Abstract word count: 205 TEMPORAL DISCOUNTING IN BULIMIA NERVOSA Increased temporal discounting

Objective: There is evidence that people with eating disorders display altered intertemporal choice behaviour (the degree of preference for immediate rewards over delayed rewards). Compared to healthy controls (HC), individuals with anorexia nervosa and binge-eating disorder show decreased and increased rates of temporal discounting (TD; the devaluation of delayed rewards), respectively. This is the first study to investigate TD in people with bulimia nervosa (BN). Method: Thirty-nine individuals with BN (2 men) and 53 HC (9 men) completed a hypothetical monetary TD task. Over 80 binary choices, participants chose whether they would prefer to receive a smaller amount of money available immediately or a larger amount available in 3 months. Selfreported ability to delay gratification (the behavioural opposite of TD) was also measured. Results: Individuals with BN showed greater TD (i.e., a preference for smaller-sooner rewards) and a decreased self-reported capacity to delay gratification relative to HC. Experimental groups did not differ in age, gender ratio, or BMI. Discussion: Increased rates of TD may contribute to some of the core symptoms of BN that appear to involve making choices between immediate and delayed rewards (i.e., binge-eating and compensatory behaviours). Altered intertemporal choice behaviour could therefore be a relevant target for intervention in this patient group. Increased temporal discounting 3 Increased temporal discounting in bulimia nervosa The pathophysiology of bulimia nervosa (BN) is poorly understood and strong evidence to guide treatment is lacking (1). Exploration of neurocognition in BN has the potential to elucidate mechanisms underpinning associated behavioural abnormalities, and to promote the development of tailored therapeutic interventions. Several neuropsychological difficulties have been observed in BN (2). For example, individuals with BN, as well as other eating disorders (EDs; anorexia nervosa [AN] and bingeeating disorder [BED]), have an increased preference for risky and disadvantageous choices in a context of uncertainty (3). There is also evidence that patients with EDs make maladaptive intertemporal choices. A reward arriving sooner is often more appealing than one arriving later, even when the later reward is larger. Thus, individuals discount the value of delayed outcomes – a phenomenon known as temporal discounting (TD). This tendency to devalue future rewards appears to be accentuated in BED (increased TD) (4) and diminished in AN (decreased TD) (5), which may underlie the disinhibited and restrictive eating that characterise these disorders. This study investigated whether individuals with BN display altered rates of TD and differences in the selfreported capacity to delay gratification relative to healthy controls (HC). MATERIAL AND METHODS Participants Participants were men and women ≥18 years with BN or no current/previous diagnosis of any psychiatric disorder (HC): their data were pooled from two larger studies conducted by our group (currently in preparation for publication). Patients with BN were recruited via online advertisements on the King’s College London (KCL) and BeatTM research recruitment webpages and through the South London and Maudsley NHS Foundation Trust ED Outpatient Service, while HC responded to online and poster advertisements at KCL. Group classification was established via Increased temporal discounting 4 self-report and checked over email/telephone: DSM-V BN diagnosis was confirmed using an edited version of the Eating Disorder Diagnostic Scale (EDDS) (6), and the absence of a psychiatric disorder in HC was confirmed using the EDDS and the Structured Clinical Interview for DSM-IV Axis I Disorders Screening Module (7). One hundred and thirty participants (BN = 55; HC = 75) completed the screening and 122 (BN = 52; HC = 70) were eligible for inclusion. Of these, 92 (BN = 39; HC = 53) completed the study and were included in the analyses. The two larger studies were approved by the KCL Psychiatry, Nursing and Midwifery Research Ethics Subcommittee and the London City Road & Hampstead Research Ethics Committee. Participants gave informed consent prior to taking part and were compensated for their time. Procedure Full procedural details are provided in Appendix A. All participants attended a testing session at the Institute of Psychiatry, Psychology & Neuroscience, KCL. Study procedures undertaken prior to the TD task were comparable between the two studies: both involved providing written consent and completing several identical questionnaires, including the Depression Anxiety and Stress Scales (DASS-21) (8) and the Delaying Gratification Inventory (DGI) (9). Additionally, in both cases the TD task was done on a laptop with an experimenter present. Data were collected between May 2014 and September 2015. Temporal discounting task TD was assessed using a computerised hypothetical monetary choice task, modelled on an established paradigm (5). On each of 80 trials, participants had an unrestricted amount of time to indicate whether they would prefer to receive a smaller amount of money immediately (smallersooner reward) or a larger amount after 3 months (larger-later reward). Two types of decision Increased temporal discounting 5 framing were employed: ‘Accelerate’ (larger-later reward remained at £100, smaller-sooner reward increased from £20 to £98 in £2 increments) and ‘Delay’ (smaller-sooner reward remained at £50, larger-later reward increased from £52 to £130 in £2 increments) (40 trials for each). The trials were pseudo-randomly interleaved, so that the two decision frames were intermixed. TD was quantified by determining participants’ discount factor (DF) – the magnitude of reduction in the present value of a future reward – for each choice set using a two-step procedure (5) (see Appendix B). Global DF was calculated as the mean of the two DFs, and was used as the primary outcome variable in this study. The value obtained can range from 0 to 1, with smaller numbers indicating greater TD (i.e., a greater tendency to choose the smaller-sooner reward). Delaying Gratification Inventory Self-reported ability to delay gratification was measured with the DGI, which requires respondents to rate the extent to which they agree with 35 statements on a 5-point Likert scale. Scores are generated for five domains of delay behaviour (Food, Physical Pleasures, Social Interactions, Money, and Achievement) and a total score (Global DGI score) is calculated. This was used as the outcome variable here. Higher values indicate a greater capacity to delay gratification. RESULTS Statistical analyses were performed using SPSS (tests were two-tailed, α = 0.05). Key sample characteristics and raw intertemporal choice data are provided in Table 1. TD data were positively skewed, therefore square-root transformations were applied and transformed values were used in all subsequent analyses. Global DFs and Global DGI scores were correlated in the sample as a whole [r = 0.33, p = .002] (i.e., the higher the rate of TD, the lower the ability to delay gratification). A one-way multivariate ANOVA showed that individuals with BN had lower Global DFs (indicating an increased rate of TD) [F(1, 90) = 5.72, p = .019] and Global DGI scores (indicating a Increased temporal discounting 6 reduced capacity to delay gratification) [F(1, 90) = 41.65, p < .001] than HC. To examine whether these group differences persisted after controlling for other possible determinants, age, gender, BMI, and DASS-21 depression, anxiety, and stress scores were entered into the model as covariates. This revealed a significant effect of group on Global DF [F(1, 84) = 5.52, p = .021] but not Global DGI score [F(1, 84) = 2.24, p = .138], due to the inclusion of DASS-21 stress scores [F(1, 84) = 5.25, p = .024]. An exploratory mixed ANOVA revealed no significant main effect of framing (Accelerate vs. Delay) or framing x group interaction on DFs [both p ≥ .654]. Bivariate correlations were used to explore relationships between Global DFs, Global DGI scores, clinical outcomes (DASS-21 and EDE-Q scores, illness duration, and frequency of bingeeating, vomiting, laxative use, and excessive exercise), and BMI. Pearson’s r and Spearman’s rho correlation coefficients were employed. In the BN group, Global DFs were not significantly related to any clinical variables [all p > .109], and Global DGI scores were also not significantly correlated with any clinical variables [all p ≥ .278] except for DASS-21 depression [r = -0.34, p = .036] and stress [r = -0.39, p = .013] scores. Neither Global DFs nor Global DGI scores were associated with BMI when the BN and HC groups were considered separately or together [all p ≥ .233].