Can electroencephalographic analysis be used to determine sedation levels in critically ill patients?

Prolonged use of sedative drugs frequently leads to oversedation of intensive care patients. Clinical assessment scales are not reliable in deeply sedated patients. Parameters obtained from spectral and bispectral analysis of electroencephalogram (EEG) records have been combined to create an index (BIS) to monitor anesthesia depth. The role of such parameters in monitoring the depth of the sedation in intensive care unit (ICU) patients has yet to be determined. We designed the present prospective study to redefine and calculate available spectral and bispectral parameters from raw EEG records and estimate their clinical relevance for the diagnosis of under- or oversedation levels in ICU patients. Forty adult patients receiving continuous midazolam and morphine sedation were included. We obtained 167 clinical evaluations of sedation level using Ramsay and COMFORT scales along with an EEG record of 300 s. Six spectral parameters-relative power of 4 frequency bands (beta, alpha, Theta, and delta), 95th percentile of the power spectrum (SEF95), and 50th percentile of the power spectrum (SEF50) and four bispectral parameters, real triple product, bispectrum (Bispectrum), bicoherence, and ratio 10-were calculated. The relevance of each of these parameters and combinations in predicting too light (Ramsay 1 and 2) or deep (Ramsay 5 and 6) sedation levels was assessed. These calculations were performed before and after exclusion of the agitated patients, whose COMFORT 4 score was above 2. The most relevant parameters for predicting levels of deep sedation (Ramsay 5 and 6) were ratio 10 (area under the curve = 0.763; 95% confidence interval, 0.679-0.833) and SEF95 (area under the curve = 0.687; 95% confidence interval, 0.597-0.767). The most relevant parameters for predicting light levels of sedation (Ramsay 1 and 2) were also ratio 10 (area under the curve = 0.829; 95% confidence interval, 0.695-0.917) and SEF95 (area under the curve = 0.798; 95% confidence interval, 0.650-0.898). There is a modest improvement in relevance of their linear combination in predicting sedation level. Results were similar after exclusion of agitated patients. We conclude that various calculated EEG descriptive parameters exhibited large interindividual variability. There was a strong correlation between EEG spectral and bispectral parameters. Bispectral analysis slightly improves the predictive power of simple spectral analysis in distinguishing too light or deep sedation levels in ICU patients.

Spectral edge frequency 95 and Ratio 10 are the most relevant electroencephalogram (EEG) indexes for monitoring the level of sedation in intensive care unit patients but calculated EEG values exhibited large interindividual variability. Bispectral analysis of EEG provides a slight improvement over simple spectral analysis.