Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Mar 10.
Published in final edited form as: J Child Psychol Psychiatry. 2024 Oct 31;66(3):378–389. doi: 10.1111/jcpp.14066

Food choice and neural reward systems in adolescents with anorexia nervosa and atypical anorexia nervosa

E Caitlin Lloyd 1,2, Jonathan Posner 3, Janet Schebendach 1,2, Alexandra F Muratore 1,2, Susie Hong 3, Jessica Ojeda 1,2, Elizabeth Rafanello 3, Joanna E Steinglass 1,2,†, Karin Foerde 4,†
PMCID: PMC11892378  NIHMSID: NIHMS2061248  PMID: 39479886

Abstract

Background:

Adolescence is a critical developmental period for the study of anorexia nervosa (AN), an illness characterized by extreme restriction of food intake. The maturation of the reward system during adolescence combined with recent neurobiological models of AN led to the hypothesis that early on in illness, restrictive food choices would be associated with activity in nucleus accumbens reward regions, rather than caudate regions identified among adults with AN.

Methods:

Healthy adolescents (HC, n = 41) and adolescents with AN or atypical AN (atypAN, n = 76) completed a Food Choice Task during fMRI scanning. Selection of high-fat foods and choice-related activation in nucleus accumbens and anterior caudate regions-of-interest (ROIs) were compared between individuals with AN/atypAN and HC. Associations were examined between choice-related activation and choice preferences among the AN group. Exploratory analyses examined associations between choice-related activation and psychological assessments among the patient group.

Results:

Adolescents with AN or atypAN selected fewer high-fat foods than HC (t = −5.92, p < .001). Counter to predictions, there were no significant group differences in choice-related activation in the ROIs. Among individuals with AN or atypAN, choice-related neural activity in the anterior caudate was significantly negatively associated with high-fat food selections in the task (r = −.32, p = .024). In exploratory analyses, choice-related anterior caudate activation was positively associated with psychological measures of illness severity among patients (p’s < .05, uncorrected).

Conclusions:

In this large cohort of adolescents with AN/atypAN, there was no evidence of altered reward system engagement during food choice. While there was no group difference in choice-related caudate activation, the associations with choices and psychological measures continue to suggest that this neural region is implicated in illness. Longitudinal analyses will clarify whether neural variability relates to longer-term course.

Keywords: Anorexia nervosa, adolescence, neuroimaging, reward, caudate

Introduction

Dieting behavior, defined here as changes in food or calorie intake with the intent to lose weight, is common among healthy adolescents. However, for some individuals dieting behavior escalates to become pathological. In anorexia nervosa (AN), what often begins as a dieting-like change in eating becomes an extreme restriction of calorie intake, as demonstrated across laboratory meal studies (Mayer, Schebendach, Bodell, Shingleton, & Walsh, 2012; Sysko, Walsh, Schebendach, & Wilson, 2005) and studies of eating in the environment (Hadigan et al., 2000; Huse & Lucas, 1984). Longitudinal investigations have shown that restrictive eating persists after acute weight restoration treatment (Mayer et al., 2012) and relates to longer-term course (Schebendach et al., 2008; Schebendach, Mayer, Devlin, Attia, & Walsh, 2012; Steinglass et al., 2023). The resulting underweight state has serious medical and psychological sequelae (Arcelus, Mitchell, Wales, & Nielsen, 2011) and restriction of high-fat food intake is maladaptive in this context. Probing the neural mechanisms of restrictive eating among adolescents with AN, who are at the developmental stage in which illness commonly begins, may provide a window into the development of AN and yield a better understanding of illness pathways for identification of more effective treatments.

The neurobiological mechanisms that reinforce maladaptive (restrictive) food choice in AN are puzzling, and it has been posited that perhaps restrictive eating itself is initially rewarding (Walsh, 2013). Adolescence is a time of heightened reward sensitivity (Ernst et al., 2005; Van Leijenhorst et al., 2010), and it may be that for some people this involves a sensitivity to individual-specific rewards related to restrictive eating behavior. Several models of the pathophysiology of AN have centered on reward system abnormalities (Bergh & Södersten, 1996; Haynos, Lavender, Nelson, Crow, & Peterson, 2020; O’Hara, Campbell, & Schmidt, 2015). Adults with AN assign lower reward values to food and have lower hedonic responses to food, compared with their peers (Lloyd & Steinglass, 2018). In non-food domains, some cognitive studies show abnormalities in reward processes, such as slower reinforcement learning rates (Foerde & Steinglass, 2017; Wierenga, Reilly, Bischoff-Grethe, Kaye, & Brown, 2022), and abnormal delay discounting (Decker, Figner, & Steinglass, 2015). Across neuroimaging studies, including structural MRI (Tadayonnejad et al., 2022; Walton et al., 2022), functional (task and resting state) MRI (Cha et al., 2016; Haynos et al., 2019; Tadayonnejad et al., 2022; Uniacke et al., 2018), and positron emission tomography (Bailer et al., 2017; Broft et al., 2015), AN has been associated with abnormalities in frontostriatal reward systems. Though most studies have been conducted with adults, some studies also indicate abnormalities in reward systems among adolescents with AN (Bischoff-Grethe et al., 2013; Frank et al., 2018). In particular, disturbances in reward system activity during the viewing of disorder-relevant stimuli (i.e. images of food and bodies) have been identified among adolescents with AN (Fladung, Schulze, Schöll, Bauer, & Grön, 2013; Horndasch et al., 2018). However, whether the reward system is differentially engaged at the time of food choice among adolescents with AN has yet to be examined.

The Food Choice Task captures the maladaptive restrictive eating behavior that characterizes AN, and task behavior predicts actual food intake in a laboratory meal (Foerde, Steinglass, Shohamy, & Walsh, 2015). Results across two separate cohorts of adults have shown that food choice is associated with neural activity in the anterior caudate among patients with AN, but not healthy controls (HC; Foerde et al., 2015, 2020, 2021). Adults with AN have often been ill for many years and the observed caudate activation may therefore reflect starvation itself, sequalae of long-standing illness, or neurocognitive adaptations over time. That is, behavior such as restrictive food choice may initially be driven by expectation of rewarding outcomes (i.e. weight-loss or another individual-specific reward), presumed to be coded in the nucleus accumbens. With repetition, the behavior may be instigated by associated cues, rather than outcomes, with a corresponding shift in underlying neural mechanisms (Smith & Graybiel, 2016; Steinglass & Walsh, 2016; Walsh, 2013).

A pressing question, given the common onset of AN during adolescence, is whether the findings from studies of adults with AN apply to adolescents. To probe the neural mechanisms of food choice among adolescents with AN and characterize the role of reward systems earlier in illness course, participants in this study completed the Food Choice Task with fMRI scanning. The primary hypotheses were that (a) food choice among adolescents with AN, compared with HC, would be more strongly associated with nucleus accumbens activity and (b) among the AN group, food choice would be more strongly associated with activity within the nucleus accumbens than the caudate. Participants with and without AN or atypical AN (atypAN) were enrolled in a longitudinal, 2-year study designed to examine associations between neural mechanisms of food choice and illness persistence versus remission. The current investigation focused on baseline assessments.

