1 INTRODUCTION

The seeming ease with which individuals with anorexia nervosa (AN) resist immediate food rewards in their pursuit of long-term weight loss is often believed to reflect altered reward sensitivity and/or an excessive amount of self-control (Steinglass, Berner, & Attia, 2019). To quantify this apparent increased capacity to delay reward in AN, studies have employed delay discounting (DD, or “intertemporal choice”) tasks which provide a behavioral measure of self-controlled versus impulsive decision making (McClelland et al., 2016). DD tasks require participants to repeatedly choose between a small (usually monetary) reward delivered relatively soon (e.g., $4 today) and a larger one delivered later (e.g., $7 next week); thereby enabling calculation of the temporal reward discounting rate—that is, the degree at which a subjective reward value decreases as a function of the delay until delivery (Berns, Laibson, & Loewenstein, 2007). In contrast to other psychiatric conditions including bulimia nervosa (BN) and binge-eating disorder (BED) where patients typically show steeper (increased) DD (i.e., greater preference for immediate rewards) suggestive of more impulsive decision making relative to healthy controls (HC; Amlung et al., 2019), several studies found shallower (decreased) DD (i.e., greater propensity to delay reward) indicative of more self-controlled choice in AN (Decker, Figner, & Steinglass, 2015; Steinglass et al., 2012; Steward et al., 2017). However, others have not (Bartholdy et al., 2017; King et al., 2016; Ritschel et al., 2015) and some evidence suggests that the tendency to delay reward in AN may only be detectable in acutely underweight patients (acAN) of the restrictive clinical subtype (Amlung, Marsden, & McCabe, 2020), be dependent on metabolic state (Bernardoni et al., 2020), and disappear following weight restoration (Decker et al., 2015; Doose et al., 2020). Furthermore, no known studies have found altered DD in long-term weight-recovered AN (recAN; King et al., 2020; Wierenga et al., 2015). Thus, it is uncertain whether decreased DD is a stable trait of AN. Together, while heterogeneous findings across DD studies are likely attributable to several factors including cohort effects, limited statistical power, and/or task differences (King & Ehrlich, 2020; Lempert, Steinglass, Pinto, Kable, & Simpson, 2018), empirical evidence supporting the notion of an increased capacity to delay reward in AN is modest.

To validate or fine-tune the hypothesis of a greater willingness to delay reward in AN, it might be helpful to ask patients directly how they typically respond to various rewards in their everyday lives. A related self-report instrument, the Delaying Gratification Inventory (DGI; Hoerger, Quirk, & Weed, 2011), could be informative in this respect. The DGI measures the tendency to forgo immediate gratification in favor of long-term rewards in the following five domains: Food, Physical pleasure, Social interaction, Money, and Achievement. The only known study to employ the DGI in an AN sample found no differences in DGI total scores between acAN and HC, but individuals with BN and BED both showed less delayed gratification compared to both AN and HC (Bartholdy et al., 2017). Here we employed the DGI in a larger sample of acAN at the beginning of weight restoration therapy (acAN-T1), again after short-term initial weight gain (acAN-T2), and in an independent sample of long-term weight-recovered women with a history of AN (recAN). Importantly, given that AN is defined by forgoing food reward, we also calculated a DGI total score after removing food-related items. We tested whether differences relative to HC or longitudinal changes on any of the subscales would be evident only in the underweight state or might persist following remission, as would be expected if a greater propensity to delay gratification constitutes a stable trait of AN. Based on the previous task-based DD findings outlined above, we predicted that if group differences in DGI scores would be evident, they would be more pronounced in the acutely underweight state and weaker following weight restoration.

2 METHODS 2.1 Participants

Data from 241 female volunteers (12–29 years) collected between August 2017 and May 2021 were included in the present study: 71 acAN (clinical subtype: 54 restrictive, 14 binge/purge, 3 incomplete data), 52 recAN (clinical subtype during illness: 34 restrictive, 18 binge/purge) and 120 HC. Fifty-two acAN were reassessed after initial weight gain (minimum increase of >10% body mass index [BMI]; acAN-T2). Data were pooled from ongoing projects within the greater Saxonian Anorexia Nervosa Study lead by our research group that were approved by the Ethics Commission of the Technische Universität Dresden and carried out in accordance with the Declaration of Helsinki. All participants (and their guardians if underage) gave written informed consent.

