|Home | About | Journals | Submit | Contact Us | Français|
Eating disorders and obsessive-compulsive disorder (OCD) commonly co-occur, but there is little data for how to treat these complex cases. To address this gap, we examined the naturalistic outcome of 56 patients with both disorders, who received a multimodal treatment program designed to address both problems simultaneously.
A residential treatment program developed a cognitive-behavioral approach for patients with both OCD and an eating disorder by integrating exposure and response prevention (ERP) treatment for OCD with ERP strategies targeting eating pathology. Patients also received a supervised eating plan, medication management, and social support. At admission and discharge, patients completed validated measures of OCD severity (the Yale-Brown Obsessive-Compulsive Scale—Self Report [Y-BOCS-SR]), eating disorder severity (the Eating Disorders Examination-Questionnaire), and depressive severity (the Beck Depression Inventory II [BDI-II]). Body mass index (BMI) was also measured. Paired-sample t-tests examined change on these measures.
Between 2006 and 2011, 56 individuals completed all study measures at admission and discharge. Mean length of stay was 57 days (SD = 27). Most (89%) were on psychiatric medications. Significant decreases were observed in OCD severity, eating disorder severity, and depression. Those with bulimia nervosa showed more improvement than those with anorexia nervosa. BMI significantly increased, primarily among those underweight at admission.
Simultaneous treatment of OCD and eating disorders using a multimodal approach that emphasizes ERP techniques for both OCD and eating disorders can be an effective treatment strategy for these complex cases.
Cognitive-behavioral therapy (CBT) is effective for various psychiatric disorders (Barlow 2008). However, CBT’s efficacy has largely been demonstrated in clinical trials that recruit individuals with one primary disorder, use disorder-specific CBT protocols, and exclude or standardize concomitant medication. Such trials do not mirror routine clinical practice where many patients suffer from multiple psychiatric disorders and receive a range of medications. Findings from clinical trials must be translated into effective treatment plans for these more complex cases.
Patients with comorbid obsessive-compulsive disorder (OCD) and an eating disorder are an example of this clinical dilemma, as these disorders often co-occur (Altman & Shankman, 2009; Angst et al., 2004; Kaye, Bulik, Thornton, Barbarich, & Masters, 2004; Swin-bourne & Touyz, 2007). Randomized controlled trials have identified evidence-based treatments for both disorders. For OCD, treatments include CBT emphasizing exposure and response prevention (ERP), serotonin reuptake inhibitors (SRIs) (clomipramine and the selective SRIs), or their combination (American Psychiatric Association, 2007). For eating disorders, treatments include: CBT emphasizing monitoring eating and cognitive restructuring (for anorexia nervosa [AN] and bulimia nervosa [BN]), fluoxetine (for BN), and weight restoration (for AN; American Psychiatric Association, 2006). However, OCD trials have generally excluded patients with eating disorders (e.g., Foa et al., 2005; Goodman et al., 1996; Hollander et al., 2003; Tollefson et al., 1994) and eating disorder trials usually do not report the frequency of comorbid OCD or its potential impact on treatment outcome (e.g., Fluoxetine Bulimia Nervosa Collaborative Study Group, 1992; Goldstein, Wilson, Thompson, Potvin, & Rampey, 1995; McIntosh et al., 2005).
Naturalistic studies suggest a complex relationship between OCD and eating disorders (Cumella, Kally, & Wall, 2007). One study of 508 females receiving treatment for an eating disorder (AN, BN, or eating disorder not otherwise specified [ED-NOS]), half of whom had comorbid OCD, found a reciprocal relationship, such that symptom improvement in one led to symptom improvement in the other (Olatunji, Tart, Shewmaker, Wall, & Smits, 2010). One implication of these findings is that simultaneous treatment of both conditions might lead to synergistic treatment effects.
ERP is highly effective for OCD as monotherapy or to augment SRIs (Foa et al., 2005; Simpson et al., 2008). ERP techniques have been used in BN with some success (Bulik et al., 1998; Gray & Hoage, 1990; Kennedy et al., 1995; Leitenberg, Rosen, Gross, Nudelman, & Vara, 1988; Wilson et al., 1991). Pilot data suggest that exposure techniques may also help patients with AN (Steinglass et al., 2011; Steinglass et al., 2012). Thus, one approach to patients with both OCD and an eating disorder is to adapt ERP to address both OCD and pathological eating fears and behaviors, and to couple this as needed with other evidence-based interventions, such as weight restoration (for AN) and medication management (for OCD or BN).
