Optimal Normality Transformation Methods for the Analysis of Biomarkers

• Activity Number: 410
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Kelly Zou*+ and Ching-Ray Yu and Martin O. Carlsson and Ye Tan
• Companies: Pfizer and Pfizer and Pfizer and Pfizer
• Keywords: Biomarker ; Normality Transformation ; Normality Test ; Sensitivity ; Specificity ; Receiver Operating Characteristic Curve

Abstract

Biomarkers tend to have skewed distributions. Therefore, when analyzing such data parametrically, non-normality can be a frequently encountered challenge and must be taken into considerations. In this research, an optimal normality transformation (e.g., the Box-Cox power transformation) is applied. Various normality tests (e.g., Shapiro-Wilk) are conducted both before and after the transformation. Furthermore, the accuracy of the biomarkers is evaluated based on sensitivity, specificity, and receiver operating characteristic curve analysis. In Monte-Carlo simulations, both normally and non-normally data are generated. The performances of normality transformations are examined. These methods are illustrated using cancer marker data.