Simulation Studies on the Analysis of Interval-Censored Survival Data with Differential Assessment Times

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Isao Yokota*+ and Yutaka Matsuyama and Yasuo Ohashi
• Companies: University of Tokyo and University of Tokyo and University of Tokyo
• Keywords: differential assessment ; grouped survival analysis ; interval-censored data ; multiple imputation method ; weighting

Abstract

In standard survival analysis, time-to-event of interest is assumed to be observed exactly or right censored, meaning that it is only known the event occurred after the last observation. In some situations, however, it may only be known to have occurred within an interval of time. For example, in a randomized trial of adjuvant interferon after curative resection of HCV-related hepatocellular carcinoma, patients are seen at prescheduled visits but the event of interest may occur between visits. Additionally, there may be a possibility of differential assessment intervals between groups, because control group will receive no treatment before recurrences. In this talk, we propose a grouped survival analysis, in which event times are equally grouped into broader time interval than planned assessment one. We propose to impute exact event times from interval censored observations, and apply usual analysis methods such as the log-rank test and Cox model. A multiple imputation method is used to appropriately account for variability. Simulation studies showed our approach maintained the type I error and gave unbiased treatment effect, compared with other methods for interval-censored data.