Identification of the minimum non-inferior dose in a three-arm non-inferiority trial

Non-inferiority (NI) trials have gained recognition as an effective tool with which to search for substitutes for a standard treatment that is associated with certain undesirable features, such as adverse side-effects, an exorbitant cost, or an extremely complicated therapeutic regimen. Statistical methods have been developed for NI studies with multiple experimental treatments. However, experimental treatments sometimes represent a new treatment with various dose levels, and in such circumstances, identification of the minimum non-inferior dose is a reasonable objective of the trial. Thus, we examine several potential testing procedures with respect to this testing objective. A simulation study is conducted to evaluate their performance, and determination of the sample size is discussed. Two clinical examples are provided to illustrate the proposed method.