Interplay Between Pulse Phase Duration and Inter-Phase Gap in the Assessment of Neural Health With Electrically Evoked Compound Action Potentials

Authors: Henk A. Vink, Huib Versnel, Dyan Ramekers

Abstract

Objectives:

Following deafness, a cochlear implant (CI) can be used for the restoration of hearing. CI effectiveness relies on the condition of the auditory nerve, which typically degenerates after deafness. The nerve’s condition can be assessed with the electrically evoked compound action potential (eCAP), the whole-nerve response to an electric pulse. Changes in the eCAP following an increase in the inter-phase gap (IPG) of a biphasic pulse have been reported to be informative of neural survival. This IPG effect can be explained by the temporal separation of the hyperpolarizing phase from the depolarizing phase. We hypothesize that increasing the phase duration (PD) has a similar effect.

Design:

We investigated the PD effect in normal hearing and in ototoxically deafened guinea pigs (total N = 40) with various conditions of the auditory nerve by recording eCAPs to biphasic current pulses with alternating polarity and with varying PD and IPG. The eCAP data were obtained from both chronically and acutely implanted guinea pigs by using CI stimulation paradigms with a fixed charge and varying PD (30, 50, or 100 µs) and IPG (2.1 or 30 µs). We evaluated six eCAP five derived from the amplitude growth function and the N1 latency. We examined the relationships of PD and IPG effects with the survival of the spiral ganglion cells.

Results:

The PD effects were stronger for latency than IPG effects, but weaker for the other five evaluated eCAP measures. The PD effect did not correlate as well with neural survival as the IPG effect. The IPG effect decreased with increasing PD, and accordingly, the stronger correlations between the IPG effect and neural survival were found for a short PD. Notably, the latency increase with increasing PD was greater than 1, which indicates the second phase of the pulse significantly contributes to the eCAP. This second-phase contribution was larger for lower neural survival.

Conclusions:

The PD effect of latency has predictive power in assessing neural survival. When using other eCAP measures than latency, the best approach to assess neural survival is using the IPG effect with a short PD (around 30 µs).

INTRODUCTION

As of yet, a cochlear implant (CI) is the only therapy for hearing restoration in cases of severe to profound sensorineural hearing loss, which is mainly caused by loss of cochlear hair cells and/or loss of spiral ganglion cells (SGCs) that make up the auditory nerve. A CI bypasses the cochlear hair cells and directly stimulates the SGCs. The CI function is therefore dependent on the condition of the auditory nerve. Following deafness, SGCs degenerate (Ylikoski et al. 1974; Spoendlin 1975; Webster & Webster 1981; Leake & Hradek 1988; Van Loon et al. 2013; Kroon et al. 2017), which negatively affects hearing performance in human CI users according to recent literature (Seyyedi et al. 2014; Kamakura & Nadol 2016; Ishiyama et al. 2019). SGC degeneration can only indirectly be assessed in CI users, for instance, with imaging, psychophysics, or electrophysiology (Pfingst et al. 2015; Vos et al. 2015; Wasmann et al. 2018; Skidmore et al. 2022; Schvartz-Leyzac et al. 2023).

To find objective measures to quantify auditory nerve health, numerous studies have been performed to compare electrophysiological measures as obtained with electrically evoked compound action potentials (eCAPs) with postmortem histological analysis of SGC degeneration in guinea pigs (Prado-Guitierrez et al. 2006; Ramekers et al. 2014; Adenis et al. 2018; Schvartz-Leyzac et al. 2019, 2020a). eCAPs, whole-nerve responses to electric pulses, have been of great interest as they allow for a near-site observation of the auditory nerve. Therefore, eCAPs have been studied both in animals and in human CI users. Absolute eCAP measures have previously been reported to vary with the extent of neural survival (Miller et al. 1998; Ramekers et al. 2014). For instance, the eCAP amplitude decreases with SGC loss. However, because these absolute measures can be highly dependent on non-neural factors (Schvartz-Leyzac et al. 2020b), relative within-subject differential measures have been developed as an alternative. Increasing either the time between the two phases, the inter-phase gap (IPG), or the phase duration (PD), with a fixed charge, led to a stronger neural response, as both of these modifications of the pulse shape effectively increase the temporal separation of depolarizing, action-potential-initiating, phase from the polarizing phase, and thereby reduce the chances of a negated action potential (Van den Honert & Mortimer 1979; Shepherd & Javel 1999). The effect of increasing PD will be countered by membrane leakiness, which in particular plays a role at longer PDs (>50 µs; Chatterjee & Kulkarni 2014; Ramekers et al. 2014; Zhou et al. 2021). The changes in eCAP measures as a result of an increase in either IPG (the IPG effect) have been reported to correlate with SGC survival in guinea pigs (Prado-Guitierrez et al. 2006; Ramekers et al. 2014, 2015, 2022; Schvartz-Leyzac et al. 2019, 2020a; Vink et al. 2020, 2022). A similar correlation was observed for an increase in PD (the PD effect), and SGC survival was observed for electrically evoked auditory brainstem responses for long PDs (104 to 208 µs; Prado-Guitierrez et al. 2006), which could not be detected with short PDs recording eCAPs (20 to 50 µs; Ramekers et al. 2014). Larger IPG than PD effects were observed in clinical studies with human CI users (He et al. 2020a, 2020b). In fact, the PD effect was not significant, which can be attributed to the narrow range of PDs applied (50 to 88 µs; He et al. 2020b).

