The database search resulted in 7,211 hits. Screening of titles and abstracts identified 540 potentially eligible articles. Two additional records were identified through hand search of the references of relevant articles. After the review of relevant full text articles, 49 studies were included in this systematic review. The flow diagram of the systematic literature search, based on the preferred reporting items for systematic reviews and meta-analyses (PRISMA) template , is presented in Fig. 1.
Most studies (n = 40) were cross-sectional [2–4, 19–55], 7 longitudinal [10–12, 15–18], and 2 were database analyses [13, 14]. Included studies originated from 23 countries, with over half (n = 25) conducted in the US. The sample sizes of the included studies ranged from 31  to 77,807 , with a mean ages from 13.0  to 49.8  years. Many studies only included females (n = 18) but in those including both sexes, females were over-represented. Diagnosis of BED was made using DSM-IV criteria in 44 [2–4, 10, 12–15, 17–20, 22–30, 32–41, 43–55], using the proposed DSM-5 criteria in 3 [11, 16, 31], and using both DSM-IV and the proposed DSM-5 criteria in 2 studies [21, 42]. The general characteristics of the reviewed studies are presented in Online Resource 2.
Included studies (n = 49) fulfilled between 26.5  and 84.4 %  of the STROBE criteria. Details of the quality assessment are provided in Online Resource 3.
Only one study reported incidence rate for BED . A sample of 496 adolescent females over an 8-year period found the incidence of BED (DSM-5) to be 343 per 100,000 person-years .
Point prevalence rate for BED was reported in 22 studies [12, 14, 20–39], 12-month prevalence rate in 4 studies [2–4, 13] and lifetime prevalence rate in 11 studies [2–4, 13, 16, 17, 21, 36, 40–42]. Detailed information on the reported prevalence rates are presented in Online Resource 2. Point prevalence rates, not limited to population-based estimates varied between 0.1 % (in a sample of pregnant women, 6–12 months prior pregnancy)  and 34.1 % (in a sample with type 2 diabetes)  using DSM-IV criteria, and 0.6 % (in a sample of female high school and university students)  to 1.7 % (in a sample of first degree relatives of BED patients)  using the proposed DSM-5 criteria. The 12-month prevalence rate of BED, not limited to population-based estimates ranged between 0.1 % (in a large community sample of people over 18 years of age)  and 1.1 % (in a sample of Latino)  using DSM-IV criteria. Lifetime prevalence for BED, not limited to population-based estimates was 0.17 % (in a sample of adult female twins)  to 8.8 % (in a sample of outpatients with bipolar disorder)  using DSM-IV criteria and 0.2 % (in a sample of adult female twins)  to 3.6 % (in a sample of first degree relatives of BED patients)  using the proposed DSM-5 criteria.
In three studies the lifetime prevalence of BED (DSM-IV) was 1.5–6 times higher in women than in men [3, 4, 21]. Hudson et al. . found no difference in the proportion of women and men regarding the DSM-IV and the proposed DMS-5 lifetime prevalence of BED (Online Resource 2). Sex difference was also observed in point and 12-month prevalence rates (Online Resource 2). The majority of BED cases occurred first in people’s lives between the ages of 12.4 and 24.7 years [3, 4, 16, 43]; however, the prevalence of BED continued to rise until 40 years old . From the reviewed articles information on ethnic or racial differences in prevalence of BED was limited and inconsistent. Azarbard et al.  found no significant differences in the prevalence of BED among Hispanic, African American and White women (Online Resource 2). However, Perez et al.  reported higher prevalence for BED among Hispanic (2.3 %) and Black individuals (1.6 %) compared to White individuals (1.0 %).
Several studies examined the association of BED with physical and psychiatric comorbidities. In the study Kessler et al.  the odds ratio (OR) for BED was 2.9 for individuals with diabetes, 2.2 for hypertension, 1.6 for stroke and 1.3 for heart disease (population-based estimates). Point prevalence of BED (DSM-IV) in overweight/obese individuals was reported to be 5.9 % in a sample with serious mental illness , 13.4 % in a sample waiting for bariatric surgery  and 23.9 % in a sample seeking weight loss treatment . The OR for lifetime BED was 0.7 for BMI < 18.5, 1.0 for BMI 18.5–24.9, 1.3 for BMI 25–29.9, 3.1 for BMI 30–34.9, 3.0 for BMI 35–39.9 and 6.6 for BMI > 40 (population-based estimates) . Lifetime prevalence of BED was found to be elevated in patients with obsessive–compulsive disorder and bipolar I or II disorders [36, 40]. Compared to a non-eating disorder group, the mean scores in Beck anxiety inventory was higher for the BED group (23 vs. 13, p < 0.001) and for the Beck depression inventory (23 vs. 15, p < 0.001) . Several studies reported that BED occurred in a significant number of women during pregnancy (1.8–7 %) and in the post-partum period (2.7–3.1 %) [12, 27, 28, 32].
Mortality for BED was not reported in any of the included studies and only 3 studies described data on suicide attempt and/or ideation in BED patients [4, 22, 44]. When comparing asymptomatic and BED individuals, among a school-based sample of youth the ORs for suicidal ideation and attempt were 2.6 and 3.1, respectively . In the studies conducted by Carano et al.  and Swanson et al. , 27.5 % (in adults) to 34.4 % (in adolescents) of individuals with BED had suicidal ideations and the frequency of suicidal attempts was 12.5 % (in adults) to 15.1 % (in adolescents).
