Intraoperative Compound Action Potentials as a Predictor of Postoperative Cortical Auditory Evoked Potentials in Cochlear Implant Users

Cochlear implants (CIs) are the standard of care for severe-to-profound sensorineural hearing loss. It is well documented that CIs can significantly improve hearing outcomes [1‒3]. Nevertheless, there is significant variation between CI users as to the degree of benefit they derive from their implants [4]. Identifying factors that predict hearing outcomes is advantageous for effective counseling and management of recipient expectations. Etiological and demographic factors, such as the age at implantation, have a strong influence on postoperative hearing outcomes [5]. Likewise, preoperative behavioral hearing tests also have some degree of predictive power for postoperative outcomes [6]. Another possible method for predicting hearing outcomes is using objective measures, which employ direct readouts of neural activity at different sites within the ascending auditory pathway. These measures are useful because they can be used with individuals for whom behavioral tests are challenging, such as young children or people with concomitant intellectual or cognitive impairments. These objective methods include compound action potentials (CAPs) and cortical auditory evoked potentials (CAEPs).

CAPS are neural responses generated by the auditory nerve. CAPS can be electrically evoked via the CI and are called electrically evoked compound action potentials (ECAPs). ECAPs do not require a patient’s conscious perception, and so can be elicited while they are under general anesthesia. As such, ECAPs provide perhaps the earliest objective measure of CI function, as they can be elicited during surgery. ECAPs also have the advantage in that they require no external equipment beyond that needed for typical CI programming, and current CI fitting software allows for swift and reliable stimulation and recording of ECAPs [7‒9]. The generation of an ECAP response indicates the successful transmission of auditory stimuli from the external CI processor to the auditory nerve. They are characterized by a negative peak which occurs between 0.2 and 0.4 ms poststimulation, followed by a positive peak at around 0.5 ms [10].

ECAP transmission is dependent on the integrity of the auditory nerve and the spiral ganglion neurons (SGNs) in the cochlea. Residual SGN density is vital because it is strongly correlated with word recognition word scores [11]. SGN density can be impacted by trauma induced during implantation. This includes translocation of the electrode array into the scala vestibuli, which is more common with perimodiolar array designs compared to lateral wall arrays [12]. However, surgical approach and array design may not fully explain variability of SGN density and survival. Histologic assessment of SGN density is difficult in clinical practice, as the implanted cochlea must be studied postmortem. There are several objective ECAP measurements associated with SGN density, namely, amplitude growth function (AGF) and threshold levels (TLs). These are discussed below.

The slope of the AGF may be used as a proxy measure for the density of functional SGN and could serve as a predictor of CI hearing outcomes. It has been demonstrated that a higher SGN density has been associated with steeper ECAP AGF slopes [13]. However, the literature is conflicting. Published studies differ on outcomes of factors influencing AGF including pediatric versus adult CI recipients, lateral wall versus perimodiolar array placement, differing surgical techniques, and apical versus basal cochlea regions of SGN density.

In the pediatric CI population, there are a number of studies demonstrating different outcomes. Research with CI users with fully inserted, lateral wall electrode arrays reported (1) moderate positive correlation between postoperative AGF slope of apical electrodes and monosyllabic word scores and (2) low-moderate negative correlations between postoperative AGF slopes of all electrode contacts and signal-to-noise ratio loss in speech understanding in noise results [14]. The authors suggested that greater AGF slopes in apical regions compared to medial or basal regions may be attributable to a higher density or survival rate of SGN in apical electrodes. Conversely, separate research has reported an absence of any significant relationship between intraoperative ECAP AGF slopes and speech performance in pediatric CI users, although surgical approach and electrode array design were not controlled for in the analysis [15]. Similarly, a recent study in an adult population found no correlation between intraoperative ECAP TLs or AGF slopes obtained through ART and AutoART with speech intelligibility measures at 6 months of postimplantation [16]. Despite differences in factors that may influence AGF slopes, it remains unclear why ECAP AGF slopes appear to predict CI speech performance in some studies and not others.

The link between ECAP TLs and CI outcomes has also been investigated. The relationship between ECAP TL and SGN density is weaker and more readily influenced by distance between the electrode and medial wall. However, given the ease of ECAP TL measurement relative to SGN density and electrode position, if ECAP TLs could predict CI speech outcomes, this would be extremely valuable.

