Background: In 1986, Jass and colleagues claimed to have improved on Dukes' classification of prognosis for carcinoma of the colon and rectum. To have clinical relevance, such results should be reproducible and confirmed by other institutions.
Study design: Retrospective clinicopathologic study of 312 carcinomas of the colon and rectum to determine whether or not Jass' classification is superior to that of Dukes' as assessed by their relative reproducibility and prognostic significance.
Results: Dukes' classification had excellent intraobserver and interobserver reproducibility (kappa values of 0.86 and 0.93, respectively). In contrast, the reproducibility of variables assessed by Jass showed only slight to fair agreement (lymphocytic infiltration: intraobserver and interobserver kappa values of 0.08 and 0.05, respectively, growth pattern: intraobserver and interobserver kappa values of 0.37 and 0.41, respectively). Dukes' stage and patient age were the most important prognostic variables on multivariate regression analysis. Tumor differentiation, nuclear polarity, tubule configuration, and lymphocytic infiltration remained significantly related to survival in the presence of Dukes' stage and age. The model which best predicted prognosis was a combination of Dukes' stage, patient age and tumor differentiation. Further addition of the variables assessed by Jass to this model did not significantly improve the prediction of prognosis.
Conclusions: Dukes' classification is of greater prognostic value and more reproducible than the components of Jass' classification. The continued use of Dukes' classification is, therefore, warranted for prognostic and therapeutic decisions in patients with carcinoma of the colon and rectum.