Research indicates that glial fibrillary acidic protein (GFAP), part of the astroglial skeleton, could be a marker of traumatic brain injury (TBI). S100B, an astroglial protein, is an acknowledged marker of TBI. Our goal was to analyze the relationship of GFAP/S100B to brain damage and outcome, and to compare the accuracy of GFAP/S100B for prediction of mortality after TBI. Our prospective study included 92 patients admitted <12 h after TBI (median injury severity score 25, median Glasgow Coma Scale 6). TBI was verfied by computerized tomography. GFAP/S100B were measured immunoluminometrically at admission and daily in the intensive care unit (average 10 days, range 1-21 days). We compared GFAP/S100B in non-survivors versus survivors, accuracy for mortality prediction according to receiver operated characteristic curve analysis, correlation between GFAP and S100B, relationship of GFAP/S100B to computerized tomography, cerebral perfusion pressure (CPP), mean arterial pressure (MAP) and 3-month Glasgow Outcome Score (GOS). GFAP (p < 0.005) and S100B (p < 0.0005) were higher in non-survivors than survivors. Both GFAP and S100B were accurate for mortality prediction (area under curve 0.84 versus 0.78 at <12 h after TBI). GFAP and S100B release correlated better later than 36 h after TBI (r = 0.75) than earlier (r = 0.58). GFAP was lower in focal lesions of <25 mL than in shifts of >0.5 cm (p < 0.0005) and non-evacuated mass lesions of >25 mL (p < 0.005). S100B was lower in focal lesions of <25 mL than in non-evacuated mass lesions (p < 0.0005) and lower in swelling than in shifts of >0.5 cm (p < 0.005). GFAP and S100B were lower in ICP < 25 than ICP > or = 25 (p < 0.0005), in CPP > or = 60 than CPP < 60 (p < 0.0005), in MAP > 70 than MAP < or = 70 mm Hg, and in GOS 4-5 than GOS 1 (p < 0.0005). Both measurement of GFAP and S100B is a useful non-invasive means of identifying brain damage with some differences based on the pattern of TBI and accompanying multiple trauma and/or shock.