From Models to Medicine: Statistical Methods in Cancer Survival

Advances in cancer research increasingly rely on sophisticated statistical methods that can accommodate the complexity, heterogeneity, and rapidly evolving nature of oncology data. As therapeutic innovations accelerate and precision medicine becomes central to clinical decision-making, the statistical community faces both unprecedented opportunities and novel methodological challenges. This session is especially timely given the rapid evolution of cancer therapies and the corresponding need for statistical tools that can capture their complexity. Modern immunotherapies, for example, are known for delayed treatment effects, non-proportional hazards, cure fractions, and survival curve crossing, all of which pose significant challenges for standard Cox models. Precision medicine has further transformed oncology by emphasizing the integration of genomic, proteomic, radiomic, imaging, and pathology-derived biomarkers into survival modeling frameworks, requiring advanced techniques to handle high-dimensional predictors, nonlinear functional forms, and interactions that influence prognosis and treatment response. At the same time, the proliferation of real-world data sources, including electronic health records, registries, claims, wearable devices, home-based sensors, and patient-reported outcomes, has created both opportunities and difficulties for survival analysis, including missingness, time-varying confounding, immortal time bias, and inconsistent follow-up. Cutting-edge statistical and machine learning approaches such as dynamic survival models, Bayesian joint models, causal inference frameworks, and machine-learning-augmented survival methods are increasingly critical for extracting insights from such data sources. Moreover, collaboration among statisticians, clinicians, translational researchers, and data scientists is essential to ensure that methodological innovations lead to clinical impact, and this session is intentionally structured to foster that interdisciplinary dialogue