Objectives: To investigate the effect of combination antiretroviral therapy on plasma HIV-1 RNA as measured by HIV RNA PCR and to assess their safety and tolerability. Design: A randomized, multicentre, double-blind, placebo-controlled trial. Setting: Multicentre study in eight European countries, Australia and Canada. Patients: Antiretroviral naive patients (n = 103) with CD4 cell counts between 150 and 500 x 10(6)/I. Intervention: Patients were randomly assigned to zidovudine (ZDV; 200 mg three times per day) plus lamivudine (3TC; 150mg twice per day) or to ZDV + 3TC + indinavir (IND; 800 mg q8h) for 52 weeks. Main outcome measures: Degree and duration of reduction of plasma HIV-1 RNA as measured by RNA PCR; Development of drug-related toxicities sufficiently severe to warrant dose modification, interruption or permanent discontinuation. Results: ZDV + 3TC + IND reduced plasma HIV-1 RNA (P < 0.001) and increased CD4 cell count significantly (P = 0101) more than ZDV + 3TC. The addition of IND to ZDV + 3TC as initial therapy markedly increased the proportion of patients with plasma HIV-1 RNA values 500 copies/ml (31/52, 60%) or 20 copies/ml (24/52, 46%) as compared with ZDV + 3TC (9/50, 18% or 2/50, 4% respectively) at week 52 in an intention-to-treat, missing = failure analysis. Assessment of time to virological rebound (> 0.5 log(10) copies/ml above nadir) showed that patients who attained a minimum plasma HIV-1 RNA of less than or equal to 20 copies/ml were less likely to rebound than those who did not reach this threshold. The addition of IND to ZDV + 3TC did not result in any significant increase in adverse experiences. Conclusion: ZDV + 3TC + IND resulted in a considerable improvement compared with the double combination, in reduction in plasma HIV-1 RNA, increase in CD4 cell count and proportion of patients with HIV RNA below the limit of detection. Despite an average 3 log(10) decrease in plasma HIV-1 RNA on triple therapy, however, maximal suppression (less than or equal to 20 copies/ml) was only attained in about one-half of the patients in an intent-to-treat analysis. (C) 2000 Lippincott Williams & Wilkins