Correspondence : Youl-Ri Kim, MD, Eating Disorders Clinic, Department of Neuropsychiatry, Inje University Seoul Paik Hospital, 85 Jeo-dong, 2-ga, Jung-gu, Seoul 100-032, Korea
Tel : +82-2-2270-0970, Fax : +82-2-2270-0344, E-mail : youlri.kim@gmail.com

Location of Work : Eating Disorders Unit PO 59, Institute of Psychiatry, King's College London, De Crespigny Park, SE5 8AF, London, UK

Introduction

Anorexia nervosa (AN) has a heterogeneous course.^1,2,3,4 In some instances, the disorder is short-lived, whereas in others it becomes entrenched.^5,6,7,8,9,10 A transition to the bulimic form of the disorder occurs frequently.^{11,12,13,14,15,16,17} This heterogeneity in the outcome of the disorder complicates the interpretation of aetiological studies and treatment planning. There is uncertainty in the course of AN about the factors associated with either its maintenance or its transition into another form of the disorder.
The developmental factors, which may reflect the effects of genes or the environment, or both, may supplement the descriptions of the phenotype in terms of the ability to predict the course and prognosis of the illness. The clinical features at presentation have been the main variables examined as the predictors of the course of the illness,^2,18 whereas there has been very little research which attempted to examine how the developmental factors contribute to the prognosis.
The aim of this study was to examine which developmental features play a role in predicting the course of AN. The assumption of this study was that there are 3 broad courses of AN; continued AN, crossover to a bulimic spectrum disorder, and remission. This outcome classification made it possible to define 3 broad, homogenous categories for the purpose of this investigation.

Methods

Participants
A total of 400 female participants with a lifetime history of AN (restricting or binge purging subtype) were recruited from clinical settings (inpatient and outpatient services) and the community, by way of a volunteer database maintained by the eating disorder research unit. The database was established in 1999 as a means of (i) informing research participants about the results of studies in which they participated, and (ii) as a confidential record of individuals who, when asked, expressed an interest in participating in future research projects.
Information about the register was provided to all patients within the South London and Maudsley National Health Service Trust when they were offered an assessment appointment, or at a later date in their treatment. Several other eating disorders services across the UK, both public and private, also provided their patients with information about the register. All of the individuals on the register were sent newsletters twice yearly which presented recent news and research findings from the eating disorders research unit. At this time, individuals were also sent information regarding ongoing research projects that they might like to participate in and were sometimes invited to complete a questionnaire-based study. The participants experienced a range of different treatment services.
Written informed consent was obtained from all of the participants. The study was approved by the South London and Maudsley National Health Service Trust and the Institute of Psychiatry research ethics committee.

Procedure
In the first phase of the screening at the time of recruitment to the register, lifetime worst diagnoses, assigned according to a formal diagnostic hierarchy,¹⁹ were determined for each participant according to the Diagnostic and Statistical Manual of Mental Disorders (4th edition)(DSM-IV) criteria. Lifetime diagnosis of AN was made using selected behavioural items, the current weight and height, and the current and past menstrual functioning items of the Eating Disorders Examination Questionnaire Version (EDE-Q)²⁰ adapted to capture episodes of maximum severity of symptoms in the past, in addition to the current symptoms (The questionnaire can be found at http://www.eatingresearch.com). A subgroup of 60% of those within AN sample also completed a semi-structured clinical interview using the EATATE interview²¹ as part of another ongoing study in the research unit. The EATATE instrument demonstrates good inter-rater reliability in terms of diagnoses (kappa =0.82-1.0) and illness history variables (kappa=0.80-0.99). The assessment of the diagnostic validity (compared to the clinical notes) yielded kappa values of between 0.77 and 1.0 for the sequential diagnoses. The mean age of the participants with AN at recruitment was 31.1±10.9 years and their mean duration of illness was 14.4±10.2 years. Their mean body mass index (BMI, wt[kg]/ht[m]²) was 15.65±1.7.
The assessment for this study was performed 3.1±1.1 years after they had entered the register, using a questionnaire pack mailed to all of them, to which 55% (n=220) replied. 30 % of the mail was returned undelivered, because of changes of address. The questionnaire pack included items developed from measures used to ascertain the childhood features. The diagnoses for the purpose of this study were made using the Eating Disorder Diagnostic Scale,²² a self-report questionnaire derived from validated structured interviews for assessing eating disorders. The threshold and sub-threshold diagnoses were generated as outlined by DSM-IV. Those respondents who returned incomplete questionnaires were excluded. Those whose ages were over 60 were excluded to minimize age-related recall bias. The respondents and non-respondents did not differ significantly in age, ethnicity, educational status, marital status, or in the illness history variables (onset, duration, and hospitalization)(data available on request). The 20 females with subthreshold AN were eliminated from the analyses due to the instability of their course. Thus, a total of 142 female participants completed assessments for the study.

