AEP “Click Detection” Failure: May Be, May Be Not!

To the Editor: In their article, Schmidt et al. (1) reported failure of the “Click Detection” algorithm (Alaris AEP™ monitor, SW version 1.5, Alaris Medical Systems Inc., San Diego, CA) to identify headphone disconnection during intraoperative auditory evoked potential (AEP) monitoring. They also proposed that the brainstem response (first 10 ms of the AEP), being unaffected by anesthetics, is useful in determining the quality of AEP signal. We believe they have misinterpreted the data. The ability to detect headphone disconnection using the present “Click Detection” algorithm was said to be poor during anesthesia and was somewhat better in the awake patient. However, there was no statistical test directly comparing the performance between the two groups. More interestingly, the reported sensitivity presumably refers to the proportion of patients registering “NO/LOW AEP” warning following headphone disconnection, was actually higher in the ANESTHESIA group compared with the AWAKE group throughout the study period. Their results suggested that “Click Detection” correctly indicated headphone disconnection in all anesthetized patients with a delay of 2 min, whereas 12% of the awake patients (n ≥ 2) were missed. It is even more puzzling as how the specificity and the receiver operating characteristics curve were derived in this study. In the present context, “specificity” at any given time implies the proportion of patients who are free from “NO/LOW AEP” warning (negative test) when the headphone is accurately placed (negative event). We believe it is impossible to calculate the specificity on the basis of data collected after headphone disconnection. Nevertheless, the long delay in identifying headphone disconnection was well known to the inventor (2). The algorithm has been revised and the current “noise window” has extended to include both the middle latency and the brainstem AEP (0–80 ms). This was implemented in the latest version of the monitor (AEP monitor/2, Danmeter A/S, Odense, Denmark) and was released in January 2003. Similar to previous version, the signal-to-noise ratio (SNR) is calculated by comparing the maximum peak-to-trough amplitude of the synchronous to asynchronous averaging of N sweeps (3). In the presence of auditory stimuli, the synchronized averages add together to produce a signal of large amplitude. On the other hand, the asynchronized averages cancel out each other resulting in smaller peaks (usually a fifth of the signal). Therefore, the amplitude of asynchronized averages is a measure of the background noise. Currently, a signal with SNR < 1.45 is considered as contaminated. The “AEP” signal bar will be replaced by an “EEG” message indicating the signal is inadequate for calculation of autoregressive AEP index (AAI). The subsequent AAI is based on β ratio and the extent of burst suppression (4). To determine the efficacy of the revised signal-rating algorithm, we have repeated the headphone disconnection experiment in 20 healthy women, aged 23–52 yr, and undergoing elective laparoscopic surgery. The study was approved by the Clinical Research Ethics Committee, and all patients gave written informed consents. We measured the changes in AAI values before and after headphone removal, when the patients were awake and after anesthesia with propofol 3 μg/mL and remifentanil 5 ng/mL. The status of the signal quality bar was also recorded. When the patient is awake, headphone disconnection resulted in a significant decrease in mean (±SD) AAI from 89 (±8) to 61 (±4) (analysis of variance with repeated measures, F = 84.4; P = 0.001). We believe this is partly related to the removal of “acoustic startle response” (5). There was no significant change in AAI values during anesthesia following headphone disconnection (F = 1.4; P = 0.81). Nonetheless, when headphone was removed, the calculation of AAI is based purely on the electroencephalographic parameters. The clinical correlation of this index remains to be defined (3). The mean (95% confidence intervals) time to the first AEP signal quality warning during anesthesia, 51 s (CI, 48–53), was similar to the awake recordings, 50 s (CI, 48–52) (Fig. 1; log-rank test, P = 0.47). Our data suggested that the revised algorithm is capable to detect headphone disconnection in both the awake and anesthetized patients with shorter delay. Clearly, this response time is dependent on the SNR cutoff value. Thus, a high cutoff value results in early recognition of “poor” signals. However, it will reject many “good” records at the same time. Currently, there is no agreement on the optimal SNR cutoff value.Figure 1.: Kaplan-Meier incidence curve of time from headphone disconnection to auditory evoked potential (AEP) signal warning.Rating of AEP signals has been a difficult task (6). We believe the current algorithm is a useful tool to the anesthetists who is inexperience in picking AEP artifacts. When the signal quality warning comes up, the attending anesthetists should exclude poor electrode contact, electromyographic interference and misplaced headphone. Matthew T. V.