Have case-register studies of eating disorders incidence reached their use-by date?

Studies of the epidemiology of eating disorders (EDs) have a chequered history (Patton and Szmukler, 1995). Serious interest in the field began only in the 1960s. Prompted by clinical speculation that the prevalence of anorexia nervosa (AN) might be increasing, researchers began to examine trends in the incidence of this disorder, relying on psychiatric case registers and hospital records. This research in turn stimulated the development of explicit diagnostic criteria. During the 1970s, the occurrence of purging behaviours, namely, self-induced vomiting and misuse of laxatives and/or diuretics, was noted in some AN patients. Subsequently, the observation that such behaviours were being employed by women of normal weight led to the description of bulimia nervosa (BN) as an ED distinct from AN. Speculation that the prevalence of BN might also be increasing followed. The relatively higher base rates of BN and variants of this disorder permitted a shift from studies of treated incidence to population- and primary-care-based studies of both incidence and prevalence. More recently, attention has turned to ED characterised by binge eating in the absence of extreme weight-control behaviours, culminating in the inclusion of binge eating disorder (BED) as a formal diagnosis in Diagnostic and Statistical Manual of Mental Disorders–Fifth Edition (DSM-5).

Methodological challenges inherent in psychiatric epidemiology research are such that answers to apparently simple questions, such as whether the prevalence of a given disorder has increased, remain elusive. In the case of epidemiological studies of ED, two key challenges account for much of the difficulty. First, the population prevalence of ED characterised by low body weight, namely, AN and variants of AN, and of cases of BN meeting formal diagnostic criteria, is low, less than 2% combined. Even in large general population surveys, therefore, the number of these cases identified is small (Mond et, al., 2007). The problem of low base rates is exacerbated by the known tendency for individuals with ED to be over-represented among non-respondents in epidemiological studies.

For these reasons, epidemiological studies of AN and, to a lesser extent, BN have relied on clinical samples of individuals – typically the ‘high-risk’ demographic of adolescent and young adult females – receiving specialist treatment. The problem with this approach is that many, if not most, individuals with these disorders never receive treatment for an eating problem from a mental health professional (Mond et al., 2007). This includes individuals who receive specialist treatment for a comorbid mental health problem but not an ED, those who receive treatment for an ED and/or other mental health problem in primary care or in private practice only and those who receive treatment only for eating- or weight-related medical complications. Given the stigma associated with mental health problems, the denial associated with AN, the shame and secrecy associated with BN and BED and poor ‘ED mental health literacy’ more generally, each of these proportions is likely to be substantial. These patterns of health service utilisation constitute the second key challenge in epidemiological studies of ED.

Superimposed on these issues are challenges common to all epidemiological studies of mental health problems. These include variability across studies and over time with respect to population characteristics and sampling methods, diagnostic criteria and operational definitions of these, methods of assessment, service availability and/or cost and referral, registration and treatment practices. Changes in diagnostic criteria have been particularly problematic in ED research, as the architects of classification schemes have sought to adapt to changes in the presentation of ED over time and respond to concerns that as many as half of individuals with EDs, in both community and clinical samples, fall into residual categories such as ‘atypical eating disorder’ and ‘Eating Disorders Not Otherwise Specified’.

In view of these multiple challenges, it is not surprising that the issue of whether ED became more common in the latter half of the 20th century was never satisfactorily resolved. What could reasonably be concluded, however, was that there was a substantial and progressive increase in the incidence of adolescent and young adult females with AN receiving primary care and/or specialist treatment for an ED or other mental health problem in the latter half of the 20th century, peaking in the 1970s before levelling off in the 1980s. A similar increase, and subsequent levelling off, in the treated prevalence of BN occurred following its emergence in the late 1970s. Whereas population-level change may have played a part in these trends, increased awareness and understanding of ED, and consequent increases in help seeking and improvements in case identification, probably played a larger part.

What then to make of the research of Gammelmark et al. (2015) in this issue? For the reasons outlined above, research of this kind is of little value in informing population-level change in ED occurrence. For this purpose, consecutive population-based surveys of symptom prevalence and/or primary-care-based incidence studies are a better bet. Even as estimates of the treated incidence of AN, figures based on psychiatric-registers are underestimates because many individuals with AN do not receive specialist treatment for an ED. Research of this kind is also of limited benefit in informing change in service use among males with ED and among individuals with ED other than AN, individuals with BED in particular. This is surely a problem given that BED is the most common ED and relatively common in males (Mond et al., 2014).

Have case-register studies of ED incidence reached their use-by date though? For me, the answer to this question is ‘no’. While these studies do not inform population-level change in ED occurrence and relate to only a small proportion of all ED cases, research of this kind is nevertheless valuable in informing change over time in the number of cases of certain ED receiving specialist treatment and in suggesting possible reasons for this. The authors’ conclusion that the most plausible explanation for the observed increases in treated incidence is improved public awareness and understanding of the nature and adverse consequences of ED is notable because there has been relatively little interest in ED mental health literacy thus far and little in the way of evidence that this may have improved in the same way that improvements in mental health literacy relating to other mental health problems have been documented (Mond, 2014).

If studies of this kind contribute, albeit indirectly, to improved ED mental health literacy, then they are worthwhile for that reason alone. Indeed, if community awareness and understanding of the individual and community health burden of ED is in fact improving, then the limitations of case-register studies in terms of underrepresentation of males with ED and individuals with BED may gradually diminish, as more of these individuals seek and receive mental health care. Of course, increases in demand for mental health care, possibly combined with population-level increases in prevalence, will inevitably place pressure on already strained services and focus attention on the costs and benefits of treatment. Whether and to what extent current treatments for ED are cost-effective, or even effective, is another debate.