Regular Article

Psychother Psychosom 2007;76:234241 DOI: 10.1159/000101502

Childhood Eating and Weight in Eating Disorders: A Multi-Centre European Study of Affected Women and Their Unaffected Sisters
N. Micali a, b J. Holliday b A. Karwautz e M. Haidvogl e G. Wagner e F. Fernandez-Aranda f A. Badia f L. Gimenez f R. Solano f M. Brecelj-Anderluh g R. Mohan c D. Collier d J.L. Treasure b
Department of Child and Adolescent Psychiatry and Eating Disorders Research Unit, Institute of Psychiatry, Kings College London and b Department of Academic Psychiatry (Guys Hospital), Eating Disorders Research Unit, Kings College London, c Section of Cultural Psychiatry and d SGDP, Institute of Psychiatry, London, UK; e Department of Child Neuropsychiatry, University of Vienna, Vienna, Austria; f Department of Psychiatry, Ciutat Sanitaria Universitaria de Bellvitge, Barcelona, Spain; g Department of Psychiatry, University of Ljubliana, Ljubliana, Slovenia
a

Key Words Childhood eating Feeding problems Eating disorders Sister pairs, discordant

Abstract Background: Previous studies have suggested that childhood eating and weight problems may be risk factors for eating disorders. Robust evidence is still lacking. Aims: To investigate whether childhood eating and weight problems increase the risk of eating disorders in affected women compared to their unaffected sisters. Methods: Women (150) with anorexia (AN) or bulimia nervosa (BN) recruited from clinical and community samples were compared to their unaffected sister closest in age on maternal reports of childhood eating and weight. Results: Women with BN were significantly more overweight at the ages of 5 and 10 (both OR = 2.8, p ! 0.01), ate a lot (OR = 1.3, p ! 0.01), were less picky (OR = 0.6, p ! 0.05) and ate quickly (OR = 2.3, p ! 0.05) between the ages of 6 and 10 compared to their healthy sisters. Significantly more women with AN were described as having a higher weight at 6 months (OR = 0.8, p ! 0.01) and 1 year (OR = 0.6, p ! 0.01) compared to their healthy sisters. Childhood eating was comparable in the women with AN and

their unaffected sisters. Conclusions: Traits of childhood overeating were more common in bulimic women compared to their unaffected siblings. Subjects with AN did not differ from their sisters on eating variables. The increased risk of BN due to childhood overweight suggests that prevention strategies for childhood obesity and overweight may therefore be applicable in BN.
Copyright 2007 S. Karger AG, Basel

Introduction

The link between eating and weight problems in childhood and later eating disorders (ED) remains ambiguous [1]. Longitudinal studies have provided some evidence in this respect. Marchi and Cohen [2] prospectively investigated a large group of children (659) from the age of 1021 using maternal reports. Pica and digestive problems between the age of 1 and 10 were related to bulimic symptoms between the ages of 12 and 20. Picky eating and digestive problems during childhood (ages 110) predicted anorexic symptomatology between the ages of 9 and 18. However, these findings were not replicated at later time points. In the same cohort at the age of 27, eating

2007 S. Karger AG, Basel 00333190/07/07640234$23.50/0 Fax +41 61 306 12 34 E-Mail karger@karger.ch www.karger.com Accessible online at: www.karger.com/pps

Dr. Nadia Micali Child Psychiatry Department, PO Box 85, Institute of Psychiatry Kings College London, De Crespigny Park Denmark Hill, London SE5 8AF (UK) Tel. +44 207 848 0468, Fax +44 207 708 5800, E-Mail n.micali@iop.kcl.ac.uk

