Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2025 Apr 19;33(5):1005–1016. doi: 10.1002/erv.3194

The Prevalence of Excessive Exercise in Eating Disorders: A Systematic Review and Meta‐Analysis

Connor Campbell 1,✉, Xander Greig 1, Jessica Griffiths 1, Devon Hashman 1, Thomas Sottile 1, Masanori Isobe 2, Anees Bahji 3, Gina Dimitropoulos 4, Dan Devoe 1
PMCID: PMC12319126  PMID: 40252210

ABSTRACT

Objective

Individuals with eating disorders (EDs) often present with maladaptive behaviours such as excessive exercise (EE). The consequences of EE include physical injuries, increased risk of anxiety and depression, and impaired social functioning. No systematic reviews have been conducted on the prevalence of EE in EDs. This study aimed to assess the prevalence of EE in EDs and by ED type.

Method

An electronic database search of the peer‐reviewed literature was conducted from inception to October 2024. Review eligibility was restricted to research studies reporting prevalence data for EE in individuals diagnosed with EDs.

Results

Fifty‐six studies met the inclusion criteria (n = 21,518; mean age: 22.34 years). The current prevalence of EE in all EDs was 48%. Current prevalence was highest in AN (48%), followed by BN (45%), OSFED (38%), and BED (11%). The lifetime prevalence of EE in all EDs was 63%. Lifetime prevalence was highest in AN (72%), followed by BN (57%) and OSFED (21%).

Conclusions

Nearly half of individuals with an ED engage in EE. High heterogeneity across the included studies likely influenced the prevalence found in this study. Data suggest clinical screening and longitudinal monitoring of EE in those with EDs. Future research into early intervention and treatment for EE in those with EDs is recommended.

Trial Registration

PROSPERO: CRD42023464148; Open Science Framework: https://doi.org/10.17605/OSF.IO/MYVXW

Keywords: eating disorders, excessive exercise, mental health, meta‐analysis

Summary.

  • The meta‐analysis revealed that 48% of patients with eating disorders engage in excessive exercise upon presentation at treatment centres, clinics, and hospitals. The number climbs to 63% across the lifetime.

  • Patients with anorexia nervosa showed the highest prevalence of excessive exercise compared to other eating disorder types.

  • A unified definition of excessive exercise is desperately needed to future research into its association with eating disorders.

1. Introduction

Eating disorders (EDs) are severe and life‐threatening psychiatric conditions that involve over‐regulating one's eating and controlling weight through restrictive or purging behaviours (Treasure et al. 2020). These disturbances have a notable impact on physical health, such as electrolyte imbalances, gastrointestinal and renal problems, and cardiovascular issues (Himmerich et al. 2021; Robinson et al. 2016). For example, a recent studies have found that 32% of patients with EDs suffered from electrolyte imbalances (Solmi et al. 2024), more than 50% of patients suffered from abdominal pain, gastric distension, and constipation (Sato and Fukudo 2015), and up to 85% of patients with AN specifically suffered from some form of cardiovascular abnormalities (Friars et al. 2023). Furthermore, EDs are associated with high risk of mortality (Smink et al. 2012) and an estimated 3.3 million healthy life years were taken away by EDs and ED‐related symptomatology (van Hoeken and Hoek 2020). An exhaustive 2019 systematic review reported on the various prevalence estimates from 2000 to 2018, reporting an average lifetime prevalence for women at 8.4% and 2.2% for men (Galmiche et al. 2019). EDs are associated with psychiatric comorbidities, including anxiety disorders, mood disorders, post‐traumatic stress disorder (PTSD), and substance abuse disorders (Devoe et al. 2022; Keski‐Rahkonen and Mustelin 2016; Momen et al. 2022). Individuals with EDs also incur considerable personal, familial, and societal costs as they often require long‐term treatment (Meneguzzo et al. 2021; Samnaliev et al. 2014).

Excessive exercise (EE) is defined as exercise that is driven, rigid, avoids aversive outcomes, interferes with important activities, occurs at problematic times or in inappropriate settings, continues despite injury, is an activity that one feels compelled to engage in, and is accompanied by feelings of guilt if postponed (Colledge et al. 2020; Dittmer et al. 2018; White and Halliwell 2010). This comprehensive definition was constructed for the purposes of this paper and utilised multiple sources in order to capture the most available data on EE. Early research into EE in the context of ED focused on the intensity, duration, and frequency of exercise (Scharmer et al. 2020). However, as noted by White and Halliwell (2010), evidence has shown that exercise frequency is not a good predictor of EE, as it is not significantly associated with poorer mental health, lower self‐esteem, or problematic eating behaviour. As such, recent research has since moved away from such descriptors and instead examined the psychological relationship an individual has with exercise (Scharmer et al. 2020), such as the definition proposed by Dittmer et al. (2018) whose main criterion contains the psychological elements of rigidity and avoidance of aversive emotions. Dittmer et al. (2018) notes that the component of rigidity relates to how individuals keep to strict and fixed exercise routines; routines that, when missed, can cause distress. This highlights both the behavioural and cognitive aspects that exist in those with EE. Furthermore, more modern research has included elements such as dependence and compulsion in the examination of EE (Berczik et al. 2012). In the context of EE, dependence is associated with a pleasure‐seeking craving to exercise and avoidance of exercise withdrawal, whereas compulsion is linked to exercising with the goal of remove anxiety or negative emotion (Di Lodovico et al. 2019; Scharmer et al. 2020). However, despite the definition presented in this paper, no unified definition of EE has been accepted by the eating disorder field. Moreover, the term EE was selected for this paper because it is the most used term in the literature. Other terms seen in the literature include compulsive exercise (CE), driven exercise (DE) and over‐exercise (OE).