AcAN participants were admitted to eating disorder programs at the University Hospital Carl Gustav Carus Dresden and were initially assessed within 96 hr after beginning behaviorally oriented nutritional rehabilitation. To be included in the acAN group at admission (acAN-T1), participants needed to have a BMI less than 17.5 kg/m2 if they were older than 15.5 years or a BMI in the 10th age percentile or lower if they were younger than 15.5 years. AcAN participants were invited for reassessment during the last 2 weeks of inpatient treatment. The average time between assessments was 97 days (SD = 27) and the average BMI increase was 23.8% (SD = 6.1; range = 12.9–39.6%). To be considered “recovered,” recAN participants had to (a) maintain a BMI > 18.5 (for participants 18 years and older) or above the 10th age percentile (if younger than 18 years); (b) menstruate; and (c) refrain from binge eating, purging, or other substantially restrictive eating patterns for at least 12 months prior to the study. The HC group consisted of normal-weight (BMI ≥15th and ≤ 85th age percentile if under 18 years, BMI ≥ 18.5, and ≤25 if older than 18), eumenorrhoeic girls and women with no history of psychiatric illness recruited via contact list provided by the local municipality and advertisements at high schools and university campuses. Exclusion criteria are detailed in the Supporting Information Material.

2.2 Procedure

Clinical assessments including structured interviews for current or past eating disorders (Fichter & Quadflieg, 1999), psychiatric comorbidity (Margraf, 1994), anthropometric measurements to calculate BMI, questionnaires to assess eating disorder (EDI-2; Paul & Thiel, 2005), depression (BDI-II; Hautzinger, Keller, & Kühner, 2009), and anxiety symptoms (STAI; Spielberger, Gorsuch, & Lushene, 1970) are described in the Supporting Information Material.

2.3 Delaying gratification inventory

The main variable of interest in the current study, self-reported propensity to delay reward, was assessed using the DGI. The DGI consists of 35 statements that respondents agree or disagree with on a five-point Likert scale, yielding one global score and five domain-specific subscale scores: Food, Physical pleasures, Social interactions, Money, and Achievement. Higher scores indicate a higher tendency to postpone reward. For participants under 16 years old, we omitted two items belonging to the Physical pleasures subscale due to their sexual nature and adapted the calculation of scores accordingly. Internal consistencies (Cronbach's α) were acceptable to good in the overall sample for the total score (.86) and all subscales (≥.73) except for the Physical pleasures subscale (.43). Given the poor reliability of the Physical pleasures subscale and the problematic nature of food-related items in an eating disorder defined by a propensity to forgo food, we also calculated a total score after excluding each of the respective subscales (α = .84 and .86). Similarly acceptable to good internal consistencies in the individual groups (except for poor reliability of the Food subscale in recAN) are summarized in Table S1.

2.4 Statistical analysis

Demographic and clinical measures were analyzed using analyses of variance and paired t tests in the cross-sectional and longitudinal samples, respectively. Our main analyses of the DGI data used (a) a MANCOVA covarying for age to test for cross-sectional group differences and (b) a repeated-measures MANOVA to test for longitudinal change in the respective subscales and total scores, followed by Bonferroni corrected pairwise comparisons. The results of additional sensitivity analyses covarying for depression (BDI-II scores) and anxiety (STAI-S) symptoms, clinical subtype, and based on age-matched groups are presented in the Supporting Information Material. IBM SPSS software (Version 27) and R Studio (RStudio Team, 2020) were used for all analyses.

3 RESULTS

Demographic variables, clinical measures, and DGI scores of the cross-sectional and longitudinal samples are summarized in Tables 1 and 2, respectively.

