Analysis

Analyses were primarily descriptive. Where duplicates were reported, data were used from the most complete data-set, or the data-set closest to the reporting date. Incidence was calculated as number of cases/total children of that age in the population (UK plus ROI) expressed as per 100 000 population, using data for the total population of children for 2005 aged 5–13 obtained from the Office of National Statistics (UK) and the Central Statistics Office (Ireland). Nutritional status was determined by calculating body mass index (BMI) from weight and height, then converting to % of median BMI for age and gender using UK BMI reference data. ^{Reference Cole, Freeman and Preece21} Underweight was defined as <85% BMI for age and gender, based on the criteria used for diagnosis of anorexia nervosa ¹⁸ and the World Health Organization definition of malnutrition. ²² To investigate the validity of eating disorder diagnoses in this age group, principal-components analysis was used to identify clusters of symptoms, using eigen values greater than one as a threshold. Analyses were performed using Microsoft Excel and SPSS version 15 on Windows.

Incidence

Notification of 208 cases during the study period provides an overall incidence estimate of 3.01/100 000 (95% CI 2.6–3.5).Figure 2 shows incidence by age for the overall group and for each eating disorder category. There was a clear relationship between incidence and increasing age. There were no cases identified for individuals aged 5 years. The overall incidence age 6–7 years was 0.21/100 000 (95% CI 0.01–0.88); at age 7–8 years was 0.21/100 000 (95% CI 0.01–0.89); at 8–9 years was 0.92/100 000 (95% CI 0.28–1.66); at 9–10 years was 1.64/100 000 (95% CI 0.89–2.84); at 10–11 years was 3.56/100 000 (95% CI 2.3–4.99); at 11–12 years was 4.46/100 000 (95% CI 3.12–6.17); and at 12–13 years was 9.51/100 000 (95% CI 7.5–11.9).

Using the operational definitions in the Appendix, 76 individuals were classified as having anorexia nervosa (overall incidence 1.09/100 000); 3 with bulimia nervosa (incidence 0.04/100 000); and 89 with eating disorder not otherwise specified (EDNOS) (incidence 1.22/100 000), of which 6 showed binge eating but no compensatory behaviours, hereafter called binge eating disorder (incidence 0.09/100 000) (Fig. 2). Forty individuals met the case definition, but did not have weight and shape concerns or fear of weight gain (defined as ‘other’; incidence 0.58/100 000).

Demographics and medical status

Of the 208 individuals identified, 171 (82%) were female and 37 (18%) male. Of the anorexia nervosa group 12% were boys, rising to 18% of the EDNOS group and to 33% of the ‘other’ group. In total, 86% were White British, 2.5% Irish, 5% Asian/Asian mixed ethnicity and 1% Black/Black mixed ethnicity. The mean age was 11.5 years (s.d. = 1.3), with a reported mean duration of illness of 8.3 months prior to presentation.

Of all the boys and girls, 59% were fully prepubertal, 27% in early/mid-puberty, 8% were in late puberty and 6% unknown. Menarcheal status was known in 161 of 171 girls; 123 (76%) were premenarcheal. Of the 38 girls known to be postmenarcheal, 23 had amenorrhea, 8 did not and the remainder was unknown or not reported.

Fig. 1 Flow diagram of case ascertainment.

Fig. 2 Incidence by age at onset and diagnosis.

Other eating disorders include all other eating disorders (bulimia nervosa, eating disorder not otherwise specified, binge eating disorder); ‘other’ category includes individuals meeting definition for clinically significant eating difficulty, not currently classifiable as an eating disorder.

By definition, individuals with anorexia nervosa were underweight (mean 77% BMI for age), but % BMI in this group was not lower than for the ‘other’ group, whose mean % BMI was 79%. The six individuals classified with binge eating disorder were overweight, with a mean % BMI of 144%.

Family and early history

For 44% of individuals there was a close family member with a history of psychiatric disorder and for 12% more than one family member was affected. The proportion did not vary across diagnostic groups. Anxiety and depression accounted for almost 70% of the family history of mental health problems. Obsessive–compulsive disorder (OCD), anxiety disorder, psychosis, substance misuse and neurodevelopmental disorders were also represented.

Of 186 individuals in whom gestation was reported, only 2 were less than 36 weeks premature. A history of early feeding problems was reported in around 21% (66% none, 13% unknown), across all diagnoses (range 17–26%). The majority of feeding difficulties reported (as free text) were picky, fussy or faddy eating, and restrictive eating (eating small quantities). Comments were made highlighting specific features or severity (for example ‘fussy to 9 years’; ‘milk only until age 3’; ‘at 3–4 still eating liquids, could not tolerate lumps, not aware of hunger’).

