Anorexia nervosa (AN) is a life-threatening eating disorder typically beginning in adolescence and characterized by an intense fear of weight gain and a relentless pursuit of weight loss, mostly by self-starvation. Only less than half of AN patients fully recover from the disorder, one-third shows improvement with residual symptoms, and one fifth remains long-term chronically ill [1]. The underlying mechanisms of AN are still largely unclear and a better understanding of factors contributing to the development and progression of this enigmatic disorder would be the foundation for improving therapeutic interventions. Recently, the gut–brain axis, which describes the bidirectional communication between gut and brain via neurons, immune mediators, gut hormones, and influence of the gut microbiota [2], has been discussed as a possible key factor in the pathophysiology of AN.

Peptide YY (PYY) belongs to the neuropeptide Y family of biologically active peptides that are considered important mediators of the gut–brain axis [2]. It is a 36-amino acid peptide that is thought to be a satiety factor inhibiting food intake, gastrointestinal motility, and secretion, and may interact with the gut microbiota [2,3,4,5,6]. More recently, PYY has also been implicated as a possible regulator of glucose homeostasis [7]. PYY is expressed by endocrine cells of the digestive system, mostly by the endocrine L cells of the gastrointestinal tract in co-secretion with the incretin hormone glucagon-like-peptide 1 but also by endocrine cells in the pancreas and enteric neurons of the stomach [2, 7]. Its release from L cells into the blood stream is triggered by meal intake in proportion to energy intake and depending on meal composition, as well as stimulated by gastric acid secretion and a number of metabolites (for example, cholecystokinin or short chain fatty acids produced by the gut microbiota) [2, 8]. After its release, part of the parent peptide PYY1–36 is cleaved enzymatically by dipeptidyl peptidase 4 (DPP-4) to the main circulating form PYY3–36 [2, 9]. This cleavage is accompanied by a shift in the stimulated Y receptor subtypes leading to possibly divergent actions of PYY1–36 and PYY3–36 regarding energy and glucose homeostasis [2, 7, 10, 11]. Out of the two human forms of PYY, PYY3–36 is considered to be the one relevant to energy homeostasis [7]. The effects of PYY3–36 are mostly mediated through agonism at Y2 receptors on neurons both in the gastrointestinal tract and in the central nervous system [2]. Translational research implicates the arcuate nucleus of the hypothalamus and certain brain stem regions as key areas of central appetite-regulating circuits influenced by PYY3–36 [12]. Batterham et al. [13] found that peripherally infused PYY3–36 leads to a significant reduction in appetite and food intake in healthy, non-obese volunteers. Evidence from functional magnetic resonance imaging (fMRI) studies points to an influence of PYY concentration on neural activity in brain regions involved in reward processing (such as the orbito-frontal cortex), implicating that PYY may modulate the reward value of food [14, 15].

Relatively few studies have investigated PYY concentration in AN yielding heterogeneous results, as highlighted in Table 1, and only one study has investigated PYY in peripheral blood in long-term recovered AN patients. While a number of studies found a normal fasting PYY concentration in acutely underweight AN [16,17,18,19,20], some studies reported an elevated PYY concentration [21,22,23,24,25]. From a physiological perspective, one would expect the anorexigenic PYY to have low levels in this underweight and undernourished population. An elevated PYY concentration would be a maladaptive response as it may facilitate restrictive eating in AN. However, when interpreting previous studies, some confounding factors should be taken into consideration. For instance, physical exercise was shown to significantly increase the peripheral PYY concentration. Suggestive of short-term effects, Larson-Meyer et al. [26] reported elevated PYY concentrations following a 60-min run which gradually returned to baseline over the span of 120 min. Short aerobic exercise (3 × 10 min of rope skipping or cycling) also led to a brief PYY increase [27]. Similarly, Broom et al. [28] found an increase in PYY concentration after aerobic exercise (60-min run) but no difference in PYY levels among the exercise and non-exercise groups 7 h post-exercise. The majority of acutely underweight AN patients engage in excessive physical activity (reported prevalences range from 31 to 80% [29]) and it seems possible that exercise effects bias the assessment of PYY concentrations in AN. Furthermore, only a few studies assessed the biologically more relevant isoform PYY3–36 instead of total PYY [16, 21, 24, 30].

