Survival-Related Prognostic Threshold on Quantitative Biomarkers

• Activity Number: 294
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Mental Health Statistics Section
• Author(s): Xinhua Liu*+ and Zhezhen Jin
• Companies: Columbia University and Columbia University
• Keywords: Biomarkers ; Prognosis ; Sensitivity ; Specificity ; Survival ; Threshold

Abstract

In biomedical research and practice, quantitative biomarkers are often used for diagnostic or prognostic purposes, with a cut point established on the measurement to aid binary classification. When prognosis is regarding survival time, single cut point may not be very informative. It is also challenging to select threshold using the data from a follow up study where the survival time is subject to random censoring due to dropouts. Using survival time dependent sensitivity and specificity, we extend classification accuracy based objective function to allow for survival dependent threshold. To estimate optimal threshold for a range of survival rate we adopt a non-parametric procedure, which produces satisfactory result in a simulation study. We also apply the method to estimate survival related threshold on the biomarkers of serum albumin and bilirubin for prognosis of primary biliary cirrhosis, along with the associated sensitivity and specificity. The profile shows variation in the estimated thresholds for a range of survival rate, which may help clinicians picture the prognostic role of the threshold.