The systematic review followed the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) standard (Page et al., 2021). Additionally, the systematic review protocol was structured in accordance with the PRISMA-P guidelines. A literature search was carried out by the chief librarian of Aalborg University Hospital, Denmark on 17 October 2023, in the databases PubMed, PsycInfo, and Embase. No time restrictions were imposed, and all identified relevant articles found in the databases were included in the systematic review. The search phases combined controlled vocabularies (e.g., Mesh, thesaurus, and Emtree) and free-search terms relevant to the Population (individuals with an ED), Comparison (presence or absence of ED), and Outcome (prevalence of NTE) components of the PICO framework. Besides these keywords, other reviews were excluded by filters in PubMed and Embase in the search. A comprehensive list of the search terms, synonyms, and controlled vocabularies, used in the different databases can be found in Supplementary Table 1. The extraction files were then uploaded into Rayyan:¹ an online reference management software. After removing duplicates by Rayyan’s duplicate-identification strategies the articles were screened for relevance, language, and availability. All searches, starting from the initial screening of the title and abstract to the final decision on inclusion through full-text reading, were conducted independently by two reviewers (K.M.J., K.F.N.). To reduce the risk of errors made in the selection of studies, the reviewers were blinded to each other. Discrepancies between the two reviewers were resolved through consensus. If persistent disagreements arose, a senior researcher (G.K.T.) was consulted as a third party to reach a final decision. This approach ensured a rigorous screening process and minimized the potential for bias in study selection.

All studies collected from the literature search were screened based on the predetermined inclusion and exclusion criteria. To be included in the systematic review, studies had to be peer-reviewed and have a quantitative research design. Additionally, the prevalence had to be extractable or calculable from the articles’ data. Participants in the studies were required to have or have had a clinical diagnosis of ED according to the International Classification of Diseases (ICD-10) or the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV and DSM-5), to be included. Studies on comorbid disorders and medical conditions were not excluded from the systematic review. This decision stemmed from the well-established overlap between EDs, and diverse medical and psychiatric conditions documented in existing literature (Hudson et al., 2007). By incorporating individuals with comorbidity, a more nuanced depiction of the complex presentation of EDs was achieved. Additionally, the studies had to include NTE among participants with a current or previous ED diagnosis. Inclusion criteria specified that studies had to have a sample size of 10 or more individuals diagnosed with an ED.

Studies that only reported an overall category of trauma were excluded, as NTE could not be differentiated from interpersonal traumatic events. Studies were also excluded if they had other research designs than empirical study designs. Moreover, studies in languages other than Danish, English, Swedish, or Norwegian were excluded.

In the data collection process, the following information was extracted from the included studies: (a) author names, (b) year of publication, (c) country, (e) population frame (e.g., in-patients, out-patients, etc.), (f) the different ED subcategories and assessment tools, (g) trauma type and measurement, (h) main findings regarding prevalence among individuals with and without ED. This form was used to summarize and synthesize the overall findings in relation to the objectives of the systematic review. The primary outcome of interest of the included studies was prevalence. If data were raw or numerical, the percentage was calculated. Only the relevant control groups for the objectives of this study were stated in the form. Throughout the article, we employed the term ‘control group’. This designation encompassed both reference groups as well as matched control groups without ED. We contacted the corresponding authors by mail if the prevalence of NTE could not be extracted from the articles but was expected to be in the possession of the authors. Articles were excluded if no response was received.

All included studies were assessed using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Studies Reporting Prevalence Data, to critically evaluate the methodological quality of the included studies (Munn et al., 2020). This tool was used to evaluate different items in the studies, with responses categorized as “Yes,” “No,” “Unclear,” or “Not applicable.” JBI checklist was applied as it offered tailored evaluation criteria for prevalence studies, ensuring a thorough assessment of methodological quality in the included studies in a systematic and standardized way. With inspiration from Rasmussen et al. (2023), the total score for the JBI checklist for prevalence studies ranged from 1 to 9. The overall evaluation of the quality of the studies clarified the impact of the studies.

A total of 7,390 articles were retrieved in the initial search. From these, 2,624 came from PubMed, 3,160 from Embase, and 1,606 from PsycInfo. Five articles were identified under the subheading PsycTest that were excluded in the initial search process. Only references under the subheadings PsycInfo and PsycArticles were included. The removal of duplicates resulted in 2471 articles, and another 16 duplicates were subsequently removed. The following screening by title and abstract resulted in 64 articles, of which seven articles could not be retrieved. The agreement rate between authors in the independent screening was 97.9%. 57 articles were assessed according to the predetermined exclusion and inclusion criteria, of which 16 were included. Figure 1 depicts a detailed description of the study selection in a flow diagram.

