Anorexia Nervosa in vivo cytokine production: a systematic review

Introduction: The underlying psychobiology that contributes to Anorexia Nervosa (AN) onset and disease progression remains unclear. New research is emerging suggesting a possible link between inflammation and a variety of mental illnesses. Alterations of cytokines may play a role in the pathogenesis of AN. Some studies have found differences in the cytokine profile of those with AN compared to healthy controls, but results are heterogeneous. The aim of this work was to systematically review existing studies investigating in-vivo cytokine production in those with AN before and after weight restoration compared to controls. Methods: A comprehensive literature search of four electronic databases (PubMed, PsychInfo, EMBASE and CINAH) was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines to identify human in-vivo studies investigating the relationship between AN and cytokine production. Data extracted from included studies related to population characteristics (e.g. age, gender, mean mBMI/%IBW), cytokine measurement and relevant findings. Confounding factors (e.g. smoking status, co-morbid mental illness, menstruation status) were also collected. Results: 36 studies were eligible for this systematic review of which the majority were conducted in Europe (77.8%) and involved female subjects (97.2%). Those with AN ranged in age from 13 to 47 years and had an illness duration of 3 months to 24 years. 15 candidate cytokines and 3 receptors were identified (TNF-alpha, IL-6, IL-1B, CRP, IL-2, IL-7, IL-10, IFN-γ , TNF-R2, IL-1 α , IL-15, TNF-R1, IL-17, IL-18, TGF-B1, IL-12, IL-6R and TGF-B2) exploring in-vivo levels in patients with AN and comparing to controls. TNF-alpha and IL-6 were the most extensively studied with IL-6 being significantly elevated in 4 out of 8 (50%) of longitudinal studies when comparing AN patients at baseline compared to post weight restoration. Following weight restoration, there was no difference in IL-6 levels when comparing to HC in 7 of 8 (87.5%) longitudinal studies examined. Conclusions: The most promising cytokine potentially involved in the pathogenesis of AN appears to be IL-6, and possibly TNF-alpha pathways. The heterogeneity of clinical and methodology factors impedes the generalizability of results. Future studies may wish to address these methodological shortcomings as alterations in cytokine levels in AN could act as therapeutic targets assisting with weight restoration and psychopathology and may offer diagnostic potential.


