The adolescent onset anorexia nervosa study (ANABEL): Design and baseline results

Abstract

The anorexia nervosa adolescent longitudinal biomarker assessment study (ANABEL) is a 2‐year longitudinal study.

Objective

Evaluate several clinical, biochemical, immunological, psychological, and family variables and their interactions in adolescent onset eating disorders (EDs) patients and their 2‐year clinical and biological outcome. This article illustrates the framework and the methodology behind the research questions, as well as describing general features of the sample.

Methods

A longitudinal study of 114 adolescents with EDs seeking treatment was performed. Only adolescents were selected during 4 years (2009–2013). The variables were collected at different times: baseline, 6, 12, 18, and 24 months of the start of treatment. Diagnoses were completed through the semi‐structured Kiddie‐Schedule for Affective Disorders and Schizophrenia interview.

Results

At baseline, the mean age was 15.11 (SD = 1.36). The mean ED duration was 10 months (SD = 5.75). The mean body mass index was 16.1 (SD = 1.8). The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition diagnosis at baseline for restrictive anorexia nervosa was 69.6%, 17.4% for purgative anorexia nervosa, and 24.3% for other specified feeding disorder. At 12 months, 19.4% were in partial remission, whereas at 24 months, 13.8% had fully recovered and 29.2% had partially recovered.

Conclusions

There was an acceptable physical and psychopathological improvement during the first year of treatment, with recovery being more evident during the first 6 months.

1. INTRODUCTION

Eating disorders (EDs) are severe mental illnesses that frequently begin and develop during adolescence. The impact of anorexia nervosa (AN) on adolescents is especially severe as it is the third most prevalent chronic illness in this age range (Lucas, Beard, O'Fallon, & Kurland, 1991; Seitz et al., 2015). Exhaustive reviews (Steinhausen, 2002; Steinhausen, 2009) concluded that AN prognosis was cause of concern despite improvements made in diagnosis and treatment, although prognoses at younger adolescent ages tend to be better than those at older ages. Patient recovery has an intermediate recovery course (4–6 years) depending on the definition used for the patient recovery criteria relying on the recuperation definition (Strober, Freeman, & Morrell, 1997; Zerwas et al., 2013). Furthermore, the adolescents with AN require early, intensive, and prolonged care (Lock & Le Grange, 2005; Treasure & Russell, 2011). Hence, appropriate treatment designs require investigating the variables that are associated with good outcomes, both in the short and in the long term (Castro‐Fornieles et al., 2007).

Previously, treatment outcome was mainly gauged by the recovery of menstruation cycles, weight gain, and recovery from the psychiatric comorbidity specifically associated with AN. Moreover, the staging of ED in partial and total remission, recovery, and relapse is the result of including clinical severity criteria. Recent studies have shown that when psychological symptoms (behavioral, cognitive, and emotional) are included in the clinical assessment, the patient's follow‐up outcome category can change by up to 20% (Clausen, 2004; Couturier & Lock, 2006; Pla & Toro, 1999). As a result of new research, the criteria and stages of ED are becoming more complex and will incorporate new biomarkers throughout the illness's development (Baker, Brownley, Thornton, Bardone‐Cone, & Bulik, 2018; Marcos, 1997; Nova, Lopez‐Vidriero, Varela, Casas, & Marcos, 2008). Regarding these predictive illness course variables, studies have identified the following factors of interest: (a) psychological symptoms such as severe and persistent body image disturbances, obsessive concerns, and low self‐esteem (Culbert, Racine, & Klump, 2015; Keel, Dorer, Franko, Jackson, & Herzog, 2005); (b) psychiatric comorbidity such as depression, anxiety, and personality disorders (Spindler & Milos, 2007); and (c) family factors associated with the organization and interaction among family members as expressed emotion (Anastasiadou, Medina‐Pradas, Sepulveda, & Treasure, 2014; Fernández‐Aranda et al., 2007; Kyriacou, Treasure, & Schmidt, 2008) and parents' psychopathology (Sepúlveda, Anastasiadou, Del Río, & Graell, 2012; Sepulveda et al., 2012).

Regarding biological variables, little is known about their implication in the illness's etiology and clinical course. In a previous longitudinal study with AN patients, changes in parameters such as ferritin, transferrin, complement factors C3 and C4, and alkaline phosphatase have been reported that may be associated with anabolic/catabolic processes during hospital refeeding and outpatient follow‐up (Nova et al., 2004). Moreover, low levels of T3 have been described as the most important underweight and refeeding progress' biomarker (Aschettino‐Manevitz et al., 2010; Swenne, Stridsberg, Thurfjell, & Rosling, 2009) whereas hypoercortisolemia has been found in anorexia as a response mechanism of adaptation to malnutrition (Misra & Kibanski, 2010). Cortisol level is normalized with the nutrition and, in addition, the increase of this hormone correlates with depressive and anxious symptoms (Lawson et al., 2009).

