This editorial was prompted by the paper written by Watson & Andersen (1) in this issue on weight and amenorrhoea as diagnostic criteria for anorexia nervosa (AN). However, there is much else to question about the classification of eating disorders. An ideal classification should consist of categories that are mutually exclusive and collectively exhaustive. Its entities should be discreet and together they should cover the ground. The classification of eating disorders in the two main systems, ICD 10 and DSM IV (2, 3) measures up to these standards rather poorly (4). In each, the canon contains only two major categories – AN and bulimia nervosa (BN). AN has low-weight and BN has binge-eating as an essential criterion. The two disorders share the criterion of what in broad terms might be described as an over-concern about body weight and size although some would see an important difference in degree or emphasis in the typical ideas held by sufferers from AN and BN. In DSM-IV, AN takes precedence over BN in the sense that the presence of the former bars the diagnosis of the latter but the reverse is true in ICD10. There is in DSM-IV, however, a new subclassification of AN into binge-purging and pure restricting subtypes. The rules in these sets of criteria represent different responses to the fact that low-weight and bingeing occur together commonly and that the cardinal features of AN and BN may co-exist even in cross-section. When longitudinal course over time is considered, then the overlap becomes even more striking. In many series, a substantial minority of BN sufferers have a past history of AN. The reverse transition from BN to AN is less common, but it does occur. Thus AN and BN are far from being entirely discreet disorders and can be made to seem so only by the use of somewhat abitrary rules of definition. Thus the classification of the eating disorders fails to provide discreet entities that are truly, mutually exclusive. However, it fails even more to cover the ground. Many people present with clinically relevant eating disorders that fulfil criteria for neither of the two main eating disorders. How are these to be classified?
DSM-IV provides a single catch-all diagnosis for eating disorders that are neither AN nor BN, namely eating disorder not otherwise specified (EDNOS). This is essentially a diagnosis of exclusion although the diagnosis of binge eating disorders (BED) is included as a provisional subcategory of EDNOS ‘for further study’. However, in practice, BED has come already to be accorded the status of a diagnosis in its own right at least in the USA. ICD10 by contrast provides the additonal categories of ‘atypical AN’ and ‘atypical BN’ together with a number of rarely used entities and its own residual category ‘other eating disorders’.
EDNOS is common. In many clinical series it is the single most common diagnosis, and in some eating disorder services EDNOS patients form the majority of cases. Many people suffer from clinical eating disorders without ever fulfilling criteria for either AN or BN. Furthermore, a considerable cross-over occurs from AN to BN, and also a smaller amount of cross-over from BN to AN and EDNOS may be the only permissible diagnostic entity in the interim periods. What is the clinical meaning of this? There are many unanswered questions. For instance, is EDNOS in a patient, changing from AN to BN the same as that of changing from BN to AN? And what about course and outcome in the long-term? There are clear differences in mortality between typical AN and typical BN, AN having a well-documented excess mortality in treatment seeking samples (but not in population-based studies) (5–7). In BN, the evidence on mortality is inconclusive at present. Is this true of atypical cases? Likewise, for AN there is evidence of a significantly increased fracture risk from a population-based study (8) and a nationwide register study (9). The dire consequences with respect to procreation, education and social life is amply documented for former AN-patients, both in clinical and population-based studies (7). Is the same true of EDNOS or rather, which cases of EDNOS are at particular risk? And what should be offered as treatment to people suffering from EDNOS? The research literature tends to select patients for inclusion in trials using the main diagnostic categories. It is largely silent on the appropriate treatment of EDNOS.
ICD10 and DSM IV are the products of committees informed by tradition. To be useful, our classifications need to have some stability but we should be critical of them. We should not believe in them too much. The categories they advocate are useful tools and we cannot and should not do without them. However, the use of either tool box has problems. They share the most egregious defect of excluding from clear categorization at least a substantial minority of those who suffer from a clinically important eating disorder. There is room for a measure of conservatism but we cannot be satisfied until the EDNOS issue is more adequately addressed.