A prospective phase II evaluation of regional FUDR chemotherapy using a totally implantable drug infusion pump was conducted in 81 patients with colorectal metastases to the liver. The survival results were compared to a historical control group of 129 patients with isolated liver metastases. The two groups were comparable with respect to their dominant prognostic factors. The pump patients received their continuous chemotherapy on an outpatient basis and had an 88% response rate, as evidenced by a fall in their serum CEA levels by one-third or greater after two cycles of chemotherapy. By four criteria, the regional chemotherapy patients had an improved survival rate compared to the control series. First, the 1 year survival and median survival was better for the entire group of pump patients vs. controls (82% vs. 36%, 26 months vs. 8 months, p <0.0001). The survival for the regional chemotherapy patients was not influenced by the extent of tumor involvement, whether previous systemic 5-FU was given, or whether the patient had symptomatic disease. Second, the entire group of regional chemotherapy patients (including nonresponders) had a greater 1 year survival compared to the most favorable subgroup of control patients with the following characteristics: normal liver function tests, no symptoms, and only one lobe involved (82% vs. 66%, p = 0.009). Third, a subgroup of 49 pump patients, whose initial treatment for metastatic disease was regional chemotherapy (within 3 months of diagnosis) had a better 1 year survival than an exactly matched group of 49 control patients (67% vs. 30%, p = 0.000003). Fourth, the actuarial survival for all 81 pump patients was significantly better than predicted by a mathematical model constructed to predict the patient's clinical course based upon the seven dominant prognostic variables identified in a multifactorial analysis (82% survival at 1 year vs. 33% predicted survival). While liver metastases could be controlled in most patients, the major cause of death was tumor progression in extrahepatic sites, particularly lung metastases and abdominal carcinomatosis. Although regional chemotherapy with an implantable pump appears to prolong life by 12 to 18 months more than matched historical controls, these results must be confirmed by a randomized (phase III) prospective clinical trial.