A comprehensive updated cross-sectional and longitudinal meta-analysis of cytokines in eating disorders

A comprehensive updated cross-sectional and longitudinal meta-analysis of cytokines in eating disorders | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A comprehensive updated cross-sectional and longitudinal meta-analysis of cytokines in eating disorders Johanna Keeler, Charlotte Bovenberg, Hubertus Himmerich, Janet Treasure, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6547856/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 01 Oct, 2025 Read the published version in Communications Medicine → Version 1 posted You are reading this latest preprint version Abstract Background Prior research has found altered levels of immune signalling proteins, such as cytokines, in people with eating disorders (EDs). This study is an update of a previously published meta-analysis. Methods This systematic review and meta-analysis assessed cross-sectional and longitudinal studies from four databases (PubMed, Web of Science, MEDLINE and PsycINFO) reporting cytokine concentrations in people with EDs. Random-effects models were utilised for all meta-analyses. Results Twenty-four new studies were incorporated, resulting in a total of 43 studies included in the meta-analyses. Interleukin (IL)-6 and IL-15 were higher, and IL-7 was lower, in AN compared with HC. When controlling for outliers, concentrations of tumour necrosis factor (TNF)-α, IL-1β, IL-4, IL-8, IL-10, IFN-γ, monocyte chemoattractant protein (MCP) and transforming growth factor (TGF)-β were similar between AN and HC. Longitudinally, IL-6 was lower in AN at follow-up compared to baseline, whereas TNF-α and IL-1β did not change. There were largely no differences in IL-6 and TNF-α in BN and were insufficient studies to perform meta-analyses for binge eating disorder (BED) or other EDs. Conclusion In acute AN, concentrations of IL-6 and IL-15 are elevated and IL-7 is decreased, with evidence for normalisation of IL-6 over the course of weight restoration. Concentrations of other cytokines considered to broadly have pro-inflammatory functions were not increased in AN. In people with BN, there is less evidence for increases in pro-inflammatory cytokines, but the evidence base is limited. Methodological considerations for future studies are recommended. Biological sciences/Neuroscience/Neuroimmunology Biological sciences/Immunology/Cytokines Biological sciences/Psychology/Human behaviour Health sciences/Diseases/Psychiatric disorders Health sciences/Diseases/Nutrition disorders/Malnutrition Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Eating disorders (EDs) including anorexia nervosa (AN), bulimia nervosa (BN), binge eating disorder (BED), avoidant/restrictive food intake disorder (ARFID) and other specified feeding and eating disorder (OSFED), involve alterations in feeding and eating patterns and in some cases compensatory behaviours as an attempt to prevent weight gain or offset calorie intake. This group of disorders results in secondary consequences that impact physical systems within the body. One of these is the immune system; both the food restriction seen in EDs such as AN and patterns of overeating seen in binge-type EDs may lead to changes in the production and functioning of immune cells. Additionally, recent research has implicated the presence of autoimmune conditions as a risk factor for the development of an ED, and vice versa ( 1 ). Prior research has examined peripheral concentrations of cytokines in ED populations as indicators of changes in immune signalling. Cytokines are signalling proteins produced by immune cells, endothelial and epithelial cells, adipocytes and connective tissue, which can aid communication between cells in order to modulate the immune response. For example, by stimulating or slowing down the immune system or stimulating the movement of cells toward sites of infection or inflammation. They can have systemic as well as local effects and are pleiotropic in that a single cytokine can produce multiple biological functions. Common groups of cytokines include chemokines, which induce chemotaxis (cell migration); interleukins, which act as chemical signals between white blood cells; interferons, which help the body resist viral infections and cancers; and the tumor necrosis factor (TNF) family, which have a wide range of effects on immune functioning, including in tissue modeling and remodeling, and neuronal development. They are often categorised as pro-inflammatory or anti-inflammatory ( 2 ), although as aforementioned most cytokines exhibit pleiotropy and therefore may have either effects depending on factors such as the site of expression and/or cell target. Blood-borne cytokines are able to pass the blood-brain barrier to enter cerebrospinal fluid and interstitial fluid spaces of the brain ( 3 ), although they are also produced in the brain by various cells such as neurons, astrocytes and microglia. Cytokines have been implicated in eating disorders due to their effects on appetite, body weight and eating behaviours, which may be mediated by their effects on brain areas such as the hypothalamus ( 4 ). Three meta-analyses have quantified differences in concentrations of cytokines between ED populations and healthy controls ( 5 – 7 ), one of which examined only AN samples ( 5 ). When comparing AN to healthy controls, these studies have found small-to-moderate sized increases in levels of the pro-inflammatory cytokines IL-1ß, IL-6 and TNF-α ( 5 , 6 ) and TNF-receptor-II ( 5 ) in AN. In a meta-analysis of inflammatory markers in EDs as a broader group, it has also been found that levels of TNF-α, and IL-1β are elevated in comparison with healthy control individuals (HCs) ( 7 ). This has led to propositions that increases in pro-inflammatory cytokines may be a factor implicated in the aetiology of AN ( 8 ) and neuroinflammation is often cited as a possible aetiological factor in EDs (e.g., ( 9 )). However, this is not without debate, as several recent studies in both adolescents and adults with AN have failed to fully replicate these results and sources of this heterogeneity (moderators) are thus far unclear ( 10 – 14 ). This heterogeneity may be explained by differences in sample characteristics (e.g., age, illness duration) and study factors (e.g., controlling or not controlling for factors such as age, tobacco use, use of pharmacological medications) ( 6 ). Most of the studies included in the 2015 and 2018 meta-analyses ( 5 , 6 ) did not control for confounding clinical and lifestyle factors that may affect cytokine concentrations, such as age, menstruation, smoking status, medication usage, recent food intake, exercise, body fat, recent illness and concurrent medical (particularly inflammatory) or mental health diagnoses ( 15 – 20 ). Since our last meta-analysis in 2018 ( 6 ), several papers have been published that control for, or consider, such variables, which have had contrasting findings. Therefore, it is conceivable that the inclusion of newer studies controlling for such influences, or examining the effect of study quality, may modulate the overall effect sizes. A prior meta-analysis found no changes after weight gain (BMI ≥ 17.5kg/m 2 ) in IL-6, TNF-α or IL-1β in people with AN followed longitudinally across three studies ( 5 ). However, several studies with longitudinal designs over the course of weight restoration have since been published, enabling analyses with improved power. In conditions such as AN, it is possible that concentrations of cytokines are modified as dietary intake is reinstated ( 8 ), and it has also been suggested that reductions in neuroinflammation may contribute to weight gain ( 9 ). This study aims to build upon a prior meta-analysis published by our group ( 6 ), asking the following additional research questions: Including new evidence, do cytokine concentrations differ between people diagnosed with an ED and healthy individuals? Using machine learning approaches, are there any moderators that explain the observed heterogeneity between studies? How do between-group effect sizes vary as a function of study quality or publication date? In ED participants, do cytokine concentrations change as a function of weight gain and/or symptom improvement? Methods and materials The reporting of this systematic review was guided by the standards of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) Statement ( 21 ) and a PRISMA checklist can be found in the Supplementary Materials. The quality assessment of the studies was conducted according to a modified version of the Newcastle-Ottowa Scale (NOS; ( 22 )). The full quality appraisal of the studies can be found in Table S1 . The study protocol was prospectively registered on the Open Science Framework, which can be found at https://osf.io/g6d3f . 2.1. Literature search Two reviewers (JLK and CB) independently searched four electronic databases (PubMed, Web of Science Core Collection, MEDLINE and PsycINFO) from the date of the last review, 4th May 2018, until 10th November 2024. An identical search strategy from the previous review was replicated for this review, which can be found at Dalton et al. ( 6 ). In short, searches included the following keywords, mapped to Medical Subject Heading with the Explode function where possible: eating disorder*, anorexia nervosa, bulimi* or binge eat* in combination with cytokine* chemokine*, inflammat*, interleukin, interferon, IFN, tumor necrosis factor, TNF, transforming growth factor or TGF. Searches were supplemented using the ascendancy approach (hand-searches of reference lists of relevant papers and reviews) and the descendancy approach (citation tracking in Google Scholar, Crossref). 2.2. Eligibility criteria Studies in any language of any study design that assessed cytokine concentrations in the serum, plasma or cerebrospinal fluid (CSF) of individuals with a Diagnostic and Statistical Manual of Mental Disorders (DSM) ( 23 , 24 ) or International Statistical Classification of Diseases (ICD) ( 25 , 26 ) diagnosis of ED were eligible for inclusion. Studies were included if they reported cross-sectional comparisons of cytokine concentrations between an ED group and HCs, or longitudinal assessments at a minimum of two time-points, with BMI or ED symptoms assessed at both time points. Studies were excluded if: i) they did not report group comparisons or longitudinal measurements of cytokine concentrations; ii) participants had an organic cause for their disordered eating e.g., cancer, immunological conditions, genetic disorder, etc.; iii) the sample was comprised of animals; iv) they measured cytokine production or genetic expression but did not assess cytokine concentrations; or v) there was significant reported or suspected study overlap between publications, defined as the same patient and or control group. Review articles, meta-analyses, conference proceedings/abstracts, book chapters, and unpublished theses were also not included. 2.3. Source selection Titles and abstracts of publications were imported into Endnote, where duplicates were removed. During an initial screening stage, titles and abstracts were assessed against basic eligibility criteria and those deemed likely to be irrelevant (e.g., studies in animal samples, studies of other medical or psychiatric conditions, studies with ineligible study designs such as reviews) were discarded. Full-texts of the remaining articles were then assessed against the full eligibility criteria, with the reasons for study exclusion documented (Fig. 1 ). The search process was conducted independently by two reviewers (JLK and CB). 2.4. Data extraction Two reviewers (JLK and CB) conducted the data extraction process and the extracted data was checked by one reviewer (JLK). At this stage, the references of the eligible publications were uploaded into an electronic summary table used in the previous study ( 6 ) and databases were combined for extraction. Extracted data included publication identifiers (study title, year, author list); sample characteristics, including sample size, demographics (e.g. mean ± SD age), diagnostic criteria and clinical characteristics (e.g. mean ± SD illness duration, mean ± SD BMI), and medication status; parameters of interest, measurement methods, and concentrations of cytokines (mean ± SD), time interval between measurements (longitudinal studies, in weeks or days), content of intervention between measurements (longitudinal studies); potential confounds and control for confounds including presence/absence of smoking as an inclusion criteria, number of participants taking psychotropic medications (and types); main and additional study outcomes. Authors of potentially eligible publications were contacted where data were not reported (n = 13). In total, eight of these authors replied and provided data ( 11 , 13 , 27 – 32 ). For studies that reported AN subtype data separately, the means and standard deviations were pooled for the AN meta-analyses. Standard errors were converted to standard deviations using the formula SD = \(\:\sqrt{N}\:x\:SE\) where N represents the sample size. Where there were fewer than four studies for a given cytokine or raw data could not be obtained from authors, results were narratively synthesised. 2.7. Synthesis of results Where studies were clinically homogeneous they were considered for pooling in a meta-analysis. Meta-analyses were performed for each cytokine where four or more studies were available for each ED separately. 2.6. Statistical analysis Individual meta-analyses for each cytokine were conducted in Stata 18.0 ( 33 ) using the ‘metan’ command. The main outcome measure was cytokine concentrations (pg/ml or ng/ml). Random effects-models using the DerSimonian & Laird method ( 34 ) were used for all meta-analyses to pool the standardised mean difference (SMD) of the studies relative to the study sample size. The SMD of each study reflects the size of effect between the comparator groups in each study (i.e., ED vs. HC, or ED baseline vs. follow-up) relative to the variability observed in the study. All longitudinal meta-analyses utilised the two available time-points, or where multiple time-points were available, utilised the follow-up time-point representing weight restoration or discharge from care. Statistical significance of the overall SMD was ascertained according to the p < 0.05 threshold. Outliers were identified when the confidence intervals (CI) of the individual study SMD did not overlap with the CI of the pooled effect (i.e., where the upper CI bound of the individual study effect is smaller than the lower CI bound of the pooled effect, or the lower CI bound of the individual study effect is bigger than the upper CI bound of the pooled effect). Sensitivity analyses removing identified outliers were conducted and SMDs were re-estimated. Publication bias was explored using the Egger’s test for small study effects and funnel plots. The Duval and Tweedie trim and fill method ( 35 ) was used to identify smaller studies causing funnel plot asymmetry, adjust for this asymmetry, and re-estimate SMDs after adjusting for missing studies. 2.6.1. Cumulative meta-analysis To explore how the effect sizes evolved over time for the cross-sectional comparisons of the main pro-inflammatory cytokines previously synthesised (IL-6, TNF-α and IL-1β) between AN and HC, cumulative meta-analyses were performed using NOS study quality total score, and year of publication as ordering variables. This enables an examination of how the effect size changes when adding new studies, starting from a) the newest studies to the oldest studies, and b) the highest quality studies to the lowest quality studies. Cumulative meta-analyses were displayed as forest plots (Figures S1 -4) and were interpreted by visual inspection. 2.6.2. Between-study heterogeneity Between-study heterogeneity was assessed with the Higgins I 2 statistic, based on the Cochrane’s Q (chi-square) statistic, which represents the proportion of total variation across studies that is attributable to heterogeneity rather than chance. Heterogeneity was considered to be substantial if above 50%, and considerable if above 75%. Tau 2 , which estimates the amount of true variability between the SMDs of the included studies, was also reported. Subgroup analyses were conducted with AN subtypes, where the study reported data for each subtype (i.e., AN-restricting type [-R] and AN-binge-eating/purging type [-BP]) individually. To further explore sources of heterogeneity, a random-effects MetaForest analysis was conducted in R version 4.4.2 (R Core Team, 2021). MetaForest is a machine-learning based exploratory approach to meta-analysis that mitigates overfitting and assesses the relative importance of potential moderators (36) (see SM section 6 for further details). Analyses were only performed where there were 10 or more studies available. Pre-specified theoretical and methodological moderators are listed in Table 1. Table 1. Pre-specified moderators coded for meta-analysis Moderator Codes Region of study origin Asia Europe North America Year - NOS total score - Age group Adolescents Adults Mixed Sample fasting status Fasted Not fasted/not reported Type of blood sample Serum Plasma Use of psychotropic medication in the ED sample (%) - Percentage of smokers in sample (%) - Mean age of ED group (years) - Mean age of HC group (years) - Mean BMI of ED group (kg/m 2 ) - Mean BMI of HC group (kg/m 2 ) - Mean duration of illness in ED group (years) - Abbreviations: BMI = body mass index; ED = eating disorder; HC = healthy controls; NOS = Newcastle-Ottowa Scale. Results 3.1. Characteristics of included studies An additional 18 studies not included in the previous meta-analysis (6) were identified, resulting in a total of 43 studies from 11 countries included in the meta-analysis. Of these, 40 were included in the cross-sectional meta-analyses comparing cytokine concentrations between AN and HCs, seven in the cross-sectional meta-analyses comparing BN with HCs, and 11 in the meta-analyses examining cytokine concentrations longitudinally in AN. Table 2 details the study and sample characteristics for studies included in the meta-analyses. Findings of studies not included in the meta-analyses, due to insufficient data reporting or insufficient number of studies for meta-analysis (including cross-sectional comparisons between BED and HCs and recovered AN and HCs), are summarized in Table S3 and in the supplementary materials. Cytokines included in the cross-sectional meta-analyses between AN and HC were TNF-α, IL-1β, IL-6, IL-4, IL-7, IL-8, IL-10, IL-15, IFN-γ, TGF-β and MCP. Cytokines included in the cross-sectional meta-analyses between BN and HC were IL-6 and TNF-α. Longitudinally in AN, cytokines available for meta-analysis included TNF-α, IL-1β, and IL-6. Other cytokines were also measured in eligible studies, however sufficient data were not available to perform meta-analyses. Across studies, 31 studies measured cytokine concentrations peripherally in serum, 12 in plasma, and none in CSF. The mean and SD age of participants with EDs and HCs was 21.7±8.1 (n = 32 studies) and 21.4±6.7 (27 studies), respectively. All studies included only female participants. The mean BMI of the AN participants was 15.7±2.3 (n=34 studies), of BN participants was 31.7±10.4 (n=4 studies), and of HCs was 21.2±3.0kg/m 2 (n=29 studies). Diagnosis was based on the DSM-V (n=14), DSM-IV (n=28) or DSM-III (n=1). Mean illness duration for ED participants, reported in 16 studies, was 5.2±7.2 years. Medication usage of people in the ED groups was reported in 32 studies, and in 20 studies participants were confirmed medication-free as part of assessment or eligibility procedures. In the remaining 12 studies, medications used included anti-depressants, anti-psychotics, sedatives, neuroleptics and anxiolytics. For the longitudinal studies in AN, the mean BMI at baseline was 15.6±1.9kg/m 2 (n=8 studies) and at follow-up was 18.4±2.0kg/m 2 (n=7 studies). The interval between baseline and follow-up assessments used in the meta-analyses ranged from 2 to 12 months. The intervention used across studies included inpatient weight restoration treatment (n=5), specialist eating disorder treatment (n=1), unspecified weight gain treatment (n=1) and with cognitive behavioural therapy (CBT; n=1), CBT (n=1), and CBT with pharmacological therapy (n=1). In one study, treatment was unspecified. 3.2. Quality assessment of included studies The rating of studies depends on three characteristics: the selection of study groups, the comparability of groups and the ascertainment of the outcome. Of the included studies, 26 were good quality, 24 were fair quality and three were poor quality (see SM 1.1 for the full description of the quality assessment). 