conflicts, struggles around meals and unpleasant meals (ages 110) were predictive of a diagnosis of broadly defined anorexia nervosa (AN) [3]. Eating too little was significantly protective for developing bulimia nervosa (BN). In contrast to these, a case-control study comparing adolescent girls with AN and controls [4] and a small clinical study on women with BN [5] found low rates of childhood eating and feeding difficulties in the ED groups. The relationship between childhood weight problems and ED is not fully understood either. Some case-control studies have reported an association between BN and binge eating disorder and childhood overweight, although both relied on self-report [6, 7]. In 3 longitudinal studies, higher weight was found to be predictive of bulimic behaviours [8] and ED caseness [9, 10], but subsequent longitudinal studies have not shown an effect of body mass index (BMI) in predicting ED or bulimic symptoms [11, 12]. Moreover no studies have accounted for the role of family values and attitudes in the variations in childhood eating and weight problems as antecedents for ED. We aimed to clarify the role of previously identified childhood eating and weight problems as precursors of adult ED in affected women compared to their unaffected sisters, using maternal reports. Based on risk factors identified in the literature, we hypothesized that in sisters discordant for AN feeding problems during the first year of life (selective eating) and childhood (i.e. picky eating, digestive problems, difficulties in making the child eat, not eating enough at mealtime), and childhood underweight would be more frequent for the affected sister. We also hypothesized that pica, digestive problems, low rates of picky eating, low rates of not eating enough and childhood overweight would distinguish the sibling with BN from the healthy one.

Methods
Design We used a case-control study design with healthy sisters as controls. Maternal reports of childhood eating and weight patterns were compared between women with AN or BN and their unaffected sisters. Subjects Four centres participated in the study: the Eating Disorders Research Unit, Institute of Psychiatry, London, UK; the Department of Child Neuropsychiatry, University of Vienna, Austria; the Department of Psychiatry, Ciutat Sanitaria Universitaria de Bellvitge, Barcelona, Spain, and the Department of Child Psychiatry, University of Ljubliana, Slovenia.

Women with ED were invited to participate in a collaborative multi-centre study across Europe, investigating risk factors for ED if they had a sister close in age. The cases were recruited from specialist services for ED in their respective countries and from the community, amongst women who had participated in previous ED research, through media articles and user group organizations. Individuals were included in the study on the basis that they had a lifetime diagnosis of DSMIV AN or BN [13]. Longitudinal changes in diagnosis present problems in research where a single diagnosis is required. Among individuals with a short duration of illness, particularly restrictive AN, it is quite likely that a good proportion will go on to develop bulimic symptoms. This raises the question of whether it is appropriate to classify these individuals as having AN, for research purposes. We diagnosed AN and BN based on a lifetime ED status in accordance with the criteria used by the Price Foundation [14], on the basis of a hierarchical model of diagnosis: women with BN irrespective of whether they had a previous history of AN are classified as having BN. This approach is based on the idea that women with BN are phenotypically different from women with AN. Diagnoses of individuals who had had a singular diagnosis of restrictive AN less than 2 years prior to ascertainment were excluded from the study, because of the possibility that they might develop a different form of ED later on. The participants had to have a healthy sister who did not have any form of eating disorder [neither AN, BN, nor ED not otherwise specified (EDNOS)], no co-morbid psychotic disorder or learning difficulties, and be a native speaker. The sister pairs had to be less than 10 years apart and to have lived in the same family for a minimum of 8 years, to account for differences in shared environment. The use of sisters aimed to control for socio-demographic factors, such as age and social class, but more importantly for family environment and values such as attitudes to weight and eating, parental eating problems and family cultural values like the importance of thin ideals and eating. If the ED case had more than 1 sister, the sister closest in age was approached. The sisters who agreed to participate were interviewed in order to exclude any lifetime ED diagnosis (AN, BN or EDNOS). The unaffected sisters had to be 16 or older to take part in the study, to minimize the possibility that they might develop an ED later on in life. Across the 4 centres, 262 women were recruited. Initially 224 probands met the study criteria, 17 (7.6%) sister pairs were excluded because the probands sister met the criteria for a full or sub-threshold ED following a diagnostic interview. Another 4 sister pairs (1.7%) were excluded because the unaffected sisters were younger than 16. So 203 sister pairs were eligible and agreed to participate in the study. Six did not complete the diagnostic assessment and were therefore excluded. The study sample consisted of 197 sister pairs. Questionnaires were sent to 197 mothers. A total of 150 questionnaires (76%) were received, and 150 women and their sisters were included in the study. These patients were representative of the 203 patients eligible, as shown by comparing the groups across different centres on all available socio-demographic and clinical variables (age, current BMI, education) with parametric and non-parametric tests as appropriate. Ethical approval was obtained by the Institute of Psychiatry and South London and Maudsley Trust Ethics Committee, as well as by local ethics committees in the 3 collaborating centres. Signed informed consent for participation was obtained from all the subjects included in the study.

