The systematic literature search was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement 
Any randomized (RCT) and non-randomized controlled trials (non-RCT) or clinical trials (CT) meeting the subsequent specifications were included. Trials were included if the mean age of participants (males and females) was older than 18 years (adults, no upper age limit) and if the mean BMI at baseline was above 25.0 kg.m−2
. Studies with one or more cohorts participating in aerobic or resistance exercise (physical activity) were eligible for inclusion in the meta-analysis. Physical activity was defined as a program that included voluntary aerobic or resistance exercise at “low to moderate” or “vigorous” intensity for at least two sessions per week. Study inclusion was limited to four discrete measurements of visceral adipose tissue: 1) Computerized Tomography (CT scan), 2) Magnetic Resonance Imaging (MRI), 3) dual energy x-ray absorptiometry (DXA), and 4) Ultrasound (US). These tests were selected because of their documented validity and reliability for assessments as well as reported prevalence in the literature.
The studies were expected to conduct a (supervised) physical activity intervention without dietary intervention and to include all information needed for further meta-analysis in order to be considered for inclusion. Diet-only and supplementation-only studies or studies not meeting the inclusion criteria (e.g. physical activity intervention less than 8 weeks) were excluded. If a study examined the effects on visceral adipose tissue in diet and exercise groups, only the data of the exercise intervention arm was included in the final analysis.
Data sources and search strategies
Databases that were systematically searched were Pubmed, SPORTDiscus, Pedro and Cochrane. The following search strategy was conducted (adapted for each database): (Overweight OR Obesity) AND (Exercise OR “Physical activity” OR “Exercise therapy” OR “Resistance training” OR “Aerobic training”) AND (“Visceral adipose tissue” OR “Intra-abdominal fat”).
Studies published in English, German, French and Dutch were included. The date range was from 1990 to August 2012.
Reference lists were checked for any topic-related relevant studies.
Hand searching and screening for abstracts and citations from annual scientific conferences relating to exercise science were not performed.
The corresponding author of a study was contacted if needed to obtain any missing information or data. If authors could not be reached or if the data were no longer available, the trial was not included in the meta-analysis.
Screening and data-extraction form
All citations identified by electronic databases were organized and the duplicates were deleted. Initially, two investigators independently screened the results from the electronic searches in order to select potentially relevant citations based on titles and abstracts. The kappa statistic was used to evaluate the chance-adjusted inter-reviewer agreement (Kappa
0.94). Inter-reviewer disagreements about study eligibility were resolved through consensus. For articles with relevant citations or with titles/abstracts that were not sufficient for deciding on inclusion/exclusion, the full-text articles were retrieved and evaluated. All studies selected at the first screening step were read and abstracted independently by three reviewers. Differences between the reviewers were resolved by consensus or referred to the third reviewer if necessary.
The following study characteristics were extracted from the articles: publication year, journal, study design, BMI, gender, type of intervention, study size, study duration, volume of physical activity, intensity of physical activity and change in VAT. Missing information was requested from authors by email.
The quality of each eligible study was assessed in duplicate. Disagreements were resolved by mediation, if necessary with input from a third investigator.
The Critical Review Form for Quantitative Studies (Mc Master University 1998)
was used for quality assessment, resulting in a maximum score of 15. Only studies with a score of 8 or higher were included.
A meta-analysis with a random-effects model (specified a priori), accounting for possible heterogeneity between the studies, was used to examine the overall effect size of physical activity on visceral adipose tissue.
Effect sizes (change in VAT) of the uncontrolled and controlled studies were calculated as standardized mean differences and expressed as Hedge's g to correct for overestimating the true effect. The 95% confidence intervals [95%CI] were calculated for the individual studies and the overall estimate.
Subgroup analyses were conducted to assess the influence of different co-variates, such as the intensity of physical activity, on the overall estimate of VAT change. Meta regression was used to assess the possible influence of the duration (expressed as weeks) of intervention on the effect sizes of the 15 studies under investigation.
The Cochran's Q statistic and I2 were calculated to assess the degree of heterogeneity across studies. Publication bias was assessed using visual analysis of the funnel plot and by formal testing for funnel plot asymmetry using the ‘trim and fill’ and the ‘fail ’n safe' algorithms. For all analyses, P values less than 0.05 were considered significant. All calculations and plots were conducted using the CMA-2 software (Comprehensive Meta-Analysis 2nd version, Biostat, Englewood, NJ, USA).
For purpose of clinical interpretation the overall estimate of a meta-analysis on a subgroup of five controlled studies which used the same measurement scale (cm2
) was re-expressed in the original units following the guidelines as described in the Cochrane handbook for systematic reviews of interventions
. Baseline data of the McTiernan study
was used to calculate a pooled standard deviation for the female and male experimental and control groups as well as for the combined gender groups.