Neutrophils to Lymphocytes Ratio (NLR) Variation in Relationship with Childhood Maltreatment in Patients with Anorexia Nervosa: A Retrospective Cohort Study


 Purpose: Anorexia nervosa (AN) is a serious mental illness that is frequently accompanied with a history of childhood maltreatment (CM) which may constitute a specific ecophenotype of the eating disorder necessitating special assessment and management. Chronic stress inflicted by CM in patients with AN may manifest through increased low grade inflammation reflected by an increase in ratios from white blood cells count such as neutrophils to lymphocytes ratio (NLR) platelets to lymphocytes ratio (PLR) and monocytes to lymphocytes ratio (MLR). Methods: N=206 participants were evaluated in an eating disorders daycare unit in Montpellier, France between March 2013 and January 2020. CM was assessed using childhood trauma questionnaire (CTQ). AN clinical severity was assessed using eating disorder examination (EDE-Q). The MINI was used to assess other clinical characteristics. Results: NLR has been found to be higher in patients with AN and a history of CM (p=0.029) as well as in patients with AN and a history of emotional abuse (p=0.021) when compared to patients with AN and without a history of CM. In the multivariate analysis, emotional abuse (β=0.17; p=0.027) contributed significantly to NLR’s variability. Conclusion: NLR is a low grade inflammation marker that is influenced by multiple sociodemographic, clinical and biological factors in patients with AN. It is more directly affected by certain subtypes of CM, especially emotional abuse than by the presence or absence of a history of CM. Future studies may focus on mediators between CM and increased inflammation such as interoceptive awareness, emotional dysregulation, food addiction and stress-sensitization.Level of evidence: Level III: Evidence obtained from well-designed cohort or case-control analytic studies


Introduction
Anorexia nervosa (AN) is a serious mental illness estimated to affect 0.5%-1% of women during their lifetime and approximately one tenth as many men with a crude mortality rate of approximately 5.6% per decade of illness [1]. It is characterized by a restriction in energy intake, body-image disturbance, undue in uence of body image on self-evaluation, and an intense fear of weight gain [2]. Childhood maltreatment (CM) is the abuse (emotional, sexual or physical) and neglect (emotional and physical) that occurs to children under 18 years of age which results in actual or potential harm to the child's health, survival, development, or dignity [3]. CM is known to be a non-speci c risk factor for eating disorders in general and AN in particular [4,5]. Emotional dysregulation, post-traumatic stress disorder, depression and food addiction have been considered as potential mediators between CM and all subtypes of eating disorders including AN [6-10].
Patients with AN and a history of CM might constitute a different ecophenotype than patients with AN without CM, possibly due to chronic activation of the hypothalamo-pituitary adrenal (HPA) axis [11,12]. As a matter of fact, emerging evidence has demonstrated that childhood trauma exposure has persistent detrimental effects on the HPA axis reactivity to a psychosocial stressor in adults with AN [11]. Moreover, it seems that the ecophenotype of patients with AN and a history of CM has been more exposed to a speci c type of CM such as emotional abuse [11,12]. From another perspective, it is becoming more accepted that stress in uences the HPA axis and autonomous nervous system and, ultimately, the immune system which may increase proin ammatory cytokine levels and lymphocytes activation mainly through norepinephrine secretion, oxidative stress and sleep disturbance [13][14][15]. During the last decade, preclinical and clinical studies have started to emerge in which, aside from HPA axis disturbances and proin ammatory cytokine levels, the neutrophils to lymphocytes ratio (NLR) has been evaluated as a marker of the impact of chronic stress on the immune system [16][17][18]. It appears that chronic stress may increase the NLR due to an increase in neutrophils and/or a reduction in lymphocytes in peripheral blood [16][17][18]. In addition to its possible implication in the in ammatory reaction to chronic stress related to psychological trauma, NLR, platelets to lymphocytes ratio (PLR) and monocytes to lymphocytes ratio (MLR) have all been evaluated as potential markers of low grade in ammation in several psychiatric disorders such as schizophrenia, bipolar disorder, suicidal behavior, opioid use disorder, attention de cit and hyperactivity disorder, etc. [19][20][21][22][23][24][25][26][27].
