The reporting will be in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines for protocols (PRISMA-P) (40). The study protocol is also registered in the International Prospective Register of Systematic Reviews (PROSPERO, Registration ID: CRD42020173434).
Search strategy
The relevant articles will be identified in the following databases up to August 2020: PubMed/MEDLINE, EMBASE, ISI (Web of Science), Scopus, and Google Scholar using Medical Subject Heading (MeSH) and non-MeSH keywords. We will not apply any language or other restrictions. In addition, we will check the reference lists of all relevant studies to identify additional relevant articles. Unpublished studies will be identified by searching the websites indexing the preprints such as Research Square (https://www.researchsquare.com/) and the registered clinical trials approved by the World Health Organization (WHO). All abstracts of interest will be evaluated for further information by contacting the authors. The search strategies that will be applied in PubMed, Scopus, and ISI are provided in supplementary table 1.
Study selection
Two investigators will independently perform the study selection. All articles from electronic searches will be imported into the EndNote software (version: desktop, X7; Thompson Reuters, New York, USA) and duplicate studies will be deleted. Titles, abstracts and full-text articles will be screened and cross-checked according to the eligibility criteria for study inclusion independently by 6 reviewers (Z.Y, S.S, S.B, SH.R, S.MT, and T.Z). Any disagreements will be resolved by discussion and consensus. The PRISMA flow chart will be presented to describe the process of the study selection.
Articles selected for full text review must meet the following criteria:
- Participants must be a minimum 18 years of age and older and have a body mass index (BMI) ≥ 25 kg/m2 (pregnant and lactating women will be excluded);
- Interventions must contain one arm in which participants receive an exercise intervention (i.e. aerobic or resistance) with a weight-loss (i.e. hypocaloric) diet and one arm where participants only receive a weight-loss diet (Exactly according to the diet considered for the intervention group);
- Interventions must be randomized or non-randomized controlled clinical trials with either a parallel, cross-over, or factorial design with at least two weeks of follow-up.
Data Extraction
Data extraction and management
The following data will be extracted using a predefined data extraction form by two independent investigators from the eligible studies and any discrepancy will be resolved by a third author:
Study and participant’s characteristics: The participants age, number of males and females, number of participants in the intervention and control group/period, the geographical location of the study and the health condition of participants.
Intervention details: The study design (parallel/cross-over/factorial), number of study arms, the intervention duration, funding source(s), amount of calorie restriction, type of diets and exercise programs, intensity, frequency, compliance, and delivery of each exercise used for the intervention group.
Outcome measures: Data on baseline, post-intervention or change from baseline mean ± standard deviation (SD) for energy intake, body weight, anthropometrics and body composition measures, blood glucose control markers (serum/plasma fasting glucose, insulin, C-peptide, insulin resistance markers including HOMA-IR and hemoglobin A1C), lipid profile [serum total cholesterol, triglycerides, low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C), and apoproteins], systolic and diastolic blood pressure, sexual hormones (testosterone and estradiol), SHBG, serum/plasma inflammation (hs-CRP, IL-6, and TNF-a), bone health markers, liver and kidney enzymes, depression, anxiety, and quality of life, will be extracted for the intervention and control groups/periods. P-values for within-group and between-group comparison will also be collected to calculate the change values.
Assessment of risk of bias in individual studies
The eligible studies will be assessed using the Cochrane collaboration’s risk of bias assessment tool considering 7 domains: (i) random sequence generation (selection bias), (ii) allocation concealment (selection bias), (iii) blinding of participants and personnel (performance bias), (iv) blinding of outcome assessment (detection bias), (v) incomplete outcome data (attrition bias), (vi) selective reporting (reporting bias), and (vii) the dietary compliance as another possible source of bias in dietary interventions. Each study will be judged as low risk of bias, high risk of bias, or unclear risk of bias according to the mentioned domains (41). The overall quality of studies will be classified as low risk (low risk for all domains), unclear risk (unclear for at least one domain), and high risk (high risk for at least one domain).
Data analysis
The mean change values from baseline for the intervention (weight loss diet + exercise) and control group/period (weight loss diet alone) and their standard deviations (SDs) will be used to calculate the raw mean difference and standard error (SE) between the intervention and control. The hedges’ g (bias corrected standardized mean difference) statistic and corresponding SD will be calculated for outcome variables reported in different scales. The mean difference will be used as the effect size for meta-analysis. If the change values were not reported, we will calculate SD for the change values by selecting 0.5 as the reference correlation coefficient between baseline and end point values (r = 0.5) and to make sure that the meta-analysis was not sensitive to the selected correlation coefficient, all analyses will be repeated using 0.2 and 0.8 as the correlation coefficient. The weighted mean difference (WMD) and its corresponding 95% confidence intervals (CIs) will be derived using the random effects model which takes the between-study heterogeneity into account (42). All statistical analyses will be performed using STATA, version 11.2 (Stata Corp, College Station, TX) and a two-sided P-value less than 0.05 will be considered as statistically significant. If data cannot be meta-analyzed, we will summarize the articles and conclude on high-quality studies.
Between study heterogeneity and subgroup analysis
The heterogeneity will be checked using Cochran’s Q test and I-squared statistic (I2 is an estimate for between study variation to total meta-analysis variation ratio ranging from 0-100%,)(43). P values less than 0.05 for Cochran’s Q test and I2 ≥ 25% will be considered as high level of heterogeneity. To examine the potential sources of between-study heterogeneity, several subgroup analyses based on follow-up duration, the health status of the participants, the diet used for weight loss, the exercise, participants’ sex and other possible variables will be conducted.
Sensitivity analysis
The sensitivity analysis will be done by sequentially removing individual studies included in the meta-analyses to assess the robustness of the meta-analyses (44).
Publication bias
In case there are fewer than 10 studies in a meta-analysis, we will construct a funnel plot to investigate the potential for publication bias for the primary outcome by visual inspection for asymmetry. If our meta-analysis involves 10 or more studies, publication bias will be evaluated by inspecting Begg’s funnel plots and Egger’s and Begg’s asymmetry tests (45). Duval and Tweedie’s trim and fill analysis will be conducted if the publication bias becomes evident (46).
Dealing with missing data
If data are missing, we will attempt to contact the authors through e-mails to obtain missing data or additional information twice, one week apart. The impact of missing data will also be evaluated in the sensitivity analysis. Additionally, we will describe the possible influences of missing data in the “Discussion” section of the resulting publications.
Confidence in cumulative evidence
Overall quality of the evidence will be assessed by using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) tool (47) with GRADEprofiler (GRADEpro) V.3.6 software, identifying the quality of evidence for each outcome as the extent to which one can be confident that an estimate of effect is near to the quantity of certain interest (45). There are four levels used to rate the quality of evidence across trials in the GRADE system: very low, low, moderate, and high. Randomized clinical trials are categorized as high quality but can be downgraded due to limitation in study design, indirectness of evidence, imprecision of results, unexplained heterogeneity or inconsistency of results or high probability of publication bias (47).