OBJECTIVES: To investigate the potential impact of non-adherence resulting from multiple daily dosing on cost-utility of gastrointestinal prophylaxis strategies in preventing NSAID associated gastrointestinal complications. METHODS: Two decision analytical models (TreeAge 2005) were developed - one model incorporating non-adherence resulting from multiple daily dosing and another model that does not. Each model evaluates the cost-utility of gastrointestinal prophylaxis in a hypothetical cohort of patients' age = 65 years beginning a 6 month course of NSAIDs. Patients entering each model were treated with either: (1.) No Prophylaxis, (2.) Proton Pump Inhibitors (generic omeprazole od) (3.) misoprostol (200ug qid), (4.) ranitidine (300mg bid). Average adherence to therapy: od = 79% ± 14%, bid = 69% ± 15% and qid = 51% ± 20% were obtained from a systematic review of studies using electronic monitoring devices. Costs were from the perspective of a 3rd party payer of a Canadian provincial seniors' drug plan. Cost-Utility was measured in terms of cost per quality adjusted life year (QALY) gained, relative to no prophylaxis. Probabilistic sensitivity analysis using Monte Carlo Simulation was used to generate uncertainty of results, along with cost-effectiveness acceptability curves. RESULTS: Misoprostol (200ug qid) has a greater likelihood (69.2%) than generic omeprazole (30.8%) of being the optimal cost-effective strategy at a willingness to pay (WTP) of $50,000 per QALY in the model that does not incorporate non-adherence resulting from multiple daily dosing. In the model that incorporates non-adherence resulting from multiple daily dose regimens, generic omeprazole has a greater likelihood (94.8%) than misoprostol (5.2%) of being the optimal cost-effective strategy at a WTP of $50,000 per QALY. CONCLUSION: Cost-utility results are sensitive to non-adherence resulting from multiple-daily dosing. Previous economic models in this area have not incorporated this in the analysis. Markov modeling or discrete event simulation may be better suited to incorporate non-adherence rates of therapies.