OBJECTIVE: To identify key drivers of non-persistence on therapy, using multivariate regression techniques, based on IMS Disease Analyzer Germany (DA) data. METHODS: DA is a longitudinal patient database containing de-identified patient records collected from 400 practices (290 GPs and 110 Internal Specialists) treating currently +/-1.3 million patients. We identified a cohort of hypertensive patients being prescribed valsartan, metoprolol, ramipril, enalapril, lisonopril, amlodipine or hydrochlorothiazide (HCTZ), within the period 2003-2004. Patients had at least 12 months of data prior to, and following their index date (date of first prescription of the study drug within the window). History of hypertension was defined as the difference between diagnosis date and index date. Patients who start on monotherapy of study drugs and those who start on combination therapy involving at least one of the study drugs were analyzed separately. Medication persistence was defined as the total time on a drug, from initiation of therapy to the end of the last supplied prescription for that drug without intervening discontinuation. RESULTS: A total of 42,991 patients were analysed, of whom 61.7% discontinued study medication within 12 months post-index. Cox-regression showed that heart failure (Hazard Rate to discontinue therapy; HR=1.04), renal disease (HR=1.072), and not being treated with valsartan (HR=1.07 to 1.359) independently and significantly increased the probability of being non-persistent. Older age (HR=0.998 per year), male gender (HR=0.948), >6mo history of hypertension (HR=0.87), dyslipidemia (HR=0.96), recent stroke (HR=0.891), and receiving combination therapy (HR=0.92) were significantly associated with improved persistence. CONCLUSION: Different demographic and clinical factors are independently associated with persistence to antihypertensive therapy. Patients on valsartan were 7 to 36% more likely to persist compared to patients taking other frequently used antihypertensive medications. Considering such factors is important to identify patients most likely to discontinue therapy and to avoid achieving suboptimal therapeutic outcomes.