BACKGROUND: Carotid endarterectomy (CEA) has been an evidence-based treatment for symptomatic severe carotid stenosis since 1991. Surgical techniques and patient selection have changed over the years. The results of CEA in a single centre over a 23-year period were reviewed. METHODS: Prospectively gathered preoperative, operative, postoperative and long-term follow-up data were analysed. Routine intraoperative shunting and patch closure has been used since 1988. Data were analysed using the chi2 test or by logistic regression, adjusting for age at operation and date of operation. RESULTS: Five hundred and seventy-three CEAs (37 bilateral and three repeat procedures) were carried out on 533 patients. Trainees performed an increasing proportion of CEAs from 1996 to 1998 (15, 50 and 56 per cent respectively). The perioperative death rate was 0.8 per cent and the rate of any perioperative neurological deficit was 6.9 per cent. Other causes of morbidity included nerve injury (5.1 per cent) of which the commonest was to the hypoglossal nerve (2.7 per cent). During follow-up (median 4 (range 0-22) years) there were 81 neurological events (15.9 per cent) which included 35 ipsilateral (6.6 per cent) and 18 contralateral (3.4 per cent) strokes. There was no significant difference in outcome for grade of surgeon, intraoperative shunting or patch closure. Major causes of death were cardiac death (74; 14.6 per cent) followed by stroke (23; 4.5 per cent) and cancer (20; 3.9 per cent). CONCLUSION: The introduction of routine intraoperative shunting and patch closure, as well as allowing surgical trainees to perform supervised CEAs, has not affected perioperative morbidity and mortality rates or long-term outcome.