p3state.msm: Analyzing Survival Data from an Illness-Death model

The second paper in the special issue at Journal of Statistical Software is by Luis Meira-Machado and Javier Roca-Pardinas and is about the package p3state.msm. This package allows non and semi-parametric estimation for three-state progressive or illness-death models subject to right-censoring. Much of the functionality of p3state.msm such as Cox-Markov and Cox-semi-Markov models is available elsewhere in mstate. The main niche of p3state.msm is the ability to estimate transition probabilities without making either Markov or semi-Markov assumptions applying the methodology of Meira-Machado et al (2006, Lifetime Data Analysis). Note that while the Aalen-Johansen estimator is consistent for state occupancy probabilities (Datta and Satten 2001 S&PL, Glidden 2002 Biometrics), this doesn't extend to general transition probabilities. Currently estimates are only available for the case of no covariates, though the authors state they are considering extensions to allow Cox-type models for such non-Markov models. Similarly, they are planning to allow a greater range of models, e.g. a progressive k state model.