Survival Analysis

Treatment effects will be estimated using survival probability differences and restricted mean survival time differences over 1-year and 5-year windows. The protocol acknowledges that cloning and inverse probability weighting reduce but do not eliminate confounding. The target sample size is at least 600 participants to satisfy both survival and logistic regression modelling requirements. The study assesses and reports assumptions for its survival models. Multistate survival regression models transitions among diagnosis, treatment, recurrence or remission, and death. Artificial censoring is handled with inverse probability of censoring weights because censoring is informative. The study applies Cox proportional hazards regression and parametric survival methods.