Using the Additive Hazards Model with Two-Phase Sampling in Atherosclerosis Risk in Community Study

• Activity Number: 483
• Type: Topic Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Jie Kate Hu* and Kwun Chuen Gary Chan and Norman Breslow and David Couper
• Companies: The Climate Corporation and University of Washington and University of Washington and The University of North Carolina at Chapel Hill
• Keywords: Two-phase Sampling ; Risk Prediction ; Additive Hazards Model ; Coronary Heart Disease ; Calibration ; Biomarker

Abstract

Atherosclerosis Risk in Communities (ARIC) study used stratified sampling to select participants within the main cohort for biomarker studies of the risk of Coronary Heart Disease (CHD). Previous studies showed one inflammation marker high-sensitivity C-reactive protein (hs-CRP) was an independent predictor of increased risk of CHD and this biomarker acted additively on the baseline risks of CHD. To make efficient use of the data and perform individual risk prediction, we developed a new method for fitting additive hazards model to two-phase sampling data. Our method jointly estimated the regression coefficients and the baseline hazard function in the model so that we are able to make risk profiles for individuals with different levels of hs-CRPs. Our method further improved inference and prediction precision by calibration techniques. These risk profiles made upon the standard risk factors and the new biomarker information that is only available for selected samples may help physicians to discover new patients at high risk of CHD. Our new tools are not restricted to a particular two-phase study and they were implemented in the R software for immediate use.