A mathematical model was developed to investigate the source and effect of bias in data from dental clinical trials when the status of some teeth is misclassified. Methods of obtaining unbiased estimates of the true attack rate and other parameters are suggested.