ABSTRACT
In drug development, phase III clinical trials are crucial to demonstrate the effectiveness of investigational treatment. However, they usually are extremely long in study duration and expensive in cost. To mitigate the potential risk of negative outcomes, sponsors have developed various metrics and approaches to predict the phase III study results through early phase clinical trials data. Survival outcomes, e.g., progression-free survival or overall survival, are commonly used primary endpoints in many disease areas, especially in oncology. The Cox proportional hazards model is commonly used for survival regression problems when incorporating patient information. Incorporating patient’s subject-level data offers the potential to make the survival inference more informative and the predicted survival results more accurate. As more machine learning methods become available, there have been many alternatives to some of these stringent parametric assumptions.
