2.6.1. Assumptions of the NMA

The frequentist model was utilized for the NMA, combining direct and indirect evidence using R version 4.1.0 (A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria)) using the Netmeta package [29]. There are several assumptions for NMA, such as connectivity, homogeneity, transitivity, and consistency [30]. Connectivity was visually verified by connecting each network node with a line using a network plot. Homogeneity was assessed using the Cochrane Q statistic or the I<sup>2</sup> score. In our study, a random effect model was applied, as it was judged that there was heterogeneity between studies due to differences in study design or interventions [30,31]. When evaluating transitivity, it is necessary to explore the distribution of effect modifiers and determine their effects on the effect size. In our study, we qualitatively compared the sample size, age, sex, disease duration, severity, treatment dosage, and period for transitivity assessment [30]. Consistency is a quantitative statistical evaluation of transitivity. Consistency was statistically evaluated using the net-splitting method [32].