Health-related quality of life burden
Sixteen studies reported data on the HRQoL burden of BED (Table 1) [10, 11, 14, 15, 18, 19, 34, 43, 45–52].Various questionnaires were used to measure HRQoL in the studies. All but one study used validated HRQoL questionnaires (Table 1). Perez et al.  used a self-developed instrument. General HRQoL measures included the Extended Satisfaction With Life Scale (ESWLS) , EuroQol Five Dimensional Questionnaire (EQ-5D-3L) , Short Form-36 (SF-36) [11, 19, 45–48, 50], Short Form-12 (SF-12)  or World Health Organization Brief Quality of Life Assessment (WHOQOL-BREF) [15, 49]. The SF-36 (n = 7) was the most commonly used HRQoL questionnaire. An obesity-specific measure was used in 5 studies (Impact of Weight on Quality of Life Lite (IWQOL-LITE) [18, 45, 51, 52] and Obesity Related Well-Being (ORWELL) ).
HRQoL of patients with BED was significantly lower than in control subjects (Table 1) [14, 34, 45–47, 49, 51, 52]. BED was associated with marked impairment compared to general population norms in both the physical component summary (PCS) and the mental component summary (MCS) scores of SF-36 (Masheb et al.  PCSBED: 47.3, MCSBED: 39.7; Padierna et al. . PCSBED: 36.5, MCSBED: 32.0 vs. PCSpopulation norm = 50.0, MCSpopulation norm = 50.0). In the study conducted by Grenon et al.  the EQ-5D-3L mean score of overweight/obese women with BED (0.77) was significantly lower than for a US community sample of women with a similar mean age (0.89).
Where studies evaluated HRQoL for the different eating disorders, there were no significant differences among AN, BN and BED (Table 1) [47, 49, 50]; but, decreased physical HRQoL appeared to be most evident in patients with BED. In the study by Mond et al.  physical health measured by SF-12 was poorer in BED (PCS: 40.2) than in AN (PCSrestricting type: 45.4, PCSpurging type: 46.8) and BN (PCS: 49.3); however, most BED patients were obese (BMI ≥ 30) and had a higher mean age in that patient group [mean age (standard deviation); AN restricting type: 19.31 (4.22); AN purging type: 25.53 (9.77); BN: 23.48 (6.25); BED: 34.33 (7.37)]. These findings were consistent with those of Padierna et al. , who reported more impaired physical health (SF-36) in BED patients (PCS: 36.5); nevertheless the differences compared to AN (PCSrestricting type: 44.0, PCSpurging type: 43.5) and BN (PCS: 43.0) were not statistical significant (Padierna et al.  presented their results only in figure form, SF-36 scores were extracted from the figure).
Among subgroups of BED, obese BED patients had significantly worse HRQoL than non-obese BED patients (Table 1) [14, 34, 49]. In addition, BED was associated with more impairment in HRQoL than obesity without BED (IWQOL-LITE BED: 74.0 vs. No BED: 61.2; higher scores indicating greater impairment) . Obesity status more strongly affected the physical dimension of HRQoL, whereas BED status affected mental health and social functioning HRQoL . Among BED patients, obese BED subjects (PCS: 45.3) had significantly lower scores in PCS scores on the SF-36 compared to non-obese (PCS: 53.6) . Depressive symptoms were also reported to be significantly associated with lower HRQoL in patients with BED (measured with EQ-5D-3L, after controlling for age and BMI) .
Healthcare utilization in BED was assessed only in 8 studies [2–4, 13, 43, 53–55]. Healthcare costs were reported only in two studies [43, 53]. Reported healthcare utilization and healthcare costs data are presented in Table 2 and Table 3.
Striegel-Moore et al.  analyzed the healthcare utilization in women with BED using emergency room visits, outpatient physician visits for medical care, outpatient psychotherapy visits and days spent in the hospital over the last 12 months. Compared to healthy individuals (mean total service days: 3.4–8.4), BED was associated with increased total health service use (mean total service days: 11.8–21.4), but resource utilization in BED was similar to other major psychiatric conditions (mean total service days: 6.9–18.4) (Table 2). There was no significant difference in resource utilization between BED group and non-eating Axis I psychiatric disorders group.
In the included studies, only a minority of BED patients received treatment specifically for their eating disorder. Swanson et al.  found that the lifetime service use for treatment for eating or weight problems among BED patients was only 11.9 %. In the study conducted by Kessler et al.  38.3 % of patients with a lifetime diagnosis of BED had received treatment specifically for an eating disorder. Mond et al.  reported that in their study sample, 58.1 % of BED patients had received treatment for an eating problem, but 83.9 % received treatment for general mental health and 87.1 % for weight loss. Only 12.9 % of BED patients were treated by a mental health professional specifically for their eating disorder.
Marques et al.  examined mental health service utilization for differences among ethnic groups. They found mental health service use was higher for non-Latino Whites (78.9 %) than for Latinos (54.1 %), Asians (54.5 %) or African Americans (71.1 %).
The reported annual direct healthcare costs per BED patient ranged from a low of $2,372  to a high of $3,731  (Table 3). In one Canadian study , the annual healthcare cost of overweight/obese women with BED was 36.5 % higher than the estimated health expenditure per capita for women in the same age group. Another US study  compared healthcare utilization between women with BED (DSM-IV) and recurrent binge eating (individuals with at least 1 day per week with an objective bulimic episode over a 3-month period, with no periods of binge free time greater than 2 weeks, and did not meet DSM-IV criteria for BN or BED), and evaluated the effect of the number of binge eating days on healthcare costs. BED and recurrent binge eating groups did not differ significantly in total annual healthcare costs adjusted for age, BMI and self-reported depression.