A comprehensive study by M. Meyer et al. (2021) examined the occurrence of mental disorders in individuals with EE. In a sample of 156 individuals, 32 were identified as having EE (M. Meyer et al. 2021). Of the 32 with EE, 56.3% were diagnosed with a depressive disorder, 46.9% with a personality disorder, 31.3% with obsessive‐compulsive and related disorders, 28.1% with an anxiety disorder, and 15.6% with an ED (M. Meyer et al. 2021). The study also found that as the severity of EE increased, so too did the number of diagnosed mental disorders (M. Meyer et al. 2021). Despite the occurrence of EE in multiple categories of mental disorders, it has typically been studied in the context of EDs (M. Meyer et al. 2021). Regardless of the disorder, it may be that EE is a maladaptive coping mechanism to combat psychological distress (Marques et al. 2019). Specifically in the context of EDs, EE poses greater psychopathological impairments such as OCD symptomatology and anxiety, and anhedonia (Scharmer et al. 2020). Moreover, these issues are present in adolescent, young adult, and older adult populations (as seen in Fietz et al. [2014] and Lampe et al. [2022]), as well as in both female and male populations (as seen in M. Meyer et al. [2021]).

The medical consequences of EE include joint and muscle injuries, impaired physiological functioning, negative social outcomes, problematically low weight, and an increased risk of developing an ED (Bereda 2023). Those engaging in EE are also more prone to mental health issues such as anxiety and depression, which is often the result of over‐training, as an individual with EE will typically show signs of irritation, disorientation, and exhaustion, which can ultimately lead to them being unhappy about their overexertion (Bereda 2023). Moreover, physical activity typically increases hunger (Bereda 2023). However, excessive levels of exercise can negatively affect satiety, which may result in severe weight loss issues (Bereda 2023). Individuals with EE also tend to suffer socially and have poor social relationships (Bereda 2023; Dittmer et al. 2018; Lichtenstein et al. 2017). Exercise becomes time‐consuming, which can jeopardise an individual's romantic partnerships, family relationships, and friendships (Bereda 2023; Dittmer et al. 2018; Lichtenstein et al. 2017).

Martenstyn et al. (2022) suggested that EE develops in ED for two reasons: (1) to regulate negative emotions and (2) to burn calories to reduce weight and change body shape. There is a strong association between EE and EDs, with some research suggesting that those with EDs are nearly four times as likely to suffer from EE than those without EDs (Berczik et al. 2012; Trott et al. 2021). Research has also established a clear inverse relationship, in which individuals affected by EE often display significant concerns about body weight and shape and engagement in dietary control, all of which are prominent symptoms of EDs (Berczik et al. 2012).

There is very little data to provide a clear indication of the prevalence of EE in those with EDs (C. Meyer et al. 2008). Possible explanations for the minimal empirical focus are the lack of consensus on an EE definition and bundling EE with other compensatory behaviours such as purging or fasting (Harris et al. 2020; C. Meyer et al. 2008). Moreover, there is also some uncertainty regarding EE as a symptom in patients with EDs (Mond and Gorrell 2021). For example, despite evidence that EE is found in the majority of individuals with anorexia nervosa (AN), neither the DSM nor ICD have included EE as diagnostic criteria (Mond and Gorrell 2021; Peñas‐Lledó et al. 2002). These discrepancies and the apparent lack of consensus in the field of EDs have led to a lack of understanding about the impact EE has on individuals with EDs. The prevalence studies that exist in the literature have shown considerably varied results, ranging anywhere from 30% to 80% in adults (Dalle Grave et al. 2008; Dittmer et al. 2020; Shroff et al. 2006), and upward of 85% in adolescents (Fietz et al. 2014). Moreover, EE rates have been shown to vary by ED type (Shroff et al. 2006).

The vast majority of research on EE in ED has focused on the treatment of ED and its symptoms (including EE) or the treatment of EE in those with ED. In both instances, the proportion of patients with ED with EE has been given; however, these prevalence estimates are limited to single studies. In a 2014 systematic review, Fietz et al. did report on more widespread prevalence estimates, but these results were limited to adolescent populations. As such, the primary gap in the research on EE in ED is the lack of an encompassing meta‐synthesis on the prevalence estimates of EE in ED. Moreover, an analysis of the prevalence of EE in all the different ED types and ED subtypes is needed to gain an appropriate understanding of the specific impact EE has on the different ED types and subtypes. The main goal of this study is to provide a meta‐analytical review of the peer‐reviewed literature regarding the prevalence of EE in EDs and the differences in the prevalence of EE in ED types and ED subtypes. A comprehensive synthesis of EE prevalence estimates will not only contribute to the current state of knowledge but also inform clinicians and exercise professionals about the present dangers of EE.

2. Method

2.1. Protocol and Guidelines

This systematic review and meta‐analysis was registered a priori with PROSPERO (CRD42023464148) and made available on Open Science Framework, and followed both PRISMA (preferred reporting for systematic reviews and meta‐analyses) and MOOSE (meta‐analysis of observational studies in epidemiology) guidelines.

2.2. Search Strategy

An evidence‐based electronic search was conducted in accordance with PRESS guidelines for systematic reviews. A comprehensive search of the literature was conducted in the following online databases: CINAHL, Embase, ERIC, MEDLINE, APA PsycINFO and SCOPUS from their inception to October 31st, 2024, with no geographical or language restrictions. The keywords included two concepts: (1) eating disorders ([EDs]; anorexia nervosa [AN], bulimia nervosa [BN], binge eating disorder [BED], other specified feeding or eating disorder [OSFED], and avoidant/restrictive food intake disorder [ARFID]) and (2) excessive exercise (EE) terms. Database searches and an exhaustive list of key terms are provided in Supporting Information S1.

2.3. Selection Criteria

Peer‐reviewed studies were selected by two reviewers (C.C. and one of reviewers X.G., J.G., D.H., or T.S.) for inclusion in this review if they met the following criteria: (1) participants with EDs (AN, BN, BED, ARFID or OSFED), (2) participants with EDs with EE or any of its alternative terms (compulsive exercise, exercise dependence, maladaptive exercise), (3) study designs that were either treatment based, qualitative, survey, cross‐sectional, or longitudinal, (4) specifically for the meta‐analysis component: individuals with EDs and data pertaining to EE, (5) articles written in any language, (6) no exclusions due to age and (7) no exclusions due to geographical location. In addition, the following exclusion criteria were applied: (1) study designs classified as case reports, review articles, opinion pieces, books or book chapters, and editorials without original data, (2) articles with insufficient data for the meta‐analysis (e.g., studies with data on patients with EDs and not EE, or vice versa, or studies where data did not indicate the prevalence of EE in those with EDs), (3) articles where ED diagnosis was self‐reported or where ED diagnosis could not be confirmed and (4) articles that did appropriately or correctly define EE (i.e., studies that did not utilise elements of the comprehensive definition used in this paper), or where the symptom of EE was self‐reported by participants.

