Using Logistic Model with Measurement Errors in Covariates to Assess Baseline Biomarkers Association with Treatment Response

• Activity Number: 238
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Dion Chen*+ and Jiandong Lu
• Companies: Centocor R&D, Inc. and Johnson & Johnson
• Keywords: Logisitc regression ; Measurement error ; Biomarker dada analysis ; Clinical response

Abstract

Logistic regression model is often used in assessing biomarkers association with treatment response. However, biomarkers may not be accurately ascertainable, because of measurement errors due to imperfect instruments and biological daily fluctuation. Considering the additive measurement model, one observes X*=X+error, instead of X. In this case, ignoring the measurement error and using X* as X in the estimation may results in substantial bias (Stefanski and Carroll, 1985), and the bias-adjusted estimate is not consistent, and in small sample size it could be over- or under-adjusted. The authors proposed two new estimates, which can be easily implemented in SAS and R. The finite sample properties of the bias-adjusted estimates have been studied with numerical simulation. The results show that the bias is reduced substantially. A real life example of biomarker analysis is presented.