In their study of 72 CI users aged 1–75 years, Wu et al. [17] found that lower postoperative ECAP TLs were (1) associated with improved speech and tone recognition, and (2) more common with perimodiolar electrode arrays than straight arrays. While lower intraoperative ECAP TLs were associated with lower postoperative ECAP TLs, no significant correlation was observed with speech outcomes and intraoperative ECAP TL. Similarly, Kim et al. [18] observed an inverse relationship between intraoperative ECAP TLs and categories of auditory performance scores at 12 months of postimplantation in a study of 71 adult and 151 pediatric implanted ears with a mix of both lateral wall and perimodiolar electrode arrays. In contrast, Basiony et al. [14] reported no statistically significant relationship between postoperative ECAP TLs and speech perception in 21 CI users aged 7–17 implanted with lateral wall electrode arrays. While ECAP TLs may be impacted by electrode array design [19], it does not fully account for the variation in the reported relationship between ECAP TLs and speech performance.

A systematic review of 25 studies which investigated relationships between ECAP parameters and hearing outcomes found considerable heterogeneity among these studies and concluded that there is, as yet, insufficient evidence for the efficacy of ECAPs as predictors of CI speech outcomes [20].

The varied results from research into ECAP measures and CI outcomes might be partially explained by ECAPs predominantly reflecting peripheral auditory function, thereby offering limited insight into centrally activated cortical response. While central processing is contingent upon upstream peripheral responses, the extent to which peripheral responses predict central processing remains unclear. In people with normal hearing [21], found that deficits in auditory temporal processing and speech-in-noise intelligibility were associated with decreased central inhibition (measured through the Stroop color and word test), rather than a change in ECAP AGF slope. This would indicate that speech-in-noise performance would be more accurately reflected by central auditory performance than peripheral. However, Scheperle and Abbas [22] argued that while cross-subject differences could not be explained entirely by differences in peripheral input, utilizing both peripheral and central responses improved their model of speech performance prediction [23]. Thus, while peripheral responses may not fully explain variation in CI performance, understanding their relationship with central responses may provide insight for CI programming decisions and counseling.

CAEPs are measured through surface EEG and consist of a positive peak (P1) with a latency of approximately 50 ms poststimulation, followed by a negative peak (N1) at approximately 100 ms, and another positive peak (P2) at approximately 180 ms [24]. These peaks represent neural processing at both pre-cortical and cortical levels [25, 26]. The presence of CAEPs in response to speech stimuli provides evidence that speech is audible to the listener [27]. The relationship between speech perception scores and CAEP responses has been well documented; present CAEP peaks and latencies closer to those of normal hearing individuals are correlated with improved speech perception [28‒30]. Hence, CAEPs may serve as a proxy measure of conscious sound perception.

Like CAPs, CAEPs can also be evoked via a CI (termed electrically evoked CAEPs [ECAEPs]). ECAEP responses have similar characteristics to those evoked via acoustic stimulation [31]. ECAEPs have a disadvantage in that they cannot be reliably elicited in patients under general anesthesia and therefore cannot give reliable insight into whether the CI can induce auditory perception at the earliest stage of CI use (i.e., intraoperatively). It is therefore useful to investigate if intraoperative ECAPs may themselves serve as a proxy measure for postoperative ECAEPs and therefore provide predictive insights into hearing outcomes.

The primary aim of this retrospective study was to investigate the relationships between intraoperative ECAP response parameters and postoperative ECAEPs, specifically the relationships between ECAP TLs and AGF slopes with ECAEP peak latencies and amplitudes. Given that prior studies which have indicated that the relationship between ECAP responses and hearing outcomes varies based on the cochlear region stimulated (apical, medial, or bases), we also analyzed the relationships between ECAP and ECAEP parameters separately by cochlear region. It was hypothesized that lower ECAP TLs and steeper AGF slopes would correlate with shorter ECAEP latencies and greater peak amplitudes.

This retrospective study was designed and conducted in accordance with the Declaration of Helsinki. Ethics approval was obtained from the South Metropolitan Health Service, Human Research Ethics Committee (reference number 3258). Written informed consent was obtained from all subjects.

Clinical data were obtained from CI users treated at Audiology Department of Fiona Stanley Hospital. Participants were included in the study if they had recordings of both intraoperative ECAPs and postoperative ECAEPs available. These measurements were completed as part of routine surgical and rehabilitative procedures at the clinic. All participants used devices from MED-EL.

