Between April 1988 and April 1992 we conducted a study of DICE (dexamethasone 10 mg q6h, ifosfamide 1 g/m2, cisplatin 25 mg/m2, etoposide 100 mg/m2, and mesna uroprotection daily x 4 given every 28 days) in thirty six patients with intermediate or high-grade non-Hodgkin's lymphomas (NHL's). All patients had disease refractory to anthracycline-based chemotherapy and were ineligible for our dose intensive salvage chemotherapy program including autologous bone marrow transplant (ABMT). Twenty six men and 10 women aged 21-79 years (median 61) were given 100 courses of treatment in hospital. Twenty three patients had stage IV, 7 stage III and 6 stage II disease. Fourteen patients had bone marrow involvement and 11 had B symptoms. Thirty two patients had received only 1 previous chemotherapy regimen. After a follow-up of 1-70 months (median 7), 8 of the 34 (23%) patients evaluable for response had complete remission (CR) of their disease for 8-70+ months (med 24) and 15/34 (44%) partial remission (PR) for 2-55+ months (med 13), giving an overall response rate of 67% (23/34). Eight of the 36 patients are alive 34-70 months (median 48) after being enrolled on study, 6 of whom relapsed and responded to subsequent treatment. The estimated survival rates for all patients at 1 and 2 years are 50% and 36% respectively (Kaplan Meier). Response to prior chemotherapy was the only baseline characteristic statistically predictive (p = 0.04) of response to DICE by logistic regression analysis. Myelosuppression was the dose limiting toxicity.(ABSTRACT TRUNCATED AT 250 WORDS)