Childhood Eating and Weight in Eating Disorders

Psychother Psychosom 2007;76:234241

235

Measures EATATE Phenotype Interview (Part 1) This is a semi-structured interview, comprising a European adaptation of the Longitudinal Interval Follow-up Evaluation (LIFE) [15] and the Eating Disorders Examination (EDE) [16]. The interview is used to obtain a lifetime history of ED symptoms, which are then plotted on a lifeline. This instrument has been validated and has demonstrated good inter-rater reliability in terms of diagnoses ( 0.821.0) and illness history variables (0.800.99). Diagnostic validity (compared to clinical notes) yielded values between 0.77 and 1.0 for sequential diagnoses [17]. Each sister was interviewed separately. Age, educational level, current BMI, and lowest and highest ever BMI at current height were obtained as part of the research interview. Infant and Childhood Feeding and Eating Questionnaire This is a self-report questionnaire for mothers, devised by the authors, about infant feeding, childhood eating and weight history of their daughter(s). The questions were devised by the authors following an extensive literature review to identify factors that may be associated with the risk of developing an ED and based on existing well-validated childhood feeding and eating questionnaires: the Child Feeding Questionnaire [18] and Preschooler Feeding Questionnaire [19]. The questionnaire comprised 29 questions, divided into 4 sections which covered (i) weight and growth: birth weight in grammes; the child being underweight at the ages of 6 months, 1 year, 5 years and 10 years compared to most children their age; the child being overweight at the ages of 6 months, 1 year, 5 years and 10 years compared to most children their age; (ii) breastfeeding and feeding in infancy, i.e. selective eating in the first year of life (defined as having a strong preference for a few foods or refusing solids); (iii) weaning and moving to solid food, and (iv) eating problems during childhood (between the ages of 15 and 610): picky eating (being a fussy eater more than twice a month); digestive problems (frequent, i.e. twice a month or more, stomach aches and vomiting); difficulties in making their child eat; the child not eating enough at mealtimes compared to most children their age; pica (defined as eating non-food items: earth, dust, chalk, plaster, paint grass, etc.) and eating quickly compared to most children their age. Most questions had a 4-point scale answer from 1 (never) to 4 (most days, always). The questionnaires were translated into and back-translated from the 4 different languages, and difficulties in translation were addressed by discussions in research meetings involving principal investigators from the participating countries until a consensus was obtained. For the purpose of this study, only variables that had already been identified as risk factors for ED in previous studies were coded and analyzed, i.e. sections i and iv of the questionnaire and 1 question, selective eating in the first year of life, from section ii. Data Analysis Parametric (2, Fishers exact and t tests) and non-parametric tests (Mann-Whitney) were used as appropriate, after testing for normality, to compare socio-demographic variables between subjects and healthy sisters. The data were analyzed using SPSS (version 10 for Windows). Conditional logistic regression analyses, where groups were matched on belonging to the same family, were used to analyze

the data on eating problems in pairs of sisters discordant for AN or BN and to evaluate the role of eating behaviours in predicting ED status. Odds ratios (OR) were calculated using STATA (version 8 for Windows). All statistical tests presented are 2-tailed. Statistical significance was defined as p ! 0.05.

