|
Abstract:
|
Interpretation of hazard ratios can be difficult, in particular when the outcome is subject to competing risks. For example, in prostate cancer studies, prostate cancer-specific death is subject to the competing risk of death from other causes. The difference in restricted mean times lost (RMTL) can be used as an alternative method to quantify effect sizes. It has a clinically relevant interpretation as the expected number of life years lost due to the cause of interest. We describe a simple doubly robust estimator for the difference in RMTL for competing risks data. The approach also adjusts for potential confounders. We construct the doubly robust estimator by combining an inverse-probability weighting estimator with a direct adjustment model, both estimated via pseudo-observations. We assessed the performance and robustness of the proposed approach by a simulation study under multiple scenarios. We illustrate the method by comparing the RMTL for prostate cancer-specific death between two groups of age at diagnosis in men diagnosed with de novo metastatic prostate cancer with data from the Surveillance, Epidemiology, and End Results registry.
|
