OBJECTIVETo compare the accuracies of cerebral state index (CSI) and bispectral index (BIS) in sedation monitoring during target control infusion of midazolam.

METHODSTwenty informed adult male volunteers were intravenously administered with midazolam through plasma target control infusion from 30ng/ml (in increments of 10ng/ml every time) until they became unresponsive to tactile stimulation (i. e., mild prodding or shaking). The BIS and CSI were continuously recorded simultaneously. Sedation was assessed using the Observers' Assessment of Alertness/Sedation (OAA/S) scale at each time when Ct equaled to Ce. The electroencephalogram (EEG) parameters were correlated with the OAA/S scores using nonparametric Spearman's correlation analysis. The prediction probabilities were calculated at the points of lost of verbal contact (LVC) and lost of responses to stimulus (LOR). BIS05, BIS50, BIS95, and CSI05, CSI50, CSI95 were also calculated for LVC and LOR.

RESULTSBIS and CSI were well correlation with OAA/S scales during both the onset and recovery phases. When the sedation level increased, BIS and CSI progressively decreased. The prediction probabilities of BIS and CSI were 84%, 74% for LVC and 79%, 68% for LOR, while the BIS05, BIS50, and BIS95 as well as CSI05, CSI50, and CSI95 were 85.5, 60.6, and 35.7 (for BISs) and 82.2, 65.2, and 30.3 (for CSIs) at the point of LVC and 79.7, 47.6, and 15.6 (for BISs) and 75.9, 43.4, and 11 (for CSIs) at the point of LOR.

CONCLUSIONSBoth CSI and BIS seem to be useful parameters for assessing midazolam-induced sedation. BIS is superior in the prediction of LVC and LOR.