Making Inference about a Biomarker by Using Information from Different Biomarkers in Time-Dependent ROC Estimation for Censored Data

Objective: Time-dependent reciever operating characteristics (ROC) curves are statistical methods which can be used when the specified outcome is an event which can take place at any time after the diagnostic test has been measured and may be right censored. This work presents an approach for making inference related to the performance of prognostic biomarker which can be measured from smaller number of patients, by borrowing information from the other biomarker(s) which can be measured from larger number of patients for right censored survival data. Material and Methods: Simulation studies were performed to see the performance of the proposed modification. We evaluated estimators related to the time dependent ROC function and the area under the curve (AUC) in terms of efficiency and unbiasedness to see whether proposed modification provides benefit over the original method. Results: It is observed that proposed approach yielded smaller bias, mean square error and standard deviation values for most scenarios in the simulation studies. Conclusion: The proposed approach, which combines information from different samples with different biomarkers may be useful to make inference related to the biomarker of interest which is measured from sample with a smaller size.