2.5. Statistical analysis
The odds ratios (ORs) with 95% confidence intervals (CIs) for dichotomous outcomes and standardized mean differences (SMDs) or mean differences (MDs) with 95% CIs for continuous outcomes were calculated.
This network meta-analysis was performed within a frequentist framework using the STATA 16.0 (StataCorp, Texas, USA) command ‘mvmeta’[20]. Firstly, the network geometry plot for each outcome was established, which provided a visual and concise description between pairs of interventions[21]. Secondly, statistical consistency was evaluated by the node-splitting method and loop inconsistency model. P-value ≥ 0.05 or the 95% CI for each closed-loop contained 0 means direct comparison and indirect comparison were considered consistent[22]. Thirdly, a comparison-adjusted funnel plot was used to evaluate publication bias. A symmetrical graph indicated a low influence of publication bias and an asymmetric graph meant possible publication bias. Finally, the forest plot was constructed to report the results for the mixed comparison between interventions and placebo, and the league table was performed to illustrate all head-to-head comparisons. We assumed that 95% CIs not containing 0 were considered statistically significant for SMDs or MDs, and those not containing 1 were considered statistically significant for ORs. The two-dimensional graph is presented to visualize the comprehensive comparisons of drugs to placebo. The point which lies to the lower-left portion of the coordinate system and does not intersect with the dark grey dashed line indicates that this pharmacological intervention is superior to placebo in terms of both postoperative pain intensity and the incidence of PONV. Additionally, the ranking probabilities were estimated for all interventions of being at each possible rank of each intervention[21]. By using the ranking probabilities, the treatment hierarchy was summarized and reported as the surface under the cumulative ranking curve (SUCRA)[21]. The larger the SUCRA value is, the better are the rank of the treatment for outcomes.