Anxiety in anorexia nervosa and its management using family-based treatment


  • Tom Hildebrandt,

    Corresponding author
    1. Eating and Weight Disorders Program, Department of Psychiatry, Mount Sinai School of Medicine, New York, NY, USA
    • Eating and Weight Disorders Program, Mount Sinai School of Medicine, One Gustave L. Levy Place, Box 1230, New York, NY 10029, USA.
    Search for more papers by this author
  • Terri Bacow,

    1. Eating and Weight Disorders Program, Department of Psychiatry, Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author
  • Mariana Markella,

    1. Eating and Weight Disorders Program, Department of Psychiatry, Mount Sinai School of Medicine, New York, NY, USA
    Search for more papers by this author
  • Katharine L. Loeb

    1. Eating and Weight Disorders Program, Department of Psychiatry, Mount Sinai School of Medicine, New York, NY, USA
    2. School of Psychology, Fairleigh Dickinson University, Teaneck, NJ, USA
    Search for more papers by this author


Anorexia nervosa (AN) is characterized by its similarity to anxiety disorders, especially obsessive-compulsive disorder (OCD). Family-based treatment (FBT) has shown promising initial results for treatment of AN in adolescents, yet the precise mechanisms of action are unknown. We present a theoretical argument and model, suggesting that FBT may work via exposure (and habituation) to food and its consumption. First, we review the evidence for pathological anxiety in AN, and suggest a framework for identifying specific anxious triggers, emotions (fear and worry) and avoidance strategies. Second, we briefly review evidence indicating that cognitive-behavioural therapy (CBT) and specifically exposure in its various forms is most effective for treating anxiety disorders in youth. Third, we consider distinct approaches to exposure therapy based on the pattern of triggers, anxious emotions and avoidance. We conclude that the interventions utilized in FBT share clear similarities to exposure with response prevention, a type of exposure therapy commonly used with OCD, and may work via facilitating habituation to food and eating in one's natural environment. We also highlight how parents facilitate this process in between sessions by effectively coaching their children and facilitating naturalistic exposure to food and related triggers. Options for future research are considered. Copyright © 2010 John Wiley & Sons, Ltd and Eating Disorders Association.


There is increasing interest in family-based treatment (FBT) for adolescents with anorexia nervosa (AN) as the literature in support of its efficacy continues to grow. While family therapy has a long history in the context of AN (e.g. Minuchin's writings on ‘psychosomatic’ families; Minuchin, Rosman, & Baker, 1978), FBT refers specifically to an intervention developed and tested over 25 years ago at the Maudsley Hospital in England (Russell, Szmukler, Dare, & Eisler, 1987) and more recently manualized (Lock, Le Grange, Agras, & Dare, 2001), disseminated (Loeb et al., 2007), adapted for different eating disorder populations (Le Grange, Crosby, Rathouz, & Leventhal, 2007; Le Grange & Lock, 2007; Lock, Le Grange, Forsberg, & Hewell, 2006; Loeb & Le Grange, 2009) and subjected to additional randomized controlled trials (Eisler, Dare, Hodes, Russell, Dodge, & Le Grange, 2000; Le Grange, Eisler, Dare, & Russell, 1992; Lock, Agras, Bryson, & Kraemer, 2005), including several ongoing single and multi-site studies. FBT is distinctive in its ability to accomplish weight restoration on a purely outpatient basis (Eisler et al., 2000; Le Grange, Eisler, Dare, & Russell, 1992; Lock et al., 2005; Loeb et al., 2007) in as little as 10–20 sessions (Lock et al., 2005), while also being useful as a post-hospitalization intervention to complete weight gain and prevent relapse (Russell et al., 1987); follow-up studies show that both applications yield promisingly durable results (Eisler, Dare, Russell, Szmukler, Le Grange, & Dodge, 1997; Eisler, Simic, Russell, & Dare, 2007; Lock, Couturier, & Agras, 2006).

FBT has some important limitations that have been documented in the literature. Patients with AN are more likely to have a poorer outcome in FBT if their parents are overly critical of them (Smukler, Eisler, Russell, & Dare, 1985; Uehara, Kawashima, Goto, Tasaki, & Someya, 2001); studies have shown that these patients fare better in a separated format of FBT in which the patient and parents are seen separately in the same session (Eisler et al., 2007; Le Grange, Eisler, Dare, & Hodes, 1992). This suggests that FBT may not always work as well when parent level of expressed emotion is high. Further, with regard to family status, one study using a moderator analysis found that patients with AN with a non-intact family status, e.g. divorced, separated, single parent households, did better with a longer duration of FBT (e.g. 12 months) than a shorter, more typical duration (e.g. 6 months) (Lock et al., 2005).

FBT is considered a novel treatment compared to more traditional approaches. FBT actively enlists parents in the process of their adolescent's recovery, temporarily giving them the authority to make decisions about their child's food intake and facilitate nutritional rehabilitation, much as an inpatient staff might, until resistance and AN symptoms begin to abate. Importantly, control over eating is transferred back to the adolescent as soon as it is clinically indicated to do so, ideally 2 months after treatment commences. Interestingly, FBT does not directly target the cognitive features of AN, namely catastrophic cognitions about weight gain, disturbance in the experience of shape and weight and over-valuation of shape and weight; rather, it views weight restoration as paramount and necessary—even sufficient—to effect changes in these domains.