For title and abstract screening, each article required two reviewers to determine that article's inclusion or exclusion. Five blinded reviewers (C.C., X.G., J.G., D.H., and T.S.) independently performed title and abstract screening using the online Covidence systematic review software (Covidence). Reviewer C.C. reviewed every title and abstract, and one of the other four reviewers made the second determination. For full‐text screening, each article also required two reviewers to determine inclusion or exclusion. Three blinded reviewers (C.C., X.G. and J.G.) independently performed full‐text screening also using Covidence. Reviewer C.C. reviewed every full text, and either X.G. or J.G. provided the second determination. Disagreements regarding inclusion or exclusion for both title and abstract screening and full‐text screening were discussed in a consensus meeting, and C.C. made the final decision.

2.4. Data Extraction

Data extraction was completed independently by three reviewers in duplicate (C.C., X.G. and J.G.). Reviewer C.C. extracted data from every full text, while reviewers X.G. and J.G. each extracted data from half of the full text.

For the systematic review, the following study characteristics were extracted: (1) author, (2) year of publication, (3) country, (4) study type, (5) how EE is referred to, (6) scale used to measure exercise, (7) ED type or ED subtype, (8) demographic information (age [mean ± SD], sex, gender, gender identity, sexual orientation, socioeconomic status, race, ethnicity, and nationality) and (9) percent female (number of females). For the meta‐analysis, the following data was extracted: (1) author, (2) year of publication, (3) how EE is referred to, (4) the scale used to measure exercise, (5) ED type or ED subtype, (6) the ratio or number representing participants with EE for the numerator, (7) the number of participants with EDs for the denominator (e.g., AN, BN) or ED subtype (e.g., AN‐R, AN‐BP), age (mean ± SD), (8) percent female and (9) whether the relevant data represented current (e.g., upon presentation at a clinic, treatment centre, etc.) or lifetime prevalence.

2.5. Risk‐of‐Bias Assessment

The studies included were independently evaluated for quality by three reviewers (C.C., X.G. and J.G.). All studies included in this review were examined using the RoB‐PrevMH, a brief tool for assessing risk‐of‐bias in studies that measure the prevalence of mental health disorders (Tonia et al. 2023). The RoB‐PrevMH consists of four questions (Tonia et al. 2023). One is simply a yes‐or‐no question regarding whether the target population was clearly defined (Tonia et al. 2023). The next two address selection bias and the fourth question examines information bias (Tonia et al. 2023). For each study, each of the three questions is given a judgement rating of either high, low, or unclear (Tonia et al. 2023). The RoB‐PrevMH was designed to meet the adaptability needs of systematic reviews and prevalence studies (Tonia et al. 2023). Most risk‐of‐bias assessment tools are designed for specific types of studies. However, the questions utilised in the RoB‐PrevMH can be applied to any kind of study (Tonia et al. 2023).

2.6. Data Synthesis and Analysis

A detailed description of the studies is provided in tabular form and the results are described in terms of percentages. Because of the expected heterogeneity of the studies included, DerSimonian and Laird (DerSimonian and Laird 1986) random‐effects meta‐analyses were performed to estimate pooled‐effect sizes and 95% confidence intervals (CIs) for each study. Outcomes were then organised for each analysis.

All meta‐analyses in this study employed the Freeman‐Tukey double arcsine transformation, which calculates the weighted pooled estimate using a variance stabilising transformation and then executes a back‐transformation on the pooled estimate. This approach is preferred in scenarios of zero count data, as it prevents such studies from being excluded from the analyses that would create a bias in prevalence estimates. The exact CI method was utilised for this study as it is more conservative. Additionally, as is common practice, all meta‐analyses were weighted by sample size.

Two or more independent studies with similar observations were required to be included in the aggregated meta‐analysis. Where applicable, scales were inverted to match the direction of effect of the majority of scales. To avoid double‐counting, when two or more studies reported similar observations in the same sample, the data from the largest sample was used. When data permitted, subgroup meta‐analyses were performed based on geography and stratified by age where applicable.

Statistical heterogeneity was examined using the I 2 statistic, with I 2 ≥ 50% deemed moderate and I 2 ≥ 75% deemed high. An I 2 value is produced only when there are two or more studies in a meta‐analysis. All analyses were performed in R v.4.3.2 (R Core Team 2023). The data obtained using R calculated prevalence by dividing the number of people with EE by the number of individuals with EDs. Moreover, in the subgroup analyses, prevalence was calculated by dividing the number of people with EE by the number of individuals with a specific ED type (e.g., AN or BN). A series of forest plots were produced to visually represent the proportion of those with EE in individuals with EDs and types of EDs.

3. Results

3.1. Search Yield

Electronic database searches identified 6638 records. Fifteen additional records were included from a previous exploratory search conducted by the authors. After duplicates were removed, a total of 4410 abstracts and titles were screened. Upon resolution of inconsistencies between reviewers, 160 studies were retrieved and reviewed in full text. Full‐text articles were excluded from the final analysis if they fell into one or more of the following three categories: (1) different outcomes (e.g., no prevalence data), (2) different patient population (disordered eating populations) or (3) wrong study design (e.g., review articles or case studies). Additionally, a grey literature search was conducted through OpenGrey, New York Academy of Medicine's Grey Literature Report, and TRIP PRO databases; however, no further sources were identified. Overall, 56 studies met the inclusion criteria and were included in the meta‐analysis, see Figure 1.

FIGURE 1.

FIGURE 1

Open in a new tab

PRISMA screening flow diagram.

