Abstract: Sometimes assessment of equivalence / noninferiority between two new drugs and medical devices involves comparisons of the response rates between paired binary endpoints in clinical trials.Statistical procedures have been developed for inferring equivalence / noninferiority often concentrated on comparing the rates of two independent binary responses in the past few years.Procedures for assessing equivalence/ noninferiority between the paired binary endpoints have not been well studied.In this paper, we introduce the determination of sample size and hypothesis testing of equivalence / noninferiority for paired binary data based on restricted maximum likelihood estimation.An equivalence / noninferiority clinical trial of ultrasound diagnostic device is used to illustrate the proposed methods.Discussions on some important issues are provided to help understand and conduct these trials.

Key words: sample size, hypothesis testing, equivalence, noninferiority, matched-pair design, binary response, paired binary data