OBJECTIVES: The use of antiretroviral therapy has enormous success in slowing down disease progression in HIV/AIDS. Combinations of therapeutics have resulted in declining rates for opportunistic infections and mortality. The high costs of these medications have prompted questions about their cost-effectiveness. Most studies estimating cost-effectiveness of combination antiretroviral therapy (HAART) use data from a short period of time. Data is often collected during six months. In this study, the impact of using an extended study period of two years to assess the long-term effects of HAART is investigated. METHODS: To estimate the impact of a longer period of data collecting, data from the HIV-database from the University Hospital of Groningen were used. This database contains data on among others hospital admission data, opportunistic infections and serologic markers of over 300 patients since 1996. A Markov transition model for disease progression was developed. Using data from a 2-year period, the Markov transition probabilities and the final Markov reward was calculated. This was compared with an analysis in which data from a 6-month period were used. RESULTS: When compared with progression rates based on 6-month data, smaller progression rates were found when the model was based on a 2-year data. The use of a 2-year data period for Markov modelling was associated with better health outcomes than with a model based on 6-month data. Since the modelling of outcomes is closely associated with costs, a decline in costs was also seen. Overall, the cost-effectiveness of HAART was more favourable when the model was based on data from a 2-year period. CONCLUSIONS: The use of an extended study period for modelling the long-term effect of HAART leads to significantly different results. It is likely that the use of a relatively short period for the gathering of data leads to an underestimation of the cost-effectiveness ratio of HAART.