The systematic literature search resulted in 7211 hits. The screening of titles and abstracts identified 540 potentially eligible articles. Finally, 69 studies were included in this review. Two records were identified through a hand search of the references of the relevant articles. A 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. The characteristics of the reviewed studies are detailed in Online Resource 2.
Health-related quality of life burden
Data on the HRQoL of AN, BN, BED, and relatives/caregivers of ED patients were reported in 41 studies: 18 for AN [13–30], 17 for BN [13–15, 17, 20–25, 27, 29–34], 18 for BED [17, 21, 22, 24, 35–48], and 5 for the relatives/caregivers of ED patients [49–53]. Seventeen types of HRQoL instruments were applied in the included studies. The Medical Outcome Study Short Forms (i.e., Short Form 36 [SF-36] and Short Form 12 [SF-12]) were the most commonly used (n = 19) HRQoL questionnaires and were administered alone [16, 17, 19, 21, 24–27, 32, 33, 36, 38, 40, 42, 43] or in combination with other HRQoL questionnaires [13, 22, 23, 37] (Table 1). The reported HRQoL data are summarized in Online Resource 3.
Patients with AN, BN, and BED were shown to have significantly lower HRQoL than the general population [18, 34, 37, 39, 40, 43]. Many of the included studies investigated the differences between the HRQoL impact of AN, BN, and BED/EDNOS. One of these studies evaluated patients with AN or BN and compared them to healthy subjects . Both ED groups showed significantly more impairment than controls in the health domains of the Nottingham Health Profile (NHP), and the AN patients had significantly reduced mobility compared with the BN patients and the healthy controls. Three studies compared patients with AN, BN, or BED. One of them showed no differences in HRQoL, but demonstrated a significant relationship between HRQoL and the severity of eating symptomatology . Another study found a more heterogeneous picture using the SF-36 . Patients with BN or BED had lower scores than the non-ED subjects on the SF-36 Mental Component Summary (MCS) (mean [SD]: 43.76 [9.35], 40.83 [9.00] and 49.83 [16.47], respectively), but no differences were found on the SF-36 Physical Component Summary (PCS) between the ED groups and the non-ED group (mean [SD]: 43.40 [9.02] for AN, 47.20 [9.47] for BN, 46.31 [9.12] for BED, and 48.08 [16.67] for non-ED subjects). Regarding the various domains, the BN patients had lower scores on the Role Emotional, Social Functioning, Mental Health, Vitality, and General Health scales than the non-ED subjects, and the BED patients had lower scores on the Role Emotional, Mental Health, and Vitality scales than the non-ED subjects . Mond et al.  presented similarly heterogeneous findings: significantly lower MCS scores in the BN and purging AN groups compared to the restrictive AN group (the MCS mean [SD] scores were: 27.60 [9.44] vs. 26.96 [7.44] vs. 38.38 [11.05], respectively), and significantly lower World Health Organization Brief Quality of Life Assessment (WHOQOL-BREF) Social Relationship Scale (QoLS) scores in the BED and purging AN groups compared with the restrictive AN group (mean [SD]: 2.20 [0.98] vs. 2.58 [1.06] vs. 3.58 [1.01], respectively). Rie et al.  could not show any HRQoL difference on comparing AN, BN, and EDNOS groups to each other, but all of the ED patients had significantly poorer HRQoL than the normal reference group. Focusing on EDNOS , no differences were detected between the EDNOS and BN groups on the SF-36 General Health and Healthy Status scales, but AN was associated with lower scores on the Social Functioning, Vitality, and Physical Functioning scales than EDNOS. Bamford et al.  found lower Psychological and Physical/Cognitive Scale scores (using the Eating Disorder Quality of Life [EDQoL] instrument) in AN than in BN and EDNOS. Latner et al.  found that SF-36 scores were worse in those with subjective bulimic episodes, food avoidance, laxative abuse, and self-induced vomiting, and that PCS scores were worse in those with subjective bulimic episodes and food avoidance. However, these authors did not find differences in HRQoL between AN, BN, BED, and EDNOS.
Eight studies investigated the separate and joint effects of BED and obesity (defined as BMI ≥30 kg/m2) on HRQoL [36, 37, 40, 41, 43–46]. The results of Masheb et al.  showed that BED patients with obesity had significantly lower scores on several SF-36 subscales (i.e., Physical Functioning, Bodily Pain, General Health, Vitality, Social Functioning, and PCS) than those without obesity (the PCS mean [SD] scores in patients with BMI ≥30 kg/m2 and BMI <30 kg/m2 were: 45.3 [9.6] vs. 53.6 [9.4], respectively, P = 0.001). Among obese individuals, all but two studies [41, 45] concluded that BED patients had significantly reduced HRQoL compared with individuals without BED [36, 37, 46], even in a population with extreme obesity (defined as BMI ≥40 kg/m2)  (Online Resource 3). Perez et al.  found that obesity was more strongly related to physical HRQoL variables, but BED was more predictive of the mental health and social functioning HRQoL variables. In contrast, Ricca et al.  did not find any significant differences in HRQoL between those obese patients without BED, those with threshold, and those with subthreshold BED (defined by a minimum average binge eating [BE] frequency of once a month for a minimum duration of 6 consecutive months). In the study conducted by Kolotkin et al. , BED was shown to be associated with more impaired HRQoL in obese individuals (the IWQOL mean (SD) total scores in the BED and non-BED patients were: 51.5 (21.9) vs. 65.3 (19.8), respectively, P < 0.001), but BED did not prove to be an independent factor of weight-related quality of life after controlling for BMI, demographic variables, and psychological variables.
