Changes in Appetite-Regulating Hormones Following Food Intake are Associated with Changes in Reported Appetite and a Measure of Hedonic Eating in Girls and Young Women with Anorexia Nervosa

Abstract

Background:

Females with anorexia nervosa (AN) have higher ghrelin and peptide YY (PYY) and lower brain-derived neurotropic factor (BDNF) levels than controls, and differ in their perception of hunger cues. Studies have not examined appetite-regulating hormones in the context of homeostatic and hedonic appetite in AN.

Objective:

To examine whether alterations in appetite-regulating hormones following a standardized meal are associated with homeostatic and hedonic appetite in young females with AN vs. controls.

Methods:

68 females (36 AN, 32 controls) 10–22 years old were enrolled. Ghrelin, PYY and BDNF levels were assessed before, and 30, 60 and 120 minutes following a 400-kilocalorie standardized breakfast. Visual Analog Scales (VAS) assessing prospective food consumption, hunger, satiety, and hedonic appetite were administered before and 20 minutes after breakfast. A Cookie Taste Test (CTT) was conducted after a snack as a measure of hedonic eating behavior ~3 hours after breakfast.

Results:

AN had higher fasting ghrelin and PYY, and lower fasting BDNF (p=0.001, 0.002 and 0.044 respectively) than controls. Following breakfast (over 120 minutes), ghrelin and PYY area under the curve (AUC) were higher, while BDNF AUC was lower in AN vs. controls (p=0.007, 0.017 and 0.020 respectively). Among AN (but not controls), reductions in ghrelin and increases in PYY in the first 30-minutes following breakfast were associated with reductions in VAS scores for prospective food consumption. AN consumed fewer calories during the CTT vs. controls (p<0.0001). In AN (particularly AN-restrictive subtype), BDNF AUC was positively associated with kilocalories consumed during the CTT

Conclusions:

In young females with AN, changes in ghrelin and PYY following food intake are associated with reductions in a prospective measure of food consumption, while reductions in BDNF are associated with reduced hedonic food intake. Further studies are necessary to better understand the complex interplay between appetite signals and eating behaviors in AN.

Introduction

Anorexia nervosa (AN) is characterized by sustained restriction of energy intake, an intense fear of weight gain, and altered body image (DSM-5) ¹. Individuals with AN may report alterations in hunger cues, or that hunger cues are absent entirely, despite persistent caloric restriction ^2,3. Extensive research exists on the neuroendocrine dysfunction associated with AN, including changes in various hormones that regulate appetite and satiety, such as the orexigenic growth hormone secretagogue, ghrelin ^4–9, and the anorexigenic hormones, peptide YY (PYY) ^9,10 and brain derived neurotrophic factor (BDNF) ^11,12. High levels of ghrelin and low levels of BDNF might represent adaptation to chronic starvation to increase the appetite signal, while high levels of PYY may contribute to maintenance of food restriction. However, the relationship between levels of these hormones and appetite and/or eating behavior in AN has not been demonstrated.

Ghrelin stimulates appetite immediately following its secretion or administration in healthy individuals ¹³, and circulating ghrelin is highest immediately preceding a meal, followed by a decrease 30 to 60 minutes after food intake ^14,15. Ghrelin levels are higher in AN compared to normal-weight controls ^5–9,16, likely an adaptive response to low body weight in AN to increase hunger and, therefore, caloric intake. However, it is unclear whether these higher ghrelin levels are consciously perceived by individuals with AN as an increase in appetite. A study in adolescent AN demonstrated changes in certain appetite-regulating hormones and in reported appetite following a mixed meal, but did not report associations between hormone levels and reported appetite ¹⁶. Whereas administration of low dose ghrelin infusion have been demonstrated to increase caloric energy intake by 20% in a lean control sample ¹⁷, at least one study in adults reported a weak but positive association between ghrelin levels and reported hunger in the fasting state in controls, but not in women with AN ¹⁸, and another reported no significant impact of a ghrelin infusion on appetite levels in women with AN ¹⁹. Thus, AN may be a ghrelin-resistant state in which the adaptive increase in ghrelin from low body weight fails to increase hunger. Alternately, this population may indeed experience the hunger cues generated by ghrelin, but effectively suppress this signal and continue in restrictive eating behaviors. Interestingly, one study of five women with the restrictive form of AN reported that twice daily ghrelin infusion increased hunger sensation and daily energy intake in these women ²⁰.