3.3. Cross-sectional meta-analysis results The results of the cross-sectional meta-analyses comparing participants with AN to HC, and participants with BN to HC, are displayed in Table 3. Table 3. Summary of comparative outcomes and heterogeneity for all conducted cross-sectional meta-analyses. Cytokine ( n studies) Sample N (ED, HC) SMD 95% CI z p Heterogeneity Egger’s test Anorexia nervosa vs healthy controls TNF-α (n=30) 872, 764 0.27 a 0.01, 0.54 2.06 0.040 Tau 2 = 0.43 I 2 = 84.05% z = 1.63 p = 0.103 IL-6 (n=25) 822, 774 0.36 b 0.11, 0.61 2.82 0.005 Tau 2 = 0.32 I 2 = 81.81% z = 3.03 p = 0.002** IL-1β (n=14) 354, 244 0.53 c -0.06, 1.12 1.75 0.080 Tau 2 = 1.14 I 2 = 90.59% z = 0.57 p = 0.570 IL-10 (n=7) 278, 250 0.41 d -0.17, 0.99 1.39 0.166 Tau 2 = 0.50 I 2 = 88.09% z = 3.57 p < 0.001** IFN-γ (n=6) 171, 146 0.37 e -0.41, 1.15 0.92 0.356 Tau 2 = 0.83 I 2 = 89.94 z = 2.60 p = 0.009** IL-8 (n=5) 235, 229 -0.16 f -0.78, 0.47 -0.49 0.626 Tau 2 = 0.44 I 2 = 88.83% z = -0.21 p = 0.831 MCP (n=5) 235, 229 -0.19 -0.44, 0.06 -1.50 0.135 Tau 2 = 0.03 I 2 = 33.07% z = -2.29 p = 0.022* TGF-ß (n=5) 240, 183 0.61 g -0.90, 2.11 0.79 0.431 Tau 2 = 2.75 I 2 = 96.76% z = 3.39 p = 0.001** IL-4 (n=4) 124, 105 -0.01 -0.28, 0.25 -0.09 0.925 Tau 2 = 0.00 I 2 = 0.00% z = 1.29 p = 0.196 IL-7 (n=4) 211, 188 -0.58 -1.07, -0.09 -2.32 0.020 Tau 2 = 0.18 I 2 = 75.55% z = -0.55 p = 0.582 IL-15 (n=4) 183, 126 0.67 0.07, 1.26 2.19 0.029* Tau 2 = 0.30 I 2 = 82.74% z = 1.96 p = 0.050 Bulimia nervosa vs healthy controls IL-6 (n=5) 163, 115 0.73 h -0.28, 1.73 1.42 0.157 Tau 2 = 1.21 I 2 = 92.76% z = -0.48 p = 0.632 TNF-α (n=5) 145, 123 2.34 i 0.02, 4.66 1.98 0.048* Tau 2 = 6.84 I 2 : 97.88% z = 4.91 p < 0.001** Notes. a When outliers removed SMD = 0.06, p = 0.475 (TNF-α AN); b When outliers removed SMD = 0.32 (IL-6 AN); c When outliers removed SMD = 0.23 (IL-1β AN); d When outlier removed SMD = –0.04 (IL-10 AN); e When outlier removed SMD = –0.11 (IFN-γ AN); f When outlier removed SMD = 0.13 (IL-8 AN); g When outliers removed SMD –0.15 (TGF-β AN); h When outlier removed SMD = 0.30 (IL-6 BN). i When outlier removed SMD = 0.06, p = 0.475 (IL-6 TNF-α).* Significant at the p <0.01 threshold; ** Significant at the p <0.005 threshold. Abbreviations: CI = confidence intervals; ED = eating disorder; HC = healthy control; IFN-γ = interferon-gamma; IL = interleukin; MCP = monocyte chemoattractant protein; p = p -value; SMD = standardized mean difference; TNF-α = tumour necrosis factor-alpha; z = z-score. 3.3.1. Anorexia nervosa 3.3.1.1. Tumour-necrosis factor-α (TNF-α) Across 30 studies with a total of 1636 participants (AN n = 872; HC n = 764), six found elevated levels of TNF- α and none reported reduced levels (Figure 2). Across all studies there was a small difference in TNF-α concentrations between AN and HCs whereby AN had elevated levels (SMD = 0.27; 95% CI 0.01, 0.54; p = 0.040). Heterogeneity estimates were high ( I 2 =84%; Table 3) and although there was no evidence of publication bias from the Egger’s test, six outliers were identified (12, 32, 40, 57, 60, 69). After removing these outliers, the effect size became marginal and non-significant (SMD = 0.06; 95% CI –0.10, 0.22; z = 0.71; p = 0.475) and the I 2 heterogeneity estimate reduced to 47.26%. A total of 15 studies reported AN subtype data for TNF-α, which indicated no subgroup difference between the AN-R and AN-BP subtypes ( p = 0.690; Figure S25). Again, when including only these studies, neither meta-analyses showed significant differences between the individual subgroups (i.e., AN-R or AN-BP) and HCs. Cumulative meta-analyses were conducted using year and NOS total study quality score as sorting variables in a descending fashion. A visual inspection highlighted that newer studies were associated with lower SMDs between AN and HC for TNF-α (Figure S1), although the pattern for the NOS score was less clear (Figure S2). To further investigate sources of heterogeneity, a random-effects MetaForest analysis was conducted. A random effects MetaForest analysis provided no evidence for associations between the entered moderators and the SMD for TNF-α, even when removing study outliers (SM 6.1). 3.3.1.2. Interleukin-6 (IL-6) Across 25 studies including a total of 1596 participants (AN n = 822, HC n = 774), nine exhibited elevated levels and two exhibited reduced levels of IL-6 in AN (Figure 3). The pooled mean concentrations of IL-6 were significantly higher in AN compared to HC with a small effect (SMD = 0.36; 95% CI 0.11, 0.61; p = 0.005; Table 3; Figure 2). This analysis showed high heterogeneity ( I 2 = 82%) and the Egger’s test for small study effects was significant. Five studies were identified as outliers following an inspection of the 95% CIs (11, 12, 32, 40, 60). When removing these studies the SMD was slightly lower but remained significant (SMD=0.32; p = 0.001); the I 2 statistic lowered to 65.1% and the Egger’s test was no longer significant (z = 1.98, p = 0.050). A total of 15 studies reported AN subtype data, which indicated no subgroup difference between the AN-R and AN-BP subtypes ( p = 0.790; Figure S26). With this fewer number of included studies, neither meta-analysis showed significant differences between the individual subgroups (i.e., AN-R or AN-BP) and HCs. In cumulative meta-analyses, both forest plots indicated that higher quality studies and newer studies had overall lower SMDs between AN and HC (Figures S3 and S4). After removing study outliers, a random effects MetaForest analysis identified the following moderators as the most important in explaining the effect size: study year, NOS score, study region, age group, percentage of sample using medication, mean age, BMI, and illness duration of the AN sample, and mean age and BMI of the HC sample (relative variable importance shown in Figure S34). These moderators were entered into meta-regressions (Table S2). Only study country was significant, with studies conducted in North America showing a greater difference between AN and HC compared to studies in Europe (B = 0.78; SE = 0.29; z = 2.72; p = 0.007; 95% CI 0.22, 1.34). 3.3.1.3. Interleukin-1β (IL-1β) Across 14 studies with a collective sample size of 598 participants (AN n = 354, HC n = 244), there was a non-significant moderate-sized increase in pooled concentrations of IL-1β between AN and HC groups (SMD = 0.53, 95% CI -0.06, 1.12; p = 0.080; I 2 = 91%; Table 3; Figure 4). After exploring the heterogeneity three studies were identified as outliers(40, 60, 64). The effect size was smaller and remained non-significant after removing these studies (SMD = 0.23; 95% CI –0.04, 0.50; z = 1.67; p = 0.095; I 2 = 44%. Subgroup analyses according to AN subtype were not possible, as few studies reported data per subtype. A random effects MetaForest analysis provided no evidence for associations between the entered moderators and the SMD for IL-1b (SM 6.3). 3.3.1.4. Other cytokines available for meta-analysis Eight additional cytokines had at least four studies available and thus were meta-analysed (IL-4, IL-7, IL-8, IL-10, IL-15, IFN-γ, MCP and TGF-β; Figures S27 and S28, Table 3). Only concentrations of IL-15 and IL-7 were significantly different between AN and HC, whereby IL-15 was higher in the AN group with a moderate effect size (SMD = 0.67; 95% CI 0.07, 1.26; p = 0.029; I 2 = 83%) and IL-7 was lower in AN with a moderate effect size (SMD = -0.58, 95% CI -1.07, -0.09; p = 0.020; I 2 = 76%). Concentrations of IL-4, IL-8, IL-10, IFN-γ, MCP and TGF-β were not significant between AN and HCs. One study was identified as an outlier in the meta-analyses of IL-10 and IFN-γ (27), and the removal of this study reduced the SMDs to –0.04 (95% CI –0.28, 0.20; p = 0.761; I 2 = 32%) and –0.11 (95% CI –0.51, 0.28; p = 0.571; I 2 = 59%), respectively. One outlier was identified in the IL-8 meta-analysis (32), and the removal of this study increased the SMD to 0.13 (95% CI –0.40, 0.67; p = 0.626). Two studies (40, 60) were identified as outliers in the TGF-β meta-analysis, and the removal of these lowered the SMD to –0.14 (95% CI –0.11, 0.84; p = 0.783). According to I 2 , estimated heterogeneity was high in the meta-analyses of IL-7, IL-8, IL-10, IL-15, IFN-γ and TGF-β. The trim and fill method resulted in two imputed studies in both the IL-4 and MCP analyses (Figures S21 and S22); adjusting for these imputed studies altered the SMDs to –0.09 and –0.10, respectively (both still non-significant). Funnel plots for meta-analyses are included in the supplementary materials (Figures S8-15). 3.3.2. Bulimia nervosa 3.3.2.1. Interleukin-6 (IL-6) The meta-analyses of the studies in BN populations indicated higher concentrations of IL-6 in participants with BN compared with HCs, although this was not significant (Table 3; Figure 5).One study (66) was deemed an outlier. Removing this study from the IL-6 meta-analysis reduced the pooled SMD to 0.30 (95% CI -0.25, 0.86; p = 0.287; I 2 = 65%). 3.3.2.2. Tumour necrosis factor-α (TNF-α) There were significantly higher concentrations of TNF-α in participants with BN compared with HCs (Table 3; Figure 5). The heterogeneity estimates were considerably high for both analyses. The funnel plot indicated asymmetry (Figure S17) which was confirmed by a significant Egger’s test for small study effects (z = 4.91, p < 0.001). One study (66) was deemed an outlier in and removing this study reduced the overall effect size (SMD = 0.72) rendering the difference between groups non-significant (95% CI -0.14, 1.58; p = 0.101) and slightly reducing the I 2 heterogeneity statistic to 84%. After removing this study the Egger’s test was also no longer significant (z = 1.30, p = 0.192). 3.4. Longitudinal meta-analyses in AN There were sufficient data to meta-analyse differences between baseline and follow-up in individuals with AN undergoing weight restoration for the cytokines IL-6, IL-1β and TNF-α. All meta-analyses utilised the two available time-points, or where multiple time-points were available, utilised the follow-up time-point representing weight restoration or discharge from care. Insufficient studies were available to investigate longer-term follow-ups. Forest plots for the analyses can be seen in Figures S29-31 and full results of the analyses can be seen in Table 4. As sensitivity analyses, studies were removed from analyses where either ≤10% weight gain occurred, where participants didn’t reach at least 80% ideal body weight, or where the follow-up group mean BMI was ≤17kg/m 2 , depending on what data was reported in the study. There were insufficient studies to explore moderators or conduct subgroup analyses. Table 4. Summary of comparative outcomes and heterogeneity for longitudinal meta-analyses in AN samples. Cytokine ( n studies) Sample N (T0, T1) SMD 95% CI z p Heterogeneity Egger’s test TNF-α (n=9) 267, 240 -0.05 a -0.22, 0.13 -0.50 0.617 Tau 2 = 0.00 I 2 = 0.00% z = 0.64 p = 0.524 IL-6 (n=8) 279, 239 0.21 c 0.01, 0.42 2.03 0.042* Tau 2 = 0.02 I 2 = 22.06% z = 2.66 p = 0.008** IL-1β (n=5) 125, 104 0.003 b -0.37, 0.37 0.01 0.988 Tau 2 = 0.07 I 2 = 38.61% z = -1.48 p = 0.139 Notes. a When studies with insufficient weight gain removed SMD = –0.04 (TNF-α); b When studies with insufficient weight gain removed SMD = -0.10 (IL-1β); c When studies with insufficient weight gain removed SMD = 0.19 (IL-6). * Significant at the p <0.01 threshold; ** Significant at the p <0.005 threshold. Abbreviations: CI = confidence intervals; IL = interleukin; p = p -value; SMD = standardized mean difference; T0 = baseline time-point; T1 = follow-up time-point; TNF-α = tumour necrosis factor-alpha; z = z-score. 3.4.1. Tumour necrosis factor-α (TNF-α) The meta-analysis of TNF-α included 267 AN participants at baseline and 240 at follow-up, finding no significant difference between time-points (SMD = -0.05, 95% CI -0.22, 0.13; p = 0.617 I 2 = 0%; Table 4; Figure S29). The trim and fill procedure imputed two missing studies (Figure S23), increasing the SMD to –0.07 (95% CI –0.24, 0.10). The estimated SMD remained similar when removing three studies (28, 65, 70) where weight increase was insufficient (SMD = –0.04; 95% CI –0.23, 0.15; z = - 0.43 ; p = 0.670). 3.4.2. Interleukin-6 (IL-6) Across 8 studies with 279 AN participants at baseline and 239 participants at follow-up collectively, concentrations of IL-6 were significantly higher at baseline than follow-up with a small effect size (SMD = 0.21; 95% CI 0.01, 0.42; p = 0.042; I 2 = 22%; Table 4; Figure S30). The Egger’s test for small study effects was significant (Table 4) and the trim and fill procedure identified three missing studies (Figure S24). When imputing these studies, the re-estimated SMD lowered to 0.10 and was non-significant (95% CI –0.13, 0.32). Additionally, when removing two studies where the BMI at follow-up was insufficient (28, 70), the SMD reduced to 0.19 and again was non-significant (95% CI –0.05, 0.43; z = 1.52; p = 0.129). 3.4.3. Interleukin-1β Five studies with 125 AN participants at baseline and 104 participants at follow-up were available for the meta-analysis of IL-1β concentrations. There was no significant difference between the baseline and follow-up time-points in IL-1β concentrations (SMD = 0.003; 95% CI -0.37, 0.37; p = 0.988; I 2 = 39%; Table 4; Figure S31). After removing two studies where there was insufficient weight increase (28, 70), the SMD increased to –0.10 but remained non-significant (95% CI –0.68, 0.48; z = -0.34; p = 0.733). Discussion This updated systematic review examined cytokine concentrations in people with AN and people with BN compared with controls, and in AN across multiple time-points (largely, after weight restoration treatment). A total of 53 studies were included, 43 of which were entered into meta-analyses with a combined total of 2,533 participants. Cross-sectionally, in people with AN, concentrations of IL-6 and TNF-α were significantly higher with a small effect size, although differences in TNF-α became smaller and non-significant when removing study outliers. These analyses did not differ according to AN subtype. Cumulative meta-analysis plots visually indicated that higher quality and more recently published studies produced smaller effect sizes. Concentrations of IL-15 were moderately higher in AN, and concentrations of IL-7 were moderately lower, compared with controls. Several other cytokines (IL-1β, IL-4, IL-8, IL-10, IFN-γ, MCP and TGF-β) effects were not significantly different between AN and controls. For the cross-sectional analyses of IL-6, TNF-α, and IL-1β, machine learning was used and did not identify any consistently important moderator. Longitudinally, meta-analyses were performed to examine differences in concentrations of IL-6, TNF-α and IL-1β from baseline to follow-up in AN, finding significant but small decreases over time only for IL-6. In BN, sufficient data were available for IL-6 and TNF-α, whereby TNF-α was significantly higher than controls with a large effect size, although this effect size substantially decreased and became non-significant when removing a study outlier. The findings of this updated review broadly align with and extend the results of Dalton, Bartholdy (6). However, they also indicate the absence of a systemic inflammatory profile in AN, given that the majority of pro-inflammatory cytokines did not show elevations, and meta-analyses of other proteins that are indicative of inflammation, such as C-reactive protein, show decreases in AN (71). Nevertheless, it is apparent that concentrations of IL-6 may be slightly elevated in the acute stages of AN and decrease longitudinally with weight restoration. Interestingly, the pleiotropic functions of IL-6 include both pro-inflammatory and anti-inflammatory effects. Secreted by macrophages in response to pathogen-associated molecular patterns (PAMPs), it is known to mediate fever and the acute phase immune response, but also has inhibitory effects on TNF (72, 73). Relevant to AN, IL-6 has also been found to stimulate energy mobilization (74) and as a myokine is elevated up to one hundred times basal rate during exercise in response to muscle contraction (75). Therefore, the factors contributing to increased IL-6 in AN could be manifold both as a result of the behavioural symptoms associated with AN and the adaptation of the body to starvation. For the first time, concentrations of IL-15 and IL-7 showed moderately-sized increases and decreases in the acute stages of AN, respectively. Furthermore, one study included in this meta-analysis in adolescents with AN found elevated IL-15 at baseline, discharge (weight restoration) and after a 1-year follow-up (11). IL-15 induces the proliferation of natural killer cells in the innate immune system (76) but has both pro- and anti-inflammatory effects depending on its expression and site of action. It has been suggested that increases in IL-15 have a pathophysiological role in AN, specifically pertaining to its anabolic role in maintaining muscle mass in the case of acute illness and starvation (11). This adaptation hypothesis would however not explain the elevated concentrations seen after weight restoration and after long-term follow-up in the aforementioned study (11). Although, it could be related to exercise levels as IL-15 is upregulated after acute and chronic exercise (77). IL-15 also modulates serotonergic transmission (78) and synaptic GABAergic transmission in the hippocampus, impairing short- and long-term episodic memory (79), and thus may be involved in the mood and memory disturbances often seen in AN. IL-7 has essential roles in providing survival and growth signals for different immune cells, and low levels can produce immune deficiencies due to lymphopenia (80), which is seen in AN and is thought to be a result of malnutrition (81, 82). Indeed, reduced levels of IL-7 are not observed in constitutional thinness where individuals are underweight without a change in eating behaviour (83), and in AN, a longitudinal study has found increases in IL-7 with weight gain (29). It is notable that in the seven years since the previous meta-analysis was published (28), only two new studies published in BN samples were eligible for inclusion in the meta-analyses, despite calls for more such studies in this population and in people with BED. The results including these new studies were broadly aligned with the previous meta-analysis, when removing one study that constituted a clear outlier. Overall, the pro-inflammatory cytokines TNF-α and IL-6 were similar between people with BN and controls. Moreover, one additional study found a decrease in the anti-inflammatory cytokine IL-10 in BN compared to BMI-matched controls (66). In BED, there were insufficient studies to perform meta-analyses. A single study found a decrease in IL-10 in people with BED, although this was also found in non-BED individuals with obesity (27). In another study, TNF-α was found to be elevated in BED compared to non-BED controls with a similar BMI (84), although IL-6 (84) and IL-1ß (27) were similar between BED and controls. Therefore, decreases in IL-10 in BN and increases in TNF-α in BED may be BMI-independent findings in these populations, although more research is necessary. 4.1. Strengths and limitations This study was a well-controlled and pre-registered systematic review and meta-analysis of cytokine concentrations including the most studies of any meta-analysis published thus far. Authors were contacted where data were unavailable in publications, enabling the inclusion of eight additional studies. We were able to expand on the previous meta-analysis both by (a) including new studies of cytokines that have previously been meta-analysed; (b) meta-analysing several new cytokines that have not been previously examined; (c) examining changes in pro-inflammatory cytokines longitudinally in people with AN; and (d) employing various approaches, including machine learning, to examine potential sources of heterogeneity. For transparency, these moderator analyses employing machine learning were decided for after pre-registration. There are several limitations associated with the study design and conduct of the study. Firstly, because of the nature of cytokine data, data distributions across samples often showed skewedness, and therefore authors publish median rather than mean averages for their samples. Where this occurred, authors were contacted for mean and standard deviation values for samples to enable inclusion in the meta-analysis. One acknowledgement of this approach is that meta-analyses of cytokines including such studies may be inherently biased, although an advantage of including these studies is that the analyses have increased power. Likewise, removing study outliers may introduce bias into results as it is possible that extreme results may constitute real biological anomalies. Therefore, we opted to report results before and after removing outliers. Finally, the results of the cumulative meta-analyses were interpreted visually, therefore the conclusions drawn from these analyses should be taken with caution. Other limitations relating to the included studies themselves pertain mostly to a lack of data reporting, as aforementioned. Not all studies, for example, reported the fasting status of their sample, or the age range of their sample, meaning that several moderators had missing data. Therefore, the predictive capability of the machine learning approach to moderator identification, and the follow-up meta-regressions, was limited. 