TABLE 1. Demographic variables, clinical measures, and DGI scores of the cross-sectional sample Group MANCOVA analysis acAN (n = 71) recAN (n = 52) HC (n = 120) η2P F p Posthoc differences Age (years) 16.1 ± 2.5 22.3 ± 3.9 18.9 ± 4.5 .242 38.2 <.001 acAN < HC***acAN, HC < recAN*** Current BMI (kg/m2) 14.6 ± 1.5 21.1 ± 1.8 21.3 ± 2.2 .717 303.5 <.001 acAN < HC, recAN*** Current BMI-SDS −3.2 ± 1.2 −0.4 ± 0.7 −0.0 ± 0.7 .732 327.1 <.001 acAN < HC, recAN***recAN < HC* Minimal lifetime BMI 14.2 ± 1.5 14.5 ± 1.5 20.4 ± 1.9 .761 366.9 <.001 acAN, recAN < HC*** BDI-II 26.2 ± 10.9 7.6 ± 8.6 4.3 ± 4.7 .603 177.8 <.001 acAN ˃ HC, recAN***recAN ˃ HC* STAI-state 49.6 ± 12.2 40.4 ± 10.6 34.3 ± 8.5 .296 48.64 <.001 acAN ˃ HC, recAN***recAN ˃ HC** STAI-Trait 48.5 ± 13.1 38.7 ± 11.0 33.7 ± 8.1 .274 43.40 <.001 acAN ˃ HC, recAN***recAN ˃ HC* EDI-2 total 221.4 ± 44.0 157.9 ± 41.8 131.3 ± 25.0 .549 140.4 <.001 acAN ˃ HC, recAN***recAN ˃ HC*** Drive for thinness 25.0 ± 10.6 27.4 ± 11.1 11.7 ± 4.2 .440 92.81 <.001 acAN, recAN ˃ HC*** Bulimia 10.7 ± 4.2 10.6 ± 3.8 9.3 ± 2.0 .044 5.40 .005 acAN ˃ HC* Bodydissatisfaction 34.3 ± 12.2 36.7 ± 11.6 20.4 ± 7.7 .360 65.84 <.001 acAN, recAN ˃ HC** DGI food 29.5 ± 4.3 25.3 ± 3.3 24.1 ± 3.9 .289 48.59 <.001 acAN ˃ HC, recAN*** DGI physical pleasures 25.8 ± 3.5 24.9 ± 2.9 24.3 ± 3.2 .080 10.41 <.001 acAN ˃ HC***acAN ˃ recAN** DGI social 28.8 ± 3.2 30.3 ± 3.3 29.3 ± 3.2 .003 .344 .709 — DGI money 30.3 ± 4.3 30.7 ± 3.7 29.4 ± 4.1 .019 2.32 .101 — DGI achievement 27.4 ± 4.1 27.2 ± 4.1 26.6 ± 4.4 .012 1.43 .242 — DGI Total 141.8 ± 11.7 138.5 ± 11.2 133.7 ± 13.0 .108 14.48 <.001 acAN ˃ HC***acAN ˃ recAN*** DGI Total excluding physical pleasures 116.0 ± 10.1 113.6 ± 9.6 109.3 ± 11.0 .093 12.31 <.001 acAN ˃ HC***acAN ˃ recAN* DGI total excluding food 112.4 ± 10.2 113.2 ± 9.8 109.6 ± 10.9 .029 3.53 .031 acAN ˃ HC* Note: Data are presented as mean ± SD. The mean duration of the current episode of illness in acAN was 12.67 ± 13.41 months. The mean duration since weight restoration in recAN was 58.02 ± 39.23 months (range: 12–168 months). Fifty-four acAN participants were of the restrictive subtype and 14 were of the binge/purge subtype (incomplete data). Thirty-two recAN participants were of the restrictive subtype and 18 of the binge/purge subtype during illness. Two acAN and nine recAN participants were receiving SSRIs. Fourteen acAN participants received one or more comorbid Axis I psychiatric diagnoses during treatment (six major depression, seven anxiety disorder [five social