The Radiation Therapy Oncology Group (RTOG) conducted a Phase I/II study in hepatocellular cancer that closed on September 9, 1987 and some results presented previously. Here, 17 patient characteristics are evaluated to identify any of prognostic significance. Two hundred sixteen patients were entered and 198 (74% with metastases and/or previous chemotherapy) were evaluable. Treatment began with an induction regimen of external beam radiotherapy to the liver (21.0 Gy, 3.0 Gy/Fx, 10 MV photons, 4 days per week) with low-dose chemotherapy (5-Fluorouracil (FU), 500 mg, i.v.; Doxorubicin, 15 mg, i.v.) on treatment Days 1, 3, 5 and 7. In the later stages of these studies, 56 patients received external beam radiotherapy as hyperfractionated treatment (1.2 Gy twice daily, 4 hours separation, 5 days per week, 24.0 Gy total) with similar chemotherapy. One month following induction therapy, cycles of radiolabeled antibody therapy were given every 2 months. Each cycle was derived from a different species of animal and consisted of 30 mCi I-131 antiferritin, Day 0, and 20 mCi, Day 5. On Day -1, 5-FU, 500 mg, and Adriamycin, 15 mg, were administered. The overall median survival for the entire group, including previously treated patients, was 4.9 months. The median survival for alpha-fetoprotein (AFP) - patients not previously treated was 10.5 months. Median survival for all AFP - patients was 8.5 months and for all AFP + patients was 4.6 months (p = 0.006). Of the 17 pretreatment characteristics investigated for prognostic value Karnofsky Performance Score (KPS) (80-100 vs. less than 80) (p = 0.0001), presence/absence of ascites (p = 0.0002), bilirubin level (less than 1.5 vs. greater than or equal to 1.5) (p = 0.018), SGOT (less than or equal to 35 vs. greater than 35) (p = 0.001); alkaline phosphatase (less than or equal to 95 vs. greater than 95) (p = 0.008) were found to be significant independently using a multivariant regression model. The relative risk of dying for the unfavorable component of each of these characteristics was 2.2, 2.0, 1.5, 1.9 and 1.7, respectively. Good and poor prognostic groups were then defined and compared to a similar patient population (RTOG study 83-19) with confirmation of the validity of the model. When stratification for these overpowering clinical factors was incorporated, AFP status was again significant with a relative death rate 1.80 times higher for AFP+ patients. Our recommendations for structuring future prospective randomized trials are discussed and include stratification by AFP status.