Patients with AN manifest increased secretion and concentration of the in ammatory cytokines which leads to disturbances in neutrophils and lymphocytes proliferation and activity [28]. However, it is still unclear whether the witnessed disturbances in peripheral white blood cells is related to a primary immunological predisposition in patients with AN or to a non speci c end result due to malnutrition [28]. To date, the only study evaluating NLR in patients with AN has demonstrated that it is inversely correlated to bone mineral density due to in ammation's effect on bone mineralization in adolescent girls [29]. To our knowledge, there is no study that assesses the impact of CM on the immune system of patients with AN as re ected by the NLR. It is hypothesized that the chronic stress in icted by CM in patients with AN may determine a speci c ecophenotype characterized by heightened in ammation markers that might be re ected by disturbances in NLR. The primary objective of this study is to evaluate whether NLR is higher in patients with AN and a history of CM compared to patients with AN without a history of CM. The secondary objective of this study is to explore whether the effect of CM on the NLR in patients with AN manifests speci cally in some types of maltreatment.

Materials And Methods
Participants: All outpatients with AN according to the DSM-5 (Diagnostic and Statistical Manual of Mental Disorders) criteria who were assessed in an eating disorders unit in Montpellier, France, between March 2013 and January 2020 were eligible for the study (30). Patients with AN are referred to this daycare unit for multidisciplinary assessment, diagnostic con rmation, and management. Signed informed consent was obtained from all the participants (or from the parents of underage participants). Inclusion criteria were as follows: age superior to 15 years (15-70 years), speaking French, and having a diagnosis of AN according to DSM-5 criteria. Exclusion criteria were as follows: 1) refusing to consent, 2) having a mental disability such as intellectual de ciency or a psychotic disorder, 3) having a severe physical comorbidity, 4) suffering from an autoimmune disease or a chronic infectious disease that may alter the white blood count. The data utilized here are drawn from two large studies approved by the Ethics Committee of CPP Sud-Mediterranée IV for the rst one (reference number 11-04-SC) and CPP Sud-Est 6 (reference number: AU13-13) for the second study. The two cohorts were recruited in the same way between 2013 and 2017 for the rst one and 2017-2020 for the second one. The study was conducted in compliance with the World Medical Association Declaration of Helsinki.

Measures:
The multidisciplinary clinical assessment was carried out during a full day at the outpatient unit by experienced mental health professionals. The AN diagnosis was established on the basis of a nonstructured clinical assessment by psychiatrists, psychologists, and nutritionists, as well as a structured evaluation with the Mini-International Neuropsychiatric Interview (MINI, Version 5.0.0). Body weight and height were collected in a standardized way during the clinical examination. Fasting peripheral blood samples were collected from the cephalic vein between 8:00 and 9:00 a.m. in EDTA tubes. Complete blood counts were performed using a Sysmex XN-10/XN-20 Hematology Analyzer (Norderstedt, Germany). NLR was calculated by subdividing the total neutrophils count by the total lymphocytes count. In addition, participants completed the following questionnaires: The eating disorder examination (EDE-Q) is a 28-item screening tool used to evaluate ED symptoms. It measures disordered eating over a 28-day period and is scored across four subscales that explore the four core clinical dimensions of EDs: Restraint, eating concern, body shape concern, weight concern [31]. The standardized Cronbach's alpha coe cient was 0.825 for the items of the EDE-Q in this sample (0.861 for restraint, 0.766 for eating concern, 0.902 for body shape concern and 0.834 for weight concern). The Child Trauma Questionnaire (CTQ) is a 28-item self-report instrument for the retrospective assessment of trauma exposure during childhood. The CTQ consists of ve subscales representing different types of trauma (physical abuse, emotional abuse, sexual abuse, physical neglect, and emotional neglect) with multiple items according to a ve-point Likert scale ranging from 1 (never true) to 5 (very often true). A higher score on a subscale indicates more severe childhood trauma [32]. In line with the recommendations from Bernstein and Fink, thresholds or cut-off scores were set for each type of CM at four levels of maltreatment: none, low, moderate and severe. These were further dichotomized as moderate/severe vs. none/low, in order to minimize false identi cation [33,34]. Participants were classi ed in the group with moderate/severe CM [CM(+)] in case they presented at least one moderate/severe trauma subtype or else they have been classi ed in the group CM(-). In this sample, the standardized Cronbach's alpha coe cient was 0.838 for emotional abuse, 0.893 for physical abuse, 0.946 for sexual, 0.898 for emotional neglect and 0.501 for physical neglect. The low Cronbach's alpha coe cient for physical neglect's subscore reduces the reliability of analysis including this CM subtype but we will show them for better comprehensiveness.

