OBJECTIVE: The objective of this study was to compare the medical severity of adolescents who had eating disorders not otherwise specified (EDNOS) with those who had anorexia nervosa (AN) and bulimia nervosa (BN).
METHODS: Medical records of 1310 females aged 8 through 19 years and treated for AN, BN, or EDNOS were retrospectively reviewed. Patients with EDNOS were subcategorized into partial AN (pAN) and partial BN (pBN) when they met all Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria but 1 for AN or BN, respectively. Primary outcome variables were heart rate, systolic blood pressure, temperature, and QTc interval on electrocardiogram. Additional physiologically significant medical complications were also reviewed.
RESULTS: A total of 25.2% of females had AN, 12.4% had BN, and 62.4% had EDNOS. The medical severity of patients with EDNOS was intermediate to that of patients with AN and BN in all primary outcomes. Patients with pAN had significantly higher heart rates, systolic blood pressures, and temperatures than those with AN; patients with pBN did not differ significantly from those with BN in any primary outcome variable; however, patients with pAN and pBN differed significantly from each other in all outcome variables. Patients with pBN and BN had longer QTc intervals and higher rates of additional medical complications reported at presentation than other groups.
CONCLUSIONS: EDNOS is a medically heterogeneous category with serious physiologic sequelae in children and adolescents. Broadening AN and BN criteria in pediatric patients to include pAN and pBN may prove to be clinically useful.
WHAT'S KNOWN ON THIS SUBJECT:
Few studies have focused on any medical sequelae in adolescents with eating disorders not otherwise specified.
WHAT THIS STUDY ADDS:
This study provides evidence that serious medical complications can occur in children and adolescents with eating disorders who do not meet full DSM-IV criteria for anorexia nervosa or bulimia nervosa.
According to current diagnostic criteria, most pediatric patients with disordered eating receive a diagnosis of eating disorders not otherwise specified (EDNOS), defined in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) as “disorders of eating that do not meet the criteria for any specific eating disorder,”1 a categorization that has long troubled practitioners.2–9 Although a handful of studies have examined bone density, fracture rates, or electrocardiograms of patients with EDNOS,10,–,12 most studies on medical sequelae of disordered eating have focused on bulimia nervosa (BN) or anorexia nervosa (AN), and little work has documented the severity, frequency, or clinical significance of EDNOS in young people.4,8,13,–,15
Studies of patients with EDNOS have focused on psychiatric features, comparing adult patients who had partial AN (pAN) or partial BN (pBN) with patients who met full DSM-IV criteria. Patients with pAN and pBN typically have similar psychological profiles to those who meet full criteria for AN and BN, whereas pAN and pBN differ significantly from each other despite that both are subgroups of EDNOS.4,16,–,30
Numerous medical organizations, including the American Academy of Pediatrics, agree that patients with eating disorders and severe malnutrition, bradycardia, hypotension, hypothermia, or orthostasis are critically ill and require hospitalization.31,32 No study has examined how DSM-IV diagnostic criteria correlate with medical severity, and there are no data to validate the commonly held tenet that EDNOS is associated with lower medical severity.
This article reviews current diagnostic criteria and discusses their utility in predicting the medical severity of patients with eating disorders (EDs). Our goal is to describe a large group of pediatric patients with EDNOS and compare them with pediatric patients with AN and BN. In addition, we compare the medical severity of adolescents who have EDNOS with pAN or pBN with those who meet full diagnostic criteria. We predicted that those with EDNOS and pAN or pBN would be less medically compromised than those with full DSM-IV syndromes. Furthermore, we predicted that pAN and pBN would differ significantly from one other with respect to meeting hospitalization criteria.
All 1310 female patients who were aged 8 to 19 years and had been diagnosed with AN, BN, or EDNOS in an academic pediatric ED program from January 1997 through April 2008 were identified. All patients initially received a clinical diagnosis from a board-certified psychiatrist or psychologist with expertise in the assessment of children and adolescents with ED, after diagnostic interviews with both patients and parents or guardians, and as part of a comprehensive evaluation by a multidisciplinary team. Both inpatients and outpatients were included. Because of small within-gender cell sizes that prevented adequate assessment of potential gender differences, male patients were excluded from analyses, as were patients who were found not to have a DSM-IV–diagnosable ED during their evaluation or treatment. A waiver of informed consent and a Health Insurance Portability and Accountability Act–compliant waiver of individual authorization were granted; all data collection protocols were approved by our Panel on Medical Research in Human Subjects and compliant with the Health Insurance Portability and Accountability Act of 1996.