4.2. Conclusions and future directions Overall, it can be concluded that increases in IL-6 in the acute stages of AN are more robust than increases in TNF-α, and increases in IL-15 and decreases in IL-7 in AN are a new finding. Concentrations of IL-6 decrease longitudinally in people with AN after some weight recovery. When considering the effects of outliers, other cytokines (IL-1β, IL-4, IL-8, IL-10, IFN-γ, MCP, TGF-β and TNF-α) were not altered in AN. There were no differences between people with BN and healthy controls in IL-6 and TNF-α. Heterogeneity was generally high across analyses, the source of which remains unclear. Notably, not all studies reported key data that enabled more in-depth moderator analyses examining key variables known to influence immune functioning, such as smoking status (19), use of psychotropic medications (18) and fasting status/dietary habits (85). With this in mind, in Table 5 we have listed recommendations to enable standardised reporting across studies to enable future high-quality and well-powered meta-analyses, individual participant data meta-analyses, and detailed moderator analyses of cytokine and other biological data in ED populations. Table 5. Recommendations for standardised reporting of biological data in people with EDs Methodological consideration Description Open-access data sharing Publish anonymised participant-level data or group-level data open-access according to Open Science principles. Reporting on data distribution Report information (or plots) on the distribution of the data across samples. Reporting age Report the mean and standard deviation age of the samples and subgroups, including specifying the age range. Reporting other variables Report the mean and standard deviation BMI, body fat percentage, illness duration. Additional variables that ideally should be reported include: medication usage (and type), number of smokers in each sample, mean and standard deviation values of psychopathological measures (e.g., the EDE-Q for ED symptoms, DASS for depression and anxiety symptoms), alcohol use (or exclusion based on alcohol misuse). Fasting status Ideally ensure that blood samples are collected in a standardised manner, for example after an overnight fast. If fasting has not occurred, report this transparently. Immunological status Consider removing participants who have had recent infection or who have a comorbid inflammatory/autoimmune disorder. If included, report data separately for these individuals in the supplementary materials. AN subtype Report data separately according to AN subtype (e.g., in the supplementary materials). Menstrual cycle Report menstrual phase for female participants if possible, and/or levels of hormones such as estradiol and progesterone. Exercise Consider including self-report or objective measures of exercise. We reiterate from our previous publication (6) that studies in ED diagnoses under-represented in this field of research, such as people with BN, BED, ARFID, and atypical AN, are needed. Studies should also carefully consider eligibility criteria for the research to minimise the confounding influence of comorbid illness (e.g., recent infection or inflammatory/autoimmune conditions) or lifestyle factors (e.g., smoking, alcohol misuse), as well as methodological factors, and transparently report key variables in their samples (Table 5). As a final remark, alterations in the concentration of other immune molecules such as immunoglobulins and C-reactive protein have also been suggested to be associated with AN and other EDs. For example, immunoglobulin G (IgG) seems to play a role as α-melanocyte stimulating hormone (α-MSH)-binding protein in the blood, and its levels have been found to be low in patients with AN at hospital admission but to increase alongside weight recovery (86). Alterations in IgG autoantibodies directed to ghrelin and leptin might also be involved in the development of EDs (87). Additionally, immunoglobulin A (IgA) concentrations have been found elevated in the saliva of patients with AN (88). We already mentioned earlier that C-reactive protein shows decreases in AN (71). Thus, for a comprehensive understanding of changes in immune molecules within patients with EDs, the complex relationships between cytokines, IgG, IgA, C-reactive protein and other molecular and cellular immune markers should be considered. Principal component analyses or machine learning approaches might help identifying subtypes with shared immunological profiles. The identification of such subtypes in future research might be more meaningful for the development of individually tailored biological treatments for people with EDs than a comparison of mean cytokine levels. Declarations Acknowledgements: This study represents independent research part funded by the NIHR Maudsley Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. Author contributions: JLK contributed to the conceptualization and management of the study, the database searches, the data extraction, management and curation, the formal analysis, the interpretation of results, the writing of the protocol and manuscript, and the manuscript submission; CB contributed to the database searches, the data extraction, management and curation, the interpretation of results, and the writing of the manuscript; HH, JT, BC and US contributed to the review and editing of the final manuscript; BD contributed to the conceptualization and supervision of the study, the data curation, the interpretation of results and the review and editing of the final manuscript. Competing interests: The authors declare no competing interests. Data availability statement: The authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files. References Hedman A, Breithaupt L, Hübel C, Thornton LM, Tillander A, Norring C, et al. Bidirectional relationship between eating disorders and autoimmune diseases. Journal of Child Psychology and Psychiatry. 2019;60(7):803-12. Himmerich H, Patsalos O, Lichtblau N, Ibrahim MA, Dalton B. Cytokine research in depression: principles, challenges, and open questions. Front Psychiatry. 2019;10:423189. Banks WA, Kastin AJ, Broadwell RD. Passage of cytokines across the blood-brain barrier. Neuroimmunomodulation. 1995;2(4):241-8. Wong S, Pinkney J. Role of cytokines in regulating feeding behaviour. Current Drug Targets. 2004;5(3):251-63. Solmi M, Veronese N, Favaro A, Santonastaso P, Manzato E, Sergi G, et al. Inflammatory cytokines and anorexia nervosa: A meta-analysis of cross-sectional and longitudinal studies. Psychoneuroendocrinology. 2015;51:237-52. Dalton B, Bartholdy S, Robinson L, Solmi M, Ibrahim MA, Breen G, et al. A meta-analysis of cytokine concentrations in eating disorders. J Psychiatr Res. 2018;103:252-64. Dani C, Tarchi L, Cassioli E, Rossi E, Merola GP, Ficola A, et al. A transdiagnostic and diagnostic-specific approach on inflammatory biomarkers in eating disorders: A meta-analysis and systematic review. Psychiatry Res. 2024;340:116115. Brooks S, Dahl K, Dudley-Jones R, Schiöth H. A neuroinflammatory compulsivity model of anorexia nervosa (NICAN). Neuroscience & Biobehavioral Reviews. 2024;159:105580. Rantala MJ, Luoto S, Krama T, Krams I. Eating disorders: an evolutionary psychoneuroimmunological approach. Frontiers in psychology. 2019;10:2200. Breithaupt L, Holsen LM, Ji CN, Hu J, Petterway F, Rosa-Caldwell M, et al. Identification of State Markers in Anorexia Nervosa: Replication and Extension of Inflammation-Associated Biomarkers Using Multiplex Profiling. Biol Psychiatr-Glob Open Sci. 2024;4(5):11. Käver L, Voelz C, Specht HE, Thelen AC, Keller L, Dahmen B, et al. Cytokine and Microbiome Changes in Adolescents with Anorexia Nervosa at Admission, Discharge, and One-Year Follow-Up. Nutrients. 2024;16(11). Keeler JL, Patsalos O, Chung R, Schmidt U, Breen G, Treasure J, et al. Reduced MIP-1β as a Trait Marker and Reduced IL-7 and IL-12 as State Markers of Anorexia Nervosa. J Pers Med. 2021;11(8). Nilsson IAK, Millischer V, Göteson A, Hübel C, Thornton LM, Bulik CM, et al. Aberrant inflammatory profile in acute but not recovered anorexia nervosa. Brain Behav Immun. 2020;88:718-24. Specht HE, Mannig N, Belheouane M, Andreani NA, Tenbrock K, Biemann R, et al. Lower serum levels of IL-1β and IL-6 cytokines in adolescents with anorexia nervosa and their association with gut microbiota in a longitudinal study. Front Psychiatry. 2022;13:920665. Dugué B, Leppänen E, Gräsbeck R. Preanalytical factors and the measurement of cytokines in human subjects. International Journal of Clinical and Laboratory Research. 1996;26:99-105. Goebel MU, Mills PJ, Irwin MR, Ziegler MG. Interleukin-6 and tumor necrosis factor-α production after acute psychological stress, exercise, and infused isoproterenol: differential effects and pathways. Biopsychosocial Science and Medicine. 2000;62(4):591-8. Haack M, Hinze-Selch D, Fenzel T, Kraus T, Kühn M, Schuld A, et al. Plasma levels of cytokines and soluble cytokine receptors in psychiatric patients upon hospital admission: effects of confounding factors and diagnosis. J Psychiatr Res. 1999;33(5):407-18. Munzer A, Sack U, Mergl R, Schönherr J, Petersein C, Bartsch S, et al. Impact of antidepressants on cytokine production of depressed patients in vitro. Toxins. 2013;5(11):2227-40. Rom O, Avezov K, Aizenbud D, Reznick AZ. Cigarette smoking and inflammation revisited. Respiratory physiology & neurobiology. 2013;187(1):5-10. Zhang Y, Wang J, Ye Y, Zou Y, Chen W, Wang Z, et al. Peripheral cytokine levels across psychiatric disorders: A systematic review and network meta-analysis. Progress in Neuro-Psychopharmacology and Biological Psychiatry. 2023;125:110740. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. bmj. 2021;372. Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2000. American Psychiatric Association. Washington, DC: American Psychiatric Association. 1990. American Psychiatric Association. Diagnostic and statistical manual of mental disorders DSM-5. 5th ed. Washington, DC: American Psychiatric Association; 2013. World Health Organization. The ICD-10 classification of mental and behavioural disorders: clinical descriptions and diagnostic guidelines: World Health Organization; 1992. World Health Organization. ICD-11: International classification of diseases: World Health Organization; 2022. Caroleo M, Carbone EA, Greco M, Corigliano DM, Arcidiacono B, Fazia G, et al. Brain-Behavior-Immune Interaction: Serum Cytokines and Growth Factors in Patients with Eating Disorders at Extremes of the Body Mass Index (BMI) Spectrum. Nutrients. 2019;11(9). Dalton B, Campbell IC, Chung R, Breen G, Schmidt U, Himmerich H. Inflammatory Markers in Anorexia Nervosa: An Exploratory Study. Nutrients. 2018;10(11). Dalton B, Leppanen J, Campbell IC, Chung R, Breen G, Schmidt U, et al. A longitudinal analysis of cytokines in anorexia nervosa. Brain Behav Immun. 2020;85:88-95. Roubalova R, Prochazkova P, Dvorak J, Hill M, Papezova H, Kreisinger J, et al. Altered Serum Immunological and Biochemical Parameters and Microbiota Composition in Patients With AN During Realimentation. Front Nutr. 2021;8:680870. Ruiz-Guerrero F, Gomez Del Barrio A, de la Torre-Luque A, Ayad-Ahmed W, Beato-Fernandez L, Polo Montes F, et al. Oxidative stress and inflammatory pathways in female eating disorders and borderline personality disorders with emotional dysregulation as linking factors with impulsivity and trauma. Psychoneuroendocrinology. 2023;158:106383. Tyszkiewicz-Nwafor M, Jowik K, Paszynska E, Dutkiewicz A, Słopien A, Dmitrzak-Weglarz M. Expression of immune-related proteins and their association with neuropeptides in adolescent patients with anorexia nervosa. Neuropeptides. 2022;91:102214. StataCorp. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC; 2023. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials. 1986;7(3):177-88. Duval S, Tweedie R. Trim and fill: a simple funnel-plot–based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455-63. van Lissa CJ. Package ‘metaforest’. 2024. Agnello E, Malfi G, Costantino A, Massarenti P, Pugliese M, Fortunati N, et al. Tumour necrosis factor alpha and oxidative stress as maintaining factors in the evolution of anorexia nervosa. Eating and Weight Disorders-Studies on Anorexia, Bulimia and Obesity. 2012;17:e194-e9. Ahrén-Moonga J, Lekander M, Von Blixen N, Rönnelid J, Holmgren S, af Klinteberg B. Levels of tumour necrosis factor-alpha and interleukin-6 in severely ill patients with eating disorders. Neuropsychobiology. 2010;63(1):8-14. Allende, Corell, Manzanares, Madruga, Marcos, Madroño, et al. Immunodeficiency associated with anorexia nervosa is secondary and improves after refeeding. Immunology. 1998;94(4):543-51. Amerio A, Escelsior A, Martino E, Strangio A, Giacomini C, Montagna E, et al. Dysfunction of Inflammatory Pathways and Their Relationship with Anti-Hypothalamic Autoantibodies in Patients with Anorexia Nervosa. Nutrients. 2023;15(9). Brambilla F, Bellodi L, Brunetta M, Perna G. Plasma concentrations of interleukin-1β, interleukin-6 and tumor necrosis factor-α in anorexia and bulimia nervosa. Psychoneuroendocrinology. 1998;23(5):439-47. Brambilla F, Monti D, Franceschi C. Plasma concentrations of interleukin-1-beta, interleukin-6 and tumor necrosis factor-alpha, and of their soluble receptors and receptor antagonist in anorexia nervosa. Psychiatry Res. 2001;103(2-3):107-14. Campanile AA, Eckel LA, Keel PK. Elevated interleukin-6 in women with binge-eating spectrum disorders. Int J Eat Disord. 2024;57(7):1510-7. Caso JR, Graell M, Navalón A, MacDowell KS, Gutiérrez S, Soto M, et al. Dysfunction of inflammatory pathways in adolescent female patients with anorexia nervosa. Prog Neuropsychopharmacol Biol Psychiatry. 2020;96:109727. Corcos M, Guilbaud O, Chaouat G, Cayol V, Speranza M, Chambry J, et al. Cytokines and anorexia nervosa. Biopsychosocial Science and Medicine. 2001;63(3):502-4. Dolezalova R, Lacinova Z, Dolinkova M, Kleiblova P, Haluzikova D, Housa D, et al. Changes of endocrine function of adipose tissue in anorexia nervosa: comparison of circulating levels versus subcutaneous mRNA expression. Clin Endocrinol. 2007;67(5):674-8. Elegido A, Gheorghe A, Sepúlveda AR, Andrés P, Díaz-Prieto LE, Graell M, et al. Adipokines, cortisol and cytokine alterations in recent onset anorexia nervosa. A case-control study. Endocrinol Diabetes Nutr (Engl Ed). 2019;66(9):571-8. Gabriel T, Massoubre C, Hanachi M, Doré J, Lambert C, Germain N, et al. Association of gut-specific non-inflammatory T lymphocytes with chronic anorexia nervosa and constitutional thinness. Eur Eat Disord Rev. 2023;31(1):76-86. Jiskra J, Haluzik M, Svobodova J, Haluzikova D, Nedvidkova J, Parizkova J, et al. Serum leptin levels and soluble leptin receptors in female patients with anorexia nervosa. Casopis Lekaru Ceskych. 2000;139(21):660-3. Karczewska‐Kupczewska M, Kowalska I, Nikolajuk A, Adamska A, Otziomek E, Gorska M, et al. Hyperinsulinemia acutely increases serum macrophage inhibitory cytokine‐1 concentration in anorexia nervosa and obesity. Clin Endocrinol. 2012;76(1):46-50. Karczewska‐Kupczewska M, Adamska A, Nikołajuk A, Otziomek E, Górska M, Kowalska I, et al. Circulating interleukin 6 and soluble forms of its receptors in relation to resting energy expenditure in women with anorexia nervosa. Clin Endocrinol. 2013;79(6):812-6. Keeler JL, Bahnsen K, Wronski ML, Bernardoni F, Tam F, Arold D, et al. Longitudinal changes in brain-derived neurotrophic factor (BDNF) but not cytokines contribute to hippocampal recovery in anorexia nervosa above increases in body mass index. Psychol Med. 2024;54(9):2242-53. Krizova J, Papezova H, Haluzikova D, Parizkova J, Jiskra J, Kotrlikova E, et al. Soluble leptin receptor levels in patients with anorexia nervosa. Endocrine research. 2002;28(3):199-205. Křížová J, Dolinková M, Lacinová Z, Sulek S, Doležalová R, Housová J, et al. Adiponectin and resistin gene polymorphisms in patients with anorexia nervosa and obesity and its influence on metabolic phenotype. Physiol Res. 2008;57(4):539-46. Misra M, Miller KK, Tsai P, Stewart V, End A, Freed N, et al. Uncoupling of cardiovascular risk markers in adolescent girls with anorexia nervosa. The Journal of pediatrics. 2006;149(6):763-9. e1. Nagata T, Yamada H, Iketani T, Kiriike N. Relationship between plasma concentrations of cytokines, ratio of CD4 and CD8, lymphocyte proliferative responses, and depressive and anxiety state in bulimia nervosa. Journal of psychosomatic research. 2006;60(1):99-103. Nakai Y, Hamagaki S, Takagi R, Taniguchi A, Kurimoto F. Plasma concentrations of tumor necrosis factor-α (TNF-α) and soluble TNF receptors in patients with anorexia nervosa. The Journal of Clinical Endocrinology & Metabolism. 1999;84(4):1226-8. Nakai Y, Hamagaki S, Takagi R, Taniguchi A, Kurimoto F. Plasma concentrations of tumor necrosis factor‐α (TNF‐α) and soluble TNF receptors in patients with bulimia nervosa. Clin Endocrinol. 2000;53(3):383-8. Nogueira J, Maraninchi M, Lorec A, Corroller AB, Nicolay A, Gaudart J, et al. Specific adipocytokines profiles in patients with hyperactive and/or binge/purge form of anorexia nervosa. Eur J Clin Nutr. 2010;64(8):840-4. Ostrowska Z, Ziora K, Oświęcimska J, Marek B, Świętochowska E, Kajdaniuk D, et al. Selected pro-inflammatory cytokines, bone metabolism, osteoprotegerin, and receptor activator of nuclear factor-kB ligand in girls with anorexia nervosa. Endokrynol Pol. 2015;66(4):313-21. Ostrowska Z, Ziora K, Oświęcimska J, Świętochowska E, Marek B, Kajdaniuk D, et al. TGF-β1, bone metabolism, osteoprotegerin, and soluble receptor activator of nuclear factor-kB ligand in girls with anorexia nervosa. Endokrynol Pol. 2016;67(5):493-500. Pomeroy C, Eckert E, Hu S, Eiken B, Mentink M, Crosby RD, et al. Role of interleukin-6 and transforming growth factor-β in anorexia nervosa. Biological psychiatry. 1994;36(12):836-9. Roczniak W, Mikołajczak-Będkowska A, Świętochowska E, Ostrowska Z, Ziora K, Balcerowicz S, et al. Serum interleukin 15 in anorexia nervosa: Comparison to normal weight and obese girls. World J Biol Psychiatry. 2020;21(3):203-11. Ruiz Guerrero F, González Gómez J, Benito Gonzalez P, García García J, Berja Miguel A, Calcedo Giraldo G, et al. Low levels of proinflammatory cytokines in a transdiagnostic sample of young male and female early onset eating disorders without any previous treatment: A case control study. Psychiatry Res. 2022;310:114449. Shimizu T, Satoh Y, Kaneko N, Suzuki M, Lee T, Tanaka K, et al. Factors involved in the regulation of plasma leptin levels in children and adolescents with anorexia nervosa. Pediatrics international. 2005;47(2):154-8. Tabasi M, Anbara T, Siadat SD, Kheirvari Khezerloo J, Elyasinia F, Bayanolhagh S, et al. Socio-demographic Characteristics, Biochemical and Cytokine Levels in Bulimia Nervosa Candidates for Sleeve Gastrectomy. Arch Iran Med. 2020;23(1):23-30. Terra X, Auguet T, Agüera Z, Quesada IM, Orellana‐Gavaldà JM, Aguilar C, et al. Adipocytokine levels in women with anorexia nervosa. Relationship with weight restoration and disease duration. Int J Eating Disord. 2013;46(8):855-61. Vaisman N, Hahn T, Karov Y, Sigler E, Barak Y, Barak V. Changes in cytokine production and impaired hematopoiesis in patients with anorexia nervosa: the effect of refeeding. Cytokine. 2004;26(6):255-61. Víctor VM, Rovira‐Llopis S, Saiz‐Alarcón V, Sangüesa MC, Rojo‐Bofill L, Bañuls C, et al. Involvement of leucocyte/endothelial cell interactions in anorexia nervosa. European journal of clinical investigation. 2015;45(7):670-8. Yasuhara D, Hashiguchi T, Kawahara K, Nakahara T, Harada T, Taguchi H, et al. High mobility group box 1 and refeeding-resistance in anorexia nervosa. Mol Psychiatr. 2007;12(11):976-7. Xu C, Mutwalli H, Haslam R, Keeler JL, Treasure J, Himmerich H. C-reactive protein (CRP) levels in people with eating disorders: A systematic review and meta-analysis. J Psychiatr Res. 2024. Schindler R, Mancilla J, Endres S, Ghorbani R, Clark SC, Dinarello CA. Correlations and interactions in the production of interleukin-6 (IL-6), IL-1, and tumor necrosis factor (TNF) in human blood mononuclear cells: IL-6 suppresses IL-1 and TNF. 1990. Matthys P, Mitera T, Heremans H, Van Damme J, Billiau A. Anti-gamma interferon and anti-interleukin-6 antibodies affect staphylococcal enterotoxin B-induced weight loss, hypoglycemia, and cytokine release in D-galactosamine-sensitized and unsensitized mice. Infection and Immunity. 