Assessment of inadequacy of anesthesia: A European multi-center approach

No single parameter has been denned for assessment of depth of anesthesia. Numerous studies have investigated calculated parameters of the electroencephalogram (EEC) and somatosensory (SEP) or auditory evoked responses (AEP) as measures of depth of anesthesia. The goal of this European multi-center study was to evaluate EEC and middle latency AEP parameters (MLAEP) which may predict movement or hemodynamic responses following skin incision during isoflurane/nitrous oxide anesthesia. Following IRB approval 106 ASA I-II patients undergoing abdominal surgery were enrolled in this study. 7.5 mg midazolam was given l h before anesthesia. Anesthesia was induced with 1.5 mg propofol, 0.1 mg fentanyl and 1.5 mg/kg succinylcholine. The lungs were intubated. Using overpressure, the patient was rapidly equilibrated to 0.6% end expiratory isoflurane in 50% nitrous oxide. Mechanical ventilation was set to maintain end-tidal PCO₂ of 35-38 mmHg. Single EEC was recorded from electrodes placed at the forehead and linked mastoids (sample rate: 1 kHz, bandpass: 0.5-1000 Hz). Impedance was kept below 5 MOhm. After amplification and digitization all data were stored on disk. Binaural auditory stimuli were applied at a rate of 6.8 Hz. Mean arterial blood pressure, heart rate, end-tidal PCÜ2, and temperature were recorded every minute. Coughing, movements and tears were also noted. All data were recorded from 10 min before until 10 min after skin incision. Parameters of the EEG power spectra and amplitudes/latencies of AEP components Pa and Nb were analyzed before and after skin incision. Fourier transformation was used for calculation of classical power spectrum densities. In addition EEG data were subjected to Bispectral EEG analysis. Discriminant analysis based on the results of this analysis was performed to investigate if patients who moved on skin incision show a higher phase coupling than spectra from non-movers. In addition, MLAEP were subjected to a wavelet transform (Gabor wavelet) to detect time-dependent changes in frequencies of the non-stationary signals. Following this, the calculated parameter set was classified using a neural net (Kohonen net). Mann-Whitney U and Wilcoxon signed rank tests were used for statistical analysis. P < 0.05 was considered significant. Fifty-two patients (49%) moved after skin incision and 44% showed increases in heart rate and/or blood pressure. No significant correlations between changes in hemodynamics (> 20%) and electrophysiologic parameters were found. Also, no changes related to skin incision were found for median frequency, spectral edge frequency, power densities in delta, theta or beta bands. In patients who moved after skin incision alpha power was significantly increased (p < 0.05). Calculated over all patients MLAEP amplitudes and latencies did not change following skin incision. However, in movers PaNb amplitude was significantly increased. Using bispectral EEG 125 frequency combinations were analyzed. Significant phase couplings (p < 0.001) for discrimination between movers and non-movers were calculated for activities in the 3-8 Hz and 3-11 Hz bands. In addition, 6 bicoherence variables and 15 frequency combinations have higher mean values (p < 0.05) for movers when compared to non-movers. The results of the discriminant analysis shows that based on the EEC before skin incision a correct classification of patients who will move on skin incision can be achieved in 77% of all cases. Six cases were classified in the wrong category. However, in only 2 (7.7%) cases movement occurred when no movement was predicted ('worst case'). Classification of MLAEP based on the wavelet-transform and neural net analysis was correct in approximately 80% for prediction of movers and non-movers. In the present study a prediction of inadequate anesthesia (move, hemodynamic response) could not be performed by analysis of 'classical' EEC and MLAEP parameters. However, predictive values for movers versus non-movers could be achieved in approximately 80% by bispectral EEG analysis and MLAEP data subjected to wavelet-transform and classified by neural net. Our data indicate that these mathematical tools hold promise for a definition of measures which may predict inadequate anesthesia for skin incision.