OBJECTIVES: Simulation models are an essential tool in estimating the impact of vaccination, screening and treatment on cancer rates. Model calibration is the process of identifying reasonable values for model parameters, such that the outputs of the model are close to values observed in a real population. The purpose of this work was to calibrate an existing model for cervical cancer using Irish data and Markov Chain Monte Carlo (MCMC) in a Bayesian framework. This is compared and contrasted with a previous random search calibration.   METHODS: An existing microsimulation model for cervical disease which was coded in C was embedded in a loop running in R. MCMC, which is an iterative algorithm was implemented in parallel on multiple desktop machines and the results were collated for analysis. The calibration method used differs from pure optimisation strategies and identifies a probability distribution on the parameter space, which is of benefit for models requiring probabilistic sensitivity analysis. RESULTS: Estimates of the model parameters were obtained from both MCMC and from the fitting of existing reference parameter sets resulting from a random search of the parameter space. These are compared on the basis of goodness of fit statistics (the sum of squared errors between targets and fitted values). Of 20 MCMC chains that were run, 5 of them gave better fits than the best fit sets for the random search method. However, 8 of the 20 chains had not reached parameter sets that gave good fits when compared with the best 135 fitted sets from the random search method.   CONCLUSIONS: MCMC is a useful technique which provides probabilistic estimates of the parameters of interest in a calibration exercise. Care is needed with starting values and proposal distributions to ensure that the chains have converged and that the parameter space is properly explored.