Tekrarlı Çapraz Geçişli Deney Uygulamalarında Kişi Sapan Değeri Bulmak için Kullanılan Hotelling T2 Yönteminin Performansının Değerlendirilmesi

Abstract

Detection of outlying subject in bioavailability/bioequivalence studies is an important component of data screening process before statistical analysis. Several methods have been proposed for detecting outliers based on standard crossover design (2x2) used in bioequivalence. Hotelling T2 test is one of these methods. Another study design in bioequivalence is replicated crossover design used for high variability drugs. Detection of outlying subject for this design was considered only in the way of adjusting Hotelling T2 to fully replicated crossover design and has been found as powerful as 2x2. In this study, we have applied and extended Hotelling T2 test to partial and fully replicated crossover design for two types of outlying subject. For the first type, a subject has extreme response to only one formulation and for the second type, a subject has extreme responses to both formulations. We have simulated 1000 data sets that have 0.5 correlation between responses of one subject, 10%-40% intrasubject variability (CV) with various sample sizes and means. According to the simulation results, it is concluded that it is always a right choice to use Hotelling T2 for determination of a first type outlying subject or of a subject that has too much extreme responses to both formulations in replicated crossover design. However, considering the use of replicated crossover design when CV is greater than 30%, since CV has a masking effect for determination of a second type outlying subject, the right choice will be to use another testing procedure different than Hotelling T2.