Intraoperative ECAPs were generated and recorded via the “AutoART” or “ART” functions of the MAESTRO fitting software via the MAX programming interface (all MED-EL, Innsbruck, Austria). Stimulation and measurement parameters are given in Table 1. ECAP peak amplitudes, TLs, and AGF slopes were determined automatically via MAESTRO software. All responses were obtained from a recording electrode adjacent to the stimulating electrode.

ECAEPs were recorded via stimulation at three cochlear locations: basal (electrodes 1 or 2), medial (electrode 6), and apical (electrodes 10, 11, or 12). The participants were stimulated at their clinical MCL. Stimulation was performed via the programming interface. A trigger cable connected the programming interface to an auditory evoked potential (AEP) recording device (Bio-logic NavPRO; Natus, Middleton, WI, USA) which allowed the synchronization of sampling and stimulation. The stimulation and recording setup is illustrated in Figure 1.

Prior to stimulation, impedance tests were conducted to ensure recording electrode impedance of <5 kΩ and impedance differences between recording electrodes of <2 kΩ. ECAEP stimuli consisted of a 40 μs pulse burst with a burst duration of 70 ms, a burst rate of 1 kHz, and a cycle rate of 0.9 Hz. Stimulation continued until 100 non-rejected averages were taken. The active electrode was placed on the vertex (or forehead if good contact could not be made). The reference electrode was placed on the contralateral mastoid. The ground electrode was placed on the low forehead. All electrodes were adhesive tab electrodes.

Signals were processed using a low frequency filter at 0.3 Hz and a high frequency filter at 100 Hz, as well as artifact rejection and a notch filter if noise was present. The time window for the recording was 533 ms. During stimulation, a display showed the number of rejected and accepted recordings; if this gave an indication of high noise levels, clinicians stopped recording and troubleshooted to reduce noise, with that trace reattempted. Results were analyzed using the Bio-logic AEP software. Three clinicians marked responses for each participant individually, with final latencies and amplitudes resulting from the average position across the three markings. If at least two of three clinicians did not conclude that a peak was present, it was labeled as a nonresponse.

Statistical analyses were performed using R Statistics and RStudio version 4.3.0 (R 2013). A mixed-effects model was conducted using the “lme” function from the “lme4” package [32]. Each ECAEP measure (peak latencies and peak-peak amplitudes) was used as the outcome variable, while electrode location, time between ECAP and ECAEP measurements, ECAP TL, and ECAP AGF slope were fixed effects, and patient number was a random effect. Post hoc analysis via Spearman’s rank correlation was performed on the relationships found to be significant (p < 0.05) using the “cor.test” function from RStudio “stats” package. Differences in ECAP and ECAEP responses between electrode locations were assessed using Friedman rank sum test via the “friedman.test” function from RStudio “stats” package.

Participant demographics are shown in Table 2. ECAP and ECAEP responses from an apical (E1, E2, or E3), medial (E6), and basal (E10, E11, or E12) electrode were analyzed for 42 participants. However, data were not recorded from all three cochlear regions for some participants due to intra-cochlear recordings with missing responses and electrical artifacts. Both missing responses and electrical artifacts result in the detection algorithm of the ART program reporting an ECAP TL of zero and the trend line being negative. Consequently, 23 data points (7 apical, 7 medial, and 9 basal) were removed from analysis in the same manner as missing responses. The mean (±SD) interval between ECAP and ECAEP recording was 2.26 (±2.20) years.

An example ECAP trace obtained via MAESTRO’s ART and AutoART tasks (MED-EL, Innsbruck, Austria) is available in Figure 2. Ninety-three ECAP responses were obtained, consisting of 30 from apical electrodes, 32 from medial, and 31 basal. Friedman rank sum test found no significant difference in ECAP TLs at different electrode locations (χ² = 1.1783, df = 2, p = 0.5548). However, ECAP AGF slopes did vary depending on electrode location (χ² = 26.701, df = 2, p = 0.000002). The spread of AGF slopes by electrode location is evident in Figure 3.