In this paper, we describe the outcome of 56 patients with both OCD and an eating disorder (AN, BN, or ED-NOS), who entered a residential treatment program that offered a cognitive-behavioral program that integrated standard ERP for OCD with ERP strategies specifically adapted for eating disorder pathology. Patients also received a structured supervised eating plan, medication management, and other psychosocial supports. Using validated outcome measures, we investigated how effective this multimodal program was at treating such complex cases. On the basis of the literature for both disorders suggesting efficacy of cognitive-behavioral approaches and on the conceptual overlap in the psychopathology underlying these conditions, we hypothesized that the treatment program would lead to clinically significant change in both OCD and eating disorder severity.
The data came from patients in the Comorbid Program, a residential treatment program at Rogers Memorial Hospital in Oconomowoc, WI. Patients provided written informed consent for their data to be used for research. The Rogers Center for Research and Training and the Institutional Review Board of the University of Wisconsin-Milwaukee approved the study.
The Comorbid Program is an eight-bed, residential facility designed to treat adults (≥18 years of age) with OCD and coexisting eating disorders who have not adequately responded to outpatient treatment. Patients are not admitted with active psychosis, mania, homicidality, or suicidality; substance abuse or electroconvulsive therapy within the past 30 days; or mental retardation. There is no minimum body mass index (BMI) required at admission. Some patients are transferred from an inpatient setting where they may have achieved partial weight restoration. The program accepts patients who self-pay or have private insurance.
The hallmark of this program is the use of ERP techniques to treat both OCD and pathological eating fears and behaviors. Standard ERP procedures are used for OCD symptoms (Kozak & Foa, 1997). Briefly, therapists help patients to generate a hierarchy of situations that trigger their anxiety or distress from least to most feared. Then, therapists help patients face these situations (“exposures”) to promote habituation to the anxiety that these situations trigger, starting with easier triggers and moving up the OCD hierarchy. In parallel, patients are asked to stop avoidance behaviors and rituals (“response prevention”) to break the connection between rituals and anxiety relief. Together, these procedures help disconfirm patients’ irrational beliefs. Therapists practice these steps with patients in session and assign specific exercises for between-session practice.
ERP procedures are also used to address eating disorder symptoms. As is standard in eating disorder programs, there is a minimum caloric goal per day for each patient to either restore weight if the patient is underweight (e.g., up to 90 calories/kg of body weight/day) or to maintain an appropriate weight (i.e., 50 calories/kg of body weight/day). However, what differs in this program is that a hierarchy of feared situations related to the eating disorder is also created, and the therapist helps patients face these situations in graduated steps of increasing difficulty. In addition, the menu plan is consistent with the ERP hierarchy: the initial food prescription is comprised entirely of foods low on the individual’s hierarchy and then changes as the patient habituates to the anxiety at one level and moves stepwise up the hierarchy. While this may resemble the introduction of challenge foods that are common in eating disorders treatment, it is unusual in the detail with which food fears are rated and in how the fear hierarchy shapes the menu plans. In this way, the approach more closely resembles ERP than classic CBT for eating disorders. Specific exposures to feared foods and eating situations (e.g., eating in public) are also done. To address fears about body shape, body image exposures may be conducted, including viewing disliked body parts in a mirror or wearing normally fitting rather than excessively loose clothing. Patients are asked to do exposures without engaging in compensatory behaviors related to their eating disorder (e.g., restricting during a future meal, vomiting, excessive exercise, or overeating in reaction to negative emotions). Two case histories are provided (see Table 6) that illustrate how such ERP procedures were applied to individual patients.
This ERP program is delivered each week-day in a three-hour block. During this block, patients participate in (1) individual therapy sessions (each lasting 30–45 minutes) with a behavioral specialist (typically a master’s-level clinician); (2) supervised exposures with other staff; and (3) self-exposures. They are also assigned 90 minutes of homework each day to practice assigned exposures without ritualizing. The goal is for patients to reach at least 70% up their hierarchy of fear prior to discharge and to continue to the top as an outpatient. To complement these ERP techniques, cognitive restructuring techniques are used to challenge irrational thoughts related to OCD and the eating disorder, constituting about 10% of treatment time.