Introduction
Anorexia Nervosa (AN) is a mental illness associated with significant morbidity and the highest mortality rate of all psychiatric disorders (Harris, Barraclough, 1998;Nielsen, 2001;Sullivan, 1995).With a lifetime prevalence of AN in females ranging from 1.2% to 2.2%, AN is a relatively common disorder (Jagielska and Kacperska, 2017) characterized by restrictive eating, significantly low body weight, body image distortion and an irrational fear of weight gain.It has been estimated amongst surviving AN patients, less than one-half fully recover, one-third improve and approximately 20% remain chronically ill (Steinhausen, 2002).There is little evidence to suggest outcomes for AN have improved over the second half of the last century, reminding us of the lack of robust and effective treatments.
Efforts to develop effective treatments for AN remain elusive and has contributed to an urgent and increased examination of putative biological mechanisms that may underly the disorder.Emerging research suggests the immune system may play a potential role in the pathophysiology of AN (Gibson and Mehler, 2019).While protein caloric malnutrition is the most common cause of human immunodeficiency (Słotwińska, Słotwiński, 2017) leading to increased susceptibility to infection (Chandra and Kumari, 1994) underweight AN patients unusually are not at increased risk of viral infections (Winston, Jan et al., 2023) and indeed in the presence of infection, they often exhibit fewer symptoms such as fever in response to viral infection (Dobner and Kaser, 2018).The extent to which alterations in the immune system and cytokines play a role in the pathophysiology of AN and response to infections is yet unknown.In addition, genome wide association studies provide additional evidence for a dysregulated immune system with the finding of shared loci of interest between AN and multiple autoimmune disorders (Duncan et al., 2017).Furthermore, those with autoimmune and autoinflammatory diseases are at an increased risk for the development of eating disorders (ED) such as AN (Zerwas et al., 2017) suggesting immune system disturbance may play a role in the pathogenesis and/or maintenance of AN.
New research is evolving suggesting a possible link between inflammation and a variety of mental illnesses that are highly co-morbid with AN such as depression (Beurel, Toups, and Nemeroff, 2020) and anxiety (Peirce and Alviña, 2019).Inflammation of the body can cause changes in how the brain works, which in turn causes changes in mood, cognition and behavior (Bullmore, 2019).Compelling evidence shows, inflammation-associated anorexia, reduced appetite and food intake, observed during acute and chronic inflammatory states in humans and animals, have been attributed to pro-inflammatory factors, primarily cytokines, on the nervous system.Such evidence demands consideration of pro-inflammatory pathways and AN pathophysiology (Gautron, Layé, 2010).
Cytokines are soluble proteins that regulate the immune response through complex and dynamic networks.A comprehensive review by Kronfol and Remick (2000), documented the emerging relevance of cytokine biology to clinical psychiatry suggesting cytokines may act as both immunomodulators and neuromodulators contributing to the pathogenesis of a variety of mental illnesses.Injection of Interleukin 6 (IL-6), Tumor Necrosis Factor alpha (TNF-alpha) and Interleukin 1β (IL-1β) into experimental animals induces actions that mimic aspects of AN such as fatigue, apathy, decreased appetite and depressed mood (Brambilla, Monti, and Franceschi, 2001).Expanding these studies into humans is more challenging for researchers and clinicians (Aziz, 2015) although progress is being made using both in-vitro and in-vivo studies.
In-vitro cytokine studies occur within a controlled environment, such as a test tube, whereas in-vivo cytokine studies are performed within a living being such as a person or laboratory animal.In-vitro studies allow for greater control of the chemical environment and larger numbers of samples can be analyzed at a lower cost.However, replication of living cell conditions is crucial, therefore contributing to the superiority of the more costly and time consuming in-vivo studies, where results are viewed as more specific and reliable given biological effects are observed directly in the test subject (Liu et al., 2021).
The existent literature on in-vivo inflammatory cytokine production in AN has been carefully and systematically reviewed in two metaanalyses led by Solmi (Solmi et al., 2015) and Dalton (Dalton et al., 2018).Similarities between meta-analyses (Solmi, andDalton et al., 2015, 2018) include elevated circulating concentrations of proinflammatory cytokines TNF-alpha and IL-6 amongst AN patients in comparison to healthy controls (HCs) however only Solmi et al., 2015, showed an elevation in IL-1β amongst AN patients compared to HCs.Solmi et al., 2015, examined longitudinal levels in cytokine concentration and showed no significant difference in IL-6 and TNF-alpha in patients affected by AN compared to HCs following weight gain.Missing data limited this analysis in the more recent review by Dalton et al., 2018.Additional studies have been added since 2018 (Caroleo et al., 2019;Caso et al., 2020;Dalton et al., 2018;Elegido et al., 2019;Himmerich et al., 2021;Keeler et al., 2021;Mariani et al., 2018;Nilsson et al., 2020;Roczniak et al., 2020;Dalton et al., 2019;Roubalova et al., 2021) allowing further exploration of the data.
The current study aims to update the synthesis of literature by conducting a systematic review of cross-sectional and longitudinal studies assessing in vivo-cytokine production in those with AN compared to healthy controls.As well, this study will review in-vivo cytokine production before and after weight restoration in those with AN.Our aim is to address the following two research questions: (1) Does in-vivo cytokine production differ between those with AN compared to Healthy Controls (HC)?(2) Does in-vivo cytokine production change longitudinally in those with AN following weight gain?

Methods
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Moher et al., 2009).

Study selection
Included were only studies in English that (1) reported on in-vivo serum levels of pro-inflammatory or anti-inflammatory cytokines (2) in patients with AN, diagnosed by Diagnostic and Statistical Manual of Mental Disorders (DSM) III/IV/V or ICD-10 criteria and (3) comparing AN patients with HCs (not other psychiatric samples), and (4) with or without comparisons within patients affected by AN before and after weight gain.
Studies were excluded if: (1) they did not use clear diagnostic criteria for AN, (2) they assessed in-vitro cytokine concentration or genetic expression but did not assess cytokine production in-vivo; (3) they did not measure or did not report quantitative cytokine levels in both patients and HCs (4) they did not statistically compare AN patients with HCs or before and after weight gain; (5) they were not in English or full text was not available or (5) if the sample was comprised of animals.If HCs were included in longitudinal studies, only the baseline data were used as part of the cross-sectional comparisons.Review articles, metaanalyses, conference proceedings/abstracts, editorials, letters, book chapters and unpublished theses were excluded once further searches failed to generate publications.