Symptom profiles

Determined food avoidance was the most common symptom, occurring in 96% of all individuals. Other symptoms were morbid preoccupation with food (84%), fear of weight gain (71%), preoccupation with body weight (67%), preoccupation with body shape (51%), excessive exercising (43%), somatic complaints (31%), self-induced vomiting (19%), binge eating (5%) and laxative or diuretic use (1%). There was 71% concordance in preoccupation with body weight and with body shape. Weight concerns were more common in the absence of shape concerns than vice versa.

Principal-components analysis of the nine key symptoms found three components with an eigen value of over one, together accounting for 60% of the variance (30% for component one; 18% component two; 12% component three).Figure 3 shows the symptoms weighting each component. Component one was distinguished by fear of weight gain, preoccupation with body weight and shape and excessive exercise. Component two comprised underweight and somatic complaints but not weight and shape concerns and negatively for binge eating. Component three comprised self-induced vomiting and binge eating and was negative for underweight.

Comorbidity and medication

The three main comorbidities (depression, anxiety and OCD) were reported as possible or definite in 86 (41%) individuals, absent in 98 (47%) and the remainder unknown. Comorbidity was fairly evenly distributed across diagnoses, anxiety being the most common. For 13 individuals comorbid autism-spectrum or other pervasive developmental disorder was mentioned, as were a number of other neurodevelopmental problems such as tic disorders and attention-deficit hyperactivity disorder. Other comorbidities noted included trauma and grief reactions (n = 7), chronic fatigue syndrome and other somatoform disorders (n = 3), psychosis (n = 2) and learning disability (n = 2).

Management

In total, 104 (50%) were admitted to hospital (71% of those with anorexia nervosa). Most admissions were to paediatric wards for all diagnostic groups (37% of total, 73% of all admitted); 18% of the total number required psychiatric admission. Twenty-three (11%) individuals were fed by nasogastric tube, all from the diagnostic groups of anorexia nervosa, EDNOS or ‘other’. A total of 13% were prescribed psychotropic medication, the highest proportion being in the EDNOS categories. Most commonly used were a selective serotonin reuptake inhibitor (fluoxetine for nine individuals, sertraline for four, and one not stated) or an atypical antipsychotic (olanzapine for four individuals and risperidone for eight), with one diazepam and one melatonin.

Fig. 3 Principal-components analysis showing three main components with differing symptom profiles.

FA, food avoidance; FWG, fear of weight gain; PBS, preoccupation with body shape; PBW, preoccupation with body weight; XE, excessive exercise; SIV, self-induced vomiting; SC, somatic complaints; U, underweight; BE, binge eating.

1-year outcome

Some follow-up data were reported on 178/208 individuals (86%), although of these 11% were reported as unknown. No diagnostic group was over- or underrepresented in the follow-up data. There were no deaths at 1-year follow-up. In total, 130 (73%) were reported as improved, 11 (6%) were worse and 18 (10%) unchanged. Overall, 96 (60%) were still receiving treatment, 48 (30%) were not and for the remainder this was unknown. It is not known whether those not receiving treatment had recovered or had dropped out of treatment. Of 178 at follow-up, only 67 were still under the care of the same clinician.

At a mean of 11.5 months after measuring initial weight, most individuals had gained weight (mean 7 kg, range 0.2–24.9). Of the nine who had lost weight, the range was from 0.2 to 14 kg loss (mean 4.6). Mean change in BMI at follow-up was +2.1 kg/m².

The average length of admission reported at follow-up was 76 days (32 days for paediatric admissions), and seven individuals had been in hospital for more than 200 days by the time of follow-up. By 1 year only six children had been readmitted or admitted for the first time to hospital – six to paediatric wards and three to psychiatric in-patient units. Most children were in school, with 16 attending part time and 8 not in school. Comorbidity at follow-up was almost identical to that at baseline – data not shown.

Incidence

Direct comparison with previous UK data is limited because of differences in the age bands used and methodology. The 1999 Office for National Statistics study reported a prevalence of 0.3% for eating disorders in 11- to 15-year-olds, based on a single screening question, with no one in the 5- to 10-year-old age group. ^{Reference Ford, Goodman and Meltzer23}

Incidence data from a general practice registry study in the early 1990s found an incidence estimate for anorexia nervosa of 0.3/100 000 in 0- to 9-year-olds and 17.5/100 000 in the 10- to 19-year-olds. For bulimia nervosa the rates were 0/100 000 and 20.5/100 000 respectively. ^{Reference Turnbull, Ward, Treasure, Jick and Derby24} Our data for 0- to 9-year-olds was two and a half times higher (7.5/100 000). Retrospective studies from the USA and Denmark have suggested higher figures for anorexia nervosa, for example 9–27 per 100 000 for 10- to 14-year-old girls and 3.7 per 100 000 for boys. ^{Reference Pfeiffer, Lucas and Ilstrup25,Reference Joergensen26} Previous studies have been limited to anorexia nervosa and bulimia nervosa, and excluded EDNOS and other atypical presentations, which together accounted for 62% of our sample.

