|
Abstract:
|
Increasing knowledge in cancer biology leads to more accurate diagnosis, prognosis assessment and development of targeted therapy. To maximize patients benefit in apply those achievements, an important issue is to identify predictive marker in treatment selection, i.e., to treat patients according to the level of the predictive marker. In this paper, we propose a graphic approach to determine the cutoff points for a continuous marker, according to the treatment effects in intervals of the marker value. Most important part is determined at which range the new treatment is no better than standard treatment/placebo. Relative issues in determining the optimal cutoff points are discussed. Finally, we apply the method in categorized biomarkers for treatment selection in two NCIC CTG lung cancer trials.
|