Search strategy
Four electronic databases (PubMed, PsychInfo, EMBASE and CINAHL) were searched from data base inception until 20/06/2022, using the following keywords and mapped to the subject headings with the explode function where possible: "anorexia nervosa" in combination with cytokine* OR inflammat* OR interleukin* OR interferon* OR IFN OR tumor necrosis factor OR TNF OR TNF-X OR IL-1B OR IL-6 OR CRP.Reference lists of included articles and those relevant to the topic were hand-searched for identification of potentially relevant articles.
Titles and abstracts of retrieved publications were imported into EndNote, duplicates were removed, and papers that were deemed unlikely to be relevant were disregarded.Full-text versions of the remaining articles were then obtained and screened according to the pre-specified eligibility criteria described above.All papers that did not meet inclusion criteria were excluded, with the reasons documented (see flow chart in Fig. 1).The entire search process was conducted independently by two reviewers (K.M. and E.J.) and disagreements at the final stage were resolved by consulting a third reviewer (F.M.).

Data extraction
Two reviewers (K.M. and E.J.) extracted data from all included studies into electronic summary tables which were then checked by two other reviewers (F.M. and E.M.).Information collected related to population characteristics (e.g.age, gender, mean BMI/%, IBW), cytokine measurement methods and relevant findings.As cytokine measurement is affected by a range of confounding factors (e.g.smoking status, menstruation status, co-morbid mental illness), details regarding these were also collected and are shown in Table 1.

Characteristics of included studies
An initial search yielded 1134 studies (see Fig. 1.For PRISMA flow diagram).36 studies met inclusion criteria, consisting of 25 crosssectional studies and 11 longitudinal studies.All included studies assessed in-vivo cytokine production in participants with AN.Studies were conducted in Europe (28 studies, 77.8%), the United States of America (3 studies, 8.3%), Japan (4 studies, 11.1%) and Israel (1 studies, 2.7%).The age range of patients with AN and HC was 13-47 and 13-54 year with only one study (2.77%) including both male and female subjects.Diagnostic criteria used for AN differed across studies, the majority used DSM-IV (32 studies, 88.8%), one used DSM-III (, 2.7%), and three DSM-V (8.33%).Body Mass Index (BMI) in AN patients at baseline was reported in 32 studies (88.8%), ranging from 10.6 to 19.8.BMI in HCs was reported in 30 studies (83.3%) and ranged from 15.7 to 25.0 The range of illness duration of AN participants was 3-24 months, and reported in 21 studies (58.3%).
The following cytokines were assessed in the included studies: TNF- The Joanna Briggs Institute (JBI) critical appraisal checklist for prevalence studies was used to assess the methodological quality of each study (Munn et al., 2014).Based on this assessment, 18 studies (50%) were considered of high quality whereases 18 studies (50%) were consider of medium quality (See Tables 2 and 3).Overall, it was felt that data analysis was conducted with sufficient coverage of the identified sample and appropriate statistical analysis was carried out.Where some studies fell short, was due to lack of detail in description of control subjects and absence of reporting of confounding factors.
The clinical relevance of included studies was also assessed based on the extent to which the study might help us to better understand the etiology of AN, play a role in diagnostics or offer a possible therapeutic target.It was felt that 25 studies (68.4%) offered a medium degree, 10 studies (27.7%) a high degree and one study (2.7%) a low degree of clinical relevance (See Tables 2 and 3).

Reporting on confounding factors (see Table 1)
Age matching between AN and HCs was reported in 22 studies (61.1%).Smoking status was reported on (i.e.smoker vs. non-smoker) in 10 studies (27.8%).Amenorrhea in those with AN was commented on in 11 studies (30.6%), sampling during the follicular stage of the menstrual cycle in HC was mentioned in 8 studies (22.2%) and in 21 studies (58.3%) there was no mention of menstrual status of participants.
Those with AN and HC were confirmed medication free in 16 studies (44.4%).Psychopharmacological medication was administered to AN patients as part of their treatment in 5 studies (13.9%). 1 study (2.8%) mentioned some participants were on the oral contraceptive pill.Medication status of patients was not reported in 12 studies (33.3%).Those with AN received psychotherapy in 9 studies (25%), nutritional refeeding in 7 studies (19.4%) and in 21 studies (58.3%) intervention(s) received were not reported.Seventeen studies (47.2%) reported that HCs were free from psychiatric disorders.An additional 6 studies (16.6%) reported that HCs were free from a history of an eating disorder, depression, substance use disorder psychotic disorder or any psychiatric disorder resulting in cachexia.Three studies (8.3%) confirmed AN patients had psychiatric co-morbidity such as depression or bipolar affective disorder whereas 10 studies (27.7%) stated there was no psychiatric co-morbidity in AN patients.There was no mention of possible psychiatric co-morbidity amongst AN patients in 23 studies (63.8%).
In relation to physical health disorders, 18 studies (50%) reported all participants were free from medical illness.HCs were reported to be free from physical illness in 3 studies (8.33%).Thirteen studies (36.1%) did not report on presence or absence of physical health disorders amongst participants.Additionally, all participants were reported to be free of infection or infectious disease in 10 studies (27.7%) and free from an autoimmune or inflammatory disease in 11 studies (30.5%).

