This prospective clinical study was conducted to determine whether S 100 B is a reliable serum marker for traumatic brain injury (TBI) with and without multiple trauma. Fifty-five trauma patients (Injury Severity Score [ISS] > or = 24 and Glasgow Coma Score [GCS] < or = 8) were classified by radiography, computer tomography, ultrasound, and neurology as TBI without multiple trauma (n = 23), TBI with multiple trauma (n = 23), or multiple trauma without TBI (n = 9). S 100 B was measured initially after trauma and daily for a maximum of 21 days. Both survivors and nonsurvivors had markedly increased S 100 B initially. All survivors returned to normal or moderately increased S 100 B levels within the first 48 h after trauma. In contrast, all nonsurvivors of isolated TBI had S 100 B values that either increased consistently or dropped and then increased again 48 h after the initial increase after trauma. There was no relationship between localization, extent, or severity of TBI and S 100 B. According to receiver operating characteristic curve analysis and calculation of the area under the curve (AUC), S 100 B is equally accurate for mortality prediction at 24, 48, and 72 h after trauma and is most accurate >84 h after trauma. Sensitivity/specificity for mortality prediction are more accurate in TBI without multiple trauma (AUC 0.802-0.971) than in TBI with multiple trauma (AUC 0.693-0.783). Thus, though S 100 B may be a reliable marker of brain damage in TBI without multiple trauma 24 h after trauma and thereafter, it appears to be less reliable in TBI with multiple trauma.