Receiver Operating Characteristic Curve Method for Evaluating Surrogate Endpoints in a Left Ventricular Assist Device (LVAD) Trial

• Activity Number: 637
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Hong Wang*+ and Kevin Najarian
• Companies: HeartWare and HeartWare
• Keywords: Medical Device ; Receiver Operating Characteristic ; C statistics ; Left Ventricular Assist Device System ; Surrogate Endpoint

Abstract

The receiver operating characteristic (ROC) measures the relationship between sensitivity and specificity. First developed by engineers, it is now used to model data in a number of disciplines, such as, psychology, medicine, radiology and biometrics. The ROC curve was considered to assess the effect of blood pressure (mean arterial pressure or MAP) on the clinical outcomes of stroke (ICVA and HCVA) in a left ventricular assist device (LVAD) trial for patients eligible for heart transplant. The area under the ROC curve (or c-statistic), generally considered a measure of the predictive accuracy of a logistic regression model, was calculated with the associated 95% confidence interval. The optimal point estimate for the c-statistic was determined by using the probability cutoff method. The likelihood reduction factor (LRF) and proportion of information gain (PIG) were used to confirm the correlation of MAP and stroke.