EQUIVALENCE IN TEST ASSAY METHOD COMPARISONS FOR THE REPEATED-MEASURE, MATCHED-PAIR DESIGN IN MEDICAL DEVICE STUDIES: STATISTICAL CONSIDERATIONS1*

摘要

In medical device clinical studies, including therapeutic device and in vitro test assay method studies, the investigator is frequently interested in demonstrating the equivalence of clinical response, generally in continuous measurements, between a standard assay method and a new assay method over various occasions or times. The new assay method may be less invasive or more convenient or cheaper to use than the standard assay method. In this paper, several statistical approaches are discussed, including various repeated-measure regression models, the simultaneous 95 confidence interval for paired mean differences derived from Hotelling's multivariateT2analysis for repeated-measure, paired data, repeatability and reproducibility studies, and concordance correlation coefficient.