Results

Demographic and Clinical Characteristics Of the 150 pairs, 76 were recruited in Austria, 37 in Spain, 28 in the UK and 9 in Slovenia. Of the sample, 64% (n = 94) were recruited from participating tertiary referral centres in their respective countries and 36% from community samples. The sources of recruitment differed significantly (Fishers exact = 44.1, p ! 0.05) amongst samples from the different countries. All of the Spanish and 61% of the Austrian samples were recruited from a clinical setting, whereas in the UK and Slovenia, probands from clinical samples accounted for a lower proportion of those recruited (36 and 17%, respectively). Of the total sample, 99% (n = 149) were white Caucasian. Ethnic distribution did not differ between the countries (table 1). The affected sisters had a median age of 24 years (range 1552), with a median age of 24 for the unaffected sisters (range 1650). The sisters were similar in age, with a mean age gap of 3.4 years (82.3); the median age of the healthy sisters was comparable to that of the probands (table 1). The affected sister was the eldest of the sibling pair in 47.2 % of cases. Educational status was comparable amongst discordant sisters. The probands had a median BMI of 19.5 (range 12.5 31.5), the unaffected sisters one of 21.8 (range 1743.3). The BMI differed significantly between affected and unaffected sisters (Mann-Whitney U = 5,822.5; p ! 0.001). Of the probands, 82 (55%) had a lifetime diagnosis of AN, 68 (45%) a lifetime diagnosis of BN. The percentages of subjects with AN or BN recruited in each of the 4 centres differed significantly; Slovenia and the UK recruited more women with AN, whereas Spain recruited a higher percentage of women with BN (table 2). The countries of recruitment did not differ on mean age, education, age of onset or duration of illness of AN or BN probands. The median age of first established symptoms of an ED for subjects with a lifetime diagnosis of AN (index age) was 16 years (range 1136). The median duration of illness was 4 years (range 131). At the time of interview they had a mean BMI of 18.8 (SD 2.7). Women with BN had a median index age of 16 (range 1024) and a median duration of illness of 7 years (range 122). They had a median BMI of 20.7 (range 17.531.5) at interview. The 2
Micali et al.

236

Psychother Psychosom 2007;76:234241

Table 1. Socio-demographic characteristics: comparisons between ED women and unaffected sisters

ED group (n = 150) Mean age 8 SD, years Mean BMI at interview 8 SD Number attended secondary/university education 25.586.6 19.983.0 111 (75.5%)

Unaffected sisters (n = 150) 26.287.7 22.884.3 98 (73.1%)

p NS* <0.05* NS**

* p values shown were obtained by comparing median values with Mann-Whitney U test. ** p value obtained by 2 test.

Table 2. Sample description: ED women

AN = 82 (55%) Mean lowest ever BMI 8 SD Mean highest ever BMI 8 SD Mean age of onset 8 SD, years Mean duration of illness 8 SD, years Country of origin Austria UK Spain Slovenia 14.282.2 20.882.3 17.084.4 6.185.7 39 (51.3) 22 (78.6) 13 (35.1) 7 (77.8)

BN = 68 (45%) 17.383.3 24.184.3 16.883.0 7.484.9 37 (48.7) 6 (21.4) 24 (64.9) 2 (22.2)

p <0.005* <0.005* NS <0.05*

<0.05**

Mean values are shown for ease of description. The figures in parentheses represent the percentage. * p values obtained by comparing median values with Mann-Whitney U test. ** p value obtained by 2 test.

proband groups (AN and BN women) had comparable ages of onset, but differed significantly on duration of illness (Mann-Whitney U = 2,113.5; p ! 0.05). The mean minimum and maximum ever BMIs significantly differed according to proband diagnosis (t = 6.5, p ! 0.001; t = 5.4, p ! 0.001), as did BMI at interview (Mann-Whitney U = 1,314.5; p ! 0.001; table 2). Eating and Weight Problems during Childhood Weight Analyses of birth weight and weight gain throughout the first year of life showed no difference with respect to proband status, both in pairs discordant for AN [mean birth weight: 3.1 kg (3.03.3) for AN women and 3.2 kg (3.03.3) for healthy sisters] and for pairs discordant for BN [mean birth weight: 3.2 kg (3.03.4) for both BN women and healthy sisters]. In addition 42.5% of the women with AN and 27.5% of their sisters, and 45.1% of the women with AN and 29.1% of their sisters were deChildhood Eating and Weight in Eating Disorders