Two surveillance studies have used similar methodology to ours. Madden et al ^{Reference Madden, Morris, Zurynski, Kohn and Elliot27} reported an annual incidence of 1.4 per 100 000 children in Australia, half that of this UK/ROI study. The Australian study was confined to in-patients for 1 of 3 study years, and included few child and adolescent psychiatrists (2% of respondents). Our study used a dedicated child and adolescent psychiatric surveillance system. Unpublished data from the Canadian Paediatric Surveillance Program report 2.6/100 000, similar to our study. ²⁸

Importantly, our data provide a baseline from which to measure whether the incidence of eating disorders in children are really increasing as Van Son et al ^{Reference van Son, van Hoeken, Bartelds, van Furth and Hoek4} suggest, but do not at this point provide evidence for a true rise in incidence in children.

Demographics and medical status

As expected, the majority of individuals with eating disorders were female. Boys made up 18% of the overall sample, increasing to 33% of the ‘other’ category. This compares with 27% males overall in the Australian study. ^{Reference Madden, Morris, Zurynski, Kohn and Elliot27} This may be as a result of the broader case definition used in the Australian study, resulting in a higher proportion of atypical cases. Atypical presentations and lack of awareness that these conditions can arise in young children and boys can lead to delay in referral, diagnosis and treatment. ^{Reference Jacobs and Isaacs6,Reference Bryant-Waugh, Lask, Shafran and Fosson29} Three-quarters of girls were premenarcheal, similar to findings among in-patients in the Australian Paediatric Surveillance Unit study. ^{Reference Madden, Morris, Zurynski, Kohn and Elliot27}

We found that around 50% of individuals needed to be admitted to hospital. This contrasts with hospitalisation rates of over 70% in the Australian study, a sample biased towards in-patients. Our study provides a more representative picture in terms of the burden on hospital resources for younger individuals with eating disorders. The majority (73%) of those admitted in our study were admitted to paediatric units rather than psychiatric or specialist eating disorders units. The National Inpatient Child & Adolescent Psychiatric Survey ³⁰ showed that eating disorders were the most common diagnosis among child and adolescent psychiatric in-patients in England and Wales, 9.2% of whom were <13 years. The National Inpatient Child & Adolescent Psychiatric Survey did not include children on paediatric wards. Ours are the first published data on the scale of paediatric resource used by this patient group in the British Isles. Clinical experience suggests wide variation in the confidence among paediatricians over the management of these cases. This study provides the baseline figures needed to determine whether increases in hospitalisation rates are real or perceived, and can aid planning of age-appropriate services. Furthermore, there are currently no national paediatric guidelines for the care of individuals with eating disorders.

Diagnostic categories

Diagnosis and nosology are the subject of much discussion and debate, with revisions to DSM–IV and ICD–10 pending. In particular, the category of EDNOS has been widely criticised for being too heterogeneous, ^{Reference Thomas, Vartanian and Brownell31} accounting for 77.4% of all individuals diagnosed with adolescent eating disorders in the community compared with only 13% for anorexia nervosa and 10% for bulimia nervosa. ^{Reference Machado, Machado, Goncalves and Hoek32} Particular problems with the application of current diagnostic criteria for children and adolescents have also been highlighted. ^{Reference Bravender, Bryant-Waugh, Herzog, Katzman, Kreipe and Lask20,Reference Nicholls, Chater and Lask33} This study provided an opportunity to explore the nosology of early-onset cases using a symptom-based approach. Principal-components analysis, based on nine key symptoms, found that three main clinical profiles in this age group in our study accounted for 60% of the variance. Our findings support those who argue that weight status is a continuum and does not distinguish between clinical cases on other criteria. This is particularly so in children; weight status is often a reflection of adult interventions (i.e. whether the child had been fed) rather than a measure of severity of illness.

Those individuals who present with determined food avoidance in the absence of weight and shape concerns, 20% of our sample, are of less clear nosological status. Variously described as food avoidance emotional disorder ^{Reference Nicholls and Bryant-Waugh34} or non-fat phobic anorexia nervosa, ^{Reference Becker, Thomas and Pike35} these cases might possibly be classified as anorexia nervosa on GP registers. Identification of these individuals has important implications for the development of clinical approaches to intervention, currently lacking, and for the scope of eating disorders provision.

