Estimating cumulative treatment effects in the presence of nonproportional hazards.

摘要: Often in medical studies of time to an event, the treatment effect is not constant over time. In context Cox regression modeling, most frequent solution apply a model that assumes either piecewise or varies smoothly time, i.e., nonproportional hazards model. This approach has at least two major limitations. First, it generally difficult assess whether parametric form chosen for correct. Second, presence hazards, investigators are usually more interested cumulative than instantaneous (e.g., determining if and when survival functions cross). Therefore, we propose estimator aggregate hazards. Our based on treatment-specific baseline estimated under stratified No functional nonproportionality need be assumed. Asymptotic properties proposed estimators derived, finite-sample assessed simulation studies. Pointwise simultaneous confidence bands can computed. The method applied data from national organ failure registry.