Data synthesis and analyses

We conducted a narrative synthesis based on the primary and secondary outcomes of this review. The primary outcome measures were pooled and calculated using the statistical software RevMan 5.3, according to the Cochrane Handbook for Systematic Reviews of Interventions [16]. Attempts to contact the authors to obtain the raw data for data analysis was made but to no avail. The results of the BMI z-score, waist circumference and body fat percentage were analysed using weighted or standardized mean differences as a measure of effect size, with 95% confidence intervals. Since the participant demographics and clinical settings differed among studies, we assumed the presence of heterogeneity a priori. Therefore, we used a random-effects model to pool the results. We assessed heterogeneity using the Cochran’s Q statistic of heterogeneity with reported p-value and the degree of inconsistency across studies was quantified using I². In studies with three arm RCTs, each of the intervention group was analysed independently and compared with the control group. A funnel plot was performed to determine the presence of potential publication bias using the statistical software RevMan 5.3.