1995;63(4):1158-64. Kistner TM, Pedersen BK, Lieberman DE. Interleukin 6 as an energy allocator in muscle tissue. Nat Metab. 2022;4(2):170-9. Muñoz‐Cánoves P, Scheele C, Pedersen BK, Serrano AL. Interleukin‐6 myokine signaling in skeletal muscle: a double‐edged sword? The FEBS journal. 2013;280(17):4131-48. Carson WE, Giri JG, Lindemann M, Linett ML, Ahdieh M, Paxton R, et al. Interleukin (IL) 15 is a novel cytokine that activates human natural killer cells via components of the IL-2 receptor. The Journal of experimental medicine. 1994;180(4):1395-403. Khalafi M, Maleki AH, Symonds ME, Sakhaei MH, Rosenkranz SK, Ehsanifar M, et al. Interleukin-15 responses to acute and chronic exercise in adults: a systematic review and meta-analysis. Front Immunol. 2024;14:1288537. Wu X, Hsuchou H, Kastin AJ, He Y, Khan RS, Stone KP, et al. Interleukin-15 affects serotonin system and exerts antidepressive effects through IL15Rα receptor. Psychoneuroendocrinology. 2011;36(2):266-78. Di Castro MA, Garofalo S, Mormino A, Carbonari L, Di Pietro E, De Felice E, et al. Interleukin-15 alters hippocampal synaptic transmission and impairs episodic memory formation in mice. Brain, Behavior, and Immunity. 2024;115:652-66. Winer H, Rodrigues GO, Hixon JA, Aiello FB, Hsu TC, Wachter BT, et al. IL-7: comprehensive review. Cytokine. 2022;160:156049. Bowers TK, Eckert E. Leukopenia in anorexia nervosa: Lack of increased risk of infection. Archives of Internal Medicine. 1978;138(10):1520-3. Saito H, Nomura K, Hotta M, Takano K. Malnutrition induces dissociated changes in lymphocyte count and subset proportion in patients with anorexia nervosa. Int J Eating Disord. 2007;40(6):575-9. Germain N, Viltart O, Loyens A, Bruchet C, Nadin K, Wolowczuk I, et al. Interleukin-7 plasma levels in human differentiate anorexia nervosa, constitutional thinness and healthy obesity. PLoS One. 2016;11(9):e0161890. Caldas ND, Braulio VB, Brasil MAA, Furtado VCS, de Carvalho DP, Cotrik EM, et al. Binge eating disorder, frequency of depression, and systemic inflammatory state in individuals with obesity-A cross sectional study. Arch Endocrinol Metab. 2022;66(4):489-97. D'Esposito V, Di Tolla MF, Lecce M, Cavalli F, Libutti M, Misso S, et al. Lifestyle and dietary habits affect plasma levels of specific cytokines in healthy subjects. Front Nutr. 2022;9:913176. Seitz J, Lahaye E, Andreani NA, Thomas B, Takhlidjt S, Chartrel N, et al. Long-term dynamics of serum α-MSH and α-MSH-binding immunoglobulins with a link to gut microbiota composition in patients with anorexia nervosa. Neuroendocrinology. 2024;114(10):907-20. Fetissov SO, Déchelotte P. The putative role of neuropeptide autoantibodies in anorexia nervosa. Current Opinion in Clinical Nutrition & Metabolic Care. 2008;11(4):428-34. Paszynska E, Dmitrzak-Weglarz M, Tyszkiewicz-Nwafor M, Slopien A. Salivary alpha-amylase, secretory IgA and free cortisol as neurobiological components of the stress response in the acute phase of anorexia nervosa. The World Journal of Biological Psychiatry. 2016;17(4):266-73. Table 2 Table 2 is available in the Supplementary Files section. Additional Declarations There is NO Competing Interest. Supplementary Files SupplementarymaterialscytokineMAv1.0.pdf Supplementary materials Table2.docx Cite Share Download PDF Status: Published Journal Publication published 01 Oct, 2025 Read the published version in Communications Medicine → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6547856","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":454909360,"identity":"1f52dcdd-7845-4bb0-89ac-208bd05cfe40","order_by":0,"name":"Johanna Keeler","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABBUlEQVRIie2RMUvDQBiG3xDQ5YLrhYP0L1wIBKShv0UI1KWLm+BgIHBZTlzrP3FwuPCBXUpdhSy6uFpxCdTBK+2gwzWODvdMH/fx8L4fB3g8/5CgYruB29kAhcTx7kH+RYFVphLhgAL8VmhYCZub9vPifoK4qY35eHjKTmoE6x6UOYvpVSnuliUEezxr529dzglhrEG5U5nPpIhUiITPJDHTFdIWE7ZhcUDJNpG6RjJ6X9OXWW2VcDOg5DaFIDgDwZjcKkfbFHcxvczHkVqwWE9lq02ZcQrUqZbnzvPTRmddpK4SvqDXl95M0tumpuf+cpxWLmW/YD+Tq4MfOXKvPB6Px7PnGyWGUIvU2E1DAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-3905-9857","institution":"King's College London","correspondingAuthor":true,"prefix":"","firstName":"Johanna","middleName":"","lastName":"Keeler","suffix":""},{"id":454909361,"identity":"62fa142f-a1b4-4838-a8ed-71854ec82843","order_by":1,"name":"Charlotte Bovenberg","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Charlotte","middleName":"","lastName":"Bovenberg","suffix":""},{"id":454909362,"identity":"524a7aba-b8da-4cfa-a746-3e06911df409","order_by":2,"name":"Hubertus Himmerich","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Hubertus","middleName":"","lastName":"Himmerich","suffix":""},{"id":454909363,"identity":"c7669a8c-5d3b-4fad-a399-03b07e818176","order_by":3,"name":"Janet Treasure","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Janet","middleName":"","lastName":"Treasure","suffix":""},{"id":454909364,"identity":"86745665-137f-4330-8981-8cf7af0c522f","order_by":4,"name":"Ben Carter","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ben","middleName":"","lastName":"Carter","suffix":""},{"id":454909365,"identity":"e319a236-fbb7-47f9-b17d-d414398ddb8c","order_by":5,"name":"Ulrike Schmidt","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Ulrike","middleName":"","lastName":"Schmidt","suffix":""},{"id":454909366,"identity":"1392e718-37ed-4363-b1bd-4527d07321f7","order_by":6,"name":"Bethan Dalton","email":"","orcid":"","institution":"King's College London","correspondingAuthor":false,"prefix":"","firstName":"Bethan","middleName":"","lastName":"Dalton","suffix":""}],"badges":[],"createdAt":"2025-04-28 12:25:41","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6547856/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6547856/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s43856-025-01122-z","type":"published","date":"2025-10-01T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":82797162,"identity":"ac79baa6-998a-4c49-ad8b-222990f15cf3","added_by":"auto","created_at":"2025-05-15 10:42:08","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":412299,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA flow diagram. Abbreviations: ED = eating disorder; HC = healthy control; ICD = International Classification of Diseases; DSM = Diagnostic and Statistical Manual of Mental Disorders\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/89a4fe5fa1b2881473195cf1.png"},{"id":82797161,"identity":"40049b46-1459-4fb9-b39e-5d13c9df1ff8","added_by":"auto","created_at":"2025-05-15 10:42:08","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":366907,"visible":true,"origin":"","legend":"\u003cp\u003eForest plot of standardized mean difference in TNF-α between AN participants and HCs from n = 30 studies (AN n = 872, HC n = 764). Zero is the line of no effect, and studies to the right of zero indicate an elevation of TNF-α in AN compared to HCs. When removing outliers (12, 32, 40, 60, 69) the SMD reduced to 0.06 and became non-significant (\u003cem\u003ep\u003c/em\u003e = 0.475). Abbreviations: SMD = standardized mean difference; CI = confidence intervals.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/f819fe98448077fb1fcffa67.png"},{"id":82798448,"identity":"c9117058-7e2b-4f3b-b699-74ccc24a5e79","added_by":"auto","created_at":"2025-05-15 10:50:08","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":347625,"visible":true,"origin":"","legend":"\u003cp\u003eForest plot of standardized mean difference in IL-6 between AN participants and HCs from n = 25 studies (AN n = 822, HC n = 774). Zero is the line of no effect, and points to the right of zero indicate an elevation of IL-6 in AN compared to HCs. When removing outliers (11, 12, 32, 40, 60) the SMD reduced to 0.32 but remained significant (\u003cem\u003ep\u003c/em\u003e = 0.001). Abbreviations: SMD = standardized mean difference; CI = confidence intervals.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/20130880117b256ea20371d4.png"},{"id":82798449,"identity":"77d529c7-c35e-4f1d-9de2-7d558be49c38","added_by":"auto","created_at":"2025-05-15 10:50:08","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":238219,"visible":true,"origin":"","legend":"\u003cp\u003eForest plot of standardized mean difference in IL-1β between AN participants and HCs from n = 14 studies (AN n = 354, HC n = 224). Zero is the line of no effect, and points to the right of zero indicate an elevation of IL-6 in AN compared to HCs. When removing outliers (40, 60, 64) the SMD reduced to 0.23 and remained non-significant (\u003cem\u003ep\u003c/em\u003e = 0.095). Abbreviations: SMD = standardized mean difference; CI = confidence intervals.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/0a2708c21daa6b1acb7c14f2.png"},{"id":82799899,"identity":"03ea1803-151f-4b66-9f46-aa6b8ff78455","added_by":"auto","created_at":"2025-05-15 11:06:08","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":227321,"visible":true,"origin":"","legend":"\u003cp\u003eForest plot of standardized mean difference in IL-6 and TNF-α between BN participants and HCs. Zero is the line of no effect, and points to the right of zero indicate an elevation of the cytokine in BN compared to HCs. When removing a study outlier (66), effect sizes reduced to 0.30 and 0.72, respectively; both analyses were then non-significant. Abbreviations: CI = confidence intervals; IL = interleukin; SMD = standardized mean difference; TNF-α = tumour necrosis factor-alpha.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/b2daeedc7be1d40e9ce14cb2.png"},{"id":92637766,"identity":"5b0a377e-b45d-4580-b97e-e425fdb4373c","added_by":"auto","created_at":"2025-10-02 07:07:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2553782,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/fbcfe7b3-37ef-4ec9-a1bd-1c6710b662a1.pdf"},{"id":82797175,"identity":"c035a016-ac65-472e-a528-2c86fd4553cf","added_by":"auto","created_at":"2025-05-15 10:42:08","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":8105974,"visible":true,"origin":"","legend":"Supplementary materials","description":"","filename":"SupplementarymaterialscytokineMAv1.0.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/092a6f8225203fb5b92d965e.pdf"},{"id":82798447,"identity":"5f48b20b-6cf1-46b5-84b2-6864d5de22c6","added_by":"auto","created_at":"2025-05-15 10:50:08","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":124223,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-6547856/v1/904cb67a635f2e97a12045b1.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"A comprehensive updated cross-sectional and longitudinal meta-analysis of cytokines in eating disorders","fulltext":[{"header":"Introduction","content":"\u003cp\u003eEating disorders (EDs) including anorexia nervosa (AN), bulimia nervosa (BN), binge eating disorder (BED), avoidant/restrictive food intake disorder (ARFID) and other specified feeding and eating disorder (OSFED), involve alterations in feeding and eating patterns and in some cases compensatory behaviours as an attempt to prevent weight gain or offset calorie intake. This group of disorders results in secondary consequences that impact physical systems within the body. One of these is the immune system; both the food restriction seen in EDs such as AN and patterns of overeating seen in binge-type EDs may lead to changes in the production and functioning of immune cells. Additionally, recent research has implicated the presence of autoimmune conditions as a risk factor for the development of an ED, and vice versa (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003ePrior research has examined peripheral concentrations of cytokines in ED populations as indicators of changes in immune signalling. Cytokines are signalling proteins produced by immune cells, endothelial and epithelial cells, adipocytes and connective tissue, which can aid communication between cells in order to modulate the immune response. For example, by stimulating or slowing down the immune system or stimulating the movement of cells toward sites of infection or inflammation. They can have systemic as well as local effects and are pleiotropic in that a single cytokine can produce multiple biological functions. Common groups of cytokines include chemokines, which induce chemotaxis (cell migration); interleukins, which act as chemical signals between white blood cells; interferons, which help the body resist viral infections and cancers; and the tumor necrosis factor (TNF) family, which have a wide range of effects on immune functioning, including in tissue modeling and remodeling, and neuronal development. They are often categorised as pro-inflammatory or anti-inflammatory (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e), although as aforementioned most cytokines exhibit pleiotropy and therefore may have either effects depending on factors such as the site of expression and/or cell target. Blood-borne cytokines are able to pass the blood-brain barrier to enter cerebrospinal fluid and interstitial fluid spaces of the brain (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e), although they are also produced in the brain by various cells such as neurons, astrocytes and microglia. Cytokines have been implicated in eating disorders due to their effects on appetite, body weight and eating behaviours, which may be mediated by their effects on brain areas such as the hypothalamus (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThree meta-analyses have quantified differences in concentrations of cytokines between ED populations and healthy controls (\u003cspan additionalcitationids=\"CR6\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e), one of which examined only AN samples (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). When comparing AN to healthy controls, these studies have found small-to-moderate sized increases in levels of the pro-inflammatory cytokines IL-1\u0026szlig;, IL-6 and TNF-α (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e) and TNF-receptor-II (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) in AN. In a meta-analysis of inflammatory markers in EDs as a broader group, it has also been found that levels of TNF-α, and IL-1β are elevated in comparison with healthy control individuals (HCs) (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). This has led to propositions that increases in pro-inflammatory cytokines may be a factor implicated in the aetiology of AN (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e) and neuroinflammation is often cited as a possible aetiological factor in EDs (e.g., (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)). However, this is not without debate, as several recent studies in both adolescents and adults with AN have failed to fully replicate these results and sources of this heterogeneity (moderators) are thus far unclear (\u003cspan additionalcitationids=\"CR11 CR12 CR13\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis heterogeneity may be explained by differences in sample characteristics (e.g., age, illness duration) and study factors (e.g., controlling or not controlling for factors such as age, tobacco use, use of pharmacological medications) (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Most of the studies included in the 2015 and 2018 meta-analyses (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e) did not control for confounding clinical and lifestyle factors that may affect cytokine concentrations, such as age, menstruation, smoking status, medication usage, recent food intake, exercise, body fat, recent illness and concurrent medical (particularly inflammatory) or mental health diagnoses (\u003cspan additionalcitationids=\"CR16 CR17 CR18 CR19\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). Since our last meta-analysis in 2018 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e), several papers have been published that control for, or consider, such variables, which have had contrasting findings. Therefore, it is conceivable that the inclusion of newer studies controlling for such influences, or examining the effect of study quality, may modulate the overall effect sizes.\u003c/p\u003e \u003cp\u003eA prior meta-analysis found no changes after weight gain (BMI\u0026thinsp;\u0026ge;\u0026thinsp;17.5kg/m\u003csup\u003e2\u003c/sup\u003e) in IL-6, TNF-α or IL-1β in people with AN followed longitudinally across three studies (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). However, several studies with longitudinal designs over the course of weight restoration have since been published, enabling analyses with improved power. In conditions such as AN, it is possible that concentrations of cytokines are modified as dietary intake is reinstated (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e), and it has also been suggested that reductions in neuroinflammation may contribute to weight gain (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study aims to build upon a prior meta-analysis published by our group (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e), asking the following additional research questions:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eIncluding new evidence, do cytokine concentrations differ between people diagnosed with an ED and healthy individuals?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eUsing machine learning approaches, are there any moderators that explain the observed heterogeneity between studies?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eHow do between-group effect sizes vary as a function of study quality or publication date?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eIn ED participants, do cytokine concentrations change as a function of weight gain and/or symptom improvement?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e"},{"header":"Methods and materials","content":"\u003cp\u003eThe reporting of this systematic review was guided by the standards of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) Statement (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e) and a PRISMA checklist can be found in the Supplementary Materials. The quality assessment of the studies was conducted according to a modified version of the Newcastle-Ottowa Scale (NOS; (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e)). The full quality appraisal of the studies can be found in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThe study protocol was prospectively registered on the Open Science Framework, which can be found at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/g6d3f\u003c/span\u003e\u003cspan address=\"https://osf.io/g6d3f\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Literature search\u003c/h2\u003e \u003cp\u003eTwo reviewers (JLK and CB) independently searched four electronic databases (PubMed, Web of Science Core Collection, MEDLINE and PsycINFO) from the date of the last review, 4th May 2018, until 10th November 2024. An identical search strategy from the previous review was replicated for this review, which can be found at Dalton et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). In short, searches included the following keywords, mapped to Medical Subject Heading with the Explode function where possible: eating disorder*, anorexia nervosa, bulimi* or binge eat* in combination with cytokine* chemokine*, inflammat*, interleukin, interferon, IFN, tumor necrosis factor, TNF, transforming growth factor or TGF. Searches were supplemented using the ascendancy approach (hand-searches of reference lists of relevant papers and reviews) and the descendancy approach (citation tracking in Google Scholar, Crossref).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Eligibility criteria\u003c/h2\u003e \u003cp\u003eStudies in any language of any study design that assessed cytokine concentrations in the serum, plasma or cerebrospinal fluid (CSF) of individuals with a Diagnostic and Statistical Manual of Mental Disorders (DSM) (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e) or International Statistical Classification of Diseases (ICD) (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e) diagnosis of ED were eligible for inclusion. Studies were included if they reported cross-sectional comparisons of cytokine concentrations between an ED group and HCs, or longitudinal assessments at a minimum of two time-points, with BMI or ED symptoms assessed at both time points.\u003c/p\u003e \u003cp\u003eStudies were excluded if: i) they did not report group comparisons or longitudinal measurements of cytokine concentrations; ii) participants had an organic cause for their disordered eating e.g., cancer, immunological conditions, genetic disorder, etc.; iii) the sample was comprised of animals; iv) they measured cytokine production or genetic expression but did not assess cytokine concentrations; or v) there was significant reported or suspected study overlap between publications, defined as the same patient and or control group. Review articles, meta-analyses, conference proceedings/abstracts, book chapters, and unpublished theses were also not included.