To determine whether the production of experimental hepatic metastases in athymic nude mice by human colorectal carcinomas (HCC) correlated with the clinical outcome in patients, we harvested colorectal carcinomas from 82 patients, dissociated the tumors with collagenase and DNase, and injected them into groups of nude mice, either in the flank to assess experimental tumorigenicity or into the spleen to produce experimental metastasis in the liver. Growth in mice was then associated with clinicopathological factors and clinical outcome. Growth of HCC in either the flanks or the livers of nude mice was associated with the time to recurrence in a Wilcoxon analysis. Analysis of the outcome data in a Cox proportional hazards model suggested that there was an interaction between tumorigenicity and metastatic potential of HCC in nude mice and serum CEA concentration in the patient and stage of disease. A univariate analysis indicated that both tumorigenicity and metastatic potential of HCC in nude mice were significantly associated with the serum CEA concentration of the patient but not with the other variables of stage of disease, mucin production, local tissue invasion, state of differentiation, or sex. A subset of 57 patients was operated upon for cure and followed prospectively for up to 61 months. Tumorigenicity and, to a lesser extent, experimental metastatic potential were associated with disease recurrence in 23 of these patients. Seventy-eight % of the subset of patients who were operated upon for cure developed liver metastasis as one site of their progressive disease. Thus, the ability of HCC cells isolated from surgical specimens to grow in athymic nude mice correlates with the development of advanced disease in patients.