Adjunctive treatment is also provided. Dieticians see patients at least weekly to assess the caloric prescription and to monitor weight and adherence with the structured meal plan that follows the food hierarchy (as described above). Meals are initially supervised by staff. As patients improve, they are expected to eat more independently. Psychiatrists see patients at least weekly to monitor medications. Nurses see patients as needed regarding general medical issues. Social workers meet with patients individually for an hour once or twice a week (to provide supportive psychotherapy and discharge planning) and in a group format every weekday for an hour (to provide psychoeducation about OCD and eating disorders and to facilitate peer support). Social workers also conduct weekly family sessions (in person or by phone). These sessions are used to provide psychoeducation, to discuss how family members can support the patient’s treatment goals, and to update the family on patient progress.
Admission diagnoses were made by board-certified psychiatrists according to DSM-IV-TR criteria. To determine these diagnoses, these psychiatrists, all experts in OCD and anxiety disorders, completed in-depth reviews of patients’ previous records as well as an interview with patients upon admission.
At admission and discharge, patients were asked to complete self-report outcome measures; BMI (kg/m2) was also calculated. The self-report measures were (1) the Yale-Brown Obsessive-Compulsive Scale—Self Report (Y-BOCS-SR; Steketee et al., 1996); (2) the Eating Disorder Examination Questionnaire (EDE-Q; Luce & Crowther, 1999); and (3) the Beck Depression Inventory-II (BDI-II; Beck et al., 1996). The Y-BOCS-SR consists of a checklist of obsessions and compulsions and a scale (0–40) that assesses their severity in the last week. To facilitate comparison with prior OCD clinical trials (Simpson et al., 2006), we also calculated the proportion of patients at discharge who met OCD response criteria (Y-BOCS-SR decrease ≥ 25%) and who had mild-to-minimal OCD symptoms (Y-BOCS-SR ≤ 12). The EDE-Q consists of 38 questions about behaviors and cognitions relating to eating, body shape, and weight in the prior 28 days; scores from four subscales (eating restraint, eating concerns, shape concerns, and weight concerns) are averaged to provide a global score. In a healthy young adult female population, the mean global EDE-Q (standard deviation [SD]) is approximately 1.6 (1.3) (Mond et al., 2006; Welch, Birgegard, Parling, & Ghaderi, 2011). The BDI-II is a 21-item measure of depressive severity in the prior week. Finally, for the Y-BOCS-SR and the EDE-Q, we used Jacobson and Truax (1991) methods to calculate the proportion of patients who met reliable change (defined as Y-BOCS-SR change of at least 7 points and an EDE-Q change of at least .81) and the proportion who achieved end-state functioning within the nonpatient distribution (defined as within one SD of a normal population1: Y-BOCS-SR ≤ 14; EDE-Q < 2.3). Normative data for the Y-BOCS-SR were taken from Steketee et al. (1996), in which the mean (SD) and test–retest reliability were 8.0 (6.6) and r = .88. Normative data for the EDE-Q global scale were taken from Ro et al. (2010), in which the corresponding values were 1.17 (1.11) and .93.
Treatment outcome was evaluated with paired-sample t-tests applied to admission and discharge scores for each continuous outcome measure (Y-BOCS-SR, EDE-Q, BMI, and BDI-II). Separate one-way analyses of variance (ANOVAs) explored whether the three eating disorder groups (AN, BN, and ED-NOS) differed in change scores on the four outcome measures. Significant group effects were followed by post-hoc t-tests for each pairwise comparison. Similar methods were used to explore whether treatment outcome differed between those at admission with low weight versus without low weight (BMI < 18.5 kg/m2 vs. BMI ≥ 18.5 kg/m2). Unless otherwise specified, statistical tests were two-tailed with level of significance α = .05.
One hundred eight patients diagnosed with OCD and a comorbid eating disorder were admitted to the Comorbid Program between June 2006 and July 2011. Fifty-six (52%) completed all study measures (Y-BOCS-SR, EDE-Q, BDI-II, and BMI) at both admission and discharge and therefore are included in this report. Completeness of data was the only inclusion criteria, not duration of stay, circumstances of discharge, or response to treatment. The mean length of stay (SD) for these 56 people was 57 days (27, range 12–131).