Supplementary Table 2 shows data from the 16 included research articles. The number of participants in the ED group ranged from n = 21 (Thornley et al., 2016) to n = 4,524 (Backholm et al., 2013). For the control groups included in the studies, the sample sizes ranged from n = 245 (Lejonclou et al., 2014) to n = 35,709 (Convertino et al., 2022). Thirty-seven-point 5 % of the studies involved only adult participants (>18 years; Reyes-Rodríguez et al., 2011; Backholm et al., 2013; Tagay et al., 2014; Thornley et al., 2016; Longo et al., 2019; Brewerton et al., 2020) while 6.25% included child/adolescent participants (<18 years; Groth et al., 2020). When including Kjaersdam Telléus et al. (2021), who also investigated adolescence, but in the age range from 15 to 19.5, the percentage increased to 12.5%. Half of the studies either included both age groups or did not clarify the age range (Tagay et al., 2010; Degortes et al., 2014; Lejonclou et al., 2014; Vieira et al., 2018; White et al., 2018; Longo et al., 2020; Lie et al., 2021; Convertino et al., 2022). Results indicated a varied prevalence of NTE in the child/adolescent and adult age groups, with no consistent trend observed between the two groups.

The included studies predominantly originated from the United States and Europe, with minimal to no representation from other regions such as Asia or Africa. Of the included studies six and four studies investigated only out- and inpatients, respectively. When comparing the two groups, inpatients exhibited a higher prevalence of illness or injury ranging from 20 to 59% (Thornley et al., 2016; Longo et al., 2019; Brewerton et al., 2020; Longo et al., 2020), while outpatients showed a range of 6.5–19.4%, when Lejonclou et al. (2014) study was excluded from the outpatient range due to significantly higher results (e.g., 60% had been in the hospital and 70% had experienced having a family member in the hospital; Tagay et al., 2010; Degortes et al., 2014; Vieira et al., 2018; White et al., 2018; Convertino et al., 2022).

In general, different measurements were used to measure eating pathology and exposure to traumatic events, including both structured clinical interviews and self-report questionnaires. Supplementary Table 3 provides a list of all the measurements used in the studies. Overall, 31% of the studies used the Life Events Checklist (LEC) to assess the prevalence of traumatic events in patients with an ED (Backholm et al., 2013; Thornley et al., 2016; Longo et al., 2019; Brewerton et al., 2020; Longo et al., 2020).

Despite using the same measurement instrument, LEC, no consistency was observed in the prevalence of NTE subtypes across the five studies, but rather a significant variation ranging from 0–59%. For instance, in Backholm et al. (2013), only 1.57% of patients with ED had experienced life-threatening illness or injury, whereas Thornley et al. (2016) reported a much higher prevalence of 43% for this variable.

Two of the 16 studies presented the prevalence of an overall category of NTE in patients suffering from ED. In these, the prevalence was, respectively, 75.2% (Convertino et al., 2022) and 14.3% (Longo et al., 2020). Across the 16 studies, NTE could be grouped under five broad subheadings: accidents, injury, illness, war, and natural disaster. Overall, the prevalence of the different NTEs in patients with an ED across all included studies ranged from 0 to 72%. The results were ambiguous, but generally, the prevalence was between 0 and 30%. The studies that showed the highest prevalence were Brewerton et al. (2020), Lejonclou et al. (2014), Tagay et al. (2014) and Thornley et al. (2016). Generally, illness and injury showed the highest prevalence compared to the other subtypes of NTE (Backholm et al., 2013; Degortes et al., 2014; Lejonclou et al., 2014; Tagay et al., 2014; Thornley et al., 2016; Longo et al., 2019; Convertino et al., 2022). When differentiating these, illness had the highest prevalence of the two variables (Convertino et al., 2022). In contrast to these findings, Tagay et al. (2010) and Brewerton et al. (2020) found that accidents were the subtype with the highest prevalence. However, most of the studies showed that accidents had a lower prevalence than illness and injury, but a higher prevalence than natural disasters and war (Reyes-Rodríguez et al., 2011; White et al., 2018).

