A latent health process model for dynamic risk prediction with application to an HPV study

Individuals with an elevated risk of disease (ex., HPV infection) are often monitored with increased screening along with the collection of additional biomarkers to improve risk assessment at each screening visit. These individuals may experience disease progression, but may also transition back to a low-risk status (ex., clearing an infection). However, standard joint modeling approaches for dynamic risk prediction for these at-risk individuals have several limitations. First, it is challenging to flexibly link a noisy longitudinal biomarker to both interval-censored disease progression and remission without restrictive assumptions on the association structure and survival process. Furthermore, for non-terminal events, standard joint modeling approaches often ignore biomarkers collected after the event, meaning biomarker models do not account for the event status. We propose to model the health status of at-risk individuals as an underlying stochastic process constrained between two thresholds: an upcrossing representing a transition to low-risk status and a downcrossing representing disease progression. We link the change in the health process to a longitudinal biomarker whose trajectory can change based on the event. Simulations and a real data set demonstrate the flexibility of our proposed approach, showing substantial improvements in risk estimates by capturing an event's impact on the biomarker trajectory and utilizing two thresholds to account for the competing risks of disease progression and remission.