scribed as heavier compared to most children their age at 6 months and 1 year of age, respectively. Weight status at 6 months and 1 year was significantly predictive of developing AN, with an OR of 0.6 (p ! 0.01, 95% CI: 0.40.9) and 0.6 (p ! 0.01, 95% CI: 0.40.8; table 3). Fifty percent of the BN women and 37.9% of the unaffected sisters at the age of 5 and 49.2% of the women with BN and 37.9% of the sisters at the age of 10 were described as heavier compared to most children their age. Being heavier at 5 and 10 years was significantly associated with developing BN, OR = 2.8 (p ! 0.05, 95% CI: 1.17.8) and OR = 2.8 (p ! 0.05, 95% CI: 1.17.7; table 3). Feeding and Eating Difficulties Retrospective reports of feeding and eating difficulties are summarized in table 4. Selective eating in the first year of life was not associated with a diagnosis of AN or BN. Frequent picky eating, not eating enough, struggles to get the child to eat, eating quickly or slowly compared
Psychother Psychosom 2007;76:234241

237

Table 3. Weight during infancy and childhood: ORs for conditional logistic regression comparisons between ED women and unaffected sisters

AN (n = 82) versus unaffected sisters (n = 82) Low weight at 6 months Low weight at 1 year Low weight at 5 years Low weight at 10 years 0.6 (0.40.9)** 0.6 (0.40.8)** 0.6 (0.31.5) 0.6 (0.31.4) High weight at 6 months High weight at 1 year High weight at 5 years High weight at 10 years

BN (n = 68) versus unaffected sisters (n = 68) 1.0 (0.71.3) 1.1 (0.81.5) 2.8 (1.17.8)* 2.8 (1.17.7)*

Values represent OR, and 95% CI are shown in brackets; * p < 0.05; ** p < 0.01.

Table 4. Childhood feeding and eating difficulties: ORs for conditional logistic regression comparisons between

ED women and unaffected sisters AN (n = 82) versus unaffected sisters (n = 82) Selective eating in first year of age Eating quickly 15 years Eating quickly 610 years Picky eating 15 years Picky eating 610 years Pica 15 years Pica 610 years Vomiting 15 years Vomiting 610 years Stomach aches 15 years Stomach aches 610 years Eating enough 15 years (a lot of food vs. about right, not enough) Eating enough 610 years (a lot of food vs. about right, not enough) Struggle to get child to eat 15 years Struggle to get child to eat 610 years 1.0 (0.6-1.5) 1.1 (0.81.6) 1.1 (0.71.6) 1.1 (0.81.4) 0.9 (0.71.2) not testedb not testedb 0.9 (0.61.4) 1.5 (0.92.6) 1.1 (0.71.6) 1.1 (0.71.6) 0.8 (0.51.2) 1.0 (0.61.6) 0.8 (0.61.2) 0.9 (0.61.3) BN (n = 68) versus unaffected sisters (n = 68) not testedb 1.7 (1.03.0) 2.3 (1.33.8)* 0.5 (0.31.0) 0.6 (0.40.9)* 1.4 (0.72.9) not calculateda 0.7 (0.31.5) 0.6 (0.21.3) not calculateda 0.8 (0.41.9) 1.2 (1.01.3) 1.3 (1.11.5)** not testedb not testedb

Values represent OR, and 95% CI are shown in brackets; * p < 0.05; ** p < 0.01. a The predictor was absent in all subjects tested, it was therefore not possible to calculate an OR. b These factors were not hypothesized as relevant and were therefore not tested.

to most children their age, frequent vomiting and stomach aches between the ages of 15 and 610 were not associated with the development of AN. Picky eating between the ages of 6 and 10 was reported significantly less often by mothers for their daughters with BN (6.2%) compared to their healthy daughters (19.7%), with an OR of 0.6 (p ! 0.05, 95% CI: 0.40.9). There was a trend for low picky eating between the ages of 1 and 5 to be associated with later BN. However, the difference did not reach statistical significance. The women with BN were more
238
Psychother Psychosom 2007;76:234241

often described as eating a lot between the ages of 6 and 10 compared to their sisters (BN: 64.7%, sisters: 52.9%; OR = 1.3; p ! 0.05, 95% CI: 1.11.5) and between the ages of 1 and 5 (BN: 57.4%, sisters: 50%), although this difference was not statistically significant. Pica and early digestive problems showed no relationship to a diagnosis of BN. Eating quickly between the ages of 5 and 10 was also commoner in the probands with BN (19.2%) compared to their healthy sisters (4.4%) and was significantly associated with a BN diagnosis (OR = 2.3, p ! 0.005, 95% CI:
Micali et al.

