OBJECTIVE To assess the efficacy of salvage therapy containing ritonavir and saquinavir after failure of indinavir- or nelfinavir-containing regimens, and to determine correlates of success or failure. METHODS Retrospective chart review. SETTING. The Moore Clinic - the HIV clinic of Johns Hopkins Hospital. METHODS Forty-one HIV-infected patients were identified through physician contacts, referrals from other providers, and review of a comprehensive clinical database. METHODS To determine response to salvage therapy, HIV-1 viral RNA (absolute and log10-transformed) was measured using the Roche Amplicor quantitative HIV-1 RNA assay after initiation of the salvage regimen. Potential correlates of response included: viral RNA at the time of switch; viral RNA at the time of switch as a percentage of baseline viral RNA; magnitude of decline in viral RNA; and the interval between virologic failure of single protease inhibitor therapy and switch to the salvage regimen. RESULTS Thirteen (56.5%) of 23 patients failing indinavir responded to salvage therapy (HIV RNA < 400 copies/ml) with persistence throughout the follow-up period (median of 37 weeks; range 18-67 weeks). Mean absolute viral RNA at the time of switch was 20 238 copies/ml (median, 9281) compared with 42 953 copies/ml (median, 24 650) for the 10 non-responders. Mean log10 viral RNA at switch was 3.804 for responders versus 4.405 for non-responders (P = 0.040). Among four responders who had failed nelfinavir, mean viral RNA was 9634 copies/ml and mean log10 viral RNA was 3.749 at the time of switch. Two non-responders had a mean viral RNA of 21 551 and a mean log10 viral RNA of 4.037 at switch. CONCLUSIONS In contrast with previous reports, salvage regimens containing ritonavir and/or saquinavir can be effective and durable following the failure of combination regimens containing either indinavir or nelfinavir. Salvage therapy may be more likely to succeed when it is initiated early in failure at low viral loads.