The objective of this study was to investigate stay-on-therapy patternsover 3 years among patients prescribed different classes of antihypertensivedrugs for the first time. A retrospective analysis of informationrecorded in the drugs database of the Local Health Unit of Ravenna(Italy) was carried out on 7312 subjects receiving a first prescriptionfor diuretics, beta-blockers, calcium channel blockers, angiotensin-convertingenzyme (ACE) inhibitors or angiotensin II antagonists between 1 Januaryand 31 December 1997. Patients were followed up for 3 years. Allprescriptions of antihypertensive drugs filled during the follow-upperiods were considered. The patients continuing or discontinuingthe initial treatment, the duration of treatment, and the doses takenwere all calculated, as well as main factors influencing the persistencerate. The drugs prescribed were predominantly ACE-inhibitors, followedby calcium channel blockers, diuretics, beta-blockers and angiotensinII antagonists. A total of 57.9% of patients continued their initialtreatment during the 3-year follow-up period, 34.5% discontinuedthe treatment, whilst 7.6% were restarted on a treatment in the thirdyear. Persistence with treatment was influenced by: age of patient(persistence rate increasing proportionately with advancing years),type of drug first prescribed (persistence rate higher with angiotensinII antagonists, progressively lower with ACE-inhibitors, beta-blockers,calcium channel blockers and diuretics), gender of patient (persistencewas better in males), age of general practitioner (GP) (the youngerthe GP, the better the persistence rate) and gender of GP (betterstay-on-therapy rate with male GP prescribing). In the case of patientstreated continuously, mean daily dose increased progressively overthe 3 years. With adequate markers, helpful data can be collectedfrom prescription claims databases for the purpose of monitoringthe persistence of patients in continuing their medication, and thequality of antihypertensive treatment in a general practice setting.