OBJECTIVES: Portuguese guidelines for treatment of antiretroviral (ARV) therapy-naïve HIV-patients include the use of combination ARV regimens with ≥ three drugs, such as two nucleoside/nucleotide reverse transcriptase inhibitors (NRTIs) and a ritonavir-boosted protease inhibitor. Ritonavir-boosted atazanavir (ATV/r) and lopinavir (LPV/r) are among recommended first-line therapies. This study compared lifetime clinical benefits and cost-effectiveness of ARV therapy with ATV/r or LPV/r in treatment-naïve HIV patients at different baseline CD4 levels from the Portuguese National Healthcare System perspective. METHODS: A microsimulation model compared the average patient population of the CASTLE trial, treated with either ATV/r or LPV/r (median CD4 level 205 cells/m3), with the subgroup of patients with a CD4 level of ≥200 cells/m3 when enrolled in the trial. Virological response was modelled as reduction in viral-load (<50 copies/ml), impacting CD4+ cell count as well as risk of AIDS Defining Events (ADEs) and cardiovascular events. Both treatment-specific adverse events (AE) and long-term toxicities were included. Relative drug efficacy and AEs were based on data from the CASTLE study; other inputs came from published literature. RESULTS: When compared to the average patient population, initiation of ATV/r- or LPV/r-based treatments (with CD4+ cell ≥200 cells/m3 at baseline) in treatment-naïve patients were cost-effective strategies resulting in more life-years and quality-adjusted life years (QALYs) gained, at lower overall costs. Initiation of ATV/r treatment provided greater effectiveness (additional 1.69 life-years and 1.77 QALYs) and lower costs (€162,460 vs. €177,038; 2010 values) compared with LPV/r (5% discounting for costs and effects). CONCLUSIONS: Initiating treatment with ARV agents when CD4+ cell ≥200 cells/m3 was predicted to lower overall treatment costs, lead to increased survival and fewer ADEs. Similarly, initiating treatment with ATV/r rather than LPV/r represented a cost-effective strategy in treatment naïve patients with HIV-1 infection. Probabilistic sensitivity analyses were performed and confirmed the robustness of the base-case findings.