Malnutrition associated to AN elicits important adaptive biological mechanisms. In this sense, fat mass depletion is associated with alterations in the plasma levels of fat tissue derived hormones (adipokines), such as leptin and adiponectin. These adipokines display high and low values respectively in severe AN patients and tend to normalize their values with weight gain (Brichard, Delporte, & Lambert, 2003; Holtkamp et al., 2004; Pannacciulli et al., 2003). Adipoquines play an essential role in metabolic regulation and energy balance as well as participating in the regulation of the immune system (Birmingham et al., 2003). Likewise, the severity of hematological changes (blood cell counts, leukocytes, and neutrophils) in AN is widely correlated with insufficient body mass index (BMI). The most significant alterations encountered are leukopenia, granulocyte function alterations, and a depletion of T lymphocytes, both helper (CD4+) and cytotoxic (CD8+) T cells, as well as natural killer cells (CD16+/56+). Furthermore, a decrease of the CD4/CD8 ratio, which is considered a good marker of nutritional status, may be observed (Elegido et al., 2017; Marcos, Varela, Santacruz, Muñoz‐Vélez, & Morandé, 1993). However, the interaction between biological processes and psychological/psychiatric development predictors remains unclear.

The study of neurobiological variables, especially regarding the psychopathological state, is of significant interest to AN. Alterations in the neurotransmitters, hormones, and cytokines exhibiting hypothalamic actions may be subsidiary to the patient's malnutrition (Dwarkasing, Marks, Witkamp, & van Norren, 2016). They may be the triggering factors of the patient's anorexia and, therefore, may be associated with personality traits or psychiatric comorbidities such as depression or anxiety prior to weight loss (Connan, Campbell, Katzman, Lightman, & Treasure, 2003). These are AN etiology issues that remain to be answered (Amitani et al., 2013)

The majority of the published studies that have assessed biomarkers in AN patients have compared patients' status before and after weight recovery (Modan‐Moses et al., 2007; Solmi et al., 2015). There are, to our knowledge, very few studies in which biological variables are monitored along the patient's stages of psychological and psychiatric recovery.

The study of anorexia in adolescence allows us to frame the age group in which these disorders usually appear. This would mean being able to establish hypotheses at the beginning of the clinical picture that could be contrasted longitudinally. The shortage of longitudinal studies with a sufficient number of cases to be statistically valid (Jacobi, Hayward, de Zwaan, Kraemer, & Agras, 2004) makes it necessary to increase data that allows a representative spread of the course of the disorder.

The child and adolescent onset AN study (ANABEL) is a longitudinal study, designed to evaluate clinical, familial, social, biochemical, and immunological variables in adolescent onset of AN. Our main objective is to assess the clinical characteristics, prognostic factors, diagnostic specificities, and bio‐psycho‐social interactions over the course of 2 years of specialized treatment in a sufficiently large sample of adolescents diagnosed with an anorexia nervosa and other specified feeding disorder (OSFED). This article describes the methodology, procedure, and assessment and also presents the demographic and clinical characteristics of the sample.

2. METHOD

2.1. Subjects

The ANABEL sample included 114 female adolescents aged 12 through 17 (M = 14.5, DT = 1.5) years diagnosed with restrictive AN (AN‐R), purgative AN (AN‐P), or OSFED, as defined by the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM‐5; American Psychiatric Association, 2013). The inclusion criteria for patients were female sex, age 12 or older and 17 or younger (at the time of first diagnosis), no specific EDs treatment, and presence of AN‐R, AN‐P, or OSFED diagnosis as described by the DSM‐5 at the start of the study.

Adolescents were excluded if they presented diabetes, vaccination 6 weeks before evaluation, diagnosis of psychosis or autism spectrum disorder or if they were not Spanish speakers. During the recruitment period, 116 patients met the inclusion criteria. However, one was excluded because, during the follow‐up, she presented psychotic symptoms and another participant left the study without any specific reason. Both, fathers and mothers were invited to take part in the study. The parents' sample was composed of 88 fathers and 94 mothers.

Taking into account the ED pathology assessed through the Eating Disorders Inventory (Garner, Olmsted, & Polivy, 1983) as the main outcome of the intervention, in a recent meta‐analysis about the predictors of treatment outcome in this pathology, Vall and Wade (2015) found a mean effect size of r = 0.23. Based on that mean effect size, the odd ratio was calculated (OR = 2.35). The G*Power program (Faul, Erdfelder, Lang, & Buchner, 2007) was used in order to calculate the sample size needed to detect this effect, obtaining an estimated sample size of 100. Furthermore, other authors such as Fichter, Quadflieg, and Hedlund (2006) used a similar sample size (N = 103) and found that, at 2 years, the Eating Disorders Inventory results showed differences. Therefore, we can conclude that it is a sensitive measure.

The transition of the drop‐out was 114 (T1)–96 (T2)–87 (T3)–63 (T4)–66 (T5). At the end of the evaluation, the response rate was 60% of the patients.

2.2. Procedure

The participants were recruited by verbal invitation among the patients at the Eating Disorder Unit of the Hospital Infantil Universitario Niño Jesús in Madrid (Day Hospital and Inpatient ward). This Unit and the Immunonutrition group of the ICTAN‐CSIC (GIN‐CSIC) also in Madrid have extensive experience in evaluation and treatment with semi‐structured interviews, clinical scales, and systematic therapeutic interventions. The recruitment period was from 2009 to July 2013.

Participation was voluntary and a signed consent form was obtained from all of the participants and from their parents or legal guardians before their inclusion in the study. This study was approved by the Ethics Committee of the Hospital del Niño Jesús (ref code. R‐0020/08) and the corresponding University Committee (UAM, CEI 27‐673).