A standard course of FBT for AN consists of approximately 20 outpatient sessions divided into three phases. Phase I, when the parents are in charge of their child's eating regimen, is when the majority of weight gain occurs. Once the patient is minimally weight restored, and, importantly, exhibits a market reduction in resistance to eating and returning to a healthy weight, s/he gradually assumes more autonomy around symptom management in Phase II until the eating disorder is resolved. At that point, Phase III, which briefly deals with broader issues of adolescent development, commences. Ideally, the entire family attends FBT, and siblings are assigned a supportive role. Methods for achieving weight gain are actively discussed with the therapist in FBT, although parents are empowered to idiosyncratically select and implement particular strategies in accordance with their general parenting style, their adolescent's premorbid temperament and their practical constraints, provided parents work as a team to advance health without blaming or criticizing their child, or catering to the ‘wishes’ of the illness. The FBT therapist is, therefore, authoritative and directive in providing parents with their mission statement and guiding them to remain in the framework of FBT principles, while deferring to parental judgment about many of the details of implementation. Parents in FBT are non-punitive in their tactics and never use force. Instead, parents accomplish their goals with a judicious blend of firmness in their creation of a ‘zero tolerance’ environment for disordered eating, and empathy for their child's suffering and plight. To promote this balance, the illness is externalized and characterized as beyond the adolescent's direct control. Parents are often surprised at how effective they are in FBT, especially if they had prior failed attempts—typically in the form of trying to convince their child to eat rather than removing the option of self-starvation—to accomplish weight gain or prevent post-hospitalization relapse at home.

FBT is especially unique in its atheoretical stance concerning the aetiology of AN; moreover, FBT's framework regarding variables maintaining the illness are at most implied by virtue of its intervention strategies, and are only recently explicitly theorized (Loeb, Le Grange, Lock, & Greif, 2010). By extension, little is known about the therapeutic mechanisms of FBT. The literature shows that FBT is effective for many, but not all adolescents with AN, and that it may be more potent for the physiological criteria for AN than the psychological (Loeb et al., 2007). Additional research is needed to discover which specific components of FBT contribute to its efficacy. In this paper, we posit a novel mechanism of action for FBT, namely, that FBT accomplishes a full course of food/eating exposure in the generalized context of the affected individual's home. We begin by reviewing evidence for the overlap between AN and the anxiety disorders and argue that (1) there are unique set of triggers for those with AN that trigger anxious emotions, (2) fear, worry and associated avoidance are central to AN pathology, (3) the interventions most effective in ameliorating pathological forms of anxiety and avoidance include various forms of exposure therapy and (4) parent-facilitated weight restoration in the context of FBT shares many commonalities with exposure therapy. We conclude by proposing a model postulating that changes in anxiety over eating and associated avoidance strategies mediate outcomes in FBT.

Anorexia nervosa and its relation to anxiety

Comorbidity and anorexia nervosa

Anxiety disorders and AN share common temperamental elements and personality characteristics. For example, temperamental characteristics of perfectionism, rigidity, compulsivity, harm avoidance and trait anxiety are elevated among individuals diagnosed with AN (Collier & Treasure, 2004; Fassino, Abbate-Daga, Amianto, Leombruni, Boggio, & Rovera, 2002; Jacobs et al., 2009; Karwautz, Troop, Rabe-Hesketh, Collier, & Treasure, 2003) and are also identified as potential risk factors for its development (Fairburn, Cooper, Doll, & Welch, 1999). These traits also appear to be elevated among family members (Sohlberg & Strober, 1994; Woodside et al., 2002) suggesting a perfectionistic/anxious interpersonal environment as one possible aetiological factor involved in the illness, along with the heritability of temperamental characteristics. The existing genetic research supports these observations (Bulik et al., 2007; Halmi et al., 2005) although ongoing large-scale work will hopefully identify the relevant genes that contribute to this pathology more conclusively (Kaye et al., 2008). The perfectionistic/anxious temperament common to AN may manifest in unrelenting standards for thinness, intolerance for defects or mistakes and an often catastrophic view of how behaviours, such as eating, weight fluctuations and interpersonal feedback, will affect current state as well as future outcomes. Frequently, these perceived catastrophic outcomes pertain to weight gain and increased fatness associated with increased food consumption, but may generalize to any aspect of control over one's body including rituals around food, exercise or other compensatory behaviours. These perfectionistic/anxious temperamental characteristics predate and persist after recovery of AN (Kaye et al., 2003) and are difficult to separate from the pathogenesis of AN (Godart et al., 2003).

Further, among those with AN, there is also high rate of anxiety disorders in general (Kaye, Bulik, Thornton, Barbarich, & Masters, 2004), and the highest comorbidity exists with obsessive-compulsive disorder (OCD), social phobia and specific phobia (Bulik, Sullivan, Fear, & Joyce, 1997; Godart, Flament, Perdereau, & Jeammet, 2002; Kendler, Walters, Neale, Kessler, Heath, & Eaves, 1995; Walters & Kendler, 1995). The patterns of comorbidity among those with AN suggests possible genetic heritability of factors that contribute to anxiety (i.e. fear, worry and disgust; Bulik, Sullivan, Joyce, & Carter, 1995; Halmi et al., 2003; Klump, Bulik, Pollice, Halmi, & Fichter, 2000; Strober, 2004). Twin and family studies (Keel, Klump, Miller, McGue, & Iacono, 2005) further suggest that genetic liability for anxiety disorders and AN may overlap, supporting Strober's (2004) hypothesis, that AN, anxiety disorders and anxious temperaments are aetiologically related phenotypes, sharing inherited risk factors in common, including functional abnormalities in neural systems regulating emotional behaviour. Thus, comorbidity and heritability studies generally suggest a common anxiety-based mechanism underlying AN with core behavioural features closely related to obsessions and compulsions (Pallister & Waller, 2008).