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Source selection\u003c/h2\u003e \u003cp\u003eTitles and abstracts of publications were imported into Endnote, where duplicates were removed. During an initial screening stage, titles and abstracts were assessed against basic eligibility criteria and those deemed likely to be irrelevant (e.g., studies in animal samples, studies of other medical or psychiatric conditions, studies with ineligible study designs such as reviews) were discarded. Full-texts of the remaining articles were then assessed against the full eligibility criteria, with the reasons for study exclusion documented (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The search process was conducted independently by two reviewers (JLK and CB).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Data extraction\u003c/h2\u003e \u003cp\u003eTwo reviewers (JLK and CB) conducted the data extraction process and the extracted data was checked by one reviewer (JLK). At this stage, the references of the eligible publications were uploaded into an electronic summary table used in the previous study (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e) and databases were combined for extraction. Extracted data included publication identifiers (study title, year, author list); sample characteristics, including sample size, demographics (e.g. mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD age), diagnostic criteria and clinical characteristics (e.g. mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD illness duration, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD BMI), and medication status; parameters of interest, measurement methods, and concentrations of cytokines (mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD), time interval between measurements (longitudinal studies, in weeks or days), content of intervention between measurements (longitudinal studies); potential confounds and control for confounds including presence/absence of smoking as an inclusion criteria, number of participants taking psychotropic medications (and types); main and additional study outcomes. Authors of potentially eligible publications were contacted where data were not reported (n\u0026thinsp;=\u0026thinsp;13). In total, eight of these authors replied and provided data (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan additionalcitationids=\"CR28 CR29 CR30 CR31\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFor studies that reported AN subtype data separately, the means and standard deviations were pooled for the AN meta-analyses. Standard errors were converted to standard deviations using the formula SD = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sqrt{N}\\:x\\:SE\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003ewhere N represents the sample size. Where there were fewer than four studies for a given cytokine or raw data could not be obtained from authors, results were narratively synthesised.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.7. Synthesis of results\u003c/h2\u003e \u003cp\u003eWhere studies were clinically homogeneous they were considered for pooling in a meta-analysis. Meta-analyses were performed for each cytokine where four or more studies were available for each ED separately.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6. Statistical analysis\u003c/h2\u003e \u003cp\u003eIndividual meta-analyses for each cytokine were conducted in Stata 18.0 (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e) using the \u0026lsquo;metan\u0026rsquo; command. The main outcome measure was cytokine concentrations (pg/ml or ng/ml). Random effects-models using the DerSimonian \u0026amp; Laird method (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e) were used for all meta-analyses to pool the standardised mean difference (SMD) of the studies relative to the study sample size. The SMD of each study reflects the size of effect between the comparator groups in each study (i.e., ED vs. HC, or ED baseline vs. follow-up) relative to the variability observed in the study. All longitudinal meta-analyses utilised the two available time-points, or where multiple time-points were available, utilised the follow-up time-point representing weight restoration or discharge from care. Statistical significance of the overall SMD was ascertained according to the \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 threshold.\u003c/p\u003e \u003cp\u003eOutliers were identified when the confidence intervals (CI) of the individual study SMD did not overlap with the CI of the pooled effect (i.e., where the upper CI bound of the individual study effect is smaller than the lower CI bound of the pooled effect, or the lower CI bound of the individual study effect is bigger than the upper CI bound of the pooled effect). Sensitivity analyses removing identified outliers were conducted and SMDs were re-estimated.\u003c/p\u003e \u003cp\u003ePublication bias was explored using the Egger\u0026rsquo;s test for small study effects and funnel plots. The Duval and Tweedie trim and fill method (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e) was used to identify smaller studies causing funnel plot asymmetry, adjust for this asymmetry, and re-estimate SMDs after adjusting for missing studies.\u003c/p\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.6.1. Cumulative meta-analysis\u003c/h2\u003e \u003cp\u003eTo explore how the effect sizes evolved over time for the cross-sectional comparisons of the main pro-inflammatory cytokines previously synthesised (IL-6, TNF-α and IL-1β) between AN and HC, cumulative meta-analyses were performed using NOS study quality total score, and year of publication as ordering variables. This enables an examination of how the effect size changes when adding new studies, starting from a) the newest studies to the oldest studies, and b) the highest quality studies to the lowest quality studies. Cumulative meta-analyses were displayed as forest plots (Figures \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e-4) and were interpreted by visual inspection.\u003c/p\u003e\u003cp\u003e\u003cem\u003e2.6.2. Between-study heterogeneity\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eBetween-study heterogeneity was assessed with the Higgins I\u003csup\u003e2\u003c/sup\u003e statistic, based on the Cochrane\u0026rsquo;s Q (chi-square) statistic, which represents the proportion of total variation across studies that is attributable to heterogeneity rather than chance. Heterogeneity was considered to be substantial if above 50%, and considerable if above 75%. Tau\u003csup\u003e2\u003c/sup\u003e, which estimates the amount of true variability between the SMDs of the included studies, was also reported.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSubgroup analyses were conducted with AN subtypes, where the study reported data for each subtype (i.e., AN-restricting type [-R] and AN-binge-eating/purging type [-BP]) individually. To further explore sources of heterogeneity, a random-effects MetaForest analysis was conducted in R version 4.4.2 (R Core Team, 2021). MetaForest is a machine-learning based exploratory approach to meta-analysis that mitigates overfitting and assesses the relative importance of potential moderators (36) (see SM section 6 for further details). Analyses were only performed where there were 10 or more studies available. Pre-specified theoretical and methodological moderators are listed in Table 1. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 1. Pre-specified moderators coded for meta-analysis\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eModerator\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCodes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eRegion of study origin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eAsia\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eEurope\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eNorth America\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eYear\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eNOS total score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eAge group\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eAdolescents\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eAdults\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eMixed\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eSample fasting status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eFasted\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eNot fasted/not reported\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eType of blood sample\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eSerum\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003ePlasma\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eUse of psychotropic medication in the ED sample (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003ePercentage of smokers in sample (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eMean age of ED group (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eMean age of HC group (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eMean BMI of ED group (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eMean BMI of HC group (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 312px;\"\u003e\n \u003cp\u003eMean duration of illness in ED group (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAbbreviations: BMI = body mass index; ED = eating disorder; HC = healthy controls; NOS = Newcastle-Ottowa Scale.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e3.1. Characteristics of included studies\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAn additional 18 studies not included in the previous meta-analysis (6) were identified, resulting in a total of 43 studies from 11 countries included in the meta-analysis. Of these, 40 were included in the cross-sectional meta-analyses comparing cytokine concentrations between AN and HCs, seven in the cross-sectional meta-analyses comparing BN with HCs, and 11 in the meta-analyses examining cytokine concentrations longitudinally in AN. Table 2 details the study and sample characteristics for studies included in the meta-analyses. Findings of studies not included in the meta-analyses, due to insufficient data reporting or insufficient number of studies for meta-analysis (including cross-sectional comparisons between BED and HCs and recovered AN and HCs), are summarized in Table S3 and in the supplementary materials.\u003c/p\u003e\n\u003cp\u003eCytokines included in the cross-sectional meta-analyses between AN and HC were TNF-\u0026alpha;, IL-1\u0026beta;, IL-6, IL-4, IL-7, IL-8, IL-10, IL-15, IFN-\u0026gamma;, TGF-\u0026beta; and MCP. Cytokines included in the cross-sectional meta-analyses between BN and HC were IL-6 and TNF-\u0026alpha;. Longitudinally in AN, cytokines available for meta-analysis included TNF-\u0026alpha;, IL-1\u0026beta;, and IL-6. Other cytokines were also measured in eligible studies, however sufficient data were not available to perform meta-analyses. Across studies, 31 studies measured cytokine concentrations peripherally in serum, 12 in plasma, and none in CSF.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe mean and SD age of participants with EDs and HCs was 21.7\u0026plusmn;8.1 (n = 32 studies) and 21.4\u0026plusmn;6.7 (27 studies), respectively. All studies included only female participants. The mean BMI of the AN participants was 15.7\u0026plusmn;2.3 (n=34 studies), of BN participants was 31.7\u0026plusmn;10.4 (n=4 studies), and of HCs was 21.2\u0026plusmn;3.0kg/m\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e\u0026nbsp;\u003c/sub\u003e(n=29 studies). Diagnosis was based on the DSM-V (n=14), DSM-IV (n=28) or DSM-III (n=1). Mean illness duration for ED participants, reported in 16 studies, was 5.2\u0026plusmn;7.2 years. Medication usage of people in the ED groups was reported in 32 studies, and in 20 studies participants were confirmed medication-free as part of assessment or eligibility procedures. In the remaining 12 studies, medications used included anti-depressants, anti-psychotics, sedatives, neuroleptics and anxiolytics.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;For the longitudinal studies in AN, the mean BMI at baseline was 15.6\u0026plusmn;1.9kg/m\u003csup\u003e2\u003c/sup\u003e (n=8 studies) and at follow-up was 18.4\u0026plusmn;2.0kg/m\u003csup\u003e2\u003c/sup\u003e (n=7 studies). The interval between baseline and follow-up assessments used in the meta-analyses ranged from 2 to 12 months. The intervention used across studies included inpatient weight restoration treatment (n=5), specialist eating disorder treatment (n=1), unspecified weight gain treatment (n=1) and with cognitive behavioural therapy (CBT; n=1), CBT (n=1), and CBT with pharmacological therapy (n=1). In one study, treatment was unspecified.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e3.2. Quality assessment of included studies\u003c/p\u003e\n\u003cp\u003eThe rating of studies depends on three characteristics: the selection of study groups, the comparability of groups and the ascertainment of the outcome. Of the included studies, 26 were good quality, 24 were fair quality and three were poor quality (see SM 1.1 for the full description of the quality assessment). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e3.3. Cross-sectional meta-analysis results\u003c/p\u003e\n\u003cp\u003eThe results of the cross-sectional meta-analyses comparing participants with AN to HC, and participants with BN to HC, are displayed in Table 3.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 3. Summary of comparative outcomes and heterogeneity for all conducted cross-sectional meta-analyses.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCytokine\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(\u003cem\u003en\u0026nbsp;\u003c/em\u003estudies)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSample \u003cem\u003eN\u0026nbsp;\u003c/em\u003e(ED, HC)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSMD\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e95% CI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ez\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHeterogeneity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEgger\u0026rsquo;s test\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"8\" style=\"width: 624px;\"\u003e\n \u003cp\u003e\u003cem\u003eAnorexia nervosa vs healthy controls\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTNF-\u0026alpha;\u0026nbsp;\u003c/strong\u003e(n=30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e872, 764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.27 \u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e0.01, 0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e2.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.040\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.43\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 84.05%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 1.63\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.103\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-6\u0026nbsp;\u003c/strong\u003e(n=25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e822, 774\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.36 \u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e0.11, 0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e2.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.32\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 81.81%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 3.03\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.002**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-1\u0026beta;\u0026nbsp;\u003c/strong\u003e(n=14)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e354, 244\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.53\u003csup\u003e\u0026nbsp;c\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.06, 1.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e1.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.080\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 1.14\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = \u0026nbsp; \u0026nbsp; 90.59%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 0.57\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.570\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-10\u0026nbsp;\u003c/strong\u003e(n=7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e278, 250\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.41 \u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.17, 0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e1.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.166\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.50\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 88.09%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 3.57\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep\u0026nbsp;\u003c/em\u003e\u0026lt; 0.001**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIFN-\u0026gamma;\u0026nbsp;\u003c/strong\u003e (n=6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e171, 146\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.37\u0026nbsp;\u003csup\u003ee\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.41, 1.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.356\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.83\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 89.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 2.60\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.009**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-8\u0026nbsp;\u003c/strong\u003e(n=5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e235, 229\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e-0.16 \u003csup\u003ef\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.78, 0.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e-0.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.626\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = \u0026nbsp;0.44\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 88.83%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = -0.21\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.831\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMCP\u003c/strong\u003e (n=5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e235, 229\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e-0.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.44, 0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e-1.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.135\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.03\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 33.07%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = -2.29\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.022*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTGF-\u0026szlig;\u0026nbsp;\u003c/strong\u003e(n=5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e240, 183\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.61\u003csup\u003e\u0026nbsp;g\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.90, 2.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.431\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 2.75\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 96.76%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 3.39\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.001**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-4\u003c/strong\u003e (n=4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e124, 105\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e-0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.28, 0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e-0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.925\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.00\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 0.00%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 1.29\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.196\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-7\u003c/strong\u003e (n=4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e211, 188\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e-0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-1.07, -0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e-2.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.18\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 75.55%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = -0.55\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.582\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-15\u003c/strong\u003e (n=4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e183, 126\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e0.07, 1.26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e2.