3.2. Study Characteristics

All studies included in this systematic review are described in detail in Supporting Information S1. Studies were published between 1987 and 2022. Most studies were conducted in either North America (n = 25) or Europe (n = 23), followed by Australia (n = 5), Asia (n = 2) and Argentina (n = 1). Thirty‐five studies recruited individuals from a hospital setting or a specific eating disorder programme, six were recruited from outpatient clinics, six utilised medical records, five used databases, and four were recruited from university settings.

3.3. Participant Characteristics

A total of 21,518 individuals were identified as having an ED, with sample sizes ranging from 12 to 9074 participants with EDs in individual studies. The mean age of those with EDs was 22.34 years (range 11.3–45), and the percentage of females was 95%. Participant SES data is described in detail in Supporting Information S1.

3.4. Risk‐of‐Bias

All studies included in this systematic review were evaluated using the RoB‐PrevMH instrument (Tonia et al. 2023). All included studies clearly defined the target population based on a yes/no question in the RoB‐PrevMH (Tonia et al. 2023). Twenty‐seven of the studies earned a low‐risk label for all three bias assessment questions, while 24 studies earned low‐risk labels for two of the bias assessment questions and one label for unclear. Five studies had one question labelled high risk, with the other two being labelled low risk. There was also one study with two questions labelled unclear and the other question being low risk. A single study also had one question with low risk, one unclear, and one labelled high risk. All 56 studies earned low‐risk labels for the representativeness of the sample frame. In comparison, 41 studies were given low‐risk ratings for representativeness of the responder (14 unclear and one high risk), and 40 studies were labelled low risk for measurement of the condition (12 unclear and four high risk). A full breakdown of the risk of bias can be found in Supporting Information S1.

3.5. Prevalence of EE in EDs

In the studies that looked at the current prevalence of EE, random effects pooled estimates demonstrated a 48% prevalence in those with EDs (CIs: 0.44–0.51; I 2 = 94%; k = 50, N = 19,810). In studies that looked at the lifetime prevalence of EE, random effects pooled estimates showed a 63% prevalence in individuals with EDs (CIs: 0.52–0.73; I 2 = 93%; k = 9, N = 1670). The difference between the current and lifetime prevalence of EE in those with EDs was significant (χ 1 2 = 7.08, df = 1, p < 0.01). All prevalence data for EE in individuals with EDs are detailed in Figure 2.

FIGURE 2.

FIGURE 2

Open in a new tab

Current prevalence and lifetime prevalence of EE in all EDs.

3.6. Prevalence of EE by ED Subtypes

For individuals with AN (k = 24, N = 4577), the current prevalence of EE was 48% (CIs: 0.41–0.55; I 2 = 94%). In those with BN (k = 15, N = 3.980), the current prevalence of EE was 45% (CIs: 0.38–0.51; I 2 = 86%). In OSFED individuals (k = 7, N = 5910), the current prevalence of EE was 38% (CIs: 0.28–0.48; I 2 = 96%). For those with BED (k = 2, N = 670), the current prevalence was 11% (CIs: 0.08–0.14; I 2 = 0%). The current prevalence of EE in the different ED types is detailed in Figure 3. Additionally, the subtypes of AN were analysed. The current prevalence of EE in those AN‐R was 40% (CIs: 0.16–0.68; I 2 = 98%; k = 6, N = 940). The current prevalence of EE in individuals with AN‐BP was 38% (CIs: 0.32–0.44; I 2 = 39%; k = 4, N = 503). See Supporting Information S1 for more details. Additionally, no studies included reported on the current prevalence of EE in those with ARFID.

FIGURE 3.

FIGURE 3

Open in a new tab

The current prevalence of EE in ED types.

For the lifetime prevalence of EE in those with AN (k = 6, N = 843), the prevalence was 72% (CIs: 0.54–0.87; I 2 = 93%). In those with BN, the lifetime prevalence of EE was 57% (CIs: 0.47–0.67; I 2 = 0%; k = 2, N = 98). The lifetime prevalence of EE in AN and BN is detailed in Supporting Information S1. Of note is that there was only one study that reported lifetime prevalence for OSFED, and therefore, it was not analysed. Moreover, no studies reported lifetime prevalence for BED or ARFID.

3.7. Prevalence of EE in Specific Subgroup Populations

Three studies reported on EE in males with ED, two of which reported lifetime prevalence. Random effects pooled estimates showed a 61% prevalence across those two studies (CIs: 0.50–0.71; I 2 = 52%; k = 2, N = 183). Moreover, some studies in this analysis reported EE specifically in adolescents with ED. The current prevalence of EE in adolescents with ED was 57% (CIs: 0.47–0.66; I 2 = 97%; k = 9, N = 5125). The prevalence of EE in adolescents with AN was 54% (CIs: 0.26–0.80; I 2 = 97%; k = 4, N = 1336). Forest plots of these three analyses are shown in Supporting Information S1.

3.8. Sensitivity Analyses

In addition to looking at the prevalence of EE in those with EDs and ED subtypes, analysis of the prevalence of EE in EDs, based on the different scales used to measure EE, were performed. This analysis was run only for current prevalence as there was insufficient data for lifetime prevalence. Moreover, scales that were utilised in less than three studies were combined under one subgroup called ‘other’. For results, see Supporting Information S1.

Furthermore, to account for Berkson's bias, another analysis was conducted comparing the prevalence of EE in EDs in studies that looked at inpatients versus outpatients or database studies. For results, see Supporting Information S1.

3.9. Publication Bias

Egger's test was utilised to test for publication bias. For studies looking at the current prevalence of EE in those with EDs, there was a nonsignificant result (very slight positive skew), indicating no publication bias, see Table 1. Studies looking at the current prevalence of AN in individuals with EDs yielded a significant result (slightly positive skew), which indicates a possible publication bias. Studies of the current prevalence of BN in those with EDs showed a nonsignificant result (slight negative skew), indicating no publication bias. Details of Egger's tests for studies looking at the current prevalence of AN and BN in those with EDs are detailed in Table 2. Funnel plots for these three tests are shown in Supporting Information S1.

TABLE 1.

