Objective: Spectral EEG analysis has been claimed to reduce subjective vari
ability in EEG assessment of hepatic encephalopathy and to allow the detect
ion of mild encephalopathy.
Method: To test such assumptions, 43 digital EEG were recorded in 32 cirrho
tics without overt encephalopathy or with grade 1 overt encephalopathy; 7 p
atients were re-tested (2-5 times) in their follow up. All patients underwe
nt psychometric assessment. Nineteen controls were considered. EEG were bli
ndly evaluated by two electroencephalographers and by spectral EEG analysis
performed according to 3 different techniques.
Results: The reliability of the classification based on spectral analysis (
biparietal technique) was higher than that based on a three-degree qualitat
ive visual reading (concordance/discordance = 58/4 versus 46/16 P < 0.01) a
nd comparable with that of semiquantitative visual assessment based on post
erior basic rhythm (concordance/discordance = 55/7 P = 0.5). The accuracy o
f spectral EEG analysis was higher than that of qualitative visual EEG read
ings alone (90 versus 75%) and comparable to semi-quantitative visual asses
sment (87%), however, statistical significance was not reached. In the foll
ow-up, the variations of theta and delta relative power were found to be si
gnificantly correlated with psychometric variations.
Conclusions: In conclusion, spectral EEG analysis may improve the assessmen
t of mild hepatic encephalopathy by decreasing interoperator variability an
d providing reliable parameters correlated with mental status. (C) 1999 Els
evier Science Ireland Ltd. All rights reserved.