Anxiety triggers in anorexia nervosa

Anxiety is undoubtedly a central psychological feature of AN, manifesting primarily in response to triggers in five domains including: (1) food, (2) eating, (3) interoceptive cues, (4) shape and weight and (5) social evaluation (see Table 1). The food domain includes any specific type, class or aspect of food that reliably yields an anxious response. In some instances, it may be a specific feature of food such as its symmetry, caloric content, smell, texture or the novelty of the food. Eating triggers involve any aspect of the eating process that reliably produces anxiety. These triggers may include speed of eating, the context of eating or the uncontrollability of the eating environment (e.g. choice over food, its preparation or the amount eaten). Interoceptive (physiological) triggers include the internal sensations (hunger, fullness, stomach pain, pressure, etc.) that produce anxious responses. For example, delayed gastric emptying, a symptom which has been shown to occur in AN, likely exacerbates anxiety due to increased and prolonged gastric fullness, and is in turn reinforced by food avoidance. These types of interoceptive cues are an often overlooked source of anxiety among individuals with AN, and these physical feelings manifest as feeling sick, heavy, bloated or full, among other sensations. Shape and weight triggers include any aspect of appearance or change in appearance that produces anxiety. For instance, shape and weight triggers may include the proximal trigger of current weight, the number that appears on a scale or the appearance of oneself in the mirror. However, these triggers may also include uncertainty about future weight gain and the effects of weight gain on one's later appearance. In contrast to the previous three domains, shape and weight triggers are often associated with distal consequences or exposure as opposed to proximal triggers included in food, eating and interoceptive cues. Finally, social evaluation includes any aspect of a social situation or anticipation of a social situation that yields anxiety where the individual perceives or objectively experiences evaluation. This may be in the context of eating or in anticipation of events involving food such as social gatherings. As with shape and weight triggers, a significant portion of these social evaluation triggers will be distal.

Table 1. Phenomenology of anxious stimuli and hypothetical relationship to fear, disgust, and worry in anorexia nervosa
 Forbidden foods++ 
 Food of uncertain caloric or macronutrient content++ 
 Novel foods++ 
 Lack of symmetry or imperfect food + 
 Eating in front of others+  
 Eating too quickly+  
 Eating when feeling full++ 
 Eating a certain times of day+  
 Uncertainty about quantity, duration, or frequency of eating+ +
 Digestion of food   
Introceptive cues   
 Bloating, distension, or water retention (site specific or global) ++
 Feeling fat ++
 Fullness ++
 Stomach pain ++
Shape and weight   
 Uncertain symmetry or exactness of body shape or body parts++ 
 Weight change+ +
 Reflection or pictures of oneself ++
 Weight threshold or limit+ +
Social evaluation   
 Being seen by certain people+ +
 Social situations where one ‘dresses up’+ +
 Shopping for clothes or food  +

The nature of these triggers raises some important issues. First, proximity of the trigger may vary according to the context. For instance, one's scale weight may initially be anxiety provoking because it reflects a change in weight (i.e. a proximal trigger), but also stimulate anxiety about future threats, including continuous or uncontrollable weight gain (i.e. distal trigger). These complex triggers, thus, have the possibility of being almost constantly stimulating and involve objective information (i.e. scale weight) and cognitions about the future (i.e. projected weight gain). Second, triggers may not be obvious clinically. Patients may be reluctant to disclose them, have developed such complex avoidance patterns that the trigger is unclear or be so overwhelmed by the emotion that they cannot recognize the specific trigger. Thus, the specificity and proximity of the triggers have important implications for the types of anxious emotions experienced by patients and their ability to be targeted in treatment. Table 1 summarizes some potential relationships between specific triggers and different anxious emotions.

Fear and anorexia nervosa

Fear is the emotion most commonly linked to AN, and a dysregulation in fear conditioning (i.e. learning fearful associations) has a relatively straightforward application to individuals with AN (Strober, 2004). Despite this theoretical link, limited research exists documenting the relationship between fear and other core behaviours of AN. Unlike simple phobia, where fear can be linked evolutionarily to dangerous stimuli (e.g. spiders, snakes, heights, etc.), food and eating are natural reinforcers which directly engage reward circuitry in the brain (Epstein, Leddy, Temple, & Faith, 2007). In this light, fearing food can be potentially dangerous as it is crucial for survival, unless this mechanism prevents contamination from unsafe foods. Recent evidence using psychophysiological measures suggest that patients with AN have a reduced appetitive response to food stimuli, suggesting a disruption in the motivational system to act in response to salient reinforcers (Friedrich et al., 2006). As such, the failure of food or eating cues to stimulate a drive to eat may further disrupt this system, leaving food and eating cues to be more readily associated with threat or danger. Evidence by Sysko, Walsh, Schebendach, and Wilson (2005) suggests that patients with AN report significant anxiety when presented with food of unknown quantity, type or caloric content and that they responded with significant food avoidance in this context, even after weight restoration. The failure for food to be rewarding, in conjunction with a pattern of food avoidance and other safety behaviours, may reduce access to opportunities for reversal learning and facilitate the acquisition of conditioned fear.

An extension of this fear-based model of AN, proposed by Steinglass and Walsh (2006), suggests a specific role of habit learning. In this model, the deficits in reward processing observed among those with AN serve to facilitate rigidity and adherence to habits. Thus, those with AN not only experience more fear associations, but once they have engaged in a ritual or developed an compulsion to manage this association, they have difficulty breaking this pattern. This type of dysregulation can also be observed among those with OCD and involves specific deficits in frontostriatal neural circuits (Hoenig, Hocherin, Quednow, Maier, & Wagner, 2005) that monitor, initiate and inhibit behavioural responses. Thus, increased levels of fear and the repetitive nature of these behaviours that are used to prevent, reduce or compensate for fear provide for a significant source of phenomenological overlap between AN and OCD.

