Risk-Stratified and Personalized Cancer Screening with Bidirectional Multistate Markov Models

Objectif

Multistate Markov Models (MMM) are powerful tools for analyzing the progression of diseases through different stages, particularly in oncology. Recent studies have modeled stage transition probabilities as functions of multimodal, time-dependent biomarkers to enable predictions based on histopathological features. However, current MMM approaches often overlook the reverse relationship—that is, the influence of the disease stage on the evolution of time-dependent biomarkers. At worst, this omission leads to misspecification bias, at best we miss the opportunity of a better modeling. In this proposal, I aim to address this gap by developing MMM methodologies that explicitly account for the bidirectional relationship between stage progression and associated time-dependent covariates, I call them bidirectional-MMM (bi-MMM).

This research is particularly relevant in cancer screening contexts, where a patient's cancer stage influences treatment decisions and, consequently, the trajectory of their biomarkers, which might be high-dimensional and missing at random. A more accurate representation of cancer progression enables the design of personalized screening protocols, thereby improving the harm-to-benefit ratio and reducing the risk of overdiagnosis –when a patient is heavily treated for a cancer that would not have caused symptoms or harm during their lifetime. Overdiagnosis is particularly prevalent in prostate cancer screening compared to other cancer screening programs, even though prostate cancer is the most common cancer among men in Western countries.
The supervisors of this project, Prof. Auvinen and Prof. Nevalainen, are currently leading an innovative prostate cancer screening initiative at Tampere University, in alignment with European health policy goals. This initiative is the primary motivation and data source for the project. A secondment in France is also planned with Prof. Proust-Lima, a leading expert in MMM, to enhance the project's impact.