OBJECTIVES: Bayesian network meta-analysis (NMA) allows for the estimation of relative treatment effects in a connected evidence network. We propose the use of single armed observational data to enrich the evidence base where RCT data alone does not form a connected network to allow for pairwise comparisons between treatments. The approach is presented using a case study in relapsed or refractory multiple myeloma (rrMM).

METHODS: A systematic literature review of RCT evidence reveals two disconnected evidence networks. Non-comparative observational studies are matched based on study level covariates to bridge separate networks. Since such methods are prone to bias, we capture the additional uncertainty to reduce the risk of over-confident interpretation of results. Uncertainty is captured by exploring a range of possible matches to bridge the networks.

RESULTS: CONCLUSIONS: The analysis illustrates how observational evidence can be used to bridge the gaps in existing RCT evidence; allowing for the indirect comparison of a large number of treatments which cannot be achieved using standard NMA methods. We stress the importance of incorporating additional uncertainty to avoid interpretation of results as if obtained from clinical trials. Appropriate communication of uncertainty is an advantage, compared to a naive assumption of equal efficacy between certain interventions, which has been done in the past.