Background: We compared two spectral entropies, state entropy (SE) and response entropy (RE), based on the irregularity of the EEG, to measure loss of response to verbal command (LOR(verbal)) and noxious stimulus (LOR(noxious)) with the bispectral index (BIS) during propofol infusion with and without remifentanil.
Methods: Three groups of 20 patients received an effect-site controlled propofol infusion (Ce(PROP)) starting at 1 microg ml(-1) and increased in steps of 0.5 microg ml(-1) at 4 min intervals. In addition, a remifentanil infusion was maintained at a group-dependent, fixed effect-site target concentration (Ce(REMI)) (0, 2 or 4 ng ml(-1)). The ability of BIS, SE or RE to predict LOR(verbal) and LOR(noxious) were compared with the changes in BIS, SE and RE using logistic regression, prediction probability (P(K)), and sensitivity/specificity.
Results: In all groups, BIS, SE and RE decreased with increasing Ce(PROP). However, BIS decreased more smoothly than SE and RE at deeper levels of sedation. At LOR(verbal), BIS(50), SE(50) and RE(50) increased with increasing Ce(REMI). BIS, SE and RE all detected LOR(verbal) accurately but BIS performed better at 100% sensitivity. Sensitivity/specificity for detection of LOR(verbal) decreased for all methods with increasing Ce(REMI). LOR(noxious) was poorly described by all measures.
Conclusion: LOR(verbal) was detected accurately by BIS, SE and RE except for 100% sensitivity, where BIS performed better. Though BIS, SE and RE were influenced by remifentanil during propofol administration, their ability to detect LOR(verbal) remained accurate. None of the measures predicted LOR(noxious).