Demographic and clinical features of these 56 people are provided in Tables 1 and and2.2. At admission, they had clinically significant OCD, with a mean Y-BOCS-SR (SD) of 24.6 (7.2). Somewhat more were diagnosed with AN than with BN or ED-NOS. The mean EDE-Q global score (SD) was 3.8 (1.5), reflecting significant pathology (Fairburn & Beglin, 1994). Mean BMI (SD) at admission was 21.7 (7.4), but ranged from 13.8 to 43.1, as would be expected in this heterogeneous group. The mean (SD) BDI-II score at admission was 32.6 (13.4).
Fifty-two people admitted to the program were not included in the analyses because they were missing data either at admission or discharge. Those with admission data did not significantly differ on these clinical features when compared to the study sample (i.e., mean Y-BOCS-SR scores [SD] = 25.5 [7.7], t(96) = .60, p = .55; mean EDE-Q global scale [SD] = 3.5 [1.7], = t(90) = −1.0, p = .31; mean BMI [SD] = 21.0 [5.1], t(106) = −.66, p = .46; and mean BDI-II [SD] = 30.4 [14.5], t(98) = −.78, p = .44). The 52 not included in the study also did not differ in mean length of stay (53 days , range 3–161, t(106) = −.71, p = .48).
Patients experienced significant changes in all four outcome measures at discharge (Table 2). There were significant decreases in OCD severity, as measured by the Y-BOCS-SR. There were also significant decreases in eating disorder severity, as measured by the global EDE-Q score and reflected in all four EDE-Q subscales (eating restraint, eating concerns, shape concerns, and weight concerns). There were also significant increases in BMI and significant decreases in depressive severity, as measured by the BDI-II. All changes remained significant even with Bonferroni correction for eight comparisons (new α = .006).
At discharge, 45 of 56 (80%) met response criteria for OCD (≥25% Y-BOCS-SR reduction) and 28 of 56 (50%) had mild or minimal OCD symptoms (Y-BOCS-SR ≤ 12). Following the methods of Jacobson and Truax (1991), 40 of 56 (71%) met the reliable change index (RCI) for the Y-BOCS-SR (≥7 points), and 32 of 56 (57%) had clinically significant change (Y-BOCS-SR ≤ 14). At discharge, 36 of 56 (64%) met the RCI for the EDE-Q (≥.81 points), and 33 of 56 (59%) had clinically significant change (EDE-Q < 2.3).
As shown in Table 3, individuals with AN, BN, and ED-NOS did not significantly differ at admission in OCD, eating disorder, or depressive severity (all p-values > .25). Length of stay in days also did not differ (AN = 56.8 [29.8]; BN = 47.4 [20.6]; ED-NOS = 63.4 [25.5], F(2, 53) = 1.5, p > .23). Groups differed at admission on BMI, F(2, 53) = 6.2, p < .01. Post-hoc pairwise comparisons indicated individuals with AN had lower BMI scores than those with either BN or ED-NOS (both p-values < .01), those with BN did not differ from those with ED-NOS (p > .69). As shown in Table 3, individuals with AN, BN, and ED-NOS all showed reductions in OCD, eating disorder, and depressive severity. At the same time, BN patients had significantly more improvement than AN patients on the Y-BOCS-SR, EDE-Q, and BDI-II, and AN patients had significantly greater increases in BMI.
As shown in Table 4, individuals with and without low BMI did not significantly differ at admission in OCD, eating disorder, or depressive severity (all p-values > .40). Length of stay in days also did not differ (low BMI: 62.1 [24.9]; without low BMI: 52.9 [27.5], F(1, 54) = 1.7, p > .20). As shown in Table 4, patients with and without low BMI both showed large reductions in OCD, eating disorder, and depressive severity, with no significant group differences. Those with low BMI at admission had a significantly greater increase in BMI at discharge; this was expected given that they received a weight restoration (versus a weight maintenance) diet.