Methods

Participants were adolescents with AN/atypAN (n = 89) and healthy peers (n = 42) between the ages of 12 and 18 years. Participants were recruited via media notices and referrals from health professionals. Inpatient and/or outpatient treatment was available at the New York State Psychiatric Institute (NYSPI) at no cost for all patients. Individuals were included if they were assigned female sex at birth, had no major medical or neurologic illness, had normal (or corrected to normal) vision, had an estimated IQ above 80 (assessed via Weschler Abbreviated Scale of Intelligence; WASI; Wechsler, 1999), and had no food allergies that interfered with participation. Psychotropic medications within 4 weeks of the study were exclusionary for patients, except for antidepressants (i.e. SSRIs were not exclusionary); all psychotropic medications were exclusionary for HC. Patients were excluded if they had a co-occurring diagnosis that required specialized treatment (e.g. substance use disorder, bipolar, or psychotic illness), or were at high/imminent risk for suicide. HC had no current or lifetime psychiatric illness and a body mass index (BMI) between the 5th and 85th percentile for sex and age. Individuals who provided data for the current study had no contraindications to MRI.

Eating disorder diagnoses were established using the Eating Disorder Diagnostic Assessment for DSM-5 (Sysko et al., 2015). Patients met criteria for AN if BMI at the time of EDA-5 assessment had been below the 10th percentile for age and sex within the past 3 months; patients met criteria for atypAN if all other diagnostic criteria for AN were satisfied but BMI was not, or had not been, below the 10th percentile. Co-occurring diagnoses were assessed by the Schedule for Affective Disorders and Schizophrenia for School-Age Children (Chambers & Puig-Antich, 1978). Duration of illness and dietary restriction were assessed via Age of Onset interview (Ranzenhofer et al., 2022). Pubertal stage was assessed using the Tanner staging method (Marshall & Tanner, 1969): participants identified their breast and pubic hair development stage and ordinal scores were averaged to provide the Tanner stage.

Ethical considerations

The study was approved by NYSPI Institutional Review Board. Written informed consent, or participant assent and guardian consent, was obtained. Data were collected between May 2017 and April 2022.

Procedures

Participants received a standardized lunch at 12 pm (~550 kcal; turkey or peanut butter and jelly sandwich, Nutrigrain bar, 8 oz water) and were instructed to have nothing to eat or drink, except water, between lunch and the laboratory multi-item meal at 5 pm. The Food Choice Task with fMRI scanning took place at 2 pm. Height and weight were measured on the day of participation. The study day is described in full here: https://github.com/Columbia-Center-for-Eds/Longitudinal-Assessment-of-Teens-with-Anorexia-Nervosa.

Food Choice Task (Figure 1)

Figure 1.

Figure 1

Open in a new tab

Food choice task. Participants completed the Food Choice Task with fMRI scanning. The task comprises two rating phases (A, B) and a choice phase (C). During the Choice phase, on each trial, participants indicated their preference for a changing food item (shown on the right) relative to a repeating neutral-rated reference item (on the left)

The Food Choice Task (Steinglass, Foerde, Kostro, Shohamy, & Walsh, 2015; freely available at https://github.com/Columbia-Center-for-Eds/Food-Choice-Task) consists of three blocks: healthiness rating, tastiness rating, and choice (Figure 1). The number of high-fat food choices in the Food Choice Task is reliably associated with actual food intake at a multi-item meal (Foerde et al., 2015, 2020). In each block, participants rated 76 images of food on a five-point scale. Half of the foods were ‘high fat,’ defined as >30% of calories derived from fat. Images, macronutrient content, image properties, and normative ratings are available here: https://doi.org/10.7916/d8-497c-2724.

After completing healthiness and tastiness rating phases (order counterbalanced across participants), a ‘Reference’ item was selected, rated by the participant as ‘Neutral’ on both healthiness and tastiness. Eight individuals had a Reference item rated as non-neutral for tastiness. In the choice block, participants indicated a preference for each of 75 trial-unique food items, with the Reference item presented on the left. To ensure participants responded according to true preferences, they were instructed that one trial would be randomly selected, and that they would be served a snack-sized portion of the chosen food as a snack later that day.

On each trial, participants had 4 s to make their response. The inter-trial interval (ITI) was jittered (mean ITI = 2.3 s, median = 2 s, range = 1–10 s, across all three phases). A fixation cross was presented during the ITI. Each run lasted 480 s. All task phases were presented using Matlab and the Psychophysics toolbox (Brainard, 1997).

fMRI data acquisition

Whole-brain imaging was conducted on a 3.0 T GE MRI system, using a SENSE head coil. Structural images were collected using a high-resolution T1-weighted Spoiled Gradient-Recalled (SPGR, BRAVO) pulse sequence (1 mm3 voxel size). Functional images were initially (n = 14 HC; 24 AN) acquired with a multi-echo T2*-weighted echoplanar (EPI) sequence. This sequence was changed due to a scanner upgrade part-way through the study (n = 27 HC; 52 AN), and replaced with a gradient (single) echo T2*-weighted EPI sequence.

Multi-item meal

The laboratory meal followed standardized procedures (Sysko, Steinglass, Schebendach, Mayer, & Walsh, 2018) and consisted of items of varied fat and calorie content. Food was weighed before and after the meal (Sartorius 8200 electronic balance, readability 0.1 g), and intake in grams, kcal, and macronutrient content (e.g. % kcal from fat) was calculated. The association between the number of high-fat food choices in the task and meal intake was examined to confirm task validity among adolescents.

Psychological assessments

Questionnaire measures of the primary constructs of interest – restrictive eating, habit, and reward – were obtained in addition to self-reported eating disorder severity. Restrictive eating was assessed using the Three Factor Eating Questionnaire, Restraint subscale (TFEQ; Stunkard & Messick, 1985). Eating disorder severity was assessed with the Eating Disorder Examination Questionnaire (EDE-Q) global score (Fairburn & Beglin, 1994). Habit strength was assessed with the Self-Report Habit Index (SRHI), adapted for use in AN (Davis, Walsh, Schebendach, Glasofer, & Steinglass, 2020). Reward sensitivity was assessed with the relevant subscale of the (parent-report) Sensitivity to Reward and Punishment Questionnaire for Children (SPSRQ-C; Colder & O’Connor, 2004). Exploratory analyses examining the association between these questionnaire assessments and choice-related activation in regions of interest (ROIs) were undertaken to provide secondary validation of the relationship between the constructs of interest (i.e. restrictive eating, habit and reward) and neural activation patterns in ROIs.

