¹Virginia Institute for Psychiatric and Behavioral Genetics, Department of Psychiatry,
Virginia Commonwealth University, Richmond, Virginia

²Department of Psychological Medicine, Christchurch School of Medicine
Christchurch, New Zealand

Accepted 29 December 1997

Abstract Objective: To examine characteristics of individuals who show a rapid and sustained response to cognitive-behavioral therapy (CBT) for bulimia nervosa (BN). Method: As part of a randomized clinical trial designed to dismantle CBT for BN, we compared 19 individuals who exhibited complete abstinence from binging and purging after only eight sessions of CBT and maintained abstinence throughout the duration of treatment and the 1-year follow-up interval, to 79 individuals who had a more variable response to treatment. Demographics, baseline clinical characteristics, lifetime comorbidity, and personality disorder symptoms and profiles from the Temperament and Character Inventory were examined.. Results: In univariate analyses, frequency of binging and purging at baseline, low scores on the Eating Disorders Inventory (EDI) Bulimia subscale, lower harm avoidance, and higher self-directedness were associated with rapid response. In multiple regression analyses, frequency of binging at baseline (OR = 0.87; 95% C1 0.77-0.98) and self-directedness (OR = 1.12; 95% C1 1.04-1.21) independently predicted rapid and sustained treatment response. Conclusion: Frequency of binging and character quality of self-directedness may be useful predictors of those individuals who are likely to respond positively to a brief course of CBT for BN ©1999 by John Wiley & Sons, Inc. Int J Eat Disord 26: 137-144, 1999.

Key words: cognitive-behavioral therapy; bulimia nervosa; abstinence

INTRODUCTION

An important question in treatment research on bulimia nervosa is the identification of
predictors of treatment nonresponse. Although several studies have identified specific predictors of poor outcome, little consensus has been reached (Blouin et al. 1994 Collings & King, 1994: Fairburn, Peverler, Jones, Hope, & Doll, 1993: Freeman, Beach, Davis, & Solyom, 1985: Herzog & Sacks, 1993: Herzog, Keller, Lavori, & Sacks, 1991; Hsu & Holder, 1986, Keller, Herzog, Lavori, Bradburn, & Mahoney, 1992: Mitchell et al., 1988; Olmstead, Kaplan, & Rockert, 1994: Turnbull et al., 1997). In this study, we propose a complementary approach that attempts to identify factors associated with a rapid and sustained response to cognitive-behavioral therapy (CBT) for bulimia nervosa.

Support for this approach comes from a reanalysis of the data performed by Wilson (personal communication, 1997) from two randomized clinical trials (Fairburn et al., 1991; Walsh et al., 1997) for bulimia nervosa in which early response was associated with good outcome. In this reanalysis, in the comparative trial of CBT, interpersonal psychotherapy, and behavioral therapy (Fairburn et al., 1991), the majority of clinical change was seen by the eight psychotherapy sessions. Similarly, in the randomized clinical trial by Walsh et al. (November, 1997) patients who had not shown 50% or more clinical change by Session 5 of CBT tended to be nonresponders. Thus, rapid response may reflect important underlying dimensions that
portend positive and sustained clinical change. Elucidation of these underlying factors could

ultimately assist with identification of individuals who would be most likely to respond to a brief psychotherapy trial and perhaps even to less intensive therapies such as self-help or minimal primary care intervention (Treasure et al., 1994, Waller et al., 1996).
The goal of this study was to compare the characteristics of individuals who showed a rapid response to eight sessions of CBT (defined as a 2-week abstinence from binging and purging), and who remained abstinent throughout treatment and a 1-year follow-up, with individuals who did not evidence rapid and sustained response.