The HRQoL burden of EDs is determined not only by the deficit in HRQoL of the patients themselves, but also by the potential deficit of close relatives who live with the patient or caregivers and take care of an ED patient. A study found that 80 % of the siblings of adolescent ED patients reported that their HRQoL was negatively affected by the onset of their siblings’ ED . De La Rie et al.  showed that the caregivers of ED patients had worse HRQoL than normal controls (mean SF-36 scale scores: 56.2 [Vitality] to 90.4 [Physical Functioning] vs. 68.6 [Vitality] to 84.0 [Social Functioning], respectively). The caregivers’ perceived burden improved significantly over the first year of follow-up, but no further improvement was observed with longer follow-up . Many factors were associated with higher caregiver burden, such as higher anxiety and depression, purgative behaviors, lower patient HRQoL , and low-level education .
Twenty studies reported data on healthcare utilization: 14 for AN [3, 54–66], 12 for BN [2, 3, 34, 54, 58–62, 64–67], and 8 for BED [2, 3, 39, 59, 65, 66, 68, 69]. The healthcare utilization data extracted from the included studies are provided in Online Resource 4.
AN (78 %), BN (88 %), and BED (73 %) were associated with increased health service use (for any treatment, lifetime) compared with individuals without an eating disorder (44 %) . In the study conducted by Striegel-Moore et al. , the number of 12-month total health service days (i.e., inpatient care, outpatient care, and emergency care) of BED patients (11.8–21.4) was higher than in the healthy comparison group (3.4–8.4), but it was similar to the utilization found in other major psychiatric disorders (6.9–18.4). The health service use in AN was reported to be equal to or higher than in BN or BED [3, 54, 58–62, 64, 65]. The difference in healthcare utilization between the AN and BN patients was largest in the case of hospitalization. The length of hospital stay was found to be much longer for patients with AN (15.0–52.7 days) than for those with BN (9.0–45.7 days) [54, 58, 60, 61, 64].
Only one study evaluated the effect of obesity (defined as BMI ≥30 kg/m2) on health service use in patients with BED. In the study by Striegel-Moore et al. , the number of 12-month total health service days (i.e., inpatient care, outpatient care, and emergency care) was found to be higher for obese BED women (mean [SD] total service days: 21.4 [28.1]) than for nonobese BED women (mean [SD] total service days: 11.8 [21.8]); however, the association between obesity and health service use was not significant.
In the assessed studies, patients with AN, BN, or BED rarely received treatment for their ED, but received more frequent treatment for comorbid psychiatric symptoms and/or weight loss [2, 3, 59, 66]. In the study conducted by Mond et al.  on AN, BN, and BED, 22, 14, and 23 % of the patients, respectively, were treated by a mental health professional, specifically for their ED in their lifetime.
Data on healthcare costs were reported in 17 studies: 9 for AN [58, 60, 61, 70–75], 11 for BN [31, 34, 58, 60, 61, 73, 75–79], and 2 for BED [39, 68]. In the reviewed studies, the annual healthcare costs for AN, BN, and BED ranged from €2993  to €55,270 , €888  to €18,823 , and €1762  to €2902 , respectively (Table 2). Detailed information on the reported healthcare costs is presented in Online Resource 5.
The healthcare costs of AN, BN, and BED were not compared in any of the included studies. However, 5 studies contrasted AN with BN [58, 60, 61, 73, 75] (Online Resource 2). In most of these studies, AN was found to be associated with considerably higher annual healthcare costs than BN (€5952 vs. €1460 , €18,587 vs. €11,406 , and €5445 vs. €3386 ). However, Mitchel et al.  estimated higher costs for BN (€6279) than AN (€3221). Striegel-Moore et al.  assessed the cost of EDs (i.e., AN, BN, and EDNOS) and compared these costs with the costs of other mental disorders (i.e., obsessive–compulsive disorder and schizophrenia). AN (€6590 for female, €2993 for male) and BN (€3229 for female, €4235 for male) patients had higher costs than obsessive–compulsive disorder patients (€2104 for female, €1965 for male) . The healthcare costs for men with AN and BN were slightly less than the costs for men with schizophrenia (€5552); however, the costs were higher for women with schizophrenia compared to BN (€5259). The healthcare costs for patients with BN were comparable to the costs associated with the treatment of EDNOS (female €3229, male €4235 vs. female €3496, male €2360 , and €6279 vs. €6514 ). In one of the two studies that examined the costs of BED, the annual healthcare cost of overweight/obese BED women (€1,762) was reported to be 36.5 % higher than that of the age- and sex-matched national average . Grenon et al.  found no association with BMI (after controlling for age) and the total healthcare costs in BED. However, this finding is in contrast to the study by Dickerson et al.  that showed a significant association with higher BMI and higher mental health medication and total medication costs in women with BED or recurrent BE.
In AN, premature death in the young patient population leads to significant indirect costs to society (63 % of the overall costs), mainly due to the loss of production . Symptom-related costs for food are considerable expense factors for patients with BN. Crow et al.  found that costs associated with BE and purging in BN accounted for 32.7 % of the total food costs; the mean total annual food cost per BN patient was €4735. No study was identified for BED-related food costs.