We present the clinical and biological outcome of a longitudinal study covering from the onset of the treatment until two follow‐up years. During this period, the more intensive treatment intervention was implemented in severe ED adolescents. The participants were assessed five times during the study, at baseline (T1), at 6 months (T2), again at 12 (T3), at 18 months (T4), and 24 months (T5), at the end of the study. Their parents/guardians were assessed at the start of the study. During the study's follow‐up period, the adolescent ED patients receive a comprehensive and integral treatment program based on medical, psychological, familial, and social interventions in hospitalization, day hospital, and outpatient settings.

Diagnoses were resigned according to DSM‐5 criteria (American Psychiatric Association, 2013) for EDs. Interviews and clinical scales were administered individually to the adolescents and their parents/guardians by experienced child psychiatrists and psychologists with specific training in the semi‐structured interview, as described in the ED Unit of Hospital del Niño Jesús protocol. Diagnose instruments and scales are described below. All participants had been previously diagnosed when recruited for this study. In addition, a blood test was obtained during the course of diagnosis to evaluate the biological markers and all evaluation points. These data were evaluated by the GIN‐CSIC. Biochemical, hematological, and immunological data were compared with the laboratories' reference values from healthy adolescents of the same age range. The analysis of hematological, immunological, and hormonal variables was conducted by GIN‐CSIC.

2.3. Assessment scales and variables

2.3.1. Diagnostic interview

The semi‐structured interview Kiddie‐Schedule for Affective Disorders and Schizophrenia, Present and Lifetime Version (K‐SADS‐PL; Kaufman et al., 1997; Ulloa et al., 2006) was used. It was designed to evaluate psychopathology in children and adolescents, according to Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria in the current and in the past, although in this research, patients were later reclassified according to DSM‐5 diagnoses. The diagnoses were codified as definitely or absent in each moment. The original version had good inter‐rater reliability. In addition, an EDs interview designed at the hospital was used to evaluate EDs specifically.

2.3.2. Psychological and familiar variables

The evaluation protocol is presented in Appendix 1. It is a comprehensive protocol with several scales, 14 for patients and 11 for parents. The description and validation of each instrument can be seen there.

2.3.3. Biological variables

A physical examination and a laboratory analysis of blood markers related to nutritional and immunological status were assessed at all evaluation time points:

1. Anthropometric variables: weight, height, BMI. The BMI (kg/m²) was calculated with the following formula: BMI = weight (kg)/height (m²).

2. Assessment of the patient's menstrual state by specific interview.

3. Blood cell analysis (15 ml): Red blood cell counts and indexes, hemoglobin, and white blood cells and differential were assessed in an automated cell counter at the hospital lab. Blood samples were taken after an overnight fast. A Blood Sampling Questionnaire was previously answered in order to not perform the extraction if the patient presented an ongoing infection or had received a vaccination shot in the previous 6 weeks.

4. Biochemical variables: The following parameters were measured in serum: glucose, urea, uric acid, creatinine, pre‐albumin, albumin, total protein, LDH, GOT, GPT, GGT, alkaline phosphatase, total bilirubin, total cholesterol and its fractions (HDL, LDL), triglycerides, Apo A1, Apo B, ions (Ca, P, Na, K, Cl, Fe), transferrin, ferritin, vitamin B12, folic acid, retinol binding protein with an automated analyzer using colorimetric and nephelometric techniques and by electric potential using selective electrode (Na, K).

5. Immunological variables: Adaptive immunity: (a) lymphocyte subsets in pheripheral blood by membrane marker assessment including CD3, CD4, CD8+, CD19+, CD16 + 56+, CD45RO+, and CD45RA+ (flow cytometry). (b) Inmunoglobulins (IgA, IgG, IgM, IgE; nephelometry). (c) Nonspecific immunity: complement factors C3 and C4, C‐reactive protein (CRP). (d) cytokine levels in serum: IL‐1β, IL‐6, TNF‐α, and IL‐2 (xMAP Technology for immunoassay of multiple analytes).

6. Blood collected in EDTA‐K3 vacumtainers was used for lymphocyte subset analysis. Immediately after collection, 1 ml of blood was mixed with an equal volume (1 ml) of preservative solution (Streck Cell Preservative™ CE. Streck, USA) and sent in cool temperature within 2–6 days to ICTAN laboratory for processing and flow cytometry analysis.

7. Neuroendocrine variables: Serum determination of hormones: free T4, T3, TSH, cortisol, insulin, FSH, LH, oestrogens, testosterone, progesterone, prolactin, leptin, soluble leptin receptor, adiponectin, peptide YY (PYY), by RIA, ELISA, and xMAP Technology for immunoassay of multiple analytes (Millipore).

2.4. Data analyses

The data were archived pseudonymized in a computerized database. Continuous variables have been described using centralization indices: mean, standard deviation, minimum, and maximum. Categorical variables were described by percentages. The normality of the variables was verified by the Kolmogorov–Smirnov goodness‐of‐fit test.

