Article Text

Download PDFPDF

Medical instability and growth of children and adolescents with early onset eating disorders
  1. Lee D Hudson1,2,
  2. Dasha E Nicholls2,
  3. Richard M Lynn3,4,
  4. Russell M Viner1
  1. 1Department of General and Adolescent Paediatrics, UCL Institute of Child Health, London, UK
  2. 2Department of Feeding and Eating Disorders, Great Ormond Street Hospital, London, UK
  3. 3Research Division, Royal College of Paediatrics and Child Health, London, UK
  4. 4Centre for Epidemiology, UCL Institute of Child Health, London, UK
  1. Correspondence to Lee D Hudson, UCL Institute of Child Health, General and Adolescent Paediatrics, 30 Guilford Street, London WC1N 1EH, UK; l.hudson{at}


Objective Little is known about the physical burden of early onset eating disorders (EOEDs). Most published data on physical instability and growth in malnutrition come from specialist centres, or from the developing world where aetiology differs. The authors present data on physical status at presentation from population-based surveillance systems in the UK and Ireland.

Design Prospective surveillance study.

Participants All suspected cases of EOED in children under 13 years of age reported by paediatricians and psychiatrists via the British Paediatric Surveillance System (BPSU) and Child and Adolescent Psychiatric Surveillance System (CAPSS) in the UK and Ireland from March 2005 to May 2006 (15 months).

Results 208 cases were identified (24% reported by paediatricians). Median age was 11.8 years (IQR 1.74). 171 (82%) were female (78% premenarcheal and 60% prepubertal). 74% of males were prepubertal. 35% of cases had medical instability at presentation (60% bradycardia, 54% hypotension, 34% dehydration, 26% hypothermia). 52% of cases required admission at diagnosis (73% to a paediatric ward). 41% of cases with medical instability were not underweight, that is, they had body mass index (BMI) z-scores above −2.0 (2nd centile). Sensitivities for identifying medical instability with BMI z-score <−3 or 70% median BMI were 31% and 15%, respectively. Menarcheal status did not predict risk of medical instability.

Conclusions EOEDs present with severe levels of physical instability and frequently to paediatricians. As anthropological indices alone are poor markers for medical instability, clinical assessment is essential. Doctors providing care for children have a central role in both the recognition and management of EOEDs.

View Full Text

Statistics from


Early onset eating disorders (EOEDs), defined here as onset before 13 years of age, are known to cause profound levels of medical instability secondary to malnutrition1,,3 and impact upon growth and development.4 Psychological pathologies such as EOEDs have increased as a proportion of the burden of paediatric illness in developed countries as part of the epidemiological transition from communicable to non-communicable childhood disease.5 It is vital that those involved with the healthcare of children and young people are able to recognise and respond to these emerging challenges.

Although there are a number of consensus guidelines to assist the management of medical complications in children presenting with EOEDs, research is lacking.6,,8 Much of the literature on underweight in this age group comes from resource poor settings9,,11 or from patients with malnutrition secondary to physical, often chronic, illness.12,,14 Such groups are likely to be poor models for underweight in eating disorders, with malnutrition in the developing world usually due to reduced access to food, set upon a background of poorer living standards and greater co-existing childhood morbidity. In chronic illness, nutritional requirements may be increased due to the demands of the condition itself, and intake can be limited by the condition. The covert nature of eating disorders and associated behaviours to avoid food and weight gain can lead to a delay in recognition,15 increasing the risk for physical consequences.

What is already known on this topic

  • Early onset eating disorders (EOEDs) are known to present with medical instability, but physical status in the UK has until now remained unknown.

  • Previous studies have suggested greater potential risk for medical instability in those presenting younger and at earlier pubertal stage.

What our study adds

  • Children with EOEDs present with a high level of medical instability.

  • More than 40% of cases with medical instability in our study had body mass index z-scores above −2 (2nd centile).

  • Cases of EOED presenting before menarche were no more likely to have medical instability than those presenting after menarche.

To address gaps in knowledge about EOEDs, we conducted a prospective surveillance-based study in the British Isles to estimate incidence, describe physical and psychological burden at presentation across the spectrum of EOEDs, and identify services issues (including time to presentation, whom patients presented to and initial management). The principal psychiatric findings from this study are published separately.16 Here we present data on medical status at presentation and initial management.