Egger's test for current prevalence of EE in those with EDs.

Intercept Confidence intervals t p
0.67 −0.86 to 2.19 0.86 0.40
Open in a new tab

Abbreviations: ED, eating disorder; EE, excessive exercise.

TABLE 2.

Egger's test for current prevalence of EE in those with AN and BN.

ED type Intercept Confidence intervals t p
AN 2.81 0.54–5.07 2.43 0.02
BN −1.24 −3.01 to 0.52 −1.38 0.19
Open in a new tab

Abbreviations: AN, anorexia nervosa; BN, bulimia nervosa; ED, eating disorder; EE, excessive exercise.

There were only nine studies that looked at the lifetime prevalence of EE in those with EDs. Therefore, no testing for funnel plot asymmetry was conducted, as suggested by the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al. 2024). Abiding by this suggested rule‐of‐thumb, any analysis in this study with nine or fewer studies was not accompanied by Egger's test for publication bias, nor was a funnel plot produced.

4. Discussion

4.1. Summary

This systematic review and meta‐analysis has contributed to the research literature knowledge base by being the first to examine the pooled prevalence of EE in those with EDs. It was estimated that nearly half (48%) of patients with ED are engaging in EE upon presentation to a specialised ED clinic or when seeking treatment for their ED elsewhere (e.g., hospitals). This number climbs to 63% for patients with any ED across the lifetime. The highest rates of EE, in both current prevalence and lifetime prevalence estimates, were seen in AN patients, which is consistent with previous literature on the subject (Dalle Grave et al. 2008; Dittmer et al. 2018). Current prevalence of AN subtypes (AN‐R and AN‐BP) were lower than overall AN. This was due to some studies reporting on AN, some reporting AN and subtypes, and others reporting only AN‐R and AN‐BP. The quality of the literature assessed was quite strong; however, there were a few studies that contained a high risk of bias for certain criteria. Overall, this study demonstrates that EE is a common occurrence in those with EDs, and clinicians should be made aware of the frequency with which EE appears in patients with ED.

4.2. Clinical Implications

The results of this meta‐analysis are particularly alarming, especially when compared to the prevalence of EE in non‐ED populations. Research on exercise addiction (EA), which uses interchangeable definitions with EE or CE, highlights that general populations and sport populations have shown rates of 14.2% in endurance disciplines (e.g., running, cycling, swimming), 10.4% in sport settings, 8.2% in fitness settings, 6.4% in bodybuilding and CrossFit disciplines, and 3.0% in the general population (Di Lodovico et al. 2019). Another study, looking specifically at elite athletes, demonstrated a 7.6% risk of EA (Lichtenstein et al. 2021). The prevalence of EA in non‐ED populations highlights just how staggering the rates of EE in those with EDs are. One study included in the systematic review by Di Lodovico et al. (2019) reported the prevalence of EA in a sample of 1285 triathletes at 20%, which is still less than half than the prevalence of 48% of EE in those with EDs found in this study. Important to note is that, much like research on EE in those with EDs, studies examining the prevalence EE in non‐ED contexts utilise a variety of definitions (Berczik et al. 2012). For example, Di Lodovico et al. (2019) uses the term EA, which the authors describe as a synthesis of dependence and compulsion. Moreover, Lichtenstein et al. (2021) also utilises the term EA and uses measures that examine how exercise affects participants both psychologically and physically, specifically highlighting elements such as how individuals feel when exercise is missed or postponed and how exercise has negatively impacted their social lives. Given that the definitions used for EE both within and outside of ED contexts share such an emphasis on the psychological relationship an individual has with exercise, it is stunning how much higher the prevalence of EE is in those with EDs. This perhaps suggests that it is the escalation of appetitive emotions and distorted perceptions of one's body that may lead to higher prevalence of EE in those with EDs.

There has been minimal research into the treatment of EE (Lichtenstein et al. 2017), with most intervention studies published in 2016 or later (Martenstyn et al. 2022). Early treatment of EDs proposed the complete elimination of exercise; however, recent research has shown that structured exercise can have positive effects on patients with ED, such as decreasing depressive symptoms and pathological eating (Martenstyn et al. 2022). Current treatments for EE often incorporate a gradual introduction of exercise in those with EDs (Martenstyn et al. 2022). EE is commonly addressed with cognitive‐behavioural therapy ([CBT]; Martenstyn et al. 2022; Mathisen et al. 2018). Behavioural monitoring and psychoeducation are also utilised (Dittmer et al. 2020; Martenstyn et al. 2022). In a recent systematic review and meta‐analysis of the treatment of EE (termed CE in their study) in EDs, Martenstyn et al. (2022) reported that the treatment methods utilised had moderate‐to‐large reductions in EE when analysed absent of control groups. However, those reductions decreased when compared to control groups (Martenstyn et al. 2022). Important to highlight is that most treatments employed a multi‐component approach (e.g., CBT plus structured exercise) was used (Martenstyn et al. 2022). This multi‐component approach appears to be the best method for treating EE in ED, although the optimal blend of treatment approaches has yet to be determined (Martenstyn et al. 2022). The results of this study highlight the necessity of selecting appropriate and useful treatment methods for patients with ED engaging in EE. Moreover, given the general lack of efficacy of CBT for those with AN (Argas 2019) and the high prevalence of EE in AN found in this study, future research should examine whether CBT would an appropriate treatment for patients with AN who engage in EE.

4.3. Excessive Exercise Terminology and Definitions

EE was by far the most used term in the literature, with CE being the second‐most used term. Collective research has typically utilised one of these two terms, and while their varying definitions tend to feature distinct elements, the overlap in their definitions is notable. Traditionally, the definition of EE has included elements such as inappropriate and harmful frequency, duration, and intensity of exercise. However, numerous studies included in this meta‐analysis incorporated features such as ‘driven to exercise’ and ‘accompanying guilt when postponed’ in their definitions of EE; these features are typically associated with CE. Similar overlaps in defining features were found when studies used the term CE. With such an overlap of features found in the literature, and considering the findings of this study, it appears that prevalence data will be similar whether the term EE or CE is used. If that is indeed the case, perhaps EE and CE (and other terms such as DE and OE) should be merged under one banner. Whether that banner be named EE or CE is less important than developing a unified definition. Efforts to develop such a unified definition already exist (e.g., Colledge et al. 2020; Dittmer et al. 2018). A unified definition would both aid in the validity of future research on EE in EDs as well as provide the field of EDs a common diagnostic and assessment template.

