Abstract
BACKGROUND: S100B protein is a potential biomarker of central nervous system insult. This study quantitatively compared two methods for assessing serum concentration of S100B. METHODS: A prospective, observational study performed in a single tertiary medical center. Included were fifty two consecutive adult patients undergoing surgery for meningioma that provided blood samples for determination of S100B concentrations. Eighty samples (40 pre-operative and 40 postoperative) were randomly selected for batch testing. Each sample was divided into two aliquots. These were analyzed by ELISA (Sangtec) and a commercial kit (Roche Elecsys(R)) for S100B concentrations. Statistical analysis included regression modelling and Bland-Altman analysis. RESULTS: A parsimonious linear model best described the prediction of commercial kit values by those determined by ELISA (y = 0.045 + 0.277*x, x = ELISA value, R2 = 0.732). ELISA measurements tended to be higher than commercial kit measurements. This discrepancy increased linearly with increasing S100B concentrations. At concentrations above 0.7 microg/L the paired measurements were consistently outside the limits of agreement in the Bland-Altman display. Similar to other studies that used alternative measurement methods, sex and age related differences in serum S100B levels were not detected using the Elecsys(R) (p = 0.643 and 0.728 respectively). CONCLUSION: Although a generally linear relationship exists between serum S100B concentrations measured by ELISA and a commercially available kit, ELISA values tended to be higher than commercial kit measurements particularly at concentrations over 0.7 microg/L, which are suggestive of brain injury. International standardization of commercial kits is required before the predictive validity of S100B for brain damage can be effectively assessed in clinical practice.