OBJECTIVE: The objective of this study was to quantify the clinical and economic effects of non-adherence with triple therapy in treatment-naïve HIV patients. METHODS: A Markov model was developed to project quality-adjusted life expectancy and direct medical costs for HIV patients (mean age = 37 years) on an initial regimen of highly active antiretroviral therapy (HAART) with efavirenz, lamivudine and extended release stavudine given once daily. Each month, patients faced a risk of transition to AIDS that was a function of viral load, CD4 count, and adherence to drug therapy. Patients were assumed to change to another triple-therapy regimen if their viral load was = 400 copies/mL or if they transitioned to AIDS. After four regimens, patients followed the natural history of the disease. We compared clinical and economic outcomes for two adherence scenarios: "clinical trial" (representing ideal utilization observed in clinical trials) and "typical" (based on observational studies in actual practice). Costs were derived from the HIV/AIDS Costs and Services Utilization Survey and average wholesale drug prices. Future costs and QALYs were discounted 3%. RESULTS: Mean discounted quality-adjusted life expectancy was 8.6 and 10.0 QALYs under the typical and clinical trial adherence scenarios, respectively. Lifetime direct medical costs in the typical and trial scenarios were $295,000 and $336,000, respectively. Compared with typical adherence, clinical trial adherence conferred an average gain of 1.4 QALYs at an incremental cost of $29,000 per QALY gained. Up to $1650 per patient per year could be spent on an intervention to improve adherence from typical to trial levels, and the incremental cost-effectiveness of doing so would remain below $50,000 per QALY gained. CONCLUSIONS: Typical adherence with HAART reduces quality-adjusted life expectancy by 14% compared to adherence levels in clinical trials. Programs to improve adherence appear to be a highly cost-effective use of resources.