OBJECTIVES: Recently, network meta-analysis of survival data with a multi-dimensional treatment effect was introduced using fractional polynomial (FP) distribution. With these models, the hazard ratio is not assumed to be constant over time, thereby reducing the possibility of violating consistency in indirect comparisons. However, beyond the FP models it is challenging to assess parametric distributions often used for cost-effectiveness models in health technology assessments (HTA). We aim to develop a two-step network meta-analysis (NMA) for time-to-event data.

METHODS: First, for each arm of every randomized controlled trial (RCT) connected in the network of evidence simulated patient data were fit to alternative parametric distributions, including exponential, Weibull, Gompertz, log-normal, log logistic, gamma, and generalized gamma. For each distribution, the resulting scale and shape parameters per arm were then included in a multivariate NMA, which preserved randomization and accounted for the correlation between the parameters. Results were compared to FP models to validate results and evaluate any differences.

RESULTS: An illustrative analysis is presented for a network of RCTs evaluating interventions for advanced melanoma. The NMA was assessed using alternative distributions, which were compared using Akaike information criterion, which would facilitate model averaging to propagate structural uncertainty in a cost-effectiveness analysis. Based on the generalized gamma distribution, the difference in mu, Q, and sigma parameters for each treatment versus dacarbazine (DTIC) were: Non-DTIC: 0.41 (-0.19,1.01), 0.12 (-0.08,0.32), 0.39 (-0.7,1.49); DTIC+ Interferon (IFN): 0.24 (-0.15,0.69), -0.07 (-0.24,0.11), 0.45 (-0.87,1.66); DTIC+non-IFN: 0.22 (-0.13,0.65), -0.05 (-0.21,0.11), -0.17 (-1.13,1.1).

CONCLUSIONS: A two-step NMA of survival data allows for a straightforward comparison of alternative parametric distributions in terms of goodness of fit by avoiding the need to specify each distribution in WinBUGS. This approach provides a more generalizable evidence synthesis framework for HTA.