Inference on Competing Risks under Non-Proportional Hazards

0506832 - ÚI 2020 RIV BE eng A - Abstract
Valenta, Zdeněk - Skrabaka, D. - Owczarek, A. J. - Kolonko, A. - Król, R. - Więcek, A. - Ziaja, J.
Inference on Competing Risks under Non-Proportional Hazards.
40th Annual Conference of the International Society for Clinical Biostatistics. Book of abstracts. Leuven: ISCB, 2019. s. 364-365. ISBN 978-94-6165-287-4.
[Annual Conference of the International Society for Clinical Biostatistics /40./. 14.07.2019-18.07.2019, Leuven]
Institutional support: RVO:67985807
Keywords : Competing Risks * Model Diagnostics * Schoenfeld Residuals * Cumulative Incidence * Cause-Specific Hazards * Kidney Transplantation
OECD category: Statistics and probability

IN: 40th Annual Conference of the International Society for Clinical Biostatistics. Book of abstracts. Leuven: ISCB, 2019. s. 364-365. ISBN 978-94-6165-287-4. [Annual Conference of the International Society for Clinical Biostatistics /40./. 14.07.2019-18.07.2019, Leuven]. ABSTRACT: We analyse data on Silesian patients after kidney transplantation under competing events scenario where time to death and time to graft rejection are considered as absorbing (terminal) events. Objectives: To use model diagnostics, incl. Schoenfeld residuals, in identifying violations of proportionality assumption under the framework of subdistribution and cause-specific hazards, respectively. Methods: We use Fine-Gray model for subdistribution hazards (cumulative incidence, CI) implemented in the ‘cmprsk’ package and ‘survival’ library. Under the cause-specific hazards (CSH) scenario we use the Cox PH model and Gray’s time-varying coefficients (TVC) model [JASA 1992] and available model diagnostics. Results: We show that violation of proportional subdistribution hazards assumption may be conveniently identified using residual diagnostics and properly accounted for by involving time interactions with the corresponding model predictors. We observed that non-proportional effects on cumulative incidence would not necessarily translate in those on cause-specific hazards. Rather, they appeared to have attenuated under cause-specific hazards scenario.
Permanent Link: http://hdl.handle.net/11104/0297986