DSM-IV criteria for EDs are guidelines and allow for latitude in their application in clinical settings; however, this study was designed to answer a primary research question of how DSM-IV diagnostic criteria predict medical outcomes. A systematic retrospective review of all medical records was conducted, therefore, by 2 independent assessors and reviewed by the primary investigator to note relevant clinical parameters at presentation. When indicated after this comprehensive review, patients were recategorized from their original clinical ED diagnosis, by using strict DSM-IV criteria (Fig 1). In premenarchal females, AN was diagnosed when weight and psychiatric criteria were met as per DSM-IV guidelines.
Variables and Outcomes
Predictor variables for primary analyses were categorical diagnoses of EDNOS, AN, and BN. To examine each separate criterion for AN and BN in the DSM-IV,1 we further categorized patients with EDNOS into nonoverlapping pAN and pBN categories:
pBN-binge/purge: patients who binge-ate and purged (defined by self-induced vomiting only or laxative abuse) in the month before presentation but with less frequency than defined in the DSM-IV;
pBN-binge only: patients who binge-ate with no purging behaviors, similar to binge-eating disorder but with any level of frequency of binge eating;
pBN-purge only: patients who purged with no binge-eating behaviors;
pAN-low weight/menstruating: patients who met weight criteria for AN but not menstrual criteria;
pAN-low weight/not menstruating: patients who met menstrual and weight criteria for AN but did not openly acknowledge psychiatric criteria, although exhibiting denial of the severity of their underweight along with weight and shape concerns by parental report were sufficient to diagnose a clinical eating disorder;
pAN-<90%: patients who met menstrual criteria for AN and weighed >85% median body weight (MBW) but <90%; or
pAN-25%: patients who were not in other categories of pAN or pBN but had lost >25% of premorbid weight at presentation; the DSM-III suggested that patients with this degree of weight loss be eligible for the diagnosis of AN even if they were not <85% MBW,33 although this convention was dropped for the DSM-IV.
Medical outcome variables are defined in Table 1 on the basis of national guidelines for acute hospitalization of adolescents with EDs.2,32,34 Primary outcomes were heart rate, blood pressure (BP), temperature, and QTc interval. Severe malnutrition was not a primary outcome in this study because pAN and pBN categories were partly defined by weight. Secondary outcome variables included rates of admission within 2 weeks of presentation, length of disease, complications that were attributed to the ED before presentation, and complications that occurred during the first hospital stay if the hospitalization occurred within 2 weeks of presentation. There were no deaths in this series during the first hospital stay.
Heart rates (measured manually) and BPs (using a sphygmomanometer) were taken after lying supine for 5 minutes, and standing heart rate and BP were taken after standing for 2 minutes. When heart rates or BPs were low supine or when significant dizziness was reported, standing vital signs were not obtained. Temperatures were obtained orally by using a digital thermometer. Electrocardiograms were performed by trained staff members by using a standard 12-lead method. Weights were recorded in gowns with no clothing, and heights were obtained by using a stadiometer. Because electrocardiograms and laboratory values were performed clinically rather than as part of a research protocol, the majority of but not all patients had these tests performed (Table 2).
Percentage Median Body Weight
BMI was calculated by using the equation BMI = weight in kg/(height in m)2. MBW was calculated by using gender-specific 2000 Centers for Disease Control and Prevention BMI-for-age growth charts for children and adolescents aged 2 to 20 years (www.cdc.gov/growthcharts). The 50th percentile BMI for exact age at presentation on the Centers for Disease Control and Prevention chart was used to calculate an MBW, together with the height at presentation.
Rate of Weight Loss
Reported maximum weights were extracted from the medical record. Total weight loss before presentation was defined as the maximum weight minus the weight at presentation. Total percentage weight loss was defined as the total weight loss divided by the maximum weight, multiplied by 100. The rate of weight loss was defined as total percentage weight loss divided by the months from the date of maximum weight to the date of presentation. When the maximum weight was the weight at presentation, the total weight loss was 0, as was the rate of weight loss.
