競合リスクが存在する場合の累積発生確率についての検定(Grayの検定)のサイズに対する検討

概要

論文の詳細を見る
Objective: The event of an individual may be one of several distinct types or causes. If the first event precludes the occurrence of other types of events under investigation, competing risks problems occur. Censoring due to such an event is generally not independent of the time to the event of interest. It is known that treating the events of the competing causes as censored observation will lead to a bias in the Kaplan-Meier estimate. It is shown by Freidlin and Korn that the cause-specific log-rank test does not maintain the nominal level of the test if the event times of the two types are correlated. The cumulative Incidence Function (CIF) approach is recommended in competing risks situation. The estimate of the CIF can be obtained by the method of Kalbfleisch et al., and the test of CIF is performed by Gray's test. We explore the size of Gray's test by simulation if the event times of the two types are correlated and there are censored observations.Method: We perform simulation study under various conditions. Results: It is found that Gray's test maintain the nominal level of the test under the conditions examined. Conclusions: It is thought Gray's test is useful in the presence of competing risks.
2012-10-00