In-vivo cytokine production
This systematic review identified 15 candidate cytokines and 3 receptors exploring in-vivo levels in patients with AN and comparing to healthy controls.TNF-alpha and IL-6 being the most extensively studied with both cross sectional and longitudinal data available.Other candidate cytokines and receptors were reported in smaller cross sectional studies.

Comparison on patients with AN with healthy control subjects.
Data from 4 of 16 (25%) cross-sectional studies (Agnello et al., 2012;Nakai et al., 1999;Ostrowska et al., 2015;Caso et al., 2020) and baseline values from 2 of 9 (22.2%)longitudinal studies (Nakai et al., 1999;Roubalova et al., 2021) showed TNF-alpha was significantly higher in patients with AN than HCs.However, in 2 of the 16 (12.5%)cross-sectional studies (Nogueira et al., 2010;Elegido et al., 2019) TNF-alpha was found to be significantly lower in AN than HCs.The remaining 17 of 25 (68%) cross-sectional and longitudinal studies showed no difference in TNF-alpha levels between those with AN compared to HCs (Tables 2 and 3).(Nakai et al., 1999;Roubalova et al., 2021) showed TNF-alpha was significantly higher in patients before compared to after weight gain.The remaining 7 of 9 (77.8%)longitudinal studies showed no change in TNF-alpha in those with AN before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and HCs.
Only 1 out of 9 (11.1%)longitudinal studies (Nakai et al., 1999) showed significantly higher levels of TNF-alpha in those with AN after weight gain compared to HCs.The remaining 8 of 9 (88.9%)longitudinal studies showed no significant difference in TNF-alpha in those with AN post weight gain compared to HCs (Table 3).

Comparison on patients
with AN before and after weight gain.4 out of 8 (50%) longitudinal studies (Dalton et al., 2019;Komorowska-Pietrzykowska et al., 2001;Misra et al., 2006;Pomeroy et al., 1994) showed significantly higher baseline values of IL-6 in low weight AN compared to after weight gain.The remaining 4 of 8 (50%) longitudinal studies did not show any difference between IL-6 measures before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and HCs.
Only 1 of the 8 (12.5%) longitudinal studies (Komorowska-Pietrzykowska et al., 2001) showed significantly higher levels of IL-6 in those with AN following weight gain compared to HCs.The remaining 7 of 8 (87.5%) longitudinal studies showed no significant difference in IL-6 in those with AN post weight gain compared to HCs (Table 3).

Comparison between patients with AN after weight gain and HCs.
Across 3 longitudinal studies (Brambilla et al., 1999;Dalton et al., 2019;Vaisman et al., 2004) no significant differences were found in IL-β amongst patients after weight gain compared to HC's (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 4 of 7 (57.1%)cross-sectional studies (Dolezalova et al., 2007;Haluzíková et al., 2009;Lawson et al., 2007;Mariani et al., 2018) and 1 of 1 (100%) longitudinal studies (Misra et al., 2006) showed CRP was significantly lower in patients with AN than HCs.The remaining 3 of 8 (37.5%) cross-sectional and longitudinal studies showed no difference in CRP levels between those with AN compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal study (Misra et al., 2006) showed no significant difference in CRP in those with AN before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and HCs.
Results from 1 of 1 (100%) longitudinal study (Misra et al., 2006) showed significantly lower CRP in those with AN following weight gain compared to HCs.(Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 1 or 5 (20%) cross-sectional studies (Elegido et al., 2019) showed IL-2 was significantly higher in patients with AN than HCs while 1 of 5 (20%) cross-sectional studies (Corcos et al., 2001) showed IL-2 was significantly lower in patients than HCs.The remaining 4 of 6 (80%) cross-sectional and longitudinal studies showed no difference in IL-2 between those with AN compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-2 before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and HCs.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-2 in those with AN following weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 2 or 5 (40%) cross-sectional studies (Germain et al., 2016;Keeler et al., 2021) showed IL-7 was significantly lower in patients with AN than HCs.The remaining 4 of 6 (66.6%) cross-sectional and longitudinal studies showed no difference in IL-7 between those with AN compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed IL-7 was significantly decreased in those with AN before compared to after weight gain (Table 3).