Table 1 PYY research on anorexia nervosa in humans

Gaining a broader understanding of the role of gut hormones such as PYY in the etiology and maintenance of AN holds the potential of identifying prognostic biomarkers and opening up new avenues for treatment. To date, no study on PYY in AN has taken physical activity into account, even though it is likely to be of great importance as a confounding factor in this patient sample. Furthermore, there is a considerably research gap regarding the persistence of possible changes after short-term and long-term weight recovery. Thus, the objective of the current study was to investigate PYY3–36 concentrations of patients with acutely underweight AN, before and after partial weight recovery (longitudinal sample), as well as in long-term recovered patients in comparison to healthy control participants (cross-sectional sample). The inclusion of long-term recovered patients allows the differentiation of state markers (i.e., related to acute undernutrition) from factors that might confer vulnerability towards AN. To control for possible exercise-related bias on PYY in AN, blood samples were obtained shortly after patients were admitted to an intensive treatment program which helps patients resist the drive to exercise.

Materials and methods


A total of 140 female subjects participated in this study: 47 underweight patients with acute AN (acAN, 12–23 years old), 35 patients long-term recovered from AN (recAN, 15–28 years old), and 58 healthy control participants (HC, 12–26 years old). For the longitudinal arm of our study, 32 acAN patients were reassessed after short-term weight rehabilitation. Participants with acAN were admitted to intensive treatment of an eating disorder program at a child and adolescent psychiatry and psychosomatic medicine department of a tertiary care university hospital. Newly admitted patients spent their first days under close supervision by specialized nursing staff and continuous heart rate monitoring at night. Therefore, we can ensure an exercise-free time period of at least 18 h before blood sample collection. Diagnosis of AN was established using the expert form of the Structured Interview for Anorexia and Bulimia Nervosa (SIAB-EX) [31] and required a body mass index (BMI) below the 10th age percentile (if younger than 15.5 years) or below 17.5 kg/m2 (if older than 15.5 years). RecAN participants were required to have had a diagnosis of AN in the past and, for at least 6 months before the study, (1) to have maintained a BMI above the 10th age percentile (if younger than 18 years) or above 18.5 kg/m2 (if older than 18 years), (2) to menstruate, and (3) to not have binged, purged, or engaged in substantial restrictive eating patterns. HC had to be of normal weight, eumenorrhoeic, and without any history of psychiatric illness. All HC were assessed with the SIAB-EX [31] and excluded if they showed any abnormal eating behavior. HC were recruited through advertisement among middle school, high school, and university students. While recAN and HC participants were only assessed once, acAN participants were assessed within 96 h of admission to intensive treatment (timepoint 1, acAN-T1) and reassessed after short-term weight rehabilitation (timepoint 2, acAN-T2; after BMI increase of at least 13%) for the longitudinal arm of this study.

Information regarding exclusion criteria and possible confounding variables, including menstrual cycle and use of contraceptive medication, was obtained from all participants (acAN, recAN, and HC) using the SIAB-EX [31], supplemented by our own semi-structured interview and medical records. Comorbid diagnoses were taken according to standard practice from medical records and confirmed by an expert clinician with over 10 years of experience. Participants of all groups were excluded if they had a history of any of the following diagnoses: organic brain syndrome, schizophrenia, substance dependence, psychosis not otherwise specified, bipolar disorder, bulimia nervosa, or binge-eating disorder. Further exclusion criteria for all participants were an IQ below 85; current substance abuse; inflammatory, neurologic, or metabolic illness; chronic medical or neurological illness that could affect appetite, eating behavior or body weight; clinically relevant anemia; pregnancy or breast feeding. Psychotropic medication within 4 weeks before the study was an additional exclusion criterion for all groups, with the exception of selective serotonin reuptake inhibitors in the acAN and the recAN group.