4.4. Strengths and Limitations

There are several strengths to this study. First, to the author's knowledge, this is the first systematic review of EE in those with EDs and the first meta‐analysis. Second, the quality of the majority of studies included in the meta‐analysis was strong. Third, this study adhered to PRISMA, MOOSE, and PRESS guidelines and was comprehensive and methodologically rigorous.

However, certain limitations should be considered in the appraisal of the evidence presented by this study. Most notable is the lack of a unified definition of EE. While this study utilised a comprehensive definition constructed from multiple sources, the results presented would be more accurate and robust if a unified definition to measure EE in those with EDs was employed by the various studies analysing this association. The varying definitions of EE utilised by the various included studies is one reason why such high heterogeneity was seen across studies for nearly all estimates. Other reasons for high heterogeneity include the wide range of instruments used to measure EE, the range of study years, and the different terms used to refer to EE (e.g., EE or CE). Moreover, the high heterogeneity seen across the included studies likely influenced the overall prevalence found by this study. Such high heterogeneity in this context could reduce the accuracy of the pooled prevalence estimates as well as lead to wider confidence intervals. As such, the results of this study should be considered in this light. Another limitation is the lack of studies assessing the lifetime prevalence of EE in those with EDs. Consequently, only two studies looked at lifetime EE in BN. This may be due to the fact that EE has been studied in the context of AN for much longer than other EDs (Scharmer et al. 2020). Moreover, there were only two studies that looked at current prevalence in BED patients. Although the nature of BED would appear unsuitable for EE research, due to the fact that BED patients do not typically engage in purging or restrictive behaviours (i.e., EE), it should still be considered a limitation.

As this is a meta‐analysis, a limitation of the method is that individual characteristics were not explored. There was also a limited representation of geographical regions, which limits this study's ability to estimate the prevalence of EE in those with EDs globally. A sensitivity analysis was conducted looking at the prevalence of EE in EDs by geographical region, but there was no notable difference between the overall prevalence of all studies combined versus when they were separate by region. There was limited data about patients with BED and none about ARFID, which means that overall prevalence estimates of EE are not truly representative of all EDs but rather a combination of AN, BN and OSFED. Moreover, only 19 of the 56 studies reported on EE in males with EDs. Three of these 19 studies reported only on males, and males in the other 16 studies represented anywhere from three to 22% of the sample size. The lack of data on EE in diagnosed with BED or ARFID, as well as EE in male ED populations, affects the generalisability of this study. More research is required into these specific populations to increase the generalisability of EE in EDs. Further, as with all meta‐analyses, this study was limited by the quality and quantity of existing studies. As such, the results reflect only what is available in the existing literature.

4.5. Future Research

Future research is needed to explore the prevalence of EE in different subgroup populations of patients with ED, particularly males, children and adolescents, and older adults. Such research could shed light on any differences in the prevalence of EE between adolescent and adult populations. While this study did examine EE in males with ED, the analysis only included two studies reporting lifetime prevalence. As such, more research is needed to determine the prevalence of EE in males with EDs. Some of the studies included in this analysis also reported on the prevalence of EE in adolescents with ED. However, the number of studies was limited, and more research on this population is needed. Future directions might also involve exploring early intervention and treatment options for ED individuals engaging in EE. One possible avenue of exploration for early intervention would be further investigation into utilising the gradual introduction and monitoring of exercise (Martenstyn et al. 2022). Future research would also benefit from examining how many patients with EDs experienced EE before the onset of their ED. Moreover, research into whether EE persists during ED treatment and recovery would also shed light on potential connections between EE and EDs. While research into EE has been increasing, particularly in the context of EDs, there has been minimal research into what types of exercise individuals with EDs are engaging in excessively. When studies do mention or identify the type of exercise, it is almost always running or similar endurance‐related exercises. However, as shown by Di Lodovico et al. (2019), EE certainly occurs outside of endurance disciplines. Future research into EE should consider identifying and reporting on the types of exercise being utilised by patients with ED. Moreover, the development of a measure/instrument to appropriately determine what kind of EE patients with ED are engaging is recommended.

5. Conclusions

This is the first meta‐analysis on the prevalence of EE in those with EDs. The analyses found high prevalence, with nearly half of patients with ED engaging in EE and over 60 percent engaging in EE at some point in their lifetime. Although causal conclusions cannot be drawn, considering the dangers of EE detailed in this paper, clinical screening and longitudinal monitoring of EE in those with EDs is strongly recommended.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information S1

ERV-33-1005-s001.docx (219.9KB, docx)

Handling Editor: Sarah Maguire

Funding: The authors received no specific funding for this work.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

