Bispectral index predicts deaths within 2 weeks in coma patients, a better predictor than serum neuron-specific enolase or S100 protein

Abstract

We assessed the ability of bispectral index (BIS) to predict clinical outcome (dead or alive within 2 weeks). In total, 90 coma patients with severe brain injuries underwent BIS monitoring, and serum neuron-specific enolase (NSE) and S100 protein levels were assayed within the first 3 days of admission. Receiver operator characteristic (ROC) curve analysis was used to assess the performance of BIS values for predicting death within 2 weeks. A cutoff value was calculated using the Youden index. A significant negative correlation was found between BIS value and serum NSE and S100 levels. The area under the curve for BIS value was 0.841 (p < 0.001, 95 % CI = 0.751-0.931), and higher than for NSE (0.713) (p = 0.002, 95 % CI = 0.582-0.844) or S100 (0.790) (p < 0.001, 95 % CI = 0.680-0.899). The optimal cutoff of BIS was 32.5. Serum NSE and S100 protein levels and the mortality rate were significantly lower in patients with a BIS value >32.5 than in patients with a BIS value ≤32.5. Bispectral index values may reflect degree of brain injury, and BIS is an objective and noninvasive monitoring method for helping clinicians to predict death in patients with a BIS value ≤32.5.