OBJECTIVES: In contrast to commonly used discrete-time Markov models, we developed a more realistic continuous-time multi-stage Markov model to evaluate long-term clinical and economic outcomes of raltegravir in treatment naïve HIV-1 patients. METHODS: The multi-stage cost-effectiveness model incorporating 3 lines of therapy was developed using differential equations and was solved in Mathematica® 6.0. The analysis was conducted from the perspective of the National payer in Portugal.  A typical patient enters the model in a given health state, transitions to another health state, can develop acquired immunodeficiency syndrome (AIDS)/ coronary heart disease (CHD)/other adverse events or die. Eighteen health states were defined based on CD4 and HIV RNA levels. We used the maximum likelihood method to estimate matrices of instantaneous transition rates corresponding to the efficacy of the included therapies. Six multi-stage treatment strategies depicting clinical practice in Portugal were evaluated. The model outputs included projected number of AIDS and CHD events, life expectancy and incremental cost-effectiveness ratios (ICER). The model was evaluated for internal and external validity and extensive sensitivity analyses were conducted. RESULTS: The predicted prevalence of patients in different health states over time provided a good fit to the clinical trial data. Strategies where raltegravir was included in the initiating therapy followed by an NNRTI or PI based regimen against efavirenz or PI based initiating therapies followed by raltegravir resulted in longer undiscounted life expectancy [20.07 Yrs. vs. 18.88 Yrs] and demonstrated cost-effectiveness [ICERs: €1,078 - €33,406/QALY]. CONCLUSIONS: Continuous-time Markov modeling based on transition matrices estimated using robust statistical methodology is a more realistic and sound approach. Results suggest initiating therapy on raltegravir generated lower costs and higher survival versus saving the drug for later lines of therapy in Portugal.