Assessments

Eating Disorder Diagnostic Scale
The diagnoses were made using eating disorder diagnostic scale (EDDS), a self-report questionnaire derived from validated structured inter-views used for assessing eating disorders. The threshold and sub-threshold current diagnoses of AN, bulimia nervosa (BN), and binge eating disorder (BED) were made according to the criteria defined by Stice et al.²² The sub-threshold diagnoses required the presence of all the symptoms of the disorder, at least one of which should have a sub-diagnostic severity (e.g. binge-eating only once a week). Sub-threshold AN was diagnosed if the subject met all of the following criteria; a BMI of less then '18.5'; she missed at least 'one' period in the past 3 months or had no missed periods but was on the pill. Sub-threshold BN was diagnosed if the subject met all of the following criteria; during the last six months the participant experienced overeating with a loss of control; the above eating patterns occurred 'one or more' times per week over the past 3 months; over the past three months the sum of the compensatory behaviour episodes per week was equal to 'one or more'; over the past three months the participant's weight or shape 'slightly' influenced how they judged themselves as a person. Sub-threshold BED was diagnosed if the subject met all of the following criteria; during the last six months the participant experienced overeating with loss of control; over the past six months the participant's eating followed the above pattern on 'one' or more days per week; during these episodes of overeating the participant experienced 'three or more' of the binge eating features; the participant felt very upset about her uncontrollable eating or resulting weight gain; over the past three months the sum of the compensatory behaviour episodes per week was equal to 'zero'. In those cases where the participant's eating behaviours were still maladaptive, even if they did not meet the criteria for subthreshold eating disorders, one of the sub-threshold diagnoses was determined using National Institute of Mental Health (NIMH) criteria used in the ongoing NIMH genetics study. These sub-threshold diagnoses were considered as eating disorders not otherwise specified according to DSM-IV.

Childhood Features
A broad range of childhood factors compiled by a thorough review of the literature were assessed to identify the putative course predictors, including items from the McKnight Risk Factor Survey IV,^23,24 Wardle's Food Scale,²⁵ Dutch Eating Behaviour Questionnaire,²⁶ Childhood Feeding Questionnaire,27 Food Attitude Survey^28,29 and Health and Taste Attitudes Test.³⁰ Additional items were obtained from working with a focus group of patients with AN, where the proposed instrument was discussed. Most of the items have four or five point scale answers ranging from 0 (never) to 3 (a lot) or 4 (always or totally). These questions were categorized a priori into domains reflecting certain types of risk factors; (1) Social support (2) Perfectionism (3) Emotional overeating (4) Parents or adults concerned with thinness (5) Ritualized eating (6) Emotional undereating (7) Anxiety (8) High parental expectations (9) Avoidance of eating with others and (10) Picky eating.