Behavioral data analysis

Analyses were completed in Rstudio. Behavioral and neuroimaging analyses implemented methods that were robust to the unequal size of the AN and HC groups (i.e. methods that do not assume homogeneity of variance): Huber–White standard errors (White, 1980) were calculated for fixed effects models (for non-repeated measures data); mixed effects models estimated associations in repeated measures data (Pinheiro, 2014). Demographic characteristics were compared between diagnostic groups (HC vs. AN/atypAN) using independent samples t-tests. Age and IQ were included as covariates in neuroimaging analyses given their association with striatal activation during reward learning and decision making (Barkley-Levenson & Galván, 2014; Civai, Hawes, DeYoung, & Rustichini, 2016; Ripke et al., 2015; Van Leijenhorst et al., 2010). Pubertal (Tanner) stage was not included as a covariate because it was not associated with brain activation (nucleus accumbens r = .01, p = .89; caudate r = −.03, p = .75; see Table S7 for results with Tanner stage as a covariate).

Food choice task.

Healthiness and tastiness ratings were averaged for high- and low-fat food items to generate mean rating scores. Choice ratings were converted to binary Yes/No preference responses for the trial-unique food item. Neutral responses were omitted and the proportion of choices of the trial-unique food calculated separately for high-fat and low-fat foods. Ratings and choices were analyzed using mixed effects linear regression (including a within-subject effect of Food-type [Low-fat/High-fat], a between-subject effect of Group [HC/AN], and a random intercept per participant).

To understand the influence of healthiness and tastiness ratings on choices, the binary choice variable was regressed onto continuous healthiness and tastiness ratings (z-scored for each participant), Group (HC/AN), and their interaction, in a multilevel logistic regression model. Multilevel linear regression analysis assessed the association between tastiness and healthiness ratings: tastiness ratings were regressed onto z-scored healthiness ratings. Multilevel models included a random intercept and slope per participant and a random intercept per food.

FMRI data analysis

Pre-processing.

Multi-echo data were preprocessed using multi-echo independent components analysis pipeline (ME-ICA; version 3.2; Kundu, Inati, Evans, Luh, & Bandettini, 2012). Single echo images were preprocessed using fMRIprep (version 20.2.1; Esteban et al., 2019). All images were aligned to the MNI152asym2009c template; multi-echo images were down-sampled to match the voxel size of single-echo images (3 mm).

First level analysis.

In all models, each choice event was convolved with a canonical hemodynamic response function and entered into a general linear model (GLM). The GLM included the regressors: (i) onsets for each trial on which a response was made, (ii) onsets for each trial in which a response was made parametrically modulated by the (demeaned) choice rating on that trial, (iii) onsets for each trial in which a response was made parametrically modulated by response time (RT; demeaned) on that trial, and (iv) onsets for missed trials. The event duration was specified as the trial RT for regressors i–iii, and the trial length (4 s) for regressor iv.

Variables capturing motion in the x, y, z, planes (rotation and translation) were included as nuisance regressors in GLMs (single and multi-echo scans). For single-echo images, motion derivatives, framewise displacement (FD) and derivative of root mean square variance parameters, were included as additional motion confounds, and single time-point regressors modeled out motion spikes (volumes per run: MAN = 13.4, MHC = 20).Multi-echo scans underwent more extensive motion correction during preprocessing (including removal and replacement of motion spikes), rendering additional motion regressors redundant. Across single-echo and multi-echo scans, mean FD was similar between AN and HC groups (MAN = 0.10 ± 0.09, MHC = 0.12 ± 0.20).

Harmonization.

NeuroCombat (Fortin et al., 2018) harmonized parameter maps produced by level-one analyses to remove effects of scan-type (i.e. single vs. multi-echo). The harmonization model included age, IQ, BMI percentile, group, and proportion of high-fat selections in the Food Choice Task. Harmonized data were used in all analyses.

Second-level analyses.

Region of Interest (ROI) analyses: The right nucleus accumbens and anterior caudate ROIs were obtained from the Harvard–Oxford probabilistic atlas, thresholded at 25% probability (Figure 2). Following previous studies (Foerde et al., 2015, 2020, 2021), the anterior caudate was defined as the area anterior to y = 0. The z-statistic describing the standardized association between onsets and rating-related neural activation was extracted and averaged across ROI voxels for each individual.

Figure 2.

Figure 2

Open in a new tab

Food choice task behavior. (A) Low-fat items were rated higher on healthiness than high-fat items. Individuals with AN or atypAN rated food, and high-fat foods in particular, as lower in healthiness compared to HC. (B) The AN group rated high-fat foods lower in tastiness relative to HC. (C) The AN group was less likely to choose high-fat foods than the HC group; groups did not differ on low-fat selections. (D) Choice was influenced more by tastiness among HC relative to individuals with AN or atypAN, whereas choice was influenced more by healthiness among individuals with AN or atypAN relative to HC. (E) Tastiness and healthiness ratings were more strongly associated among individuals with AN or atypAN than HC. (F) Associations between the proportion of high-fat food choices on the task and actual caloric intake at the multi-item meal. * indicates p < .01

Choice-related activation in the ROIs was compared between groups using multiple linear regression. To compare relative activation in nucleus accumbens and anterior caudate ROIs, linear mixed-models were completed for AN and HC groups separately, with a random intercept for each participant and a fixed effect of region. To assess group differences in relative activation, group and group by region interaction terms were added to statistical models. These analyses were repeated excluding individuals with atypAN (which did not change the pattern of results; Table S8).

The primary analyses probing group differences in choice-related activation within striatal ROIs were subject to Bayesian model averaging analyses (Hinne, Gronau, van den Bergh, & Wagenmakers, 2020) to compare the probability of models with and without a group term (see Supporting Information). This approach allows for accepting as well as rejecting the null hypothesis (Keysers, Gazzola, & Wagenmakers, 2020).

Whole-brain analyses:

To explore group differences beyond a priori ROIs, individuals with AN or atypAN (the AN group) and HC were compared in whole-brain analyses (Figure S2).

Association between brain and behavior:

Partial correlation analysis (Pearson estimation) assessed the relationships between choice-related activation (in the nucleus accumbens and anterior caudate) and selection of high-fat foods, separately for AN/atypAN and HC groups. p-values for these four tests were corrected for multiple comparisons using the Bonferroni method. Correlations of nucleus accumbens and anterior caudate activation with food choice were compared within groups using the Steiger z-test.

Correlation analysis examined relationships between choice-related activation in striatal ROIs and psychological measures among individuals with AN or atypAN; as these tests were exploratory, multiple comparison correction was not applied.

The main analyses of this study were repeated excluding individuals with atypAN, which did not change the pattern of results (see Table S8).