Worry and anorexia nervosa

The emotional state and associated phenomena of AN also map well onto the experience of worry. In this model, triggers may signal the distal possibility of danger and this emotional arousal prepares the individual for this possibility of future harm. The hallmark features of worry include the increased subjective experience of generalized arousal and hypervigalence (Borkovec, Alcaine, & Behar, 2004). However, attempts to link worry to increased autonomic or physiological arousal have yielded little connection (Thayer, Friedman, & Borkovec, 1996), with the exception of muscle tension (Hazlett, McLeod, & Hoehn-Saric, 1994). General arousal and physiological arousal are overlapping but distinct phenomena, with the former pertaining to activation of the brain areas involved in processing information and coordinating or inhibiting behaviour and the latter involving increased energy expenditure necessary for fighting or running away (i.e. the ‘fight or flight’ response). The disconnect between subjective arousal and physiological arousal in pathological worry has lead to an increased focus on the cognitive components of this emotion. For instance, worriers experience a state of hypervigalence, which suggests a heightened state of information processing used to identify threat, but a suppression of behaviour. Subjectively, this experience equates to an increase in worry-thoughts, but a suppression of behavioural responses. In this conceptualization, the natural or phasic physiological changes in arousal such as hear rate are often cognitively misinterpreted as threatening (Wells & Carter, 1999) despite little evidence for change in the autonomic system. This cognitive misinterpretation is facilitated by initial positive beliefs about such arousal (Roemer, Orsillo, & Barlow, 2002) and its value in identifying harmful things. The consequence of misinterpretation is a further increase in general brain arousal, where the individual is primed to identify threats, but s/he experiences a suppression or attenuation of autonomic arousal. Brain arousal and autonomic arousal are often correlated, but not in circumstances where the situation requires an alert mind but not an immediate behavioural response. When the trigger is distal, this physiological state (i.e. high general arousal, low autonomic arousal) is ideal because it allows for an increased sensitivity to information relevant to the trigger while conserving the energy needed for a behavioural response. In this context, worry may also become functional to the patient (while dysfunctional from the standpoint of physical health) as it increases his/her attention and preparedness for finding certain signals associated with danger like foods of unknown caloric density or changes in appearance. Thus, from the patient's perspective, worry may be considered a positive and necessary state to prevent perceived danger and those with pathological worry may have difficulty recognizing this emotion as problematic (Behar, DiMarco, Hekler, Mohlman, & Staples, 2009). Among individuals with AN, this chronic state of worry can manifest as the hypervigilance associated with shape and weight changes or the worry that weight will ‘sneak up’ on him or her over time. Thus, natural fluctuations in weight become signals of future harm, even when the weight is considered in a ‘safe’ place.

Avoidance behaviours in anorexia nervosa

Avoidance is the natural behavioural response to anxiety. It is actually quite complex, as there are several types of avoidance, each sharing some functional link to the specific trigger and specific emotional state. For example, avoidance may manifest as a safety behaviour, a prevention strategy or a compulsion. Avoidance behaviours are extremely common among those with AN and include a wide range of behavioural responses (Halmi et al., 2003; Pallister & Waller, 2008; Wildes, Ringham, & Marcus, 2010). The nature of this avoidance is most often equated to the avoidance patterns seen with OCD (Jimenez-Murcia et al., 2007; Swinbourne & Touyz, 2007), due to the ritualized, rigid or repetitive nature of many behaviours. For example, the individual with AN may repetitively cut food into tiny pieces with hopes of either preventing immediate calorie absorption or to delay having to complete the meal. Body checking and avoidance may also occur. For example, the individual with AN with co-occurring body image concerns may avoid going to the pool for fear of being seen in a bathing suit and perceived by others as ‘fat’. Depending upon the context, these avoidance strategies have higher likelihoods of being associated with specific triggers. For instance, vomiting will almost always be a compulsive avoidance strategy designed to immediately remove the effects of a specific trigger. However, it may also be used in attempts to prevent future weight gain and thus functionally linked to worry reduction.

Functional relationship of triggers, anxiety and avoidance

Research on the phenomenology of anxious emotions breaks them down into sequential processes including the appraisal of threat, the perception of uncertainty, underlying biological responses and behavioural consequences (see Figure 1). As with anxiety disorders, threat appraisal is a defining aspect of AN pathology supported by observations of attentional bias towards threatening stimuli common among those with eating disorders (Pallister & Waller, 2008). Patients with AN will report appraising triggers from the five domains described in Table 1 (e.g. food, eating, shape and weight, etc.) as threats to their shape or weight, health and/or social status. For example, individuals with AN might interpret a piece of pizza as dangerous because s/he believes that eating the pizza will lead to the immediate effects of physical discomfort (e.g. feeling fat). Learned associations between food (e.g. calories or fat grams) and weight gain or changes in shape might facilitate this type of threat appraisal and the associated emotional response (e.g. fear, if pizza is immediately present). The outcome of this learned association is food avoidance (e.g. the safety behaviour of patting oil off pizza). This avoidance can be subsequently reinforced by weight loss (i.e. confirming the belief that pizza yields weight gain, when absence of pizza seems to yield weight loss). Thus, food avoidance limits contradictory experience or reversal learning by facilitating a successful escape from the perceived threat signalled by food. In families, this avoidance may be reinforced by parents who encourage avoidance of ‘unhealthy’ foods or inadvertently through expression of negative feelings about fatness or weight.

Figure 1.

Fear, worry, and disgust in anorexia nervosa

Little is known about the perception of uncertainty among those with AN, but it is likely linked to worry. The relevance of uncertainty to anxiety has been developed primarily in models of GAD (Lee, Orsillo, Roemer, & Allen, 2009), but has also been included in models of social phobia (Boelen & Reijntjes, 2009) and OCD (Holaway, Heimberg, & Coles, 2006). Experimental manipulations of uncertainty indicate that those who have greater intolerance of uncertainty, experience more worry when completing a given task that is high in unpredictability (Ladouceur, Gosselin, & Dugas, 2000). In the case of AN, when a task involves a high degree of uncertainty (e.g. food of unknown caloric density or predictive impact on future weight status), the consequence is increased brain arousal (see O'Doherty, Buchanan, Seymour, & Dolan, 2006). At higher degrees of uncertainty, brain arousal increases and there is increased experience of worry (Buhr & Dugas, 2009). Intolerance of uncertainty may also facilitate the reliance upon habits or rituals, as adherence to behavioural patterns are used to reduce day-to-day uncertainty among those with AN (Steinglass & Walsh, 2006). Thus, uncertainty is likely to be a key component of an anxiety trigger and perhaps facilitates worry (as opposed to fear) when salient.

