*2.4. Data Synthesis and Statistical Analysis*

We compared moxibustion with control, acupuncture with control, and moxibustion plus acupuncture with control. The primary outcome was the fetal presentation at birth, and the secondary outcome was adverse events.

Data were analyzed using Review Manager Software (version 5.3.5). Dichotomous outcomes were extracted from each study to compute the RR with a 95% CI. The pooled RR and the associated 95% CI were estimated by the Mantel-Haenszel method. Numbers needed to treat (NNT) were calculated from the formula (NNT = 1/absolute risk reduction). We assessed clinical heterogeneity by comparing the methodologies and study designs of the included studies. Statistical heterogeneity of effect sizes between studies was assessed using the I<sup>2</sup> statistic and Q statistic with an X<sup>2</sup> test. We defined statistical heterogeneity using *<sup>p</sup>* <sup>≤</sup> 0.1 for the X<sup>2</sup> test or I<sup>2</sup> <sup>≥</sup> 50%. In the meta-analysis, a fixed-effect model was performed when there was no significant heterogeneity, and a random-effects model was performed when the heterogeneity was significant. A funnel plot was produced to detect possible publication bias. Sensitivity analysis was performed to test the robustness of results by excluding trials that used low-quality methodologies. To assess between-group differences and explain heterogeneity, we carried out a subgroup analysis. Because regional differences may exist, we reported treatment effects on breech presentation separately.
