Many eating disordered patients migrate between diagnostic categories, particularly from the initial diagnosis 12. A diagnostic crossover from Anorexia Nervosa (AN) to Bulimia Nervosa (BN) is one of the most commonly observed 23. Particularly for patients with initial diagnoses of AN restricting-type (AN-R), treatment is said to be difficult because many develop binging and vomiting. Previous studies of patients with an initial diagnosis of AN-R have reported that 62% changed to the AN binge-eating/purging type (AN-BP) 4 and 21~36% changed to BN 56. These crossovers usually occurred within the first five years after onset of the eating disorder. Fichter et al. 7 found that patients crossed over from AN-R to AN-BP at a rate of 16.7% after two years, 10.3% after six years, and 3.7% after 12 years, and that patients crossed over from AN-R to BN at a rate of 13.3% after two years and 3.4% after six years.

These lines of evidence suggest that predicting the possibility of diagnostic crossover from AN-R at an early stage would be useful in treatment, because bulimia and purging behavior predict an unfavorable prognosis 89 and an increase in time to remission 5. Only a few studies have examined the psychological factors possibly related to such crossovers from AN-R. Tozzi et al. 6 found that higher levels of 'Parental criticism' (a subscale of the Multidimensional Perfectionism Scale [MPS] for measuring perfectionism) and lower levels of 'Self-directedness' (a subscale of the Temperament and Character Inventory [TCI] for measuring personality) predicted crossover from AN-R to BN.

However, a relationship between these factors and depression has been documented 1011. Depression is generally prevalent among women with eating disorders 1213. Furthermore, several studies have suggested that patients with AN-BP are more depressive than those with AN-R 1213, and that patients with BN are more depressive than those with AN 14. Thus, it is possible that the degree of depression may be a major psychological factor related to the crossover from AN-R to BN. Indeed, the necessity of considering state variables such as depression in examinations of personality traits has been shown 1015.

In addition, regarding the factors of body mass index (BMI) and obesity, patients with BN were higher than patients with AN-R on maximum BMI in the past 1617 and childhood obesity is a reported risk factor for BN 1. The relationship of a tendency to gain weight easily and the crossover from AN-R should be investigated.

Therefore, the aim of this study was to examine psychological factors and the tendency toward obesity related to the crossover from AN-R, and to examine the relationship between these psychological factors and depression.

In comparison with patients with persistent AN-R, those who developed BN were significantly higher in both current BMI and maximum BMI in the past. They also scored significantly higher in 'Parental criticism' and lower in 'Self-directedness', but both these trait factors disappeared when depression scores were used as a co-variant. No significant differences were observed for any of the personality traits surveyed among patients with persistent AN-R irrespective of their duration of illness.

A tendency toward obesity among patients who cross over from AN-R to BN is reported by previous studies 1617. Because childhood obesity is a predisposition of BN 1, the present results suggest that patients with AN-R at onset who were overweight in the past are inclined to gain weight and will develop BN over the course of their illness. This finding is compatible with recent genetic studies 24. The chromosomal region 10p, a susceptibility factor for BN, has been implicated in obesity 24 and the preproghrelin gene single nucleotide polymorphisms, another susceptibility factor for BN 18, were associated with higher current and maximum BMI in the past among female students 1825. In addition, the above genetic polymorphisms were related to higher 'Drive for Thinness-Body Dissatisfaction' scores measured by the Eating Disorder Inventory (EDI) 1825.

Consistent with previous findings 6, high 'Parental criticism' and low 'Self-directedness' were associated with crossover from AN-R to BN in Japanese patients, indicating no ethnic or cultural effects on the psychological factors related to crossover from AN-R to BN. The finding of higher 'Parental criticism' scores of patients who crossed over from AN-R to BN is consistent with previous findings: lack of expressed empathy/affection from the parents predicted binge eating in AN-R patients 5; stronger perception of deficits in parental nurturance in BN patients than in AN patients 26; and 'Parental criticism' scores predicting bulimia scores measured by the EDI during a stressful situation for normal female high school students 27. Families in which members are criticized may be weak in support for their members undergoing treatment, substantially inhibiting the treatment process and inclining restrictors to begin binging and vomiting. Moreover, maternal critical comments 28 and lack of parental care 29 are associated with poor outcomes in eating disorder patients. Thus, this study further suggests the importance of a supportive family system for the disease course of eating disordered patients.