Statistical Analyses
Quantitative variables signi cantly departing from normality assumptions (as assessed by Shapiro-Wilk test and quartile-quartile (Q-Q plots) were expressed as medians with interquartile ranges (IQR: Q1-Q3). All the variables, except for body mass index (BMI) were not normally distributed.
In a rst analysis, the NLR has been compared between groups created by categorical variables. Moreover, it has been correlated with continuous variables using Spearman's rho correlation. The bivariate comparison between the groups' characteristics was conducted using the Mann-Whitney U test and Pearson's chi-square test. Nominal variables having categories with small size (N<30) were regrouped by merging categories. Cronbach's alpha was computed for EDE-Q total score and CTQ subscales.
In a secondary analysis, Log(NLR) has been considered as the dependant variable. An independent variable was considered in the nal regression model in case it showed an in uence on NLR in the bivariate analysis (p<0.15). Retained continuous independent variables were classi ed in categories that respected their original distribution in the nal regression models. Independent variables were added and retained in a forward fashion to the regression model while examining the adjusted R 2 of the ANOVA, the variance in ation factor (VIF) and the degree of freedom (df) of the model. Retained independent variables were those that assured the highest adjusted R 2 and df as well as the lowest VIF. For the purpose of demonstrating the study's hypothesis, CM(+/-), emotional abuse and physical neglect were considered to be essential independent variables, each constituting a model of linear regression. Other retained independent variables were age, sex, relationship status and the restriction subscore of EDE-Q.
Age and EDE-Q have been categorized according to quartiles. The Akaike Information Criterion (AIC) was calculated for all models created for the linear regression in order to compare the relative quality of these models according to the following formula: AIC = n x ln (SSres) + 2K. The model with the lowest value of the AIC was considered as the model which best represents the variability of NLR. Finally, all the statistical analyses were conducted using JASP 0.14.1.0 and SPSS 25.0. The rst bivariate analysis corresponded to a comparison of NLR, PLR and MLR between the groups CM (+) and CM (-) (Table 2, Figure 1). Among all three ratios, only NLR was signi cantly different between these two groups (p=0.029), and was the only lowgrade in ammatory marker to be further evaluated in bivariate analysis. Accordingly, NLR was correlated with age (Spearman's rho=0.201, 95% Con dence Interval (CI) = [0.06-0.33], p=0.005) and to the restriction subscore of EDE-Q (Spearman's rho= -0.147, 95%CI = [-5.26x10 -4 -(-0.28)], p=0.049). Moreover, median NLR level signi cantly differed according to the presence or absence of a current history of drug and/or alcohol abuse and/or dependency (p=0.034). NLR also differed signi cantly in participants with and without a history of moderate to severe emotional abuse (p=0.021) and physical neglect (p=0.005) (Figure 1). Finally, NLR did not present a statistical signi cance when compared between both genders (p=0.087), both professional status categories (p=0.087) and between single and married/divorced groups (p=0.123). However, these variables were considered in the multivariate analysis because of their relatively low p value.