Comparison between patients with AN after weight gain and HCs.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-7 in those with AN after weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 1 of 4 (25%) cross-sectional studies (Caroleo et al., 2019) showed IL-10 was significantly lower in patients with AN than HCs.The remaining 4 of 5 (80%) cross-sectional and longitudinal studies showed no difference in IL-10 between those with AN compared to HCs (Tables 2  and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-10 before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and HCs.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-10 in those with AN after weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 1 of 4 (25%) cross-sectional studies (Caroleo et al., 2019) showed IFNγ was significantly lower in patients with AN than HCs.In contrast, baseline values from 1 of 2 (50%) longitudinal studies (Komorowska-Pietrzykowska et al., 2001) showed IFNγ was significantly increased in patients with AN compared to HCs.The remaining 4 of 6 (66.6%) cross-sectional and longitudinal studies showed no difference in IFNγ between those with AN compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 2 (50%) longitudinal studies (Komorowska-Pietrzykowska et al., 2001) showed IFN-y was significantly decreased in those with AN before compared to after weight gain.The remaining 1 of 2 (50%) longitudinal studies (Dalton et al., 2019) showed no change in IFN-y in patients before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and HCs.
Results from 1 of 2 (50%) longitudinal studies (Komorowska-Pietrzykowska et al., 2001) showed IFNγ was significantly increased in patients after weight gain compared to HCs (Table 2).The remaining 1 of 2 (50%) longitudinal studies (Dalton et al., 2019) showed no change in IFNγ in patients after weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 1 of 3 (33.3%)cross-sectional studies (Nakai et al., 1999) and baseline values from 1 of 2 (50%) longitudinal studies (Nakai et al., 1999) showed TNF-R2 was significantly higher in patient with AN than HCs.The remaining 3 of 5 (60%) cross-sectional and longitudinal studies showed no difference in TNF-R2 levels between those with AN compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 2 of 2 (100%) longitudinal studies (Nakai et al., 1999;Terra et al., 2013) showed no significant difference in TNF-R2 before and after weight gain in AN patients (Table 3).

Comparison between patients with AN after weight gain and HCs.
Results from 1 of 2 (50%) longitudinal studies (Nakai et al., 1999) showed TNF-R2 was significantly increased in patients after weight gain compared to HCs (Table 3).The remaining 1 of 2 (50%) longitudinal studies (Terra et al., 2013) showed no change in TNF-R2 in patients after weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 1 of 3 (33.3%)cross-sectional studies (Caroleo et al., 2019) showed IL-1α was significantly higher in patients with AN than HCs.The remaining 3 of 4 (75%) cross-sectional and longitudinal studies showed no difference in IL-1α levels between those with AN compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-1α before and after weight gain in AN patients (Table 3).

Comparison between patients with AN after weight gain and
HCs. Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-1α in those with AN following weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 3 of 3 (100%) cross-sectional studies (Dalton et al., 2018;Himmerich et al., 2021;Roczniak et al., 2020) showed IL-15 was significantly higher in patients with AN than HCs.Baseline values from 1 of 1 (100%) longitudinal studies (Dalton et al., 2019) showed no difference in IL-15 levels between those with AN compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-15 before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and
HCs. Results from 1 of 1 (100%) longitudinal study (Dalton et al., 2019) showed no significant difference in IL-15 in those with AN following weight gain compared to HCs (Table 3).