Case ascertainment

Cases of EOED between the ages of 5 and 12 years (inclusive) were ascertained over a 15-month period (March 2005 to May 2006) in the UK and Ireland by two routes. Paediatricians were asked to report possible cases via the established card-reporting surveillance system of the British Paediatric Surveillance Unit (BPSU) in the UK and Ireland,17 while child and adolescent psychiatrists were asked to report cases via the Child and Adolescent Psychiatric Surveillance System (CAPSS), a similar but separate surveillance system to the BPSU, established for the purposes of this study and described in detail elsewhere.18 Instructions guiding notification of potential cases were provided and are listed in figure 1A. When possible cases were notified, data sheets were sent to reporting paediatricians and psychiatrists requesting more detailed information. This information was used to assess diagnostic eligibility for inclusion and to provide information regarding physical and psychological status at presentation, and initial management.

Figure 1

(A) Instructions to clinicians for notification of potential cases and (B) modified DSM-IV and ICD-10 diagnostic criteria used to classify cases into type of early onset eating disorder. BMI, body mass index; EDNOS, eating disorder not otherwise specified; EOED, early onset eating disorder; IBW ideal body weight.

Case definitions

Returned data forms were reviewed for confirmation of an EOED, allowing inclusion as a case. Confirmation was established using modified DSM-IV and ICD-10 diagnostic criteria and cases classified as to type of EOED (see figure 1B). Cases not meeting criteria and cases with insufficient information for classification were excluded. Duplicate reporting was encouraged, and where duplicate cases existed, those with the most complete data sets were included. Medical instability was defined as a reported presence of one or more of the following: hypotension (systolic blood pressure <80), bradycardia (<50 bpm), hypothermia (temperature <35.5°C) or any additional clinical finding related to instability (eg, signs of poor peripheral perfusion or dehydration). Clinicians were asked to report female cases as either pre- or postmenarcheal and also to report pubertal stage as pre, early/mid or late/adult stage for both sexes. As many reporting physicians were likely to be psychiatrists not necessarily trained in pubertal assessment, it was thought unrealistic to request Tanner staging. Cases reported as postmenarcheal were defined as late/adult pubertal stage regardless of the clinician report of pubertal stage. Amount of weight loss for each case was calculated using known premorbid weight (if available). Rate of weight loss was derived from weight loss divided by reported duration of symptoms.


Of a total of 505 notifications, 208 discrete cases were analysed after removal of duplicates, reporting errors and cases with insufficient data. Paediatricians reported 24% of all notifications and psychiatrists 76%. Overall, 19% of final cases were reported by paediatricians, 65% were reported only through CAPSS and 16% were reported by both. Data were not available in all domains for all cases; information on the actual number of cases for each domain analysed can be found in tables 13.Body mass index (BMI) was calculated and z-scores for height, weight and BMI were calculated using UK 1990 growth reference data.19 ,20 We also calculated each subject's % median (or expected) BMI for age and sex (%BMI),21 commonly used in clinical practice as an indicator of underweight.4 ,8 ,22

Table 1

Data for age, sex, auxological parameters, medical instability and duration of illness of different eating disorder categories

Table 2

Auxology and medical instability by pre or postmenarcheal status in female patients

Table 3

Data on age, sex, auxological parameters and duration of illness for all cases, and cases with and without medically instability

Statistical analysis was performed using SPSS version 18. Anthropometric data were mostly skewed towards lower values and were analysed using the Mann–Whitney U test. Normally distributed data were analysed by t test. ORs were calculated to identify associations between categorical variables using logistic regression.


Patient characteristics

We have previously reported the age, sex and disease type of reported cases.16 In summary, the median age of cases was 11.8 years (IQR 1.74). There were no cases of EOED aged 5 or below. In total, 171 (82%) cases were female. No deaths were reported during the surveillance period. The estimated incidence of EOED in the UK and Ireland was 3.01/100 000 (95% CI 2.6 to 3.5). In order of frequency, 83 (40%) of cases were classified as eating disorder not otherwise specified (EDNOS), 76 (36%) as anorexia nervosa (AN), 40 (19%) as ‘other’, six (3%) as binge eating disorder and three (1%) as bulimia nervosa. Note that ‘other’ cases were characterised by significant weight loss and determined food avoidance in the absence of weight and shape concerns.

