Background
Binge-eating disorder (BED), a recent addition to the DSM-5 [
1], is marked by recurrent episodes of binge eating in the absence of recurrent inappropriate compensatory behaviors. Global DSM-IV lifetime prevalence estimates for BED are approximately 1.9 % in adults [
2]. An internet-based study of 22,397 adults [Validate Attitudes and Lifestyle Issues in Depression ADHD and Troubles with Eating (VALIDATE)], reported lifetime prevalence estimates of DSM-5 BED projected to the US population to be 2.03 %, (1.41 % for men and 2.61 % for women) [
3]. Age at onset is typically late teens to early twenties [
2,
4]. The illness can be protracted and can wax and wane [
2,
4‐
8]. We explored fundamental features of BED including clinical characteristics, diagnostic flux, and psychiatric comorbidities, using rich, extensive, Swedish population register data.
Diagnostic flux (i.e., migration from one eating disorders presentation to another) commonly occurs between anorexia nervosa (AN) and bulimia nervosa (BN) [
5,
6,
8‐
11]; however, less is known regarding flux in BED. Crossover between BED and BN [
5,
6] and between BED and eating disorder not otherwise specified (EDNOS) [
5] has been reported; however, large-scale, longitudinal, population-based explorations are lacking.
Psychiatric comorbidity burden is considerable in individuals with BED with the most common co-occurring conditions being mood [
12‐
20], anxiety [
13‐
20] and substance use disorders [
12,
14,
16,
17,
20]. Using national patient records, we are able to broaden the inquiry to include more psychiatric disorders. An important question addressed in the literature is the extent to which morbidity and comorbidity in BED is attributable to the presence of obesity. Hudson et al. [
4] have shown that risk for components of metabolic syndrome are elevated in BED independent of the effects of BMI. We further explored the extent to which observed psychiatric comorbidity in BED was attributable to obesity.
Register-based studies that include longitudinal follow-up and that capture all clinical contacts in a nation’s health care system allow for intensive characterization of a disorder. By linking rich longitudinal data from the Swedish national eating disorders quality registers and national patient registers, we explored clinical characteristics at evaluation, diagnostic flux, and psychiatric comorbidity in all individuals presenting for evaluation for BED across the country.
Discussion
Using the extensive register data available for the entire country of Sweden, we were afforded the unique opportunity to explore clinical course and psychiatric comorbidity of all individuals seeking treatment for BED, yielding a sample of 850 individuals. Briefly, BED shows considerable diagnostic fluidity with other eating disorders, carries high psychiatric comorbidity burden, and is associated with elevated suicide risk.
The reported age range at first diagnosis was broad (from 14–72), encouraging vigilance for BED across the age spectrum. Less than 5 % of the sample was male, which could reflect a lower prevalence of BED in males in Sweden (contrast the projected lifetime prevalence of 1.41 % projected for men in the US [
3]), or, that males with BED are less likely to seek assistance for their disordered eating than females [
32]. A recent evidence based review reported that across 48 reviewed randomized controlled trials, the percentage of females participating ranged from 67 to 100 % [
33] suggesting that relative to epidemiological prevalence estimates, females do indeed appear to be more likely to seek treatment. Therefore, treatment seeking (not just in Sweden) may be less common in males with BED, underscoring the importance of psychoeducation that reaches males as well as clinicians and other positioned to detect BED in males.
