An Application of the Best Linear Unbiased Estimators (BLUE) for Missing Biomarkers Values in the Analysis of Repeated Measurements in Oncology Clinical Trials

• Activity Number: 646
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Santosh Sutradhar*+ and D. Das Purkayastha
• Companies: Novartis and Novartis
• Keywords: Biomarker ; missing values ; Best Linear Unbiased Estimator ; error-component ; repeated measurements ; oncology clinical trial

Abstract

In clinical trials, especially in oncology, diagnostic and prognostic biomarkers are collected in order to investigate possible correlation with a specific disease. However, it is very common to have missing values of the biomarkers during the follow-up.. In this presentation the Best Linear Unbiased Estimates (BLUE) of missing data will be applied. Unlike traditional methods of estimation of missing values assumption such as 'MAR' is not required in this approach. The unexplained variations between individual patients in a given time period and for a specific patient over time are captured using error-component modeling which systematically takes into account the mean effect of other related observed series of patient-specific data in the model. In a simple empirical framework the variance-covariance matrix of the residuals is estimated after appropriate transformations. The BLUE estimates of the missing observations will be obtained by minimizing the trace of the variance-covariance matrix. Thus, this method allows obtaining more accurate and precise estimates of missing assessments. The method will be applied to estimate the missing values of certain biomarkers.