An example ECAEP trace is evident in Figure 4. There were 69 P1 peaks identified by the clinicians (24 at apical electrodes, 24 at medial, and 21 at basal), 71 N1 peaks (24 at apical electrodes, 27 at medial, and 20 at basal), and 71 P2 peaks (23 at apical electrodes, 27 at medial, and 21 at basal). Consequently, P1-N1 peak-to-peak amplitude was calculated 67 times (23 at apical electrodes, 24 at medial, and 20 at basal), and N1-P2 was calculated 69 times (23 at apical electrodes, 26 at medial, and 20 at basal). Friedman rank sum test found no significant electrode location dependent difference in ECAEP P1 latency (χ² = 3.25, df = 2, p = 0.197), N1 latency (χ² = 2.48, df = 2, p = 0.289), P2 latency (χ² = 2.25, df = 2, p = 0.325), P1-N1 peak-to-peak amplitude (χ² = 0.609, df = 2, p = 0.738), nor N1-P2 peak-to-peak amplitude (χ² = 4.261, df = 2, p = 0.119).

Findings from mixed-effects model analysis are evident in Table 3. Statistically significant (p > 0.05) relationships were found between ECAEP P1-N1 peak-to-peak amplitude and time between ECAP and ECAEP measurements, ECAEP N1-P2 peak-to-peak amplitude and ECAP TLs, and ECAEP N1-P2 peak-to-peak amplitude and time between ECAP and ECAEP measurements. Statistically significant (p > 0.05) relationships were found between ECAEP P1-N1 peak-to-peak amplitude and time between ECAP and ECAEP measurements, ECAEP N1-P2 peak-to-peak amplitude and ECAP TLs, and ECAEP N1-P2 peak-to-peak amplitude and time between ECAP and ECAEP measurements.

Post hoc analyses via Spearman’s rank sum test were performed on the three significant relationships. These relationships are plotted in Figure 5. A weak positive relationship was found between intraoperative ECAP thresholds and ECAEP N1-P2 peak-to-peak amplitude (r = 0.301, p = 0.005). Time between ECAP and ECAEP measurements was weakly correlated with P1-N1 peak-to-peak amplitude (r = 0.321, p = 0.002) and ECAEP N1-P2 peak-to-peak amplitude (r = 0.340, p = 0.001).

There were 31 instances across the 126 electrodes analyzed where results were excluded from analysis because ECAPs responses were not produced. In 24 of these cases (77.4%), at least two components of the ECAEP P1-N1-P2 complex were recorded. Conversely, 74 out of 95 electrodes (77.9%) that produced an ECAP also produced at least two components of the ECAEP P1-N1-P2 complex.

In the present study, relatively weak or nonexistent correlations were observed between intraoperative ECAP response parameters and postoperative ECAEP response parameters. The only statistically significant intraoperative predictor of ECAEP responses was ECAP TLs, which showed a weak positive correlation (r = 0.301, p = 0.005) with ECAEP N1-P2 peak-to-peak amplitude. Furthermore, it should be noted that time between ECAP and ECAEP measurements had a stronger relationship with ECAEP N1-P2 peak-to-peak amplitude (r = 0.340, p = 0.001) and was also correlated with ECAEP P1-N1 peak-to-peak amplitude (r = 0.321, p = 0.002). Overall, these findings suggest that intraoperative ECAP responses offer negligible predictive insight into later cortical responses and sound perception in CI users.

The current study found no significant correlation between intraoperative ECAP AGF slope and postoperative ECAEP response parameters. This is in agreeance with findings from Wu et al. [15], who found that intraoperative ECAP TLs and AGF slopes were not correlated with any postoperative functional auditory outcomes. In contrast, with a cohort of CI users with fully inserted lateral wall electrode arrays, Basiony et al. [14] reported (1) moderate positive correlation between postoperative AGF slope of apical electrodes and monosyllabic word scores and (2) low-moderate negative correlations between postoperative AGF slopes of all electrode contacts and signal-to-noise ratio loss in speech understanding in noise results. The discrepancy in reported relationship could be due to the time of ECAP recording, with postoperative ECAP recordings allowing more time for peripheral function to recover and stabilize compared to intraoperative ECAPs.

The current study adds to the body of literature reporting weak or insignificant correlations between intraoperative ECAP TLs and postoperative central performance. Studies from Wu et al. [17] and Maged El Shennawy et al. [33] align with the current study’s reporting on weak or insignificant correlations between intraoperative ECAP TLs and postoperative auditory performance. However, research by Kim et al. [18] has reported significant inverse relations between intraoperative ECAP TLs and postoperative auditory performance across 22 electrode locations. While electrode design can influence ECAP TLs [17], it was not a distinguishing factor between these studies and thus cannot explain the differing findings.