The functional significance of this sequence is illustrated in Figure 1. As information about the triggers is processed, the degree to which it is threatening and signals uncertainty is appraised. This process involves retrieving learned information (e.g. caloric probability with pizza, familial or peer attitudes towards pizza, etc.) as well as identifying relevant aspects of an environmental context (e.g. quantity of pizza available). If the trigger is distal or uncertainty is particularly salient, then the individual experiences worry and the individual is more likely to engage in avoidance behaviours aimed at preventing a threat from occurring. If the trigger is proximal and/or uncertainty is low, then the individual is more likely to experience fear and engage in avoidance strategies used to manage the emotional arousal. Over time, repeated avoidance reinforces the anxiety because of limited access to competing information and anxiety generalizes to other triggers through associative learning. For example, all greasy foods or restaurants where greasy food is served are avoided or endured with a suitable degree of anxiety.

Psychological treatment of pathological fear and worry using exposure therapy

Exposure therapy as a mechanism of action in psychological treatments such as CBT

Cognitive-behavioural therapy (CBT) has received considerable empirical support for the treatment of anxiety disorders and is generally the preferred approach for pathological fear and worry. One hypothesized mechanism of action that is arguably shared across empirically supported treatments for different anxiety disorders is exposure therapy (the technique in which the patient is encouraged to directly face his/her fears or worries and habituation to the anxiety). There appears to be a growing consensus that exposure is an essential ingredient of CBT for most anxiety disorders, and that it is one of the most widely used behaviour therapy techniques in the treatment of anxiety symptoms (Mineka & Thomas, 1999; Moscovitch, Antony, & Swinson, 2009). Evidence from the child and adolescent literature on CBT for anxious youth generally support those data found in adults regarding the benefit of exposure therapy in symptom reduction (Kendall, Flannery-Schroeder, Panichelli-Mindel, Southam-Gerow, Henin, & Warman, 1997; Silverman, Kurtines, Ginsburg, Weems, Rabian, & Serafini, 1999; Prins & Ollendick, 2003; Chu & Harrison, 2007).

Of note, many CBT treatments that incorporate the specific ingredient of exposure with anxious children and adolescents also have a parent component. From a clinical standpoint, anecdotal evidence suggests that parents, in addition to participating in some sessions with their anxious child, are often helpful outside of the therapy office in facilitating between-session exposure ‘homework’ assignments assigned by the CBT therapist, either by helping the reluctant and anxious child take the next ‘step’ in his/her hierarchy or by providing support when exposures are to be completed. It is not uncommon for parents to either accompany the child to an exposure practice (e.g. in the case of phobia of heights, to a high floor in a tall building) outside of the therapy office or give encouragement, praise or a functional reward when the task is completed. An example of a model that includes parent involvement in treatment for child and adolescent anxiety that has received empirical support is the ‘transfer of control’ model (i.e. a gradual transfer of knowledge, skills and methods from therapist to parent and then from parent to child). Specifically, a RCT (Khanna & Kendall, 2009) comparing individual CBT (ICBT), family CBT (FCBT) and a family-based education control (FESA) explored the relation between use of specific parent training strategies and outcome. Results showed that ‘transfer of control’ and parental anxiety-management techniques significantly contributed to improvement on clinician and parent ratings of child global functioning within FCBT.

Different approaches to exposure therapy

Exposure to fear

As Antony and Barlow (2002) point out, exposure to feared objects and situations is both necessary and sufficient for addressing the symptom of fear, particularly in the context of treating specific phobias. Studies have found that in vivo exposure is a more effective method in reducing fear than exposure in imagination, and that in vivo exposure is effective for treating a wide range of feared stimuli (Antony & Barlow, 2002). While there is variability across studies regarding the length of treatment (with some evidence that merely a single session of exposure may be effective), the extent of therapist involvement and the age of participants, the positive effects of exposure for fear reactions appear to be robust. Research suggests that exposure in multiple contexts and to varying stimuli can reduce the rate of relapse, indicating that for maximal effectiveness, individuals should be exposed to phobic stimuli in a variety of locations and situations (e.g. Mineka & Thomas, 1999) and to varied stimuli (Rowe & Craske, 1998). Thus, fear-based exposure requires accuracy in the stimulus selected as well as exposure to the stimulus in different contexts to be most potent. This may be particularly true as the feared stimuli get more complex and difficult to identify, as in OCD. In gradual exposure, a fear and avoidance hierarchy is typically created, and the rate of exposure to the feared stimulus varies depending on the nature and severity of the fear. With more complex fear reactions (e.g. OCD, post-traumatic stress disorder, etc.), items on the hierarchy may need to be approached at a slower rate and repeated practice of each level may be necessary before the next item is attempted.

There are nuanced differences in exposure-based protocols tailored for OCD which may inform an exposure-based perspective in the treatment of AN, a disorder in which rituals around food and eating are often apparent (e.g. compulsive compensatory exercise, calorie counting, measuring portions). From a clinical standpoint, treatment for OCD is typically more intensive than treatment for other anxiety disorders in that both obsessions and compulsions are targeted. Exposure and response prevention techniques (in which the patient is exposed to the feared situation that is the target of the obsession and instructed to block the ritual that s/he feels driven to perform) may be more complicated than exposure for simple phobia. More specifically, EXRP may require both greater frequency (the optimum number of exposure sessions for OCD appears to be about 20; Steketee & Barlow, 2002), intensity (exposures may occur as often as daily), and sessions may be longer. For example, treatment may involve doing non-sensical tasks in exposure with the patient, such as touching a dirty toilet and refraining from hand-washing; this practice may need to be conducted a large number of times for habituation to occur. March and Muller (1998) suggest that satiation (an exposure-based method for habituating to obsessions by repeating them over and over again) and massed practice (similar to satiation except it is the compulsion that is practiced until performing it loses any sense of urgency) may be required when using EXRP for OCD and consequently, more time may be required to devote to these practices in and out of the therapy office.