Data on age, sex, auxology, medical instability and duration of illness by eating disorder category are shown in table 1. By definition, cases with a diagnosis of AN had lower BMI z-scores (p<0.01, z=−10, r=−0.83) and %BMI (p<0.01, z=−10, r=−0.86) at presentation compared to those with a diagnosis of EDNOS. BMI and %BMI were also lower in those with ‘other’ classification, albeit by small amounts (−2.0 vs −2.3 BMI z-score; 78 vs 80% mean %BMI; and effect sizes derived in analyses were small: r=0.2). AN cases did not differ from other disorder types in age, duration of illness or rate of weight loss. Height z-scores were lower in AN than EDNOS (difference in means 0.41, 95% CI 0.05 to 0.76, p=0.03).

Cases with AN were more likely to have medical instability compared to those with EDNOS (OR 2.8, 95% CI 1.4 to 5.8, p<0.01) and cases of ‘other’ (OR 3.4, 95% CI 1.4 to 8.2, p=0.01). Risk for hypothermia was greater in cases with AN compared to EDNOS (OR 4.78, 95% CI 1.27 to 17.90, p<0.01) and risk for bradycardia was greater in cases with AN compared to ‘other’ (OR 4.03, 95% CI 1.27 to 12.82, p<0.01).

Growth and pubertal status

Twenty-five males (74%) were prepubertal and nine (26%) in early/mid puberty; none were late/adult stage. Among females, 98 (60%) were prepubertal, 48 (30%) in early/mid puberty and 16 (10%) at late or adult stage. Most females were premenarcheal (n=131, 78%).

Auxology and medical instability by menarcheal status in females is shown in table 2. Premenarcheal cases were significantly more underweight than postmenarcheal cases, but were no more likely to have medical instability.

Medical status at presentation

Information regarding the presence or absence of medical instability was available for 185 (89%) of all cases. Of these 185 cases, 65 (35%) were reported as medically unstable at presentation. In the 65 cases with medical instability, 60% (39) were bradycardic, 54% (35) hypotensive, 34% (22) dehydrated, 26% (17) hypothermic and 11% (7) poorly peripherally perfused. In the 65 cases with medical instability, 49% (32) had only one feature, 28% (18) had two features, 15% (10) had three features and 8% (5) had four features. In addition, 22/208 (11%) were reported as having lanugo hair (10 of those in the medical instability group, 12 of those stable); five cases had reported damaged skin.

Full anthropometric data were available on 60 of the 65 cases with medical instability: 41 (69%) had a BMI z-score above −3 (the WHO cut-off for severe thinness and thinness grade 3),22 36 (61%) were above −2.67 (0.4th centile) and 24 (41%) were above −2.0 (2nd centile; WHO cut-off for thinness and thinness grade 2). Expressed as %BMI, 51 (82%) of those with medical instability were greater than 70%BMI, 30 (48%) greater than 80%BMI, 12 (19%) greater than 90%BMI and 3 (5%) greater than 100%BMI. Sensitivities and specificities of using z-scores to identify medical instability were 31.7% and 93%, respectively, for below −3 (WHO severe thinness) and 60% and 71%, respectively, for below −2 (WHO thinness). Sensitivities and specificities were 15% and 97%, respectively, below 70%BMI; 50% and 72.3%, respectively, below 80%BMI; and 80% and 55.5%, respectively, below 90%BMI.

Data on age, sex, auxology and duration of illness for cases with and without medical instability are shown in table 3. Cases with medical instability had lower raw BMI (p=0.01, Z=−3.26, r=−0.25), %BMI (p<0.01, Z=−4.24, r=−0.33) and BMI z-scores (difference in means 1.0, 95% CI 0.60 to 1.44, p<0.01). The reported rate of weight loss was greater in cases with medical instability (p=0.02, Z=−2.36, r=−0.3).

Initial admission and management

Information on hospital admission was available for 204 cases. One hundred and four (51%) cases required admission at or shortly after diagnosis, of whom 76 (73%) were admitted to a paediatric ward, and the remainder to inpatient psychiatric units. Those with medical instability were more likely to be admitted to hospital than those without medical instability (86% vs 36%; OR 11.3, 95% CI 5.1 to 25.0, p<0.01). Cases of AN were more likely to be admitted to hospital than cases of EDNOS (71% vs 38%; OR 3.8, 95% CI 2.0 to 7.5, p<0.01) or cases of ‘other’ (71% vs 44%; OR 3.1, 95% CI 1.4 to 7.0, p<0.01).