Most patients (n = 50 [89%]) were receiving psychiatric medications at admission. The proportion of patients receiving different classes of medication is shown in Table 5(a), with the most common class being antidepressants. At discharge, there was a small increase in the use of each class of medication (Table 5(b)). Only a few patients changed their medication, with the addition of a medication being more common than its removal (Table 5(c)). Antipsychotic medication was the class most commonly added (25% of patients); benzodiazepine was the class most commonly removed (9% of patients). There were no significant differences at admission in the percentage of patients with the three eating disorder diagnoses who were prescribed antidepressants, antipsychotics, mood stabilizers/anticonvulsants, or benzodiazepines (all χ2 < 1.9, all p-values > .40). At discharge, there were no differences among the three eating disorders in the percentage of patients prescribed antidepressants, antipsychotics, or mood stabilizers/anticonvulsants (all χ2 < 2.5, all p-values > .42), but there was a trend toward a differences in benzodiazepines (AN = 38%, BN = 12%, and ED-NOS = 63%; χ2 (2) = 5.9, p < .06). There were no significant difference at admission or at discharge in the percentage of patients in the two BMI groups (BMI < 18.5 kg/m2 vs. BMI ≥ 18.5 kg/m2) who were prescribed any of the four classes of psychiatric medications (antidepressants, antipsychotics, mood stabilizers/anticonvulsants, or benzodiazepines; all χ2 < 2.3, all p-values > .13).
This study investigated the outcome of patients diagnosed with OCD and an eating disorder, who entered a residential treatment program designed to address both disorders simultaneously. The uniqueness of the program was the use of ERP that focused not only on OCD fears and rituals but also on fears and rituals around food and body image. At discharge, the patients had clinically meaningful reductions in OCD severity, eating disorder symptoms, and depressive severity, and increases in BMI. These data suggest that simultaneous treatment of OCD and eating disorders with a multimodal approach can be highly effective.
Despite the presence of comorbid eating disorders, patients in this program experienced robust decreases in OCD severity at discharge. At discharge, the rate of OCD response (80%) and proportion of patients with mild to minimal OCD symptoms (50%) in this sample were similar to those reported in ERP trials that recruited OCD patients without comorbid eating disorders and had similar OCD severity at baseline (Foa et al., 2005; Simpson et al., 2006; Simpson et al., 2008). Significant reductions in OCD severity were observed across all eating disorder diagnoses regardless of BMI. At the same time, BN patients showed more improvement in OCD symptoms than AN patients.
Despite the presence of comorbid OCD, patients in this program also experienced significant reductions in eating disorder severity as measured by the EDE-Q. Improvements occurred across all the eating disorder diagnoses (although BN patients showed more improvement than AN patients) as well as in those with and without low BMI at admission. Shape concern, although improved, remained more elevated at discharge than other EDE-Q subscales, consistent with other eating disorder studies where shape concerns show less improvement or take longer to improve (Fairburn et al., 2009; Rosen, 1996). There was also a significant increase in weight in those with AN and in those with low BMI at admission.
These improvements in OCD symptoms and eating disorder severity occurred despite the high level of depression experienced by many on admission. Moreover, this sample also reported significant reduction in depressive symptoms at discharge, even though the CBT program did not systematically target depression. Since few (7%) started antidepressant treatment during the hospital stay, the reduction in depression was likely due to the psychosocial elements of the program or secondary to improvement in OCD or eating disorder psychopathology. Prior studies found that treatment of OCD or of eating disorders can improve depression (Attia et al., 1998; Foa et al., 1992; Meehan et al., 2006; Storch et al., 2010).
Because of the naturalistic design and multimodal treatment plan, we cannot determine whether the observed improvements in OCD or eating disorder severity were due to the integrated ERP component, other specific aspects of the program (e.g., structured eating), nonspecific effects (e.g., residential setting, social support, and time), or medication changes that occurred during the program. However, some tentative hypotheses can be made. First, the significant decreases in OCD and eating disorder severity were unlikely only due to medication, since many more patients responded to this treatment program than were newly prescribed antidepressants or antipsychotics, two classes of medications with known effects in these conditions as well as known side effects of weight gain. Importantly, patients with AN or low BMI at admission showed the greatest increases in BMI, but were not more likely to be on antipsychotics or antidepressants at admission or discharge than the other groups. Second, the robust decreases in OCD severity are most likely due to the ERP component, given the very low response rates in OCD to nonspecific interventions (Huppert et al., 2004; Simpson et al., 2008). Moreover, although eating disorder treatment on its own may reduce OCD symptoms (Cumella et al., 2007; Olatunji et al., 2010), no data indicate that it can do so when OCD is as severe as it was in this sample of patients. Finally, the decrease in OCD severity could be responsible in large part for the decrease in depressive severity, as this degree of improvement in comorbid depression after successful ERP treatment has been observed (Foa et al., 1992). Of note, the greatest improvement in depression occurred in those with BN, who also had the largest decreases in OCD severity.
