OBJECTIVE: To compare five choices to initiate antihypertensive pharmacotherapy as a function of persistence with treatment and drugs cost. METHODS: An administrative database, held by the Local Health Unit of Ravenna, recording pharmacy claims was used to perform an observational study of patients receiving antihypertensive drugs for the first time. All new users, 20 years-old or over, receiving a prescription for amlodipine, atenolol, fosinopril, indapamide or losartan between January 1st, 1997 and December 31st, 1997 were enrolled. The follow up period lasted 365 days. A bayesian cost-effectiveness analysis based on Markov Chain Monte Carlo simulations was performed. Persistence with treatment (a duration of therapy on any antihypertensive drug more than 273 days) at a correct dosage (within the therapeutic range recommended by the JNC VI) was assumed as a proxy to define effective therapies. Drugs cost was evaluated at NHS purchase prices. RESULTS: A total of 4614 subjects was enrolled (22.8% on amlodipine, 44.0% on atenolol, 18.8% on fosinopril, 6.7% on indapamide, and 7.8% on losartan). The annual average cost of treatment ranged between €44.27 (95% CI 38.18 - 52.06) for those started on atenolol to €175.65 (95% CI 150.37 - 205.10) for those started on losartan. The average probability of an effective antihypertensive treatment ranged from 8.40% for those initiated on fosinopril to 18.36% for those initiated on losartan. As a consequence, the choice to initiate antihypertensive treatment on losartan displayed a probability of being more cost-effective equal to 0.99 compared to fosinopril, 0.94 to amlodipine, 0.84 to indapamide, and 0.75 to atenolol. CONCLUSION: The low reproducibility of findings from experimental studies to clinical practice should encourage health care providers to account and assess both resources and associated outcomes in a "real world" setting: the availability of a linkage between these variables could affect the cost-effectiveness of selected choices.