CE4 Expanding Evidence Base VS Introducing Heterogeneity in Networks for Network Meta-Analyses: A Simulation Study


      Network meta-analyses (NMAs) are widely used to estimate the relative treatment effect between treatments that have not directly been compared in clinical trials. Challenges arise in precision medicines, where only small networks of evidence are available. This study aimed to explore the trade-off between increasing the evidence base by including additional studies and increasing the level of heterogeneity in the network considered.


      Data were simulated to reflect a small network including four treatments. Four scenarios were simulated with increasing number of studies per direct comparison. Each study was assumed to add an additional level of heterogeneity in the network, that was partly explained by an observable covariate. Scenario 1 represented a small network with a low level of heterogeneity while scenario 4 was a bigger network with a higher heterogeneity level. Standard NMAs were conducted and the mean relative difference (MRD) between the NMA estimates and the true treatment effects were used to compare the scenarios. For the two largest networks, meta-regressions were also conducted to adjust partially on the heterogeneity.


      From scenarios 1 to 3, including additional studies in the networks while increasing the heterogeneity level reduced the MRDs when conducting standard NMAs (respectively 48.2%, 21.3% and 18.6%). However, a higher MRD was obtained for the largest network (scenario 4: 24.4%) compared to scenarios 2 and 3. Conducting meta-regressions markedly reduced the MRDs for the scenario 3 (7.2% with meta-regression vs. 18.6%) and scenario 4 (9.1% with meta-regression vs. 24.4%).


      This simulation study shows that including additional studies in case of a small network in a NMA context can be relevant, especially if it allows for a meta-regression to be conducted.