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.029*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.30\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 82.74%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 1.96\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.050\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"8\" style=\"width: 624px;\"\u003e\n \u003cp\u003e\u003cem\u003eBulimia nervosa vs healthy controls\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-6\u0026nbsp;\u003c/strong\u003e(n=5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e163, 115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e0.73 \u003csup\u003eh\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e-0.28, 1.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e1.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.157\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 1.21\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 92.76%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = -0.48\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.632\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTNF-\u0026alpha;\u003c/strong\u003e (n=5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79px;\"\u003e\n \u003cp\u003e145, 123\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 58px;\"\u003e\n \u003cp\u003e2.34 \u003csup\u003ei\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 86px;\"\u003e\n \u003cp\u003e0.02, 4.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e1.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.048*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 6.84\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e: 97.88%\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 4.91\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep \u0026lt;\u0026nbsp;\u003c/em\u003e0.001**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eNotes.\u003c/em\u003e \u003csup\u003ea\u003c/sup\u003e When outliers removed SMD = 0.06, \u003cem\u003ep\u003c/em\u003e = 0.475 (TNF-\u0026alpha; AN); \u003csup\u003eb\u003c/sup\u003e When outliers removed SMD = 0.32 (IL-6 AN); \u003csup\u003ec\u003c/sup\u003e When outliers removed SMD = 0.23 (IL-1\u0026beta; AN); \u003csup\u003ed\u003c/sup\u003e When outlier removed SMD = \u0026ndash;0.04 (IL-10 AN); \u003csup\u003ee\u003c/sup\u003e When outlier removed SMD = \u0026ndash;0.11 (IFN-\u0026gamma; AN); \u003csup\u003ef\u003c/sup\u003e When outlier removed SMD = 0.13 (IL-8 AN); \u003csup\u003eg\u0026nbsp;\u003c/sup\u003eWhen outliers removed SMD \u0026ndash;0.15 (TGF-\u0026beta; AN); \u003csup\u003eh\u003c/sup\u003e When outlier removed SMD = 0.30 (IL-6 BN). \u003csup\u003ei\u003c/sup\u003e When outlier removed SMD = 0.06, p = 0.475 (IL-6 TNF-\u0026alpha;).* Significant at the \u003cem\u003ep\u003c/em\u003e\u0026lt;0.01 threshold; ** Significant at the \u003cem\u003ep\u003c/em\u003e\u0026lt;0.005 threshold. \u0026nbsp;Abbreviations: CI = confidence intervals; ED = eating disorder; HC = healthy control; IFN-\u0026gamma; = interferon-gamma; IL = interleukin; MCP = monocyte chemoattractant protein; \u003cem\u003ep\u003c/em\u003e = \u003cem\u003ep\u003c/em\u003e-value; SMD = standardized mean difference; TNF-\u0026alpha; = tumour necrosis factor-alpha; z = z-score.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e3.3.1. Anorexia nervosa\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3.3.1.1. Tumour-necrosis factor-\u0026alpha; (TNF-\u0026alpha;)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAcross 30 studies with a total of 1636 participants (AN n = 872; HC n = 764), six found elevated levels of TNF-\u003cem\u003e\u0026alpha;\u0026nbsp;\u003c/em\u003eand none reported reduced levels (Figure 2). Across all studies there was a small difference in TNF-\u0026alpha; concentrations between AN and HCs whereby AN had elevated levels (SMD = 0.27; 95% CI 0.01, 0.54; \u003cem\u003ep\u0026nbsp;\u003c/em\u003e= 0.040). Heterogeneity estimates were high (\u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e=84%; Table 3) and although there was no evidence of publication bias from the Egger\u0026rsquo;s test, six outliers were identified (12, 32, 40, 57, 60, 69). After removing these outliers, the effect size became marginal and non-significant (SMD = 0.06; 95% CI \u0026ndash;0.10, 0.22; z = 0.71; \u003cem\u003ep\u003c/em\u003e = 0.475) and the\u003cem\u003e\u0026nbsp;I\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e heterogeneity estimate reduced to 47.26%.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA total of 15 studies reported AN subtype data for TNF-\u0026alpha;, which indicated no subgroup difference between the AN-R and AN-BP subtypes (\u003cem\u003ep\u0026nbsp;\u003c/em\u003e= 0.690; Figure S25). Again, when including only these studies, neither meta-analyses showed significant differences between the individual subgroups (i.e., AN-R or AN-BP) and HCs.\u003c/p\u003e\n\u003cp\u003eCumulative meta-analyses were conducted using year and NOS total study quality score as sorting variables in a descending fashion. A visual inspection highlighted that newer studies were associated with lower SMDs between AN and HC for TNF-\u0026alpha; (Figure S1), although the pattern for the NOS score was less clear (Figure S2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo further investigate sources of heterogeneity, a random-effects MetaForest analysis was conducted. A random effects MetaForest analysis provided no evidence for associations between the entered moderators and the SMD for TNF-\u0026alpha;, even when removing study outliers (SM 6.1).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3.3.1.2. Interleukin-6 (IL-6)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAcross 25 studies including a total of 1596 participants (AN n = 822, HC n = 774), nine exhibited elevated levels and two exhibited reduced levels of IL-6 in AN (Figure 3). The pooled mean concentrations of IL-6 were significantly higher in AN compared to HC with a small effect (SMD = 0.36; 95% CI 0.11, 0.61; \u003cem\u003ep\u003c/em\u003e = 0.005; Table 3; Figure 2). This analysis showed high heterogeneity (\u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e = 82%) and the Egger\u0026rsquo;s test for small study effects was significant. Five studies were identified as outliers following an inspection of the 95% CIs (11, 12, 32, 40, 60). When removing these studies the SMD was slightly lower but remained significant (SMD=0.32; \u003cem\u003ep\u0026nbsp;\u003c/em\u003e= 0.001); the I\u003csup\u003e2\u003c/sup\u003e statistic lowered to 65.1% and the Egger\u0026rsquo;s test was no longer significant (z = 1.98, \u003cem\u003ep\u003c/em\u003e = 0.050).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA total of 15 studies reported AN subtype data, which indicated no subgroup difference between the AN-R and AN-BP subtypes (\u003cem\u003ep\u0026nbsp;\u003c/em\u003e= 0.790; Figure S26). With this fewer number of included studies, neither meta-analysis showed significant differences between the individual subgroups (i.e., AN-R or AN-BP) and HCs.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;In cumulative meta-analyses, both forest plots indicated that higher quality studies and newer studies had overall lower SMDs between AN and HC (Figures S3 and S4).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAfter removing study outliers, a random effects MetaForest analysis identified the following moderators as the most important in explaining the effect size: study year, NOS score, study region, age group, percentage of sample using medication, mean age, BMI, and illness duration of the AN sample, and mean age and BMI of the HC sample (relative variable importance shown in Figure S34). These moderators were entered into meta-regressions (Table S2). Only study country was significant, with studies conducted in North America showing a greater difference between AN and HC compared to studies in Europe (B = 0.78; SE = 0.29; \u003cem\u003ez\u003c/em\u003e = 2.72; \u003cem\u003ep\u003c/em\u003e = 0.007; 95% CI 0.22, 1.34).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3.3.1.3. Interleukin-1\u0026beta; (IL-1\u0026beta;)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAcross 14 studies with a collective sample size of 598 participants (AN n = 354, HC n = 244), there was a non-significant moderate-sized increase in pooled concentrations of IL-1\u0026beta; between AN and HC groups (SMD = 0.53, 95% CI -0.06, 1.12; \u003cem\u003ep\u003c/em\u003e = 0.080; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 91%; Table 3; Figure 4). After exploring the heterogeneity three studies were identified as outliers(40, 60, 64). The effect size was smaller and remained non-significant after removing these studies (SMD = 0.23; 95% CI \u0026ndash;0.04, 0.50; z = 1.67; \u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.095; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 44%.\u003c/p\u003e\n\u003cp\u003eSubgroup analyses according to AN subtype were not possible, as few studies reported data per subtype. A random effects MetaForest analysis provided no evidence for associations between the entered moderators and the SMD for IL-1b\u0026nbsp;(SM 6.3).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3.3.1.4. Other cytokines available for meta-analysis\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eEight additional cytokines had at least four studies available and thus were meta-analysed (IL-4, IL-7, IL-8, IL-10, IL-15, IFN-\u0026gamma;, MCP and TGF-\u0026beta;; Figures S27 and S28, Table 3). Only concentrations of IL-15 and IL-7 were significantly different between AN and HC, whereby IL-15 was higher in the AN group with a moderate effect size (SMD = 0.67; \u0026nbsp;95% CI 0.07, 1.26; \u003cem\u003ep \u0026nbsp;=\u0026nbsp;\u003c/em\u003e0.029; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 83%) and IL-7 was lower in AN with a moderate effect size (SMD = -0.58, 95% CI -1.07, -0.09; \u003cem\u003ep\u003c/em\u003e = 0.020; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 76%). Concentrations of IL-4, IL-8, IL-10, IFN-\u0026gamma;, MCP and TGF-\u0026beta; were not significant between AN and HCs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOne study was identified as an outlier in the meta-analyses of IL-10 and IFN-\u0026gamma; (27), and the removal of this study reduced the SMDs to \u0026ndash;0.04 (95% CI \u0026ndash;0.28, 0.20; \u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.761; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 32%) and \u0026ndash;0.11 (95% CI \u0026ndash;0.51, 0.28; \u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.571; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 59%), respectively. One outlier was identified in the IL-8 meta-analysis (32), and the removal of this study increased the SMD to 0.13\u0026nbsp;(95% CI \u0026ndash;0.40, 0.67; \u003cem\u003ep\u003c/em\u003e = 0.626). Two studies\u0026nbsp;(40, 60)\u0026nbsp;were identified as outliers in the TGF-\u0026beta; meta-analysis, and the removal of these lowered the SMD to \u0026ndash;0.14 (95% CI \u0026ndash;0.11, 0.84; \u003cem\u003ep\u003c/em\u003e = 0.783).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAccording to I\u003csup\u003e2\u003c/sup\u003e, estimated heterogeneity was high in the meta-analyses of IL-7, IL-8, IL-10, IL-15, IFN-\u0026gamma; and TGF-\u0026beta;. The trim and fill method resulted in two imputed studies in both the IL-4 and MCP analyses (Figures S21 and S22); adjusting for these imputed studies altered the SMDs to \u0026ndash;0.09 and \u0026ndash;0.10, respectively (both still non-significant). Funnel plots for meta-analyses are included in the supplementary materials (Figures S8-15).\u003c/p\u003e\n\u003cp\u003e3.3.2. Bulimia nervosa\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3.3.2.1. Interleukin-6 (IL-6)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe meta-analyses of the studies in BN populations indicated higher concentrations of IL-6 in participants with BN compared with HCs, although this was not significant (Table 3; Figure 5).One study (66) was deemed an outlier. Removing this study from the IL-6 meta-analysis reduced the pooled SMD to 0.30 (95% CI -0.25, 0.86; \u003cem\u003ep\u003c/em\u003e = 0.287; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 65%).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e3.3.2.2. Tumour necrosis factor-\u0026alpha; (TNF-\u0026alpha;)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThere were significantly higher concentrations of TNF-\u0026alpha; in participants with BN compared with HCs \u0026nbsp;(Table 3; Figure 5). The heterogeneity estimates were considerably high for both analyses. The funnel plot indicated asymmetry (Figure S17) which was confirmed by a significant Egger\u0026rsquo;s test for small study effects (z = 4.91, \u003cem\u003ep\u0026nbsp;\u003c/em\u003e\u0026lt; 0.001). One study (66) was deemed an outlier in and removing this study reduced the overall effect size (SMD = 0.72) rendering the difference between groups non-significant (95% CI -0.14, 1.58; \u003cem\u003ep\u003c/em\u003e = 0.101) and slightly reducing the I\u003csup\u003e2\u003c/sup\u003e heterogeneity statistic to 84%. After removing this study the Egger\u0026rsquo;s test was also no longer significant (z = 1.30, \u003cem\u003ep\u003c/em\u003e = 0.192).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e3.4. Longitudinal meta-analyses in AN\u003c/p\u003e\n\u003cp\u003eThere were sufficient data to meta-analyse differences between baseline and follow-up in individuals with AN undergoing weight restoration for the cytokines IL-6, IL-1\u0026beta; and TNF-\u0026alpha;. All meta-analyses utilised the two available time-points, or where multiple time-points were available, utilised the follow-up time-point representing weight restoration or discharge from care. Insufficient studies were available to investigate longer-term follow-ups. Forest plots for the analyses can be seen in Figures S29-31 and full results of the analyses can be seen in Table 4.\u003c/p\u003e\n\u003cp\u003eAs sensitivity analyses, studies were removed from analyses where either \u0026le;10% weight gain occurred, where participants didn\u0026rsquo;t reach at least 80% ideal body weight, or where the follow-up group mean BMI was \u0026le;17kg/m\u003csup\u003e2\u003c/sup\u003e\u003csub\u003e,\u0026nbsp;\u003c/sub\u003edepending on what data was reported in the study. There were insufficient studies to explore moderators or conduct subgroup analyses.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Table 4. Summary of comparative outcomes and heterogeneity for longitudinal meta-analyses in AN samples. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCytokine\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(\u003cem\u003en\u0026nbsp;\u003c/em\u003estudies)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSample \u003cem\u003eN\u0026nbsp;\u003c/em\u003e(T0, T1)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 59px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSMD\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 87px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e95% CI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ez\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHeterogeneity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEgger\u0026rsquo;s test\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTNF-\u0026alpha;\u0026nbsp;\u003c/strong\u003e(n=9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e267, 240\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 59px;\"\u003e\n \u003cp\u003e-0.05\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 87px;\"\u003e\n \u003cp\u003e-0.22, 0.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e-0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.617\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.00\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 0.00%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 0.64\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.524\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-6\u0026nbsp;\u003c/strong\u003e(n=8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e279, 239\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 59px;\"\u003e\n \u003cp\u003e0.21\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 87px;\"\u003e\n \u003cp\u003e0.01, 0.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e2.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.042*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = \u0026nbsp; \u0026nbsp; 0.02\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 22.06%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = 2.66\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.008**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIL-1\u0026beta;\u0026nbsp;\u003c/strong\u003e(n=5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e125, 104\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 59px;\"\u003e\n \u003cp\u003e0.003\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 87px;\"\u003e\n \u003cp\u003e-0.37, 0.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 50px;\"\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.988\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eTau\u003csup\u003e2\u003c/sup\u003e = 0.07\u003c/p\u003e\n \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e = 38.61%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 88px;\"\u003e\n \u003cp\u003ez = -1.48\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ep =\u0026nbsp;\u003c/em\u003e0.139\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eNotes.\u003c/em\u003e \u003csup\u003ea\u003c/sup\u003e When studies with insufficient weight gain removed SMD = \u0026ndash;0.04 (TNF-\u0026alpha;); \u003csup\u003eb\u003c/sup\u003e When studies with insufficient weight gain removed SMD = -0.10 (IL-1\u0026beta;); \u003csup\u003ec\u003c/sup\u003e When studies with insufficient weight gain removed SMD = 0.19 (IL-6). * Significant at the \u003cem\u003ep\u003c/em\u003e\u0026lt;0.01 threshold; ** Significant at the \u003cem\u003ep\u003c/em\u003e\u0026lt;0.005 threshold. \u0026nbsp;Abbreviations: CI = confidence intervals; IL = interleukin; \u003cem\u003ep\u003c/em\u003e = \u003cem\u003ep\u003c/em\u003e-value; SMD = standardized mean difference; T0 = baseline time-point; T1 = follow-up time-point; TNF-\u0026alpha; = tumour necrosis factor-alpha; z = z-score. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e3.4.1. Tumour necrosis factor-\u0026alpha; (TNF-\u0026alpha;)\u003c/p\u003e\n\u003cp\u003eThe meta-analysis of TNF-\u0026alpha; included 267 AN participants at baseline and 240 at follow-up, finding no significant difference between time-points (SMD = -0.05, 95% CI -0.22, 0.13; \u003cem\u003ep\u003c/em\u003e = 0.617 \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 0%; Table 4; Figure S29). The trim and fill procedure imputed two missing studies (Figure S23), increasing the SMD to \u0026ndash;0.07 (95% CI \u0026ndash;0.24, 0.10). The estimated SMD remained similar when removing three studies (28, 65, 70) where weight increase was insufficient (SMD = \u0026ndash;0.04; 95% CI \u0026ndash;0.23, 0.15; \u003cem\u003ez = -\u003c/em\u003e0.43\u003cem\u003e; p\u0026nbsp;\u003c/em\u003e= 0.670).\u003c/p\u003e\n\u003cp\u003e3.4.2. Interleukin-6 (IL-6)\u003c/p\u003e\n\u003cp\u003eAcross 8 studies with 279 AN participants at baseline and 239 participants at follow-up collectively, concentrations of IL-6 were significantly higher at baseline than follow-up with a small effect size (SMD = 0.21; 95% CI 0.01, 0.42; \u003cem\u003ep\u003c/em\u003e = 0.042; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 22%; Table 4; Figure S30). The Egger\u0026rsquo;s test for small study effects was significant (Table 4) and the trim and fill procedure identified three missing studies (Figure S24). When imputing these studies, the re-estimated SMD lowered to 0.10 and was non-significant (95% CI \u0026ndash;0.13, 0.32). Additionally, when removing two studies where the BMI at follow-up was insufficient (28, 70), the SMD reduced to 0.19 and again was non-significant (95% CI \u0026ndash;0.05, 0.43; \u003cem\u003ez\u003c/em\u003e = 1.52; \u003cem\u003ep\u003c/em\u003e\u0026nbsp; = 0.129).