A comparison of means between each variable at different times using the repeated measures analysis of variance was carried out. In order to calculate the size of the effect, the “Cohen d” was used to indicate the magnitude of the effects. The interpretation criteria for the values obtained are: d2 < 0.4: small effect, d ≥ 0.4: moderate effect, d ≥ 0.75: large effect. For multiple comparisons, Bonferroni adjust was used in order to deal with the multiple testing problem. To know the association between the psychological, familial, and neurobiological variables, the Pearson correlation coefficient was calculated for each stage. Separate univariate and multivariate regression models (in successive steps) were used.

During data collection, part of the sample did not complete each assessment. A comparison of respondents in each assessment and nonrespondents in each assessment was performed with the objective of knowing if potential nonresponse bias exists. Considering that data are not available for all time points, we assumed the missing at random assumption. We fitted the mixed models using the maximum likelihood method with Mplus 7, which provides estimates that are valid under the missing at random assumption.

Statistical analyses were carried out using the statistical software SPSS 21 in its version for Windows. Statistical analyses were performed with a significance level of 0.05.

3. RESULTS

3.1. Demographic characteristics

Age, duration of illness, presence of amenorrhea, and socioeconomic status at baseline are shown in Table 1. Mean age was 15.11 (DT = 1.36), and the average of the duration of illness was 10.07 months (DT = 4.5). Seventy‐nine percent presented amenorrhea. The majority of patients (42.6%) belonged to the highest social status, following Hollingshead and Redlich scale (Hollingshead & Redlich, 1958).

Table 1.

Demographic characteristics at baseline

	TCA
	M	SD
Age (11,76–17,59)	15.11	1.36
ED duration (months) (1–36)	9.75	5.71
Months of treatment (0–24)	3.33	4.52
Days of hospitalization (7–91)	37.55	12.16
Lost weight (kg) (1, 2–30)	12.03	5.96

	N	%
Amenorrhea
Yes	91	79.1
No	7	6.1
Irregular menstruation	15	13
Socioeconomic status
I (lowest)	21	18.3
II	7	6.1
III	22	19.1
IV	16	13.9
V (highest)	49	42.6

3.2. Clinical characteristics

The results in each stage are shown in Table 2. The mean BMI at baseline was 16.1 (SD = 1.77), at 6 months it was 18.57 (SD = 1.39), at 12 months it was 18.91 (SD = 2.07), at 18 months it was 18.97 (SD = 2.35), and at 24 months it was 19.41 (SD = 2.57).

Table 2.

Mean and standard deviation of ANABEL biological variables and diagnostic spectra during the period

Variables N = 114	T1 mean (SD) N = 114	T2 mean (SD) N = 96	T3 mean (SD) N = 87	T4 mean (SD) N = 63	T5 mean (SD) N = 66
BMI (Min–Max)	16.1 (1.77) (12.4–21)	18.56 (1.64) (14.45–25.15)	18.91 (2.07) (15.11–24.76)	18.97 (2.35) (15.3–26.77)	19.41 (2.57) (16.17–29.87)

Neurohormones	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Cortisol (Min–Max)	17.25 (5.05) (6–34.3)	13.64 (4.25) (5.1–23.6)	14.65 (4.94) (5.2–35)	13.01 (4.68) (1.47–21.2)	15.08 (5.42) (1.48–27.4)
T3 (Min–Max)	0.98 (0.19) (0.65–1.46)	1.14 (0.21) (0.69–1.59)	1.15 (0.22) (0.65–1.74)	1.13 (0.25) (0.65–1.74)	1.13 (0.22) 0.71–1.82)
T4 (Min–Max)	6.41 (0.86) (4.97–8.95)	6.37 (0.73) (5.11–8.06)	6.4 (0.74) (4.92–8.54)	6.42 (0.97) (4.68–9.54)	6.59 (0.98) (8.36–9.18)

Diagnosis	(N/%)	(N/%)	(N/%)	(N/%)	(N/%)
AN‐R	80 (69.6%)	36 (37.5%)	29 (25.2%)	21 (30%)	14 (21.5%)
AN‐P	20 (17.4%)	1 (1%)
BN	—	2 (2.1%)	1 (0.9%)	1 (1.4%)
BED	—	1 (1%)			1 (1.53%)
OSFED	14 (13%)	49 (51%)	34 (29.6%)	27 (23.5%)	22 (33.8%)
Partial remission		7 (7.3%)	20 (19.4%)	15 (21.4%)	19 (29.2%)
Full remission			3 (2.6%)	6 (5.2%)	9 (13.8%)

Note. AN‐P: anorexia nervosa‐purging type; AN‐R: anorexia nervosa‐restricting type; BED: Binge eating disorder; BMI: body mass index; BN: bulimia nervosa; FED‐NEC: feeding or eating disorder not elsewhere classified; T1: baseline; T2: 6 months; T3: 12 months; T4: 18 months; T5: 24 months.

The results of the neuroendocrine variables showed that cortisol changed during the course of the disease (T1 > T2 < T3 > T4 < T5) while T3 increased for the first 12 months (T1 < T2 < T3) and stabilized later (T4 and T5). On the other hand, similar results were found for T4 during the follow‐up.

Diagnoses during the period of the evaluation are also shown in Table 2. The most common diagnosis during the course of the disease was AN‐R, with a percentage at the baseline of 70%, at 12 months of 25.2% and at 24 months of 21.5%. The amount of OSFED increased over time. At 12 months, 19.4% had partially recovered, at 18 months, 21.4% and, at 24 months, 29.2%. On the other hand, at the last assessment, 13.8% had fully recovered.