Data were described with standard mean and frequency statistics and analyzed by using χ2 testing, Student's t testing, and analysis of variance with Tukey's posthoc comparisons testing on SPSS 17.0 software (SPSS, Inc, Chicago, IL). To guard against type I error in analysis of the primary aims, we used a Hochberg modified Bonferroni procedure.35 To assess further the relationships between primary predictor and outcome variables, we added age and length of disease as covariates by using analysis of covariance.
Demographic and clinical characteristics are presented in Table 2. Table 3 outlines medical differences between DSM-IV ED categories at presentation. The medical severity of patients with EDNOS fell between that of patients with AN and BN in most criteria examined. Differences were statistically significant for all primary outcomes. Pediatric patients with EDNOS had similar age, length of disease, and rate of weight loss as those with AN, but, otherwise, posthoc testing revealed that most differences in secondary outcomes were significant among all 3 diagnostic categories.
Medical outcomes were compared between patients with pAN and AN, pBN and BN, and pAN and pBN (Table 4). Patients with pAN did not differ from those with AN in sexual maturity rating (SMR), but patients with pBN were slightly less pubertally mature than their BN counterparts (SMR breast: 4.5 vs 4.7 [t = 2.6, P < .05]; pubic hair: 4.5 vs 4.7 [t = 2.8, P < .01]). All differences that were noted in primary analyses and detailed in Tables 3 and 4 retained significance after the Hochberg modified Bonferroni correction was applied, except for those related to temperature differences between patients with pBN and BN. Of note, all relationships between primary predictor and outcome variables remained significant after controlling for age and months of disease.
In exploratory analyses, pAN subgroups were compared with AN, and pBN subgroups were compared with BN. Patients with pAN-low weight/menstruating were significantly older, whereas those who did not meet psychiatric criteria were the youngest (15.8 vs 14.3 years: pAN-low weight/not menstruating; 14.8: pAN-<90%; 15.0: pAN-25%; and 15.3: AN [F = 9.7, P < .001]). Patients with pAN-low weight/not menstruating had the longest QTc intervals (394 vs 393: pAN-low weight/menstruating; 386: pAN-<90%; 378: pAN-25% [F = 3.0, P < .05]). The pAN-25% group, despite being nearly at their MBW (97.7% vs 77.7%: pAN-low weight/menstruating; 75.7%: pAN-low weight/not menstruating; 87.2%: pAN-<90%; and 75.8%: AN [F = 198.8, P < .001]), demonstrated the highest percentage of weight lost (34.0% vs 19.6%, 16.6%, 19.2%, and 23.0% [F = 32.9, P < .001]) at the fastest rate (3.5% vs 2.6%, 2.8%, 2.4%, and 2.4%/mo [F = 3.9, P ≤ .005]). They also had higher rates of bradycardia (43.5% vs 22.5%, 28.5%, 28.9%, and 38.5% [χ2 = 14.4, P < .01]) and orthostasis by heart rate (57.1% vs 52.7%, 37.0%, 32.4%, and 32.8% [χ2 = 18.0, P ≤ .001]) than all other pAN subgroups and were more likely than all except patients with AN to meet any admission criteria, excluding weight (76.1% vs 66.7%, 61.5%, 59.5%, and 73.0% [χ2 = 11.8, P < .05]). Patients with AN were most likely to have hypotension (16.2% vs 2.7%: pAN-low weight/menstruating; 6.9: pAN-low weight/not menstruating; 5.0: pAN-<90%; and 6.5: pAN-25% [χ2 = 26.6, P < .001]) and had the lowest phosphorus levels (3.7 vs 3.9, 4.0, 3.9, and 3.8 [F = 3.7, P ≤ .005]). There were no differences noted among pAN subgroups and AN with regard to rates of hypothermia, orthostatic hypotension, hypokalemia, serious hospital complications, or complications before presentation.
Most pBN subgroups did not show significant differences from each other in medical hospitalization criteria, although small cell sizes precluded meaningful analyses of some categorical outcomes. Patients with BN were older (16.4 vs 15.8 years: pBN-binge/purge; 15.6: pBN-purge only; and 15.6: pBN-binge only [F = 6.4, P < .001]) and had disease longer than all pBN subgroups (26.6 mo vs 19.4, 16.5, and 18.7 months [F = 8.5, P < .001]). Patients with pBN-purge only had lost weight faster than patients with BN (2.4% vs 1.5%/month: pBN-binge/purge; 1.9%/month: pBN-binge only; and 1.4%/month: BN [F = 3.0, P < .05]). There were no differences between groups in mean percentage weight loss, BP, orthostatic changes, potassium and phosphorus levels, length of stay if hospitalized, and complication rates.