Multivariate analysis:
The variability of Log(NLR) has not been found to be affected by the presence of a CM (p=0.076). Nonetheless, both emotional abuse (p=0.027) as well as physical neglect (p=0.008) signi cantly contributed to the variability of NLR. After comparing both models using the AIC, physical neglect seemed to be slightly better than emotional abuse as a subtype of CM explaining the variability of NLR (Table 4). However, due to the internal consistency of the physical neglect subscore of the CTQ, the variability of NLR in due to the presence of a history of emotional abuse has been considered more robust.

Discussion
In this study, we have demonstrated that chronic stress due to CM may leave an in ammatory mark detected in white blood cells count of patients with AN. As a matter of fact, NLR is signi cantly higher in patients with AN and a history of CM compared to patients with AN without history of CM. This nding is in line to what has been recently demonstrated during pregnancy in women who have experienced childhood sexual abuse [35]. In this last study, NLR was signi cantly higher in pregnant women with a history of childhood sexual abuse independently from prenatal depression [35]. The current medical literature points towards a role of epigenetics via DNA methylation on the HPA axis dysregulation and immune system activation after CM [36]. Besides the consistent nding of HPA axis disturbances leading to lower morning cortisol levels and a attening of the diurnal cycle in adults with a history of CM, an accumulating body of evidence supports the fact that the immune system is chronically activated after CM [36 -41]. Classically, C-reactive protein (CRP) and interleukine-6 (Il-6) have been shown to be increased in adults with CM [39][40][41][42]. However, after adjustment on age, sex, marital status and EDE-Q restriction score, the effect of CM on NLR was no more found in our sample. This demonstrates that the global effect of CM on NLR is multifactorial and that the variability of NLR in this population is related to the variability of other signi cant factors such as age, marital status and the level of eating restriction. As a matter of fact, NLR has already been shown to increase with age [43]. According to our data, eating restriction seems to inversely impact NLR variability. Indeed, the quality of consumed food may have an in uence on NLR from one side, and eating restriction may induce cachexia and a subsequent hypercatabolic state from another side [44][45][46][47]. This might explain how eating restriction may exert a complex effect on the immune system independently from affecting the BMI. Nevertheless, the effect of eating restriction on NLR warrants to be studied in future studies. Finally, NLR seems to be in uenced by the marital status with single participants being at a higher NLR than married/divorced ones. This might be in line with what has already been shown regarding the fact that being widowed, separated or never married may be accompanied with a signi cant modi cation in NLR levels [48]. This nding implicates that several lifestyle factors and behaviors related to marital statuses may explain immunologic differences, especially in patients with AN, that warrant further studying.
In our study, emotional abuse was associated with a higher NLR in the bivariate as well as in the multivariate analysis. Childhood abuse, in general, is known to be associated with loss of control on eating and depressive mood as symptoms leading to eating disorder's psychopathology [49]. Although anxiety and depression might be the cause for in ammation in AN, this nding has not been con rmed according to available studies on the subject in the literature [50]. However, the medical literature points towards a certain role of emotional abuse as a risk factor of in ammation. In this regard, it has been demonstrated that women who reported more emotional abuse had greater percent change in CRP over time [51]. The link between emotional abuse and the psychopathology of AN may be related to the fact that patients exposed to emotional abuse may develop emotional dysregulation which inhibits the emotions of anger and disgust as a part of maladaptive attitudes [52]. In a mixed-model investigation using network analysis, the childhood emotional abuse experience has been identi ed as the common node mediating the association between each type of CM and eating disorder's core symptoms [53]. In this same study, patients with AN restrictive type have shown that interoceptive awareness was included in the shortest pathway between emotional abuse and drive to thinness [53]. In the study conducted by Monteleone et al., emotional trauma in patients suffering from eating disorders contributed to the heightened cortisol production when exposed to stress [12]. Moreover, we have previously demonstrated that emotional abuse might be affecting the clinical severity of eating disorders through food addiction [10]. Accordingly, emotional abuse might be a subtype of CM determining a speci c ecophenotype of maltreated patients with AN characterized, on the biological level by an increased HPA axis disturbances as well as an immune activation and on the clinical level by an increased interoceptive awareness, emotional dysregulation and food addiction. In this regard, emotional abuse as a CM may be perceived as a precursor event leading to emotional dysregulation as a coping mechanism which is a risk factor for food addiction and interoceptive awareness all of which being precipitated by HPA axis dysregulation while increasing low grade in ammation at the end of the cascade [54][55][56].
