Age over 65 years is a risk factor per se for doxorubicin administration, and coexisting diseases pose additional problems. There is still controversy whether chemotherapy regimens for elderly patients with aggressive NHL should be full-dose doxorubicin containing or whether development of non-anthracycline containing regimens is warranted. In this prospective study, 47 patients aged over 65 years with diffuse large cell NHL clinical Stage I/IE bulky-IV and no other initial exclusion criteria were randomized to receive either BCNU 120 mg/m2 d. l, VP 16 60 mg/in2 d.2-4, procarbazine 85 mg/m2 d. 2-8 (arm A, 27 patients) or mitoxantrone 6 mg/m2 d. l. with VP16 and procarbazine in the same dosage and schedule (Arm B, 20 patients). Partial responders received additional irradiation treatment if feasible. Arms were well balanced according to age, sex, clinical stage and performance status. Ten patients from arm A and 13 from arm B had PS 2 or 3; 14 patients from arm A and 8 from arm B had clinically significant antecedent and/or concomitant disease (SACD: cardiac, vascular, cerebrovascular, neurological, renal or other). On the intent-to-treat basis, the results were the following. ARM A: median number of cycles 3 (range 1--6); early death 3 patients; 16/27 responses (59%), 7 complete (30%). ARM B: median number of cycles 3 (range 1-6); early death 4 patients; 12/20 responses (60%), 3 complete (15%). There was no difference either in response rate or survival between the two arms, and pooled results from the two arms displayed a plateau on the survival curve from the 20-th month onwards on the probability level of 0.40. Clinical stage of NHL, bulky disease, age and sex did not influence survival. Initial performance status did influence survival at the significance level of p = 0.045. Although presence of SACD did not influence initial performance status, it had a strong negative impact on survival p = 0.0004). The results point to the existence of two prognostic categories of elderly patients with large cell NHL, one with a poor survival, the other achieving a significant response rate and relapse free survival. Comorbidity (SACD) apparently accounts for the poor survival in a subpopulation of elderly patients. Clinical trials with elderly patients with NHL with PS 0 or 1 and no serious coexisting disease as inclusion criteria, analyzed on an evaluable patients basis, target only to a prognostically better subpopulation among these patients.