Warfarin and absolute risk of hemorrhagic stroke

We read with great interest the CMAJ article by Gomes and colleagues,1 which highlights the importance of quantifying the risk of bleeding episodes associated with warfarin now that new therapeutic options are becoming available.

We have just completed a population-based cohort study using the Health Improvement Network, a primary care database from the United Kingdom, aimed at estimating the incidence rate of hemorrhagic stroke in the general population.2 We identified individuals aged 20 to 89 years between 2000 and 2008 and followed them for an average of 5.8 years. Overall, we ascertained 1797 cases of intracerebral hemorrhage and 1340 cases of subarachnoid hemorrhage. We wanted to explore if we could reproduce the estimates of hemorrhagic stroke risk among new users of warfarin reported by Gomes and colleagues using our data from the UK.

We found that, among those aged 65 years and older who were prescribed warfarin for the first time, the incidence rate of hemorrhagic stroke after starting warfarin was 294.6 cases per 100 000 person-years (95% CI [confidence interval] 118.5–606.1) within the first 30 days and 165.6 cases per 100 000 person-years (95% CI 136.6–199.2) during the overall follow-up. These figures correspond to annual rates of 0.29% per person-year for the 30-day period, and 0.17% per person-year for the overall follow-up period. Both estimates are extremely similar to those reported by Gomes and colleagues1 for intracranial hemorrhage: 0.4% for the initial 30 days and 0.2% overall.

In conclusion, we have been able to reproduce similar estimates of absolute risk to those reported by Gomes and colleagues1 in a different population (UK) when using an analogous design despite different sources of information. Along with their study, our results serve to underline the importance of performing this type of study in a population-based setting, which mimics the real-life scenario of a clinician better than estimates based on clinical trial data. It is therefore important that newer anticoagulant drugs that have emerged in recent years should also be subjected to population-based studies performed with valid sources of information and valid methods in years to come.