OBJECTIVES: International databases for prescription and use of pharmaceuticals suggest that adherence and persistence to antihypertensive treatment is low and decreasing over time. Furthermore, persistence rates seem to differ between antihypertensive drug classes. The objective of this analysis is to assess the implications of poor persistence on the cost-effectiveness of alternative antihypertensive drug treatments. METHODS: A stochastic Markov model was developed to estimate resource use, morbidity and mortality as a result of published differences in persistence rates. All classes of antihypertensives are assumed to have identical blood-pressure lowering effects, and a given reduction in blood pressure thus generates identical reductions in the probability for stroke, myocardial infarction, and congestive heart failure. Consequently, differences in CV-related morbidity and mortality are solely driven by differences in persistency rates. The model employs a lifetime perspective for a 50-year-old patient in base-case. Patients that discontinue treatment are assumed to lose the protective drug effect, and transit from a well-controlled health state to health states associated with moderate or serious hypertension. In the different Markov states patients experience CV-events subject to event rates calculated on the basis of data from the literature. RESULTS: In a hypothetical comparison between two cohorts of 1000 completely persistent and completely non-persistent 50-year-olds, the model predicts a reduction of 75 cardiovascular events and generates a maximum of 0.46 LYG (discounted at 3%). The ICER (per LYG) of ARBs compared to alternative drug classes, is highest when compared to ACE-inhibitors (SEK 105,000, €1 = SEK 9.25) and lowest when compared to thiazides (SEK 68,000). The cost-effectiveness decreases with increasing start age or diminishing time perspective. CONCLUSIONS: Model results suggest that differences in persistence rates have a non-negligible impact on the relative cost-effectiveness between antihypertensive drug treatments. Validation of model predictions via real-life epidemiological studies is warranted.