The pathological grading and staging of rectal cancer

New techniques for the pathological grading and staging of large bowel cancer are being advocated, but the traditional methods have rarely been subjected to detailed scrutiny by modern statistical methods. A comprehensive set of discrete histopathological parameters was submitted for survival analysis, and subsequently to multivariate analysis using the Cox proportional hazards regression model. Parameters were selected that yielded the best-fitting model and awarded integer scores proportional to their regression coefficients. A superior grading system was derived. These variables were then allowed to compete in a model that included stage-related parameters. Only lymphocytic infiltration was selected along with the number of involved lymph nodes and extent of tumour spread. The clinical pathological prognostic classification was compared to Dukes, Astler-Coller and the TNM classification. Finally, the contribution made by DNA content was independent but very small indeed. Ploidy may be ignored if grading and staging are performed with meticulous care.