The results of psychological and familial variables in each stage are shown in Table 3. At the time of first assessment, the mean EAT‐26 score was 28.67 ± 18.51 (cutoff: 20), the mean Children Depression Inventory was 16.02 ± 9.43 (cutoff: 19), the mean STAIC‐T score was 37.05 ± 10.85 and the mean STAIC‐S score 37.05 ± 10.85. All patients' results reduced during the course of the follow‐up.

Table 3.

Mean and standard deviation of ANABEL psychological and familial variables

Psychological patient	T1 mean (SD)	T2 mean (SD)	T3 mean (SD)	Psychological familial	T1 mean (SD)
CDI (Min–Max)	16.02 (9.43) (1–43)	15.43 (9.98) (0–42)	15.02 (9.30) (1–38)	BDI Father Mother	6.22 (4.62) 9.39 (5.08)
STAIC‐S (Min–Max)	37.05 (10.85) (20–59)	33.92 (9.78) (20–60)	32.5 (9.47) (20–60)	STAI‐S Father Mother	21.68 (9.32) 28.83 (10.47)
STAIC‐T (Min–Max)	40.01 (9.35) (22–58)	37.77 (8.80) (21–56)	36.91 (8.19) (20–58)	STAI‐T Father Mother	16.14 (7.10) 20.75 (9.34)
EAT‐26 (Min–Max)	28.67 (18.51) (1–62)	22.88 (17.80) (1–63)	22.12 (15.85) (0–63)

Note. BDI: Beck Depression Inventory; CDI: Children Depression Inventory; EAT‐26: eating attitudes test; STAIC‐R: trait anxiety for children; STAIC‐S: state anxiety for children; STAI‐S: state anxiety; STAI‐T: trait anxiety; T1: baseline; T2: 6 months; T3: 12 months.

Regarding family variables, across all the scales, mothers had more pathological results. On the BDI scale, the mean for parents was 6.22 (SD = 4.62) and for mothers was 9.39 (SD = 5.08). On the STAIC‐S scale, the mean for parents was 21.68 (SD = 9.32) and for mothers was 28.83 (SD = 10.47). On the STAIC‐T scale the mean for parents was 16.14 (SD = 7.10) and for mothers was 20.75 (SD = 9.34).

4. DISCUSSION

Answering questions about the outcome and heterogeneous course of EDs has become important nowadays due to severity and the prevalence of this disorder among the adolescent population. The main aim of this study is to evaluate the clinical outcome and the prognostic variables in a large enough sample of adolescents with EDs. Most of the recruited patients were in the middle adolescence (14–15 years); this finding is in agreement with previous studies, which shows an increase of seeking treatment during adolescence (Hudson, Hiripi, Pope Jr., & Kessler, 2007). In addition, a high percentage belonged to the highest social status, which is in concordance with the findings of Striegel‐Moore and Bulik (2007) about the commonly presence of AN among individuals of this social status. Moreover, the fact that the adolescent ED had less than 1 year of evolution allowed the study of patients with less psychiatric complications (i.e., comorbidities and personality disorders) and the analysis of the natural history of the ED disorder. The main ED diagnosis of adolescent patients seeking specialized treatment in the studied sample was AN‐R following by AN‐P and OSFED in similar proportion. Eighty‐five percent were inpatients at the time of baseline assessment, which shows a high level of disorder severity of the sample.

The clinical outcome following DSM‐5 diagnosis criteria, nutritional, and psychological assessment shows an acceptable physical and psychopathological improvement during the first year of specialized treatment, the recovery was more evident during the first 6 months. This data are similar to other short‐term outcome studies (Lock, Agras, Bryson, & Kraemer, 2005). Longitudinal studies show similar courses of T3 and cortisol (Neto et al., 2018), in concordance with their dependence of BMI as an adaptative mechanism of malnutrition. The correlation between neurohormones (leptin) as a short‐term prognostic marker depending on the severity of the anorexia will be analyzed.

In the ANABEL study, we followed the patients for 2 years, studying several and varied variables that allowed us to posit a prognosis model. To our knowledge, this is the first study that investigates the relationship between variety of biomarkers and psychological parameters along the clinical course of the ED (Rybakowky, Slopien, & Tyszkiewicz‐Nwafor, 2014). In forthcoming reports, we will present detailed analyses of the interactions among psychological and biological variables.

The strengths of this study include the evaluation from measurements of both self‐report and biomarkers, medical records, and other external sources of information such as parents, which increases validity, as it does not depend only on self‐report measures.

The main limitation of ANABEL is the reduction of the follow‐up sample sizes, which decreases the accuracy of possible estimations. It was difficult to contact the parents whose children did not continue their treatment at the hospital. Several papers are currently in progress to address the questions raised in this study. The use of these data in future research will help with the interpretation of longitudinal studies in EDs.

CONFLICT OF INTEREST STATEMENT

The authors do not have any conflict of interest to declare.