Partly due to these findings, the focus in the existing literature has mainly been on the IPG. But because an increase in PD is in essence similar to an increase in IPG (i.e., temporally separating depolarization from (re)polarization), we hypothesize that the PD effect is just as valuable in predicting neural health as the IPG effect. In the present study, we therefore examined the magnitude of the PD effect in animals with various conditions of the auditory nerve, using a larger range of PDs (30 to 100 µs) than in earlier-mentioned eCAP studies (Ramekers et al. 2014; He et al. 2020b). We questioned whether a predictive relationship between the PD effect and neural survival exists. Furthermore, we examined how PD influences the well-established IPG effect and its relation to neural survival.

MATERIALS AND METHODS

Animals and Experimental Design

The data presented in this paper were obtained from 21 acutely and 19 chronically implanted guinea pigs (total N = 40). The acutely implanted animals (normal hearing, NH, N = 9; and six-week deaf, 6WD, N = 12) have previously been described by Vink et al. (2020); part of the seven-week deaf (7WD, N = 8) chronically implanted animals have previously been described by Ramekers et al. (2022). Note that the NH animals did not receive any ototoxic treatment before eCAP recordings. In the present study, we have analyzed unpublished electrophysiological data from all the aforementioned animals (N = 29). The remaining 11 chronically implanted guinea pigs have not yet been reported.

A total of forty female albino guinea pigs (Dunkin-Hartley; Hsd Poc: DH; 250 to 350 g) were obtained from Envigo (Horst, the Netherlands) and kept under standard laboratory conditions (food and water ad libitum; lights on between 00 A.M. and 00 P.M.; temperature 21°C; humidity 60%). All animals had normal hearing before any experimental procedure, as assessed with click-evoked auditory brainstem responses (ABRs). Figure 1 shows a schematic overview of the experimental groups from the different studies, which consisted of one NH control group and three groups that were deafened via systemic co-administration of kanamycin and furosemide. The animals in all groups were sacrificed shortly after the eCAP recordings presented here. For the deafened groups, this was at either 3, 6, or 7 weeks following the deafening procedure. The 3WD and 7WD animals (3WD: N = 5; 7WD: N = 14) received an intracochlear electrode array 4 weeks before deafening. Half of these chronically implanted animals received chronic electrical stimulation. Because this did not significantly affect the electrophysiological outcomes, we did not divide these animals into subgroups. The 6WD group and NH group were acutely implanted with an electrode array just before termination to allow for eCAP recordings. The 6WD and 7WD groups were treated separately in our analyses because, apart from the 1-week difference in duration of deafness, the animals were from two studies, with the largest difference being the acute insertion of the electrode array in the 6WD group (Vink et al. 2020) and chronic implantation of the array in the 7WD group (Ramekers et al. 2022). Tissue growth in chronically implanted cochleas will be larger (Swiderski et al. 2020); furthermore, peripheral processes have been found to be larger in acutely implanted cochleas (Ramekers et al. 2020). The implanted cochleas were processed for histology immediately following termination. All surgical and experimental procedures were approved by the Dutch Central Authority for Scientific Procedures on Animals (CCD: 11500201550 and 1150020174315).

Surgical Procedures

Chronic Implantation

The right ears of the animals in the two chronic groups (3WD, 7WD) were implanted with an intracochlear electrode array (see Ramekers et al. 2022 for details). In short, animals were anesthetized by intramuscular injection of dexmedetomidine (0.25 mg/kg) and ketamine (40 mg/kg). Click-evoked ABRs were recorded to confirm normal hearing (see 2.3.1), with thresholds <40 dB peak equivalent SPL considered to indicate normal hearing. The skull was exposed, and six transcranial screws were positioned to anchor the head connector and to be applied as active or reference electrodes for ABR or eCAP recordings. A 0.4-mm cochleostomy was hand-drilled into the scala tympani of the basal turn, through which a custom-made three-contact electrode array (MED-EL GmbH, Innsbruck, Austria) was inserted. The head connector of the electrode array was positioned approximately on the bregma and fixed with cold-curing dental cement (ProBase Cold; Ivoclar Vivadent AG, Schaan, Liechtenstein).

Deafening Procedure

Anesthesia was induced with dexmedetomidine and ketamine as described earlier (2.2.1), followed by preoperative analgesia by subcutaneous injection of carprofen (4 mg/kg), and click-evoked ABRs were recorded to confirm normal hearing (see 2.3.1). Deafening was done by subcutaneous injection of kanamycin (Sigma-Aldrich, St. Louis, MO, USA; 400 mg/kg) followed by infusion of furosemide (Centrafarm, Etten-Leur, the Netherlands; 100 mg/kg) into the external jugular vein, leading to substantial hair cell loss (West et al. 1973; Versnel et al. 2007). In addition, as the 6WD animals were originally used as a negative control group in a previous study (see Vink et al. 2020), these animals underwent an additional surgical procedure 2 weeks after deafening, in which they received a saline-soaked gelatin sponge on the round window membrane of their right cochlea.

Preparation for Terminal Recordings

For all animals, before the eCAP recordings, anesthesia was induced by injection of Hypnorm (Vetapharma; 0.5 mL/kg i.m.) and subsequent administration of a gas mixture of 2% isoflurane evaporated in O2 and N2O (1:2) via a mouth cap (for details see Vink et al. 2020). For animals receiving acute cochlear implantation (NH and 6WD groups), the skull was then exposed, and two transcranial screws were placed 1 cm bilateral to bregma to serve as reference electrodes for eCAP recordings. All animals were tracheostomized and artificially ventilated (Amsterdam infant ventilator mk3, Hoekloos, Schiedam, the Netherlands) with a gas mixture of O2 and N2O (1:2) and 1 to 1.5% isoflurane (45 to 50 cycles/min respiration rate, 1.8 to 2.3 kPa) throughout the remainder of the experiment. Acute cochlear implantation was performed by exposing the right cochlear bulla and removing the bone to expose the cochlea. A 0.5-mm cochleostomy was then hand-drilled in the basal turn through which a custom-made four-contact electrode array (MED-EL GmbH, Innsbruck, Austria) was inserted into the scala tympani. Last, to enable the eCAP recordings, the implanted electrode array of each animal was connected to a MED-EL PULSAR cochlear implant.