Comparison on patients with AN before and after weight gain.
Results from 2 of 2 (100%) longitudinal studies (Dalton et al., 2019;Roubalova et al., 2021) showed no significant difference in IL-17 before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and
HCs. Results from 2 of 2 (100%) longitudinal studies (Dalton et al., 2019;Roubalova et al., 2021) showed no significant difference in IL-17 in those with AN after weight gain compared to HCs (Table 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal study (Nakai et al., 1999) showed no significant difference in TNF-R1 before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and
HCs. Results from 1 of 1 (100%) longitudinal study (Nakai et al., 1999) showed no significant difference in TNF-R1 in those with AN after weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 1 of 3 (33.3%)cross-sectional studies (Keeler et al., 2021) showed IL-12 was significantly decreased in those with AN compared to HCs.The remaining 3 of 4 (75%) of cross-sectional and longitudinal studies showed no significant difference in IL-12 levels in patients compared to HCs (Tables 2 and 3).

Comparison on patients with AN before and after weight gain.
Results from 1 of 1 (100%) longitudinal studies (Dalton et al., 2019) showed no significant difference in IL-12 before and after weight gain (Table 3).

Comparison between patients with AN after weight gain and
HCs. Results from 1 of 1 (100%) longitudinal studies (Dalton et al., 2019) showed no significant difference in IL-12 in those with AN after weight gain compared to HCs (Table 3).

Comparison on patients with AN with healthy control subjects.
Data from 2 of 2 (100%) cross-sectional studies (Dalton et al., 2018;Himmerich et al., 2021) showed TNF-B1 was significantly decreased in those with AN compared to HCs (Table 2).

Comparison on patients with AN with healthy control subjects.
Data from 1 of 2 (50%) cross-sectional studies (Tanaka et al., 2019) showed IL-18 was significantly decreased in those with AN compared to HCs.However, in the remainder 1 of 2 (50%) cross-sectional studies (Nilsson et al., 2020) no significant difference emerged in IL-18 levels between those with AN compared to HCs (Table 2).

Comparison on patients with AN with healthy control subjects.
Data from 2 of 2 (100%) cross-sectional studies (Brambilla et al., 2001;Karczewska-Kupczewska et al., 2013) showed IL-6R was significantly decreased in those with AN compared to HCs (Table 2).

Comparison on patients with AN with healthy control subjects.
Data from 1 of 1 (100%) cross-sectional study (Corcos et al., 2001) showed TGF-B2 was significantly decreased in those with AN compared to HCs (Table 2).

Discussion
Cytokine biology remains a rapidly evolving area of science with promising application to clinical psychiatry.Despite challenges in finding candidate cytokines for AN diagnosis and treatment, efforts are worthwhile (Kronfol and Remick, 2000).We hypothesized in-vivo cytokine production differed between those with AN compared to HCs and changed over time in response to weight restoration.Results from this updated systematic review, with data from 36 studies and 2032 participants (AN = 1103 and HCs = 929) concurred with earlier findings from Solmi et al. (2015) showing significantly lower levels of CRP and IL-6R in AN patients when compared to HCs.As well, similar to both Solmi et al. (2015) and Dalton et al. (2018), IL-6 was elevated in AN patients compared to HC at baseline.As shown by Solmi et al. (2015), most of the longitudinal studies in this systematic review found no Table 4 Study Findings.

