Activity Number:
|
293
- Advances in the Analysis of Competing and Semi-Competing Risks Data in Medical Research
|
Type:
|
Invited
|
Date/Time:
|
Wednesday, August 5, 2020 : 10:00 AM to 11:50 AM
|
Sponsor:
|
Biometrics Section
|
Abstract #309423
|
|
Title:
|
Methods for Estimation and Inference of Cumulative Incidence with Missing Cause of Failure
|
Author(s):
|
Andrea Knezevic* and Sujata M Patil
|
Companies:
|
Memorial Sloan Kettering Cancer Center and Memorial Sloan Kettering Cancer Center
|
Keywords:
|
Competing risks survival analysis;
Missing data;
Multiple imputation
|
Abstract:
|
Accurate estimation of cancer survival is a challenge in the presence of considerable mortality due to other causes. Competing risks survival analysis offers a methodology to account for mortality due to other causes, but estimates can be unreliable if cause of death is unknown for a significant proportion of cases. It has been shown that the two most common analysis strategies used when cause of failure is unknown for some patients (complete case analysis and additional failure type) both lead to substantial bias in estimation of cumulative incidence. Imputation methods based on multiple imputation or inverse probability weighting have been proposed as alternatives, assuming cause of failure is missing at random. We explore various methods of dealing with missing cause of failure in the estimation and inference of cumulative incidence functions. The statistical properties of the methods are evaluated in simulation studies, and application of the methods is illustrated with real data wherein missingness is imposed on the cause of failure under various scenarios.
|
Authors who are presenting talks have a * after their name.