In cases of peripheral auditory anomalies, such as cochlear nerve deficiency and malformed cochlea, SGNs require greater electrical current to evoke CAPs [34]. Thus, it stands that ECAP TLs should provide insight into peripheral auditory status. However, poor peripheral function may not necessarily be mirrored by poor central function and vice versa. It is possible that the impact of raised ECAP TLs on central auditory performance can be reduced through altering CI stimulation parameters, such as widening pulse widths and adjusting stimulation rate. Variation in auditory pathology location and CI map adjustments could contribute to heterogeneity in the reported relationship between intraoperative ECAP TLs and postoperative ECAEP responses.

Another observation of the present study was the relatively weak association between the ability to successfully elicit an intraoperative ECAP response and a postoperative ECAEP response using the same electrode. In 24 of the 31 instances (77.4%) where no ECAP was produced by stimulating a given electrode, the same electrodes were later able to produce at least two components of the ECAEP P1-N1-P2 complex. Thus, the presence or absence of an intraoperative ECAP response at a given electrode offers little insight into whether stimulation of that electrode will later elicit a cortical response. One possible explanation is the redundancy and resilience of the auditory nerve. In animal models, it has been shown that even when substantial neural loss is present (<5% SGN survival), electrically evoked auditory brainstem responses with normal morphology can still be elicited [35]. It is possible that when neural survival is low, a sub-detection threshold ECAP can still be generated which is capable of eliciting higher responses. Furthermore, in individuals with “normal” hearing thresholds, it has been suggested that deficits in auditory temporal processing and speech-in-noise intelligibility are mediated by decreased inhibition rather than cochlear synaptopathy [21]. Findings in normal hearing individuals and studies in peripheral nerve redundancy might explain the negligible relationship observed between ECAP and ECAEP responses in this study, suggesting that central auditory processing and CI performance might not be principally influenced by peripheral response quality.

Time between response measurements was weakly positively correlated with ECAEP P1-N1 peak-to-peak amplitude (r = 0.321, p = 0.002) and ECAEP N1-P2 peak-to-peak amplitude (r = 0.340, p = 0.001). While previous research has reported no significant correlations between ECAEP peak-to-peak amplitude and CI experience [36], the current study’s findings highlight the need to control for CI experience when looking at predictive models of CI hearing outcomes.

ECAEP response latencies and amplitudes did not differ by electrode location in the current study. However, generally speaking, more peaks were identified at apical and medial electrodes than basal electrodes. This trend was also reported in the literature [36]. While this did not hold true for ECAP detection, and ECAP TLs were similar across electrode locations, ECAP AGF slopes tended to decrease from apical to basal stimulating electrodes. Similar observations have previously been reported [37, 38]. This dependence on electrode position might be attributable to the heightened survival rate of neural tissue in the apical region, which is less subject to insertional trauma than more basal regions [39].

This study has some limitations. All measurements were taken in users of MED-EL CIs. This may limit the generalizability of these findings. Additionally, due to the retrospective nature of data collection, there were inconsistencies in the methods of data collection; some ECAPs were obtained through the AutoART function of MAESTRO, while others were obtained using the older ART function. Recent research has shown ECAP TLs and AGF slopes obtained from ART and AutoART to be positively correlated in adults, with no significant difference detected in AGF slope between measurement methods [16]. However, the use of two ECAP measurement methods may well have introduced additional variability into this set of measurements. Similarly, it is possible that absent responses may have been influenced by stimulation parameter restrictions.

The present study provides evidence for, at best, weak correlations between intraoperative peripheral and postoperative cortical responses in CI users. While intraoperative ECAP responses provide perhaps the earliest objective measure of CI function, results of the current study caution against their use as a predictor of CI hearing outcomes.

Study Protocol was reviewed and approved by the South Metropolitan Health Service Human Research Ethics Committee (EC00265), Approval No. 3258. Written informed consent was obtained from all subjects.

There are no conflicts of interest, financial or otherwise.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

A.S. performed experiments, designed the study, analyzed data, and wrote the manuscript; C.B. performed experiments; M.W.R.V. analyzed data and reviewed the manuscript; P.F. reviewed the manuscript; and D.T.V. performed experiments, designed the study, and commented on the manuscript at all stages.

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available from Alexander Stutley (A.S.) upon reasonable request.