Conversely, PYY, derived from the L-cells of the intestine, increases in the periphery following food intake and induces satiety ^21,22. Studies have demonstrated higher levels of this anorexigenic hormone in AN compared to normal-weight controls ^9,10,16. This increase in PYY does not appear to be an adaptive response to the characteristic low-weight state in AN. Prior research has shown that PYY3–36 (compared with a saline solution) reduces caloric intake in lean individuals as well as those with obesity ^23,24. However, it remains unclear whether higher PYY levels in AN lead to reduced hunger.

Another hormone implicated in altered eating behaviors is BNDF, believed to be primarily anorexigenic ^11,12 and reported to be lower in those with AN compared with controls ^11,25,26. Of note, BDNF is involved in a myriad of processes ranging from neuroplasticity to reward processing ^27–29. While the precise relationship between BDNF and eating behaviors is not fully understood, several publications have established alterations in circulating BDNF within samples of individuals diagnosed with AN ^30–32. One study reported higher circulating BDNF in women with the binge-eating/purging subtype of AN (AN-BP) compared to those with the restricting subtype (AN-R), while levels in AN as a whole did not differ from controls ³⁰. These researchers posited a potential link between BDNF and altered eating behavior in AN. Others have suggested that BDNF levels reflect weight status in AN, increasing with increasing weight ^31,32. Furthermore, a rodent model examining BDNF in relation to eating patterns demonstrated that BDNF depletion may increase consumption of regular chow or highly palatable food, depending on the specific brain area where the depletion occurs ³³. Thus, changes in BDNF may impact both homeostatic and hedonic appetite in AN. Given the contradictory nature of prior research, this rodent model is of interest in that it serves as a platform to examine the potential relationship between BDNF and hedonic eating patterns in individuals with anorexia nervosa; a relationship that remains relatively unexamined.

Our goal was to examine how these hormones change following food intake in an adolescent and young adult sample of females with AN vs. healthy controls and to determine whether alterations in these hormones following food intake are associated with homeostatic and hedonic appetite. We hypothesized that changes in ghrelin, PYY, and BDNF following a small standardized meal would be associated with measures of homeostatic and hedonic appetite in AN.

Participants and Methods

Study Participants

We enrolled 68 adolescent and young adult females 10–22 years old, 36 with AN and 32 normal-weight controls, who were part of a larger study assessing the neurobiology of low-weight eating disorders. Participants with AN met diagnostic criteria based on the Schedule for Affective Disorders and Schizophrenia for School Age Children-Present and Lifetime Version (K-SADS PL) and confirmed by the Eating Disorder Examination (EDE), and were low weight as indicated by one of the following criteria: < 90% median body mass index (BMI) or < 90% of expected body weight for age or height based on Centers for Disease Control female growth charts ³⁴. Females with AN were further classified as AN-R or AN-BP based on symptomatology (present or absent recurrent binge and/or purge behaviors ¹). If a participant did not engage in any binge or purge behaviors in the preceding three months, or if binge-purge frequency was less than once a month during this period, she was included in the AN-R group. If a participant engaged in either binge or purge behaviors at least one time per month within preceding three months, she was included in the AN-BP group. Control participants were required to be between the 25^th −85^th percentiles of BMI for age, have regular menses if at least two years post-menarchal, and have no history of pubertal delay (i.e., menarche at >16 years or thelarche at >13 years). Controls were excluded if they had a history of any psychiatric disorder as assessed by K-SADS-PL. Controls were also excluded if they exercised >10 hours per week or were running >25 miles per week in the three months preceding study enrollment, due to the potential impact this activity may have on appetite-regulating hormones (e.g., cause an increase in ghrelin and PYY) ^35,36.

Participants with AN were recruited from the community through referrals from primary care providers, adolescent medicine physicians, endocrinologists, dieticians, and psychologists, as well as from eating disorder centers in Massachusetts and neighboring states. None of the participants were inpatient at the time of the study, although some were in residential programs. Thus, participants were at different stages of treatment, as intended, to minimize homogeneity within the study sample as this would otherwise have limited our ability to demonstrate associations between covariates and outcomes of interest, and also limited the generalizability of study results. Controls were recruited through flyers in primary care offices, and advertisements within the institution and on college websites.

