OBJECTIVE: To estimate the cost-effectiveness of using rt-PA in treating acute ischaemic stroke in the UK. METHOD: We used a decision-analytic model (Stroke Outcome Model) to estimate the long term health and cost consequences of administering rt-PA to patients with acute ischaemic stroke, based on the pooled results of three major trials: ECASS-1, ECASS-2 and NINDS (2044 patients in total). Long-term consequences of stroke were based on the Oxfordshire Community Stroke Project. Disability was defined using the modified Rankin scale (disabled: categories 3-5). Utilities for disability states were taken from published studies of stroke survivors. Resource use measured from a health and social service perspective, was based on published studies and a panel of clinical experts. Unit costs (1996) were obtained from national sources. Costs and health outcomes were discounted at 6% pa. RESULTS: For patients treated within 6 hours of stroke onset the probability of symptomatic intracranial haemorrhage (SICH) was 12.6% (rt-PA group) and 3.7% (placebo group). In the absence of SICH the probability of death within 90 days was 10.1% and 14.1% and the proportion of survivors who were disabled was 38.2% and 49.6% respectively. The model predicted that over 5 years from the incident stroke patients in the rt-PA group on average lost 0.04 life-years but gained 0.31 disability-free life-years and 0.13 QALYs. This resulted in cost savings of £2,500-£3,000 per treated patient, depending on how costs of initial acute care were included. Improvements in health outcome and cost savings were greater for the group treated with rt-PA within 3 hours of stroke onset. CONCLUSIONS: Trial results suggest that rt-PA is likely to be cost saving in patients treated within 6 hours of onset of acute stroke. A comprehensive analysis needs to consider the fixed costs and wider policy implications of introducing and maintaining a local thrombolysis service.