OBJECTIVES: To compare treatment patterns and healthcare costs among subjects newly treated with donepezil versus subjects newly treated with rivastigmine or galantamine. METHODS: This retrospective claims-based analysis utilized data from a large US managed care plan. Subjects were ≥50 years of age; filled ≥1 prescription for donepezil, rivastigmine, or galantamine from January 1, 2005 – December 31, 2007; and were continuously enrolled for ≥18 months. Rivastigmine and galantamine cohorts were each matched to donepezil cohort at 1:3 ratio by propensity score. Logistic regression, Cox regression, and generalized linear models with gamma distribution and log link were used to examine adherence (medication possession ratio ≥80%), persistence/discontinuation and total health care costs, as appropriate.  RESULTS: Of 488 rivastigmine subjects, 33.0% were adherent and 69.3% discontinued medication while the rates of adherence and discontinuation in the matched 1,464 donepezil users were 42.4% and 60.7%, respectively.   Compared with rivastigmine cohort, donepezil subjects were more likely to be adherent (OR: 1.395, 95% CI: 1.117-1.742, p=0.003) and had a lower risk of discontinuation (RR: 0.827, 95% CI: 0.729-0.939, p=0.003). Mean annualized total health care costs did not differ significantly between two cohorts (donepezil: $11,431; rivastigmine:$ 12,358), although donepezil appeared to be associated with lower costs in regression analysis (Cost Ratio: 0.898, 95% CI: 0.803-1.005, p=0.06). Of 597 galantamine subjects, 36.5% were adherent and 65.7% discontinued medication while the rates of adherence and discontinuation in the matched 1,791 donepezil users were 42.0% and 60.6%, respectively. Donepezil users were more likely to be adherent than galantamine users (OR: 1.261, 95% CI: 1.038-1.534, p=0.020); however, discontinuation risk was similar. Annualized health care costs were similar in the two cohorts both bivariately (donepezil: $11,644; galantamine:$ 10,876) and in regression analysis. CONCLUSIONS: Donepezil users demonstrated better adherence than each of the matched cohorts, and lower discontinuation risk than the rivastigmine cohort. Total costs were similar.