Abstract:
|
Receiver operating characteristic (ROC) curves are widely used to identify the responder thresholds that best predict classification based on an external criterion. Different approaches such as minimizing the distance to the ideal point (0, 1) on ROC curve, Youden index, and maximizing accuracy area are used to select the optimal threshold that characterizes a meaningful improvement in a clinical trial setting. For example, in chronic pain studies, the threshold that best distinguishes "Improvement" from "No Improvement" based on patient global impression of change criterion is identified for the 11-point pain intensity numerical rating scale ranging from 0=no pain to 10=worst possible pain. An evaluation of responder thresholds for pain using the ROC analysis has exposed issues related to the use of change from baseline versus percent change from baseline. Stability of the thresholds for change from baseline versus percent change from baseline is examined by sequential monitoring of the accumulating data in a simulated clinical trial. The ROC methodology using percent change from baseline is comparatively more robust than using change from baseline in the pain therapeutic area.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.