1.33.8). Eating quickly between the ages of 1 and 5 was higher in the probands with BN (11.7%) as compared to the healthy sisters (4.4%). However, the difference did not reach statistical significance (p = 0.06).

Discussion

We found that women with AN were more often described as overweight at the ages of 6 months and 1 year compared to their sisters. This has never been described in the literature. It might be that maternal reports are less precise at that stage of life or that AN women are bigger during the first year of life. These findings need replication. Bulimia Nervosa Whilst childhood undereating and digestive problems have been linked to AN, traits of childhood overweight and overeating have often been identified as risk factors for BN. We found that picky eating protects against the development of BN (OR = 0.6) in later life, consistent with previous research [2]. Interestingly we did not find a relationship between pica or digestive problems and later BN. A recent report of adult women with BN (135) showed that gastrointestinal complaints were retrospectively reported in one third of women [8]. The fact that the subjects in this study were recruited for a treatment study might point to a highly selected sample. Moreover self-report of events, such as constipation or stomach aches, might be subject to recall bias. Previous reports have provided inconsistent findings on the role of these predictors [2, 3]. This area requires more research. In our study, eating a lot (OR = 1.3), especially in late childhood, significantly predicted BN in probands. Eating too little has previously been identified as a protective factor for BN [3]. Our results confirm these reports. Surprisingly, eating quickly in late childhood also predicted a diagnosis of BN (OR = 2.3). This has not been previously reported in the literature. We found that being overweight, at 5 and 10 years of age, was significantly associated with BN in adulthood (OR = 2.8). This validates previous case-control [6, 7] and longitudinal reports [810, 24]. It lends weight to the hypothesis that BN might lie on a spectrum with overweight and obesity [25], and that weight problems might be a non-shared risk factor for ED. Jacobi et al. [1] showed the persistence of abnormal eating patterns, in particular picky eating, during childhood, suggesting that childhood picky eating might denote a pattern of inhibited eating that remains stable throughout childhood and later life. We found an association of maternal reports of phaenotypically similar behaviours, i.e. low picky eating, eating a lot and eating quickly (all clearly related to overeating) with overweight in children who later develop BN. This leads to the plausible explanation that if, as Jacobi et al. [1] suggest, picky
Psychother Psychosom 2007;76:234241

This study used a case-control design, where unaffected sisters closest in age to the affected subject were recruited as controls. This design is a very powerful tool, as it allows controlling for family and cultural factors, such as parental beliefs and their attitudes towards eating, weight and food, along with parental eating problems. These are often unmeasured and ignored, but they are very relevant to the development of ED [20]. Childhood Eating and Weight and Anorexia Nervosa We found no evidence that childhood eating difficulties, especially digestive problems and picky eating, predicted AN when comparing affected and unaffected sisters. We did not replicate the findings by Marchi and Cohen [2], although our results are in line with those of Kotler et al. [3]. In the former study [2], there was some time overlap between predictor variables (feeding problems between the ages of 1 and 10) and outcome (anorexic symptoms between 9 and 18 years). The reliability of a comparison to both studies is limited by the fact that the outcomes used were anorexic and bulimic symptomatology, with very low rates of full and partial AN diagnosis. We failed to confirm the findings by Rastam [21] of a higher prevalence of childhood gastrointestinal problems and extreme feeding difficulties in 51 adolescents with AN compared to controls. However, high rates of bulimic behaviours were reported in the probands in this study, making it difficult to conclude whether the findings reported are specific to AN or BN. Our study had well-defined AN and BN groups, an older population group and focused on differences in specific feeding/eating problems. Higher rates of childhood feeding problems were also found in inpatients with ED compared to polysubstance dependence individuals and normal controls in a recent study [22]. Our findings contrast with the results of a pilot study on risk factors for AN from our group [23], which found a higher self-reported rate of childhood feeding problems in subjects with AN compared to unaffected sisters. However, this result was based on 1 question only, and affected subjects were used as informants.
Childhood Eating and Weight in Eating Disorders

