PCN58 FOUR-STATE MARKOV MODEL FOR IMMUNO-ONCOLOGIC THERAPIES

Abstract

Immuno-oncologic (IO) therapies have the potential to provide long-term survival benefits via a cured proportion. Traditional modelling techniques for oncologic technologies use three states. This deprives models of an ability to evaluate outcomes for those who experience cure separately. Cured individuals are expected to experience survival, cost and utility profiles like the general population. To not accurately represent them could undervalue IO therapies and inappropriately impact decision making. The objective of this project was to derive survival profiles to accurately quantify the cured proportion and the movements of patients around a state transition model. A four-state markov model was developed with health states: cured, pre-progressed uncured, progressed and death. An explicit cure state was included to better represent the heterogenous response seen with IO therapies. It is assumed that a cured patient cannot progress and so experience mortality at a general population rate. Patients in the uncured state can exit according to a parametric model on time to end of pre-progressed survival. On transition out of the uncured state a proportion move to the death health state at a rate defined by a logistic regression; the remainder proceed to a post progression state. Patients in the post-progression state experience a new parametric model of overall survival conditional upon their time in the post-progression state. All parametric parts of models were in addition to the cumulative hazard predicted per patient over follow-up conditional upon age, sex and nationality using lifetables. The modelling methods presented here maintain the mechanistic link between progression and mortality which are missing from traditional oncologic modelling. With the advancement of oncology and emergence of IO therapies the four-state model provides an option for quantification of outcome differences between cured and uncured patients and an accurate representation of the whole heterogenous population.