London, United Kingdom 2013 7 Meta-analysis and reviews 17:30 - 17:40 Reliability and feasibility of ischemic stroke classification systems for large epidemiologi-cal Cerebrovasc Dis 2013; 35 (suppl 3)1-854 159 studies R. Woodfield1, K. Rannikmae2, I . Grant3, C. Sudlow4 Division of Clinical Neurosciences, Centre for Clinical Brain Sciences, University of Ed-inburgh, Edinburgh, UNITED KINGDOM1,Division of Clinical Neurosciences, Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UNITED KINGDOM2, Informa-tion and Statistics Division, National Services Scotland, Edinburgh, UNITED KINGDOM3, Di-vision of Clinical Neurosciences, Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UNITED KINGDOM4 Introduction: We sought the best approach for sub-classifying 1000s of ischaemic stroke out-comes in a very large prospective cohort study (UK Biobank). An ideal classification system should maximise assignment to a specific subtype without compromising reliability. Methods: We systematically assessed published studies of inter-rater reliability of ischaemic stroke classification systems. We extracted data from included studies on: classification meth-ods; study setting; patient and rater characteristics; inter-rater reliability; proportion of cases unassigned to a specific subtype. We also classified ~ 600 ischaemic stroke cases in our stroke register with the three most widely studied systems: TOAST (Trial of Org 10172 in Acute Stroke Treatment) and CCS (Computerised Classification System) systems require detailed in-formation from investigations to assign a presumed stroke mechanism; the OCSP (Oxfordshire Community Stroke Project) system is based on presumed vascular anatomy, providing less in-formation on stroke mechanism, but requiring far fewer investigations. Results: Among 20 included studies in our systematic review, inter-rater reliabilities (kappa) were: 0.42-0.95 for TOAST (11 studies), 0.41-0.70 for OCSP (5 studies), and 0.70-0.90 for CCS (4 studies). Reliability improved with data abstraction protocols, computerised algo-rithms, reduced system complexity, and increased rater confidence. The proportions of cases unassigned to a single subtype were 0% for OCSP, 26% for CCS, and 40% for TOAST. In the systematic review, CCS was assessed mainly in the US setting that it was developed in. It per-formed less well in our UK based cases (kappa 0.56, 67% unassigned). Conclusions: Study characteristics other than classification system account for much of the variation in reliability. Systems may perform less well outside the setting they were developed in. These factors should be taken into account in developing optimal systems for large epidemi-ological studies. 6 Meta-analysis and reviews 17:20 - 17:30 Predicting risks and benefits of treatment with aspirin in the acute stage of ischaemic stroke: an individual patient data meta-analysis of 39,166 patients in 3 large randomised controlled trials W.N. Whiteley1, L. Candelise2, Z. Chen3, P.A.G. Sandercock4 University of Edinburgh, Edinburgh, UNITED KINGDOM1,Università degli Studi di Mila-no, Milan, Italy, Milan, UNITED KINGDOM2, Clinical Trial Service Unit and Epidemiological Studies Unit, University of Oxford, Oxford, Oxford, UNITED KINGDOM3, University of Ed-inburgh, Edinburgh, UNITED KINGDOM4 BACKGROUND Aspirin modestly reduces death or dependence after ischaemic stroke. We investigated wheth-er targeting aspirin on acute ischaemic stroke patients with a high risk of recurrent thrombotic events or a low risk of haemorrhagic events led to increased absolute benefits. METHODS We obtained individual patient data from 3 large randomised controlled trials of aspirin ver-sus control in stroke (IST-1, CAST, MAST-1). We developed and evaluated statistical models to predict thrombotic events (DVT, PE, MI or recurrent ischaemic stroke) and haemorrhagic events (intracranial or gastrointestinal haemorrhage) at 14 days after stroke in IST-1. For sim-plicity in the initial analyses, we ignored allocation to heparin or streptokinase. We calculated the absolute risk reduction of death or dependence due to aspirin in patients grouped by quar-ters of the predicted risk of haemorrhagic or thrombotic events with random effects meta-analy-sis RESULTS We used data from 39,166 patients randomly allocated to aspirin or control, of whom 18,020 were dead or dependent at final follow up. Prediction models built with all baseline variables discriminated modestly between patients with and without recurrent thrombotic events (AUC 0.57,95%CI:0.57-0.57) and haemorrhagic events (0.59, 0.58-0.59) in IST-1. There was no ev-idence of greater benefit of aspirin in patients with lesser risk of haemorrhagic events or in pa-tients with a greater risk of thrombotic events. The best estimate of the effect of aspirin in each risk group was the overall absolute risk reduction of death or dependence of 1% (95%CI:0.00- 0.02). CONCLUSIONS There was no evidence that targeting aspirin to acute ischaemic stroke patients with a high predicted risk of thrombotic events or a low predicted risk of haemorrhagic events would lead to improvements in death or dependence after ischaemic stroke. The modest absolute benefit of aspirin is similar in patients with different predicted risks of haemorrhagic or thrombotic events.
Karger_ESC London_2013
To see the actual publication please follow the link above