The basic concept of 'Self-directedness' is "self-determination and 'willpower' or the ability of an individual to control, regulate, and adapt their behavior to fit the situation in accord with individually chosen goals and values"21. Patients with eating disorders who manifest low 'Self-directedness' are not capable of continuing the restriction behaviors that help them get thinner. Accordingly, these patients may begin binging and/or vomiting, which requires a change in diagnosis. Actually, 'Self-directedness' scores were lower in BN than AN-R 161730. Moreover, low 'Self-directedness' is associated with poorer outcome in AN patients 2931. Based on these lines of evidence, our findings suggest the importance of paying careful attention to 'Self-directedness' in the treatment process of AN-R patients.

We also observed the possibility that crossover from AN-R to BN is related to depression. While depression predicts future increases of binge eating in normal females 323334 and in eating disorder patients 35, Tenconi et al. 36 reported that the degree of depression at baseline of patients with an initial diagnosis of AN-R did not predict the onset of binge eating. It is possible that depression may occur during the crossover period and that the depression may be caused by binging and vomiting. This hypothesis is supported by a study showing that the symptoms of eating-related concerns prospectively predict the onset of depression in adolescent girls 37. In our study, relationships between depression and 'Parental criticism' and 'Self-directedness' were suggested. It is possible that trait factors such as 'Parental criticism' and 'Self-directedness' create a susceptibility to the development of depression, as low 'Self-directedness' was found to predict depression in normal undergraduates 38. Otherwise, it is possible that the self-evaluation of trait factors was biased by a depressive mood 1539. The design of the present study, however, does not allow for an absolute determination of causation. Longitudinal investigation is necessary to confirm the relationship between crossover and the depression of patients with an initial diagnosis of AN-R.

In contrast to the findings for the patients who developed BN, patients who crossed over from AN-R to AN-BP showed no differences in psychological factors from those with AN-R only. This finding is supported by a cross-sectional study comparing the MPS scores of AN-BP and AN-R patients, where no significant differences were observed 40. Regarding the TCI scores, however, the findings in different studies are not consistent: no significant differences between AN-R and AN-BP 1316; higher 'Novelty seeking' in AN-BP than AN-R 304142; and higher 'Self-directedness' in AN-R than AN-BP 4142. Because 'Novelty seeking' and 'Self-directedness' are reported to be associated with personality disorders 4344 and alcohol abuse 45, it is possible that the co-morbidity of Axis I and II disorders merely contributes to those differences 304142. In addition, both of these AN sub-type groups were different in 'duration of illness'; this should be considered as a possible confounding factor, although there were not adequate numbers of subjects for a proper comparison. Further studies are needed regarding psychiatric co-morbidities and duration of illness among AN-R patients who develop AN-BP.

Scores on personality measures are also reported to be influenced by the degree of recovery from eating disorders 2946. In the present study, only those patients in an active state of illness were investigated. Although personality assessment and function might be influenced by the symptoms themselves (e.g., starvation) in an actively ill state 47, it is generally reported that personality is rather consistent 4849, and temperament and character are independent of body weight in AN 41. However, the number of patients excluded from our study because of their recovery at the time of the investigations was rather small; five with AN-R only, one who crossed over from AN-R to AN-BP, and four who crossed over from AN-R to BN. It is impossible to precisely compare the influence of symptoms of those in an active state of illness and those who have recovered.

Based on the recent findings that the highest rate of crossover was observed within the first five years after the onset of AN-R 56, we compared the patients who crossed over five years or more after the onset (n = 4) and those who crossed over within five years (n = 32). There were no significant associations of any of the psychological characteristics with the duration of illness and no significant differences on the MPS, TCI, or BDI-II between the two groups (data are not shown). Therefore, the influence of duration of AN-R is rather small on the psychological characteristics of these patients in the current study.

Finally, the present findings have clinical implications for the treatment of patients with eating disorders. Investigating the tendency toward obesity or a depressive state in a clinical setting will help us to predict crossover from AN-R to BN. Examining patient characteristics such as 'Parental criticism' and 'Self-directedness' may predict not only crossover but also, at least in part, the outcome of treatment. Cognitive and behavioral treatment, as well as nutritional interventions, for such risk factors will be necessary.

The current results present the demographic and psychological characteristics of patients who cross over from AN-R to BN: a tendency toward obesity and the psychological characteristics of low self-directedness and high parental criticism. However, these psychological traits may also be associated with depression. Further research is needed to clarify the causal role of depression and its relationship to the identified psychological characteristics of patients who cross over.

The author(s) declare that they have no competing interests.

HN performed the statistical analysis and drafted the manuscript. GK conceived the study design, collected the clinical data set, and drafted the manuscript. TA participated in the design of the study and collected the clinical data set. TNakahara, TO, KK, T. Nagata, AN, YO, KOkabe, MK, CY, SS, KOhkuma, KN, TNaruo, MT, NK, TI and the Japanese Genetic Research Group for Eating Disorders collected the clinical data set. All authors read and approved the final manuscript.