Exposure to worry

Given that worry and its accompanying features may be present across the anxiety disorders, a wide variety of cognitive-behavioural interventions have been developed to address this specific symptom, particularly in the context of Generalized Anxiety Disorder (GAD). This may be due to the complexity of the symptom of worry, which is future-oriented and potentially abstract in nature. Treatments for GAD/worry vary in terms of the specific CBT elements they contain, and Roemer et al., 2002 have identified certain common elements to these approaches. These include self-monitoring of worry, applied relaxation, imaginal exposure, in vivo exposure, coping skills rehearsal, engagement in pleasurable activities, mindfulness-based strategies and cognitive restructuring. Worry exposure (WE) is considered to be a core element of CBT for GAD. This approach involves imaginal exposure to feared catastrophic outcomes and is certainly a component of many treatment packages for worry-based disorders. For example, Craske (1999) and Zinbarg, Craske, and Barlow, (1993) have produced a cognitive-behavioural protocol that includes WE; the exposure is then followed by a management phase in which cognitive and other coping strategies are employed in response to the imagined scene. The comprehensive treatment proposed by these authors also includes additional cognitive techniques, problem-solving and self-monitoring. Although one study to date suggests that WE may be as effective as applied relaxation as a stand-alone (e.g. no other CBT interventions used) treatment for GAD (Hoyer, Beesdo, Gloster, Runge, Hofler, & Becker, 2009), it is unclear whether exposure in isolation is sufficient to address worry as a symptom due to the abstract nature of cognitions and associated sequelae. In summary, it appears that exposure therapy may need to be either amplified or accompanied by adjunctive interventions as the precise emotional symptom becomes more complex, with simple fear considered the least complex and concrete to treat and worry the most complex and abstract.

Family-based therapy and anxiety

Given the cumulative evidence for the role of anxiety in AN, it seems reasonable to suggest that a successful treatment for AN would work, at least in part, by targeting the anxiety symptomatology that is unique to AN, including symptoms of fear and worry. Family therapy for adolescent AN and its subthreshold variants (e.g. Eating Disorder Not Otherwise Specified, EDNOS) arguably rely on some basic exposure principles found in individual treatments, although these are not explicitly articulated in the manuals. Figure 2 highlights how exposure might work in FBT. By replacing avoidance strategies with exposure to threatening (immediate and distal) triggers, the individual with AN gains access to incompatible evidence about the link between anxious stimuli and catastrophic outcomes. In particular, the exposure to eating foods of unknown quantity, calories or macronutrient content in the re-feeding process spearheaded by the parents allows for an improved confidence in that one can eat safely even when uncertain of the contents of the food. Providing functional incentives for these types of exposure also allow for positive associations to be made with food, eating and a healthy weight. As exposures continue, the individual is able to acquire a host of memories that are inconsistent with the fear associations that maintain AN. When these new memories are regularly accessed, they no longer produce increased arousal or the experience of fear, worry or disgust.

Figure 2.

A model for exposure in the context of family therapy for anorexia nervosa

On the familial level, the active role of parents in re-feeding the child may also reduce collusion with their child's avoidance. In this role, parents no longer support avoidance strategies such as food refusal, discussion of calories or refusing to eat in front of others. In the illness phase, these behaviours are often supported passively or out of fear of upsetting the child. In fact, the ‘transfer of control’ framework utilized in some family CBT treatments and described earlier is quite similar to the one that is inherent to the three-phase model in FBT, in which the parents are initially in charge and the child is gradually granted more and more independence around eating depending on the amount of progress made and skills developed.

The extent of avoidance behaviour found among those with AN also necessitates a focus on response prevention and ritual reduction. To accomplish this task, re-feeding not only needs to include exposure to anxious stimuli from food and eating, but also requires an active effort at limiting or preventing avoidance behaviours. In theory, the highest degree of habituation will occur when the correct anxious triggers are part of the exposure and there is a simultaneous reduction in avoidance behaviours. Given the necessity of eating multiple times per day to restore and maintain healthy weight, there are a rather large number of opportunities for exposure, but an equally large number of opportunities to engage in avoidance behaviours. As the family is empowered to identify and limit eating disorder behaviours, the child's access to avoidance behaviour (e.g. avoidance of being weighed) is reduced. The treatment of AN is thus arguably similar to the treatment of OCD given the frequency of sessions, out of session-practice, increased session length (the first two sessions of manualized FBT are 90 minutes) as well as exposure to both feared stimuli and ritual prevention in multiple contexts. The specifics of how exposure is incorporated into FBT are explicated below. This type of exposure has already been adapted to individual therapy for adults with AN by Steinglass, Sysko, Glasofer, Albano, Simpson, and Walsh (2010). Their model focuses on exposure to feared foods and prevention of rituals and safety behaviours and is conducted 1-on-1 with a therapist. Several studies have applied these types of exposure interventions in weight gain treatments for adults with AN (e.g. Goldfarb, Fuhr, Tsujimoto, & Fischman, 1987; Steinglass, Sysko, Schebendach, Broft, Strober, & Walsh, 2007). One limitation to these attempts may involve the inability to generalize from single exposures conducted in the context of a therapeutic setting (e.g. therapy office or inpatient program) to an individual's life. The use of family members in FBT and the child or adolescent's real life environment might improve the effects of this type of exposure by increasing the overall amount of exposure and the amount of naturalistic real world learning.

