OBJECTIVE: To compare one protease inhibitor (PI)-based and two PI-sparing antiretroviral therapy regimens. METHODS: International, open label, randomized study of antiretroviral drug-naive patients, with CD4 lymphocyte counts textgreater/= 200 x 106 cells/l and plasma HIV-1 RNA levels textgreater 500 copies/ml. Treatment assignment to stavudine and didanosine plus indinavir or nevirapine or lamivudine. Primary study endpoint was the percentage of patients with plasma HIV-1 RNA levels textless 500 copies/ml after 48 weeks in the intention-to-treat analysis (ITT). RESULTS: In total, 298 patients were enrolled. After 48 weeks, the percentage of patients in the indinavir, nevirapine and lamivudine arms with HIV-1 RNA textless 500 copies/ml was 57.0%, 58.4% and 58.7%, respectively, in an ITT analysis. After 96 weeks of follow-up, these percentages were 50.0%, 59.6% and 45.0%, respectively. The percentage of patients with HIV-1 RNA textless 50 copies/ml was significantly less for those allocated to lamivudine in an on-treatment analysis after 48 and 96 weeks of follow-up. Patients in the nevirapine arm experienced a smaller increase in the absolute number of CD4 T lymphocytes. There were no significant differences in the incidence of serious adverse events. CONCLUSIONS: A comparable virological response can be achieved with first-line PI-base and PI-sparing regimens. The triple nucleoside regimen utilized may be less likely to result in viral suppression to textless 50 copies/ml, while the nevirapine-based regimen is associated with a lower increase in CD4 T lymphocytes.