Statistical analysis
Multinomial logistic regression was used to determine the correlations between the individual factors and the group membership (continued AN, remission, crossover to a bulimic spectrum disorder). The relationships between the individual factors and group membership were first assessed by including the individual factor of interest. Measures of each domain were obtained by means of component factors that each subject had been exposed to. The effect of these domain exposure variables was also assessed separately by including them as independent variables in the multinomial logistic regression model. The analyses were conducted after adjusting for current age, age at onset, and duration of illness. These adjustments minimized the differences in the phase of illness and the time available for exposure. Given that we performed multiple statistical tests, we set the significance level to p≤0.01, representing a compromise between the more stringent Bonferroni correction and the exploratory nature of this study. Two-tailed tests were used throughout the analyses. Unless otherwise indicated, data are given as means±s.d.

Results

Diagnostic outcome over 3 years
Of the total of 142 females who completed assessments for the study, the 65 females who still met the criteria for AN full threshold were classified as the continued AN group. At the time of follow-up, 40 females were assigned to the bulimic outcome group, of which 29 met the BN threshold, 9 met the sub-threshold for BN, 1 met the BED threshold, and 1 met the sub-threshold criteria for BED. 37 females did not meet the criteria for any of the above eating disorders over the past 6 months, so they were classified as the remission group. The remaining 20 females with sub-threshold AN were eliminated from the analysis because of the instability of their course. The reasoning behind this predetermined classification was that AN pathology measured by BMI takes time to evolve from recovery to remission. Figure 1 presents the diagnostic distribution.
The subjects' clinical features are summarized in Table 1. As expected, the participants in the continued AN group had a pattern of lower weight over their life course than those in the other groups.

Subtype
The subtype significantly distinguished the course of AN (Table 2). The restricting subtype was 2.5-fold more common among the subjects with continued AN than it was among those who progressed to the remission group. The binge purging subtype was 3.2-fold more common among the subjects who underwent a crossover to a bulimic spectrum disorder than it was among those who progressed to the remission group.

Individual childhood factors
Most of the individual childhood factors did not discriminate between the group outcomes (Table 3). Six items from the following domains did differ between the groups; perfectionism, emotional overeating, parents or adults concerned with thinness, and high parental expectations.

Overall level of exposure to each domain
3 groups differed in their exposure to the 3 childhood factor domains, viz. parents or adults concerned with thinness, high parental expectations, and emotional overeating (Table 4). Those participants who underwent a transition to a bulimic disorder more often reported that their parents or concerned adults were over concerned with thinness [χ²(2)=13.9, p< 0.01, 95% CI: 1.5-4.7], or they had high parental expectations [χ²(2)=10.2, p< 0.01, 95% CI: 1.3-4.0], as compared with those subjects with continued AN. Those participants with continued AN reported significantly lower emotional overeating than those in the other groups [χ²(2)=9.2, p=0.01, 95% CI: 1.1-3.0].

