The prognostic value of the modifications of the Dukes' C class of colorectal cancer. An analysis of the NSABP clinical trials.

This study was carried out in an effort to resolve the dilemma created by three proposed modifications of the Dukes' C class of colorectal cancer. Each modification is based on a separate prognostic discriminant characterized by: the level of histologically positive nodes, the depth of tumor penetration, and the number of histologically positive nodes. Data were derived from 844 patients with Dukes' C lesions randomized into two prospective clinical trials of the NSABP; the mean time on study was 41 months. Analysis of the three modifications as independent variables without regard for possible confounding effects disclosed that each had a highly significant predictive capacity. When each discriminant was examined, this time adjusting for the contribution of the other two discriminants, the effect attributable to the level of positive nodes was markedly attenuated. Thus, the level of positive nodes provided little information over and above that of depth of tumor penetration and the number of positive nodes. Of the two latter discriminants, although both were significant predictors of survival, the number of positive nodes appeared to be the strongest factor. Using both depth of penetration and the number of positive nodes, a unique Dukes' C subset of patients could be identified with a prognosis at least as good as Dukes' B lesions; this group was characterized by partial tumor penetration and the presence of 1-4 positive nodes. It is concluded that both depth of penetration and the number of positive nodes represent appropriate modifications of the initial Dukes scheme, and one discriminant should not be used to the exclusion of the other. The data raise serious doubts relative to the propriety of newly proposed TNM classification schemes that fail to utilize the number of positive nodes as a predictive discriminant.