Dukes' classification revisited. Findings from the National Surgical Adjuvant Breast and Bowel Projects (Protocol R-01).

The relative prognostic value of the Dukes, Astler, and Coller and TNM staging systems was evaluated for 745 pathologically evaluable patients with rectal cancer enrolled in protocol R-01 of the National Surgical Adjuvant Breast and Bowel Projects. All three methods were found to be highly interrelated. However, the magnitude and consistency of prognostic discrimination among stages was best exhibited by the Dukes' and TNM systems. Survival was comparable among patients with Astler and Coller A and B1 and TNM T1N0M0 and T2N0M0 lesions. Since neither method improved on the predictability noted in Dukes' A cases it is suggested that the use of confusing subscripts is unnecessary. On the other hand, striking prognostic discrimination was observed when Dukes' C cases were subdivided according to depth of tumor penetration as proposed by Astler and Coller and designated as C1 and C2. Multivariate analyses revealed this feature to be independent of number of nodal metastases (1-4 versus 5+ positive), their site (near or far from the growth), or degree of tumor differentiation. The site of nodal metastases appeared to be related to numbers of nodal metastases rather than site per se. Considerations of the findings indicate that the Dukes' staging method is the simplest and most consistent algorithm related to prognosis. The only modification that would enhance its value in this regard would be the subdivision of C cases according to the criteria of Astler and Coller rather than that proposed by Dukes himself.