Electrophysiology

ABR Recordings

Click-evoked ABRs were recorded both before ototoxic deafening and cochlear implantation using either transcranial screws (see 2.2.1) or three subcutaneous needle (1) the active electrode was placed behind the right ear, (2) the ground electrode in the hind limb, and (3) the reference electrode was placed on the skull, rostral to the brain. Broadband acoustic clicks (20 µs monophasic rectangular pulses; inter-stimulus interval 99 ms) were synthesized and attenuated using a TDT3 system (Multi-I/O processor RZ6; Tucker-Davis Technologies, Alachua, FL, USA), and presented in free field, at 10-cm distance from the right ear, using a Blaupunkt speaker (PCxb352; 4 Ω; 30 W, Blaupunkt [International] GmbH & Co. KG, Hildesheim, Germany). For amplification of the ABRs, a Princeton Applied Research (Oak Ridge, TN, USA) 5113 pre-amplifier (×5000; band-pass filter 0.1 to 10 kHz) was used. The response was subsequently digitized by the TDT3 system (100 kHz sampling rate, 24-bit sigma-delta converter) and stored on a PC for offline analysis. Hearing thresholds were determined by starting at 110 dB peak equivalent SPL and subsequently decreasing the sound level by 10 dB until no further response was observed. The threshold was then defined as the interpolated sound level at which the ABR N1-P2 peak was 0.3 µV.

eCAP Recordings

The eCAP recordings (Figs. 2A–D) were performed as described previously (Ramekers et al. 2014; Vink et al. 2020). The implant was controlled by a PC via a Research Interface Box 2 (Department of Ion Physics and Applied Physics, University of Innsbruck, Innsbruck, Austria) and a National Instruments data acquisition card (PCI-6533). Custom-written stimulation/recording paradigms were executed using MATLAB (version 7.11.0; MathWorks, Natick, MA, USA). Typically, the most apical of the electrodes on the intracochlear array was used for stimulation (~4 mm inside the cochleostomy) and the most basal one was used for recording. Fixed-charge biphasic current pulses of 30, 50, or 100 µs/phase (Fig. 2E) were presented with alternating polarity to reduce stimulation artifact, and the responses to 50 pairs of these stimuli were averaged. All stimuli were presented with an IPG of 2.1 or 30 µs with 20 stimulation levels typically ranging from 1.2 to 24 nC. If the intended current to reach saturation of the amplitude growth function (AGF) exceeded the compliance limit, as measured by electrode impedance at the beginning and end of the recording session, the maximum current was lowered accordingly.

Fig. 2.: Mean eCAP characteristics for the four groups for all PDs and an IPG of 2.1 (first and third column) and 30 µs (second and fourth column). A, B, Maximum amplitude; C, D, slope; E, F, threshold; G, H, dynamic range; I, J, level50%; K, L, N1 latency. Error bars represent SE of mean (SEM). 3WD indicates three weeks deaf; 6WD, six weeks deaf; 7WD, seven weeks deaf; NH, normal hearing.

Chronic Electrical Stimulation

The chronically implanted animals underwent eCAP recordings periodically (~weekly; outcomes not reported here; see Ramekers et al. 2022) from the time of implantation onward, and 3 of the 5 3WD and 6 of the 14 7WD animals additionally received chronic electrical stimulation. The animals were stimulated by a battery-powered stimulator, either implanted subcutaneously or fixed onto a backpack that was worn continuously for 2 weeks. The stimulator switched on automatically at 7 A.M. and stimulated the three intra-cochlear electrodes for 12 h with an 80% duty cycle (4 min on; 1 min off), for 6 days per week—eCAPs were recorded on the 7th day. Stimuli consisted of charge-balanced pseudo-monophasic pulses that were delivered with variable pulse rate between 900 and 1700 Hz, which was changed every 100 ms. The current level was set just earlier than the behavioral detection threshold for each animal individually.

Histological Processing

Both the implanted and the contralateral non-implanted cochleae were extracted after termination. However, only data from the right implanted cochleas were used, as in the present study, we are interested in the correlation between eCAP outcomes and histological outcomes. The cochleas were fixed by an intra-labyrinthine infusion with a fixative of 3% glutaraldehyde, 2% formaldehyde, 1% acrolein, and 2.5% dimethyl sulfoxide in a 0.08 M sodium cacodylate buffer, as described by De Groot et al. (1987). The cochleae were decalcified, post-fixed, and embedded in Spurr’s low-viscosity resin. Staining was performed using 1% methylene blue, 1% azur B, and 1% borax in distilled water. The cochleae were subsequently divided into two halves along a standardized midmodiolar plane and then re-embedded in fresh resin. From one of these halves, two semithin (1 µm) sections were cut at least 60-µm intervals, which were used for the quantification of SGCs.

