Available treatments for Alzheimer's disease (AD) need to be evaluated in order to determine whether the clinical benefits justify their additional costs. This study evaluated the cost-effectiveness of donepezil treatment compared with no-drug treatment of mild and moderate AD from the perspective of society and the health care system in Spain. A Markov model was designed to simulate the natural history of a cohort of patients with mild and moderate AD. Monthly transition probabilities were estimated from the international literature and donepezil clinical trials. Direct medical and non-medical costs and utilities were derived from Spanish studies. Local data on tolerance and medication withdrawal rates were incorporated into the model. Incremental cost-effectiveness ratios for a range of realistic treatment options were calculated. A probabilistic sensitivity analysis was carried out using a Monte Carlo approach with 10,000 iterations. In the baseline scenario (24 months, patients initially with mild AD) incremental cost-effectiveness for direct medical costs was 20,353 euro/QALY. When all costs were taken into account, donepezil treatment was the dominant strategy. Incremental cost-effectiveness ratios vary according to the selected perspective. For the baseline scenario, donepezil treatment is cost-effective with a probability of 95% for a threshold efficiency of 25,000 euro/QALY.