Observational studies of new treatments in routine practice are clinically important but may be limited by bias. We used a Bayesian approach to interpret and sequentially combine phase 4 studies of IV-TPA within 3 h for acute ischaemic stroke to quantify the cumulative evidence for the efficacy and safety of this therapy in clinical practice. Prior probability distributions for favourable outcome (modified Rankin Score, mRS, 0-1), symptomatic intracerebral haemorrhage, and mortality 3 months after IV-TPA were derived from the NINDS trial. Phase 4 studies from observational case series and from regulator-mandated large registries were included. A cumulative analysis was performed to quantify the increase in the total evidence base over time, unadjusted and adjusted for potential bias. The cumulative analysis indicated that IV-TPA <3 h was associated with 3-month favourable outcome in 37.1% (95% credible interval, CrI, 36.1-38.0%, n = 9,578) compared to 26% (95% confidence interval, CI, 19.6-32.9%, n = 165) in placebo-treated patients in NINDS, symptomatic intracranial haemorrhage in 6.6% (95% CrI 5.9-7.4%, n = 10, 834) compared to 6.4% (95% CI 4.0-9.4%) in the NINDS IV-TPA group, and 3-month mortality in 13.6% (95% CrI 12.6-14.8%, n = 9, 901) compared to 20.5% (95% CI 16.0-25.0%, n = 312) in the NINDS placebo group. A Bayesian approach provides further confirmatory evidence of the efficacy and safety of IV-TPA for treatment of acute ischaemic stroke within 3 h in diverse clinical practice settings, after adjusting for potential observational bias. Copyright (c) 2009 S. Karger AG, Basel.