Clinical measures

Additional to the evaluation with the SIAB-EX [31], eating disorder-specific psychopathology was assessed with the German version of the self-report questionnaire Eating Disorder Inventory-2 (EDI-2) [32]. Depressive symptoms were explored using the German version of the Beck Depression Inventory-II (BDI-II) [33] and general levels of psychopathology and anxiety symptoms using the revised Symptom Checklist 90 (SCL-90-R) [34]. Pre-treatment physical activity within the 3 months before inclusion into our study was assessed using the corresponding module of the SIAB-EX as previously described by Holtkamp et al. [35] and Ehrlich et al. [36]. The intelligence quotient (IQ) was estimated with short versions of the German adaptation of the Wechsler Adult Intelligence Scale (Wechsler Intelligenztest für Erwachsene) [37] or the Wechsler Intelligence Scale for Children (Hamburg-Wechsler Intelligenztest für Kinder IV) [38] for study participants aged 15 years or younger. Demographic and clinical study data were collected and managed using the secure, web-based electronic data capture tool REDCap (Research Electronic Data Capture) [39].


Venous blood samples were collected into vacutainer tubes between 7 and 9 a.m. after an overnight fast, for the acAN group at the first timepoint within 96 h after initiating intensive treatment. Aprotinin (Sigma-Aldrich, St. Louis, Missouri, USA) was added during blood sampling to prevent protein degradation by serine proteases. Samples were immediately centrifuged (800×g for 15 min) in a pre-cooled centrifuge (5 °C), aliquoted, and stored at − 80 °C. Plasma PYY3–36 and leptin concentrations were measured according to the manufacturer's instructions. To prevent enzymatic in-vitro degradation of PYY, DPP-4 inhibitor (Millipore S.A.S, Molsheim, France) was added before thawing of samples in a concentration of 10 µl/ml. Plasma PYY3–36 concentration was analyzed in duplicate using the commercially available human PYY3–36-specific radioimmunoassay (RIA, EMD Millipore Corporation, St. Louis, Missouri, USA) with an intra-assay coefficient of variation (CV) of 6–11%, inter-assay CV of 7–15%, and a lower limit of detection of 20 pg/ml. The employed RIA utilizes 125I-labeled PYY and a PYY3–36 antiserum to determine the PYY3–36 concentrations in the research samples and is specific to human PYY3–36, while PYY1–36 is not detectable. Plasma leptin concentration was measured using a commercially available enzyme-linked immunoabsorbent assay with an intra-assay CV of 4.2%, inter-assay CV of 6.7%, and a lower limit of detection of 0.2 ng/ml (ELISA, BioVendor Research and Diagnostic Products, Brno, Czech Republic) and served as a control variable.

Statistical analysis

For all samples, PYY3–36 was measured twice and the mean was used for all statistical analyses. Due to deviations from normality, the non-parametric Kruskal–Wallis test was used to analyze PYY3–36 concentrations for the cross-sectional group comparison (acAN-T1, recAN, HC). The Wilcoxon signed-rank test was employed to analyze PYY3–36 concentrations longitudinally in the acAN group (acAN-T1, acAN-T2). Correlation analyses were performed using Spearman correlation and p values were corrected for multiple comparisons using the False Discovery Rate correction method of Benjamini and Hochberg [40]. Statistical analyses were performed using SPSS 25 (SPSS, Chicago, Illinois) and JASP for Bayesian statistics [41] with the goal to estimate the evidence in favor of the null hypothesis.


Sample characteristics

Demographic and clinical characteristics are summarized in Table 2 for the cross-sectional sample (acAN-T1, recAN, HC) and in Table 3 for the longitudinal sample (acAN-T1, acAN-T2). As expected, patients with acAN-T1 had significantly lower BMI and BMI standard deviation scores (BMI-SDS) and higher levels of psychopathology (EDI-2 total score, BDI-II total score, SCL-R-90 global severity index). RecAN patients still had some residual psychopathology. As anticipated, leptin concentration was suppressed in the acAN-T1 group, increasing significantly after short-term weight rehabilitation (acAN-T2), while it was in the expected range for the recAN and the HC groups. There was no difference in IQ among groups. Physical activity was increased in inpatients with acAN-T1 in comparison to HC and recAN, and decreased significantly over the course of short-term weight rehabilitation (acAN-T2).