These analyses reveal that in this adolescent population with ED, 62.4% received a proper diagnosis of EDNOS when current DSM-IV standards were strictly applied; however, 61.6% of these patients with EDNOS met recommended criteria for medical hospitalization and were more compromised than patients with BN in most medical outcomes. Despite their younger age, they displayed similar disease duration and rates of weight loss, QTc prolongation, orthostasis, and hypokalemia as their full diagnostic counterparts. This is despite that they weighed significantly more than patients with AN. These results do not support our initial hypothesis that EDNOS would be less medically severe than AN or BN.
We proposed new groupings of patients with pAN and pBN within the EDNOS group, with each subgroup directly challenging 1 DSM-IV criterion for AN or BN. When patients with pAN were compared with those with AN, there were few differences. Patients with pAN as a whole were less likely to have a low heart rate or BP but did not differ from patients with AN on most other medical outcomes. Adolescents with pAN were younger and weighed significantly more but had lost weight more rapidly than those with AN and had a shorter disease duration.
Of pAN subgroups, patients who had EDNOS and had lost >25% of their premorbid body weight (pAN-25%) seemed more compromised than other subgroups of pAN and even more than patients with AN in some medical outcomes. This is the case despite being at a significantly higher, near “ideal” body weight, reminding us that malnutrition is a complex disease with manifestations at multiple weights. In addition, those with pAN-low weight/menstruating were older, possibly indicating later recognition of the ED. Patients with pBN were younger, had a shorter duration of disease, weighed less, and had lost weight more rapidly than their BN counterparts; however, patients with pBN and subgroups did not differ significantly from adolescents with BN on most other medical outcomes examined.
When pAN was compared with pBN, patients with pAN had a more medically severe condition, with the exception of duration of illness and the QTc interval: patients with pBN had more months of disease and longer QTc intervals. This mirrors our comparison of BN with AN: patients with BN report nearly twice the duration of disease and longer mean QTc intervals. Patients with pAN and pBN were similar only in rates of hypokalemia, hypophosphatemia, and orthostasis. This lends credence to the idea that EDNOS is too heterogeneous a category, because patients with this diagnosis differ more from each other than they do from AN and BN, respectively. Patients who have EDNOS and narrowly miss criteria for AN and BN are often medically compromised and in need of treatment.
To our knowledge, this is the first published comparison of reported complications among adolescents with ED from all DSM-IV diagnostic groups. Although patients with AN certainly had a high rate of objective medical complications observed during their first hospital stays, the complication profiles of other patients were hardly reassuring. Patients with pAN and pBN also displayed high rates of hospital complications at ∼18% and 19%, respectively, and patients with BN and pBN reported significantly higher numbers of serious complications before presentation than their peers with AN and pAN. Although additional prospective study is required to confirm these findings, they suggest the need to delineate better the predictors of complications and medical protocols in each DSM group separately, rather than measuring each group against an AN standard.
Limitations of this study include that it is a clinical sample from a subspecialty ED program, which limits its generalizability. It is also an exclusively female sample, and although it is critical that we learn better how to treat male adolescents with EDs, this study does not inform that pursuit. Data were collected retrospectively; therefore, data may be missing for nonrandom reasons not yet identified. In addition, clinical decisions that influenced the choice of laboratory tests had been made, which may have introduced bias on the basis of medical severity. In general, most variables were missing <10% of data, but phosphorus levels, electrocardiograms, and orthostatic testing were missing for 10% to 20% of patients, thereby necessitating caution in the interpretation of these variables.
A limitation of any study of current medical hospitalization criteria for patients with ED is that they were derived from expert consensus and not from longitudinal study. Bradycardia, hypotension, orthostasis, and hypothermia have clearly been shown in studies to be strong indicators of a malnourished state and have therefore been adopted as indicators of medical severity in patients with EDs.31,32 In addition, QTc prolongation has been shown to be a risk factor for sudden cardiac death,36 which makes it the most concerning complication of the ones examined here. However, we do not have evidence that these findings mandate hospitalization and are not certain that hospitalization improves long-term medical outcomes. It is possible that in the future, outpatient treatment regimens may prove to be equally effective and safe in treating these cardiac sequelae, and additional prospective study is urgently needed to delineate the most appropriate type of interventions and when they are indicated.