The studies that incorporated a control group were conducted by Convertino et al. (2022), Lejonclou et al. (2014), Lie et al. (2021), and Thornley et al. (2016). The included control groups used as a reference consisted of populations pooled from internet samples (Thornley et al., 2016; Lie et al., 2021), national surveys regarding alcohol and related conditions (Convertino et al., 2022), and recruitment from Secondary Schools and High Schools (Lejonclou et al., 2014). The clinical group was predominantly females, whereas gender was mixed in the control groups. Most of the studies had a control group that was representative of the clinical group in terms of age. However, Thornley et al. (2016) exhibited a greater difference.

Only one study reported the prevalence of the overall category of NTE in comparison to a control group (Convertino et al., 2022). In this study, the overall prevalence of NTE in patients suffering from ED was 75.2%, whereas the prevalence in the group without ED was 62.2%. Across studies with a control group, the prevalence of the subtypes of NTE in the ED group ranged from 0 to 70%, in comparison to 0.18–72% in the control group without ED. Two studies demonstrated a generally higher prevalence of accidents in the control group compared to the ED group (Thornley et al., 2016; Lie et al., 2021). The findings in Lejonclou et al. (2014) presented varied results regarding the prevalence of multiple items related to accidents.

In all studies the prevalence of war exposure was more than twice as high in the control group, which indicated that the prevalence was considerably lower in the ED sample (Lejonclou et al., 2014; Thornley et al., 2016; Convertino et al., 2022). Likewise, ED groups demonstrated lower levels of exposure to natural disasters compared to control groups (Lejonclou et al., 2014; Thornley et al., 2016). In contrast, Convertino et al. (2022) reported that the rates of natural disaster did not differ between those with and without ED. The two variables injury and illness were sometimes combined in the measurement instruments for exposure to NTE (Thornley et al., 2016). Other studies separated the categories or only reported one of them (Lie et al., 2021; Convertino et al., 2022). The trauma measurement used in Lejonclou et al. (2014) included hospitalization as an item, which made it unclear whether it belonged to the category of injury or illness. Generally, the ED groups demonstrated a considerably higher prevalence of illness compared to the control groups (Lejonclou et al., 2014; Thornley et al., 2016; Lie et al., 2021; Convertino et al., 2022), although Lejonclou et al. (2014) showed no significant difference in the prevalence of illness when measuring hospitalization in a family member or close relative. In terms of injury there was not a significant difference between the ED- and control groups (Convertino et al., 2022). For a general overview of the main results regarding the minimum and maximum range for the prevalence of NTE, refer to Figure 2. Here, the difference in prevalence between the ED and control groups is also displayed within this interval range.

Most studies distinguished between the subtypes of ED: AN (Tagay et al., 2010; Reyes-Rodríguez et al., 2011; Lejonclou et al., 2014; Tagay et al., 2014; White et al., 2018; Longo et al., 2019; Groth et al., 2020; Longo et al., 2020; Kjaersdam Telléus et al., 2021; Lie et al., 2021; Convertino et al., 2022), BN (Tagay et al., 2010; Reyes-Rodríguez et al., 2011; Degortes et al., 2014; Lejonclou et al., 2014; Tagay et al., 2014; White et al., 2018; Groth et al., 2020; Kjaersdam Telléus et al., 2021; Lie et al., 2021; Convertino et al., 2022), BED (Degortes et al., 2014; Groth et al., 2020; Lie et al., 2021; Convertino et al., 2022) and OSFED (White et al., 2018; Groth et al., 2020; Kjaersdam Telléus et al., 2021). In terms of NTE reported in the single studies, there was not a large difference between AN, BN, BED, and OSFED. When the participants had the opportunity to report multiple lifetime ED diagnoses, it was shown that these participants (e.g., AN/BN and AN/BN/BED) had a higher prevalence than the participants with a single lifetime ED diagnosis (e.g., AN, BN or BED; Reyes-Rodríguez et al., 2011; Lie et al., 2021). The finding that the participants reporting multiple lifetime ED diagnoses, exhibits a higher prevalence of NTE compared to those with a single lifetime ED diagnosis, aligns with the findings of Rossi et al. (2024), which show an increased prevalence of the so-called diagnostic crossover in the maltreated eco-phenotype. The presence of diagnostic crossovers, such as individuals experiencing multiple forms of eating pathology over their lifetime, underscores the dynamic nature of EDs and emphasizes the need for a more nuanced understanding of ED that considers the interplay between trauma exposure, symptom expression, and diagnostic trajectories. Moreover, these findings underscore the importance of adopting a transdiagnostic approach clinically in assessment and treatment that accounts for the diverse pathways to ED development and maintenance.

