Bayesian thinking on its way into medical statistics?

Bayesian statistical analysis is a paradigm quite different from traditional statistical inference. We wanted to show the usefulness of this approach for some medical problems. We started with Bayes¿ equation as it is used for estimating the probability of illness based on a specific laboratory test. We also looked into a recent Cochrane report on mammography that accepted two studies as valid and five
others as biased. In comparison we used examples of clinical trials from other areas that have been misinterpreted by the use of a traditional statistical approach only. We found that by taking into account our prior beliefs about the likely effects on breast cancer mortality of routine radiological screening programmes, the new data fit well into an estimate of a 5% mortality reduction with a 77% chance that there is a positive effect of screening. Bayesian statistics is helpful in making decisions on the basis of experimental evidence by taking into account our prior knowledge, whereas p-values in traditional statistics only give information on how often we will end up with a false positive conclusion in the long run.