Experimental Protocol

The study was approved by our institutional review board (IRB). Written informed consent was obtained from participants at least 18 years old, and from parents of participants <18 years old; assent was obtained from participants < 18 years old. Participants were seen at the Massachusetts General Hospital Clinical Research Center and the Athinoula A. Martinos Imaging Center. At the screening visit, a detailed history was obtained, a physical examination performed and the K-SADS-PL administered. Height was measured on a wall-mounted stadiometer in triplicate, and weight was measured on an electronic scale. BMI, percent median BMI, and percent expected body weight for age and height were calculated. A blood sample was drawn to confirm eligibility for study participation (to rule out anemia and electrolyte disturbances). A urine pregnancy test was performed at the screening visit as well as the main study visit.

For the main study visit, participants were asked to fast overnight (beginning at 11:30 PM) for at least 8 hours prior to the fasting blood draw. The Eating Disorder Examination (EDE) was performed to assess current eating disorder pathology (dietary restraint, eating concern, shape concern, weight concern), and the final DSM-5-defined diagnosis was determined at this interview. After an interim history and physical examination, participants were given a solid ~400-kcal mixed breakfast standardized for nutrient content (approximately 20% calories from protein, 20% from fat, and 60% from carbohydrates) at approximately 9:00 AM and asked to eat the entire meal over 15 minutes. The standardized breakfast meal options consisted of either a bagel with peanut butter and craisins, vanilla yogurt and cereal, or fruit yogurt and toast. Upon completion, bionutrition staff weighed the meal to determine the exact caloric intake. Blood was drawn serially for hormones immediately pre-meal (fasting sample, T0), and 30, 60, and 120 minutes after the start of the breakfast-meal (T30, T60 and T120 respectively). Visual analog scales were administered to assess feelings of homeostatic and hedonic appetite 5 minutes before the meal and after meal completion (please see details under ‘Visual Analog Scales‘). After 180 minutes from the start of the breakfast meal and 30 minutes following initiation of a standard snack (approximately 210 calories, 61% carbohydrates, 26% fat and 13% protein; options were mixed meals similar to the breakfast meal) to minimize homeostatic appetite, the Cookie Taste Test was performed to measure hedonic eating behavior (please see details under ‘Cookie Taste Test‘).

Visual Analog Scales

The Visual Analog Scale (VAS) is a widely-utilized measure for assessing levels of hunger and satiety. Within our paradigm, VAS measures were administered, approximately 5 minutes before the start of a calorically standardized breakfast meal, and after meal completion (approximately 20–25 minutes from meal initiation and just before the 30-minute blood draw). Participants were asked to respond how they were feeling on dimensions of hunger and appetite at the time that they completed the questionnaire. Questions included: (i) “How much do you think you can eat?”, (ii) “How hungry do you feel?”, (iii) “How satisfied do you feel?”, and (iv) “How strong is your desire to eat your favorite food?” The first three questions are measures of prospective food consumption, hunger and satiety respectively. Researchers used the ‘satisfaction’ item on the VAS as a proxy for satiety; specifically, as the mental construct of fullness, whereas the “feeling full” item was administered as it relates to the physical sensation of fullness. The last question was used as a measure of desire or hedonic appetite ^16,37. Participants responded on an electronic scale of “0” to “100”, with 0 corresponding to a response of “not at all” and 100 referring to “extremely strong”.

Cookie Taste Test

The Cookie Taste Test was modified from a similar behavioral task ³⁸ and is designed to be a measure of hedonic eating in that it presents highly palatable foods and is administered after a meal and snack (i.e., when homeostatic appetite has been addressed). It was performed approximately 180 minutes from the start of the breakfast-meal and 30 minutes following initiation of a standard snack (to ensure that we captured hedonic and not homeostatic eating). Participants were given the instruction to taste as much of the three different cookies as they needed to in order to rate the cookies on various dimensions, such as sweetness, texture, and likability. Following the task, the amount of cookies each participant consumed was weighed and recorded, as a behavioral measure of sustained responsiveness to reward attainment, and an indicator of hedonic appetite. Further, the literature suggests that up to half of patients with AN will develop binge eating over time ^39,40. By presenting palatable foods that patients with AN often avoid altogether, the Cookie Taste Test may also be sensitive to capturing vulnerability to binge eating within the AN group.