Family-based therapy and exposure

Although FBT for AN is atheoretical in terms of both assumptions about aetiology of illness and framework for treatment strategies, we propose that exposure is at least one mechanism of action inherent to this approach—the same mechanism as what has been shown to work in CBT treatments for anxiety disorders. Exposure to eating, while not explicitly described using this term, is intrinsic to the FBT approach, from the statement of purpose in Session One, in which the parents are assigned the task of feeding their child even in the face of resistance, to the family meal in Session 2, to which families bring a picnic meal with enough food to begin reversing the adolescent's state of starvation, and the therapist coaches the parents to help their child consume at least one more mouthful than s/he intended. This technique is infused in the remainder of Phase I (Sessions 3–8), which continues the ‘re-feeding’ process at home, and in Phase II, when the adolescent must still eat to gain weight, often despite extreme discomfort, but with markedly greater independence than in the initial part of treatment. Thus, the picnic meal functions as an in-session exposure exercise in which one more bite taken by the patient represents exposure to previously avoided quantity and/or quality of food. This experience ‘empowers’ parents to persist in their efforts in the home environment, with ongoing exposure to food and eating for their afflicted offspring. In the original FBT for AN manual (Lock et al., 2001), the focus of these efforts is on feeding amounts and types of food that will achieve weight restoration with the greatest efficiency. In an adaptation of FBT for the prevention of AN in high-risk adolescents (Loeb, Le Grange, & Lock, 2005), there is a greater emphasis on overall normalization of eating habits, specifically exposing patients to ‘forbidden’ foods that cause high levels of anxiety, challenging rules about eating (e.g. calorie limits) and eliminating rituals around eating (e.g. cutting food into small pieces). This version of FBT emphasizes that the task of the parents is to address all behavioural aspects of the eating disorder, including excessive exercise, binge eating and purging, weight checking rituals and body checking rituals; to address these active behaviours (as opposed to refusal behaviours, like dietary restriction), parents are ‘expected to monitor, supervise, contain and distract their child to prevent symptom engagement’. Importantly, escape extinction is also operative in FBT, especially in Phase I, as the adolescent learns that his/her parents will unequivocally and persistently implement a program for nutritional rehabilitation at every meal and snack, despite any efforts on the part of the patient to leave the table, the room or the house. Thus, with the exception of the absence of an exposure hierarchy and the allowance for distraction during ritual prevention (which represents a minor portion of the treatment), FBT closely resembles parent-facilitated, exposure-based CBT for child and adolescent anxiety disorders.

By placing parents in charge of their adolescent's weight restoration, FBT creates a naturalistic opportunity for an intensive version of exposure in the home. Although the same mechanism may be operative in other treatment modalities, such as inpatient care, hospital settings lack generalizability to the patient's natural environment and may, therefore, have compromised durability and potency of exposure effects. Conversely, the home environment provides a learning context that contains the majority of relevant triggers; moreover, parents either directly or indirectly facilitate healthy eating in other settings that are pertinent to sustained and generalized learning, such as school, restaurants, friends' homes and vacation environments. By re-feeding, parents are effectively exposing the patient with AN to a wide variety of foods, including foods that may be anxiety-provoking (i.e. high fat foods). Some treatment programs are transparent about the macronutrient content, in particular daily caloric value, of the food the patient is expected to consume, whereas this information is explicitly not calculated or shared with adolescents in FBT, partly as a function of the imprecision in naturalistic food preparation and feeding, and partly by design. In this respect, parent-facilitated eating in FBT implicitly capitalizes on the ‘unknown’ in the exposure model, which theoretically potentiates its benefits, much like asking a patient with OCD to touch something with unquantifiable but high-likelihood bacterial content such as a toilet. Some treatment protocols for AN require patients to be weighed with their backs to the scale and proscribe sharing weight data with patients during the acute weight gain phase of treatment. We argue that this both misses an opportunity for exposure to increasing numbers and contributes to avoidance. Even for those patients who weigh themselves excessively in the home environment (akin to a checking ritual in OCD), the optimal exposure exercise is tolerating the unknown between reasonable (e.g. weekly) weighing intervals. In addition, the function of frequent weight checking is to ensure that weight has decreased, or at least not increased, thereby negatively reinforcing the highly restrictive eating patterns in AN via fleeting anxiety reduction (notably, weight checking is also intrinsically positively reinforcing). Thus, simply withholding weight data in the context of a weight gain protocol will not treat checking or avoidance. In FBT, the therapist weighs the patient at the start of each session and the information obtained is explicitly shared in the family forum. In fact, session-to-session weight is charted to visually track weight gain over time. Moreover, FBT does not set a specific target weight; rather, psychological (e.g. reduction in depression and shape and weight concerns) and physiological (e.g. resumption or onset of menses) health markers, as well as reaching a weight range ‘that the patient can sustain without undue dieting (Lock et al., 2001, p. 124)’, guide the course of treatment. This, too, functions as another exposure exercise, in that most patients have difficulty tolerating this uncertainty about a ceiling weight, an unknown which is inherent in adolescent body development even when the AN remits. Finally, as commonly employed in CBT interventions for anxious youth, FBT includes the concept of the functional incentive to enhance the adolescent's motivation for engaging in these challenging tasks such as when an adolescent is told that s/he can resume a physical activity (e.g. baseball practice) that has been effectively curtailed by the illness. Further, as in CBT for child anxiety, the parents initially guide these exposure practices in between sessions by providing naturalistic meals, coaching and making sure that the between session meals get completed, effectively curtailing avoidance.

