Gene Expression Profiling versus TNM Classification

We appreciate the opportunity to respond to the letter from Kivelä and Kujala regarding our recent article that prospectively evaluated a gene expression profile (GEP) prognostic test for uveal melanoma. One of us (J.W.H.) was a primary member of the committee that developed the current 7th edition American Joint Committee on Cancer TNM system for uveal melanoma; thus, we are quite knowledgeable of this system and how it is applied. We agree that the clinical TNM system stratifies patients with a particular cancer into multiple categories based on tumor size, location, and extent, and then arranges those categories into stages that are relatively homogeneous with respect to prognosis. A critical fact that Kivelä and Kujala failed to mention, however, is that these stage groupings are determined by retrospective analysis of large numbers of patients with known outcome and, thus, are in constant evolution as more data are collected. In contrast, the purpose of our study was to analyze the prospective predictive accuracy of the GEP test without retrospective manipulation of prognostic categories, which is the everyday situation in which clinicians use the test.