Data Analysis

eCAP Analysis

Analysis of the eCAP recordings was performed in the same way as described before (Ramekers et al. 2014, 2015; Vink et al. 2020). The eCAP amplitude—defined as the voltage difference between the N1 and P2 peaks—was determined by manual peak picking in Matlab customized software and plotted against charge, resulting in AGFs (Figs. 2F–I), which were subsequently fitted with a Boltzmann

in which V~~eCAP~~ is the amplitude in µV, I is the stimulation current in µA, and A-D are fitting parameters. From the fitting parameters, five measures were the maximum N1-P2 amplitude (B, referred to as “amplitude”), the charge to reach the half-maximum amplitude (C, “level~~50%~~”), slope at C (B/4D, “slope”), threshold (C − 2D), and the dynamic range (4D). Note that amplitude, level~~50%~~ and dynamic range are directly related to B, C, and D, respectively, and that slope and threshold are related to those former three as slope = amplitude/dynamic range; threshold = level~~50%~~ − 0.5 × dynamic range. The N1 peak latency (“latency”), averaged over the three highest current levels, was analyzed as the sixth measure in addition to these five input/output characteristics. We chose the high levels because the outcomes are robust at those levels; we averaged over three levels to further reduce variability. This approach is consistent with previous reports (Ramekers et al. 2014, 2015, 2022; Vink et al. 2020, 2022, 2023).

The IPG in the biphasic current pulses was either 2.1 or 30 µs. The IPG effect (∆IPG), being the difference in a given eCAP measure between these two IPG durations, was calculated for amplitude as ∆~~IPG~~amplitude = (amplitude~~IPG30~~ − amplitude~~IPG2.1~~)/amplitude~~IPG30~~. We applied normalization considering the absolute difference is proportional to the amplitude itself, thus partially dependent on non-neural factors, and considering consistency with our earlier papers (Ramekers et al. 2014, 2015, 2022; Vink et al. 2020, 2022, 2023). For the remaining five measures mentioned earlier, it was calculated as ∆~~IPG~~ = (~~IPG30~~ − ~~IPG2.1~~).

Similarly, for the differences in response to varying PD, the amplitude was calculated as ∆PDamplitude = (amplitude~~PDlong~~ − amplitude~~PDshort~~)/amplitude~~PDlong~~. For the remaining five measures, ∆PD was calculated as (~~PDlong~~ − ~~PDshort~~). These will be annotated as ∆~~PD50−30~~, ∆~~PD100−30~~, and ∆~~PD100−50~~ for the different combinations of PDs.

Histological Analyses

A Leica DC300F digital camera mounted on a Leica DMRA light microscope was used with a 40× oil immersion objective (Leica Microsystems GmbH, Wetzlar, Germany) to obtain micrographs of Rosenthal’s canal of the cochlear regions B1, B2, M1, M2, A1, A2, and A3 from two sections per cochlea in which the packing density and mean perikaryal area of both type I and type II SGCs was established (for details, see Vink et al. 2020). For the purpose of assessing correlations with eCAP measures, the seven resulting packing densities were first averaged into three global cochlear locations (B, M, and A), before being averaged into a single value.

In addition to SGC quantification, a sample-wise approximation of remaining cochlear hair cells was made. This was done by quantifying both inner hair cells (IHCs) and outer hair cells (OHCs) in the organ of Corti of all seven cochlear regions in a single cross-section per animal and subsequently averaging these seven in a single value. These hair cells were quantified using the same criteria used by Tisi et al. (2022): the presence of at least one of the following (1) a nucleus, (2) a cilia bundle, and (3) a clear cochlear-hair-cell-like outline.

Statistics

To examine the dependence of the eCAPs on IPG, PD, and group, two-way repeated measures analysis of variance (RM ANOVA) was performed for each of the six eCAP measures (acting as dependent variable), with IPG as a two-level factor, PD as a three-level factor, and experimental group as the between-subject factor. Thus, we conducted six RM ANOVAs, and therefore, a Bonferroni correction for testing on six eCAP measures has been applied with a significance level of 0.0085. Pearson’s correlation coefficient was used to examine the relationship between the eCAP measures and the SGC packing density. The RM ANOVAs were performed in SPSS Statistics 26.0.0.1 for Windows (IBM Corp., Armonk, NY, USA), while Pearson’s correlation coefficients were calculated in MATLAB (version 9.1.0; MathWorks, Natick, MA, USA).

RESULTS

Deafening

Before the deafening procedure, all animals were confirmed to have normal hearing. The deafened animals were successfully deafened, with a mean ABR threshold shift of 74 dB (range of 58 to 87 dB). The ABR results were supplemented by cochlear HC sampling. Following the deafening procedure, severe cochlear HC loss was observed for all groups (reported as % survival re NH; mean ± SE of mean [SEM]): 3WD (IHC: 16 ± 5, OHC: 3 ± 1), 6WD (IHC: 28 ± 7, OHC: 27 ± 6), and 7WD (IHC: 29 ± 6, OHC: 18 ± 4). SGC packing densities (as % survival re NH; mean ± SEM) were 70 ± 9 for 3WD, 39 ± 9 for 6WD and 46 ± 3 for 7WD.

Absolute eCAP Measures and the Impact of PD on the IPG Effect

Figure 3 shows the six eCAP characteristics as a function of PD for an IPG of 2.1 and 30 µs for both the NH and 6WD groups. As the eCAP N1 latency is the only eCAP measure presented in this study that is not sigmoid-fit-derived (Eq. 1) and because PD prominently affects latency, the results for this measure are described in a separate section (eCAP Latency). For the sake of visual clarity, the data from the other deafened groups are not shown in this figure, but are included in the statistical analyses. In Figure 4, the IPG effect (∆IPG) is shown for the six eCAP characteristics of all groups as a function of PD. Table 1 presents the F and p values of the RM ANOVA analyses examining the dependence of eCAP characteristics on IPG and PD.

