[Statistical analysis of pathological factors influencing prognosis of colorectal carcinoma]

The pathological and follow-up data of 503 patients with colorectal carcinoma have been analysed to assess quantitatively prognostic values of 16 pathological factors and their sub-categories with aid of computer. Besides the routine procedure, Cox's proportional hazards model, Hayashi's quantification theory II, III, Akaike's information criteria, Kendall's rank correlation, relative risk, and linear trend test were used. Regardless of methods used, liver metastasis, peritoneal dissemination, lymph nodes metastasis, depth of penetration, venous and lymphatic invasion emerged as powerful prognostic variables. Most of the analyses indicated that venous invasion contained the largest amount of prognostic information including that of long-term survival. When venous invasion is graded not only by the number of veins involved but also by the depth of invasion, its prognostic capability was comparable with that of staging systems. The values of relative risks calculated for the factors' sub-category were more closely associated with prognosis than five-year survival rates or median survival rates, thus permitting detailed comparison of different categories. We conclude the modern statistical methods are indispensable tool for prognostic factor analysis.