Discussion

This study is, to our knowledge, the first to assess comprehensively which childhood features of AN prior to their onset correlate with subsequent course of illness. The findings suggest that only a few features, such as emotional eating, parental expectations, and adults concerned with thinness, discriminate the course of AN. The register of research volunteers utilized in our study allowed a broad spectrum of cases to be recruited, with a wide range of illness symptoms and severity, and who were less biased by referral, service and treatment patterns.
Over the 3 year period, AN persisted in 39% of the subjects, while 26% shifted to a bulimic spectrum disorder. It is interesting that 23% of the chronically ill AN subjects with a long duration of illness went into remission. The relatively high rate of persistent disorder in this study might be due to the characteristics of our sample, which was comprised of subjects who were older and had a longer duration of illness than those in most clinical case series. Therefore, the findings of this study are more applicable to the course late in the illness.
We found that those participants with the restricting subtype were more likely to continue to suffer from AN, while those with the binge purging subtype were more likely to experience a later transition to a bulimic spectrum disorder. We used the strictly defined inclusion criteria for restricting AN which includes temporal elements, i.e. the subject suffered from AN without any bulimic behaviour during the first 3 years after the onset of the disorder as defined in a genetic study. This may account for the difference in outcomes within the literature where different definitions of restricting AN were used.¹¹ Our findings suggest that the clinical phenotype has implications for the course of the illness.
Emotional undereating or picky eating in childhood was not associated with the course of the illness. However, eating in response to emotional cues occurred less often in those subjects with continued AN than in the other subjects. This finding may be related to trait-related disturbances of the dopamine metabolism or brain pathway, including the anterior cingulate cortex, which are likely to play a role in food reward.^31,32 People with AN show deficits in reward processing on tasks such as visual discrimination learning tasks³³ and the Iowa gambling task.^33,34 Decision making and behaviour choice involves, to a certain degree, the automatic computation of the balance of reward and punishment.³⁵ It is possible that food is less potent as a reward in those with persistent AN.
Those participants who crossed over to a bulimic disorder were more likely to report that their parents or others were concerned with thinness before the onset of their illness. Dieting vulnerability, such as critical comments by family members or others about shape, weight, or eating, is known to be a greater risk factor for BN than for AN.³⁶ The findings of this study suggest that individuals with AN who are vulnerable to diet in their childhood are more likely to experience a later transition into a bulimic disorder.
Family factors, namely high parental expectation and parents or adults concerned with thinness, were associated with a shift to a bulimic disorder. Previous studies found parental criticism or expectations to be significant predictors of the onset of binge eating in women with AN.^3,13,37 Interventions tailored for this problem, i.e. including family in treatment, may improve the outcome in this subgroup.³⁸ However, high parental expectations could be caused by genetic factors associated with the perfectionism of the parents, which could be a risk factor for the maintenance of the illness.³⁹ These discrepancies in the role of parental perfectionism on the course of illness should be investigated in future research.
The prominent strengths of the present study include a relatively large cohort of AN followed over a 3-year period. Other strengths include the use of a new instrument to measure childhood factors. This instrument included a broad range of risk factors established from various studies in a systematic literature review.^40,41
This study has some limitations. First, the subjects have a long duration of AN with different phases of the illness. It is therefore possible that the sample was not representative of cases of AN in the general population. It is uncertain whether the developmental factors of importance identified in this group also pertain to the early stage of the disorder. Second, those individuals with continued AN had the shortest duration of illness. Therefore, additional instances of remission that were not captured in the present study could still occur. The resolution of this issue awaits future research with a longer follow up period. Third, the validity of the instrument used for the analysis of the childhood factors was not yet verified when this study was carried out. In addition, the information afforded by the instrument was retrospectively acquired. Therefore, such soft judgments as over concern or expectations may be biased by retrospective recall, or affected by the current state of the subject. Fourth, self-reporting measures were used for diagnosis. However, the level of reliability of the diagnostic information in the sample was found to be good in the group of subjects who had their diagnosis verified by interview. The final limitation is the relatively low response rate of the volunteers in the original database of 55%, which may have biased the sample remaining in our study. However, we do not think there is any systematic bias associated with the non-responders (over 30% of the mail was returned because of a change of address).
In summary, this study furthers our understanding of the predictors for the course of AN. Those participants with continued AN differed from those in remission at outcome in terms of their childhood emotional overeating. Those subjects who underwent a crossover to a bulimic spectrum disorder were more likely to have parents, who had higher expectations and report that their parents or concerned adults were over concerned with thinness, than those with continued AN. Overall, there were very few differences in the other important factors between the various courses of AN.
The findings of this study need to be tested in the early stage of the illness with a longer follow up period, in order to determine the extent to which the developmental factors contribute to the evolving course of AN.

1. Sullivan PF, Bulik CM, Fear JL, Pickering A. Outcome of anorexia nervosa: a case-control study. Am J Psychiatry 2003;155:939-946.

2. Steinhausen HC. The outcome of anorexia nervosa in the 20th century. Am J Psychiatry 2002;159:1284-1293.

3. Strober M, Freeman R, Morrell W. The long-term course of severe anorexia nervosa in adolescents: survival analysis of recovery, relapse, and outcome predictors over 10-15 years in a prospective study. Int J Eat Disord 1997;22:339-360.