METHODS

Subjects

Subjects for this study have been described in detail elsewhere (Bulik, Sullivan, Carter & Joyce, 1998) and comprise all individuals who participated in a randomized clinical trial examining the additive efficacy of exposure with response prevention (ERP) to a core of CBT and who were available for follow-up at 1 year. One hundred thirty-five women entered the clinical trial. A total of 106 women completed all phases of the clinical trial and 105 women were available for 1-year follow-up. (7 of whom had not completed all phases of treatment). This report focuses on 98 women who completed all phases of treatment and were available for 1-year follow-up.
All women met criteria outlined in the 3rd Rev. ed. Of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IIIR; American Psychiatric Association [APA, 1994) for current bulimia nervosa as assessed by the Structured Clinical Interview for DSM-III-R (SCID; Spitzer, Williams, Gibbon & First, 1992) modified to require the presence of “objective” binges (Beglin & Fairburn, 1992). All women were between the ages of 17-45. The principal exclusion criteria were the presence of current anorexia nervosa, obesity (body mass index [BMI]>30), significant medical illness clearly influencing the eating disorder (e.g., insulin-dependent diabetes mellitus), or the use of psychoactive medication that could influence bulimic symptomatology (e.g., antidepressants). Recruitment was accomplished through a wide variety of sources including referral from general practitioners or mental workers and community advertisements. This study had been approved the university and hospital ethical committees and all subjects signed informed consent prior to participation.

MEASURES

All subjects underwent a 2 to 4-hr assessment at baseline conducted by one of the authors

This assessment consisted of the SCID (Spitzer, Williams, Gibbon, & First, 1988) which we modified to obtain more information about the lifetime history of eating disorders (Spitzer, Williams, & Gibbon, 1997) and related behaviors together with SCID-II for personality disorders occurred, for this analysis we chose to examine the total number of symptoms endorsed for Cluster A (Paranoid, Schizoid, Schizotypal), Cluster B (Borderline, Histrionic, Narcissistic, and Antisocial), and Cluster C (Obsessive-Compulsive [sic], Dependent, Avoidant, and Passive-Aggressive). The interviewing clinician also completed the 17-item Hamilton Depression Rating Scale (HDRS; Hamilton, 1960), Global Assessment of Functioning Scale (GAFS; APA, 1994), and a structured interview that assessed bulimic symptomatology in the prior fortnight including the number of objective binges, the total episodes of purging via vomiting and laxative misuse, and the degree of food restriction and body dissatisfaction. At the end of eight sessions of cognitive therapy (mid treatment) , at the end of the additional eight sessions of behavioral therapy (end treatment), and at 1-year follow-up, the structured interview for bulimic symptoms, HDRS, and GAF were readministered by a clinician blind to treatment group.
At baseline, all subjects also completed a battery of self-report questionnaires designed to gather dimensional measures of eating psychopathology, personality characteristics, and psychological style. These included the eating Disorders Inventory (EDI-II; Garner, 1991) the Bulimia Cognitive Distortions Scale (BCDS: Schulman, Kinder, Powers, Prange, & Gleghorn, 1986), and the Temperament and Character Inventory(TCI; Coninger, Svrakic , & Przybeck, 1993).

Treatment Phase 1: CBT

All individuals who entered the study received eight sessions of CBT. For the first fortnight, sessions were twice weekly, followed by weekly sessions for 6 weeks. CBT consisted of psychoeducation, self-monitoring, support of patient led normalization of meals (especially breakfast), cue identification, challenging automatic thoughts, thought restructuring, chaining, and relapse prevention. Specific goals for each session were outlined clearly in the manuals and homework was assigned for each module. In order to maintain the distinction between CBT and the ERP conditions, with the exception supporting any patient-led changes in normalization of meals, no exposure techniques were incorporated into the CBT portion of treatment.
We purposely chose eight sessions of CBT to represent a minimum dose of the intervention to enable exploration of the potential additive effects of ERP. Despite this underdosing, a substantial number of women achieved remission from binging and purging after just eight sessions and remained abstinent throughout the treatment and at 1-year follow-up. The characteristics of these individuals are the focus of this paper.

Phase II: Experimental Conditions: Common Elements
After completion of CBT, subjects were randomized to either ERP to prebinge cues, ERP to prepurge cues, or a control condition of relaxation therapy. The nature of these treatments is described in Bulik et al., (1998). In all conditions, the first four sessions were twice weekly followed by four weekly sessions. In each condition, at least two sessions were performed outside the therapist’s office and participants were given homework assignments for self-guided exposure or relaxation. All sessions were a minimum of 50 min; however, ERP sessions continued until physiological arousal and self-report levels approached baseline. Session length varied across and within subjects across time (range 50 min to 3 hr). Thus, although the number of sessions was equated for ERP and relaxation, individuals in the ERP conditions had greater total contact hours with their therapists. The effect of treatment group is not examined in this report as the clinical changes observed occurred prior to randomization.