Biochemical Analysis

Our samples were immediately placed on ice following venipuncture, spun in a refrigerated centrifuge, and serum samples stored at −80 degrees Celsius until measurement. Plasma total ghrelin levels were determined by an enzyme-linked immunosorbent assay (ELISA) (EMD Millipore: Billerica, MA), with an intra-assay CV of 1.32% and inter-assay CV of 6.62%. The lower limit of detection was 50.0 pg/mL. Serum PYY levels were measured by ELISA (Millipore Corp; intra-assay CV 17–18% and inter-assay CV 12–18%; lower limit of detection 10.0 ng/mL). Plasma BDNF levels were determined by an ELISA (R&D Systems, Inc; intra-assay CV 5% and inter-assay CV 9%; lower limit of detection 20 pg/mL).

Statistical Analysis

JMP Statistical Discoveries (version 13.0; SAS Institute, Inc., Cary, NC) was used for statistical analyses. Clinical characteristics and hormone levels were compared using the Student t-test or Wilcoxon Rank Sum test depending on data distribution (parametric and non-parametric, respectively). BDNF levels were natural log converted to approximate a normal distribution. Based on published data, the study was powered at more than 80% to detect a difference of at least 0.7*standard deviation between the groups in baseline ghrelin, PYY and BDNF levels at an alpha level of 0.05 ^10,25,41. Because ghrelin, PYY and BDNF have been reported to vary with age in some studies ^42–44, we controlled for age when comparing mean hormone levels in AN vs. control groups. Associations between (i) percent changes in hormone levels over the first 30 minutes following food intake, and percent changes in Visual Analog Scales of appetite, and (ii) between hormone area under the curve (AUC) and calories consumed during the Cookie Taste Test were assessed using Pearson correlations (r) for normally distributed data and Spearman correlations (ρ) for not-normally distributed data.

We chose to assess associations of VAS measures with hormonal changes over 30 minutes following a standardized breakfast because decreases in ghrelin and increases in PYY levels are clearly evident in the first 30 minutes following food intake ^45–48. However, we assessed associations of hormone AUC (rather than changes over the first 30 minutes) with kilocalories consumed at the Cookie Taste Test because this test occurred about 3 hours after the standardized meal and 30 minutes after a standard snack, and any effect on hedonic eating this far out from breakfast would be expected to be impacted by overall hormone levels (AUC), rather than changes over the first 30 minutes. For BDNF assessments, three participants were eliminated from data analyses because they were marked outliers (confirmed with statistical methods; at least 3 S.D.s from the mean). Statistical significance was defined as a two-tailed p-value of < .05. Trends (p<.10) are also reported as preliminary findings. Data are presented as means ± SEM.

Results

Clinical Characteristics:

Sixty-eight participants were included in the statistical analyses. Clinical characteristics are presented in Table 1. Groups did not differ for race or ethnicity. Per study design, those with AN had significantly lower BMI than controls. Our AN group was slightly older than the control group, and we therefore controlled for age in subsequent analyses.

Table 1:

Clinical Characteristics of Participants with Anorexia Nervosa and Controls

	Anorexia Nervosa N=36	Controls N=32	Effect size (Cohen’s d)	P value	P value adjusted for age
Age (years)	19.6 ± 0.4	17.5 ± 0.5	0.729	0.002
Age <18 years (n)	7/36	12/32		0.010
Bone age (years)	17.4 ± 0.2	16.2 ± 0.4	0.601	0.022
Tanner stage I to IV (n)	7/36	11/32		0.163
Weight (kg)	45.9 ± 0.8	57.2 ± 1.9	−1.139	< 0.0001
Height (cm)	164.2 ± 1.2	162.9 ± 1.6	0.163	0.524
Body mass index (BMI)	17.0 ± 0.2	21.3 ± 0.4	−1.547	< 0.0001
Percent median BMI (%expected body weight based on the 50th percentile of BMI for age) 52	79.8 ± 1.0	103.2 ± 1.4	−1.706	< 0.0001
BMI z-score	−2.07 ± 0.14	0.14± 0.09	−1.713	< 0.0001
Menarchal age (years)	13.4 ± 0.2	12.8 ± 0.2	0.473	0.055
Restricting vs. binge eating/purging subtypes (%)	Restricting 69% (n=25) Binge- 30% (n=11)	n/a	n/a	n/a	n/a
Kilocalories consumed at breakfast	363.3 ± 15.5	379.3 ± 9.7	−0.207	0.400	0.051
Fasting ghrelin (pg/mL)	891.0 ± 39.5	683.4 ± 47.8	0.763	0.001	0.003
Ghrelin AUC (pg/mL*120)	85,382 ± 4,664	66,730 ± 4,715	0.687	0.007	0.014
Fasting PYY (ng/mL)	107.9 ± 6.6	79.9 ± 5.6	0.730	0.002	0.011
PYY AUC (ng/mL*120)	15,046 ± 1,094	11,751 ± 4,003	0.625	0.017	0.082
Fasting BDNF (pg/mL)	2,731.3 ± 415.3	3,792.8 ± 457.5	−0.420	0.031**	0.024**
BDNF AUC (pg/mL*120)	250,493 ± 32,288	331,768 ± 33,964	−0.447	0.036**	0.04**
Kilocalories consumed at Cookie Taste Test	70.6 ± 10.8	179.6 ± 19.6	−1.056	< 0.0001	<0.0001

^**Log converted values compared

n/a: not applicable

Pre- and Post-Breakfast Ghrelin, PYY, and BDNF Levels:

Figures 1–3 show mean ghrelin, PYY and BDNF levels at baseline (T0), and 30, 60 and 120 minutes following baseline (T30, T60 and T120), as well as the change and percent change in these hormones from baseline to 30 (T30-T0), 60 (T60-T0) and 120 (T120-T0) minutes after controlling for age. Mean ghrelin levels were higher in the AN vs. control groups at all timepoints (p<0.05). Percent change in ghrelin from baseline did not differ between groups at any timepoint. Mean PYY levels were higher in AN vs. control groups at baseline and 30 minutes and trended higher at 60 and 120 minutes. Percent change in PYY from baseline trended lower in AN vs. controls groups at 120 minutes. Mean BDNF levels were lower in AN vs. control groups at baseline and 60 minutes. Percent change in BDNF from baseline was higher in the AN vs. controls at 120 minutes.

Figure 1:

(A) Mean ghrelin levels before and following a standardized breakfast, and (B) Percent change in ghrelin following this standardized breakfast; * p<0.05. All analyses are controlled for age. Error bars represent standard error. AN = anorexia nervosa; HC = healthy controls.

Figure 3:

(A) Mean Ln BDNF levels before and following a standardized breakfast, and (B) Percent change in Ln BDNF following this standardized breakfast; * p<0.05; † p<0.1. All analyses are controlled for age. Error bars represent standard error. BDNF = brain derived neurotropic factor; AN = anorexia nervosa; HC = healthy controls.

Pre- and Post-Breakfast Scores on the Visual Analog Scale:

Compared to controls, our AN group reported a lower desire to eat their favorite foods and reported that they could eat less (on the question “How much do you think you can eat?”) prior to the breakfast meal (Table 2). However, post-breakfast, the scores did not differ across groups, and changes in reported appetite (pre- to post-breakfast) also did not differ across AN vs. control groups. Further, AN-R did not differ from AN-BP for these scores (Supplemental Table 1).

Table 2:

