BACKGROUND Stroke is an important contributor to perioperative morbidity and mortality associated with carotid endarterectomy (CEA). This investigation was designed to compare the performance of the INVOS-3100 cerebral oximeter to neurologic function, as a means of detecting cerebral ischemia induced by carotid cross-clamping, in patients undergoing carotid endarterectomy with cervical plexus block. METHODS Ninety-nine patients undergoing 100 CEAs with regional anesthesia (deep or superficial cervical plexus block) were studied. Bilateral regional cerebrovascular oxygen saturation (rSO2) was monitored using the INVOS-3100 cerebral oximeter. Patients were retrospectively assigned to one of two groups: those in whom a change in mental status or contralateral motor deficit was noted after internal carotid clamping (neurologic symptoms; n = 10) and those who did not show any neurologic change (no neurologic symptoms; n = 90). Data from 94 operations (neurologic symptoms = 10 and no neurologic symptoms = 84) were adequate for statistical analyses for group comparisons. A relative decrease in ipsilateral rSO2 after carotid occlusion (calculated as a percentage of preocclusion value) during all operations (n = 100) was also calculated to determine the critical level of rSO2 decrease associated with a change in neurologic function. RESULTS The mean (+/- SD) decrease in rSO2 after carotid occlusion in the neurologic symptoms group (from 63.2 +/- 8.4% to 51.0 +/- 11.6%) was significantly greater (P = 0.0002) than in the no neurologic symptoms group (from 65.8 +/- 8.5% to 61.0 +/- 9.3%). Logistic regression analysis used to determine if a change in rSO2, calculated as a percentage of preclamp value, could be used to predict change in neurologic function was highly significant (likelihood ratio chi-square = 13.7; P = 0.0002). A 20% decrease in rSO2 reading from the preclamp baseline, as a predictor of neurologic compromise, resulted in a sensitivity of 80% and specificity of 82.2%. The false-positive rate using this cutoff point was 66.7%, and the false-negative rate was 2.6%, providing a positive predictive value of 33.3% and a negative predictive value of 97.4%. CONCLUSIONS Monitoring rSO2 with INVOS-3100 to detect cerebral ischemia during CEA has a high negative predictive value, but the positive predictive value is low.