Appliance classes for the correction of Class II malocclusion due to mandibular deficiency in growing patients: a systematic review and Bayesian network meta-analysis

This repository contains the data and the R code used in a systematic review and network meta-analysis comparing classes of appliances for the correction of Class II malocclusion due to mandibular deficiency in growing patients. The review protocol was registered in PROSPERO (CRD420251130636). The material is shared so that the analyses reported in the article and its supplementary files can be checked, rerun and audited. Appliances evaluated in the included randomized controlled trials were grouped into seven classes: bipartite, monoblock, rigid fixed, hybrid fixed, extraoral traction, prefabricated and conventional fixed. Untreated controls formed an eighth node, used as the reference. Three cephalometric outcomes were analysed as the mean change from baseline (final minus initial, in degrees): SNB, ANB and the incisor mandibular plane angle (IMPA). The data folder contains, for each outcome, an arm-level file (one row per trial arm, with sample size, mean change, standard deviation of the change and follow-up) and a contrast-level file with the two-arm comparisons as reported in the trials. When the standard deviation of the change was not reported, it was imputed from the baseline and final standard deviations assuming a correlation of 0.5. The CINeMA input files, with risk-of-bias and indirectness judgements, are also provided. The R folder contains one self-contained script per outcome and one script for the combined ANB and SNB league table. Each outcome script includes: Data import, checks and combination of arms of the same class within a trial (Cochrane Handbook, section 6.5.2.10). Pairwise random-effects meta-analyses (REML), with Hartung-Knapp and leave-one-out analyses. Bayesian random-effects network meta-analysis with gemtc and JAGS (arm-based normal likelihood, identity link, four chains, 20,000 adaptation and 150,000 iterations, thinning of 150, fixed seeds), with convergence diagnostics, model fit, relative effects, league table, rank probabilities and SUCRA. Assessment of inconsistency by node-splitting and an unrelated mean effects model. Sensitivity analyses with a half-normal prior for the between-study standard deviation and a common-effect model, and a meta-regression on follow-up duration. A frequentist network meta-analysis with netmeta, including the design-by-treatment interaction test, net-splitting and the contribution matrix. For SNB and ANB, the scripts also refit the network under alternative classifications of appliances and alternative inclusion criteria. Results are printed in the R console and figures are displayed in the plotting window. Tables (.xlsx), figures (.tiff) and fitted models (.rds) are written to an output folder only when the user sets export <- TRUE before running the scripts. The README file describes the variables, explains how to run the code and maps each section of the scripts to the corresponding tables and figures of the article and supplementary files. The analyses were run in R 4.4.1 with meta 8.5-0, netmeta 3.3-1, gemtc 1.1-1, rjags 4-17, coda 0.19-4.1 and JAGS 4.3.2. Because MCMC sampling depends on the software versions, posterior summaries obtained with other versions may differ slightly, usually in the second decimal place. Keywords: Class II malocclusion; functional appliances; orthodontics; network meta-analysis; Bayesian analysis; systematic review; R.