239

eating is part of an inhibited eating style, eating too much, low rates of picky eating and eating quickly in childhood could be part of a more disinhibited eating style. Whether a disinhibited eating style is part of a specific bulimic phenotype is not yet clear, but it has certainly been suggested in the past [26]. Our findings further support this idea and suggest that this style is of relevance to the development of BN. Eating difficulties and under- and overweight status are common in childhood. We focused on aspects of childhood eating and weight, which are potentially modifiable early markers for ED. Individuals who develop BN are more likely to overeat and be overweight, and this disinhibited eating style seems to be a risk factor specific to the sibling who then develops BN, and it may be part of the genetic propensity to develop this disorder. These findings suggest that preventive measures for BN might usefully be embedded within preventative approaches to obesity. This study could aid future longitudinal studies focusing on the developmental patterns of feeding, eating difficulties in childhood and their relationship with ED. Strength and Limitations of the Study Design Unlike longitudinal studies, our study relies on a large sample of clinically defined cases, including many cases of AN. Moreover it relies on both clinical and community cases. The sisters were assessed with a standardized instrument to exclude the possibility of a lifetime ED. For what concerns limitations, childhood eating and weight were based on retrospective maternal accounts, which may have been subject to recall bias, decay in memory and a search for meaning. Some questions included

in the maternal questionnaire might have been subjective, although we tried to use clear behavioural definitions. Potential sources of bias include the fact that only subjects who had a sister and a mother willing to participate in the study were included. We recruited our sample from tertiary referral centres and from women who had participated in previous research. This may have led to more chronic and ill women being overrepresented. However, women with ED are often treated in tertiary referral centres irrespective of the severity and chronicity of their ED across Europe. We were not able to determine risk factors varied in each of the recruiting centres, due to the small sample size across the centres. Lastly the ratios of subjects with BN or AN differed across recruitment centres, which might have resulted from the different recruitment strategies, rather than diagnostic discrepancies, as a standardized lifetime ED diagnostic tool was used. It might also reflect the services set-up in different European countries, as women with BN are often treated in secondary referral services in the UK, whilst they are more often seen in tertiary services in other European countries.

Acknowledgements
We are grateful to Prof. E. Taylor and Prof. E. Simonoff for their very helpful comments. This study was supported by the European Commission Framework V programme (QLK1-1999916), a grant from the PPP foundation (1206/87) to J.L.T. and D.C., and partially by FIS (G3/184). The authors of this paper won the Margaret Davenport Prize at the Royal college of Psychiatrists Child and Adolescent Faculty Meeting, September 2004, for presenting this research paper.

References
1 Jacobi C, Agras WS, Bryson S, Hammer LD: Behavioral validation, precursors and concomitants of picky eating in childhood. J Am Acad Child Adolesc Psychiatry 2003; 42: 76 84. 2 Marchi M, Cohen P: Early childhood eating behaviors and adolescent eating disorders. J Am Acad Child Adolesc Psychiatry 1990;29: 112117. 3 Kotler LA, Cohen P, Davies M, Pine DS, Walsh BT: Longitudinal relationships between childhood, adolescent and adult eating disorders. J Am Acad Child Adolesc Psychiatry 2001;40:14341440. 4 Shoebridge P, Gowers SG: Parental high concern and adolescent-onset anorexia nervosa: a case-control study to investigate direction of causality. Br J Psychiatry 2000; 176: 132 137. 5 Mitchell JE, Boutacoff L, Wilson D: Absence of early feeding problems among bulimic women: observations from parental interviews. Am J Orthopsychiatry 1986; 56: 313 316. 6 Fairburn CG, Welch SL, Doll HA, Davies BA, OConnor ME: Risk factors for bulimia nervosa: a community-based case-control study. Arch Gen Psychiatry 1997; 54: 509 517. 7 Fairburn G, Doll HA, Welch SL, Hay PJ, Davies BA, OConnor ME: Risk factors for binge eating disorder: a community-based, case-control study. Arch Gen Psychiatry 1998;55:425432. 8 Killen JD, Taylor CB, Hayward C: Pursuit of thinness and onset of eating disorder symptoms in a community sample of adolescent girls: a three-year prospective analysis. Int J Eat Disord 1994;16:227238. 9 Patton GC, Johnson-Sabine E, Wood K, Carlin JB, Wolfe R: Abnormal eating attitudes in London schoolgirls a prospective epidemiological study: outcome at twelve-month follow-up. Psychol Med 1990; 20:383394.