Table 2 Cross-sectional study sample: demographic and clinical characteristics
Table 3 Longitudinal sample: demographic and clinical characteristics

PYY3–36 concentration

There was no significant difference in PYY3–36 concentration among the acAN-T1, recAN, and HC groups, H(2) = 4.04, p = 0.133 (Fig. 1). Bayesian statistics provided additional support for the null hypothesis (Supplemental Material 1.1). In our longitudinal sample of acAN participants, the PYY3–36 concentration did not change over the course of short-term weight rehabilitation, T = 178, p = 0.170, r = − 0.17 (Fig. 2).

Fig. 1
figure 1

PYY3–36 concentrations in the cross-sectional sample. acAN-T1 acute anorexia nervosa participants at timepoint 1 (admission), recAN long-term recovered anorexia nervosa participants; HC, healthy control participants. Boxplots showing the median, upper and lower quartiles, mild outliers (depicted as circles, values deviating more than 1.5 times the interquartile range from the upper or lower quartile) and extreme outliers (depicted as asterisks, values deviating more than 3 times the interquartile range from the upper or lower quartile), and the lower measurement limit indicated as a horizontal line. PYY3-36 concentrations: acAN-T1: median (Mdn) = 81.2 pg/ml, interquartile range (IQR) = 30.5, recAN: Mdn = 77.3 pg/ml, IQR = 32.9, HC: Mdn = 91.8 pg/ml, IQR = 46.2. 2/47 acAN-T1 participants, but none of the recAN or the HC participants had PYY3-36 concentrations below the lower measurement limit (29 pg/ml). The figure was created with SPSS 25 (SPSS, Chicago, Illinois)

Fig. 2
figure 2

PYY336 concentrations in the longitudinal sample. acAN-T1 acute anorexia nervosa participants at timepoint 1 (admission), acAN-T2 acute anorexia nervosa participants at timepoint 2 (after short-term weight rehabilitation). Boxplots showing the median, upper and lower quartiles, mild outliers (depicted as circles, values deviating more than 1.5 times the interquartile range from the upper or lower quartile), and extreme outliers (depicted as asterisks, values deviating more than 3 times the interquartile range from the upper or lower quartile), and the lower measurement limit indicated as a horizontal line. PYY3–36 concentrations: acAN-T1: median (Mdn) = 85.4 pg/ml, interquartile range (IQR) = 31.1, acAN-T2: Mdn = 76.9 pg/ml, IQR = 34.9. One participant had PYY3–36 concentrations below the lower measurement limit (29 pg/ml) for both timepoints. The figure was created with SPSS 25 (SPSS, Chicago, Illinois)

Associations of PYY3–36 with BMI, BMI-SDS, age, leptin concentration, pre-treatment physical activity (for the acAN group), duration of recovery (for the recAN group), and measures of psychopathology (BDI-II total score, EDI-2 total score, SCL-90-R global severity index) were tested for each group using Spearman correlations. For the acAN-T1 group, there was a negative correlation between PYY3–36 and BMI (rs = -0.45, p = 0.001, Benjamini–Hochberg corrected p = 0.009), but no significant correlation of PYY3–36 with any of the other parameters including age-adjusted BMI-SDS and pre-treatment physical activity. There were no significant associations between PYY3–36 and demographic and clinical variables for the acAN group at T2, the recAN, or the HC groups. An exploratory analysis to test a possible influence of the duration of realimentation and BMI change over the course of treatment on PYY3–36 concentrations in the longitudinal acAN study population revealed no significant influence of these factors (Supplemental Material 1.2).


In the present study, we investigated plasma PYY3–36 concentrations in patients with acute AN and individuals after long-term recovery from AN cross-sectionally in comparison to healthy control participants, and longitudinally over the course of inpatient treatment. To specifically measure the isoform most relevant to energy homeostasis [7], we chose a PYY3–36-specific RIA. There was no group difference in PYY3–36 concentrations. Our longitudinal analysis in acute AN patients also yielded no change in PYY3–36 concentration after short-term weight rehabilitation. At admission to treatment, we found a negative correlation between PYY3–36 concentration and BMI.

