Analysis of multistage disease using semi-Markov models - A semi-parametric approach

Semi-Markov processes have gained popularity as multistage disease models describing a patient‚ as a history of events over time. In this project we extend the indirect approach of Lo et al. (2009) used to model major cardiovascular endpoints of the LIPID trial by relaxing the fully parametric formulation.

The technique does not require the proportional hazard assumption to hold at each step, allows semi-parametric estimation of clinically relevant quantities such as the hazard ratio, the overall survival or absolute benefit from treatment. Applications to various trials and cohort data are envisaged.

Related unit(s): Statistics