Statistical analysis

All analyses were performed using version 2.0 of the Comprehensive Meta-Analysis software (Biostat Inc). Three separate analyses were performed to calculate the standardized mean differences (SMDs), weighted mean differences (WMD), and 95% CIs using random-effect models using the DerSimonian and Laird approach for comparing the effects of the following 1) Ex with DI, 2) Ex + DI with Ex, and 3) Ex + DI with DI on the main outcomes (leptin and adiponectin) and secondary outcome (body weight). The results were pooled using random-effects models based on the assumption that heterogeneity was likely from a clinical or methodologic perspective and may have affected the results [51]. Several subgroup analyses were performed based on the participants’ BMIs (overweight: <30; obesity: ≥30), duration of intervention (short term: ≤12 wk; medium term: >12 wk and ≤24 wk; long term: >24 wk), age of the participants (<18 y, ≤18 y to <50 y, or ≥50 y), type of exercise (aerobic, resistance, or combined), type of exercise supervision (supervised, nonsupervised, or both), PEDro Scale score (score: <5 or ≥5), and magnitude of energy restriction (mild or severe). In pediatric studies, all participants had obesity and were categorized by obesity subgroup. The interpretation of effect sizes was conducted using the Cochrane guidelines, with 0.20–0.49, 0.50–0.79, and ≥0.8 indicating small, medium, and large effect sizes, respectively. Heterogeneity was assessed using the I² statistic, which interprets I² according to the Cochrane guidelines as follows: 25%, 50%, and 75% indicates low, moderate, and high heterogeneity, respectively. Visual interpretation of funnel plots and Egger tests were used for the detection of publication biases, wherein the Egger test was used as a secondary determinant if P was <0.10. The trim-and-fill correction method was used to address the potential effects of publication biases where relevant [52].