  1. Argas, W. S. 2019. “Cognitive Behaviour Therapy for the Eating Disorders.” Psychiatric Clinics of North America 42, no. 2: 169–179. 10.1016/j.psyc.2019.01.001. [DOI] [PubMed] [Google Scholar]
  2. Berczik, K. , Szabó A., Griffiths M. D., et al. 2012. “Exercise Addiction: Symptoms, Diagnosis, Epidemiology, and Etiology.” Substance Use & Misuse 47, no. 4: 403–417. 10.3109/10826084.2011.639120. [DOI] [PubMed] [Google Scholar]
  3. Bereda, G. 2023. “Risks of Overtraining and Compulsive Exercising on Human Health.” BOHR International Journal of Current Research in Diabetes and Preventive Medicine 1, no. 1: 30–34. 10.54646/bijrdpm.005. [DOI] [Google Scholar]
  4. Colledge, F. , Cody R., Buchner U. G., et al. 2020. “Excessive Exercise—A Meta‐Review.” Frontiers in Psychiatry 11: 521572. https://www.frontiersin.org/articles/10.3389/fpsyt.2020.521572. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Dalle Grave, R. , Calugi S., and Marchesini G.. 2008. “Compulsive Exercise to Control Shape or Weight in Eating Disorders: Prevalence, Associated Features, and Treatment Outcome.” Comprehensive Psychiatry 49, no. 4: 346–352. 10.1016/j.comppsych.2007.12.007. [DOI] [PubMed] [Google Scholar]
  6. DerSimonian, R. , and Laird N.. 1986. “Meta‐Analysis in Clinical Trials.” Controlled Clinical Trials 7, no. 3: 177–188. 10.1016/0197-2456(86)90046-2. [DOI] [PubMed] [Google Scholar]
  7. Devoe, D. J. , Anderson A., Bahji A., et al. 2022. “The Prevalence of Impulse Control Disorders and Behavioral Addictions in Eating Disorders: A Systematic Review and Meta‐Analysis.” Frontiers in Psychiatry 12: 724034. 10.3389/fpsyt.2021.724034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Di Lodovico, L. , Poulnais S., and Gorwood P.. 2019. “Which Sports Are More at Risk of Physical Exercise Addiction: A Systematic Review.” Addictive Behaviors 93: 257–262. 10.1016/j.addbeh.2018.12.030. [DOI] [PubMed] [Google Scholar]
  9. Dittmer, N. , Jacobi C., and Voderholzer U.. 2018. “Compulsive Exercise in Eating Disorders: Proposal for a Definition and a Clinical Assessment.” Journal of Eating Disorders 6, no. 42: 42. 10.1186/s40337-018-0219-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dittmer, N. , Voderholzer U., Mönch C., Cuntz U., Jacobi C., and Schlegl S.. 2020. “Efficacy of a Specialized Group Intervention for Compulsive Exercise in Inpatients With Anorexia Nervosa: A Randomized Controlled Trial.” Psychotherapy and Psychosomatics 89, no. 3: 161–173. 10.1159/000504583. [DOI] [PubMed] [Google Scholar]
  11. Fietz, M. , Touyz S., and Hay P.. 2014. “A Risk Profile of Compulsive Exercise in Adolescents With an Eating Disorder: A Systematic Review.” Advances in Eating Disorders 2, no. 3: 241–263. 10.1080/21662630.2014.894470. [DOI] [Google Scholar]
  12. Friars, D. , Walsh O., and McNicholas F.. 2023. “Assessment and Management of Cardiovascular Complications in Eating Disorders.” Journal of Eating Disorders 11, no. 1: 13. 10.1186/s40337-022-00724-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Galmiche, M. , Déchelotte P., Lambert G., and Tavolacci M. P.. 2019. “Prevalence of Eating Disorders Over the 2000–2018 Period: A Systematic Literature Review.” American Journal of Clinical Nutrition 109, no. 5: 1402–1413. 10.1093/ajcn/nqy342. [DOI] [PubMed] [Google Scholar]
  14. Harris, A. , Hay P., and Touyz S.. 2020. “Psychometric Properties of Instruments Assessing Exercise in Patients With Eating Disorders: A Systematic Review.” Journal of Eating Disorders 8, no. 1: 45. 10.1186/s40337-020-00315-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Higgins, J. P. T. , Thomas J., Chandler J., et al., eds. 2024. Cochrane Handbook for Systematic Reviews of Interventions Version 6.5 (Updated August 2024). Cochrane. www.training.cochrane.org/handbook. [Google Scholar]
  16. Himmerich, H. , Kan C., Au K., and Treasure K.. 2021. “Pharmacological Treatment of Eating Disorders, Comorbid Mental Health Problems, Malnutrition and Physical Health Consequences.” Pharmacology & Therapeutics 217: 107667. 10.1016/j.pharmthera.2020.107667. [DOI] [PubMed] [Google Scholar]
  17. Keski‐Rahkonen, A. , and Mustelin L.. 2016. “Epidemiology of Eating Disorders in Europe: Prevalence, Incidence, Comorbidity, Course, Consequences, and Risk Factors.” Current Opinion in Psychiatry 29, no. 6: 340–345. 10.1097/YCO.0000000000000278. [DOI] [PubMed] [Google Scholar]
  18. Lampe, E. W. , Wons O., Taylor L. C., Juarascio A. S., and Manasse S. M.. 2022. “Associations Between Fear of Weight Gain and Exercise in Binge‐Spectrum Eating Disorders.” Eating and Weight Disorders ‐ Studies on Anorexia, Bulimia and Obesity 27, no. 6: 2121–2128. 10.1007/s40519-022-01361-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Lichtenstein, M. B. , Hinze C. J., Emborg B., Thomsen F., and Hemmingsen S. D.. 2017. “Compulsive Exercise: Links, Risks and Challenges Faced.” Psychology Research and Behavior Management 10: 85–95. 10.2147/PRBM.S113093. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lichtenstein, M. B. , Melin A. K., Szabo A., and Holm L.. 2021. “The Prevalence of Exercise Addiction Symptoms in a Sample of National Level Elite Athletes.” Frontiers in Sports and Active Living 3: 635418. 10.3389/fspor.2021.635418. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Marques, A. , Peralta M., Sarmento H., Loureiro V., Gouveia E. R., and Gaspar de Matos M.. 2019. “Prevalence of Risk for Exercise Dependence: A Systematic Review.” Sports Medicine 49, no. 2: 319–330. 10.1007/s40279-018-1011-4. [DOI] [PubMed] [Google Scholar]
  22. Martenstyn, J. A. , Aouad P., Touyz S., and Maguire S.. 2022. “Treatment of Compulsive Exercise in Eating Disorders and Muscle Dysmorphia: A Systematic Review and Meta‐Analysis.” Clinical Psychology: Science and Practice 29, no. 2: 143–161. 10.1037/cps0000064. [DOI] [Google Scholar]
