We investigated the occurrence of apoptotic cell death in 104 colorectal carcinomas by terminal-deoxynucleotidyl-transferase-mediated dUTP-diotin nick end-labeling (TUNEL) to determine whether apoptosis could be a useful prognostic factor. The apoptotic index (AI) was calculated as the percentage of positive cancer cells per 1,000 cancer cells, the median AI being 4.1, with a range of 1.9-4.7. Apoptosis was less frequently observed in tumors with higher malignant potential, such as those at advanced stages Dukes B, C and D, than those at Dukes stage A (P < 0.05); in tumors showing evidence of moderate differentiation than in well-differentiated tumors (P < 0.05); and in tumors with venous invasion or lymph node metastasis than in those without these features (P < 0.05). Moreover, the subgroup of patients with a low AI of < 4.1 had a significantly poorer survival rate than the subgroup with a high AI in tumors at Dukes stage C, the 5-year survival rates being 33% vs 68% (P < 0.05; Cox-Mantel). Our findings suggest that less apoptosis might result in a greater progression of colorectal carcinoma, and that the rate of apoptosis might be an indicator of the degree of malignancy. Thus it would appear that the frequency of apoptosis in tumor cells could be a useful prognostic factor in colorectal carcinoma.