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e3.4.3. Interleukin-1\u0026beta;\u003c/p\u003e\n\u003cp\u003eFive studies with 125 AN participants at baseline and 104 participants at follow-up were available for the meta-analysis of IL-1\u0026beta; concentrations. There was no significant difference between the baseline and follow-up time-points in IL-1\u0026beta; concentrations (SMD = 0.003; 95% CI -0.37, 0.37; \u003cem\u003ep\u003c/em\u003e = 0.988; \u003cem\u003eI\u003c/em\u003e\u003csup\u003e2\u0026nbsp;\u003c/sup\u003e= 39%; Table 4; Figure S31). After removing two studies where there was insufficient weight increase (28, 70), the SMD increased to \u0026ndash;0.10 but remained non-significant (95% CI \u0026ndash;0.68, 0.48; \u003cem\u003ez\u003c/em\u003e = -0.34; \u003cem\u003ep\u003c/em\u003e = 0.733).\u0026nbsp;\u003c/p\u003e\n"},{"header":"Discussion","content":"\u003cp\u003eThis updated systematic review examined cytokine concentrations in people with AN and people with BN compared with controls, and in AN across multiple time-points (largely, after weight restoration treatment). A total of 53 studies were included, 43 of which were entered into meta-analyses with a combined total of 2,533 participants. Cross-sectionally, in people with AN, concentrations of IL-6 and TNF-\u0026alpha; were significantly higher with a small effect size, although differences in TNF-\u0026alpha; became smaller and non-significant when removing study outliers. These analyses did not differ according to AN subtype. Cumulative meta-analysis plots visually indicated that higher quality and more recently published studies produced smaller effect sizes. Concentrations of IL-15 were moderately higher in AN, and concentrations of IL-7 were moderately lower, compared with controls. Several other cytokines (IL-1\u0026beta;, IL-4, IL-8, IL-10,\u0026nbsp;IFN-\u0026gamma;, MCP and TGF-\u0026beta;) effects were not significantly different between AN and controls. For the cross-sectional analyses of IL-6, TNF-\u0026alpha;, and IL-1\u0026beta;, machine learning was used and did not identify any consistently important moderator. Longitudinally, meta-analyses were performed to examine differences in concentrations of IL-6, TNF-\u0026alpha; and IL-1\u0026beta; from baseline to follow-up in AN, finding significant but small decreases over time only for IL-6.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn BN, sufficient data were available for IL-6 and TNF-\u0026alpha;, whereby TNF-\u0026alpha; was significantly higher than controls with a large effect size, although this effect size substantially decreased and became non-significant when removing a study outlier.\u003c/p\u003e\n\u003cp\u003eThe findings of this updated review broadly align with and extend the results of Dalton, Bartholdy (6). However, they also indicate the absence of a systemic inflammatory profile in AN, given that the majority of pro-inflammatory cytokines did not show elevations, and meta-analyses of other proteins that are indicative of inflammation, such as C-reactive protein, show decreases in AN (71). Nevertheless, it is apparent that concentrations of IL-6 may be slightly elevated in the acute stages of AN and decrease longitudinally with weight restoration. Interestingly, the pleiotropic functions of IL-6 include both pro-inflammatory and anti-inflammatory effects. Secreted by macrophages in response to pathogen-associated molecular patterns (PAMPs), it is known to mediate fever and the acute phase immune response, but also has inhibitory effects on TNF (72, 73). Relevant to AN, IL-6 has also been found to stimulate energy mobilization (74) and as a myokine is elevated up to one hundred times basal rate during exercise in response to muscle contraction (75). Therefore, the factors contributing to increased IL-6 in AN could be manifold both as a result of the behavioural symptoms associated with AN and the adaptation of the body to starvation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor the first time, concentrations of IL-15 and IL-7 showed moderately-sized increases and decreases in the acute stages of AN, respectively. Furthermore, one study included in this meta-analysis in adolescents with AN found elevated IL-15 at baseline, discharge (weight restoration) and after a 1-year follow-up (11). IL-15 induces the proliferation of natural killer cells in the innate immune system (76) but has both pro- and anti-inflammatory effects depending on its expression and site of action. It has been suggested that increases in IL-15 have a pathophysiological role in AN, specifically pertaining to its anabolic role in maintaining muscle mass in the case of acute illness and starvation (11). This adaptation hypothesis would however not explain the elevated concentrations seen after weight restoration and after long-term follow-up in the aforementioned study (11). Although, it could be related to exercise levels as IL-15 is upregulated after acute and chronic exercise (77). IL-15 also modulates \u0026nbsp;serotonergic transmission (78) and synaptic GABAergic transmission in the hippocampus, impairing short- and long-term episodic memory (79), and thus may be involved in the mood and memory disturbances often seen in AN. IL-7 has essential roles in providing survival and growth signals for different immune cells, and low levels can produce immune deficiencies due to lymphopenia (80), which is seen in AN and is thought to be a result of malnutrition (81, 82). Indeed, reduced levels of IL-7 are not observed in constitutional thinness where individuals are underweight without a change in eating behaviour (83), and in AN, a longitudinal study has found increases in IL-7 with weight gain (29).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIt is notable that in the seven years since the previous meta-analysis was published (28), only two new studies published in BN samples were eligible for inclusion in the meta-analyses, despite calls for more such studies in this population and in people with BED. The results including these new studies were broadly aligned with the previous meta-analysis, when removing one study that constituted a clear outlier. Overall, the pro-inflammatory cytokines TNF-\u0026alpha; and IL-6 were similar between people with BN and controls. Moreover, one additional study found a decrease in the anti-inflammatory cytokine IL-10 in BN compared to BMI-matched controls (66). In BED, there were insufficient studies to perform meta-analyses. A single study found a decrease in IL-10 in people with BED, although this was also found in non-BED individuals with obesity (27). In another study, TNF-\u0026alpha; was found to be elevated in BED compared to non-BED controls with a similar BMI (84), although IL-6 (84) and IL-1\u0026szlig; (27) were similar between BED and controls. Therefore, decreases in IL-10 in BN and increases in TNF-\u0026alpha; in BED may be BMI-independent findings in these populations, although more research is necessary.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e4.1. Strengths and limitations\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThis study was a well-controlled and pre-registered systematic review and meta-analysis of cytokine concentrations including the most studies of any meta-analysis published thus far. Authors were contacted where data were unavailable in publications, enabling the inclusion of eight additional studies. We were able to expand on the previous meta-analysis both by (a) including new studies of cytokines that have previously been meta-analysed; (b) meta-analysing several new cytokines that have not been previously examined; (c) examining changes in pro-inflammatory cytokines longitudinally in people with AN; and (d) employing various approaches, including machine learning, to examine potential sources of heterogeneity. For transparency, these moderator analyses employing machine learning were decided for after pre-registration. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThere are several limitations associated with the study design and conduct of the study. Firstly, because of the nature of cytokine data, data distributions across samples often showed skewedness, and therefore authors publish median rather than mean averages for their samples. Where this occurred, authors were contacted for mean and standard deviation values for samples to enable inclusion in the meta-analysis. One acknowledgement of this approach is that meta-analyses of cytokines including such studies may be inherently biased, although an advantage of including these studies is that the analyses have increased power. Likewise, removing study outliers may introduce bias into results as it is possible that extreme results may constitute real biological anomalies. Therefore, we opted to report results before and after removing outliers. Finally, the results of the cumulative meta-analyses were interpreted visually, therefore the conclusions drawn from these analyses should be taken with caution.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOther limitations relating to the included studies themselves pertain mostly to a lack of data reporting, as aforementioned. Not all studies, for example, reported the fasting status of their sample, or the age range of their sample, meaning that several moderators had missing data. Therefore, the predictive capability of the machine learning approach to moderator identification, and the follow-up meta-regressions, was limited.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e4.2. Conclusions and future directions\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eOverall, it can be concluded that increases in IL-6 in the acute stages of AN are more robust than increases in TNF-\u0026alpha;, and increases in IL-15 and decreases in IL-7 in AN are a new finding.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConcentrations of IL-6 decrease longitudinally in people with AN after some weight recovery. When considering the effects of outliers, other cytokines (IL-1\u0026beta;, IL-4, IL-8, IL-10, IFN-\u0026gamma;, MCP, TGF-\u0026beta; and TNF-\u0026alpha;) were not altered in AN. There were no differences between people with BN and healthy controls in IL-6 and TNF-\u0026alpha;. Heterogeneity was generally high across analyses, the source of which remains unclear. Notably, not all studies reported key data that enabled more in-depth moderator analyses examining key variables known to influence immune functioning, such as smoking status (19), use of psychotropic medications (18) and fasting status/dietary habits (85). With this in mind, in Table 5 we have listed recommendations to enable standardised reporting across studies to enable future high-quality and well-powered meta-analyses, individual participant data meta-analyses, and detailed moderator analyses of cytokine and other biological data in ED populations.\u003c/p\u003e\n\u003cp\u003eTable 5. Recommendations for standardised reporting of biological data in people with EDs\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 227px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMethodological consideration\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescription\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eOpen-access data sharing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003ePublish anonymised participant-level data or group-level data open-access according to Open Science principles.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eReporting on data distribution\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eReport information (or plots) on the distribution of the data across samples.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eReporting age\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eReport the mean and standard deviation age of the samples and subgroups, including specifying the age range.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eReporting other variables\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eReport the mean and standard deviation BMI, body fat percentage, illness duration. Additional variables that ideally should be reported include: medication usage (and type), number of smokers in each sample, mean and standard deviation values of psychopathological measures (e.g., the EDE-Q for ED symptoms, DASS for depression and anxiety symptoms), alcohol use (or exclusion based on alcohol misuse).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eFasting status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eIdeally ensure that blood samples are collected in a standardised manner, for example after an overnight fast. If fasting has not occurred, report this transparently.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eImmunological status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eConsider removing participants who have had recent infection or who have a comorbid inflammatory/autoimmune disorder.\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eIf included, report data separately for these individuals in the supplementary materials.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eAN subtype\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eReport data separately according to AN subtype (e.g., in the supplementary materials).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eMenstrual cycle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eReport menstrual phase for female participants if possible, and/or levels of hormones such as estradiol and progesterone.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 227px;\"\u003e\n \u003cp\u003eExercise\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 397px;\"\u003e\n \u003cp\u003eConsider including self-report or objective measures of exercise.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003eWe reiterate from our previous publication (6) that studies in ED diagnoses under-represented in this field of research, such as people with BN, BED, ARFID, and atypical AN, are needed. Studies should also carefully consider eligibility criteria for the research to minimise the confounding influence of comorbid illness (e.g., recent infection or inflammatory/autoimmune conditions) or lifestyle factors (e.g., smoking, alcohol misuse), as well as methodological factors, and transparently report key variables in their samples (Table 5).\u003cp\u003eAs a final remark, alterations in the concentration of other immune molecules such as immunoglobulins and C-reactive protein have also been suggested to be associated with AN and other EDs. For example, immunoglobulin G (IgG) seems to play a role as \u0026alpha;-melanocyte stimulating hormone (\u0026alpha;-MSH)-binding protein in the blood, and its levels have been found to be low in patients with AN at hospital admission but to increase alongside weight recovery (86). Alterations in IgG autoantibodies directed to ghrelin and leptin might also be involved in the development of EDs (87). Additionally, immunoglobulin A (IgA) concentrations have been found elevated in the saliva of patients with AN (88). We already mentioned earlier that C-reactive protein shows decreases in AN (71). Thus, for a comprehensive understanding of changes in immune molecules within patients with EDs, the complex relationships between cytokines, IgG, IgA, C-reactive protein and other molecular and cellular immune markers should be considered. Principal component analyses or machine learning approaches might help identifying subtypes with shared immunological profiles. The identification of such subtypes in future research might be more meaningful for the development of individually tailored biological treatments for people with EDs than a comparison of mean cytokine levels.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThis study represents independent research part funded by the NIHR Maudsley\u0026nbsp;Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King\u0026rsquo;s College London.\u0026nbsp;The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eAuthor contributions:\u0026nbsp;\u003c/strong\u003eJLK contributed to the conceptualization and management of the study, the database searches, the data extraction, management and curation, the formal analysis, the interpretation of results, the writing of the protocol and manuscript, and the manuscript submission; CB contributed to the database searches, the data extraction, management and curation, the interpretation of results, and the writing of the manuscript; HH, JT, BC and US contributed to the review and editing of the final manuscript; BD contributed to the conceptualization and supervision of the study, the data curation, the interpretation of results and the review and editing of the final manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eCompeting interests:\u0026nbsp;\u003c/strong\u003eThe authors declare no competing interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eData availability statement:\u0026nbsp;\u003c/strong\u003eThe authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eHedman A, Breithaupt L, H\u0026uuml;bel C, Thornton LM, Tillander A, Norring C, et al. Bidirectional relationship between eating disorders and autoimmune diseases. Journal of Child Psychology and Psychiatry. 2019;60(7):803-12.\u003c/li\u003e\n\u003cli\u003eHimmerich H, Patsalos O, Lichtblau N, Ibrahim MA, Dalton B. Cytokine research in depression: principles, challenges, and open questions. Front Psychiatry. 2019;10:423189.\u003c/li\u003e\n\u003cli\u003eBanks WA, Kastin AJ, Broadwell RD. Passage of cytokines across the blood-brain barrier. Neuroimmunomodulation. 1995;2(4):241-8.\u003c/li\u003e\n\u003cli\u003eWong S, Pinkney J. Role of cytokines in regulating feeding behaviour. Current Drug Targets. 2004;5(3):251-63.\u003c/li\u003e\n\u003cli\u003eSolmi M, Veronese N, Favaro A, Santonastaso P, Manzato E, Sergi G, et al. Inflammatory cytokines and anorexia nervosa: A meta-analysis of cross-sectional and longitudinal studies. Psychoneuroendocrinology. 2015;51:237-52.\u003c/li\u003e\n\u003cli\u003eDalton B, Bartholdy S, Robinson L, Solmi M, Ibrahim MA, Breen G, et al. A meta-analysis of cytokine concentrations in eating disorders. J Psychiatr Res. 2018;103:252-64.\u003c/li\u003e\n\u003cli\u003eDani C, Tarchi L, Cassioli E, Rossi E, Merola GP, Ficola A, et al. A transdiagnostic and diagnostic-specific approach on inflammatory biomarkers in eating disorders: A meta-analysis and systematic review. Psychiatry Res. 2024;340:116115.\u003c/li\u003e\n\u003cli\u003eBrooks S, Dahl K, Dudley-Jones R, Schi\u0026ouml;th H. A neuroinflammatory compulsivity model of anorexia nervosa (NICAN). Neuroscience \u0026amp; Biobehavioral Reviews. 2024;159:105580.\u003c/li\u003e\n\u003cli\u003eRantala MJ, Luoto S, Krama T, Krams I. Eating disorders: an evolutionary psychoneuroimmunological approach. Frontiers in psychology. 2019;10:2200.\u003c/li\u003e\n\u003cli\u003eBreithaupt L, Holsen LM, Ji CN, Hu J, Petterway F, Rosa-Caldwell M, et al. Identification of State Markers in Anorexia Nervosa: Replication and Extension of Inflammation-Associated Biomarkers Using Multiplex Profiling. Biol Psychiatr-Glob Open Sci. 2024;4(5):11.\u003c/li\u003e\n\u003cli\u003eK\u0026auml;ver L, Voelz C, Specht HE, Thelen AC, Keller L, Dahmen B, et al. Cytokine and Microbiome Changes in Adolescents with Anorexia Nervosa at Admission, Discharge, and One-Year Follow-Up. Nutrients. 2024;16(11).\u003c/li\u003e\n\u003cli\u003eKeeler JL, Patsalos O, Chung R, Schmidt U, Breen G, Treasure J, et al. Reduced MIP-1\u0026beta; as a Trait Marker and Reduced IL-7 and IL-12 as State Markers of Anorexia Nervosa. J Pers Med. 2021;11(8).\u003c/li\u003e\n\u003cli\u003eNilsson IAK, Millischer V, G\u0026ouml;teson A, H\u0026uuml;bel C, Thornton LM, Bulik CM, et al. Aberrant inflammatory profile in acute but not recovered anorexia nervosa. Brain Behav Immun. 2020;88:718-24.\u003c/li\u003e\n\u003cli\u003eSpecht HE, Mannig N, Belheouane M, Andreani NA, Tenbrock K, Biemann R, et al. Lower serum levels of IL-1\u0026beta; and IL-6 cytokines in adolescents with anorexia nervosa and their association with gut microbiota in a longitudinal study. Front Psychiatry. 2022;13:920665.\u003c/li\u003e\n\u003cli\u003eDugu\u0026eacute; B, Lepp\u0026auml;nen E, Gr\u0026auml;sbeck R. Preanalytical factors and the measurement of cytokines in human subjects. International Journal of Clinical and Laboratory Research. 1996;26:99-105.\u003c/li\u003e\n\u003cli\u003eGoebel MU, Mills PJ, Irwin MR, Ziegler MG. Interleukin-6 and tumor necrosis factor-\u0026alpha; production after acute psychological stress, exercise, and infused isoproterenol: differential effects and pathways. Biopsychosocial Science and Medicine. 2000;62(4):591-8.\u003c/li\u003e\n\u003cli\u003eHaack M, Hinze-Selch D, Fenzel T, Kraus T, K\u0026uuml;hn M, Schuld A, et al. Plasma levels of cytokines and soluble cytokine receptors in psychiatric patients upon hospital admission: effects of confounding factors and diagnosis. J Psychiatr Res. 1999;33(5):407-18.\u003c/li\u003e\n\u003cli\u003eMunzer A, Sack U, Mergl R, Sch\u0026ouml;nherr J, Petersein C, Bartsch S, et al. Impact of antidepressants on cytokine production of depressed patients in vitro. Toxins. 