A0546
Title: Optimal cut-points for screening for pre-clinical disease based on various criteria
Authors: Cuiling Wang - Albert Einstein College of Medicine (United States) [presenting]
Abstract: The selection of a cut-point for a marker of disease to identify those at high risk of developing the disease in the future is crucial in clinical practice and research. Although optimal thresholds based on various criteria can be obtained through time-dependent receiver operating characteristic (ROC) analysis, the properties of the optimal cut-points are not well known. Recently, the properties and estimation methods for the time-dependent optimal cuts based on the Youdens index have been studied. However, optimal cuts based on other criteria are rarely studied. We investigate the properties of the time-dependent optimal cut-points based on various criteria including setting the target level for sensitivity or specificity alone, the weighted sum of sensitivity and specificity which includes Youden's index as a special case, and the average overall cost. For the weighted sum and average overall cost criteria, we provide formulae to estimate the optimal cuts using survival models. The methods are applied to screening for pre-clinical Alzheimer's dementia using a well-established memory test in the Einstein Aging Study. Simulation studies are performed to evaluate the direct estimates as well as those obtained from time-dependent ROC analysis.