anxiety disorder, one panic disorder, one non-specified anxiety disorder], three obsessive–compulsive disorder, one conduct disorder, one depressive adaptation disorder). Twenty recAN had one or more active comorbidities (17 major depression, four anxiety disorder [two social anxiety disorder, two panic disorder], one obsessive–compulsive disorder, one posttraumatic stress disorder). Bonferroni corrected significance: ***p < 0.001; **p < 0.01; *p < 0.05. Abbreviations: acAN, acutely underweight anorexia nervosa patients; BDI-II, Beck Depression Inventory-II; BMI, body mass index; BMI-SDS, body mass index standard deviation scores; DGI, Delayed Gratification Inventory; EDI-2, Eating Disorder Inventory-2; HC, healthy control women; MANCOVA, multivariate analysis of covariance; recAN, long-term weight-recovered anorexia nervosa patients; STAI, State-Trait Anxiety Inventory. TABLE 2. Demographic variables, clinical measures, and DGI scores of the longitudinal sample acAN T1 acAN T2 Paired t test Cohen's d T p Age (years) 15.53 ± 1.73 15.80 ± 1.71 −21.1 51 <.001 BMI (kg/m2) 14.55 ± 1.43 19.11 ± 1.05 −28.0 51 <.001 BMI-SDS −3.19 ± 1.08 −0.57 ± 0.47 −20.5 51 <.001 BDI-2 25.77 ± 10.08 19.49 ± 12.57 3.68 51 <.001 STAI-state 49.08 ± 11.12 50.24 ± 13.22 −0.671 49 .505 STAI-Trait 50.24 ± 5.38 NA EDI-2 total 221.4 ± 39.1 220.7 ± 43.8 0.116 46 .908 Drive for thinness 33.43 ± 9.18 30.54 ± 9.47 0.709 50 .482 Bulimia 12.16 ± 5.38 9.95 ± 2.94 2.94 49 .005 Body dissatisfaction 39.21 ± 9.59 42.03 ± 10.55 −2.09 49 .041 MANOVA η2P F p DGI food 29.81 ± 4.00 30.39 ± 3.93 .021 1.08 .304 DGI physical pleasures 25.91 ± 3.59 25.00 ± 3.06 .048 2.59 .113 DGI social 28.81 ± 3.27 28.08 ± 4.47 .042 2.23 .141 DGI money 31.22 ± 3.47 30.90 ± 4.00 .009 0.47 .495 DGI achievement 27.94 ± 4.43 26.94 ± 4.84 .077 4.26 .044* DGI total 143.70 ± 12.21 141.31 ± 13.03 .045 2.39 .128 DGI total excluding physical pleasures 117.79 ± 10.48 116.31 ± 10.89 .025 1.32 .255 DGI total excluding food 113.89 ± 10.45 110.92 ± 11.89 .092 5.18 .027* Note: Data are presented as mean ± SD. Mean duration between T1 and T2 = 96 ± 27 days. Minimal lifetime BMI was 14.22 ± 1.42 for the acAN participants included in the longitudinal analysis (n = 52). *p < .05, corrected. Abbreviations: acAN, acutely underweight anorexia nervosa patients; BDI-II, Beck Depression Inventory-II; BMI, body mass index; BMI-SDS, body mass index standard deviation scores; EDI-2, Eating Disorder Inventory-2; MANOVA, multiple analysis of variance; STAI, State-Trait Anxiety Inventory.