2013;5(11):2227-40.\u003c/li\u003e\n\u003cli\u003eRom O, Avezov K, Aizenbud D, Reznick AZ. Cigarette smoking and inflammation revisited. Respiratory physiology \u0026amp; neurobiology. 2013;187(1):5-10.\u003c/li\u003e\n\u003cli\u003eZhang Y, Wang J, Ye Y, Zou Y, Chen W, Wang Z, et al. Peripheral cytokine levels across psychiatric disorders: A systematic review and network meta-analysis. Progress in Neuro-Psychopharmacology and Biological Psychiatry. 2023;125:110740.\u003c/li\u003e\n\u003cli\u003ePage MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. bmj. 2021;372.\u003c/li\u003e\n\u003cli\u003eWells GA, Shea B, O\u0026rsquo;Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2000.\u003c/li\u003e\n\u003cli\u003eAmerican Psychiatric Association. Washington, DC: American Psychiatric Association. 1990.\u003c/li\u003e\n\u003cli\u003eAmerican Psychiatric Association. Diagnostic and statistical manual of mental disorders DSM-5. 5th ed. Washington, DC: American Psychiatric Association; 2013.\u003c/li\u003e\n\u003cli\u003eWorld Health Organization. The ICD-10 classification of mental and behavioural disorders: clinical descriptions and diagnostic guidelines: World Health Organization; 1992.\u003c/li\u003e\n\u003cli\u003eWorld Health Organization. ICD-11: International classification of diseases: World Health Organization; 2022.\u003c/li\u003e\n\u003cli\u003eCaroleo M, Carbone EA, Greco M, Corigliano DM, Arcidiacono B, Fazia G, et al. Brain-Behavior-Immune Interaction: Serum Cytokines and Growth Factors in Patients with Eating Disorders at Extremes of the Body Mass Index (BMI) Spectrum. Nutrients. 2019;11(9).\u003c/li\u003e\n\u003cli\u003eDalton B, Campbell IC, Chung R, Breen G, Schmidt U, Himmerich H. Inflammatory Markers in Anorexia Nervosa: An Exploratory Study. Nutrients. 2018;10(11).\u003c/li\u003e\n\u003cli\u003eDalton B, Leppanen J, Campbell IC, Chung R, Breen G, Schmidt U, et al. A longitudinal analysis of cytokines in anorexia nervosa. Brain Behav Immun. 2020;85:88-95.\u003c/li\u003e\n\u003cli\u003eRoubalova R, Prochazkova P, Dvorak J, Hill M, Papezova H, Kreisinger J, et al. Altered Serum Immunological and Biochemical Parameters and Microbiota Composition in Patients With AN During Realimentation. Front Nutr. 2021;8:680870.\u003c/li\u003e\n\u003cli\u003eRuiz-Guerrero F, Gomez Del Barrio A, de la Torre-Luque A, Ayad-Ahmed W, Beato-Fernandez L, Polo Montes F, et al. Oxidative stress and inflammatory pathways in female eating disorders and borderline personality disorders with emotional dysregulation as linking factors with impulsivity and trauma. Psychoneuroendocrinology. 2023;158:106383.\u003c/li\u003e\n\u003cli\u003eTyszkiewicz-Nwafor M, Jowik K, Paszynska E, Dutkiewicz A, Słopien A, Dmitrzak-Weglarz M. Expression of immune-related proteins and their association with neuropeptides in adolescent patients with anorexia nervosa. Neuropeptides. 2022;91:102214.\u003c/li\u003e\n\u003cli\u003eStataCorp. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC; 2023.\u003c/li\u003e\n\u003cli\u003eDerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials. 1986;7(3):177-88.\u003c/li\u003e\n\u003cli\u003eDuval S, Tweedie R. Trim and fill: a simple funnel-plot\u0026ndash;based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455-63.\u003c/li\u003e\n\u003cli\u003evan Lissa CJ. Package \u0026lsquo;metaforest\u0026rsquo;. 2024.\u003c/li\u003e\n\u003cli\u003eAgnello E, Malfi G, Costantino A, Massarenti P, Pugliese M, Fortunati N, et al. Tumour necrosis factor alpha and oxidative stress as maintaining factors in the evolution of anorexia nervosa. Eating and Weight Disorders-Studies on Anorexia, Bulimia and Obesity. 2012;17:e194-e9.\u003c/li\u003e\n\u003cli\u003eAhr\u0026eacute;n-Moonga J, Lekander M, Von Blixen N, R\u0026ouml;nnelid J, Holmgren S, af Klinteberg B. Levels of tumour necrosis factor-alpha and interleukin-6 in severely ill patients with eating disorders. Neuropsychobiology. 2010;63(1):8-14.\u003c/li\u003e\n\u003cli\u003eAllende, Corell, Manzanares, Madruga, Marcos, Madro\u0026ntilde;o, et al. Immunodeficiency associated with anorexia nervosa is secondary and improves after refeeding. Immunology. 1998;94(4):543-51.\u003c/li\u003e\n\u003cli\u003eAmerio A, Escelsior A, Martino E, Strangio A, Giacomini C, Montagna E, et al. Dysfunction of Inflammatory Pathways and Their Relationship with Anti-Hypothalamic Autoantibodies in Patients with Anorexia Nervosa. Nutrients. 2023;15(9).\u003c/li\u003e\n\u003cli\u003eBrambilla F, Bellodi L, Brunetta M, Perna G. Plasma concentrations of interleukin-1\u0026beta;, interleukin-6 and tumor necrosis factor-\u0026alpha; in anorexia and bulimia nervosa. Psychoneuroendocrinology. 1998;23(5):439-47.\u003c/li\u003e\n\u003cli\u003eBrambilla F, Monti D, Franceschi C. Plasma concentrations of interleukin-1-beta, interleukin-6 and tumor necrosis factor-alpha, and of their soluble receptors and receptor antagonist in anorexia nervosa. Psychiatry Res. 2001;103(2-3):107-14.\u003c/li\u003e\n\u003cli\u003eCampanile AA, Eckel LA, Keel PK. Elevated interleukin-6 in women with binge-eating spectrum disorders. Int J Eat Disord. 2024;57(7):1510-7.\u003c/li\u003e\n\u003cli\u003eCaso JR, Graell M, Naval\u0026oacute;n A, MacDowell KS, Guti\u0026eacute;rrez S, Soto M, et al. Dysfunction of inflammatory pathways in adolescent female patients with anorexia nervosa. Prog Neuropsychopharmacol Biol Psychiatry. 2020;96:109727.\u003c/li\u003e\n\u003cli\u003eCorcos M, Guilbaud O, Chaouat G, Cayol V, Speranza M, Chambry J, et al. Cytokines and anorexia nervosa. Biopsychosocial Science and Medicine. 2001;63(3):502-4.\u003c/li\u003e\n\u003cli\u003eDolezalova R, Lacinova Z, Dolinkova M, Kleiblova P, Haluzikova D, Housa D, et al. Changes of endocrine function of adipose tissue in anorexia nervosa: comparison of circulating levels versus subcutaneous mRNA expression. Clin Endocrinol. 2007;67(5):674-8.\u003c/li\u003e\n\u003cli\u003eElegido A, Gheorghe A, Sep\u0026uacute;lveda AR, Andr\u0026eacute;s P, D\u0026iacute;az-Prieto LE, Graell M, et al. Adipokines, cortisol and cytokine alterations in recent onset anorexia nervosa. A case-control study. Endocrinol Diabetes Nutr (Engl Ed). 2019;66(9):571-8.\u003c/li\u003e\n\u003cli\u003eGabriel T, Massoubre C, Hanachi M, Dor\u0026eacute; J, Lambert C, Germain N, et al. Association of gut-specific non-inflammatory T lymphocytes with chronic anorexia nervosa and constitutional thinness. Eur Eat Disord Rev. 2023;31(1):76-86.\u003c/li\u003e\n\u003cli\u003eJiskra J, Haluzik M, Svobodova J, Haluzikova D, Nedvidkova J, Parizkova J, et al. Serum leptin levels and soluble leptin receptors in female patients with anorexia nervosa. Casopis Lekaru Ceskych. 2000;139(21):660-3.\u003c/li\u003e\n\u003cli\u003eKarczewska‐Kupczewska M, Kowalska I, Nikolajuk A, Adamska A, Otziomek E, Gorska M, et al. Hyperinsulinemia acutely increases serum macrophage inhibitory cytokine‐1 concentration in anorexia nervosa and obesity. Clin Endocrinol. 2012;76(1):46-50.\u003c/li\u003e\n\u003cli\u003eKarczewska‐Kupczewska M, Adamska A, Nikołajuk A, Otziomek E, G\u0026oacute;rska M, Kowalska I, et al. Circulating interleukin 6 and soluble forms of its receptors in relation to resting energy expenditure in women with anorexia nervosa. Clin Endocrinol. 2013;79(6):812-6.\u003c/li\u003e\n\u003cli\u003eKeeler JL, Bahnsen K, Wronski ML, Bernardoni F, Tam F, Arold D, et al. Longitudinal changes in brain-derived neurotrophic factor (BDNF) but not cytokines contribute to hippocampal recovery in anorexia nervosa above increases in body mass index. Psychol Med. 2024;54(9):2242-53.\u003c/li\u003e\n\u003cli\u003eKrizova J, Papezova H, Haluzikova D, Parizkova J, Jiskra J, Kotrlikova E, et al. Soluble leptin receptor levels in patients with anorexia nervosa. Endocrine research. 2002;28(3):199-205.\u003c/li\u003e\n\u003cli\u003eKř\u0026iacute;žov\u0026aacute; J, Dolinkov\u0026aacute; M, Lacinov\u0026aacute; Z, Sulek S, Doležalov\u0026aacute; R, Housov\u0026aacute; J, et al. Adiponectin and resistin gene polymorphisms in patients with anorexia nervosa and obesity and its influence on metabolic phenotype. Physiol Res. 2008;57(4):539-46.\u003c/li\u003e\n\u003cli\u003eMisra M, Miller KK, Tsai P, Stewart V, End A, Freed N, et al. Uncoupling of cardiovascular risk markers in adolescent girls with anorexia nervosa. The Journal of pediatrics. 2006;149(6):763-9. e1.\u003c/li\u003e\n\u003cli\u003eNagata T, Yamada H, Iketani T, Kiriike N. Relationship between plasma concentrations of cytokines, ratio of CD4 and CD8, lymphocyte proliferative responses, and depressive and anxiety state in bulimia nervosa. Journal of psychosomatic research. 2006;60(1):99-103.\u003c/li\u003e\n\u003cli\u003eNakai Y, Hamagaki S, Takagi R, Taniguchi A, Kurimoto F. Plasma concentrations of tumor necrosis factor-\u0026alpha; (TNF-\u0026alpha;) and soluble TNF receptors in patients with anorexia nervosa. The Journal of Clinical Endocrinology \u0026amp; Metabolism. 1999;84(4):1226-8.\u003c/li\u003e\n\u003cli\u003eNakai Y, Hamagaki S, Takagi R, Taniguchi A, Kurimoto F. Plasma concentrations of tumor necrosis factor‐\u0026alpha; (TNF‐\u0026alpha;) and soluble TNF receptors in patients with bulimia nervosa. Clin Endocrinol. 2000;53(3):383-8.\u003c/li\u003e\n\u003cli\u003eNogueira J, Maraninchi M, Lorec A, Corroller AB, Nicolay A, Gaudart J, et al. Specific adipocytokines profiles in patients with hyperactive and/or binge/purge form of anorexia nervosa. Eur J Clin Nutr. 2010;64(8):840-4.\u003c/li\u003e\n\u003cli\u003eOstrowska Z, Ziora K, Oświęcimska J, Marek B, Świętochowska E, Kajdaniuk D, et al. Selected pro-inflammatory cytokines, bone metabolism, osteoprotegerin, and receptor activator of nuclear factor-kB ligand in girls with anorexia nervosa. Endokrynol Pol. 2015;66(4):313-21.\u003c/li\u003e\n\u003cli\u003eOstrowska Z, Ziora K, Oświęcimska J, Świętochowska E, Marek B, Kajdaniuk D, et al. TGF-\u0026beta;1, bone metabolism, osteoprotegerin, and soluble receptor activator of nuclear factor-kB ligand in girls with anorexia nervosa. Endokrynol Pol. 2016;67(5):493-500.\u003c/li\u003e\n\u003cli\u003ePomeroy C, Eckert E, Hu S, Eiken B, Mentink M, Crosby RD, et al. Role of interleukin-6 and transforming growth factor-\u0026beta; in anorexia nervosa. Biological psychiatry. 1994;36(12):836-9.\u003c/li\u003e\n\u003cli\u003eRoczniak W, Mikołajczak-Będkowska A, Świętochowska E, Ostrowska Z, Ziora K, Balcerowicz S, et al. Serum interleukin 15 in anorexia nervosa: Comparison to normal weight and obese girls. World J Biol Psychiatry. 2020;21(3):203-11.\u003c/li\u003e\n\u003cli\u003eRuiz Guerrero F, Gonz\u0026aacute;lez G\u0026oacute;mez J, Benito Gonzalez P, Garc\u0026iacute;a Garc\u0026iacute;a J, Berja Miguel A, Calcedo Giraldo G, et al. Low levels of proinflammatory cytokines in a transdiagnostic sample of young male and female early onset eating disorders without any previous treatment: A case control study. Psychiatry Res. 2022;310:114449.\u003c/li\u003e\n\u003cli\u003eShimizu T, Satoh Y, Kaneko N, Suzuki M, Lee T, Tanaka K, et al. Factors involved in the regulation of plasma leptin levels in children and adolescents with anorexia nervosa. Pediatrics international. 2005;47(2):154-8.\u003c/li\u003e\n\u003cli\u003eTabasi M, Anbara T, Siadat SD, Kheirvari Khezerloo J, Elyasinia F, Bayanolhagh S, et al. Socio-demographic Characteristics, Biochemical and Cytokine Levels in Bulimia Nervosa Candidates for Sleeve Gastrectomy. Arch Iran Med. 2020;23(1):23-30.\u003c/li\u003e\n\u003cli\u003eTerra X, Auguet T, Ag\u0026uuml;era Z, Quesada IM, Orellana‐Gavald\u0026agrave; JM, Aguilar C, et al. Adipocytokine levels in women with anorexia nervosa. Relationship with weight restoration and disease duration. Int J Eating Disord. 2013;46(8):855-61.\u003c/li\u003e\n\u003cli\u003eVaisman N, Hahn T, Karov Y, Sigler E, Barak Y, Barak V. Changes in cytokine production and impaired hematopoiesis in patients with anorexia nervosa: the effect of refeeding. Cytokine. 2004;26(6):255-61.\u003c/li\u003e\n\u003cli\u003eV\u0026iacute;ctor VM, Rovira‐Llopis S, Saiz‐Alarc\u0026oacute;n V, Sang\u0026uuml;esa MC, Rojo‐Bofill L, Ba\u0026ntilde;uls C, et al. Involvement of leucocyte/endothelial cell interactions in anorexia nervosa. European journal of clinical investigation. 2015;45(7):670-8.\u003c/li\u003e\n\u003cli\u003eYasuhara D, Hashiguchi T, Kawahara K, Nakahara T, Harada T, Taguchi H, et al. High mobility group box 1 and refeeding-resistance in anorexia nervosa. Mol Psychiatr. 2007;12(11):976-7.\u003c/li\u003e\n\u003cli\u003eXu C, Mutwalli H, Haslam R, Keeler JL, Treasure J, Himmerich H. C-reactive protein (CRP) levels in people with eating disorders: A systematic review and meta-analysis. J Psychiatr Res. 2024.\u003c/li\u003e\n\u003cli\u003eSchindler R, Mancilla J, Endres S, Ghorbani R, Clark SC, Dinarello CA. Correlations and interactions in the production of interleukin-6 (IL-6), IL-1, and tumor necrosis factor (TNF) in human blood mononuclear cells: IL-6 suppresses IL-1 and TNF. 1990.\u003c/li\u003e\n\u003cli\u003eMatthys P, Mitera T, Heremans H, Van Damme J, Billiau A. Anti-gamma interferon and anti-interleukin-6 antibodies affect staphylococcal enterotoxin B-induced weight loss, hypoglycemia, and cytokine release in D-galactosamine-sensitized and unsensitized mice. Infection and Immunity. 1995;63(4):1158-64.\u003c/li\u003e\n\u003cli\u003eKistner TM, Pedersen BK, Lieberman DE. Interleukin 6 as an energy allocator in muscle tissue. Nat Metab. 2022;4(2):170-9.\u003c/li\u003e\n\u003cli\u003eMu\u0026ntilde;oz‐C\u0026aacute;noves P, Scheele C, Pedersen BK, Serrano AL. Interleukin‐6 myokine signaling in skeletal muscle: a double‐edged sword? The FEBS journal. 2013;280(17):4131-48.\u003c/li\u003e\n\u003cli\u003eCarson WE, Giri JG, Lindemann M, Linett ML, Ahdieh M, Paxton R, et al. Interleukin (IL) 15 is a novel cytokine that activates human natural killer cells via components of the IL-2 receptor. The Journal of experimental medicine. 1994;180(4):1395-403.\u003c/li\u003e\n\u003cli\u003eKhalafi M, Maleki AH, Symonds ME, Sakhaei MH, Rosenkranz SK, Ehsanifar M, et al. Interleukin-15 responses to acute and chronic exercise in adults: a systematic review and meta-analysis. Front Immunol. 2024;14:1288537.\u003c/li\u003e\n\u003cli\u003eWu X, Hsuchou H, Kastin AJ, He Y, Khan RS, Stone KP, et al. Interleukin-15 affects serotonin system and exerts antidepressive effects through IL15R\u0026alpha; receptor. Psychoneuroendocrinology. 2011;36(2):266-78.\u003c/li\u003e\n\u003cli\u003eDi Castro MA, Garofalo S, Mormino A, Carbonari L, Di Pietro E, De Felice E, et al. Interleukin-15 alters hippocampal synaptic transmission and impairs episodic memory formation in mice. Brain, Behavior, and Immunity. 2024;115:652-66.\u003c/li\u003e\n\u003cli\u003eWiner H, Rodrigues GO, Hixon JA, Aiello FB, Hsu TC, Wachter BT, et al. IL-7: comprehensive review. Cytokine. 2022;160:156049.\u003c/li\u003e\n\u003cli\u003eBowers TK, Eckert E. Leukopenia in anorexia nervosa: Lack of increased risk of infection. Archives of Internal Medicine. 1978;138(10):1520-3.\u003c/li\u003e\n\u003cli\u003eSaito H, Nomura K, Hotta M, Takano K. Malnutrition induces dissociated changes in lymphocyte count and subset proportion in patients with anorexia nervosa. Int J Eating Disord. 2007;40(6):575-9.\u003c/li\u003e\n\u003cli\u003eGermain N, Viltart O, Loyens A, Bruchet C, Nadin K, Wolowczuk I, et al. Interleukin-7 plasma levels in human differentiate anorexia nervosa, constitutional thinness and healthy obesity. PLoS One. 2016;11(9):e0161890.\u003c/li\u003e\n\u003cli\u003eCaldas ND, Braulio VB, Brasil MAA, Furtado VCS, de Carvalho DP, Cotrik EM, et al. Binge eating disorder, frequency of depression, and systemic inflammatory state in individuals with obesity-A cross sectional study. Arch Endocrinol Metab. 2022;66(4):489-97.\u003c/li\u003e\n\u003cli\u003eD\u0026apos;Esposito V, Di Tolla MF, Lecce M, Cavalli F, Libutti M, Misso S, et al. Lifestyle and dietary habits affect plasma levels of specific cytokines in healthy subjects. Front Nutr. 2022;9:913176.\u003c/li\u003e\n\u003cli\u003eSeitz J, Lahaye E, Andreani NA, Thomas B, Takhlidjt S, Chartrel N, et al. Long-term dynamics of serum \u0026alpha;-MSH and \u0026alpha;-MSH-binding immunoglobulins with a link to gut microbiota composition in patients with anorexia nervosa. Neuroendocrinology. 2024;114(10):907-20.\u003c/li\u003e\n\u003cli\u003eFetissov SO, D\u0026eacute;chelotte P. The putative role of neuropeptide autoantibodies in anorexia nervosa. Current Opinion in Clinical Nutrition \u0026amp; Metabolic Care. 2008;11(4):428-34.\u003c/li\u003e\n\u003cli\u003ePaszynska E, Dmitrzak-Weglarz M, Tyszkiewicz-Nwafor M, Slopien A. Salivary alpha-amylase, secretory IgA and free cortisol as neurobiological components of the stress response in the acute phase of anorexia nervosa. The World Journal of Biological Psychiatry. 2016;17(4):266-73.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table 2","content":"\u003cp\u003eTable 2 is available in the Supplementary Files section.\u003c/p\u003e\n"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6547856/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6547856/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003ePrior research has found altered levels of immune signalling proteins, such as cytokines, in people with eating disorders (EDs). This study is an update of a previously published meta-analysis.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis systematic review and meta-analysis assessed cross-sectional and longitudinal studies from four databases (PubMed, Web of Science, MEDLINE and PsycINFO) reporting cytokine concentrations in people with EDs. Random-effects models were utilised for all meta-analyses.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eTwenty-four new studies were incorporated, resulting in a total of 43 studies included in the meta-analyses. Interleukin (IL)-6 and IL-15 were higher, and IL-7 was lower, in AN compared with HC. When controlling for outliers, concentrations of tumour necrosis factor (TNF)-α, IL-1β, IL-4, IL-8, IL-10, IFN-γ, monocyte chemoattractant protein (MCP) and transforming growth factor (TGF)-β were similar between AN and HC. Longitudinally, IL-6 was lower in AN at follow-up compared to baseline, whereas TNF-α and IL-1β did not change. There were largely no differences in IL-6 and TNF-α in BN and were insufficient studies to perform meta-analyses for binge eating disorder (BED) or other EDs.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eIn acute AN, concentrations of IL-6 and IL-15 are elevated and IL-7 is decreased, with evidence for normalisation of IL-6 over the course of weight restoration. Concentrations of other cytokines considered to broadly have pro-inflammatory functions were not increased in AN. In people with BN, there is less evidence for increases in pro-inflammatory cytokines, but the evidence base is limited. Methodological considerations for future studies are recommended.\u003c/p\u003e","manuscriptTitle":"A comprehensive updated cross-sectional and longitudinal meta-analysis of cytokines in eating disorders","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-15 10:42:03","doi":"10.21203/rs.3.rs-6547856/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"communications-medicine","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"commsmed","sideBox":"Learn more about [Communications Medicine](http://www.nature.com/commsmed)","snPcode":"43856","submissionUrl":"https://mts-commsmed.nature.com/cgi-bin/main.plex","title":"Communications Medicine","twitterHandle":"@commsmedicine","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Communications Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"bb6cfe1e-44b9-4959-9710-9309be6090c0","owner":[],"postedDate":"May 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":48362032,"name":"Biological sciences/Neuroscience/Neuroimmunology"},{"id":48362033,"name":"Biological sciences/Immunology/Cytokines"},{"id":48362034,"name":"Biological sciences/Psychology/Human behaviour"},{"id":48362035,"name":"Health sciences/Diseases/Psychiatric disorders"},{"id":48362036,"name":"Health sciences/Diseases/Nutrition disorders/Malnutrition"}],"tags":[],"updatedAt":"2025-10-02T07:07:03+00:00","versionOfRecord":{"articleIdentity":"rs-6547856","link":"https://doi.org/10.1038/s43856-025-01122-z","journal":{"identity":"communications-medicine","isVorOnly":false,"title":"Communications Medicine"},"publishedOn":"2025-10-01 04:00:00","publishedOnDateReadable":"October 1st, 2025"},"versionCreatedAt":"2025-05-15 10:42:03","video":"","vorDoi":"10.1038/s43856-025-01122-z","vorDoiUrl":"https://doi.org/10.1038/s43856-025-01122-z","workflowStages":[]},"version":"v1","identity":"rs-6547856","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6547856","identity":"rs-6547856","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}