CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Evaluating the accuracy of prognostic biomarkers in the presence of external information Authors:  Maria Xose Rodriguez-Alvarez - BCAM, Basque Center for Applied Mathematics (Spain) [presenting]
Thomas Kneib - University of Goettingen (Germany)
Abstract: Prior to using a diagnostic biomarker in a routine clinical setting, the rigorous evaluation of its diagnostic accuracy is essential. The Receiver Operating Characteristic (ROC) curve is the measure of accuracy most widely used for continuous biomarkers. However, the possible impact of extra information about the patient (or even the environment) on diagnostic accuracy also needs to be assessed. In addition, in many circumstances the aim of a study may involve prognosis rather than diagnosis. The main difference between diagnostic and prognostic biomarkers is that, with prognostic biomarkers, a time dimension is involved. This is the case of survival studies, where the status of an individual varies with time (e.g, death and alive). To assess the accuracy of continuous prognostic biomarkers for time-dependent disease outcomes, time-dependent extensions of the ROC curve have been proposed. A novel penalized likelihood-based estimator of the cumulative-dynamic time-dependent ROC curve is presented. The proposal allows to account for the possible modifying effect of covariates on the accuracy of the biomarker. The validity of the approach is supported by simulations, and applied to the evaluation of biomarkers for early prognosis of death after discharge in patients who suffered an acute coronary syndrome.