Classifying clinical severity to help solve problems of stage migration in nonconcurrent comparisons of lung cancer therapy.

To compare the effects of stage migration in the "traditional" 3-stage TNM (tumor, node, metastasis) system with those in a new "expanded" 5-stage system, which has two additional stages for the poor prognostic groups, we used both systems to classify a cohort of 178 patients with primary lung cancer. To check for migrations, the stages in both systems were first assigned using only "old" technological information and were then reassigned using all the available "new" as well as old technological data. Although the 5-stage system had more migrations than the 3-stage system, survival rates were relatively unaffected for patients in the two new stages with poor prognosis. In both TNM staging patterns, the effects of stage migration on survival statistics were most impressive in the prognostically better (TNM I and II) stages. A solution to the migration problem is offered by the "clinical severity" (CS) staging system. Like the expanded TNM system, the CS system has 5 stages and a sharp prognostic gradient among stages. The CS system, however, had fewer technology-induced stage migrations than either TNM system, and the migrations had no substantial impact on stage-specific survival results. The excellent prognostic discrimination and secular stability of the CS system make it superior to the TNM system for comparing treatment results from different eras, especially for patients with stage I and II disease.