One thousand two hundred fifty-four cases (610 colonic and 644 rectal cancers resected during 2 years from 1984 and followed up for more than 5 years) were entered from 140 institutions in Japan and analyzed by means of Cox's proportional hazards model. The analyzed pathologic variables were the size and depth of invasion, Dukes' stage, venous invasion, lymphatic permeation, and other clinical features, such as the sex and age of the patient and location of the tumor. The extent of dissection, serum carcinoembryonic antigen (CEA) level, and the presence or absence of adjuvant chemotherapy were also analyzed. Adjuvant chemotherapy consisted of three arms for both colonic and rectal cancers. For colonic cancer, arm I was a combination of i.p. (intraportal) and i.v. mitomycin C (MMC) + p.o. 5-fluorouracil (5-FU); arm II was i.v. MMC + p.o. 5-FU; and arm III was surgery only. For rectal cancer, arm IV was a combination of i.a. (inferior mesenteric artery) and i.v. MMC + p.o. 5-FU; arm V was i.v. MMC + p.o. 5-FU; and arm VI was surgery only. As for the factors affecting the disease-free survival of the patient, multivariate analysis disclosed nodal involvement, venous invasion, an elevated CEA level, and the lower part of the rectum. The effect of adjuvant chemotherapy on the patient's survival was proven for rectal cancer but not for colonic cancer. We conclude that these factors should be considered in setting the stage of tumor pre- and postoperatively.