Challenges and considerations in non-inferiority trials: a narrative review from statisticians’ perspectives

Non-inferiority (NI) study is a popular randomized controlled trial design that aims to demonstrate whether a test treatment, considering its auxiliary benefits, is not unacceptably worse compared to a standard active control treatment. There is extensive work in the literature that discusses NI trials' merits, issues, and how certain clinical and statistical challenges can be addressed. Here, we are aimed to provide a narrative review of NI studies in terms of its design considerations, potential issues, and corresponding solutions from the perspectives of biostatisticians. We conducted a wide literature search on clinical and statistical methodology papers related to NI trials. In the "Fundamentals of NI study" section, we start from the formulation of the margin and the NI hypothesis test and then focus on the underlying two fundamental assumptions (the constancy assumption and assay sensitivity). We present experts and regulatory agencies' opinions on how certain statistical issues of NI studies are caused and how they could be addressed. We focus on key aspects of NI studies, which include formulations of an NI hypothesis test, definition of NI margins, determining historical evidence of the active control drug, checking assay sensitivity and constancy assumption, etc. We also briefly touch on topics such as comparisons between the fixed-margin method and the synthesis method for NI evaluation, analysis principle in presence of treatment non-adherence, Bayesian design of NI studies, and restricted mean survival time (RMST) as a measure for designing NI studies. Figures and examples are given throughout the article to better illustrate ideas. We believe that NI design, with its issues addressed by appropriate statistical and clinical considerations, still plays a pivotal role in clinical research by improving patients' experience and alleviating healthcare inequalities.