APPENDIX A. EVALUATION PROTOCOL: PSYCHOLOGICAL AND FAMILIAL MARKERS IN THE STUDY

Name of variables	No. of items	Original version. Descriptions and reliability	Spanish version. Reliability
Patients
Obstetric Complications Lewis–Murray Scale of Obstetric Data (Lewis, Owen, & Murray, 1989).	15 items (0–15)	The scale rates 15 complications of pregnancy, labor, and the neonatal period with thresholds for rating them as “definite” or “equivocal.” Subjects were rated as having a “definite” complication if they had suffered at least one significant complication (Lewis et al., 1989).
Vital Events Children's Life Events Inventory (Monaghan, Robinson, & Doge, 1979).	47 items for the adolescent version	Self‐reported scale that includes a set of vital events that may appear in adolescence. Each item has two parts: one to specify the age at which the event occurred, and other for the degree of importance it had (0–10).	Adapted by Mardomingo & Gonzalez (1998).
Eating Disorder Pathology Eating Disorders Examination Questionnaire for Adolescents (EDEQ‐A; Carter, Stewart, & Fairburn, 2001).	36 items (0–36)	Diagnostic items (22 items) from the assessed ED pathology (12 items;i.e., “Over the past 14 days, how many times have you made yourself sick (vomit) as a means of controlling your shape or weight?”). The scale had good reliability (Cronbach's α = 0.79). Higher scores indicate greater eating pathology.	Sepúlveda et al., (2018) translated into Spanish. The questionnaire has good internal consistency.
Eating Attitudes Test (EAT‐26; Garner et al., 1983)	26 items (0–78)	The EAT‐26 measures a broad range of symptoms and concerns and provides a total score for disturbed eating attitudes and behavior, using a 6‐point Likert scale, ranging from “always” to “never” (i.e., “Avoid eating when I am hungry”). An optimal cutoff point of 20 on the total EAT‐26 was used to indicate adolescents with disordered eating behaviors/attitudes.	The Spanish version adapted by Gandarillas et al. (2002) was used. The Cronbach's alpha for the study was 0.94 for clinical sample and 0.86 for general population.
Eating Disorder Inventory (EDI‐II; Garner, 1991)	91 items (different scores by subscales)	It consists 91 items designed for the assessment of attitudinal and behavioral dimensions relevant to ED. Rated on a 6‐point scale (from 1 [never] to 6 [always]) that are divided into 11 subscales (i.e., “I think my hips are too big,”). The first 64 original items are grouped into eight scales and additionally, 27 new items were added to form three more scales. This Inventory has good internal consistency between 0.84 and 0.92 for each scale. Higher scores indicate higher disordered eating attitudes and behaviors.	The version employed for this study was the Spanish version by Garner (1998), which has also good psychometric properties. Alpha coefficients were 0.92 for the total scale and oscillate between 0.63 and 0.88 by subscales (Gandarillas et al., 2002).
Body Dissatisfaction Body Shape Questionnaire (BSQ; Cooper, Taylor, Cooper, & Fairnburn, 1987).	34 items (34–204)	One‐dimensional questionnaire of 34 items with scores between 1 (never) to 6 (always) for each item, in order to measure personal body dissatisfaction, fear of gaining weight and the desire to be thin. A higher score indicates more body dissatisfaction. Internal consistency of 0.98.	The Spanish version adapted by Raich et al. (1996) was used, which had an internal consistency of 0.97.
Depression Level Children Depression Inventory (CDI; Kovacs, 1992).	27 items (0–54)	Self‐report designed to evaluate depressive symptoms in children and adolescents (7–17 years). It is composed of 27 items, with three options (0–2). The three responses correspond to signs of absence, mild, or severe depression symptomatology. Higher scores indicate higher levels of depression (cutoff ≥ 19). This questionnaire has good internal consistency (α = 0.86).	The Spanish version adapted by del Barrio (1999) had good internal consistency (Cronbach's alpha between 0.81 and 0.85).
Anxiety Level State‐Trait Anxiety Inventory for Children (STAIC; Spielberger, Edwards, Lushene, Montuori, & Platzek, 1973).	40 items (State: 20–60) (Trait: 20–60)	A 40‐item scale that assesses the level of anxiety at the time of evaluation (anxiety‐state) and the level of anxiety as a trait (anxiety‐trait). Each item has three options. A higher score indicates higher levels of anxiety. Alpha coefficients oscillate between 0.80 and 0.90 in state subscale and 0.80 and 0.88 in the strait subscale.	The Spanish version adapted by Seisdedos for TEA Editorial (1990) had an internal consistency of 0.89 for state subscale and 0.85 for the trait subscale.
