Statistical Methods to Build a Prognostic Model of Residual Risk for Major Cardiovascular Events in Coronary Patients

• Activity Number: 80
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Rana Fayyad*+ and Ingar Holme and Sarah Young and Chuan-Chuan Wun
• Companies: Pfizer and Center of Preventive Medicine, Oslo University Hospital, Ullevål and Pfizer and Pfizer
• Keywords: Royston's D-statistic ; Receiver Operating Characteristic ; Hosmer-Lemeshow tests ; Prognostic model

Abstract

Patients at high risk for a major cardiovascular event (MCVE) are treated with drugs that reduce their risk of a future event, however, many patients still develop events, and it is unclear whether additional lowering of atherogenic lipoproteins further reduces risk. We addressed this issue using a prognostic analysis approach in patients from two large cardiovascular/lipid-lowering studies. Using a training sample from one of the studies, we identified the relationship between risk factors, lipids, patient characteristics and the risk of MCVEs by Cox proportional hazards regression analysis. The model identified from the training sample was tested in the remaining populations with respect to calibration by Hosmer-Lemeshow tests and with respect to prognostic ability calculated by Receiver Operating Characteristics and Royston's D-statistic. A web based calculator was generated to assess 5 year risk of MCVE based on the identified prognostic variables.