The Bayesian analysis was used in this study to investigate the diagnostic value of the bronchial provocation test with methacholine in patients with asthma. The best cutoff value of accumulated concentration of methacholine administered that caused a 20 percent fall in FEV1 post-saline (PC20) in our sample, determined with a receiver operator characteristic curve, was 15 mg/ml. The interval security of the test was established by a pretest probability between 0.16 and 0.87 and the best test results were obtained when pretest probability was 0.48. The positive final diagnostic gain of the test was maximal at this pretest probability. We conclude that the application of Bayes' theorem, considering the pretest probability of asthma and the sensitivity and specificity of the individual PC20 obtained, increases the accuracy of the bronchial provocation test with methacholine in the diagnosis of asthma.