OBJECTIVE: To estimate incremental cost-effectiveness (ICE) of donepezil for mild or moderate AD as compared with option of no treatment in Polish settings. METHODS: We used Markov model developed by Naumann et al. Model assumptions: treatment duration and drug effect, 6 months; cohort of patients initially split by 50% into mild and moderate disease states; 50% of patients on 5 mg and 50% on 10 mg daily dose of donepezil. Data on costs by stage of disease including: medications, medical equipment, home help and community nurse, were taken from the Polish Alzheimer's Society. Study was conducted from a societal perspective. Three types of outcome were estimated: expected costs, years in state less than severe (SLS), and incremental costs per additional SLS per treated person over 5 years projection. RESULTS: The difference in expected costs per patient over 12, 24, and 60 months of treatment was: US $634, $520, and $408, respectively. Patients receiving donepezil were predicted to spend additional 0.123 year in less than severe AD compared with untreated patients. Incremental costs of SLS per treated person was $3,315/year. Sensitivity analysis revealed that ICE increased in parallel with the percentage of patients being initially in the mild AD. The difference in the total expected costs of treatment was: $355 and $303 when cohort comprised of 75% and 100% of patients with mild AD, respectively. In these scenarios additional time spent in SLS AD was 0.151 and 0.179 year and incremental costs of SLS was $2,364/year and $1,694/year, respectively. When the duration of treatment and effects were projected for 1 year, incremental costs of SLS was $3,852/year. CONCLUSION: Donepezil in the treatment of AD is most cost-effective in mild stage of the disease and its cost-effectiveness is reduced when treatment lasts > 6 months.