Although physical neglect's subscore presented a low reliability in our study, future studies might be interested in considering it as a factor that in uences the variability of NLR. Participants with physical neglect had higher NLR than those without this subtype of CM. In one study, early institutionalization interacted with stressful life events in early adolescence to predict elevated IL-6 at age 16 among individuals with prolonged institutional care. The study results are in favor of a stress-sensitizing effect of early childhood neglect [57]. According to the stress-sensitization theory, early life deprivation may enhance reactivity to stress in later life which manifests through low grade in ammation [58,59]. An increasing amount of evidence supports the fact that early life neglect in uences the child and/or adolescent's brain development and slows down its maturation [60-63]. As with emotional abuse, we have previously demonstrated that physical neglect may increase eating disorders' severity through food addiction possibly due to brain connectivity dysfunctions occurring early in life [12]. Accordingly, patients with AN and a history of physical neglect may become more sensible to stress from one side because of biological changes occurring to their brain and from another side because of clinical manifestations such as increased internalizing symptoms. This may result in an increased reactivity to stress later in life as manifested by a low grade in ammation and an increase in NLR.
Our study has however several limitations. First, assessments such as EDE-Q, CTQ, MINI were self-reported which might constitute a source of bias. For example, participants with more severe clinical dimensions of AN may more easily recall incidents of CM. Second, the subgroups with physical neglect and abuse have low numbers of patients which reduces the ability to generalize our ndings. Third, the Cronbach's coe cient alpha for physical neglect subtype of the CTQ shows that the participants' answers were not consistent throughout the sample which reduces the reliability of this measure. Fourth, although multiple confounding factors have been considered in the multivariate analysis, the NLR can be affected by multiple clinical and biological determinants of AN and malnutrition so other confounding factors might have been missed in our analysis.
In conclusion, this study demonstrates that NLR is higher in patients with AN and a history of CM in general as well as in patients with AN and a history of emotional abuse in particular. The effect of CM on NLR was not present in the multivariate analysis in which age, marital status and EDE-Q restriction score seemed to in uence NLR variability more signi cantly. However, in the multivariate analysis, emotional abuse contributed signi cantly to NLR's variability. Although physical neglect signi cantly plays the same role, caution has been taken to withdraw conclusions from this nding due to the low reliability of its measuring tool.
Future studies should focus on examining the relationship between emotional abuse as a psychopathological precursor of AN from one part and the mediators between CM and NLR variability from another. These mediators might possibly be interoceptive awareness, emotional dysregulation, eating restriction, food addiction and stress sensitization.
Strength and limits: The study demonstrates that CM in general and emotional abuse in particular may be responsible of an increase in the low grade in ammation marker NLR in AN. The study's limitations, in addition to method's di culties related to measuring childhood events, can be related to the presence of clinical and biological confounding factors in uencing NLR that might have been missed in our analysis. Weight concern Shape concern 3.32 (   Weight concern Shape concern Spearman's rho:  Figure 1 Comparison between NLR levels in patients with AN categorized according to their history status of CM as well as different CM subtypes. *p<0.05