Optimal dynamic treatment regimes with survival endpoints: introducing DWSurv in the R package DTRreg

Abstract

Precision medicine is an approach to health care in which treatment decisions are tailored to patient-level information. Statistical methods for the estimation of dynamic treatment regimes (DTRs) allow to uncover a sequence of personalized treatment rules for patients with chronic diseases. Of particular interest is the identification of an optimal DTR, that is, the sequence of treatment rules that yields the best expected outcome. This is a challenging task, especially when the outcome is a survival time subject to right censoring or when available data are from observational studies. Dynamic weighted survival modelling (DWSurv) has been demonstrated to be theoretically robust and is accessible to users. We describe its implementation using the DWSurv function in the R package DTRreg. We review on the theory underlying DWSurv and demonstrate its use with hypothetical, and real-life inspired, simulated data sets.