4. Eckert ED, Halmi KA, Marchi P, Grove W, Crosby R. Ten-year follow-up of anorexia nervosa: clinical course and outcome. Psychol Med 1995;25:143-156.

5. Martin F. The treatment and outcome of anorexia nervosa in adolescent: a prospective study and five-year follow-up. J Psychiatr Res 1985;19:509-514.

6. Morgan HG, Russell GM. Value of family background and clinical features as predictors of long-term outcome in anorexia nervosa: a standardized prospective study. Br J Psychiatry 1975;143:282-287.

7. Nussabaum M, Shenker IR, Baird D, Saravay S. Follow-up investigation in patients with anorexia nervosa. J Pediatr 1985;106:835-840.

8. Seidensticker JF, Tzagournis M. Anorexia nervosa: clinical features and long-term follow-up. J Chronic Disord 1968;21:361-367.

9. Touyz SW, Beaumont PJ. Anorexia nervosa: a follow-up investigation. Med J Aust 1984;141:219-222.

10. Herzog DB, Dorer DJ, Keel PK, Selwyn SE, Ekeblad ER, Flores AT, et al. Recovery and relapse in anorexia and bulimia nervosa: a 7.5-year follow-up study. J Am Acad Child Adolesc Psychiatry 1999;38:829-837.

11. Eddy KT, Keel PK, Dorer DJ, Delinsky SS, Franko DL, Herzog DB. Longitudinal comparison of anorexia nervosa subtypes. Int J Eat Disord 2002;31:191-201.

12. Bulik CM, Sullivan PF, Fear J, Pickering A. Predictors of the development of bulimia nervosa in women with anorexia nervosa. J Nerv Ment Dis 1997;185:704-707.

13. Tozzi F, Thornton LM, Klump KL, Fichter MM, Halmi KA, Kaplan AS, et al. Symptom fluctuation in eating disorders: correlates of diagnostic crossover. Am J Psychiatry 2005;162: 732-740.

14. Gillberg I, Rastam M, Gillberg C. Anorexia nervosa outcome: six-year controlled longitudinal study of 51 cases including a population cohort. J Am Acad Child Adolesc Psychiatry 1994;33:962-966.

15. Schork EJ, Eckert ED, Halmi KA. The relationship between psychopathology, eating disorder diagnosis, and clinical outcome at 10-year follow-up in anorexia nervosa. Compr Psychiatry 1994;35:113-123.

16. Smith C, Feldman SS, Nasserbakht A, Steiner H. Psychological characteristics and DSM-III-R diagnoses at 6-year follow-up of adolescent anorexia nervosa. J Am Acad Child Adolesc Psychiatry 1993;32:1237-1245.

17. van der Ham T, van Strien D, van Engeland H. A four-year prospective follow-up study of 49 eating disordered adolescents: differences in course of illness. Acta Psychiatr Scand 1994;90:229-235.

18. Fichter MM, Quadflieg N, Hedlund S. Twelve-year course and outcome predictors of anorexia nervosa. Int J Eat Disord 2006;39:87-100.

19. The Price Foundation Collaborative Group. A search for susceptibility loci for anorexia nervosa: method and sample description. Biol Psychiatry 2000;47:794-803.

20. Fairburn CG, Beglin SJ. Assessment of eating disorders: interview or self-report questionnaire. Int J Eat Disord 1994;16:363-370.

21. Anderluh MB, Tchanturia K, Rabe-Hesketh S, Treasure J. Childhood obsessive-compulsive personality traits in adult women with eating disorders: defining a broader eating disorder phenotype. Am J Psychiatry 2003;160:242-247.

22. Stice E, Telch CF, Rizvi SL. A psychometric evaluation of the Eating Disorder Diagnostic Scale: a brief self-report measure for anorexia, bulimia, and binge-eating disorder. Psychol Assess 2000;12:123-131.