Pre- and Post-Breakfast Scores on the Visual Analog Scales

	Anorexia Nervosa N=36	Controls N=32	Effect size (Cohen’s d)	P value	P value adjusted for age
How much do you think you can eat?
Before breakfast	52.9 ± 2.9	62.7 ± 3.1	−0.551	0.023	0.003
After breakfast	27.8 ± 4.0	32.7 ± 3.5	−0.222	0.370	0.183
Percent change from before to after breakfast	−43.1 ± 8.6	−46.9 ± 5.4	0.090	0.720	0.741
How hungry do you feel?
Before breakfast	52.2 ± 4.1	60.7 ± 3.4	−0.383	0.118	0.085
After breakfast	14.4 ± 3.0	18.2 ± 2.6	−0.227	0.350	0.166
Percent change from before to after breakfast	−71.6 ± 6.4	−58.5 ± 9.8	−0.278	0.262	0.116
How satisfied do you feel?
Before breakfast	35.0 ± 3.5	35.9 ± 2.9	−0.049	0.843	0.730
After breakfast	76.0 ± 3.0	74.2 ± 2.1	0.118	0.634	0.273
Percent change from before to after breakfast	467.3± 268.9	193.1± 48.8	0.245	0.327	0.490
How strong is your desire to eat your favorite food?
Before breakfast	40.5 ± 4.3	53.1 ± 3.6	−0.529	0.030	0.035
After breakfast	16.9 ± 3.8	22.1 ± 3.6	−0.239	0.328	0.273
Percent change from before to after breakfast	−64.2 ± 7.0	−55.2 ± 7.0	−0.228	0.366	0.262

Associations of Hormonal Changes with Changes in Visual Analog Scale Scores Pre- and Post-Breakfast

Pre-Breakfast:

Hormone levels at baseline were not significantly associated with VAS scores at baseline in AN or control groups (Supplemental Table 2A).

Post-Breakfast:

Within our AN group (but not in controls), we observed a positive association between percent change in ghrelin from pre-breakfast to 30 minutes post-breakfast (T30-T0) with percent change in the VAS measure “How much do you think you can eat” (r=0.435, p=0.013) (Figure 4A) i.e. the greater the percent reduction in ghrelin following breakfast, the greater the percent reduction in scores on “How much do you think you can eat”. This association was not observed in controls (r=−0.246, p=0.182).

Figure 4:

Associations of percent change in ghrelin (T30-T0) (A) and PYY (T30-T0) (B) with percent change in ‘How much do you think you can eat’ in the young females with anorexia nervosa (AN); and of (C) BDNF AUC with calories consumed during the cookie taste test in AN-restrictive subtype (AN-R). Error bars represent standard error. PYY = peptide YY; BDNF AUC = brain derived neurotropic factor area under the curve; HC = healthy controls.

Similarly, within our AN group, but not in controls, we found a negative association between percent change in PYY levels from pre-breakfast to 30-minutes post breakfast (T30-T0) and percent change in the VAS measure “How much do you think you can eat” (Spearman ρ= −0.389, p=0.031) (Figure 4B), i.e. the greater the percent increase in PYY following breakfast, the greater the reduction in scores on “How much do you think you can eat”. This association was not observed in controls (Spearman ρ= 0.269, p=0.150).

We found no association of percent changes in ghrelin, PYY or BDNF levels with percent changes in the other three VAS measures pre- and post-breakfast (Supplemental Table 2B).

Cookie Taste Test:

In comparison to our control group, AN participants consumed fewer kilocalories during the Cookie Taste Test (Table 1). Within our AN group, we observed a trend for a positive association between BDNF AUC and kilocalories consumed during the Cookie Taste Test (r= 0.398, p=0.054). No association was observed between BDNF AUC and kilocalories consumed during the Cookie Taste Test in our control group (r= −0.108 p = 0.576).

Based on a study that demonstrated higher circulating BDNF in women with AN-BP vs. AN-Rs ³⁰, we examined associations of BDNF AUC with calories consumed in the Cookie Taste test separately in AN-BP vs. AN-R groups. We observed an association between BDNF AUC and calories consumed during the Cookie Taste Test in the AN-R but not AN-BP group (AN-R: Spearman ρ= 0.687, p=0.0095: AN-BP: Spearman ρ= 0.191, p=0.574) (Figure 4C).

We found no significant relationships between ghrelin or PYY AUC and kilocalories consumed during the Cookie Taste Test in the AN or control groups (PYY AUC: AN: r=0.030, p=0.888; controls: r=−0.056, p=0.782; ghrelin AUC: AN: r=−0.151, p=0.462; controls: r=−0.342, p=0.069).

