A double-blind, randomized study of zidovudine-experienced, PI- and lamivudine-naive adults with baseline CD4 cell counts of < or =50 cells/mm3 had demonstrated that the HIV suppression achieved with zidovudine, lamivudine, and indinavir therapy was superior to that achieved with dual-nucleoside or indinavir-only regimens after 24 weeks of therapy. In a 192-week extension of the study, 371 participants received open-label indinavir with or without other antiretroviral drugs. One hundred and eight subjects were originally randomized to receive triple therapy. After 216 weeks, the proportion of subjects with HIV RNA levels of <500 copies/mL were 34%, according to a general estimating equation analysis, 92%, according to an observed data analysis, and 24%, according to an intention-to-treat analysis counting noncompleters as failures; the proportions of subjects with HIV RNA levels of <50 copies/mL were 31%, 85%, and 22%, respectively. Hyperbilirubinemia (experienced by 31% of subjects), nausea (17%), abdominal pain (14%), and nephrolithiasis (13%) were the most common drug-related adverse events during the extension.