Results

Data from 76 patients with AN (nine of whom met criteria for atypAN) and 41 HC were included in analyses after exclusions due to missing task/fMRI data (six AN, one HC) or poor scan quality (seven AN). Demographic and clinical characteristics are reported in Table 1. Two patients with AN did not identify as female. Groups were well-matched for age but differed in estimated IQ (p = .03). There were no differences between groups in self-reported ethnicity (Χ2 (1, N = 115) = 0.0, p > .99). Groups differed in self-reported race (Χ2 (3, N = 89) = 12.8, p = .005), but not in proportion who identified as White (Χ2 (1, N = 89) = 1.34, p = .25). Mean duration of illness among individuals with AN or atypAN was 7.59 months (range 0 to 59.3 months); the majority (n = 60, 78.9%) had been ill for less than 1 year. For most patients, study procedures occurred within 2 weeks of enrollment (median = 6 days).

Table 1.

Demographic and clinical characteristics

HC (n = 41)
Mean + SD 		AN (n = 76)
Mean ± SD 		P
Age (years) 15.73 ± 1.64 15.42 ± 1.54 .32
Education (years) 10.46 ± 1.60 9.85 ± 1.87 .07
Estimated IQ 112.17 ± 12.86 107.03 ± 11.2 .03
Enrollment BMI percentile 52.49 ± 23.43 11.25 ± 14.0a <.001
AN: 7.26 ± 7.40
AAN: 40.10 ± 17.60a
Procedure day BMI percentile 53.95 ± 24.49 15.96 ± 16.09a <.001
AN: 11.81 ± 11.33
AAN: 38.90 ± 21.16
Tanner stage 4.05 ± 0.93 3.61 ± 0.88 .02
Illness duration (years) NA 0.63 ± 0.93
Duration of restrictive eating (years) NA 2.06 ± 1.90
TFEQ restraintb 5.31 ± 3.48 15.73 ± 5.47 <.001
TFEQ disinhibitionb 3.82 ± 1.82 3.43 ± 2.47 .35
TFEQ hungerb 4.13 ± 2.88 3.30 ± 2.68 .14
EDE-Q, Globalb 0.52 ± 0.73 3.56 ± 1.64 <.001
SRHI totalb 3.63 ± 1.08 4.94 ± 1.01 <.001
SPSRQ-C reward sensitivityb 2.89 ± 0.57 2.97 ± 0.63 .51
Multi-item meal (kcal)b 921.44 ± 331.16 368.66 ± 334.38 <.001
Multi-item meal (% kcal from fat)b 41.14 ± 5.42 22.61 ± 15.21 <.001
Frequency (n, %)
Restricting subtype NA 53 (69.7%)
Unmedicated NA 64 (84.2%)
Psychiatric diagnoses
 AN only 59 (64.5%)
 Depressive disorder 16 (21.1%)
 Anxiety disorder 25 (32.9%)
 Obsessive compulsive disorder 4 (5.3%)
 Attention deficit disorder 1 (1.32%)
 Tic disorder 1 (1.32%)
Racec
 Asian 5 (15.63%) 14 (24.6%) <.005
 Black 8 (25.0%) 1 (1.75%)
 Multiple races 0 (0.0%) 1 (1.75%)
 White 19 (59.4%) 41 (71.9%)
Ethnicityc
 Hispanic 10 (25.0%) 18 (24.0%) >.99
 Non-hispanic 30 (75.0%) 57 (76.0%)
Open in a new tab

AAN, atypical anorexia nervosa; AN, anorexia nervosa; BMI, body mass index; EDE-Q, Eating Disorder Examination, questionnaire version; HC, healthy control; SRHI, Self-report Habit Index; SPSRQ-C, Sensitivity to Reward and Punishment Questionnaire for Children; TFEQ, Three Factor Eating Questionnaire.

a

10 individuals diagnosed with atypical AN; 66 with AN.

b

TFEQ data missing for 2 AN and 2 HC; EDE-Q missing for 3 AN and 2 HC; SRHI data for missing 1 AN and 1 HC; SPSRQ-C data missing for 16 AN and 2 HC. Meal data missing for 8 AN and 1 HC.

c

Race data missing for 19 AN and 9 HC; Ethnicity data missing for 1 AN and 1 HC.

Food choice task and multi-item meal behavior (Figure 2)

Patients with AN or atypAN rated foods as less healthy than HC, in particular high-fat foods (Group × Food-type interaction: B = −0.09, 95% CI: [−0.15, −0.04], p < .001; Figure 2A; Table S1). Patients rated high-fat foods, though not low-fat foods, as less tasty compared to HC (Group × Food-type interaction: −0.10, 95% CI: [−0.16, −0.04], p = .001; Figure 2B; Table S2).

In the choice phase, patients with AN or atypAN were less likely to choose high-fat foods compared to HC, and there were no differences for low-fat foods (Group × Food-type interaction: B = −0.08, 95% CI: [−0.10, −0.06], p < .001; Figure 2C; Table S3).

Food intake during the laboratory Multi Item Meal was significantly lower among individuals with AN or atypAN versus HC (Table 1), and intake was positively associated with the proportion of high-fat choices on the food choice task among the AN group (r = .39, p < .001), but not HC (r = .25, p = .12; Figure 2F).

Mixed effects logistic regression indicated that individuals with AN or atypAN based their choices on healthiness to a greater extent than HC (Group × Healthiness interaction: B = 0.71, 95% CI [0.50, 0.91], p < .001; Table S4). Groups did not differ in the influence of tastiness on choice (Figure 2D). A mixed effects linear regression assessing the relationship between healthiness and tastiness ratings showed that the positive relationship between these attributes was greater among individuals with AN or atypAN (Group × Healthiness interaction: B = 0.12, 95% CI [0.10, 0.14], p < .001; Figure 2E; Table S5).

Choice-related engagement in ROIs (Figure 3)

Figure 3.

Figure 3

Open in a new tab

Region of interest (ROI) analyses. (A) There were no significant differences between groups in choice-related activation of striatal regions. Evidence was strong for the absence of a difference between groups in the nucleus accumbens (left) and inconclusive for the anterior caudate (right). (B) Choice of high-fat foods was not associated with activation of the nucleus accumbens during choice (left) but was negatively associated with choice-related anterior caudate activation among individuals with AN or atypAN (right). *p < .01

There was no significant difference between groups in choice-related activation in the right nucleus accumbens (B = 0.02, 95% CI: [−0.18, 0.22], p = .841), and choice-related activation in this ROI was not associated with selection of high-fat items among individuals with AN or atypAN (r = −.09, padj = 0.99) or HC (r = .03, padj = .99).

There was no significant difference between groups in choice-related activation in the right anterior caudate (B = 0.15, 95% CI: [−0.05, 0.35], p = .139). However, choice-related activation was significantly negatively correlated with selection of high-fat foods among patients with AN or atypAN (AN: r = −.32, padj = .024; HC: r = −.02, padj = 0.99). Among patients, the magnitude of the correlation was significantly greater between high-fat choices and caudate activation than between high-fat choices and nucleus accumbens activation (t-value = 2.18, p = .033).

