Peptide YY3–36 concentration in acute- and long-term recovered anorexia nervosa

Purpose The gut–brain axis could be a possible key factor in the pathophysiology of anorexia nervosa. The neuropeptide peptide YY3–36, secreted by endocrine L cells of the gastrointestinal tract, is a known regulator of appetite and food intake. The objective of this study was to investigate peptide YY3–36 plasma concentrations at different stages of anorexia nervosa in a combined cross-sectional and longitudinal design to differentiate between effects of acute undernutrition and more enduring characteristics. Methods We measured fasting plasma peptide YY3–36 concentrations in young patients with acute anorexia nervosa (n = 47) and long-term recovered patients (n = 35) cross-sectionally in comparison to healthy control participants (n = 58), and longitudinally over the course of inpatient treatment. Physical activity was controlled as it may modulate peptide YY secretion. Results There was no group difference in peptide YY3–36 concentration among young acutely underweight anorexia nervosa patients, long-term recovered anorexia nervosa patients, and healthy control participants. Longitudinally, there was no change in peptide YY3–36 concentration after short-term weight rehabilitation. For acute anorexia nervosa patients at admission to treatment, there was a negative correlation between peptide YY3–36 concentration and body mass index. Conclusions The current study provides additional evidence for a normal basal PYY3–36 concentration in AN. Future studies should study multiple appetite-regulating peptides and their complex interplay and also use research designs including a food challenge. Electronic supplementary material The online version of this article (10.1007/s00394-020-02210-7) contains supplementary material, which is available to authorized users.


SM 1.1 Control analyses: Bayesian statistics
We employed Bayesian statistics conducted with JASP [1] to verify our results. One advantage of these tests is the possibility to test which hypothesis is best supported by the data and to specifically also test the evidence in favor of the null hypothesis (H0).
From the results of prior studies on PYY and PYY3-36 in AN, it is possible to make an informed assumption about the pair-wise comparison between acAN and HC. To determine a likely effect size for the comparison between acAN and HC, we aimed to include all studies about total PYY or PYY3-36 in AN with a representative sample size of n>20 [2][3][4][5][6]. However, Germain et al. [5] was not included since the statistical parameters necessary to determine an effect size were not reported, and Misra et al., 2006 [6] was not included as Misra et al., 2008 [4] used data from the same subjects in their analyses. Thus, we used the results of the remaining three studies [2][3][4] to calculate the Cohen's d statistic for the pair-wise comparison between AN and HC participants according to the following formula: where M1 and M2 are the means of the AN group and the HC group and SDpooled is the pooled standard deviation, calculated according to the following formula (SD1 and SD2 are the standard deviations of the AN group and the HC group): The calculated effect sizes were d=10.98 for Eddy et al. [2] who measured PYY3-36 in 75 AN and 22 HC, d=0.33 for Fernández-Aranda et al. [3] who measured total PYY in 64 AN and 80 young HC, and d=1.60 for Misra et al. [4] who measured total PYY in 34 AN and 33 HC. The mean effect size for these three studies was d=4.30, which provides strong a priori evidence for the alternative hypothesis PYY: AN>HC.
Therefore, we conducted a Bayesian two-sample t test to test whether the data best supported the alternative hypothesis PYY3-36: acAN>HC (H+) or H0. The Bayes factor BF0+ (H0/H+) suggested that the data were 11.6:1 in favor of the null hypothesis (strong evidence, Figure S1). Figure S1. Output of the Bayesian two-sample t-test performed with JASP [1]. The probability wheel visualizes the evidence that the data provide for the null hypothesis H0 and the one-sided alternative hypothesis H+ (PYY3-36: acAN > HC). The population effect size was assigned a Cauchy prior distribution with a prior width = 1 √2 , truncated to allow only positive effect size values.
For the comparison between recAN and HC, there were no previous studies of PYY in blood. So there were two possible assumptions: the increase of PYY in acute AN observed in previous studies could a) (partially) persist or b) completely normalize with weight recovery. Therefore, we conducted a Bayesian two-sample t test to test whether the data best supported the alternative hypothesis PYY3-36: recAN > HC (H+) or H0. The Bayes factor BF0+ (H0/H+) suggested that the data were 12.2:1 in favor of the null hypothesis (strong evidence, Figure S2). Figure S2. Output of the Bayesian two-sample t-test performed with JASP [1]. The probability wheel visualizes the evidence that the data provide for the null hypothesis H0 and the one-sided alternative hypothesis H+ (PYY3-36: recAN > HC). The population effect size was assigned a Cauchy prior distribution with a prior width = 1 √2 , truncated to allow only positive effect size values.

SM 1.2 Control analyses: Effects of duration of realimentation and BMI change
We conducted an exploratory analysis to test a possible influence of the duration of realimentation and BMI change from timepoint 1 (T1, admission to treatment) to timepoint 2 (T2, after short-term weight restoration) on PYY3-36 concentrations in the longitudinal AN study population (see Table S1 for statistical parameters). The median BMI change from T1 to T2 was 24.3% (interquartile range (IQR)=12.2, minimum=13.7%, maximum=47.9%). The median duration of realimentation from T1 to T2 was 84.5 days (IQR=27.8, minimum=51 days, maximum=256 days).
We divided the longitudinal study population into two subgroups: one subgroup with a duration of realimentation below the group median of 84.5 days and one subgroup with a duration of realimentation greater than or equal to the median. Furthermore, we tested if the correlation between PYY3-36 concentration and BMI-SDS in the acAN group at T2 was influenced by duration of realimentation or BMI change from T1 to T2 (see Table  S2 for statistical parameters). In the subgroup with shorter duration of realimentation, there was no significant relationship between PYY3-36 concentration at T2 and BMI-SDS at T2. Similarly, in the subgroup with longer duration of realimentation, there was no significant relationship between PYY3-36 concentration at T2 and BMI-SDS at T2. In the subgroup with a BMI change below the median of 24.3%, there was no significant relationship between PYY3-36 concentration at T2 and BMI-SDS at T2. Similarly, in the subgroup with a BMI change greater or equal to 24.3%, there was no significant relationship between PYY3-36 concentration at T2 and BMI-SDS at T2. Spearmann correlation coefficients and p values were reported. There were no statistically significant correlations.
Abbreviations: acAN-T1, acute anorexia nervosa participants at timepoint 1 (admission); acAN-T2, acute anorexia nervosa participants at timepoint 2 (after short-term weight rehabilitation); BMI-SDS, body mass index standard deviation score; IQR, interquartile range. Short duration of realimentation was defined as below the group median of 84.5 days, long duration of realimentation was defined as equal or greater than 84.5 days. Low BMI change was defined as below the group median of 24.3%, high BMI change was defined as equal or greater than 24.3%.