OBJECTIVES: Network meta-analyses (NMAs) are the gold standard method for comparing efficacy and safety of therapies in the absence of head-to-head trials. In health technology assessment, the use of NMA is essential to evidence synthesis but often repetitive and time consuming, especially for time-to-event (TTE) data where full Kaplan Meier curves need to be considered. In this work, we seek to develop a tool for improving speed and reliability of NMA analyses in a transparent and reproducible way, necessary in the HTA context.

METHODS: We chose Bayesian survival NMA models in accordance with the National Institute for Health and Care Excellence (NICE) guidelines, including Weibull, Gompertz, log-normal, log-logistic and exponential. We also added order 1 and 2 fractional polynomial models. All of the models were programmed in WinBUGS but are executed from R. We included automatic checks for validity of input/output and tweaked initial value generation for WinBUGS to avoid computational issues. Survival and hazards functions for fitted models are derived and plotted automatically. We distributed our software as an open-source R package.

RESULTS: We conducted a case study in efficacy of treatments for metastatic renal cell carcinoma (mRCC). All the models included in the package have successfully converged with default settings. We automatically generated a report with appropriate plots and summaries. In comparison to analyse data solely in WinBUGS additional checks were needed to format inputs, find suitable initial values and re-run some models to improve convergence.

CONCLUSIONS: Conducting the survival curve NMA analyses using the package provides a user-friendly alternative to running many analyses with WinBUGS. We improved speed and reliability by automating repetitive parts of the analysis and removed problems with initial value choice typical to WinBUGS. By distributing the tool as open-source, all of the analyses are reproducible and transparent