Fig. 4.: Group means of the IPG effect as a function of PD for all six eCAP characteristics. A, ∆IPGamplitude relative to the maximum amplitude; B, ∆IPGslope; C, ∆IPGthreshold; D, ∆IPGdynamic range; E, ∆IPGlevel50%; F, ∆IPGlatency. Error bars represent SEM. 3WD indicates three weeks deaf; 6WD, six weeks deaf; 7WD, seven weeks deaf; NH, normal hearing.

The amplitude (Figs. 3A, B) for the deafened groups was significantly lower than that of the NH animals across all three PDs and both IPGs [F~~(3,36)~~ = 29.9, p < 0.001]. Increasing the IPG led to an increase in amplitude for all groups and PDs [F~~(1,36)~~ = 179.9, p < 0.001]. This increase in amplitude with increasing IPG − ∆~~IPG~~amplitude in short (note: normalized by amplitude to longer IPG) did not significantly differ among the four groups [Fig. 4A; F~~(3,36)~~ = 1.8, p = 0.15]. In addition, increasing the PD also significantly affected the amplitude [Figs. 3A, B; F~~(2,72)~~ = 11.1, p < 0.001]. The magnitude of this effect was dependent on group [F~~(6,72)~~ =5.0, p < 0.001]. Furthermore, the ∆~~IPG~~amplitude decreased with a longer PD [Fig. 4A, F~~(2,72)~~ = 45.2, p < 0.001], which was independent of group [F~~(6,72~~)~~ = 1.5, p = 0.10].

The slope (Figs. 3C, D), like amplitude, considerably decreased following deafness [F~~(3,36)~~ = 39.4, p < 0.001], and generally increased with increasing IPG [Figs. 3C, D; F~~(1,36)~~ = 109.1, p < 0.001]; the magnitude of the corresponding ∆~~IPG~~slope was larger for normal hearing than for deafened animals [Fig. 4B; F~~(3,36)~~ = 10.9, p < 0.001]. The slope significantly varied with increasing PD [Figs. 3C, D; F~~(1.4,49.0)~~ = 19.7, p < 0.001], which was reflected by a decrease for the deafened animals. We observed a strong interaction between PD and IPG [F~~(1.5,52.2)~~ = 67.6, p < 0.001]; specifically, the dependence on IPG decreased with a longer PD (Fig. 4B), with the IPG effect being virtually 0 for a PD of 100 µs. This interaction effect between PD and IPG depended on group [Fig. 4B, F~~(4.3,52.2)~~ = 12.8, p < 0.001]. Most notably, the difference in ∆~~IPG~~slope between the groups decreased with increasing PD, dissipating at a PD of 100 µs (Fig. 4B).

The eCAP threshold (Figs. 3E, F) varied slightly but significantly with group [F~~(3,36)~~ = 4.9, p = 0.006]. It furthermore decreased significantly with increasing IPG [Figs. 3E, F; F~~(1,36)~~ = 367.8, p < 0.001], which was independent of group [Fig. 4C; F~~(3,36)~~ = 3.4, p = 0.16], and varied significantly with PD [F~~(1.4,50.8)~~ = 35.3, p < 0.001], which depended on group [F~~(4.2,50.8)~~ = 3.7, p = 0.010]. Interestingly, the threshold decreased with increasing PD for short IPG and it increased with PD for the long IPG (Figs. 3E, F), meaning that the ∆~~IPG~~threshold decreased with an increase of PD (Fig. 4C), confirmed by a strong interaction effect [F~~(1.7,61.9)~~ = 132.1, p < 0.001]. This interaction effect was not dependent on group [F~~(5.2,61.9)~~ = 1.7, p = 0.14].

The dynamic range (Figs. 3G, H) was generally wider for the deafened animals than for the NH group [F~~(3,36)~~ = 4.6, p = 0.008]. The range widened for all groups following an increase in IPG [Figs. 3G, H; F~~(1,36)~~ = 139.6, p < 0.001], but the extent of this widening significantly varied among groups [Fig. 4D; F~~(3,36)~~ = 7.9, p < 0.001]. Dynamic range significantly varied with PD [F~~(1.5,53.3)~~ = 6.6, p = 0.006], the extent of which was group dependent [F~~(4.4,53.3)~~ = 5.4, p < 0.001]. Furthermore, increasing the PD decreased the IPG effect of the dynamic range, indicated by a statistically significant interaction effect [Fig. 4D; F~~(2,72)~~ = 41.7, p < 0.001], which was dependent on group [F~~(6,72)~~ = 3.9, p = 0.002]. As with ∆~~IPG~~slope, the group difference in ∆~~IPG~~dynamic range diminished strongly with increasing PD.

Level~~50%~~ (Figs. 3I, J) differed among groups [F~~(3,36)~~ = 9.9, p < 0.001], with most deafened animals (3WD, 7WD) having a level~~50%~~ equal to or lower than the NH animals. Interestingly, the 6WD animals showed the highest level~~50%~~ of all groups. Increasing the IPG led to a decrease in level~~50%~~ [Figs. 3I, J; F~~(1,36)~~ = 623.6, p < 0.001], with a significant group effect [Fig. 4E; F~~(3,36)~~ = 9.0, p < 0.001]. Conversely, increasing the PD generally increased the level~~50%~~ [Figs. 3I, J; F~~(1.6,59.1~~)~~ = 42.7, p < 0.001], which was independent of group [F~~(~~4.9,59.1)~~ = 1.0, p = 0.45]. A strong interaction effect was observed between IPG and PD [Fig. 4E; F~~(2,72)~~ = 129.0, p < 0.001], showing that, similar to ∆~~IPG~~threshold, the ∆~~IPG~~level~~50%~~ becomes smaller with a longer PD irrespective of group [F~~(6,72)~~ = 0.5, p = 0.81].