Among HC, choice-related activation was stronger in the nucleus accumbens relative to the anterior caudate (B = −0.24, 95% CI: [−0.35, −0.12], p < .001), whereas there was no difference in choice-related activation between the nucleus accumbens and anterior caudate among individuals with AN or atypAN (B = −0.08, 95% CI: [−0.19, 0.02], p = .11). The relative engagement of the nucleus accumbens versus anterior caudate did not differ significantly between groups (Group × ROI: B = 0.08, 95% CI: [−0.01, 0.16], p = .068).

Bayesian model averaging indicated that the evidence for there being no difference between groups was very strong for choice-related activation of the nucleus accumbens, but inconclusive for the anterior caudate (Table S6; Figure S1).

Associations with psychological assessments

Among individuals with AN or atypAN, cognitive restraint (measured by TFEQ), and strength of illness-relevant habits (SRHI Total Score), were positively associated with choice-related activation in the anterior caudate (ps < .05, uncorrected for multiple comparisons; Figure 4). These relationships were not observed in the nucleus accumbens.

Figure 4.

Figure 4

Open in a new tab

Associations between choice-related activation in striatal regions of interest (ROIs) and psychological assessments among individuals with AN or atypAN. Associations between (A) TFEQ restraint subscale, (B) Global EDE-Q score, (C) SRHI total score, (D) SPSRQ-C reward sensitivity, and choice-related nucleus accumbens activation (left column) and choice-related anterior caudate activation (right column). Analyses were adjusted for age and IQ, p-values are uncorrected for multiple comparisons.

Discussion

This study examined the neural mechanisms of food choice among adolescents with AN (or atypAN) and age-matched HC. A priori hypotheses focused on reward regions of interest in adolescent development (nucleus accumbens; Casey, Getz, & Galvan, 2008), and dorsal striatal regions (anterior caudate) associated with restrictive food choice among adults with AN (Foerde et al., 2015, 2021). Contrary to predictions, there were no significant group differences in either ROI. There were differences in relative use of the regions such that HC (but not AN) engaged the nucleus accumbens more than the caudate. Choice-related caudate activation was associated with restrictive food choices among individuals with AN or atypAN, but not HC, whereas no relationship was found between the nucleus accumbens and food choices in either group. Moreover, among the AN group, choice-related caudate activation was associated with some measures of illness severity.

This study extends prior findings of the neural mechanisms of AN by examining the link between brain and behavior among adolescents who were early in illness. Like adults with AN, adolescents with AN or atypAN selected fewer high-fat options in the Food Choice Task, reported healthier rated foods as tastier and preferred items rated as subjectively healthy. Selection of high-fat foods in the task was related to actual intake in the laboratory meal (among the AN group), supporting the validity of this task in measuring restrictive eating. Patients were generally within a year of illness diagnosis at the time of study participation, suggesting the characteristic eating pathology of AN is present early in illness. The convergence of healthiness and tastiness ratings at such early stages of illness is particularly striking and might be examined in future studies for its role in illness maintenance.

The neural findings somewhat differ from what has been found among adults with AN. Across two previous studies, distinct cohorts of adults with longstanding AN showed choice-related caudate activation that differed significantly from HC (Foerde et al., 2015, 2020). In the larger of the two studies, choice-related caudate activation did not change following weight restoration, though patients who most increased their selection of high-fat foods following treatment also showed a decrease in this activation (Foerde et al., 2021). The lack of significant neural differences between patients and HC among adolescents raises questions. The null findings in the caudate might be due to greater variability in the neural mechanisms of food choice among adolescents with AN (relative to adults). This variability may be relevant to illness progression. It is possible that adolescents with AN with greater caudate engagement at baseline are more likely to remain ill. That is, caudate-related activation predicts sustained illness. Alternatively, caudate engagement during food choice may increase as AN progresses (Treasure, Stein, & Maguire, 2015), such that choice-related caudate activation is an outcome of long-standing illness. Of note, in this study, choice-related activity in the anterior caudate was not associated with duration of illness or duration of restrictive eating. Longitudinal analyses from this cohort will be able to test whether choice-related caudate activation relates to illness remission versus persistence over 2 years.

Variability of caudate engagement in this large sample of adolescents with AN or atypAN may have limited the sensitivity to detect true between-group differences in the current sample. To examine this possibility, Bayesian analyses were conducted, enabling inferences about the probability of a negative finding. Outcomes indicated that there was insufficient evidence to support the alternate or null hypothesis in relation to the anterior caudate. Meaning, while the group differences were not statistically significant, the lack of difference was not robust. The possibility that choice-related caudate activation relates to illness in adolescents with AN is also supported by the associations with specific measures of psychopathology: restrictive choices on the task, and (in exploratory tests) self-report measures of restrictive eating and strength of illness-relevant habits. The alignment of findings across different assessments of restrictive eating somewhat mitigates concerns that identified brain-behavior relationships may be spurious. However, multiple comparison correction was not applied to exploratory tests and findings are considered preliminary.

The Bayesian analyses in relation to the nucleus accumbens indicated that evidence was conclusive for the absence of a group difference, counter to the hypothesis that restrictive eating is underpinned by ventral reward systems early in illness. Other neuroimaging investigations of adolescents with AN have identified elevated activation within ventral striatal and frontal systems during unexpected reward receipt (DeGuzman, Shott, Yang, Riederer, & Frank, 2017; Frank et al., 2018) and reward feedback (Ehrlich et al., 2015), and in response to viewing images of thin bodies (Fladung et al., 2010), positive emotional stimuli (Horndasch et al., 2018) and food images (Kerr, Moseman, Avery, Bodurka, & Simmons, 2017). The present results suggest that alterations in reward regions may not be implicated in restrictive eating choices, per se. Notably, other studies have identified elevated activation within the dorsal striatum (caudate and putamen) during punishment receipt (Bischoff-Grethe et al., 2013) and the viewing of food images (Eddy et al., 2023), among adolescents with AN relative to HC. The current findings more directly implicate dorsal striatal systems in the restrictive eating pathology of AN.

This study has several strengths, including the large sample of teens with AN, relative to existing studies. The procedures were standardized (including the timing of food intake before the MRI scan), and the brain and behavior investigation follows the methods of several prior studies in adults, allowing for reasonable comparisons with existing literature. The enrolled participants comprised a relatively diverse group of teens, with greater representation of minoritized groups than usual for studies of AN (Burnette, Luzier, Weisenmuller, & Boutté, 2022; Udo & Grilo, 2018). Finally, patients were early on in illness, mitigating some of the challenges in disentangling mechanisms of illness from effects of persistent malnourishment.

Nonetheless, the findings must be interpreted in light of several limitations. The sample size remains small relative to data from other psychiatric illnesses, and larger-scale multi-site studies are still needed. This may be especially true for adolescents, given that considerable variability in eating habits among HC teens may make comparisons with patients difficult. Groups differed in estimated IQ (which models covaried for). A number of participants with AN or atypAN (n = 12) were taking antidepressant medication, though results were unaffected by removing these participants (see Table S9). In addition, analyses did not covary for estrogen levels (associated with reward system activation among female HC; Op de Macks et al., 2016) as levels were undetectable among many individuals in the AN group. Finally, variability related to MRI scanner changes may have remained even after statistical adjustment.

