Prognostic factor identification from continuous variable by improvement of log-rank test

Abstract

In this research, a prognostic score calculation, “Iterative Patients Partitioning (IPP)” is suggested as an effective method to identify prognostic factor. Prognostic factor can be used in patient survival prediction, drug targets, and disease mechanism study. Due to the limited robustness and sensitivity in previous studies, novel method for evaluating prognostic factor is indispensable. A critical limitation in previous studies is that patients were stratified into different risk groups based on specific threshold of continuous variable. To overcome the specific threshold problem, the Iterative Patients Partitioning that is inspired by the Receiver Operating Characteristic (ROC) curve is presented to consider various cases of risk group stratification based on different thresholds in prognostic factor evaluation. Therefore, the Iterative Patients Partitioning can be applied in the precise identification of prognostic factors.