It seems reasonable to assume that improvements in eating disorder symptoms were due in part to the non-ERP components of the program (e.g., sustained nutrition, observation after meals, and supportive psychotherapy). Reducing OCD symptoms with ERP may also have facilitated improvements in eating disorder symptoms (Olatunji et al., 2010). At the same time, the exposure techniques that targeted fear of food and eating could have directly reduced pathological eating behaviors and cognitions, thereby improving eating and facilitating treatment response. Consistent with the latter, ERP techniques have been found useful in BN (Bulik et al., 1998; Kennedy et al., 1995), and exposures focused on fear of food were found to lead to healthier eating behaviors in AN (Steinglass et al., 2012). Any of these possibilities support a multimodal program like this one for people with both OCD and an eating disorder.
The study had several limitations stemming in large part from its naturalistic design and from the fact that the data come from patients treated in a busy clinical setting. First, our sample was restricted to those who entered the residential program and completed all study assessments. This potentially could have led to a biased sample. On the other hand, study participants were not selected based on treatment response and did not differ from nonstudy patients either on clinical features at admission or on length of stay. Second, diagnoses were made by a psychiatrist during an intake evaluation; validated diagnostic instruments were not used to confirm these diagnoses and AN subtypes were not coded. Third, although medications at admission and discharge are included in the research database, exact doses were not; moreover, prior treatment history was not systematically recorded. Fourth, the outcome measures (other than BMI, which was objectively measured) were self-reports that were judged feasible to deliver in this clinical context; the focus was symptom severity. As a result, we cannot describe the frequency of different types of OCD symptoms in this sample. This would be an important area of future study. Finally, the length of stay was variable and there is no follow-up data. Future research should focus on whether the acute reductions in symptoms on this residential unit lead to improved psychosocial functioning and quality of life after discharge.
The findings from this study suggest that simultaneous treatment of OCD and eating disorders can be highly effective when a multimodal approach is used that emphasizes ERP techniques for both OCD and eating disorders and includes weight restoration and medication management. Naturalistic studies like these are an important way to begin to evaluate the effectiveness of evidence-based treatment approaches when they are applied in novel ways to real-world patients.
Dr Simpson’s effort on this study was supported by the National Institute of Mental Health (K24 MH091555) and by the New York State Office of Mental Health. Dr Steinglass was supported by the National Institute of Mental Health (K23 MH076195). Dr Riemann is the Clinical Director and Dr Weltzin is the Medical Director of the Comorbid Program described in this paper, and Dr Leonard is a Postdoctoral Fellow working in this program; all receive salary from Rogers Memorial Hospital. These data were presented in part at the Annual Meeting of the Anxiety Disorders Association of America (Baltimore, MD, 2010) and the Annual Meeting of the Association for Behavioral and Cognitive Therapies (San Francisco, CA, 2010). We thank Dr Timothy Walsh for helpful comments on an earlier version of this manuscript.
Financial Disclosures (none of which are relevant to this paper) In the prior three years, Dr Simpson has received research funds from Janssen and Transcept Pharmaceuticals and royalties from Cambridge University Press and UpTo-Date, Inc. Dr Franklin has received royalties from Springer Science and Business Media for a book on trichotillomania. The other authors report no potential financial or other relationships relevant to the subject of this article other than those listed in the Acknowledgements.
1Note Jacobson and Truax (Jacobson & Truax, 1991, method b; see p. 13) operationally define return to normal functioning when, “[t]he level of functioning subsequent to therapy should fall within the range of the functional or normal population, where range is defined as within two standard deviations of the mean of that population.” Given that a Y-BOCS score of 16 is a commonly used criterion for admission into a study of treatment for OCD, we feel the criterion of being within two SDs of the mean for a normal sample after treatment is too liberal and have therefore opted for the more conservative criterion of having a posttreatment score that is within 1 SD of the normative mean.