23. Shisslak CM, Renger R, Sharpe T, Crago M, McKnight KM, Gray N, et al. Development and evaluation of the McKnight Risk Factor Survey (MRFS-III) for assessing potential risk and protective factors for disordered eating in preadolescent and adolescent girls. Int J Eat Disord 1999;25:195-214.

24. The McKnight Investigators. The McKnight Risk Factor Survey IV. Available at: http://bml.stanford.edu. Stanford, CA: Stanford University, 2001.

25. Wardle J, Guthrie CA, Sanderson S, Rapoport L. Development of the Children's Eating Behaviour Questionnaire. J Child Psychol Psychiatry 2001;42:963-970.

26. van Strien T, Frijters JER, Bergers GPA, Defares PB. The Dutch eating behavior questionnaire (DEBQ) for assessment of restrained, emotional and external eating behaviors. Int J Eat Disord 1986;5:295-313.

27. Birch LL, Fisher JO, Grimm-Thomas K, Markey CN, Sawyer R, Johnson SL. Confirmatory factor analysis of the Child Feeding Questionnaire: a measure of parental attitudes, beliefs and practices about child feeding and obesity proneness. Appetite 2001;36:201-210.

28. Raudenbush B, van der Klaauw NJ, Frank RA. The contribution of psychological and sensory factors to food preference patterns as measured by the Food Attitudes Survey (FAS). Appetite 1995;25:1-15.

29. Frank RA, van der Klaauw NJ. The contribution of chemosensory factors to individual differences in reported food preferences. Appetite 1994;22:101-123.

30. Roininen K, Tuorila H, Zandstra EH, de Graaf C, Vehkalahti K, Stubenitsky K, et al. Differences in health and taste attitudes and reported behaviour among Finnish, Dutch and British consumers: a cross-national validation of the Health and Taste Attitude Scales (HTAS). Appetite 2001;37:33-45.

31. Frank GK, Wagner A, Achenbach S, McConaha C, Skovira K, Aizenstein H, et al. Altered brain activity in women recovered from bulimic-type eating disorders after a glucose challenge: a pilot study. Int J Eat Disord 2006;39:76-79.

32. de Zwaan M. Basic neuroscience and scanning (2nd ed). In: Treasure J, Schmidt U, Furth E, editors. Handbook of eating disorders. Chichester: John Wiley & Sons, 2003, p.89-102.

33. Lawrence AD, Dowson J, Foxall GL, Summerfield R, Robbins TW, Sahakian BJ. Impaired visual discrimination learning in anorexia nervosa. Appetite 2003;40:85-89.

34. Cavedini P, Bassi T, Ubbiali A, Casolari A, Giordani S, Zorzi C, et al. Neuropsychological investigation of decisionmaking in anorexia nervosa. Psychiatry Res 2004;127:259-266.

35. Kringelbach ML. The human orbitofrontal cortex: linking reward to hedonic experience. Nat Rev Neurosci 2005;6:691-702.

36. Fairburn CG, Cooper Z. Risk factors for anorexia nervosa. Arch Gen Psychiatry 1999;56:468-76.

37. Sights JR, Richards HC. Parents of bulimic women. Int J Eat Disord 1984;3:3-13.

38. Treasure J, Whitaker W, Whitney J, Schmidt U. Working with families of adults with anorexia nervosa. J Fam Ther 2005;27:158-170.

39. Halmi KA, Sunday SR, Strober M, Kaplan A, Woodside DB, Fichter M, et al. Perfectionism in anorexia nervosa: variation by clinical subtype, obsessionality, and pathological eating behavior. Am J Psychiatry 2000;157:1799-1805.

40. Jacobi C, Hayward C, de Zwaan M, Kraemer HC, Agras WS. Coming to terms with risk factors for eating disorders: application of risk terminology and suggestions for a general taxonomy. Psychol Bull 2004;130:19-65.

41. Stice E. Risk and maintenance factors for eating pathology: a meta-analytic review. Psychol Bull 2002;128:825-848.