The most recent American Joint Committee on Cancer/International Union Against Cancer (AJCC/UICC) staging system subgroups patients into one to three and four or more positive nodes. However, the Gastrointestinal Study Group and the National Surgical Adjuvant Breast and Bowel Project divides node-positive patients into one to four and five or greater. A Cox multi-variate retrospective analysis was done of the overall survival of node-positive colon cancer patients with the specific objective of determining the most appropriate subcategorization. Data on 306 patients with node-positive colon cancer who underwent potentially curative surgery from 1970 to 1984 were analyzed retrospectively. No patient received adjuvant chemotherapy. Also excluded were patients with synchronous resected metastatic disease or those with rectal primaries. The median follow-up was 6 years, and the median survival for the entire group was 8.6 years. By univariate analysis, the following were significant prognostic features: number of positive nodes (P less than 0.0001), degree of differentiation (P less than 0.0001), colon primary site (P = 0.009), tumor stage (P = 0.001), and tumor size (P less than 0.0001). Lymphatic/blood vessel invasion and a mucinous histology were not significant. By Cox multivariate analysis the number of positive lymph nodes remained the best discriminant of survival (P = 0.0001). The number of positive nodes was related inversely to prognosis with the optimal dichotomization between one to three (66% 5-year survival) and four or greater nodes (37% 5-year survival).