OBJECTIVES: A three-state Markov model for cost-effectiveness analysis of cancer treatments requires information on both progression-free survival (PFS) and overall survival (OS). However, data is not always available on both of these outcomes. The objective of this study is to perform a Bayesian bivariate indirect comparison meta-analysis (BICMA) to obtain estimates of both PFS and OS for use in a cost-effectiveness analysis when data on these outcomes is incomplete. METHODS: In a UK Health Technology Assessment report on cost-effectiveness assessment of docetaxel with prednisone/prednisolone for the treatment of hormone-refractory metastatic prostate cancer, a two-state Markov model was specified using OS data from a single randomised controlled trial that did not report PFS. We propose the use of a Bayesian BICMA model that jointly estimates OS and PFS, and which in turn allows for the specification of a three-state Markov model incorporating a post-progression phase. Survival data for the trials included in the BICMA were reconstructed from survival curves, presented in the articles reporting the trials, using the method proposed by Guyot et. al. (BMC Med Res Methodol 2012;12:9) using the DigitizeIt and R software. RESULTS: The Bayesian BICMA model was designed to jointly model the correlated outcomes: OS and PFS using either non-informative or informative prior distribution on the correlation between the outcomes.  An informative prior distribution on the correlation between PFS and OS was based on external evidence using prostate cancer trials presented in Halabi et. al. (Clin Oncol 2009;27(17):2766-71). Modelling the correlated outcomes jointly using this bivariate model allows prediction of PFS for the comparison of interest. CONCLUSIONS: In the absence of evidence on PFS, required for the specification of a three-state Markov model, the proposed method allows PFS to be constructed thus eliminating the need to reduce the cost-effectiveness analysis to a two-state Markov model.