The main cross-sectional analysis of DGI scores revealed significant multivariate effects of both age and group (both F > 6.4; both p < .001). Group differences were evident in the Food and Physical pleasures subscales and total scores both with and without items belonging to these subscales (all F > 3.5; all p < .031). Pairwise comparisons showed that all of these differences were driven by higher scores in acAN relative to HC (and recAN), while recAN did not differ from HC in any DGI measure (Table 1). When also accounting for depression (BDI-II scores) and anxiety (STAI-S) symptoms, the differences between acAN and HC became stronger and additional differences on the Money and Achievement subscales emerged, but still no differences between recAN and HC were detectable (Table S2). A qualitatively similar pattern of results indicative of a heightened propensity to delay reward in acutely symptomatic patients, but not following long-term remission was obtained from an alternative statistical approach comparing individually age-matched acAN-HCacAN (both n = 71; Table S3) and recAN-HCrecAN (both n = 52; Table S4) cohorts including when covarying for comorbid symptoms (Tables S5 and S6). Further sensitivity analyses found no noteworthy influence of clinical subtype (Table S7 and S8) or active psychiatric comorbidity (Table S9) or on the main cross-sectional findings (cf. Tables 1, S3, and S4).

The main longitudinal analysis did not reveal a general treatment effect on DGI scores (F > 1.7; p < .16), but they were significantly decreased following short-term weight gain relative to admission (acAN-T2–acAN-T1) on the Achievement subscale and the total score excluding food-related items (Table 2). However, these changes were no longer signficiant after covarying for depression symptoms (Table S10). Exploratory cross-sectional comparisons between acAN-T2 and age-matched HC (who completed the DGI only once) suggested that the observed group differences in DGI scores between acAN and HC at the beginning of treatment (Tables 1 and S3) were relatively unchanged following initial weight gain (Table S11), especially when accounting for depression symptoms (Table S12).

4 DISCUSSION

Supporting the notion of an increased propensity to delay reward in AN (Steinglass et al., 2012), we found symptomatic patients' DGI total scores to be elevated at the beginning of inpatient treatment (acAN-T1) relative to those of HC even after removing items belonging to the Food or Physical pleasures subscales. This group difference was even more pronounced when accounting for depression and anxiety symptoms: scores were increased in acutely underweight patients on all DGI measures except for the Social interaction subscale. Longitudinally, significant decreases in DGI scores were evident on the Achievement subscale and in the total score excluding food-related items (acAN-T1 > acAN-T2) following initial weight gain. However, both of these effects disappeared when covarying for the improvement in depression symptoms at the end of inpatient treatment. These findings, together with results of exploratory comparisons of DGI scores between acAN-T2 and HC, suggest that the increased tendency to delay reward in AN might be relatively stable and does not change with partial weight restoration alone. We speculate that the overall modest longitudinal change might reflect the fact that acAN-T2 patients were still in treatment and not making the types of reward-related decisions that one must make in everyday life. In any event, speaking against the likelihood that increased delayed gratification might constitute a hardwired trait of AN, no differences in DGI scores were evident between recAN and HC. This suggests that lower (normal) levels of delayed gratification may be an important cognitive correlate of long-term recovery in AN. Together, the current self-report data largely mirror the findings from previous DD experiments in AN samples (reviewed in the Introduction), suggesting that increased delaying of gratification (and decreased DD) may be most readily detectible in the underweight state and may be partially dependent on other factors including comorbid psychiatric symptoms (as observed here), metabolic state (Bernardoni et al., 2020) and clinical subtype (Amlung et al., 2020). Furthermore, underlining the notion that reward responding may be highly domain-specific in AN (Haynos, Lavender, Nelson, Crow, & Peterson, 2020), the current findings suggest that although the tendency to delay reward may be relatively generalized (when accounting for comorbid symptoms), it is much more pronounced for disorder-relevant reinforcers (e.g., food, body-related) than disorder-unspecific ones. This may explain why some studies using monetary rewards have not found decreased DD in AN (Bartholdy et al., 2017; Ritschel et al., 2015). Although disorder-relevant commodities (e.g., food) are not entirely unproblematic in eating disorders research, future studies of intertemporal choice might incorporate such stimuli (cf. e.g., Lim & Bruce, 2015) to help better understand delayed gratification in AN.