To summarize, for all fit-derived eCAP characteristics, a significant main effect was observed for IPG and PD. In addition, an interaction between IPG and PD was also observed for each eCAP characteristic; most notably, for all five sigmoid-fit-derived measures, the IPG effect diminished with increasing PD. For slope and dynamic range, the PD-IPG interaction was found to be dependent on group.

PD Effect

Following the large and statistically significant interaction effect between IPG and PD for each of the five fit-derived eCAP characteristics, we assessed how elongation of PD changed the eCAP measures for both IPGs. We found a highly significant main effect of PD for all these eCAP measures (Table 1).

Following the RM ANOVA, simple contrasts were explored for ∆PD~~50−30~~ and ∆PD~~100−30~~ (Table 2), being the smallest and largest ∆PD in this study. A significant PD effect was largely absent for the fit-derived eCAP measures when prolonging the PD from 30 to 50 µs, as a statistically significant main effect of ∆PD~~50−30~~ was observed only for threshold [F~~(1,36)~~ = 8.0, p = 0.008]. However, as seen in Figure 4, PD extension from 30 to 50 µs resulted in a strong change in the IPG effect for all fit-derived eCAP measures. Indeed, for ∆PD~~50−30,~~ a strong interaction effect was observed with IPG for all five measures (∆PD~~50−30~~: F > 29, p < 0.001). Extending the PD to 100 µs demonstrated a clear PD effect, with a strong main effect of ∆PD~~100−30~~ across all five measures (F > 9, p < 0.005). These main effects were observed to be dependent on IPG (F > 59, p < 0.001), as can be seen in Figure 5—which shows the ∆PD~~100−30~~ of the eCAP characteristics as a function of IPG for each of the groups—where noticeable differences can be ∆~~PD100−30~~amplitude, ∆~~PD100−30~~slope, and ∆~~PD100−30~~dynamic range (Figs. 5A, B, D, respectively) decreased with an increase in IPG, while ∆~~PD100−30~~threshold, and ∆~~PD100−30~~level~~50%~~ (Figs. 5C, E, respectively) increased with an increase in IPG. An interaction between ∆PD~~100−30~~ and group was observed for amplitude, slope, dynamic range, and latency (F > 5; p < 0.005), indicating their potential relationship with neural health.

Fig. 5.: Group means of the ∆PD100−30 effect as a function of PD for all six eCAP characteristics. A, ∆PDamplitude relative to the longest PD; B, ∆PDslope; C, ∆PDthreshold; D, ∆PDdynamic range; E, ∆PDlevel50%; F, ∆PDlatency. Error bars represent SEM. 3WD indicates, three weeks deaf; 6WD, six weeks deaf; 7WD, seven weeks deaf; NH, normal hearing.

eCAP Latency

Action potentials are initiated when sufficient charge has accumulated to depolarize the neural membrane. The moment this threshold is reached, the so-called point of critical charge accumulation (PCCA) is temporally dependent on the PD in a fixed-charge stimulation. This, in turn, means that the eCAP N1 latency is strongly related to PD; with a longer PD, the PCCA occurs later than with a shorter PD. Therefore, one expects the eCAP N1 latency to increase with increasing PD.

Figures 3K, L show the absolute mean latency as a function of PD. Differences between groups were observed [F~~(3,36)~~ = 15.3, p < 0.001; Table 1], most notably the normal hearing and 3WD animals exhibit a longer latency than the 6WD and 7WD animals by 24 to 126 µs (shown later in Figs. 6A, B). Expectedly, the latency increased strongly with an increase in PD [F~~(1.6,58.1)~~ = 1011.3, p < 0.001], which was group dependent [F~~(4.8,58.1)~~ = 5.7, p < 0.001]. Increasing the IPG significantly affected the latency [F~~(1,36)~~ = 105.6, p < 0.001], and the magnitude of this effect significantly depended on group [Fig. 4F; F~~(3,36)~~ = 11.6, p < 0.001]. Furthermore, increasing the PD reduced the IPG effect [Fig. 4F; F~~(1.6,58.5)~~ = 13.6, p < 0.001], interacting with group [F~~(4.9,58.5)~~ = 5.7, p < 0.001] and bringing their means closer together, thereby reducing the differences between them.

Fig. 6.: LIFs for both IPGs. A, B, The mean LIF from PD 50 to 100 µs for every group (duration of deafness). C, D, The LIF from PD 50 to 100 µs for all individual animals as a function of cochlear SGC survival. IPG, inter-phase gap; LIFs indicate latency increase factors; PD, phase duration; SGC, spiral ganglion cell.

Figure 5F shows the ∆~~PD100−30~~ latency per group as a function of IPG, with a notable latency difference of about 80 µs for the NH group, and around 100 µs for the deafened animals (6WD, 7WD). Indeed, and as reasoned at the beginning of this section, increasing the PD from 30 to either 50 or 100 µs increased the latency in a strongly significant manner (F > 170; p < 0.001; see Table 2), which was dependent on group for ∆PD~~100−30~~ [F~~(3,36)~~ = 7.5, p < 0.001]. When increasing the IPG from 2.1 to 30 µs, the ∆PDlatency was significantly affected (F > 13; p < 0.001): the ∆PDlatency increased for the NH and both 3WD groups, while it did not change for the 6WD and 7WD groups, showing a strong dependency on group (F > 5, p < 0.003) and reducing the differences between groups.