Study
Findings Outcome Agnello et al. (2012) Italy TNF-x was significantly higher in AN vs. HC.TNFR1 & 2 higher in AN vs. HC but nonsignificantly.Separated sample into disease duration < 1 year vs. > 1 year.Found those with longer disease duration had significantly higher levels of TNFR1&2.TNFx equally elevated in both groups.significant difference in IL-6 levels amongst those with AN post weight gain compared to HCs indicating IL-6 tended to normalize following weight restoration.Contrary to results reported by Solmi et al. (2015), this systematic reviewed showed no difference in TNF-alpha, TNF-Receptor II or IL-1β levels, and an increase in IL-15 levels in AN patients compared to healthy controls at baseline.These findings may be explained by the positive correlation between CRP and fat mass confirmed in previous studies showing those with increased adiposity and body mass index (BMI) have significantly higher CRP levels (Giles et al., 2008;Jamialahmadi et al., 2023).Thus, it would be expected that those with AN having low adiposity and BMI would therefore have lower levels of CRP compared to HCs.IL-15 has been reported to have an anabolic role, i.e., decreasing fat and increasing muscle mass (Pedersen, 2011), and raised IL-15 in underweight AN patients may be reflective of the body's effort to maintain muscle mass in the ill state (Dalton et al., 2018).Elevated levels of IL-15 have also been associated with decreased serotonergic transmission and depressive symptomology (Pan et al., 2013;Wu et al., 2011).Given those with AN commonly present with depressive symptoms, elevation of IL-15 may be negatively modulating serotonergic transmission.The imbalance seen between elevated levels of IL-6 and decreased levels of IL-6R at baseline may represent a compensatory mechanism in AN patients used to reduce the pro-inflammatory effect of IL-6.
IL-6 is a multifunctional cytokine that plays a central role in host defense through its potent ability to induce the acute phase response and subsequent inflammation, immunity and disease (Tanaka et al., 2014;Simpson et al., 1997).IL-6 has been shown to be significantly elevated in those with higher BMI (e.g.obese individuals) (Baikpour, Baikpour, Hosseini, and Sarveazad, 2017) and is cachexic and anorexigenic, stimulating decreased food consumption (Holden RJ, 1996;Langstein, N. J, 1991).Interpretation of IL-6 production in primary malnutrition suggests when controlling for infection, IL-6 seems overall to be decreased (Gibson and Mehler, 2019;Wotton, James, and Goldacre, 2016).Therefore, given adiposity, BMI and nutrition is greatly reduced in AN patients, the significant elevation in IL-6 levels in a subset of patients is intriguing.Elevated levels of IL-6 in AN patients reported by the meta-analyses ( Solmi et al., 2015 andDalton et al., 2018) and a subset of our studies (Himmerich et al., 2021;Dalton et al., 2018;Karczewska-Kupczewska et al., 2013;Lawson et al., 2007;Ostrowska et al., 2015;Dalton et al., 2019;Komorowska-Pietrzykowska et al., 2001;Misra et al., 2006;Pomeroy et al., 1994) may help contribute to decreased appetite and subsequent weight loss experienced by those with AN.
A recent systematic review and meta-analysis (Patsalos et al., 2020) examined the effect on weight and BMI of Toclizumab, an IL-6 pathway inhibitor, in the treatment of immune diseases e.g., rheumatoid arthritis.IL-6 pathway inhibitors were associated with significant increases in weight and BMI suggesting that the IL-6 pathway is involved in weight regulation.Furthermore, two case reports, involving individuals with a dual diagnosis of AN and autoimmune disease, treated with TNF-alpha pathway inhibitors also demonstrated significant weight gain and improvement in AN psychopathology several months after beginning immunosuppressive treatment (Barber et al., 2003;Solmi et al., 2013) This contributes to clinical promise, in that modulating the IL-6 and TNF-alpha pathways using pharmacological adjuncts, may be a potential therapeutic target for AN patients.However, these medications are not without adverse side-effects including increased rates of serious infection, cancer (Li et al., 2021) and depressive symptomatology amongst medically ill patients (Knight et al., 2021).
Although earlier evidence reported elevated levels of serum cytokine in those with AN, more recent studies have had contradictory findings.Ruiz Guerrero et al. ( 2022) carried out a case-control study in a sample of forty-eight young patients with early diagnosis of AN, bulimia nervosa (BN) or other specified feeding and eating disorders (OSFED).Contrary to expectations, levels of IL-1β and IL-6 were significantly lower in ED patients compared to HCs supporting the idea of an immunosuppressive status in the early stages of disease.However, it is difficult to extrapolate findings given there was no sub-analysis of AN patients.Specht et al. (2022), carried out a cohort study of 22 female adolescents with AN also at an early stage of their illness.The authors felt studying adolescents with a shorter illness duration and fewer confounding factors may help to elucidate the role of inflammation in the pathophysiology of AN.Specht et al. (2022), confirmed previous findings of elevated TNF-alpha however reduced IL-1β and IL-6 suggesting a mixed pro-and anti-inflammatory state appears to be present in adolescents, potentially due to their shorter illness duration.
There are several methodological shortcomings that may account for conflicting results we have seen in this systematic review such as small sample sizes, differing cytokine measurement tools and many confounding factors.Similar to the systematic review looking at in-vitro cytokine production in Eating Disorders carried out by Dalton et al., 2015 we found several important confounding factors such as smoking and medication status, presence of autoimmune or inflammatory disorders and concurrent mental and physical illness which may have influenced inflammation and thus cytokine results.For example, most studies included patients who were currently on at least one medication.With medication known to have an effect on inflammation and cytokine production (Zidek et al., 2009), this may have impacted results.Cigarette smoke promotes inflammation by inducing the production of pro-inflammatory cytokines, such as TNF-alpha, IL-6, IL-1 and IL-8 (Strzelak et al., 2018).Despite very few studies reporting on smoking status of participants, existing data suggest a higher rate of smoking in people with EDs compared to HCs (Anzengruber et al., 2006), suggesting that smoking may be a confounding variable and important to control for.Additionally, psychiatric co-morbidity including depression and anxiety, are commonly seen in AN (Hughes et al., 2013) and are associated with alterations in concentrations and production of cytokines (Lichtblau et al., 2013;Vogelzangs et al., 2013).With most studies not commenting on presence or absence of psychiatric co-morbidity in AN, it is uncertain whether alterations in cytokines are due to AN or symptoms of co-morbid mental illness.However, Dalton et al., examined levels of Abbreviations: HC = healthy control; AN = anorexia nervosa; AN-R = Anorexia Nervosa Restrictive Type; AN-BP = Anorexia Nervosa Binge Purge Type; T1/T2/ T3: Time 1/2/3;*sig = significant; *non-sig = non-significant general psychopathology (symptoms of depression and anxiety, as measured by the DASS-21) in patients with AN and found few inflammatory markers were significantly associated with general psychopathology symptoms.In contrast, (Komorowska-Pietrzykowska et al., 2001), found IL-6 was significantly higher in AN patients with more pronounced depressive symptoms and decreased after treatment with an antidepressant alongside weight restoration.Furthermore, the percentage of articles including healthy controls free from psychiatric illness is low and therefore can introduce further bias in results.Disease progression and duration can have a significant impact on how we observe inflammation in the body.Given we do not know the duration of AN in almost half of all patients (41.7%) along with the duration of untreated illness amongst patients included in most studies, it is difficult to ascertain the impact on patients' bodily inflammation.Furthermore, 88.8% of subjects with AN were diagnosed based on DSM-4 criteria which required the presence of amenorrhea and bodyweight at or below the 85th percentile.The removal of these criteria from DSM-5 may alter the medical severity levels between cohorts, and therefore contribute to difference in inflammatory marker findings.Furthermore, some studies included HCs with a BMI range from 15.7, indicating low body weight, which could mean they too were physically compromised.
Future studies examining the interplay between AN and cytokines may consider assessing, analyzing and reporting on these confounding factors and how they affect inflammation when designing their study.