Definition of Response Categories

We defined two levels of response to CBT based on status at mid treatment (after eight sessions of CBT), end treatment (after an additional eight sessions of behavioral therapy), and at 1-year follow-up. We were primarily interested in individuals who evidenced a rapid and sustained response to CBT. Rapid responders (n = 19) were defined as those individuals who reported being completely abstinent from binging and purging in the fortnight prior to the mid treatment assessment and who reported continued abstinence at both the end treatment and throughout the 1-year follow-up assessment. Nonrapid responders (n = 79) were not abstinent by the midtreatment assessment and showed a more variable course of treatment response. Several patterns of response existed in this group including individuals who were abstinent at the end of treatment and at 1-year follow-up (n = 28), individuals who were symptomatic both at end treatment and at follow-up (n = 34), and individuals who fluctuated between abstinence and nonabstinence across the three measurement points (n = 17). There were no differences across response groups in terms of the behavioral treatment received, providing further support for not including treatment group in subsequent analyses.

Statistical Analysis

Data management and all statistical analyses were conducted with the JMP statistics packages (SAS Institute Inc., 1994) and SAS (SAS Institute Inc., 1989, 1996). Student’s t tests and the Wilcoxon nonparametric test (for nonnormally distributed variables) were used to compare continuous variables and the chi-square test was used to compare discrete variables across the response groups. A stepwise multiple regression was then performed using baseline clinical characteristics and response on the psychometric instruments as predictors of rapid response to treatment (SAS Institute, 1989, 1996).

RESULTS

Baseline Clinical Characteristics

There were no significant differences between rapid and nonrapid responders on age, age of onset of bulimia nervosa, duration of illness, or current or desired BMI (Table 1).
At baseline, the rapid responders exhibited significantly less binging, vomiting, and total purging episodes (a composite measure of vomiting and laxatives) than nonrapid responders, suggesting a less severe clinical presentation. No differences were observed in baseline HDRS or GAFS scores. There were also no differences in lifetime history of major depression, alcohol dependence, or anorexia nervosa. Nor were there any differences in the number of symptoms of Cluster A, Cluster B, or Cluster C personality disorder symptoms as assessed by the SCID-II.

Table 1. Comparison of baseline clinical characteristics between rapid and nonrapid responders to CBT

Variable Rapid Responders
(n = 19)
Nonrapid Responders
(n = 79)
i (96)
or x² ?

Age 26.6 (7.1) 26.6 (6.0) 0.01 ns
Age of onset of BN 19.0 (4.4) 19.9 (4.7) 0.74 ns
Duration of illness (months) 7.6 (7.0) 6.8 (5.9) 0.62 ns
Binges per fortnight 5.2 (4.4) 11.3 (11.4) 6.43 .01
Vomiting episodes per fortnight 6.1 (5.1) 12.4 (12.8) 4.77 .03
Laxative episodes per fortnight 0.7 (1.9) 2.3 (5.0) 1.86 ns
Total purges per fortnight 6.8 (5.7) 14.6 (13.1) 7.94 .005
GAFS 57.6 (5.8) 55.3 (6.7) ?1.39 ns
HDRS 7.6 (5.9) 8.8 (5.2) 0.92 ns
Current BMI 22.8 (2.3) 22.3 (2.6) ?0.65 ns
Desired BMI 20.2 (1.9) 19.5 (1.8) ?1.57 ns
Major depression lifetime 7 (37%) 45 (57%) 2.49 ns
Alcohol dependence lifetime 9 (47%) 36 (46%) 0.02 ns
Anorexia nervosa lifetime 5 (26%) 25 (32%) 0.20 ns
Cluster A symptoms 3.7 (3.41) 4.2 (3.5) 0.47 ns
Cluster B symptoms 6.7 (4.6) 7.4 (5.0) 0.58 ns
Cluster C symptoms 5.1 (4.1) 6.6 (4.7) 1.26 ns

Note: BN = bulimia nervosa; CBT = cognitive-behavioral therapy; GAFS=Global Assessment of Functioning Scale;
HDRS = Hamilton Depression Rating Scale; BMI = body mass index.