Several study limitations deserve mention. First and foremost, although the overall reliability of the DGI was good in our sample, it was poor for the Physical pleasures subscale in all groups and for the Food subscale in the recAN group. Interpretation of the Physical pleasures subscale in isolation is therefore questionable and it is difficult to compare the elevated Food subscale scores in acAN with those observed in recAN. Second, although delayed gratification and DD are (inversely) related constructs (Reynolds & Schiffbauer, 2005), a previous study found no relationship between DGI scores and task-based DD (Bartholdy et al., 2017) and future tests of potential correlations between self-report and objective laboratory measures are needed. Third, given the relatively small number of patients with the binge-purge subtype in our samples, the findings suggestive of no effect of clinical subtype on DGI scores should be interpreted cautiously. Last but not least, the longitudinal results including the exploratory comparisons between acAN-T2 and HC should also be interpreted cautiously given that only patients who successfully completed treatment were re-recruited and the lack of a longitudinal HC group.

To conclude, this study provides self-report evidence of an increased tendency to delay reward in acutely underweight, but not long-term remitted patients with a history of AN. This tendency may be particularly strong in relation to disorder-relevant rewards (e.g., food, body-related) and partially overshadowed by comorbid symptoms. The relative normalization of delayed gratification reported by individuals with a history of AN suggests that the maladaptive tendency to postpone reward experienced by acutely ill individuals may be malleable by treatment.

ACKNOWLEDGMENT

This work was supported by the Deutsche Forschungsgemeinschaft (EH 367/5-1, SFB 940/2).

CONFLICT OF INTERESTS

Veit Roessner has received lecture fees from Eli Lilly, Janssen-Cilag, Medice, Novartis, and was a member of advisory boards of Eli Lilly and Novartis. The other authors declare that there are no conflict of interests.

ETHICS STATEMENT

The Ethical Review Board (Ethikkommission) of the Technische Universität Dresden approved this study (EK 14012011).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Filename Description eat23644-sup-0001-Supinfo.docxWord 2007 document , 75.1 KB

Appendix S1. Supporting Information

Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

REFERENCES

Amlung, M., Marsden, E., Holshausen, K., Morris, V., Patel, H., Vedelago, L., … McCabe, R. E. (2019). Delay discounting as a transdiagnostic process in psychiatric disorders: A meta-analysis. JAMA Psychiatry, 76(11), 1176– 1186. https://doi.org/10.1001/jamapsychiatry.2019.2102 Amlung, M., Marsden, E., & McCabe, R. E. (2020). The elusive nature of delay discounting as a Transdiagnostic process in psychiatric disorders—The devil is in the detail—Reply. JAMA Psychiatry, 77(3), 325– 326. https://doi.org/10.1001/jamapsychiatry.2019.4124 Bartholdy, S., Rennalls, S., Danby, H., Jacques, C., Campbell, I. C., Schmidt, U., & O'Daly, O. G. (2017). Temporal discounting and the tendency to delay gratification across the eating disorder Spectrum. European Eating Disorders Review: The Journal of the Eating Disorders Association, 25(5), 344– 350. https://doi.org/10.1002/erv.2513 Bernardoni, F., Bernhardt, N., Pooseh, S., King, J. A., Geisler, D., Ritschel, F., … Ehrlich, S. (2020). Metabolic state and value-based decision-making in acute and recovered female patients with anorexia nervosa. Journal of Psychiatry and Neuroscience, 45, 253– 261. https://doi.org/10.1503/jpn.190031 Berns, G. S., Laibson, D., & Loewenstein, G. (2007). Intertemporal choice—Toward an integrative framework. Trends in Cognitive Sciences, 11(11), 482– 488. https://doi.org/10.1016/j.tics.2007.08.011 Decker, J. H., Figner, B., & Steinglass, J. E. (2015). On weight and waiting: Delay discounting in anorexia nervosa pretreatment and posttreatment. Biological Psychiatry, 78(9), 606– 614. https://doi.org/10.1016/j.biopsych.2014.12.016 Doose, A., King, J. A., Bernardoni, F., Geisler, D., Hellerhoff, I., Weinert, T., … Ehrlich, S. (2020). Strengthened default mode network activation during delay discounting in adolescents with anorexia nervosa after partial weight restoration: A longitudinal fMRI study. Journal of Clinical Medicine, 9(4), 900. https://doi.org/10.3390/jcm9040900