A1155
Title: Design on three-arm noninferiority trials using parametric and semiparametric methods for censored survival data
Authors: Yi-Kang Tseng - National Central University (Taiwan) [presenting]
Abstract: A gold standard design for three-arm noninferiority trials, which includes an experimental group, a reference group and a placebo group using parametric and semiparametric models, was investigated for censored survival outcomes. In the literature so far, the sample size determination for a three-arm design with survival endpoints is limited. Therefore, to complement the literature, the purpose is to develop a statistical method to calculate the sample size for three-arm inferiority trials. Four parametric models (lognormal, Gompertz, Rayleigh, and log-logistic) and three semiparametric models (Cox, accelerated failure time and transformation models) were considered for survival data. The three groups' minimum required sample size and testing power with optimal allocation were considered the performance criteria for both parametric and semiparametric methods. Simulation studies were conducted to find out the performance of all methods. In particular, the impact of model misspecification was studied. The proposed procedure was finally applied to bladder cancer data.