Conclusion

In contrast to hypotheses, restrictive eating at early stages of AN was not related to altered reward system activity. Furthermore, while food choice behaviors were remarkably similar to adults with AN, there was not conclusive evidence that choice-related activation in the anterior caudate differed between adolescents with and without AN (as has been shown among adults). There may be greater variability among adolescents with AN or HC relative to adults, as adolescence is a period of brain development. The relationship between multiple indices of illness/restrictive eating severity and choice-related caudate activation supports exploring the prognostic utility of this brain measure in AN.

Supplementary Material

Supplementary Material

Appendix S1 Table S1. Analysis of healthiness ratings: Group (HC/AN) X Food type (Low- fat/High-fat).

Table S2. Analysis of tastiness ratings: Group (HC/AN) X Food type (Low- fat/High-fat).

Table S3. Analysis of choices: Group (HC/AN) X Food type (Low- fat/High-fat).

Table S4. Analysis of healthiness and tastiness rating influence on choice by Group (HC/AN).

Table S5. Analysis of Healthiness rating influence on Tastiness ratings by Group (HC/AN).

Table S6. Prior and posterior model probabilities for prediction of choice-related neural activation.

Figure S1. Prior and posterior distributions for effect of diagnostic term on choice-related activation in the nucleus accumbens and anterior caudate.

Table S7. Results of sensitivity analyses adjusting for Tanner stage.

Table S8. Results of sensitivity analyses excluding individuals with atypical AN.

Table S9. Results of sensitivity analyses excluding individuals taking SSRI medication at the time of the study.

Figure S2. Whole-brain analyses of Choice phase.

Key points.

  • Restrictive eating patterns, with specific limited choice of high-fat foods, are a central disturbance in anorexia nervosa and evident even among adolescents who are early in illness.

  • Anterior caudate activation is related to patterns of maladaptive restrictive eating in adolescents with anorexia nervosa or atypical anorexia nervosa.

  • Food choice-related activation did not differ between patients and healthy peers in the hypothesized reward regions (nucleus accumbens and caudate).

  • Relative engagement of these reward regions did differ between groups such that the healthy teens showed more ventral (nucleus accumbens) than dorsal (caudate) choice related activation, whereas teens with anorexia nervosa or atypical anorexia nervosa did not.

Acknowledgements

The reported research was supported by funding from NIMH R01 MH110445; K24 MH113737; T32 MH096679; Hilda and Preston Davis Foundation. The funding sources had no involvement in study design, data collection, analysis, interpretation of data or writing of the report. The authors would like to acknowledge Jordan Dworkin, PhD for his guidance with respect to biostatistical analyses. JS has received royalties from Springer, Oxford, and UpToDate and research support from Compass Pathways. JP has received research support from Takeda (formerly Shire) and Aevi Genomics and consultancy fees from Innovative Science and AlphaSights. ER receives support from Island Psychiatry, PC. The remaining authors have declared that they have no competing or potential conflicts of interest.

Footnotes

Supporting information

Additional supporting information may be found online in the Supporting Information section at the end of the article:

Ethical approval

The study was approved by NYSPI Institutional Review Board. Written informed consent, or participant assent and guardian consent, was obtained.

Conflict of interest statement: See Acknowledgements for full disclosures.

Data availability statement

The data that support the findings of this study are available upon request and in the NIMH Data Archive at https://nda.nih.gov/edit_collection.html?id=2633, reference number 2633.