As described in section 2.4, the stimulation consisted of a biphasic pulse, which was presented with alternating polarity for artifact reduction purposes. For most groups, the eCAP latency increase with PD elongation is longer than this elongation (100 − 30 µs = 70 µs), but shorter than the increase of the two that occur in one biphasic pulse (2 × 70 µs = 140 µs). Suppose that the cathodic phase elicits the eCAP response, and suppose that the excitation occurs at the end of the phase—when all the charge has accumulated—then the increase in latency, following a cathodic-first pulse, would be 1 × ∆PD. Following an anodic-first pulse, this increase would be 2 × ∆PD. As the polarity was alternated, the latency increase would therefore be, theoretically, approximately 1.5 × ∆PD, because the excitatory cathodic phase is second in the anodic-leading pulse. On the basis of this theoretical latency increase, we define a latency increase factor as LIF = ∆latency/∆PD.

In Figures 3K, L the dashed line represents the theoretical latency increase, parameterized by the LIF of 1.5. We calculated the actual LIF for all groups between a PD of 50 and 100 µs from Figures 3K, L. The groups generally showed a LIF close to 1.5. The ∆PD~~100−50~~ LIF (Figs. 6A, B) for NH and 3WD were approximately 1.3, and for 6WD and 7WD, this was around 1.6. Repeated measures ANOVA (with IPG as the repeat) revealed that the differences between groups were close to significant [F~~(3,36)~~ = 4.3, p = 0.011].

The Effect of PD on the Relation Between IPG Effect and Neural Survival

We examined whether the decrease of the IPG effect with increasing PD, as described earlier, would change its known predictive power on neural health. Figure 7 shows the IPG effect for each eCAP measure plotted against SGC survival across the entire cochlea at an individual level. ∆~~IPG~~slope (Fig. 7, top row) was selected as an example, as there is a clear decrease in correlation between the IPG effect and SGC survival with increasing PD (from R^2^ = 0.46 to R^2^ = 0.04), with the strongest correlation at a PD of 30 µs. For a comprehensible overview of the correlations, the analyses, as exemplified with ∆~~IPG~~slope, were condensed visually to only the regression lines for each PD (Fig. 7, middle and bottom row) with corresponding R^2^ and p values of these correlations shown in Table 3.

Similar to ∆~~IPG~~slope, the strongest correlation with neural survival of ∆~~IPG~~threshold (significant only for PD = 30 µs; R^2^ = 0.13, p = 0.020), ∆~~IPG~~dynamic range (significant for all PDs; R^2^ > 0.16, p < 0.01), and ∆~~IPG~~latency (significant for all PDs; R^2^ > 0.13, p < 0.02) was observed with a PD of 30 µs, with a decrease in correlation with prolonged PD. For ∆~~IPG~~amplitude and ∆~~IPG~~level~~50%~~, the correlation was strongest for a PD of 50 µs (amplitude: R^2^ = 0.26, p < 0.001; level~~50%~~: R^2^ = 0.23, p = 0.002).

To summarize, a short PD yields the strongest correlations between the IPG effect and neural survival. This correlation with neural survival becomes weaker with a longer PD.

PD Effect and Neural Survival

We examined the relationship between the PD effect (∆PD) for each of the six eCAP measures and neural survival. The ∆PD is shown as a function of SGC packing density with an IPG of 2.1 (Fig. 8) and 30 µs (Fig. 9). Analogous to the IPG effect in Figure 7, ∆PDslope is depicted as an example of individual data points and the corresponding regression lines (Figs. 8 and 9, top row). Table 4 contains all R^2^ and p values of the correlations shown in Figures 8 and 9. Note that the ∆~~PD100−50~~latency regression lines in Figures 6E, F are essentially the same as those shown in Figures 8 and 9, with the former expressed in ∆PD and the latter in µs.

Of the six eCAP measures with an IPG of 2.1 µs (Fig. 8, middle and bottom rows) ∆slope had a positive correlation with SGC packing density for ∆PD~~50−30~~ (R^2^ = 0.40, p < 0.001) and ∆PD~~100−30~~ (R^2^ = 0.23, p = 0.002), whereas both ∆dynamic range (∆PD~~100−30~~: R^2^ = 0.12, p = 0.003) and ∆latency (∆PD~~100−30~~: R^2^ = 0.23, p = 0.002; and ∆PD~~100-50~~: R^2^ = 0.36, p < 0.001; Fig. 6E) correlated negatively with SGC survival. The earlier correlations indicate that—with a short IPG − ∆PD~~100−30~~ exhibits the best predictive potential for neural survival with slope, dynamic range, and latency, while a relationship between ∆PD~~100−30~~ and neural survival for the amplitude, threshold, and level~~50%~~ was decidedly absent.

When using an IPG of 30 µs (Fig. 9, middle and lower rows), there were no significant correlations between ∆PD~~50−30~~ or ∆PD~~100−30~~ and SGC survival. A negative correlation was observed between ∆~~PD~~100−50~~ slope and neural survival (R^2^ = 0.19, p = 0.005) and ∆~~PD100−50~~ latency (R^2^ = 0.15, p = 0.013; see also Fig. 6F). This would indicate that at an IPG of 30 µs, both ∆slope and ∆latency are possible predictors for neural survival, albeit only while varying PD between 50 and 100 µs.

In short, when using ∆PD to predict neural survival, ∆PD~~100−30~~ yielded the most predictive potential results with slope, dynamic range, and latency for an IPG of 2.1 µs. While ∆PDlatency can be a suitable predictor for neural survival when using both IPGs combined with longer PDs, the overall predictive power of ∆PD disappears for long IPGs.

