A Cox Regression Model with Isotonic Regressors with Change-Point Problems: Addressing a Clinical Question

• Activity Number: 688
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): D. Das Purkayastha*+
• Companies: Novartis Pharmaceuticals
• Keywords: Isotonicity ; monotonicity ; Cox regression model ; change-point probems ; isotonic regressors

Abstract

Some clinical predictors may or may not have monotonic properties. They may have increasing or decreasing pattern or they may remain steady during a period of observation time. A clinical question is raised on how such increasing or decreasing or constant pattern of variations may have effect on survival or on clinical outcomes independently during a period of time. Isotonicity is called the simple or partial ordered monotonicity mathematically. Isotonic regression is performed under ordered monotonic restrictions criteria. Either increasing or decreasing or constant variation of a clinical predicator or of a prognostic factor generates a change-point problem at the transition from increasing order to decreasing order or vice versa or from the steady condition to either increasing or decreasing pattern of clinical predictors. The primary thrust of this paper is to address such clinical questions using a Cox Regression Model with isotonic regressors with such change-point problems. In a case study with multiple myeloma the impact of such change-point patterns of clinically significant predictors such as white blood corpuscles on patients median survival time have been explored.