A0410
Title: Improvement of midpoint imputation for estimation of median survival time for interval-censored time-to-event data
Authors: Yuki Nakagawa - Chugai Pharmaceutical Co., Ltd. (Japan) [presenting]
Takashi Sozu - Tokyo University of Science (Japan)
Abstract: Progression-free survival (PFS) is used to evaluate a treatment effect for patients with solid tumors in cancer clinical trials. The disease progression of the patients is typically determined by radiological testing at several scheduled tumor-assessment time points. This results in a discrepancy between the true progression time and the observed progression time. Considering the observed progression time as the true progression time, a biased PFS is obtained for some patients, and the estimated survival function from the Kaplan-Meier method is also biased. Although the midpoint imputation method replaces the interval-censored data with the midpoint data and it reduces the bias, it has an unrealistic assumption that several disease progressions occur at the same time point when several disease progressions are observed in the same tumor-assessment interval. We improved the midpoint imputation method, which replaces the interval-censored data with the equally spaced timepoint data based on the number of observed interval-censored data within the same tumour-assessment interval. We evaluated the bias, root mean square error of the median, and coverage probability of the 95% confidence interval of the proposed method. The proposed method provided higher performances compared with those of the midpoint imputation method.