Importantly, none of the rapid responders in comparison to 24% of the nonrapid responders received additional treatment for their eating disorder in the 1-year follow-up interval (Fisher’s Exact Test p < .02).

Psychometric Results

On the EDI, the rapid responders scored significantly lower than the nonrapid responders on the
Bulimia subscale (Table 2). There were no significant differences on cognitive distortions as measured by the BCDS. On the TCI personality dimension of harm avoidance, rapid responders scored significantly lower than nonrapid responders. On the character dimension of self-directedness, rapid responders scored significantly higher than nonrapid responders.

Stepwise Multiple Regression

We then entered all of the variables explored in the univariate analyses into a stepwise multiple regression in order to determine the independent predictors of rapid treatment response. Only two variables entered and remained in the final regression equation: baseline binges per fortnight and TCI self-directedness (Table 3). The odds ratio for binges per fortnight indicated that lower levels of binging were predictive of rapid responding. For self-directedness, the opposite was true. For each unit increase on the self-directedness scale, an individual was 1.12 times more likely to be a rapid responder.

DISCUSSION

The purpose of this paper was to identify the characteristics of those individuals who are likely to respond to a brief course of CBT and maintain treatment gains. Identification of

Table 2. Comparison of EDI and TCI scores for rapid and nonresponders to CBT

Variable Rapid Responders
(n = 19)
Nonrapid Responders
(n = 79)
t (96)
?

EDI
Drive for Thinness 12.9 (5.3) 14.5 (4.5) ?1.38 ns
Bulimia 7.3 (4.3) 10.0 (4.8) 2.35 .02
Body Dissatisfaction 17.4 (7.7) 19.0 (7.5) 0.83 ns
Bulimia Cognitive Disorders
Questionnaire 85.6 (17.0) 90.4 (16.5) 1.15 ns
TCI
Novelty seeking 21.2 (7.0) 22.0 (6.1) 0.07 ns
Harm avoidance 17.4 (5.3) 21.2 (7.0) 2.19 .03
Reward dependence 15.2 (5.6) 15.9 (4.1) 0.64 ns
Persistence 4.6 (2.2) 4.8 (1.9) 0.53 ns
Self-directedness 29.7 (7.8) 23.4 (7.9) ?3.14 .002
Cooperativeness 34.4 (5.9) 34.1 (5.8) ?0.25 ns
Self-transcendence 11.8 (6.2) 11.1 (3.3) ?0.44 ns

Note: EDI = Eating Disorders Inventory; TCI = Temperament and Character Inventory; CBT = cognitive behavioral therapy.

potential rapid responders in advance could assist with treatment planning and with informing patients of the likely trajectory of their recovery when in treatment.
In this clinical trial, 19 individuals exhibited rapid and sustained abstinence from binging and purging after only eight sessions of CBT for bulimia nervosa. Although all of these subjects received eight additional sessions of exposure or relaxation therapy, the maintenance of abstinence throughout the follow-up interval required no additional treatment to that received in the clinical trial. These 19 individuals represented 14% of those who entered the clinical trial and 18% of completers. The rapid responders were not distinguishable from individuals with more variable response on dimensions such as age, age of onset, duration of illness, current or desired BMI, lifetime Axis I comorbidity, or personality disorder symptoms as assessed by the SCID-II.
Lesser severity–reflected in lower frequencies of binging and purging at baseline and lower scores on the Bulimia subscale of the EDI– was associated with rapid response. Indeed, frequency of binging at baseline was the only eating disorder symptom that was predictive of rapid response in the multiple regression analysis. At presentation, rapid responders reported binging on average 5.2 times per fortnight and vomiting 6.1 times per fortnight, indicating that these individuals tend to be at or around the clinical cut-off criterion for the diagnosis of bulimia nervosa in the 4th ed. Of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; APA, 1994). Reflected in the standard deviation of this measure, many of these women reported a rapid decrease in binging and purging after their first contact with the treatment program. These individuals met DSM-

Table 3. Stepwise logistic regression practicing rapid treatment response in women with bulimia nervosa.

Variable Parameter Estimate

x² p Odds Ratio 95% Confidence Interval

Binges per fortnight ?0.14 5.43 .02 0.87 0.77-0.98
Self-directedness 0.11 8.24 .004 1.12 1.04-1.21