References

  1. Arcelus J, Mitchell AJ, Wales J, & Nielsen S (2011). Mortality rates in patients with anorexia nervosa and other eating disorders: A meta-analysis of 36 studies. Archives of General Psychiatry, 68, 724–731. [DOI] [PubMed] [Google Scholar]
  2. Bailer UF, Price JC, Meltzer CC, Wagner A, Mathis CA, Gamst A, & Kaye WH (2017). Dopaminergic activity and altered reward modulation in anorexia nervosa-insight from multimodal imaging. The International Journal of Eating Disorders, 50, 593–596. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Barkley-Levenson E, & Galván A (2014). Neural representation of expected value in the adolescent brain. Proceedings of the National Academy of Sciences of the United States of America, 111, 1646–1651. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bergh C, & Södersten P (1996). Anorexia nervosa, self– starvation and the reward of stress. Nature Medicine, 2, 21–22. [DOI] [PubMed] [Google Scholar]
  5. Bischoff-Grethe A, McCurdy D, Grenesko-Stevens E, L.E. (Zoe) Irvine, Wagner A, Wendy Yau W-Y, … & Kaye WH (2013). Altered brain response to reward and punishment in adolescents with anorexia nervosa. Psychiatry Research: Neuroimaging, 214, 331–340. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Brainard DH (1997). The psychophysics toolbox. Spatial Vision, 10, 433–436. [PubMed] [Google Scholar]
  7. Broft A, Slifstein M, Osborne J, Kothari P, Morim S, Shingleton R, … & Timothy Walsh B (2015). Striatal dopamine type 2 receptor availability in anorexia nervosa. Psychiatry Research: Neuroimaging, 233, 380–387. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Burnette CB, Luzier JL, Weisenmuller CM, & Boutté RL (2022). A systematic review of sociodemographic reporting and representation in eating disorder psychotherapy treatment trials in the United States. International Journal of Eating Disorders, 55, 423–454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Casey BJ, Getz S, & Galvan A (2008). The adolescent brain. Developmental Review, 28, 62–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cha J, Ide JS, Bowman FD, Simpson HB, Posner J, & Steinglass JE (2016). Abnormal reward circuitry in anorexia nervosa: A longitudinal, multimodal MRI study. Human Brain Mapping, 37, 3835–3846. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Chambers W, & Puig-Antich J (1978). Kiddie-Sads-Present and Lifetime Version (K-SADS-PL). Biometrics Research. [Google Scholar]
  12. Civai C, Hawes DR, DeYoung CG, & Rustichini A (2016). Intelligence and extraversion in the neural evaluation of delayed rewards. Journal of Research in Personality, 61, 99–108. [Google Scholar]
  13. Colder CR, & O’Connor RM (2004). Gray’s reinforcement sensitivity model and child psychopathology: Laboratory and questionnaire assessment of the BAS and BIS. Journal of Abnormal Child Psychology, 32, 435–451. [DOI] [PubMed] [Google Scholar]
  14. Davis L, Walsh BT, Schebendach J, Glasofer DR, & Steinglass JE (2020). Habits are stronger with longer duration of illness and greater severity in anorexia nervosa. The International Journal of Eating Disorders, 53, 413–419. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Decker JH, Figner B, & Steinglass JE (2015). On weight and waiting: Delay discounting in anorexia nervosa pretreatment and posttreatment. Biological Psychiatry, 78, 606–614. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. DeGuzman M, Shott ME, Yang TT, Riederer J, & Frank GKW (2017). Association of elevated reward prediction error response with weight gain in adolescent anorexia nervosa. American Journal of Psychiatry, 174, 557–565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Eddy KT, Plessow F, Breithaupt L, Becker KR, Slattery M, Mancuso CJ, … & Lawson EA (2023). Neural activation of regions involved in food reward and cognitive control in young females with anorexia nervosa and atypical anorexia nervosa versus healthy controls. Translational Psychiatry, 13, 220. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Ehrlich S, Geisler D, Ritschel F, King JA, Seidel M, Boehm I, … & Marxen M (2015). Elevated cognitive control over reward processing in recovered female patients with anorexia nervosa. Journal of Psychiatry and Neuroscience, 40, 307–315. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Ernst M, Nelson EE, Jazbec S, McClure EB, Monk CS, Leibenluft E, … & Pine DS (2005). Amygdala and nucleus accumbens in responses to receipt and omission of gains in adults and adolescents. NeuroImage, 25, 1279–1291. [DOI] [PubMed] [Google Scholar]
  20. Esteban O, Markiewicz CJ, Blair RW, Moodie CA, Isik AI, Erramuzpe A, … & Snyder M (2019). fMRIPrep: A robust preprocessing pipeline for functional MRI. Nature Methods, 16, 111–116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Fairburn CG, & Beglin SJ (1994). Assessment of eating disorders: Interview or self-report questionnaire? International Journal of Eating Disorders, 16, 363–370. [PubMed] [Google Scholar]
  22. Fladung A-K, Grön G, Grammer K, Herrnberger B, Schilly E, Grasteit S, … & von Wietersheim J (2010). A neural signature of anorexia nervosa in the ventral striatal reward system. American Journal of Psychiatry, 167, 206–212. [DOI] [PubMed] [Google Scholar]
  23. Fladung A-K, Schulze UME, Schöll F, Bauer K, & Grön G (2013). Role of the ventral striatum in developing anorexia nervosa. Translational Psychiatry, 3, e315. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Foerde K, Schebendach JE, Davis L, Daw N, Walsh BT, Shohamy D, & Steinglass JE (2020). Restrictive eating across a spectrum from healthy to unhealthy: Behavioral and neural mechanisms. Psychological Medicine, 52, 1755–1764. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Foerde K, Steinglass J, Shohamy D, & Walsh BT (2015). Neural mechanisms supporting maladaptive food choices in anorexia nervosa. Nature Neuroscience, 18, 1571–1573. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Foerde K, & Steinglass JE (2017). Decreased feedback learning in anorexia nervosa persists after weight restoration. The International Journal of Eating Disorders, 50, 415–423. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Foerde K, Walsh BT, Dalack M, Daw N, Shohamy D, & Steinglass JE (2021). Changes in brain and behavior during food-based decision-making following treatment of anorexia nervosa. Journal of Eating Disorders, 9, 48. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Fortin J-P, Cullen N, Sheline YI, Taylor WD, Aselcioglu I, Cook PA, … & Shinohara RT (2018). Harmonization of cortical thickness measurements across scanners and sites. NeuroImage, 167, 104–120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Frank GKW, DeGuzman MC, Shott ME, Laudenslager ML, Rossi B, & Pryor T (2018). Association of brain reward learning response with harm avoidance, weight gain, and hypothalamic effective connectivity in adolescent anorexia nervosa. JAMA Psychiatry, 75, 1071–1080. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Hadigan CM, Anderson EJ, Miller KK, Hubbard JL, Herzog DB, Klibanski A, & Grinspoon SK (2000). Assessment of macronutrient and micronutrient intake in women with anorexia nervosa. The International Journal of Eating Disorders, 28, 284–292. [DOI] [PubMed] [Google Scholar]
  31. Haynos AF, Hall LMJ, Lavender JM, Peterson CB, Crow SJ, Klimes-Dougan B, … & Camchong J (2019). Resting state functional connectivity of networks associated with reward and habit in anorexia nervosa. Human Brain Mapping, 40, 652–662. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Haynos AF, Lavender JM, Nelson J, Crow SJ, & Peterson CB (2020). Moving towards specificity: A systematic review of cue features associated with reward and punishment in anorexia nervosa. Clinical Psychology Review, 79, 101872. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Hinne M, Gronau QF, van den Bergh D, & Wagenmakers EJ (2020). A conceptual introduction to Bayesian model averaging. Advances in Methods and Practices in Psychological Science, 3, 200–215. [Google Scholar]
  34. Horndasch S, Roesch J, Forster C, Dörfler A, Lindsiepe S, Heinrich H, … & Kratz O (2018). Neural processing of food and emotional stimuli in adolescent and adult anorexia nervosa patients. PLoS One, 13, e0191059. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Huse DM, & Lucas AR (1984). Dietary patterns in anorexia nervosa. The American Journal of Clinical Nutrition, 40, 251–254. [DOI] [PubMed] [Google Scholar]
  36. Kerr KL, Moseman SE, Avery JA, Bodurka J, & Simmons WK (2017). Influence of visceral interoceptive experience on the brain’s response to food images in anorexia nervosa. Psychosomatic Medicine, 79, 777–784. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Keysers C, Gazzola V, & Wagenmakers E-J (2020). Using Bayes factor hypothesis testing in neuroscience to establish evidence of absence. Nature Neuroscience, 23, 788–799. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Kundu P, Inati SJ, Evans JW, Luh W-M, & Bandettini PA (2012). Differentiating BOLD and non-BOLD signals in fMRI time series using multi-echo EPI. NeuroImage, 60, 1759–1770. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Lloyd EC, & Steinglass JE (2018). What can food-image tasks teach us about anorexia nervosa? A systematic review. Journal of Eating Disorders, 6, 1–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Marshall WA, & Tanner JM (1969). Variations in pattern of pubertal changes in girls. Archives of disease in childhood, 44, 291–303. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Mayer LES, Schebendach J, Bodell LP, Shingleton RM, & Walsh BT (2012). Eating behavior in anorexia nervosa: Before and after treatment. International Journal of Eating Disorders, 45, 290–293. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. O’Hara CB, Campbell IC, & Schmidt U (2015). A reward-centred model of anorexia nervosa: A focussed narrative review of the neurological and psychophysiological literature. Neuroscience and Biobehavioral Reviews, 52, 131–152. [DOI] [PubMed] [Google Scholar]
  43. Op de Macks ZA, Bunge SA, Bell ON, Wilbrecht L, Kriegsfeld LJ, Kayser AS, & Dahl RE (2016). Risky decision-making in adolescent girls: The role of pubertal hormones and reward circuitry. Psychoneuroendocrinology, 74, 77–91. [DOI] [PubMed] [Google Scholar]
  44. Pinheiro JC (2014). Linear mixed effects models for longitudinal data. In Balakrishnan N, Colton T, Everitt B, Piegorsch W, Ruggeri F, & Teugels JL (Eds.), Wiley statsref: Statistics reference online (1st edn). Hoboken: Wiley. [Google Scholar]
  45. Ranzenhofer LM, Jablonski M, Davis L, Posner J, Walsh BT, & Steinglass JE (2022). Early course of symptom development in anorexia nervosa. The Journal of Adolescent Health: Official Publication of the Society for Adolescent Medicine, 71, 587–593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Ripke S, Hübner T, Mennigen E, Müller KU, Li S-C, & Smolka MN (2015). Common neural correlates of intertemporal choices and intelligence in adolescents. Journal of Cognitive Neuroscience, 27, 387–399. [DOI] [PubMed] [Google Scholar]
  47. Schebendach J, Mayer LE, Devlin MJ, Attia E, & Walsh BT (2012). Dietary energy density and diet variety as risk factors for relapse in anorexia nervosa: A replication. International Journal of Eating Disorders, 45, 79–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Schebendach JE, Mayer LE, Devlin MJ, Attia E, Contento IR, Wolf RL, & Walsh BT (2008). Dietary energy density and diet variety as predictors of outcome in anorexia nervosa. The American Journal of Clinical Nutrition, 87, 810–816. [DOI] [PubMed] [Google Scholar]
  49. Smith KS, & Graybiel AM (2016). Habit formation. Dialogues in Clinical Neuroscience, 18, 33–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Steinglass J, Foerde K, Kostro K, Shohamy D, & Walsh BT (2015). Restrictive food intake as a choice – A paradigm for study. International Journal of Eating Disorders, 48, 59–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Steinglass JE, Fei W, Foerde K, Touzeau C, Ruggiero J, Lloyd C, … & Walsh BT (2023). Change in food choice during acute treatment and the effect on longer-term outcome in patients with anorexia nervosa. Psychological Medicine, 54, 1133–1141. [DOI] [PubMed] [Google Scholar]
  52. Steinglass JE, & Walsh BT (2016). Neurobiological model of the persistence of anorexia nervosa. Journal of Eating Disorders, 4, 19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Stunkard AJ, & Messick S (1985). The three-factor eating questionnaire to measure dietary restraint, disinhibition and hunger. Journal of Psychosomatic Research, 29, 71–83. [DOI] [PubMed] [Google Scholar]
  54. Sysko R, Glasofer DR, Hildebrandt T, Klimek P, Mitchell JE, Berg KC, … & Walsh BT (2015). The eating disorder assessment for DSM-5 (EDA-5): Development and validation of a structured interview for feeding and eating disorders. International Journal of Eating Disorders, 48, 452–463. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Sysko R, Steinglass J, Schebendach J, Mayer LES, & Walsh BT (2018). Rigor and reproducibility via laboratory studies of eating behavior: A focused update and conceptual review. International Journal of Eating Disorders, 51, 608–616. [DOI] [PubMed] [Google Scholar]
  56. Sysko R, Walsh BT, Schebendach J, & Wilson GT (2005). Eating behavior among women with anorexia nervosa. The American Journal of Clinical Nutrition, 82, 296–301. [DOI] [PubMed] [Google Scholar]
  57. Tadayonnejad R, Majid D-A, Tsolaki E, Rane R, Wang H, Moody TD, … & Feusner JD (2022). Mesolimbic neurobehavioral mechanisms of reward motivation in anorexia nervosa: A multimodal imaging study. Frontiers in Psychiatry, 13, 806327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Treasure J, Stein D, & Maguire S (2015). Has the time come for a staging model to map the course of eating disorders from high risk to severe enduring illness? An examination of the evidence. Early Intervention in Psychiatry, 9, 173–184. [DOI] [PubMed] [Google Scholar]
  59. Udo T, & Grilo CM (2018). Prevalence and correlates of DSM-5 eating disorders in nationally representative sample of United States adults. Biological Psychiatry, 84, 345–354. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Uniacke B, Wang Y, Biezonski D, Sussman T, Lee S, Posner J, & Steinglass J (2018). Resting-state connectivity within and across neural circuits in anorexia nervosa. Brain and Behavior: A Cognitive Neuroscience Perspective, 9, e01205. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Van Leijenhorst L, Zanolie K, Van Meel CS, Westenberg PM, Rombouts SA, & Crone EA (2010). What motivates the adolescent? Brain regions mediating reward sensitivity across adolescence. Cerebral Cortex, 20, 61–69. [DOI] [PubMed] [Google Scholar]
  62. Walsh BT (2013). The enigmatic persistence of anorexia nervosa. American Journal of Psychiatry, 170, 477–484. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Walton E, Bernardoni F, Batury V-L, Bahnsen K, Larivière S, Abbate-Daga G, … & Ehrlich S (2022). Brain structure in acutely underweight and partially weight-restored individuals with anorexia nervosa – A coordinated analysis by the ENIGMA eating disorders working group. Biological Psychiatry, 92, 730–738. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Wechsler D (1999). Wechsler abbreviated scale of intelligence. APA PsycTests. 10.1037/t15170-000 [DOI] [Google Scholar]
  65. White H (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817–838. [Google Scholar]
  66. Wierenga CE, Reilly E, Bischoff-Grethe A, Kaye WH, & Brown GG (2022). Altered reinforcement learning from reward and punishment in anorexia nervosa: Evidence from computational modeling. Journal of the International Neuropsychological Society, 28, 1003–1015. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material

