(Learn more about this study)
Initial CDAS Request Approval
May 12, 2020
Additive proportional hazard model for interval-censored data
Cox proportional hazards (PH) model is inarguably one of the most popular models for analyzing time-to-event data. Typically, it is assumed that the effect of predictors on hazard function is linear. In many cases, this assumption is not correct and may lead to biased inference. To offer more modeling flexibility, we propose to use additive PH models which allow effect of predictors to be nonlinear. In PLCO program, each participant is examined regularly for the status of several cancers (e.g., Prostate, Lung, Colorectal and Ovarian), and if the cancer is detected during one visit, it actually happened between the current and previous examination. Thus, the event time is interval censored. It is challenging to develop an easy-to-implement, accurate and efficient algorithm to fit the additive PH model with interval-censored data. Our goal is to develop such an algorithm and apply it to the PLCO data.
Develop an efficient algorithm to provide accurate estimate and reliable inference for additive proportional hazards model with interval-censored data
Use the proposed method to unmask possible nonlinear effect of predictor on hazard function.