Conclusion
This systematic review found a growing number of studies commenting on cytokines in AN however the complexity of cytokine biology and AN continues to challenge the search for candidate cytokines.The most promising cytokine findings from this review relate to IL-6 (Himmerich et al., 2021;Dalton et al., 2018;Karczewska-Kupczewska et al., 2013;Lawson et al., 2007;Ostrowska et al., 2015;Dalton et al., 2019;Komorowska-Pietrzykowska et al., 2001;Misra et al., 2006;Pomeroy et al., 1994;Solmi et al., 2015;Dalton et al., 2018).The weight regulating role of IL-6 inhibitors reported in other studies further support this putative link (Patsalos et al., 2020).As the understanding of cytokine biology evolves, other candidate cytokines will no doubt emerge however research could usefully exclude some cytokines.This systematic review highlighted issues with study design; small sample sizes, variation in cytokine sampling techniques and confounding factors such as smoking status, psychiatric comorbidity, duration of illness, presence or absence of amenorrhea all contributing to study heterogeneity and potential bias in studies.Future studies may wish to further control for these confounding factors and their role in inflammation.
Further longitudinal studies may also help us to understand how cytokines and inflammation may influence disease onset and the trajectory of AN.If we can identify changes in cytokine levels, they could act as biomarkers, offering diagnostic potential, indicators of treatment response and possibly risk of relapse.Isolating further pathways involved in the inflammatory response, could emerge as possible pharmacological treatment targets.Further studies are warranted to elucidate the specific role cytokines play in that pathogenesis of AN.

Declaration of Competing Interest
None.

Table 1
Reporting of confounding factors in the included studies.

Table 1 (continued ) Study Were Age Groups Matched? Smoking Status of Participants Menstruation Status Medication Status Other Intervention Presence of Psychiatric Disorders Presence of Physical Health Disorders Presence of Infectious Disease Presence of Autoimmune or Inflammatory Disease
(continued on next page) K.Maunder et al.

Table 2
Cross-Sectional Study Characteristics.

Table 3
Longitudinal Study Characteristics.