Appendix S1 Table S1. Analysis of healthiness ratings: Group (HC/AN) X Food type (Low- fat/High-fat).

Table S2. Analysis of tastiness ratings: Group (HC/AN) X Food type (Low- fat/High-fat).

Table S3. Analysis of choices: Group (HC/AN) X Food type (Low- fat/High-fat).

Table S4. Analysis of healthiness and tastiness rating influence on choice by Group (HC/AN).

Table S5. Analysis of Healthiness rating influence on Tastiness ratings by Group (HC/AN).

Table S6. Prior and posterior model probabilities for prediction of choice-related neural activation.

Figure S1. Prior and posterior distributions for effect of diagnostic term on choice-related activation in the nucleus accumbens and anterior caudate.

Table S7. Results of sensitivity analyses adjusting for Tanner stage.

Table S8. Results of sensitivity analyses excluding individuals with atypical AN.

Table S9. Results of sensitivity analyses excluding individuals taking SSRI medication at the time of the study.

Figure S2. Whole-brain analyses of Choice phase.

NIHMS2061248-supplement-Supplementary_Material.docx (650.3KB, docx)

Data Availability Statement

The data that support the findings of this study are available upon request and in the NIMH Data Archive at https://nda.nih.gov/edit_collection.html?id=2633, reference number 2633.

RESOURCES