BACKGROUND: Among patients with substantial carotid artery narrowing but no recent neurological symptom (stroke or transient ischaemia), the balance of surgical risks and long-term benefits from carotid endarterectomy (CEA) was unclear. METHODS: During 1993-2003, 3120 asymptomatic patients with substantial carotid narrowing were randomised equally between immediate CEA (half got CEA by 1 month, 88% by 1 year) and indefinite deferral of any CEA (only 4% per year got CEA) and were followed for up to 5 years (mean 3.4 years). Kaplan-Meier analyses of 5-year risks are by allocated treatment. FINDINGS: The risk of stroke or death within 30 days of CEA was 3.1% (95% CI 2.3-4.1). Comparing all patients allocated immediate CEA versus all allocated deferral, but excluding such perioperative events, the 5-year stroke risks were 3.8% versus 11% (gain 7.2% [95% CI 5.0-9.4], p<0.0001). This gain chiefly involved carotid territory ischaemic strokes (2.7% vs 9.5%; gain 6.8% [4.8-8.8], p<0.0001), of which half were disabling or fatal (1.6% vs 5.3%; gain 3.7% [2.1-5.2], p<0.0001), as were half the perioperative strokes. Combining the perioperative events and the non-perioperative strokes, net 5-year risks were 6.4% versus 11.8% for all strokes (net gain 5.4% [3.0-7.8], p<0.0001), 3.5% versus 6.1% for fatal or disabling strokes (net gain 2.5% [0.8-4.3], p=0.004), and 2.1% versus 4.2% just for fatal strokes (net gain 2.1% [0.6-3.6], p=0.006). Subgroup-specific analyses found no significant heterogeneity in the perioperative hazards or (apart from the importance of cholesterol) in the long-term postoperative benefits. These benefits were separately significant for males and females; for those with about 70%, 80%, and 90% carotid artery narrowing on ultrasound; and for those younger than 65 and 65-74 years of age (though not for older patients, half of whom die within 5 years from unrelated causes). Full compliance with allocation to immediate CEA or deferral would, in expectation, have produced slightly bigger differences in the numbers operated on, and hence in the net 5-year benefits. The 10-year benefits are not yet known. INTERPRETATION: In asymptomatic patients younger than 75 years of age with carotid diameter reduction about 70% or more on ultrasound (many of whom were on aspirin, antihypertensive, and, in recent years, statin therapy), immediate CEA halved the net 5-year stroke risk from about 12% to about 6% (including the 3% perioperative hazard). Half this 5-year benefit involved disabling or fatal strokes. But, outside trials, inappropriate selection of patients or poor surgery could obviate such benefits.