Background. Intermittent interleukin-2 therapy for HIV-l by continuous intravenous infusion leads to sustained increase of CD4 T cells. This method of administration is, however, inconvenient and has limiting toxic effects. We did a randomised study to compare safety and efficacy of antiviral treatment alone or combined with various interleukin-2 regimens in HIV-1-infected patients. Methods. 94 symptom-free patients, naive to antiretroviral treatment, with CD4-T-cell counts of 250-550 cells/mu l at baseline were randomly assigned zidovudine and didanosine alone (n = 26) or combined with interleukin-2 administered intravenously (12 million IU/day, n = 22) or subcutaneously (3 million IU/m(2) twice daily, n = 24) for 5 days, or were given polyethyleneglycol-modified (PEG) interleukin-2 (2 million IU/m(2) intravenous bolus, n = 22) administered every 2 months from week 2 to week 50 (seven cycles). Safety and immunological and virological results were monitored until week 56. Findings. CD4-T-cell count increased to higher than baseline by a mean of 564 cells/mu L (subcutaneous group), 676 cells/mu L (intravenous group), 105 cells/mu L (PEG group), and 55 cells/mu L (antiretroviral-therapy group, p = 0.0001). 68% and 77% of patients in the subcutaneous and intravenous; groups, respectively, achieved an 80% increase of CD4 T cells (p < 0.001). In these two groups, 50% of patients restored a CD4/CD8-T-cell ratio of more than 1. The groups did not differ significantly for changes in plasma HIV-1 RNP, loads throughout the study. The duration of common side-effects of interleukin-2 was shorter in the subcutaneous group, which enabled outpatient treatment. Naive and memory CD4 T cells, CD28 expression on CD4 and CD8 T cells, and restoration of in-vitro proliferative response to mitogens and recall antigens increased in the intravenous and subcutaneous groups. Interpretation. Subcutaneous interleukin-2 is a convenient regimen that, as well as intravenous therapy, improves immunological function in HIV-l-infected patients receiving two nucleosides. Larger studies are needed to show whether immunological improvements translate into clinical benefit.