RESUMO
Intraoperative monitoring is needed to identify accurately those patients in need of a shunt during carotid endarterectomy. EEG can be used for this purpose, but there is no consensus on the variables to use. Using a database consisting of 149 EEGs recorded from patients during carotid endarterectomy under isoflurane (n=61) or propofol (n=88) anesthesia and who did or did not receive a shunt, the authors investigated which of 16 derivations (common reference, Cz) and 12 parameters (relative and absolute powers and spectral edge frequencies [SEFs]) singly or in combination could best distinguish between the shunt and the nonshunt groups for the two anesthesia regimens. Receiver operating characteristic curves were used to select derivation/parameter combinations for three types of trend computation: (1) values of relative powers and SEFs during clamping (C) only, (2) clamp minus preclamp (baseline) differences (C-B), and (3) C-B differences in absolute logarithmic power (DeltalogP). For both anesthesia regimens, C-B computation distinguished best between the shunt and nonshunt groups. For isoflurane anesthesia, SEF parameters were the best, and for propofol anesthesia the relative power parameters. Discriminant analysis, in which additional derivation/parameter combinations were added, increased the discriminative power of the DeltalogP computation but not of the C or C-B computations. For isoflurane anesthesia, SEF 90% was the best single parameter for distinguishing between patients who did and did not need a shunt and the four best derivations were F3-Cz, P4-Cz, C4-Cz, and F7-Cz. For the propofol anesthesia, the relative power (C or C-B computations) of the delta band was the best and the four best derivations were F8-Cz, T4-Cz, C4-Cz, and F4-Cz.
