Estimating multiple tumor transition rates based on data from survival- sacrifice experiments (Q1336360)

The authors consider a Markov illness-death process to describe the development of two types of tumors in an animal survival-sacrifice experiment. The model involves four unknown tumor transition rates and four unknown death rates. A semiparametric model is proposed where the tumor transition rates are assumed proportional to each other and can be described by a parametric function. Death rates are estimated nonparametrically. The authors show an ECM algorithm to estimate the parameters. The method is illustrated using the \(\text{ED}_{01}\) study, classifying tumors into liver tumors and non-liver tumors. It is shown how the effect of dosage and the independence of tumor incidence can be tested using likelihood ratio test statistics.