Severe narrowing (or stenosis) of the carotid artery is an important cause of stroke. Surgical removal of the atheromatous material from the inside of the carotid artery (endarterectomy) may reduce the risk of stroke, but carries a risk of operative complications.This review seeks to summarize the evidence from randomized trials on the balance of risks and benefits of carotid endarterectomy in adults with symptomatic carotid stenosis.We searched the Cochrane Stroke Group's Specialized Register of trials (date last searched: March 1999), supplemented by electronic searches of several databases.Randomized controlled trials comparing 'best medical treatment plus carotid endarterectomy' with 'best medical therapy' in patients with carotid stenosis and a recent transient ischaemic attack or nondisabling ischaemic stroke in the territory of that artery.Two reviewers independently selected the studies and extracted the data. An intention to treat analysis was performed.Data on death or disabling stroke were available from two trials, which included 5950 patients: the North American Symptomatic Carotid Endarterectomy Trial (NASCET), and the European Carotid Surgery Trial (ECST). The two trials used different methods to measure stenosis, but a simple formula can be used to convert between the two methods. For patients with severe stenosis (ECST > 80% = NASCET > 70%), surgery reduced the relative risk of disabling stroke or death by 48% (95% confidence interval [CI] 27 - 73%). The number of patients needed to be operated on (number needed to treat [NNT]) to prevent one disabling stroke or death over 2 to 6 years follow-up was 15 (95% CI 10 - 31). For patients with less severe stenosis (ECST 70 - 79% = NASCET 50 - 69%), surgery reduced the relative risk of disabling stroke or death by 27% (95% CI 15 - 44%). The number of patients needed to be operated on to prevent one disabling stroke or death was 21 (95% CI 11 - 125). Patients with lesser degrees of stenosis were harmed by surgery. Surgery increased the risk of disabling stroke or death by 20% (95% CI 0 - 44%). The number of patients needed to be operated on to cause one disabling stroke or death was 45 (95% CI 22 - infinity).Carotid endarterectomy reduced the risk of disabling stroke or death for patients with stenosis exceeding ECST-measured 70% or NASCET-measured 50%. This result is generalizable only to surgically-fit patients operated on by surgeons with low complication rates (less than 6%).