BACKGROUND: The brain-derived protein S100B has been shown to be a useful marker of brain injury of different etiologies. Cognitive dysfunction after cardiac surgery using cardiopulmonary bypass has been reported to occur in up to 70% of patients. In this study we tried to evaluate S100B as a marker for cognitive dysfunction after coronary bypass surgery with cardiopulmonary bypass in a model where the inflow of S100B from shed mediastinal blood was corrected for. METHODS: 56 patients scheduled for coronary artery bypass grafting underwent prospective neuropsychological testing. The test scores were standardized and an impairment index was constructed. S100B was sampled at the end of surgery, hourly for the first 6 hours, and then 8, 10, 15, 24 and 48 hours after surgery. None of the patients received autotransfusion. RESULTS: In simple linear analysis, no significant relation was found between S100B levels and neuropsychological outcome. In a backwards stepwise regression analysis the three variables, S100B levels at the end of cardiopulmonary bypass, S100B levels 1 hour later and the age of the patients were found to explain part of the neuropsychological deterioration (r = 0.49, p < 0.005). CONCLUSIONS: In this study we found that S100B levels 1 hour after surgery seem to be the most informative. Our attempt to control the increased levels of S100B caused by contamination from the surgical field did not yield different results. We conclude that the clinical value of S100B as a predictive measurement of postoperative cognitive dysfunction after cardiac surgery is limited.