Eligibility Criteria
Our systematic review will be conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and studies will be selected for inclusion in this review according to the criteria outlined below.
Study Designs
Eligible studies must assess changes in compulsive exercise from pre- to post-intervention. This comprises studies that include a comparison group (e.g., randomised controlled trials) and those that do not include a comparison group (e.g., pilot studies and case studies). If data are reported for the same intervention across multiple studies, for example, a pilot study and a subsequent randomised controlled trial, only data from the randomised controlled trial will be extracted for narrative synthesis. Literature reviews, theoretical articles, and conference abstracts will be excluded.
Participants
We will include studies that sampled adolescents, defined as age above 10 years (Sawyer et al., 2018), and/or adult participants of both sexes who have been diagnosed with either an eating disorder or muscle dysmorphia. These clinical diagnoses must be determined using established diagnostic criteria, such as the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classification of Diseases (ICD). The list of possible eating disorder diagnoses includes anorexia nervosa, bulimia nervosa, binge eating disorder, eating disorder not otherwise specified (DSM-IV), and other specified feeding or eating disorder (DSM-V). Although muscle dysmorphia is included in the DSM-V as a specifier of body dysmorphic disorder (American Psychiatric Association, 2013), it was not officially included in earlier versions of the DSM or any version of the ICD. Thus, we will also include studies carried out prior to publication of the DSM-V that used the diagnostic criteria for muscle dysmorphia outlined in Pope et al. (1997). Studies that used clinical cut-off scores from questionnaires to infer the presence of an eating disorder or muscle dysmorphia without a corresponding clinical interview will be excluded. Likewise, studies that sampled participants based on a self-reported diagnosis and not a formal diagnosis from a relevant medical professional will also be excluded. The presence of comorbidities (e.g., depression, generalised anxiety disorder, or obsessive-compulsive disorder) will not be used to determine inclusion or exclusion.
Interventions
Studies that used any type of biopsychosocial treatment and reported longitudinal changes in compulsive exercise will be considered for inclusion in this review. These interventions may consist of, but may not be limited to, cognitive-behavioural therapy, psychoeducation, sport therapy, structured exercise programs, physical exercise and dietary therapy, exposure and response prevention, pharmacotherapy, and multi-component strategies. Both individual and group interventions will be included. We will not place restrictions on the treatment setting, for example, inpatient, outpatient, self-help app, and web-based interventions are eligible for inclusion. There will also be no restrictions on the length of treatment or the number of treatment sessions provided over the course of the intervention.
Outcomes of Interest
The primary outcome of interest in this review is changes in compulsive exercise from pre- to post-intervention. Although compulsive exercise is the preferred term to describe problematic exercise behaviour in eating dis orders patients, its use did not reach consensus in a recent Delphi study (Noetel et al., 2017). Thus, a number of other related terms will also be included as primary outcomes, including exercise addiction, exercise dependence, and obligatory exercise (see Appendix A for the full list of terms used in the search strategy). We will only include studies that measured compulsive exercise or a related construct using a standardised instrument, such as the Compulsive Exercise Test (Taranis et al., 2011), Exercise Addiction Inventory (Griffiths et al., 2005), or Obligatory Exercise Questionnaire (Steffen & Brehm, 1999). We will exclude studies that only reported on the quantity or frequency of exercise performed without assessing the corresponding qualitative relationship with exercise. Secondary outcomes of interest in this review are changes in weight, eating disorder psychopathology, and other relevant psychological outcomes (e.g., mood and anxiety).
Search Strategy and Study Selection
The following five electronic databases will be searched from database inception until November 2020: (1) PsycInfo, (2) MEDLINE, (3) Embase, (4) Web of Science, and (5) Scopus. These particular databases were chosen in consultation with an experienced university librarian who specialises in systematic reviews in the psychological sciences. To find additional eligible studies, the reference lists of included articles will be manually searched. Studies citing relevant articles will also be reviewed using the ‘cited by’ function in Google Scholar. We will conduct manual searches for unpublished and ongoing randomised controlled trials in international registers (e.g., www.clinical-trials.gov). If a relevant conference abstract is returned without a corresponding full text article, publications of the lead author will be searched. Papers are required to be peer-reviewed and written in English for inclusion in this review. We did not place restrictions on the date of publication or the geographical location where the research was conducted.
We developed a preliminary search strategy in PsycInfo that was modified in response to suggestions from the aforementioned university librarian. The final search strategy (see Appendix A) comprised three clusters of terms relating to: (a) unhealthy exercise behaviour (e.g., compulsive exercise, exercise addiction, and obligatory exercise); (b) diagnostic labels for eating disorders (e.g., anorexia, bulimia, and binge eating disorder) and muscle dysmorphia; and (c) study design (e.g., intervention, treatment, and protocol).
One reviewer (JM) will screen all titles and abstracts against the eligibility criteria after removing duplicates and full texts will be obtained for papers deemed relevant or where inclusion was uncertain. Another reviewer (PA) will cross-review 20% of excluded papers, selected at random. Both reviewers (JM and PA) will review all obtained full texts against the eligibility criteria. Queries regarding eligibility of articles will be resolved through discussion with a third reviewer (SM).
Data Extraction
Data from included articles will be extracted by one author (JM) and checked for accuracy by a second author (PA). The following information will be extracted from each article: (a) study characteristics (title, lead author, year of publication, country, and language); (b) participant characteristics (sample size, age and sex distribution, ethnicity, socioeconomic status, type of disorder, diagnostic classification system, baseline body mass index, comorbidities, duration of illness, dropout rates); (c) study methodology (type of study, study setting, intervention description, comparison group description, length of intervention, and number of treatment sessions); (d) compulsive exercise assessment (questionnaire name, questionnaire domains reported, and assessment time points); and (e) study results (statistical measures used, and primary and secondary outcomes results reported across all time points).
Quality Assessment
The methodological quality of included studies will be appraised using a modified version of the Downs and Black (1998) criteria, as amended by Ferro and Speechley (2009) to encompass a broad range of study designs. This modified quality checklist excluded items specific to randomised controlled trials, such as those items assessing randomisation, dropouts, blinding, and intervention integrity. The amended quality checklist had 15 items, as opposed to the original 27, and dichotomously scored items as 0 (unable to determine/no) or 1 (yes). The amended checklist assesses four domains: (1) reporting (seven items); (2) external validity (three items); (3) internal validity (four items); and (4) statistical power (one item). Scores from these four domains can be summed to determine an overall quality score for each study. The maximum checklist score for each study was 15 with higher scores (quality score > 10) indicative of greater methodological quality (Ferro & Speechley, 2009)
Two reviewers (JM and PA) will independently score all included studies after data extraction has been completed. Inter-rater agreement will be calculated using the kappa statistic. Large discrepancies in quality evaluation will be resolved through discussion with a third reviewer (SM).
Data Synthesis
Results for this review will be reported using narrative synthesis. Meta-analysis was considered but is likely not feasible taking into account the heterogeneity of study designs, intervention approaches, compulsive exercise instruments, and assessment intervals. Studies receiving a higher overall quality score will be weighted more in the narrative synthesis than studies with a lower overall quality score.
Dissemination
Findings from this review will be published in a peer-reviewed scientific journal and may be presented by the lead author (JM) at national and international conferences. Data extracted from individual articles will be made available with the full manuscript.