Discussion

These data provide novel insight into experienced homeostatic and hedonic appetite in young females with anorexia nervosa in relation to the aberrant hormonal function that is characteristic of this condition. Consistent with prior studies, our data demonstrate that individuals with AN have elevated levels of circulating ghrelin ^5–9 and PYY ^9,10 and lower levels of BDNF ¹¹ compared with controls during the fasting state. We also demonstrate persistently higher ghrelin and PYY levels than controls following a standardized meal. For the first time, we demonstrate the impact of a standardized meal on BDNF levels in AN compared with controls, and associations of reported appetite with hormonal changes following a standardized meal in adolescents with AN.

We demonstrate that a decrease in orexigenic ghrelin and increase in anorexigenic PYY after a meal is associated with prospective food consumption (how much one thinks she can eat), consistent with the known appetite-regulating functions of these hormones. Interestingly, these associations were not observed in our controls. The interpretation of these data may provide potential insights into how ghrelin and PYY modulate perceived appetite in individuals with AN. For example, it is possible that post-prandial changes in ghrelin and PYY result in the expected impact on appetite reflected in how they score on prospective food consumption (on the VAS). However, individuals with AN are likely very effective at suppressing the feelings of hunger (e.g., by increased cognitive control), and this may explain why similar associations were not observed for VAS measures of hunger (how hungry do you feel) and satiety (how satisfied do you feel) or hedonic eating behavior assessed by the Cookie Taste Test. Of note, following ghrelin administration as an infusion, Hotta et al. reported an increase in hunger scores on the VAS in a small group of adult AN-R ²⁰. However, a study by Miljic et al. reported no change in hunger scores following a ghrelin infusion in adult women with AN ¹⁹. At this time, there are no studies of PYY administration in AN. The reason for a lack of association of ghrelin or PYY levels with measures of subjective appetite in our controls is unclear, but may reflect greater homogeneity in our control group in hormone levels, which would limit our ability to find significant associations with appetite measures in these individuals.

Total ghrelin measures both acylated/active and des-acylated/“unactive” forms. While the role of acylated ghrelin as an orexigenic hormone is well-established, less is known about desacyl ghrelin, which until recently was thought to be inactive. Some data now suggest that desacyl ghrelin may have opposing effects to acylated ghrelin on food intake. All three forms of ghrelin have been shown to be high in anorexia nervosa. In this study of females with anorexia nervosa, we have demonstrated a relationship between post-prandial decrease in total ghrelin and subjective amount one could eat; the directionality is suggestive of an orexigenic rather than anorexigenic effect. Further studies measuring acylated and desacylated forms of ghrelin and the relationship of these to subjective appetite in anorexia nervosa will be important ⁴⁹.

The positive association between BDNF AUC and kilocalories consumed during the Cookie Taste Test within AN-R suggests that exposure to BDNF may alter hedonic eating patterns. There is a growing body of literature on BDNF in relation to eating behaviors, however, there has been minimal exploration into any role it might play in hedonic patterns of eating ^{11,12,27–29,31,32}. One study in rodents has reported that BDNF depletion in the ventromedial nucleus of the hypothalamus increases standard chow intake, while its depletion in the ventral tegmental area increases consumption of highly palatable food, but not standard chow ³³ suggesting that lower BDNF expression should result in greater homeostatic and hedonic appetite. However, whether circulating BDNF levels impact appetite remains unclear. Overall, the AN group had lower BDNF levels than controls (consistent with available literature ^11,31,32), and demonstrated a greater increase in BDNF at 120 minutes vs. baseline values in AN vs. controls. The implication of the latter is uncertain, but may account for greater suppression of homeostatic and hedonic appetite in AN. These preliminary findings need to be confirmed in future studies.

However, post-prandial changes in BDNF were not associated with VAS measures in the AN group as a whole, though a trend for a positive association was observed for BDNF AUC and kilocalories consumed during the Cookie Taste Test in the AN group. In contrast to a previous study in adults with AN, in whom we reported higher BDNF levels in AN-BP vs. AN-R ³⁰, we found no differences in BDNF levels in AN-BP vs. AN-R in this study. BDNF AUC correlated with kilocalories consumed during the Cookie Taste Test in AN-R, but not AN-BP, which may reflect the small number of AN-BP participants in our AN sample, or differences in the underlying physiology. It is interesting to note that despite the absence of an association between BDNF levels and subjective hedonic appetite (using the VAS), BDNF was associated with hedonic eating (during the Cookie Taste Test), suggesting a disconnect between appetite and eating behavior in the AN-R group. Future studies with larger numbers of participants in these subgroups are necessary to better understand the impact of BDNF on homeostatic and hedonic appetite and eating behavior in this condition.

