Morphological predictors of survival in colorectal carcinoma: univariate and multivariate analysis.

In 150 curative resected adenocarcinomas of the large bowel we analyzed the morphological parameters of the primary tumor and lymph node involvement which most efficiently expressed the relationship to survival time. Using univariate survival analysis (product limit estimator according Kaplan-Meier), several macroscopical and microscopical parameters of the primary tumor and lymph node involvement significantly correlated with survival time. A multivariate survival analysis of covariates according to the Cox regression model revealed that the significant correlations of all these parameters were expressed by a set of five variables: pT stage, number of inflammatory cells, sex, age, and pN stage. As shown by the Kaplan-Meier test, this set of variables allowed a more precise prediction of survival time than mere staging according to the TNM system. Parametric multiple stepwise survival analysis was inefficient. No distinct relationship was found between the morphology of the primary tumor and the involvement of lymph nodes.