating Behavior

Despite initially regaining weight and having good responses on psychological measures after treatment, among some patients with AN, unhealthy eating behaviors do not change, Dr. Laurel Mayer, Clinical Director of the Research Unit, said. That is, patients with persistent AN continue to avoid high-fat foods and their caloric intake remains lower than that of other patients, despite achieving similar body mass indexes (BMI, kg/m²). And, even among patients who have made a good initial recovery, relapse remains a major problem, she said.

Are there any predictors of successful weight maintenance by patients after recovery? Dr. Mayer cited the results of several studies, including that of Kaplan et al. (Psychol Med 2009; 39:1037). The results of this study showed that the best predictors of which weight-restored patients with AN would maintain their weight for 6 and 12 months were: (1) the level of weight restored at the conclusion of acute treatment, and (2) avoidance of weight loss immediately after intensive treatment. In this study, which compared weight maintenance among patients in a placebo group and a group that received fluoxetine, patients with BMIs >18.5 and who remained in therapy for at least 6 months did best. Forty-five percent of participants met this criterion. In other studies by Schebendach et al., 47 women with AN achieved weight restoration to a mean BMI of 20.5 (Am J Clin Nutr 2008 87:810). The pre-discharge diet records of patients with better outcomes indicated that they had an intake of foods with higher energy density and with a higher percentage of calories from fat; they also had more variety in their diets. Thus, lower density foods and less variety of foods were associated with poor outcome.

Dr. Mayer noted that the reduced caloric content is probably related to reduced energy density and percentage of calories from fat. This dietary pattern does not seem to change substantially with treatment approaches that do not make a major impact upon weight and psychopathology, she said. She also described a recent study in which she and her colleagues directly examined the caloric intake and food choices of inpatients with AN during a test buffet lunch. The patients were studied first at low weight and then after weight was restored, and were compared with healthy controls. The pattern that emerged was, again, that despite similarities in BMI and other measures, the post-treatment intake of patients with AN remained low. Although the fat intake significantly increased in patients, it was different from that of controls, as was the intake of carbohydrate.

Dr. Mayer and her colleagues confirmed that the eating behavior of patients with AN is often difficult to alter, regardless of normalizing weight and good results on psychological tests.

Olanzapine: An Effective Agent for Persistent AN?

Patients with AN have persistent behavioral difficulties that are affected by serious symptoms such as anxiety and depression, said Dr. Evelyn Attia, Director of the Research Unit. Because some medications work to alleviate these symptoms in other settings, it was thought that patients with eating disorders would also benefit. Thus far, only one medication, olanzapine, has improved weight and psychological symptoms among patients with AN, she said.

Dr. Attia reported that she and colleagues are in the midst of a year and a half multi-site clinical trial that is evaluating the effects of 16-week outpatient treatment with olanzapine. The study, which is being done in collaboration with the National Institute of Mental Health, Weill Medical College of Cornell University, the University of Pittsburgh, Johns Hopkins University, and the University of Toronto, is testing the efficacy of olanzapine, beginning with a daily dosage of 2.5 mg, and eventually titrating the agent to a maximum dosage of 10 mg. Primary outcomes will include weight gain as well as psychological symptoms associated with AN, including obessionality, mood, and anxiety. Patients will also receive manualized support, she said, and can be receiving other mediations if they are stable and if the medication is not associated with weight gain.

Previous studies by Dr. Attia and colleagues and others have charted the effects of olanzapine and other medications over 8 weeks, and Dr. Attia noted that the 16 weeks will give the researchers a chance to study the drug’s effect over a longer period. She added that thus far there are no signs that patients with AN are experiencing the metabolic changes that have been seen in other populations, such as in patients with schizophrenia.

— MKS