Limitations of this study include its cross-sectional nature as correlations do not indicate causation. Future studies that involve administration of these hormones in physiologic doses and subsequent assessment of appetite and eating behavior are necessary to demonstrate causative effects. Our groups differed for age (though the proportion of fully pubertal participants did not differ across groups), thus we controlled for age when comparing hormone levels across groups. It should be noted that within this sample, participants with AN were in varying stages of treatment; ranging from no outpatient treatment to a residential setting. However, this degree of heterogeneity is necessary for demonstrating associations, and we did not have specific hypotheses about how treatment status may impact homeostatic or hedonic eating. Another study limitation is that the reported findings were in the context of a standardized meal and may not accurately depict the response to foods that participants typically choose to eat.

Of note, we measured total ghrelin, and not the ‘active’ (acyl-) component of ghrelin in this study. Active (acyl-) ghrelin is unstable in plasma and is degraded to “unactive” (desacyl-) ghrelin, which is stable ⁵⁰. The total ghrelin assay measures both, and therefore degradation would not be expected to have a significant impact. Ways to reduce breakdown include freezing, acidification, or use of protease inhibitors. Our samples were immediately placed on ice following venipuncture, spun in a refrigerated centrifuge, and serum samples stored at −80 degrees Celsius until measurement. The total ghrelin levels are within the expected range and the differences in ghrelin levels between participants with AN and healthy controls as well as the ghrelin pattern in response to food intake are consistent with prior reports, suggesting that the effects of degradation on measurement of total ghrelin levels were not significant. In addition, all samples were handled in a consistent manner, so if some degradation occurred this would have affected all samples and would not be expected to impact the relationship between changes in total ghrelin levels and appetite, which was the focus here. Of note, published data suggest that PYY levels change across the menstrual cycle, with levels being lower during the luteal phase ⁵¹. We did not control for menstrual cycle phase in our analysis because of difficulties assessing cycle phase in participants with AN given that many were amenorrheic or had irregular cycles, and also in controls within two years of menarche (given that many of these cycles are known to be anovulatory and not associated with a luteal phase). Previous studies have demonstrated differences between adolescents with AN and controls for PYY levels, even when the analysis did not control for cycle phase ¹⁰.

Another point to note is that we did not have data for hormones between 120 minutes following the standardized breakfast and the Cookie Taste Test, and these data may have further added to our results. Given that we use AUC and not data from a single timepoint following the standardized breakfast to determine associations with caloric intake during the Cookie Taste Test, we likely do not lose much from missing an additional timepoint. However, it would have been ideal to have data at 180 minutes as well. Further, the accuracy of VAS reporting may differ in AN vs. controls. Although one would expect individuals with AN to lack insight into their appetite, or (deliberately or not) underreport this, which in turn would minimize the variance in VAS measures rendering detection of relationships impossible, the observed findings suggest that VAS could be a valuable tool in this population. Of note, the discrepancy in relationships between BDNF and hedonic appetite vs. eating behavior emphasizes the value of additional measures. Finally, there may be inherent biologic differences between restricting subtype AN and the binge/purge subtype that could not be accounted for, and need to be addressed in a larger, future study.

To our knowledge, this is the first study to examine reported hunger and satiety in relation to hormonal changes following a standardized meal in adolescents with AN. These findings indicate the importance of understanding the intersection of the biological mechanisms that underlie disordered eating and perceptions of appetite. Hormonal studies in larger samples of individuals with AN and using functional magnetic resonance imaging (to determine activation of key brain areas involved in homeostatic vs. hedonic appetite) are necessary to comprehensively address these key questions. Results of these studies may inform future therapies for young females with AN.

Figure 2:

(A) Mean PYY levels before and following a standardized breakfast, and (B) Percent change in PYY following this standardized breakfast; * p<0.05; † p≤0.1. All analyses are controlled for age. Error bars represent standard error. PYY = peptide YY; AN = anorexia nervosa; HC = healthy controls.