It was demonstrated in Tagay et al. (2014), that participants with BN had experienced accidents and illness/injury to a higher extent compared to AN. However, other studies found that AN had the highest or same prevalence in relation to accidents and illness/injury (Tagay et al., 2010; Kjaersdam Telléus et al., 2021). When differentiating illness and injury it was shown in Convertino et al. (2022) that the highest prevalence of injury was seen in individuals with BED, in comparison to AN and BN. In contrast, BN was the ED subtype with the highest prevalence of illness. Only one study reported on war in relation to ED subtypes (Convertino et al., 2022). It was demonstrated that individuals with BN had been exposed to more war-related traumatic events, than AN and BED. However, the difference was minimal, and the percentage was very low. Findings showed that there was no consistent difference in the prevalence of natural disasters between AN- and BN subgroups in the included studies. ED subgroup comparisons showed that individuals with BED experienced fewer events related to natural disaster.

Regarding the difference between AN-R and AN-BP, it was demonstrated by Lie et al. (2021), Longo et al. (2019), and Kjaersdam Telléus et al. (2021), that individuals with AN-R had a lower prevalence of accident and illness than AN-BP. In terms of natural disasters, Longo et al. (2019) found that AN-R had the highest prevalence, in comparison to AN-BP. Similarly, Reyes-Rodríguez et al. (2011) showed a higher prevalence in individuals with AN-R compared to AN-P and AN-B in terms of accidents. However, this was not the case regarding natural disaster, where AN-P had the highest prevalence.

Seven studies either exclusively reported the prevalence of NTE, when the events occurred before the onset of the ED diagnosis or provided information on the number of individuals who reported experiencing at least one NTE prior to the onset of the ED (Reyes-Rodríguez et al., 2011; Degortes et al., 2014; Tagay et al., 2014; Vieira et al., 2018; Longo et al., 2019, 2020; Lie et al., 2021). Between 58.3–81.1% of the ED sample reported the traumatic event happened before the ED onset. Noteworthy is that Tagay et al. (2014) and Reyes-Rodríguez et al. (2011), exclusively provided information on the timing of NTE in the subset of individuals with both ED and PTSD, omitting those who did not have PTSD. Additionally, two studies only investigated the prevalence of NTE occurring prior to the onset of ED (Degortes et al., 2014; Vieira et al., 2018).

Thirteen of the 16 articles were assessed to have a quality that scored between four and six points out of nine, on the Critical Appraisal Checklist for Studies Reporting Prevalence Data in JBI. These articles were considered to have medium quality. Two studies achieved a score of eight, indicating a high quality. On the contrary, one of the included studies had a score of two, which is classified as low quality. Overall, the two high-quality studies showed a relatively low prevalence compared to the other studies, ranging from 0.15–14%.

The review was carried out following PRISMA guidelines for systematic reviews, which is in line with best practice. A strength of this systematic review was that the results were based on a large sample size of diagnosed ED subjects and controls in total. However, the sample size of ED in most of the included studies was small, which influenced the generalizability of their findings and consequently affected the quality of the systematic review. The systematic review sought to reduce bias using multiple reviewers in the screening- and extraction process, who were blinded to each other. This resulted in moderate to high interrater reliability. The review’s broad focus on ED diagnosis provided a comprehensive view of NTE in patients with ED, a topic not systematically reviewed previously. However, this breadth became a limitation due to high heterogeneity in study designs and measurement tools for assessing NTE and ED prevalence. This hindered comparison and prevented a reliable meta-analysis. Despite recognizing the potential benefits of a meta-analysis, the inherent data heterogeneity posed a significant barrier to its validity and reliability. Thus, we deliberately opted against a meta-analytic approach. Additionally, graphical representation of prevalence faced challenges due to inconsistent reporting standards and lack of confidence intervals in many studies, complicating aggregation, and visualization. The decision to exclude papers published in languages other than Danish, English, Swedish, or Norwegian was based on practical considerations. The researchers lacked proficiency in other languages, hindering their ability to assess the quality and relevance of papers accurately. While this approach may miss useful studies, it prioritizes the reliability, comprehensibility, and accessibility of the included literature.

The results underline that patients with ED have not been exposed to NTE more than individuals without ED. NTE should not be viewed as a specific and high-risk factor in the development of ED. Thus, there should not be attached importance to NTE in the clinical work with ED. In comparison, interpersonal traumatic events are evaluated to have a higher impact on the occurrence of ED, which is why events such as sexual assaults and emotional violence should be assigned a higher priority in the assessment and treatment of ED.