Patients with EOEDs in the UK present with severe levels of physical instability, with over a third of cases presenting with features of medical instability at the time of diagnosis. EOEDs are therefore a significant burden for paediatric services, especially given that the overall incidence of EOEDs is comparable to that of key paediatric conditions such as meningococcal disease in the same age group.23

Our study highlights that BMI indices alone are poor proxies for medical instability in EOEDs. Over 40% of those with medical instability had BMI z-scores above −2 (2nd centile), the WHO's cut-off z-score for thinness (also defined as grade 2 thinness) and nearly 70% were above −3, the WHO cut-off for severe thinness (grade 3 thinness).22 Both thinness z-scores and %BMI have poor sensitivity for instability, although specificity is relatively high for lower z-scores (<−3) and %BMI (<70%). Rate of weight loss was greater in cases with medical instability, where interestingly the average rate of weight loss (290 g/week) was less than the commonly advised concern-threshold of 1 kg/week.24 However, rate of weight loss was calculated linearly when in reality it may have varied, and information was available for less than a third of cases. Overall our findings indicate that a thorough clinical assessment, including cardiovascular observations, is essential in any child/young person presenting with weight loss.

A unique feature of our study is the large number of premenarcheal females. Two previous smaller studies of children with EOEDs found greater degrees of weight loss in children at earlier pubertal stage, concluding that prepubertal cases were at higher risk of medical instability.25 ,26 Another study found that younger patients with eating disorders had increased rates of weight loss, leading to lower presenting weights, putting them at higher risk of compromise.27 In our study, premenarcheal females had lower body weight at presentation, but were not more likely to have medical instability. Premenarcheal cases also had lower height z-scores. No information was available on premorbid heights or parental heights, nor were follow-up growth data available, so it is not possible to comment on whether this represented temporary delay or stunting. However, these are concerning findings given the evidence suggesting that patients with EOEDs may fail to reach their adult genetic height potential,28 ,29 and highlight the importance of regular growth and developmental assessments of all children and young people with eating disorders who have not reached adult maturity. In our sample, median height z-scores, regardless of BMI status, were close to 0 with similar IQRs, implying preserved height at time of diagnosis.

The ‘other’ group is of particular interest from a medical perspective. Variously described as ‘non-fat phobic anorexia’30 or ‘food avoidance emotional disorder’,31 this group provides an opportunity to differentiate the effect of underweight alone from underweight in AN. Both BMI z-scores and %BMI in AN were significantly lower at presentation compared to ‘other’ cases, and risk of medical instability was also greater for AN. The implication is that behaviours associated with AN may contribute both to lower BMIs and to medical instability. This may reflect chronicity of underweight, more rapid weight loss at various time points, or be secondary to specific behaviours and food restrictions. The atypical clinical features of the ‘other’ and EDNOS groups mean that medical instability may be missed.

Our study is limited by incomplete information for all cases; for example, medical stability parameters were missing for 10% of cases and full anthropometry was only available for 90% of those. Limited clinical data likely reflect the profession as well as the experience of the reporting clinician. The study is not a case–control study, so deriving statistical associations versus the broader population is not possible. Cases were ascertained from secondary care and did not include possible cases looked after in primary care. As such, our incidence may underestimate overall incidence and over-represent sicker patients. It is likely, however, that most children with eating disorders in the UK would be referred to a child and adolescent psychiatrist or paediatrician rather than be managed in primary care alone. The incidence of EOEDs in another epidemiological study from primary care was lower than in our study,32 supporting this view.

Our data emphasise the central role of paediatricians in managing EOEDs. While eating disorders in children and adolescents in the UK are managed primarily by child and adolescent mental health services, 73% of cases requiring hospitalisation were admitted to a paediatric ward and 24% of all cases were notified by paediatricians. Furthermore, paediatricians and general practitioners may encounter such patients at a time when the diagnosis is still unclear. Skills in identifying and managing the medical aspects of eating disorders, working alongside mental health services, are therefore essential for all doctors. Guidelines such as the Royal College of Psychiatrist's Junior MARSIPAN report8 are an important resource. Medical aspects of EOED should form a core part of training and continuing professional development for paediatricians, and be considered a routine component of paediatric service provision.


The authors wish to thank the BPSU, supported by the Department of Health, as well as reporting clinicians.


View Abstract


  • Contributors DN conceived the study; RL handled BPSU and CAPSS management and facilitation; DN, RV and RL undertook case selection and methodology; and LH undertook the statistical analyses and led the writing of this paper. All four authors contributed to the final manuscript.

  • Funding The Harold Hyam Wingate Foundation provided funding for this study.

  • Competing interests None.

  • Ethics approval The study was approved by the BPSU Executive and the National Research Ethics System. As BPSU studies have PIAG section 12 exemption, consent for notification was not necessary from children or families.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.