In an attempt to determine factors predictive of survival in patients seropositive for human immunodeficiency virus (HIV) with acquired immune deficiency syndrome (AIDS)-related lymphoma, the authors studied 60 such patients, all of whom were treated with curative intent. Eleven patients presented with lymphoma primary to the brain (P-CNS); the remaining 49 had systemic AIDS-related lymphoma. Patients with P-CNS lymphoma had more severe underlying HIV-related disease than did patients with systemic lymphoma as evidenced by a higher incidence of AIDS before the diagnosis of lymphoma (73% versus 37%; P = 0.04), and lower median number of CD-4-positive lymphocytes in peripheral blood at diagnosis of lymphoma (30/dl versus 189/dl; P = 0.005). Median survival of such patients was 2.5 months versus 6.0 months for patients with systemic lymphoma (P = 0.04). Forty patients with systemic AIDS-related lymphoma have died; three factors were strongly associated with shorter survival: (1) Karnofsky performance status (KPS) of less than 70% (multivariate relative survival risk [RSR] = 3.1); (2) history of AIDS before the diagnosis of lymphoma (multivariate RSR = 3.0 for opportunistic infection plus Kaposi's sarcoma); and (3) bone marrow involvement (RSR = 3.1)). All three factors (KPS of less than 70%, prior AIDS diagnosis, and marrow involvement) were associated with early demise attributed to AIDS, whereas death attributed to lymphoma per se was associated with only two factors (KPS of less than 70% and marrow involvement). In the absence of all three risk factors, a "good prognosis" group of 17 patients was defined, with a median survival of 11.3 months; the median survival of the remaining patients ("poor prognosis") was 4.0 months (P = 0.0002). Attainment of complete response to therapy (CR) was strongly related to prolonged survival in the patients in the good prognosis group (17.8 months in patients with CR versus 5.0 months in those with less than CR); however, such meaningful prolongation of survival was not seen in patients with poor prognosis who attained CR (6.3 months versus 3.4 months). The patients with poor prognosis may be unable to tolerate the insult of multiagent chemotherapy, experiencing low CR rates (25%) and death caused by lymphoma and AIDS. However, patients in either prognostic category who attained CR remained at risk for dying of AIDS while the lymphoma was in remission. Thus, it is apparent that meaningful prolongation of survival in the patient with AIDS-related lymphoma will require not only effective antineoplastic intervention, but also control of the underlying HIV infection. In addition, future therapeutic trials should stratify patients based upon the prognostic factors defined here in an attempt to clarify the results obtained.