区間打切りデータがある場合の生存時間解析における要因効果推定方法についての性能評価

概要

論文の詳細を見る
In many clinical trials, survival analysis plays an important role when the time to event is a primary variable. However, in some cases, the event of interest cannot be observed exactly but instead can be only known to lie in an interval, called the censoring interval. Such data are said to be "interval-censored" data.As common practice, if the exact date of occurrence is not known due to interval-censoring, the date of examination when the event was first observed is most often treated as the exact time of event (right-point imputation), and the standard survival analysis are performed.In this paper, we compared the performances to estimate the regression coefficient in Cox proportional hazard model with a categorical covariate after the left-, mid-, and right-point imputation by simulation.Under the simulation settings considered here, the left-point imputation showed the smallest root mean squared error (RMSE) and the highest coverage probability in most cases. The right-point imputation showed the largest RMSE and the lowest coverage probability in some cases. It underestimated the covariate effect, especially under heavy right censoring.
国立保健医療科学院の論文