Twin concordance rates provide insight into the possibility of a genetic background for a disease. These concordance rates are usually estimated within a frequentistic framework. Here we take a Bayesian approach. For rare diseases, estimation methods based on asymptotic theory cannot be applied due to very low cell probabilities. Moreover, a Bayesian approach allows a straightforward incorporation of prior information on disease prevalence coming from non-twin studies that is often available. An MCMC estimation procedure is tested using simulation and contrasted with frequentistic analyses. The Bayesian method is able to include prior information on both concordance rates and prevalence rates at the same time and is illustrated using twin data on cleft lip and rheumatoid arthritis.