Psychological Medicine, 1990, 20, 671-680 Printed in Great Britain Prevalence of three bulimia syndromes in the general population JOHN A. BUSHNELL, 1 J. ELISABETH WELLS, ANDREW R. HORNBLOW, MARK A. OAKLEY-BROWNE AND PETER JOYCE From the Departments of Community Health and General Practice, and Psychological Medicine, Christchurch School of Medicine, University of Otago, Christchurch, New Zealand SYNOPSIS Prevalence of bulimia was estimated from a cross-sectional general population survey of 1498 adults, using the Diagnostic Interview Schedule (DIS) administered by trained lay interviewers. Lifetime prevalence of the DSM-III syndrome in adults aged 18-64 was 10% and this was concentrated in young women: in women aged 18-44 lifetime prevalence was 2-6%, and 10% currently had the disorder. Based on clinicians' reinterviews of random respondents and identified and marginal cases, the prevalence of current disorder using criteria for draft DSM-III-R bulimia was 0-5%, for DSM-III it was 0-2%, and for Russell's Criteria bulimia nervosa 00%. A strong cohort effect was found, with higher lifetime prevalence among younger women, which is consistent with a growing incidence of the disorder among young women in recent years. Although elements of the syndromes were so common as to suggest that dysfunctional attitudes to eating and disturbed behaviour surrounding eating are widespread, there was little evidence of the bulimia syndrome having become an epidemic on the scale suggested by early reports. INTRODUCTION Limitations of prevalence studies Since the formulation of criteria for the definition of bulimia nervosa (Russell, 1979), and bulimia (American Psychiatric Association, 1980), the eating disorder has been the subject of a flurry of survey activity (Stangler & Printz, 1980; Johnson et al. 1982; Fairburn & Cooper, 1982, 1984; Pope et al. 1984). Early studies which attempted to establish the prevalence of the disorder obtained very high rates of bulimia (e.g. Halmi et al. 1981), fuelling media reports of an 'epidemic' of the disorder of huge proportions. Many attempts to estimate prevalence have been constrained by methodological inade- quacies. There has been a proliferation of questionnaires designed to assess the syndrome, many of which have not appraised all the features of the DSM-III syndrome and have uncertain specificity (Halmi et al. 1981; Johnson 'Address for correspondence: John Bushnell, Department of Community Health and General Practice, Christchurch School of Medicine, University of Otago, PO Box 4345, Christchurch, New Zealand. et al. 1984; Nevo, 1984; Ben-Tovim, 1989). With the exception of a few studies (Pope et al. 1985; Pyle et al. 1983, 1986), prevalence studies have generally not shown clearly whether preva- lence rates are for lifetime occurrence or current disorder. Several studies have confused incidence with prevalence (Pyle et al. 1983; Johnson et al. 1984; Gray & Ford, 1985). In a disorder in which guilt and secrecy are common features (Fairburn & Cooper, 1982; Abraham & Beumont, 1982), heavy reliance has been placed on self-report questionnaires. The use of inter- views to explore the clinical relevance of symptoms has seldom been undertaken, and where it has, the adequacy of self-report questionnaires alone has been questioned (Pyle et al. 1986; Schotte & Stunkard, 1987). Fur- thermore, the choice of population to be sampled, the narrow age-range studied, and the method of selection (Pyle et al. 1983; Crowther et al. 1985; Hart & Ollendick, 1986; Zuckerman era/. 1986) and low participation rates (Crowther et al. 1985; Pope et al. 1985; Hart & Ollendick, 1986; Drenowski et al. 1988) limit the generaliz- ability of many findings. 671