Other treatments for AN where exposure may occur in the context of the intervention

Exposure may be a mechanism not only in FBT for AN, in which adolescents are the identified patients, but in other treatment interventions for adult AN as well. Exposure may arguably take place, even if not explicated as such, on inpatient units (IP) and day hospital treatment programs (DH) for individuals with AN. Although not all IPs and day treatment programs are identical, and not all use CBT approaches, they do share a number of similarities including a biopsychosocial model, a multidiscliplinary staff, the involvement of family to at least some degree, a higher level of intensity than outpatient treatment and all IP and DH programs share the goal of weight restoration and normalization of eating behaviour as the main objective. We suggest that interventions that occur during a higher level of care for AN (i.e. inpatient, day treatment) may involve exposure to at least some degree. For example, exposure and response prevention arguably occurs several times daily as patients have meals and refrain from eating disorder symptoms. According to Olmstead, McFarlane, Carter, Trottier, Woodside, and Dimitropoulous (2010), graded task assignment is one of the bases of inpatient treatment; patients are provided with the highest level of support and the lowest level of task difficulty to start (i.e. lower calorie requirements and being confined to the unit 24 hours per day) and this difficulty is gradually increased over time. For example, as time goes on, IP/DH patients may be asked to consume larger amounts, to eat more phobic foods in the program and on meal outings and to generalize their skills with gradually lesser levels of support over time. IPs with a CBT theoretical orientation may do so more explicitly. Fairburn (2008) describes inpatient CBT-E in which CBT techniques are provided in a specialized unit for the treatment of eating disorders. The treatment lasts 20 weeks, 13 weeks of which are spent as an IP followed by 7 weeks as a day patient during which the patient resides close to the hospital and spends weekends at home. One of the techniques employed in inpatient CBT-E is ‘assisted eating’ in which an eating plan is devised for each patient that is designed to address his or her particular difficulties. Fairburn (2008) describes that this is done collaboratively and is combined with education about nutrition and energy balance (such that the patient is exposed to potentially anxiety-producing information as well as food and related triggers on an ongoing basis).

Exposure may also take place during CBT treatment for AN delivered on an outpatient basis. The high rates of relapse after hospitalization suggest that inpatient behavioural treatment alone may be insufficient to promote lasting changes in eating patterns to sustain normal weight when the structure of the hospital is ended (Wilson, Grilo, & Vitousek, 2007). Steinglass et al. (2010) suggest that CBT for AN could be enhanced through increased emphasis on behavioural techniques such as exposure and used to increase the maintenance of successes achieved in a hospital setting. Further, CBT-E is an individual focused treatment that can be used on outpatient basis, and while exposure may not be explicated in manualized CBT for AN, it is implicit in several of the interventions delivered such as the requirement of the patient to complete daily food logs, attend to daily intake and confront the feared weight gain via weekly weighing in the therapy office.

Summary and conclusions

Although FBT has long been atheoretical, we propose that exposure and habituation to food, eating and related triggers are a primary mechanism of change. This would be evident with decreased anxiety in response to anxious stimuli and elimination of avoidance behaviours present in AN. This hypothesis is based on the overlap between EXRP and FBT for AN, as well as the similarities between AN and the anxiety disorders, both in symptom presentation and aetiological factors. Clinically, it has been observed that FBT is akin to parent-facilitated exposure in a real-world setting, with the primary stimuli being food and eating, although rituals and related anxious stimuli are also targeted. FBT directly addresses avoidance behaviours. Adolescents are exposed to their weight on a weekly basis. By placing weight restoration at the forefront of treatment goals, and pushing not just for an arbitrary target weight but one that is associated with functional changes like menarche or resumption of menses, FBT regularly provides the opportunity for a pathway for recovery that generalizes to various positive functional outcomes. Parents are coached in how to facilitate exposures in a variety of salient environments, generalizing the reduction of fear to a wide range of extremely naturalistic contexts. Thus, it may be useful to explore whether we can further improve the efficacy of FBT by making the intervention even more exposure-based. If exposure is indeed a primary mechanism of action of FBT, amplifying exposure techniques in FBT may be beneficial and may enhance or further contribute to successful outcomes. We suggest some preliminary ideas of how this would be implemented in a treatment context. For example, providing psychoeducation about how avoidance of food perpetuates avoidance of fear responses may be useful. Session 1 of FBT could include components of such formal education, and with the parents explicitly assigned the task of facilitating exposure to feared stimuli. While the process may be similar to currently existing FBT, the adolescent would achieve a greater understanding of his/her emotional and behavioural reactions to feared stimuli, and a somewhat different therapy ‘language’ would be used. In a modified, more exposure-focused version of FBT, skills for coping with anxiety while confronting feared stimuli (e.g. riding the wave of the anxiety during a meal) could be taught. Family members could be enlisted as helpers to assist the adolescent with this process throughout Phase I, and food could be approached on a hierarchy with increasing levels of fear ratings. Furthermore, changes in anxiety could be systematically evaluated as a secondary outcome to weight. Such an approach may allow for more fruitful reliance on the exposure components of FBT and may further increase the efficacy of the intervention. Given the severity of AN, it is even more important to explore tools that will assist with the treatment and prevention of the illness in a maximally effective way. Further, these changes might be useful for adolescents with more treatment-resistant AN, potentially facilitating progress for those who would otherwise require a greater dose of FBT. As Lock et al. (2005) report in one trial, adolescents with severe eating-related obsessive-compulsive symptoms benefited more from longer-term treatment. An exposure-explicit family-based approach, with its more collaborative elements between parents and offspring, could also be important for late adolescents who are ‘aging out’ of FBT but who still rely heavily on their family such as college students. It would be interesting to examine whether an exposure-based approach would address AN symptoms, especially related to rituals, obsessions and compulsions in a shorter time frame. Also, FBT could formally incorporate WE interventions targeting the catastrophic thinking associated with uncertain future changes to weight, shape or social situations. Borrowing from the WE literature could provide a novel expansion of the FBT framework with more potent interventions for the chronic worry that does not subside with weight restoration. To determine whether such an approach could indeed be more effective, research is clearly needed to formally explore mediators of change in FBT. Thus this area represents a potentially exciting future direction for improving empirically supported interventions for youth with AN and opens the door to further investigations in this regard.


Tom Hildebrandt's research on this project was supported by a grant from the National Institute on Drug Abuse (NIDA): K23 024034-01A1. Katharine L. Loeb's work on this paper was supported by a grant from the National Institute of Mental Health (NIMH): K23 MH074506-01.