Fortuitously, 67 % of evaluated individuals returned for a follow-up visit allowing us to capture detailed clinical course and symptom progression, although, by the nature of the quality registers, we are unable to tether those changes to specific interventions. One quarter of the returnees received a second diagnosis of BED at follow-up and one quarter received no eating disorder diagnosis reflecting remission of symptoms. Sixteen percent of those returning for follow-up received a different eating disorder diagnosis—most commonly BN or EDNOS and occasionally AN. This novel observation is of clinical importance in encouraging clinicians to be mindful of signs of transitioning to other eating disorder presentations. Those who transitioned to EDNOS may have been on a recovery trajectory and the change in diagnosis could reflect a reduction in total symptomatology (e.g., binge frequency) reflecting a subthreshold presentation. Such changes will presumably be rectified with the inclusion of remission specifiers in DSM-5. Transitions to BN and AN do not reflect a remission trajectory, but rather reflect the emergence of recurrent inappropriate compensatory behaviors (in BN) or significant weight loss and/or restricting behaviors (in AN), both concerning clinical developments. Of note, flux was bidirectional, with patients also transitioning to BED from AN, BN, or EDNOS. Given that our design only included at maximum two visits, and also reported on individuals who only made one visit to a clinic, although we can describe considerable diagnostic flux, we are not well positioned to comment authoritatively on the long-term stability of BED as a diagnostic entity.
We confirm and extend previous reports of comorbidity in BED [
12‐
20]. The greatest lifetime comorbidity burden was with the other eating disorders with the large odds ratio dramatically highlighting the diagnostic fluidity of these illnesses and encouraging further investigation into the mechanisms underlying diagnostic instability. Genetic, neurobiological, and environmental influences should be explored and risk algorithms developed to aid in the prediction of clinical course.
Of the other psychiatric disorders, the risk of MDD and bipolar disorder were most elevated, followed by anxiety disorders and PTSD. Only risk for schizophrenia, schizoaffective disorder, and autism spectrum disorders were not elevated. Elevated risk for depression was not associated with comorbid obesity in the “pure BED” (i.e., no other eating disorder diagnoses) sample. Thus it should not be assumed that normal weight individuals with BED are at any less risk for depression. These observations underscore the substantial comorbidity burden carried by individuals with BED and highlight the clinical importance of assessing and addressing the full array of presenting problems facing these patients. Treatment research should address the extent to which interventions require tailoring based on comorbidity profile.
Of particular clinical concern is the elevated risk for suicide in individuals with BED. Even when controlling for the presence of MDD, those with BED were at significantly increased risk for suicide attempts. Our sensitivity analyses revealed that suicide risk was elevated regardless of the presence of comorbid obesity. These results extend previous reports [
34] and underscore the seriousness of BED as a psychiatric illness.
Our observations of elevated suicide risk in BED are important clinically; however, alternative designs are necessary to explicate the mechanism of the association [
35]. One possibility is that a factor such as impulsivity could represent a genetically influenced intermediate phenotype behind both BED [
36,
37] and some, but not all forms of suicidal behavior [
38‐
40]. Alternatively or complementarily, environmental risk factors (e.g., sexual and physical abuse, bullying or being bullied) could represent important shared risk factors for BED [
41] and suicidal behavior [
42,
43].
The comparison of individuals with pure BED versus those with lifetime histories of other eating disorders revealed similarly high comorbidity burden, with the exception of drug use and suicide. This observation suggests that suicide risk in BED may be particularly concerning in those who have transitioned to or from other eating disorders during the trajectory of their illness.
Our study has notable limitations. First, we only included individuals who presented for an evaluation and received a diagnosis of BED. Untreated individuals in the community may present with different patterns of clinical course, comorbidity, and suicide risk. Second, we did not screen for eating disorders in controls. As such, our analyses represent conservative comparisons and the actual associations may be stronger. This approach was taken to guard against inflated estimates caused by artificially “clean” control samples. Third, as personality disorders tend to be less reliably included in patient registers, we were unable to explore comorbidity between BED and personality disorders. Fourth, assuming variable and incalculable lags between symptom onset and diagnosis across disorders, reports on the temporality of diagnoses could reflect either true order of onset of the disorders or differential latency to treatment seeking. Fifth, 33 % of patients did not return for follow-up. With the exception of alcohol use disorders, their comorbidity profiles were no more severe. Those who did not return may not have been recommended for further treatment at a specialty service or could have opted out of treatment.
Balancing these limitations are several methodological strengths including the large sample size, standardized evaluation by trained clinicians and the longitudinal follow up afforded by the Swedish eating disorders quality registers, national healthcare records, and detailed register-based patient psychiatric and other medical history.