Discussion
The main finding of this meta-analysis was an increased risk of SICH in patients with acute ischaemic stroke and prior warfarin therapy with a subtherapeutic INR after treatment with intravenous and/or intra-arterial thrombolysis. The risk of SICH was more than two times higher in patients on warfarin therapy compared with those without. However, there was no evidence that prior warfarin therapy was associated with (un)favourable outcome or an increase in death from all causes.
Most of the included studies concerned patients treated with intravenous rtPA, and few studies were done in patients receiving endovascular treatment. A reanalysis of the MERCI (Mechanical Embolus Removal in Cerebral Ischemia) and multi-MERCI trials demonstrated that mechanical thrombectomy in patients with acute ischaemic stroke and prior warfarin use with an INR larger than 1.7 was also associated with SICH.
During acute cerebral ischaemia the blood brain barrier becomes damaged, which can lead to spontaneous haemorrhagic transformation. Disruption of the blood brain barrier is influenced by several proteolytic enzymes, such as matrix metalloproteinases (MMP). In acute ischaemic stroke, rtPA treatment is associated with an upregulation of MMP-9 and may therefore enhance the risk of SICH. Reperfusion also enhances the risk of SICH. Animal experiments have demonstrated that warfarin increases the risk of haemorrhagic transformation in acute ischaemia, which was highly dependent on reperfusion. So, a combination of reperfusion injury, upregulation of MMP-9 and strong anticoagulant effect of warfarin might explain the observed increased risk of SICH. The same effect has been observed in patients treated with antiplatelets before rtPA treatment.
On the other hand, a common problem after thrombolytic therapy is the occurrence of early reocclusion. Some have hypothesised that recanalisation rates with warfarin may be better than without warfarin, although this idea was not confirmed by two of the included studies (Table 2). A topic of interest is whether newer anticoagulants, such as dabigatran etexilate, also increase the risk of SICH. We are only aware of a single case report of an acute stroke patient on dabigatran treated with intravenous rtPA who did not develop haemorrhagic complications.
This systematic review has limitations. First, there were some small differences in the definition of SICH, which could influence the SICH rates. Moreover, the meta-analysis regarding SICH revealed a significant heterogeneity between the studies. Most studies however used a definition of intracranial haemorrhage with an NIHSS score increase of four or more points. In the sensitivity analysis, excluding the study that used another definition of SICH, the results remained the same; therefore, it seems unlikely that these differences influenced the final results. Second, we could not perform multivariate analyses with adjustment for possible confounders. Patients with warfarin were older and, as a consequence, may have had more often leucoaraiosis. Both age and leucoaraiosis may also have contributed to an increased risk of SICH. Patients with warfarin had also more often cardioembolic strokes. It is known that cardioembolic strokes are associated with larger diffusion weighted imaging lesions, which is also a strong predictor of SICH after intravenous rtPA. Other possible confounders, such as early ischaemic changes on brain CT or blood pressure values, were not included in our analyses. These imbalances can confound the relation between prior warfarin and SICH. Third, we did not have data about the INR levels from patients with and without SICH. It could be possible that there is a cut-off point for the INR at which the risk of SICH increases. Results from our centre suggest that an elevated INR >1.2 was associated with SICH. Vergouwen et al found no association between the INR level and occurrence of SICH. Fourth, only two studies documented recanalisation rates. It would be interesting to evaluate the degree of recanalisation in patients with and without prior warfarin in a larger sample size. Fifth, it may be possible that, due to publication bias, the risks of haemorrhagic complications after thrombolytic therapy in patients with warfarin treatment were underestimated.
In summary, this study suggests that patients on warfarin therapy, but with subtherapeutic INR, have an increased risk of SICH after thrombolytic therapy. However, we did not observe an association between prior warfarin therapy and a worse functional outcome or increase in death from all causes. Further studies should focus at which INR cut-off score the risk of SICH increases.