  23. Mathisen, T. F. , Bratland‐Sanda S., Rosenvinge J. H., et al. 2018. “Treatment Effects on Compulsive Exercise and Physical Activity in Eating Disorders.” Journal of Eating Disorders 6, no. 1: 43. 10.1186/s40337-018-0215-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Meneguzzo, P. , Todisco P., Calonaci S., et al. 2021. “Health‐Related Quality of Life Assessment in Eating Disorders: Adjustment and Validation of a Specific Scale With the Inclusion of an Interpersonal Domain.” Eating and Weight Disorders ‐ Studies on Anorexia, Bulimia and Obesity 26, no. 7: 2251–2262. 10.1007/s40519-020-01081-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Meyer, C. , Taranis L., and Touyz S.. 2008. “Excessive Exercise in the Eating Disorders: A Need for Less Activity From Patients and More From Researchers….” European Eating Disorders Review 16, no. 2: 81–83. 10.1002/erv.863. [DOI] [PubMed] [Google Scholar]
  26. Meyer, M. , Sattler I., Schilling H., et al. 2021. “Mental Disorders in Individuals With Exercise Addiction – A Cross‐Sectional Study.” Frontiers in Psychiatry 12: 751550. 10.3389/fpsyt.2021.751550. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Momen, N. C. , Plana‐Ripoll O., Yilmaz Z., et al. 2022. “Comorbidity Between Eating Disorders and Psychiatric Disorders.” International Journal of Eating Disorders 55, no. 4: 505–517. 10.1002/eat.23687. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Mond, J. , and Gorrell S.. 2021. “‘Excessive Exercise’ in Eating Disorders Research: Problems of Definition and Perspective.” Eating and Weight Disorders ‐ Studies on Anorexia, Bulimia and Obesity 26, no. 4: 1017–1020. 10.1007/s40519-020-01075-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Peñas‐Lledó, E. , Vaz Leal F. J., and Waller G.. 2002. “Excessive Exercise in Anorexia Nervosa and Bulimia Nervosa: Relation to Eating Characteristics and General Psychopathology.” International Journal of Eating Disorders 31, no. 4: 370–375. 10.1002/eat.10042. [DOI] [PubMed] [Google Scholar]
  30. R Core Team . 2023. R: A Language and Environment for Statistical Computing [Computer Software]. R Foundation for Statistical Computing. https://www.R‐project.org/. [Google Scholar]
  31. Robinson, L. , Aldridge V., Clark E. M., Misra M., and Micali N.. 2016. “A Systematic Review and Meta‐Analysis of the Association Between Eating Disorders and Bone Density.” Osteoporosis International 27, no. 6: 1953–1966. 10.1007/s00198-015-3468-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Samnaliev, M. , Noh H. A., Sonneville K. R., and Austin S. B.. 2014. “The Economic Burden of Eating Disorders and Related Mental Health Comorbidities: An Exploratory Analysis Using the U.S. Medical Expenditures Panel Survey.” Preventive Medicine Reports 2: 32–34. 10.1016/j.pmedr.2014.12.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Sato, Y. , and Fukudo S.. 2015. “Gastrointestinal Symptoms and Disorders in Patients With Eating Disorders.” Clinical Journal of Gastroenterology 8, no. 5: 255–263. 10.1007/s12328-015-0611-x. [DOI] [PubMed] [Google Scholar]
  34. Scharmer, C. , Gorrell S., Schaumberg K., and Anderson D.. 2020. “Compulsive Exercise or Exercise Dependence? Clarifying Conceptualizations of Exercise in the Context of Eating Disorder Pathology.” Psychology of Sport and Exercise 46: 101586. 10.1016/j.psychsport.2019.101586. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Shroff, H. , Reba L., Thornton L. M., et al. 2006. “Features Associated With Excessive Exercise in Women With Eating Disorders.” International Journal of Eating Disorders 39, no. 6: 454–461. 10.1002/eat.20247. [DOI] [PubMed] [Google Scholar]
  36. Smink, F. R. E. , van Hoeken D., and Hoek H. W.. 2012. “Epidemiology of Eating Disorders: Incidence, Prevalence, and Mortality Rates.” Current Psychiatry Reports 14, no. 4: 405–414. 10.1007/s11920-012-0282-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Solmi, M. , Fabiano N., Clarke A. E., et al. 2024. “Adverse Outcomes and Mortality in Individuals With Eating Disorder‐Related Electrolyte Abnormalities in Ontario, Canada: A Population‐Based Cohort Study.” Lancet Psychiatry 11, no. 10: 818–827. 10.1016/S2215-0366(24)00244-X. [DOI] [PubMed] [Google Scholar]
  38. Tonia, T. , Buitrago‐Garcia D., Peter N. L., et al. 2023. “Tool to Assess Risk of Bias in Studies Estimating the Prevalence of Mental Health Disorders (RoB‐PrevMH).” BMJ Mental Health 26, no. 1: e300694. 10.1136/bmjment-2023-300694. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Treasure, J. , Duarte T. A., and Schmidt U.. 2020. “Eating Disorders.” Lancet 395, no. 10227: 899–911. 10.1016/S0140-6736(20)30059-3. [DOI] [PubMed] [Google Scholar]
  40. Trott, M. , Jackson S. E., Firth J., et al. 2021. “A Comparative Meta‐Analysis of the Prevalence of Exercise Addiction in Adults With and Without Indicated Eating Disorders.” Eating and Weight Disorders ‐ Studies on Anorexia, Bulimia and Obesity 26, no. 1: 37–46. 10.1007/s40519-019-00842-1. [DOI] [PubMed] [Google Scholar]
  41. van Hoeken, D. , and Hoek H. W.. 2020. “Review of the Burden of Eating Disorders: Mortality, Disability, Costs, Quality of Life, and Family Burden.” Current Opinion in Psychiatry 33, no. 6: 521–527. 10.1097/YCO.0000000000000641. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. White, J. , and Halliwell E.. 2010. “Examination of a Sociocultural Model of Excessive Exercise Among Male and Female Adolescents.” Body Image 7, no. 3: 227–233. 10.1016/j.bodyim.2010.02.002. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information S1

ERV-33-1005-s001.docx (219.9KB, docx)

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Articles from European Eating Disorders Review are provided here courtesy of Wiley

RESOURCES