Obsesive Symptoms The Leyton Obsessional Inventory‐Child Version (LOI‐CV; Berg, Rapoport, & Flament, 1986).	20 items	Questionnaire designed to evaluate the presence or absence (using Yes/No response) of a number of obsessive concerns and behaviors. It also includes for each item a scale of 0–3 to determine the interference in the personal functioning caused by each item in which Yes has been answered. The scale had good reliability (Cronbach's α = 0.81).	The Spanish version of the test (Serrano, Barrantes, Doménech, Obiols, & Subirá, 1997) has an internal consistency of 0.90.
Perfectionism The Child‐Adolescent Perfectionism Scale (CAPS; Flett, Hewitt, Boucher, Davidson, & Murray, 2000).	22 items (22–110)	Questionnaire used to evaluate perfectionism in children between 11 and 18 years, based on a multidimensional conceptualization of perfectionism. It has 22 items divided into two subscales: self‐directed perfectionism (12 items) and socially prescribed perfectionism (10 items), using a 5‐point Likert scale (1–5). Internal consistency of 0.85.	The version employed for this study was the Spanish version by Castro et al. (2004), which has an internal consistency of 0.89.
Parental Attachment/Style Inventory for Assessing Memories of Parental Behavior (EMBU‐A; Perris, Jacobson, Linström, Von Knorring, & Perris, 1980).	64 items (64–256)	A 64 item scale designed to measure the parenting style, using a 4‐point Likert scale (1–4). The items are grouped into three subscales: rejection, emotional warmth, and control/overprotection. The psychometric properties are satisfactory.	The Spanish version (Castro, Toro, Van der Ende, Arrindell, & Puig, 1990) presents an adequate internal consistency for the three subscales: 0.89, for rejection; 0.89, for emotional warmth; and 0.77 for overprotection.
State of Change Anorexia Nervosa Stages of Change Questionnaire (ANSOQ; Rieger et al., 2000).	20 items (1–5)	Self‐reported questionnaire based on the stages of change model developed by Prochaska and DiClemente. It was designed to evaluate readiness to recover in anorexia nervosa. The items on the ANSOCQ measure aspects of (a) body shape and weight, (b) eating behaviors, (c) weight control strategies, (d) emotional difficulties, (e) problematic personality, and (f) interpersonal conflicts. Each item has five statements that represent the five stages of change: precontemplation (1), contemplation (2), preparation (3), action (4), and maintenance (5). Internal consistency good, of 0.90 and test–retest reliability over one week was 0.89 (Rieger, Touyz, & Beumont, 2002).	The Spanish version adapted by Serrano, Castro, Ametller, Martínez, and Toro (2004) had excellent internal consistency (alpha of 0.94). Test–retest reliability over 1 week was 0.92 (Serrano et al., 2004). It might also be useful as a predictor of hospitalization (Ametller, Castro, Serrano, Martinez, & Toro, 2005).
Personality Traits Millon Adolescent Clinical Inventory (MACI; Millon, 1993).	160 items format T‐F	Questionnaire created to assess the problems, concerns, and situations faced by adolescents between the ages of 13 and 19. It has 160 items with true or false answers divided in three parts: subscales of personality patterns, subscales of expressed concerns, and subscales of clinical syndromes. Differentiation by age group (13–15 and 16–19). Internal consistency of MACI scales ranged from 0.73 to 0.91.	The Spanish version (TEA Editorial, 2004) had an internal consistency between 0.54 and 0.90.
Personality Traits The Junior Temperament and Character Inventory (JTCI; Luby, Svrakic, McCallum, Przybeck, & Cloninger, 1999)	108 items, format T‐F	Self‐report questionnaire based on the personality model developed by Cloninger. It was designed to measure temperament and character traits and was composed by 108 item, true or false answer. Cronbach's α has been reported in the range 0.44 to 0.77.	The version employed in this study was the Spanish version by Pelaz, Bayón, Fernández, and Rodríguez (2010), which had an internal consistency between 0.42 and 0.76.

