As a result of high inter-patient variability, and efficacy-concentration and toxicity-concentration relationships, optimization of HIV-protease inhibitor (PI) doses based on plasma concentrations could be beneficial. During a 48-week open prospective non-randomized interventional study of 115 protease inhibitor-naïve patients initiating an indinavir/ritonavir- or lopinavir/ritonavir-, or nelfinavir-containing therapy, protease inhibitor dose was modified when plasma trough concentrations (C(trough)) at weeks 2, 8, 16 and 24 were outside predefined optimal concentration ranges. Failure of the strategy was defined as the proportions of patients with HIV-RNA above 200 copies/mL from weeks 24 to 48 and/or experiencing grades 2, 3 or 4 PI-related adverse events during the study; proportion of patients with last C(trough) measurement outside the concentration range was determined at each visit. Virological failure and/or occurrence of adverse event were observed in 37/94 assessable patients (39%; 95% CI: 29.4-50.0). In the on-treatment analysis, failure of the strategy was noted in 16% of indinavir/r- or lopinavir/r-treated patients (8/51; 95% CI: 7.0-28.6; virological failure: 2; adverse event: 6) but in 44% of nelfinavir-treated patients (11/25; 95% CI: 24.4-65.1; virological failure: 10; adverse event: 1); C(trough) concentrations outside the range were less frequent at the last measurement than at W2 (41% vs. 66%; P < 0.05), with proportions of 35% for indinavir/r- or lopinavir/r-treated patients, but 57% for nelfinavir-treated patients. The proposed strategy of therapeutic drug monitoring may be beneficial to indinavir/r- and lopinavir/r-treated patients, but failed to move concentrations into the predefined range and to produce the expected virological success for nelfinavir-treated patients.