Abstract: A counting process procedure and its bootstrap version were used to study a two-stage model with time-dependent covariates for latent period of cancer.The transition intensity process was directly semi-parameterized by multiplication,and the maximum partial likelihood estimators,the confidence interval and hypothesis testing were performed by the two methods at the same time. As an example, the model was applied to analyse a set of 12 year and 6 run screening data of cervical cancer in Jingan county, Jiangxi Province, China.