Analysis of pooled data from the randomised controlled trials of endarterectomy for symptomatic carotid stenosis

Abstract

Endarterectomy reduces risk of stroke in certain patients with recently symptomatic internal carotid stenosis. However, investigators have made different recommendations about the degree of stenosis above which surgery is effective, partly because of differences between trials in the methods of measurement of stenosis. To accurately assess the overall effect of surgery, and to increase power for secondary analyses, we pooled trial data and reassessed carotid angiograms. We pooled data from the European Carotid Surgery Trial (ECST), North American Symptomatic Carotid Endarterectomy Trial, and Veterans Affairs trial 309 from the original electronic data files. Outcome events were re-defined, if necessary, to achieve comparability. Pre-randomisation carotid angiograms from ECST were re-measured by the method used in the other two trials. Risks of main outcomes in both treatment groups and effects of surgery did not differ between trials. Data for 6092 patients, with 35000 patient-years of follow-up, were therefore pooled. Surgery increased the 5-year risk of ipsilateral ischaemic stroke in patients with less than 30% stenosis (n=1746, absolute risk reduction -2.2%, p=0.05), had no effect in patients with 30-49% stenosis (1429, 3.2%, p=0.6), was of marginal benefit in those with 50-69% stenosis (1549, 4.6%, p=0.04), and was highly beneficial in those with 70% stenosis or greater without near-occlusion (1095, 16.0%, p<0.001). There was a trend towards benefit from surgery in patients with near-occlusion at 2 years' follow-up (262, 5.6%, p=0.19), but no benefit at 5 years (-1.7%, p=0.9). Re-analysis of the trials with the same measurements and definitions yielded highly consistent results. Surgery is of some benefit for patients with 50-69% symptomatic stenosis, and highly beneficial for those with 70% symptomatic stenosis or greater but without near-occlusion. Benefit in patients with carotid near-occlusion is marginal in the short-term and uncertain in the long-term.