INTRODUCTION
Bulimia Nervosa (BN) is an eating disorder characterized by recurrent episodes of binge eating and compensatory behaviors to prevent weight gain. Since the disorder of BN was first acknowledged as a distinct syndrome from other eating disorders in the late 1970s (Russell, 1979), research has documented that individuals who engage in bingeing and purging behaviors not only manifest discrete core symptomatology but also possess personality and behavioral characteristics that are different from individuals with restrictive eating disorders. For instance, individuals who binge and purge are typically more impulsive, less emotionally stable (Vitousek & Manke, 1994), and they have a higher propensity towards substance abuse (O'Brien & Vincent, 2003).
Using the Iowa Gambling Task (GT) (Bechara et al., 1994; Bechara et al., 2000), several studies have consistently demonstrated the presence of decision-making deficits in individuals with substance dependence (Bechara et al., 2001; Bechara et al., 2002; Grant et al., 2000). Individuals with BN share many commonalities with substance abuse disorders (Vitousek & Manke, 1994) in that they continue to engage in self-damaging behaviors (binge eating/purging) that provide immediate benefit (pleasure, relief of tension), in the face of rising negative consequences (medical and psychological comorbidities). Like substance abuse disorders, BN involves a loss of control, extreme preoccupation with the behavior of choice (i.e., binge eating/purging cycle), and the use of the behavior to reduce tension or anxiety. Excessive food intake has even been found to mimic addictive physiological responses, such as down-regulation, sensitization, and withdrawal (Grigson, 2002). Given the similarities between substance abuse disorders and BN, it seems a logical extension of the literature to examine decision-making deficits with the Gambling Task in individuals with BN.
The Gambling Task
A recent attempt to quantify behaviors involved in making decisions uses a neuropsychological measure that emulates everyday decision making. The GT (Bechara et al., 1994; Bechara et al., 2000) was developed to study decision making in patients with lesions in the ventromedial (VM) region of the prefrontal cortex. Behaviorally, patients with bilateral damage to the VM prefrontal cortex consistently act according to immediate reward despite negative future consequences, suggesting that they have impaired decision-making capabilities. The GT has been found to provide a quantitative measure of such decision making deficits in patients with VM prefrontal cortex damage, and also those with substance abuse (Bechara et al., 2001; Bechara et al., 2002; Grant et al., 2000) pathological gambling (Cavedini et al., 2002) and psychopathy (Mitchell et al., 2002). This task is considered a valid measure of decision making (Bechara et al., 1994; Bechara et al., 1997), with demonstrated construct validity (Monterosso et al., 2001). For a detailed description of the GT, see Bechara et al., 1994; Bechara et al., 2000.
Rationale and Hypotheses
Like substance abusers, pathological gamblers, and patients with damage to the prefrontal cortex, bulimics show similar real-life behaviors in that they repeatedly engage in self-destructive acts that offer immediate reward irregardless of the future negative consequences that can ensue. The purpose of the present study is to test whether poor decision-making ability may be central to the disorder of BN. Our specific predictions were: (1) women with a current or past diagnosis of BN would perform significantly worse on the GT than controls; (2) bulimic symptom severity would negatively correlate with GT performance (i.e., those with more severe bulimic symptoms would demonstrate poorer decision making ability on the GT); (3) severity of bulimic symptoms would uniquely predict performance on the GT above and beyond other demographic variables and symptoms of depression; (4) despite the high co-occurrence of BN and depression, depressive symptoms severity would not significantly correlate with GT performance.
