Statistical inferences of average bioequivalnece under crossover design with incomplete data

Abstract

Two-sequence and two-period (2 x 2) crossover design is the standard design used for the bioequivalence of test and reference drugs. In this thesis, average bioequivalence tests, sample size determination, and the missing data problem related to the 2 x 2 crossover design are handled, and the statistical methods used for different variance assumptions between the drugs are compared. First, 2 x 2 crossover design is introduced, and the related statistical tests about the period, sequence, treatment, and carryover effects are explored in Chapter 1. Average bioequivalence is introduced, and the general properties of the average bioequivalence model for the standard 2 x 2 design, including Two One-Sided Test (TOST), Anderson and Hauck``s test, and confidence interval approach, are reviewed in Chapter 2. As well as these, the power function of TOST and the appropriate sample size are examined. Second, general missing data problems are reviewed. The missing data patterns including monotone and non-monotone, and the missing data mechanisms, including Missing Completely At Random (MCAR), Missing At Random (MAR), and Not Missing At Random (NMAR), are explored. Also, the general methodologies for missing data problems, including complete case, available case, weighting, imputation, and model based approaches, are reviewed in Chapter 3. Third, the statistical procedures for analyzing the 2 x 2 crossover design in bioequivalence trials are discussed when some observations in the second period are missing . In a bioequivalence trial, it is not unusual to observe a patient either withdrawing from the trial for administrative reasons, or being dropped from the trial by the investigator because of one or more protocol violations (such as poor drug compliance, unauthorized use of concomitant medication, etc). Sometimes, a patient is dropped because of adverse experiences or unsatisfactory therapeutic responses. For these and other reasons, values at one or both periods of the c...