These analyses reveal that adolescent patients with ED exist within a larger EDNOS group and are medically similar to patients with AN and BN. They provide a rationale to consider changes to the diagnostic criteria for adolescents with ED, as other authors have proposed.* For example, cut points of weights, duration of behaviors, and endocrine dysfunction are not currently evidence-based and thus may not be truly reflective of medical severity.47 Our study also suggests that current criteria for medical intervention may be most appropriate for adolescents with AN but that we may miss critical opportunities for intervention and prevention in other ED groups.
Finally, our data propose another possibility of diagnostic groupings, which are shown in Fig 2, illustrating the original percentage of patients in AN, BN, and EDNOS categories and comparing that with a new grouping in which pAN and pBN are counted as a subgroup of AN and BN, respectively. If patients with pAN and pBN are combined into AN and BN groups, then only 14.3% of patients with “true” EDNOS remain, similar to another diagnostic reclassification of adult patients with ED.38 If >60% of patients have EDNOS by DSM-IV criteria, then they are effectively forced into a diagnostic category that lacks definition, health care coverage, or medical knowledge.
In adolescents and children especially, EDs are a devastating set of diseases with multiple long-term sequelae. It is clear that a diagnosis of EDNOS does not imply a reassuring medical profile, and these findings underscore the need to intervene early, even when young patients do not meet full diagnostic criteria for AN or BN. Future studies should be directed toward better defining the best clinical criteria by which we can intervene both medically and psychiatrically in these diverse set of illnesses.
The project was funded in part by the Stanford Child Health Research Program; Ms Wilson received funding from the Stanford Medical Scholars Research Program, and Dr Lock received funding from National Institutes of Health grant K24 MH074467.
We gratefully acknowledge Dr Iris Litt for support and help with editing of the manuscript and all of the research assistants at the Stanford WEIGHT laboratory who assisted with data collection for this study.
- Accepted January 7, 2010.
- Address correspondence to Rebecka Peebles, MD, Stanford University School of Medicine, Division of Adolescent Medicine, 1174 Castro St, Suite 250A, Mountain View, CA 94040. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- EDNOS =
- eating disorders not otherwise specified •
- DSM-IV =
- Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition •
- BN =
- bulimia nervosa •
- AN =
- anorexia nervosa •
- pAN =
- partial anorexia nervosa •
- pBN =
- partial bulimia nervosa •
- ED =
- eating disorder •
- MBW =
- median body weight •
- BP =
- blood pressure •
- pBN-binge/purge =
- partial bulimia nervosathat does not meet binging and purging frequency criteria •
- pBN-binge only =
- partial bulimia nervosa, binging but not purging •
- pBN-purge only =
- partial bulimia nervosa, purging but not binging •
- pAN-low weight/menstruating =
- partial anorexia nervosathat does not meet menstrual criteria •
- pAN-low weight/not menstruating =
- partial anorexia nervosathat meets weight and menstrual criteria •
- pAN-<90% =
- partial anorexia nervosa, 85% to 90% of median body weight •
- pAN-25% =
- partial anorexia nervosa, lost >25% of premorbid weight at presentation •
- SMR =
- sexual maturity rating
- 1.↵Diagnostic and Statistical Manual of Mental Disorders, 4th ed., Text Revision (DSM-IV-TR). Arlington, VA: American Psychiatric Association; 2000
- Panagiotopoulos C,
- McCrindle BW,
- Hick K,
- Katzman DK
- Grilo CM,
- Pagano ME,
- Skodol AE,
- et al
- Keel PK,
- Gravener JA,
- Joiner TE Jr.,
- Haedt AA
- Rome ES,
- Ammerman S,
- Rosen DS,
- et al
- 33.↵Diagnostic and Statistical Manual of Mental Disorders. 3rd ed. Arlington, VA: American Psychiatric Association; 1980
- Hochberg Y
- Copyright © 2010 by the American Academy of Pediatrics