Aims: We investigated parameters which might help identify poor prognosis colorectal cancers and, in particular, we stratified Dukes' stage B carcinomas in order to identify those patients who would benefit from adjuvant therapy.
Methods and results: Histopathological parameters and DNA ploidy were analysed in a consecutive series of 256 resected colorectal cancers and their relationship with patient survival were investigated. By univariate analysis, Dukes' stage, degree of differentiation, nature of the invasive margin and DNA ploidy all correlated with prognosis. However, degree of differentiation assessed by the worst pattern seen anywhere in the tumour correlated more significantly with clinical outcome than did the predominant pattern. This was confirmed by multivariate analysis which showed only Dukes' stage and worst pattern differentiation as independent prognostic variables. Mucinous carcinomas showed no significant difference in outcome when compared to adenocarcinomas. DNA ploidy correlated with prognosis but only at a low level (P = 0.035) but this was maintained for Dukes' B carcinomas.
Conclusions: Dukes' stage and degree of differentiation provide independent prognostic information in colorectal cancer. However, differentiation should be assessed by the worst pattern and not by the predominant pattern as is currently recommended by the UKCCCR. DNA ploidy provides some prognostic information and does stratify Dukes' B cancers and thus might provide useful information on which to base decisions concerning adjuvant therapies in this difficult group. Mucinous differentiation has no prognostic significance.