Abstract: With the completion of the Human Genome Project, we are able to identify the disease-related targets and apply them to disease treatment by developing targeted drugs. Hence, targeted clinical trials are becoming increasingly popular in recent years. However, due to the imperfect diagnostic techniques the patients enrolled in targeted clinical trials with positive result might not all truly have the targets. To include patients with false positive results would result in the bias of treatment effects. This paper analyses the influence of the accuracy of diagnostic method in targeted clinical trials and figures out the bias of treatment effects. Then a new statistical method which incorporated EM algorithm and parametric bootstrap would be introduced to give an unbiased estimation of the effect size. The method provides an adequate methodological support for the targeted clinical trials.

Key words: Targeted clinical trial, Bias, EM algorithm, Parametric bootstrap