240

Psychother Psychosom 2007;76:234241

Micali et al.

10 Wichstrom L: Psychological and behavioral factors unpredictive of disordered eating: a prospective study of the general adolescent population in Norway. Int J Eat Disord 2000; 28:3342. 11 Killen JD, Taylor CB, Hayward C, Haydel KF, Wilson DM, Hammer LD, Kraemer HC, Blair-Greiner A, Strachowski D: Weight concerns influence the development of eating disorders: a 4-year prospective study. J Consult Clin Psychol 1996; 64:936940. 12 Patton GC, Selzer R, Coffey C, Carlin JB, Wolfe R: Onset of adolescent eating disorders: population-based cohort study over 3 years. BMJ 1999;318:765768. 13 American Psychiatric Association: Diagnostic and Statistical Manual of Mental Disorders. Washington DC, American Psychiatric Association, 1994. 14 Kaye WH, Lilenfeld LR, Berrettini WH, Strober M, Devlin B, Klump KL, Goldman D, Bulik CM, Halmi KA, Fichter MM, Kaplan A, Woodside DB, Treasure J, Plotnicov KH, Pollice C, Rao R, McConaha CW: A search for susceptibility loci for anorexia nervosa: methods and sample description. Biol Psychiatry 2000;47:794803.

15 Keller MB, Lavori PW, Friedman B, Nielsen E, Endicott J, McDonald-Scott P, Andreasen NC: The Longitudinal Interval Follow-up Evaluation: a comprehensive method for assessing outcome in prospective longitudinal studies. Arch Gen Psychiatry 1987; 44: 540 548. 16 Fairburn CG, Cooper Z: The eating disorder examination; in Fairburn CG, Wilson GT (eds): Binge Eating. Nature, Assessment and Treatment (ed 12). New York, Guilford Press, 1993, pp 317331. 17 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; 160:242247. 18 Birch LL, Fisher JO: Mothers child feeding practices influence daughters eating and weight. Am J Clin Nutr 2000;71:10541061. 19 Baughcum AE, Powers SW, Johnson SB, Chamberlin LA, Deeks CM, Jain A, Whitaker RC: Maternal feeding practices and beliefs and their relationships to overweight in early childhood. J Dev Behav Pediatr 2001; 22:391408. 20 Dick DM, Johnson JK, Viken RJ, Rose RJ: Testing between-family associations in within-family comparisons. Psychol Sci 2000;11: 409413.

21 Rastam M: Anorexia nervosa in 51 Swedish adolescents: premorbid problems and comorbidity. J Am Acad Child Adolesc Psychiatry 1992;31:819829. 22 Mangweth B, Hausmann A, Danzi C, Walch T, Rupp CI, Biebl W, Hudson J, Pope H: Childhood body focused behaviours and social behaviours as risk factors for eating disorders. Psychother Psychosom 2005;74:247 253. 23 Karwautz A, Rabe-Hesketh S, Hu X, Zhao J, Sham P, Collier DA, Treasure JL: Individualspecific risk factors for anorexia nervosa: a pilot study using a discordant sister pair design. Psychol Med 2001; 31:317329. 24 Graber J, Brooks-Gunn J, Paikoff R, Warren MP: Prediction of eating problems: an eightyear study of adolescent girls. Dev Psychology 1994;30:823834. 25 Collier DA, Treasure JL: The aetiology of eating disorders. Br J Psychiatry 2004;185:363 365. 26 Connan F, Stanley S: Biology of appetite and weight regulation; in Treasure J, Schmidt U, van Furth E (eds): Handbook of Eating Disorders. London, Wiley, 2003, pp 6388.

Childhood Eating and Weight in Eating Disorders

Psychother Psychosom 2007;76:234241

241

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.