CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Robust inference for non-inferiority studies Authors:  Elena Bortolato - University of Padova (Italy)
Laura Ventura - University of Padova (Italy) [presenting]
Abstract: Nowadays, the goal of many studies is to determine if new therapies have equivalent or non-inferior efficacies to the ones currently in use. These studies are called equivalence and non-inferiority studies, and the statistical methods for their analysis require simple modifications to the traditional hypotheses testing framework. The simplest and most widely used approach to test equivalence or non-inferiority is the two one-sided test (TOST) procedure, which evolves around the use of likelihood methods for testing the comparison of parameters like means, odds ratios, hazard ratios, etc. However, it is well-known that for model misspecifications or in the presence of influential observations, likelihood methods are highly unstable in many applications and to overcome this drawback, the theory of robust unbiased estimating equations may be usefully considered. The aim of this contribution is to discuss the use of robust unbiased estimating equations in the context of non-inferiority studies. In particular, we will resort on robust confidence distributions for the scalar parameter of interest, which allow deriving not only confidence intervals and p-values, but also suitable robust measures of evidence and performing robust sensitivity analysis to the preliminary choice of inferiority and equivalence margins.