Three hundred thirty-five previously untreated patients with advanced colorectal carcinoma were randomly assigned to treatment with 5-fluorouracil (5-FU) alone, 5-FU plus N-(phosphonacetyl)-L-aspartic acid (PALA), 5-FU plus high-dose thymidine, 5-FU plus levamisole, or 5-FU plus methyl CCNU, vincristine, and streptozotocin (MOF-Strept). Dosages were designed to produce definite toxicity in the majority of patients, although the nature of dose-limiting reactions varied considerably among regimens. 5-FU alone and 5-FU plus levamisole produced mucocutaneous reactions, diarrhea, and leukopenia; 5-FU plus PALA produced primarily mucocutaneous reactions and diarrhea; 5-FU plus thymidine produced leukopenia with occasional neurotoxicity and hypotension; and MOF-Strept produced substantial nausea and vomiting with both thrombocytopenia and leukopenia. Objective response rates among patients with measurable disease varied from 12% (5-FU plus PALA) to 34% (MOF-Strept), but none of the regimens were significantly superior to 5-FU alone. Both interval to progression and survival were comparable among the five regimens with no reasonable chance that any combination regimen could produce as much as a 50% improvement when compared with 5-FU alone. Whereas we observed definite modulation of 5-FU dose--toxicity relationships, particularly with the thymidine and PALA combinations, this did not result in a detectable improvement in therapeutic effect. None of the combination regimens, administered in the dosages and schedules we used, can be recommended as standard therapy of advanced colorectal carcinoma.