Targeting aspirin in acute disabling ischemic stroke: an individual patient data meta-analysis of three large randomized trials

Abstract

Background: Aspirin is of moderate overall benefit for patients with acute disabling ischemic stroke. It is unclear whether functional outcome could be improved after stroke by targeting aspirin to patients with a high risk of recurrent thrombosis or a low risk of haemorrhage.

Aims: We aimed to determine whether patients at higher risk of thrombotic events or poor functional outcome, or lower risk of major haemorrhage had a greater absolute risk reduction of poor functional outcome with aspirin than the average patient.

Methods: We used data on individual ischemic stroke patients from three large trials of aspirin vs. placebo in acute ischemic stroke: the first International Stroke Trial (n = 18,372), the Chinese Acute Stroke Trial (n = 20,172) and the Multicentre Acute Stroke Trial (n = 622). We developed and evaluated clinical prediction models for the following: early thrombotic events (myocardial infarction, ischemic stroke, deep vein thrombosis and pulmonary embolism); early haemorrhagic events (significant intracranial haemorrhage, major extracranial haemorrhage, or haemorrhagic transformation of an infarct); and late poor functional outcome. We calculated the absolute risk reduction of poor functional outcome (death or dependence) at final follow-up in: quartiles of early thrombotic risk; quartiles of early haemorrhagic risk; and deciles of poor functional outcome risk.

Results: Ischemic stroke patients who were older, had lower blood pressure, computerized tomography evidence of infarct or more severe deficits due to stroke had increased risk of thrombotic and haemorrhagic events and poor functional outcome. Prediction models built with all baseline variables (including onset to treatment time) discriminated weakly between patients with and without recurrent thrombotic events (area under the receiver operating characteristic curve 0·56, 95% CI:0·53-0·59) and haemorrhagic events (0·57, 0·52-0·64), though well between patients with and without poor functional outcome (0·77, 0·76-0·78) in the International Stroke Trial. We found no evidence that the net benefit of aspirin increased with increasing risk of thrombosis, haemorrhage or poor functional outcome in all three trials.

Conclusions: Using simple clinical variables to target aspirin to patients after acute disabling stroke by risk of thrombosis, haemorrhage or poor functional outcome does not lead to greater net clinical benefit. We suggest future risk stratification schemes include new risk factors for thrombosis and intracranial haemorrhage.

Keywords: individual patient data meta-analysis; prediction; randomized clinical trials; stratified treatment; stroke.