METHOD
Participants and Procedures
Participants were recruited through local advertisement and through introductory psychology courses at a state supported large, liberal arts university. The present study only included female participants, because BN is 10 times more prevalent in females than in males (APA, 2000). During initial screenings of participants, questionnaire data were collected from a personal history questionnaire used to assess current and past demographic, medical, and psychosocial information; the Beck Depression Inventory (BDI-II; Beck et al., 1996) was used to assess presence of severe depressive symptoms; and the Bulimia Test Revised (BULIT-R; Thelen et al., 1991) and the Eating Disorder Examination-Questionnaire (EDE-Q; Fairburn & Beglin, 1994) were used as objective measures of bulimic symptom severity. Participants who endorsed significant abnormal eating pathology on objective measures were further interviewed over the telephone to ensure that they met Diagnostic and Statistical Manual, 4th Edition (DSM-IV) diagnostic criteria for a current or past diagnosis of BN and to determine presence or absence of exclusionary criteria for inclusion in the study. Those that met diagnostic criteria for BN with the absence of exclusionary criteria were included in the BN Group. Those who endorsed minimal eating disordered behaviors with the absence of exclusionary criteria were included in the Control Group.
Exclusionary criteria for both groups included history of substance abuse or dependence, head trauma or other neurological disorder, severe psychiatric disturbance, severe depressive symptoms, history of learning disability, and for the control group, a current or past diagnosis of an eating disorder. Participants in both groups were matched on age, level of education, and verbal intelligence.
Forty (N = 40) undergraduate females were divided into 2 groups: the Control Group: women with minimal or no eating disorder symptoms (n = 20), and the Bulimic Group: women with a current (n = 16) or past (n = 4) diagnosis of BN (total n = 20). All participants were between the ages of 18 and 25. All participants were tested individually in a private room by a trained research assistant. Each participant was weighed on a standard scale and height measurements were taken using a standard tape measure. Prior to beginning the experiment, participants were provided with a complete overview of the procedures involved in order to ensure an understanding of the study's requirements and commitment to the session. In addition, each participant signed a written informed consent form approved by the local Institutional Review Board, which outlined the risks and benefits of the study.
Measures and Materials
Eating Disorders Examination Questionnaire
The Eating Disorders Examination Questionnaire (EDE-Q) (Fairburn & Beglin, 1994) is a self-report questionnaire used to assess presence of eating disorder pathology.
Bulimia Test Revised
The Bulimia Test Revised (BULIT-R; Thelen et al., 1991) is a questionnaire that assesses the frequency of bulimic behaviors. It has been established as a valid instrument for identifying individuals meeting DSM-IV criteria for BN.
Beck Depression Inventory-II
Depressive symptoms were measured with the Beck Depression Inventory-II (BDI-II; Beck et al., 1996).
Wechsler Test of Adult Reading
The Wechsler Test of Adult Reading (WTAR; Wechsler, 2001) was developed as a measure of predicting intellectual performance and was used in the present study as a measure of verbal intelligence. Overall, the WTAR displays internal consistency of 0.93 and is highly correlated with the Wechsler Adult Intelligence Scale Third Edition Full Scale IQ measure (0.73) (Wechsler, 2001).
The Gambling Task
The GT was used as a measure of decision making. The GT was scored by subtracting the number of cards chosen from Decks A & B in the last 50 trials from the number of cards chosen from Decks C & D in the last 50 trials in accordance to existing literature (Monterosso et al., 2001; Schmitt et al., 1999). Lower scores indicate poor performance, and negative scores indicate preference for the disadvantageous decks (A & B).
RESULTS
A series of analysis of variance (ANOVA) were conducted to compare women in both groups on demographic variables, body mass index, depressive symptoms, and bulimic symptom severity (See Table 1). There were no significant differences between groups on age, F(1,38) = .17, p = .68, education level, F(1,38) = .10, p = .75, and WTAR standard score, F(1,38) = .37, p = .55. Quetlet's body mass index (BMI) was calculated (kg weight per m2 height) based on laboratory measured weight and height. The two groups did not differ on mean BMI, F(1,38)=1.52, p = .23, and mean BMI for both groups fell within the normative range (National Research Council, Committee on Diet and Health, 1989). The BN Group scored significantly higher on the BDI-II than the Control Group, F(1,38)=28.41, p < .001. The BN Group also scored significantly higher on both the BULIT-R, F(1,38)=115.74, p < .001 and the EDE-Q, F(1,38) = 82.50, p < .001, in comparison to the Control Group.