Name of variables	No. of items	Original version. Descriptions and reliability	Spanish version. Reliability
Parents
Symptoms Observed by Parents Anorectic Behavior Observation Scale for parents (ABOS; Vandereycken, 1992).	30 items (0–60)	ABOS consists of 30 items in three domains: factor I—eating behavior, concern with weight and foods, and denial of problems; factor II—bulimic‐like behavior; and factor III—hyperactivity. Questions are to be answered “yes” (2 points) or “no” (0 points) if the person is certain of the answer, and “¿?” (1 point) if he or she is uncertain about it. Scores range from 0 to 60. Higher score, greater the patients' pathology. Using a cutoff score of 19, the ABOS has a sensitivity of 90% and a specificity of 89.6%.	This questionnaire has been translated into Spanish (Instituto Nacional de la Salud, 1995), but it has not been validated among Spanish sample.
Child Behavior Checklist Child Behavior Checklist (CBCL; Achenbach, 1991).	113 items (0–226)	Caregiver report designed to measure behavior problem in children between the ages of 6 and 18. It contains 113 items recorded on a 3‐point Likert scale (0–2) divided in eight subscales: withdrawn, somatic complaints, anxious/depressed, social problems, thought problems, attention problems, delinquent behavior, and aggressive behavior. Some of them grouped in second order factors: the first three in internalizing, the last two in externalizing, and the rest of them indicate total problem scores. Cronbach's alpha values ranged from 0.46 to 0.93 on the various subscales.	The Spanish version adapted by Sardinero, Pedreira, and Muñiz (1997) had adequate psychometric properties. The subscales were highly consistent, especially the subscale social competence, with internal consistency between 0.48 and 0.55.
Depression Level Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961).	21 items (0–63)	Self‐reported instrument to evaluate the severity of depressive symptoms during the previous week. Each item has three options scored from 0 to 3. The cutoff points are 0–9 (no depression), 10–18 (low), 19–29 (moderate), and 30 (severe). Internal consistency of 0.92 and test–retest reliability (r = 0.93 over 1 week).	The Spanish version (Vázquez & Sanz, 1991) had high internal consistency (α = 0.89).
Anxiety Level State and Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, & Lushene, 1970).	40 items (State: 20–60) (Trait: 20–60)	A 40‐item scale that assesses the level of anxiety at the time of evaluation (anxiety‐state) and the level of anxiety as a trait (anxiety‐trait). Each item has four options. According to the test–retest correlations provided by Spielberger et al. (1970), the State‐anxiety should have a 0.54 and the 0.86 (trait‐anxiety) correlations.	The Spanish version (Seisdedos, 1990) had internal consistency between 0.83 and 0.92.
Checklist Sympton Check List‐90‐Revised (SCL‐90‐R; Derogatis, 1977).	90 items (0–4)	It consists of 90 questions that gauge nine symptomatic dimensions of psychopathology, using a 5‐point Likert scale (0–5). The Global Severity Index (GSI), calculated by the sum of the all items divided by 90, indicates the level of psychological distress for each individual. Alpha coefficients were 0.92 for the total scale and oscillated between 0.81 and 0.90 by subscales. Higher score means higher psychological distress.	González de Rivera, De las Cuevas, Rodriguez, & Rodriguez (2002) adapted Spanish version. A cutoff point for males is 0.6 and for females 1.12. Internal consistency was 0.75.
Obsessive Symptoms Mausley Obsesive‐Compulsive Inventory (MOCI; Hodgson & Rachman, 1977)	30 dichotomous items	The Maudsley Obsessive‐Compulsive Inventory (MOCI) is one of the most used tests in clinical psychology for assessing the obsessive and compulsive symptoms in psychiatric patients and as a screening tool in nonclinical population. The original version has four subscales: checking (9 items), cleaning (11 items), slowness (7 items), and doubting (7 items). In a meta‐analysis of 51 studies, on average, the internal consistency of test scores was 0.76 for the original 30‐items version (Sánchez‐Meca et al, 2011).	The version employed in this study was the Spanish version by Guilera, Gómez‐Benito, Tomás, and Carreras (2005). They did not report coefficients alpha for the subscales, but in a recent study, Fonseca‐Pedrero, Lemos‐Giraldez, Paino‐Piñeiro, Villazon‐García, and Muñiz (2010) applied the MOCI to a sample of 508 nonclinical children and adolescents, obtaining coefficients alpha of 0.75.
Personality Profile Temperament and Character Inventory‐Revised (TCI‐R, Cloninger, 1999)	240 items	Self‐report questionnaire based on the personality model developed by Cloninger. A 240‐item scale divided into seven dimensions: four related to temperament and three to character. The response option format ranged from 1 = definitely false to 5 = definitely true. The internal consistency ranged from 0.60 to 0.85 for the temperament scales and from 0.82 to 0.87 for the character scales.	Spanish version by Gutiérrez et al. (2004) had an internal consistency of 0.87.
Family assessment
Level Expressed Emotion (LEE; Cole & KazarIan, 1988).	45 items, format T‐F	The LEE consists in 60 items to measure the emotional climate of the home environment for the relative. It is composed of four subscales: (a) intrusiveness, (b) emotional response to the patient's illness, (c) negative attitude towards illness, and (d) tolerance and expectations relating to the patient. Each of these four components included 15 true or false questions (i.e., “I say she lacks control”). Scores are calculated for each scale as well as for a total score. The internal consistency indices for the scales range from 0.84 to 0.95 (Cole & Kazarian, 1988).	Spanish version by Sepulveda et al. (2012) has been shortened to 45‐items, as some of the items did not contribute sufficiently to the total scale. The refined 45‐item LEE‐S scale consisted of four factors. Reliability was acceptable for the subscales ranged from 0.73 to 0.79 and for the total LEE‐S scale was 0.86.
Family Questionnaire (FQ; Wiederman, 2002).	20 items (10–40)	20 items assess the level of expressed emotion with 4‐point Likert, from 1 = very rarely to 4 = very often (i.e.,” I tend to neglect myself because of him/her”). It has two subscales: criticism (CC) and emotional over‐involvement (EOI). A cutoff point for CC is 23 and for EOI is 27. Higher score means higher EE. The internal consistency was 0.92 for the CC scale and 0.80 for the EOI scale.	The Spanish version adapted by Sepúlveda et al. (2014) had internal consistency between 0.83 for the CC scale and between 0.72 for the EOI scale.
Family Adaptability and Cohesion Scale (FACES‐II; Olson, Portner, & Bell, 1982).	30 items	Self‐report questionnaire designed to assess perceived family functioning. It has two subscales: adaptability and cohesion, using a 5‐point Likert‐type scale, ranging from 1 (almost never) to 5 (almost always). Internal reliability of cohesion is 0.87 and adaptability is 0.78.	The version employed in this study was the Spanish version by López (2002), which found internal consistency of 0.78 for cohesion scale and 0.70 for adaptability scale.
Short Marital‐Adjustment and Prediction Test (Locke & Wallace, 1959).	15 items	It consists of 15 questions to measure marital success and adjustment. Different score for each item related to hobbies, sexual relationship, or finances. The range of reliability has been reported as 0.72 to 0.83.	Carrobles (1989) translated into Spanish.

Graell M, de Andrés P, Sepúlveda AR, et al. The adolescent onset anorexia nervosa study (ANABEL): Design and baseline results. Int J Methods Psychiatr Res. 2018;27:e1739 10.1002/mpr.1739