Management

Of the total sample, 11% (22% of in-patients) were fed via nasogastric tube, compared with 58% of in-patients in the Australian study. Furthermore, 34% were prescribed psychotropic medication in the Australian sample, compared with only 13% in our sample. The mean duration of admission in the Australian study was only 24.7 days, compared with 76 days overall and 32 days for paediatric admissions in our sample. These differences may reflect cultural differences in treatment approach and/or differences in health service organisation, as well as the bias towards studying in-patients in the Australian study. Further research is needed to understand the implications for outcome of different approaches.

Risk factors and early intervention: psychiatric comorbidity

This study enabled examination of the first phases of an eating disorder in young children, before it has become compounded by chronicity. Increasing recognition of the neurodevelopmental contributions to risk for anorexia nervosa are reflected in the proportion of individuals in our study with comorbid autism-spectrum, psychotic and other developmental disorders. Prepubertal cases provide an opportunity to explore neurobiological risk factors before they are overshadowed by the major role that puberty plays.

In over 40% there was a family history of mental illness, most commonly anxiety or depression. In addition, a history of early feeding problems was reported in over 20% of individuals, across the diagnostic groups. Our study supports others that highlight the importance of family history and of early feeding problems, particularly fussy or picky eating, as risk factors for the development of early-onset eating disorders. ^{Reference Nicholls and Viner36}

One of the most concerning findings from this study was that the mean time between onset of symptoms and presentation to secondary care was 8.3 months. The follow-up data show that most hospitalisations occur in the period immediately after presentation, with very few once treatment is established. The length of illness prior to presentation and the severity of illness at presentation suggest an urgent need to improve early detection and management of these individuals. Lask et al ^{Reference Lask, Bryant-Waugh, Wright, Campbell, Willoughby and Waller16} found that a single GP consultation about eating behaviour or weight and shape concerns is a strong predictor of the subsequent emergence of anorexia nervosa.

1-year outcome

Our study is the first to report the short-term outcome in a large representative sample of individuals with childhood-onset eating disorders. Although outcome was reported as good for the majority, a significant minority (16%) were reported as worse or unchanged at 1 year. Almost two-thirds of individuals were still in treatment, with a small number (seven) in hospital for most of the year. These data confirm that the healthcare burden of childhood eating disorders is not solely attributable to in-patient admission, and reinforces the needs to plan for long-term out-patient care. Our data suggest that the National Institute for Health and Clinical Excellence (NICE) guideline recommendation that out-patient treatment be offered typically for 6 months in the first instance and for a year after hospitalisation can reasonably be applied to children with eating disorders as well as adolescents and adults. ¹²

Strengths and limitations

The main strengths of our study are the national representative sample, and consequently adequate cases for descriptive analysis. With the high response rate from clinicians, the inclusion of cases presenting both to paediatricians and child psychiatrists, and the encouragement of dual reporting, we are confident that we have identified over 80% of restrictive eating disorders in those under 13 presenting to secondary care in the UK and ROI in the reporting year in both the National Health Services (NHS) and independent sector.

This methodology is limited to cases seen by consultant paediatricians and child psychiatrists. Those managed by GPs, in mental health services that do not have paediatric or psychiatric involvement, by training grade or career grade clinicians, receiving alternative healthcare or those receiving purely dietetic/nutritional care would not be identified by this methodology.

Case ascertainment in surveillance studies relies on clinician judgement and accuracy of clinician reporting. This will be more important for some data, for example binge eating and comorbidity, than for others such as weight. Binge, or loss of control eating is likely to be particularly underrepresented, since this would require systematic screening and clear operational criteria to detect. Data were checked for internal consistency, especially when duplicate reports were available. For example, individualsreportedasbothinearly pubertyand as postmenarcheal were reassigned to mid-late puberty.

Implications

The 2004 NICE guidelines ¹² on the treatment of eating disorders highlight the need for developmentally appropriate care for young people with eating disorders, and the importance of coordination between professionals involved in their care. Many eating disorder services are specifically aimed at adolescents. Our study reaffirms the importance of considering the needs of children with eating disorders, in addition to and separately from those of adolescents, rather than simply lowering the age range of adolescent services. In particular, the role of paediatric services needs clarification and coordination between professionals in the management of these individuals, for example through the use of care protocols. Childhood eating disorders are not quick or easy to treat, and for a minority may be the start of a severe and enduring illness, with a mortality comparable with some forms of leukaemia. There is an urgent need to improve efforts directed towards early detection and intervention, with the expectation that this will improve prognosis. ^{Reference Katzman, Golden, Neumark-Sztainer, Yager and Strober37}

Operational criteria for diagnostic grouping for cases meeting the case definition