We assessed the efficacy of adding indinavir in patients with advanced HIV-1 infection, who were previously exposed to different reverse transcriptase (RT) nucleoside analogues. Twenty-five patients with an initial median CD4 cell count of 20 cells/mm3 (range, 0-80 cells/mm3) were treated with indinavir (800 mg three times per day) for 24 weeks. The median initial viral load was 5.4 log (range, 3.6-6.7 log). Of these patients, 56% (14 of 25) had an initial decrease in viral load of >1 log and sustained response of >0.5 log of HIV-1 RNA from baseline. Twelve of these 14 responder patients (85%) showed a sustained RNA response undetectable by NASBA assay, and no genotypic changes in protease were detected at week 24. In those with a temporary or absent response to indinavir, either resistant viruses or lack of compliance was observed. In compliant patients (15 of 16), relatively small increases in 50% inhibitory concentration (IC50) to indinavir and only two to three amino acid changes were sufficient to produce treatment failure. Phenotypic drug-resistant assays at 24 weeks revealed cross-resistance to ritonavir in all the patient isolates and to saquinavir in one third of the isolates. We observed an initial and persistent response to the addition of indinavir in patients with advanced disease and prolonged antiretroviral treatment. Therapy failure, as defined by increases in viral RNA, was associated with either lack of compliance or the development of low level indinavir-resistant virus. Clinical studies need to be designed to determine to what extent these viruses may respond to other protease inhibitors.