The results of the present study are in line with a number of studies reporting a normal PYY concentration in acute AN [16,17,18,19,20]. However, others have found elevated PYY levels in acute AN [21,22,23,24,25] and one study, measuring circadian levels of PYY, reported lower concentrations in AN in comparison to a healthy control group [30]. This heterogeneity of results may be explained at least in part by the use of assays targeting total PYY (sum of PYY1–36 and PYY3–36) instead of PYY3–36 [17,18,19,20, 22, 23, 42, 43]. The importance of differentiating between PYY1–36 and PYY3–36 is further underlined by a recent study demonstrating different physiological effects of the two isoforms [10]. Last but not least, physical activity may be a possible confounding factor especially relevant for this study population as patients in the acute stage of the disorder often engage in excessive exercise. Multiple studies have reported an exercise-induced increase of PYY concentrations in healthy subjects [26,27,28]. The duration of the exercise-related effect on PYY concentration appears to depend on the type and intensity of the physical activity, but does not seem to exceed a few hours. Given the intensive treatment setting at the time of venipuncture in our study, we were able to minimize acute effects of excessive physical activity on PYY. Our finding of a lack of correlation between the PYY concentration and pre-treatment physical activity supports this assumption.

Two longitudinal studies, which reported no significant change in PYY concentrations after weight gain, are in accordance with our findings [24, 25]. To our knowledge, there are only two studies investigating PYY in long-term recovered AN patients, with study designs that significantly limit comparability to our results [43, 44]. In line with our findings, Rigamonti et al. [43] did not report a significant difference in total PYY concentrations among acAN and recAN, but the sample size was very small (acAN n = 7, recAN n = 4) and no healthy control group was included in the study. Gendall et al. [44] also reported normal total PYY levels for recAN patients, but PYY was measured in cerebrospinal fluid instead of blood.

Our finding of a negative correlation between PYY concentration and BMI in the acute AN group before treatment is consistent with the results of a number of studies [21, 23, 25, 43, 45]. While our result should be interpreted cautiously as it was not confirmed for age-adjusted BMI-SDS, it is possible that a higher PYY concentration contributes to the severity of the disorder by reducing appetite and food intake. On the other hand, a low PYY concentration might be a protective factor against severe undernutrition in a subgroup of patients as it may counteract self-starvation by increasing appetite and food intake. In this context, Misra et al. [23] remarked a strong negative association between fasting PYY levels and fat intake in girls with AN.

When considering our findings, some limitations have to be taken into account. First, the sample size may not have been sufficient to draw definite conclusions regarding basal PYY3–36 concentrations in AN. Second, taking into account the complexity of the function of PYY as a regulator of energy homeostasis, measurement of fasting concentrations is only one approach to investigate its role in the etiology and progression of AN and future studies in acute AN should also target the PYY3–36 response to standard meals. To date, only four studies investigated PYY in response to meal intake in AN with mixed results [18,19,20, 24] and only one of these measured PYY3–36 instead of total PYY [24]. Third, because of the possibly divergent actions of the two human isoforms PYY1–36 and PYY3–36, we chose an assay that targets only the isoform PYY3–36, which is assumed to be more relevant for energy homeostasis. Thus, our results allow no conclusions regarding the physiological effects of PYY1–36 and measurement of both, total PYY (the sum of PYY1–36 and PYY3–36) and PYY3–36, yielding their ratio (i.e., total PYY/PYY3–36 or PYY1–36/PYY3–36), would allow for a more comprehensive picture of the peptide’s metabolism in AN.

In conclusion, the current study provides additional evidence for a normal basal PYY3–36 concentration in AN. Future studies should further measure both PYY1–36 and PYY3–36, explore their levels in response to food intake in AN, and take physical activity as a possible confounding factor into consideration. Another promising approach would be to simultaneously study multiple appetite-regulating peptides and their complex interplay.