Age, education level, verbal ability, body mass index, depression, and bulimic symptomatology of Groups 1 and 2
An ANOVA indicated that women in the BN Group performed significantly worse on the GT than the Control Group (i.e. the BN Group chose significantly fewer times from Decks C and D—the advantageous decks in comparison to the Control Group, F(1,38)=4.44, p = .04). On the last 50 trials, the Control Group selected an average of 34.95 (SD = 9.90) cards from advantageous decks while the BN Group selected 28.75 (SD = 9.20) from these decks.
A series of bi-variate correlations revealed that as bulimic symptom severity (measured by the BULIT-R and EDE-Q) increased, GT performance decreased (BULIT-R: [r = −4.10, p < .001]; EDE-Q: [r = −4.20, p < .01]). Table 2 shows the intercorrelations between the BULIT-R, the global EDE-Q score, subscales of the EDE-Q (Restraint, Eating Concern, Weight Concern) and the GT. All eating disorder measures were significantly negatively correlated with the GT. As would be expected, all eating disorder measures were significantly positively correlated with each other (See Table 2).
Correlations between EDE-Q, BULIT-R, and Gambling Task
A standard multiple regression analysis was performed using age, education level, verbal ability, BULIT-R score, and BDI-II score as the independent variables and GT score as the dependent variable. BULIT-R score was the only variable that was able to uniquely predict GT performance, Beta = –0.61, p < .01.
The BN Group scored higher than the Control Group on the BDI-II, F(1,38) = 28.41, p < .001. It should be noted that, whereas this difference is significant, the average BDI-II score of the BN Group was indicative of only mild depressive symptoms (M = 15.5). Because of this group difference, a bi-variate correlation was run to assess whether scores on the BDI-II were correlated with performance on the GT. Results indicated that there was no significant correlation between the BDI-II and the GT (r = −.108, p > .50), thus eliminating the need to control for depressive symptoms when comparing groups on the GT.
DISCUSSION
The present investigation is the first to demonstrate decision-making impairments via the Gambling Task in women with BN. Results of this study indicated that bulimic women demonstrated preference for the high-risk decks and continued to select cards from these decks despite high monetary losses. Further, bulimic symptoms had a significant relationship to GT score and were able to uniquely predict GT performance. Depressive symptoms were not correlated with GT performance, providing further evidence that decision-making deficits are secondary to the specific behavioral characteristics of BN.
Bulimics continually binge-eat and purge despite future negative consequences, suggesting impaired ability to weigh immediate rewards versus long-term negative consequences. Little empirical research has been conducted on decision making in BN; however, the current study offers additional support for the notion that poor decision making is characteristic of BN and generalizes the use of the GT to this population.
Bulimia and Decision Making
Together, these results demonstrate that in comparison to a control group of women, bulimic women showed a distinct impairment in their ability to make decisions on a neuropsychological task. Recurrent episodes of out of control binge eating, followed by attempts to rid of the calories consumed are central features of BN. Individuals with bulimia display striking difficulties in controlling such behaviors, despite the fact that they view their bingeing as pathological and report a desire to change their chaotic eating patterns. Thus behaviorally, bulimics tend to choose high immediate gains (bingeing and purging) even if these rewards are followed by higher future losses (distress, depression). Such actions persist even in the face of rising negative outcomes. The current study demonstrated that this pattern of poor decision-making behavior can be quantified using the GT. In general, bulimic women continued to select cards from the risky decks even though the decks resulted in great losses over the long run, mirroring the real-life behaviors that are characteristic of this disorder.
Given the results of the present study, it can be speculated that increased bulimic symptom severity and subsequent inability to inhibit binge/purge behaviors contribute greatly to making the decision of whether or not to engage in binge eating and purging. Poor performance on the GT has been linked to VM prefrontal cortex dysfunction, given that patients with such lesions have pronounced decision-making deficits. Of interest, is whether future studies can provide a possible link between the neural circuitry involvement in the pathophysiology of BN and the decision-making deficits associated with the disorder.
Continued evaluations of such cognitive constructs would help to explain the interconnectedness between these cognitive and neurobehavioral characteristics. Clearly, additional research is needed in this area in order to better understand how these cognitive processes work in BN and how they relate to the behavioral presentation of this disorder.
