BACKGROUND Regimens based on ritonavir-boosted protease inhibitors (PI/r) are often used as rescue interventions. It is unclear whether significant differences exist between distinct PI/r. METHODS All HIV+ patients who had experienced PI failure at two HIV clinics and were rescued with a regimen based on saquinavir (SQV)/r 1000/100 mg bid, indinavir (IDV)/r 800/100 mg bid, lopinavir (LPV)/r 400/100 mg bid, amprenavir (APV)/r 600/100 mg bid, atazanavir (ATV)/r 300/100 mg qd, or tipranavir (TPV)/r 500/200 mg bid were retrospectively examined. A significant virological response (VR) was defined as >1 log reduction in plasma HIV-RNA or to <50 copies/mL at week 24. RESULTS A total of 389 patients were included in the analysis: 139 on SQV/r, 35 on IDV/r, 129 on LPV/r, 35 on APV/r, 29 on ATV/r, and 22 on TPV/r. No significant differences in HIV-RNA and CD4 counts at baseline were recognized between groups. In a multivariate analysis, only the total number of protease resistance mutations was associated with a lower VR (odds ratio [OR] = 0.77, 95% CI 0.68-0.87, p < .001). The presence of <5 or > or =5 protease resistance mutations at baseline was the best threshold to discriminate the achievement of VR in any treatment group. In an intent-to-treat analysis, for individuals with 5 protease resistance mutations, the rates of VR were 64% with TPV/r, 47% with LPV/r, 46% with SQV/r, 33% with ATV/r, 25% with IDV/r, and 16% with APV/r. Adverse events leading to treatment withdrawal occurred more frequently using IDV/r (22.8%) than others (p = .03). CONCLUSIONS The rate of VR in salvage therapy using PI/r-based regimens is relatively high in PI-experienced patients. The efficacy is greatly influenced by the number of baseline protease resistance mutations; 5 mutations is the best threshold to predict the chances of VR to any PI/r-based regimen.