DISCUSSION

A significant effect of elongating the PD of the biphasic current pulse was observed for all six eCAP measures, which was large for the N1 latency and small for the five AGF-derived measures. This PD effect was correlated to neural survival for two of the six eCAP measures (slope and N1 latency), fading with the long IPG. The IPG effect was generally larger than the PD effect, but it decreased with increasing PD.

A Hypothesized Mechanism for the IPG and PD Effect

The IPG effect is the change in response to electrical pulse stimulation (here, a change in the eCAP) caused by an increase in the IPG. We assume that with biphasic pulse stimulation, one phase functions as an excitatory phase and the other as an inhibitory phase. A longer IPG increases the temporal separation of the opposing charge buildup of both phases, diminishing the effect of the inhibiting phase and thereby reducing the chances of a negated action potential (van den Honert & Mortimer 1979). Similarly, the PD effect essentially arises from the separation of the time PCCA to evoke a neural response (see Fig. 10 for a schematic). Increasing the PD separates the PCCAs of the first and second phases. Indeed, we found that a longer PD resulted in a stronger response (Fig. 3), which we assume is caused by the reduced effect of the inhibiting phase, by the same underlying mechanism as the IPG effect. Crucially, however, increasing the PD has a smaller effect on neural excitation than increasing the IPG (Fig. 3). We ascribe this to two mechanisms that counter the effect of the separation of the opposing charges. First, the cell membrane leakiness reduces the efficacy of charge buildup, in particular for PD >50 µs, a mechanism known as leaky integrator (Zhou et al. 2021; Skidmore et al. 2022); this effect is reflected by the amplitudes, which slightly decreased when the PD was increased from 50 to 100 µs (Figs. 3A, B). Second, the neural discharges are spread in time by the increase of PD, which leads to desynchronization, reducing the eCAP amplitude. Further increasing IPG or PD does not linearly increase stimulus effectiveness. For IPG, this was illustrated by Ramekers et al. (2014; their Fig. 5) as increasing IPG from 2.1 to 10 µs had a stronger effect than from 10 to 20 µs, or from 20 to 30 µs. Likewise for PD, increasing PD from 30 to 50 µs generally had a larger effect than from 50 to 100 µs (Fig. 3), probably enhanced by the earlier-mentioned leaky integrator and desynchronization of neural discharges. The overall smaller effect for the 50 to 100 µs increase in PD agrees with the absence of a PD effect in the human CI study of He et al. (2020b) applying a 50 to 88 µs change in PD.

The LIFs, the remarkable eCAP latency increases with increasing PD, are informative towards mechanisms of the IPG and PD effect. Because we applied alternating polarity, the excitatory phase alternates between the first and second phases, and the LIF will depend on which of the two contributes most. Figure 10 illustrates three scenarios with PCCA at different points in the at the start (Fig. 10A), in the middle (Fig. 10B), and at the end (Fig. 10C). As most LIF values were found around 1.5, we consider LIF = 1.5. This LIF can be explained by a PCCA in the middle of the phase, with only the second phase contributing to the eCAP, or by a PCCA at the end of the phase, with both first and second phases equally contributing. Most likely, it would be between those two scenarios, with low-threshold neurons responding around the middle and high-threshold neurons toward the end of the phase. Note that the membrane leakiness will have contributed to the elongation of the latency when increasing PD to 100 µs, putting the PCCA around the end of the phase, favoring the second scenario. Because LIF >1 in the present data, we conclude that the second phase significantly contributes to the neural responses. In deafened animals with LIF >1.5, the contribution of the second phase must be larger than that of the first phase, even considering PCCA at the end of the phase.

The PD Effect and Its Predictive Potential on Neural Health

The PD effect was the strongest for the largest ∆PD tested (70 µs; Table 2; Fig. 3). We observed a significant interaction between PD and group for this ∆PD. Such an interaction was not found by Ramekers et al. (2014), most likely because their PD range (20 to 50 µs) was too small, considering we did not find an interaction for a PD change from 30 to 50 µs (Table 2).

For a short IPG (2.1 µs), ∆PDslope, ∆PDdynamic range, and ∆PDlatency correlated significantly with SGC survival when increasing the PD from 30 to 100 µs, with ∆PDslope decreasing and both ∆PDdynamic range and ∆PDlatency increasing with SGC loss. The IPG effect of these measures showed a similar directional correlation when increasing the IPG from 2.1 to 30 µs (Fig. 7). The PD increase (70 µs) is more than twice that of the IPG increase (28 µs) to reach a similar effect. This means that separating excitatory and inhibitory PCCAs by lengthening the PD has a weaker effect than by lengthening the IPG. This explains that He et al. (2020b) observed no changes in neural responsiveness in human CI users when increasing the PD from 50 to 88 µs (38 µs increase) at an IPG of 7 µs.

The strong interactions between PD and IPG are well expressed by the positive ∆PD effect of slope for NH and 3WD animals for the short IPG, and a negative ∆PD effect for the long IPG (Fig. 5B). Furthermore, the change in threshold when increasing PD from 30 to 50 µs was reversed from short to long IPG for both NH and deafened animals (Figs. 3E, F). These strong PD-IPG interactions, also generally observed for the other eCAP measures (Table 2, Fig. 5), can be explained by the opposing effects of (1) temporal separation of the PCCAs which reduces the effect of the inhibitory phase and (2) the inherently detrimental consequences of a long PD—“leakiness” of the membrane (Skidmore et al. 2022) and desynchronization of neural responses. “Leakiness” here refers to neural membranes being an imperfect insulator, resulting in charge being gradually lost across the membrane, making long PDs less effective (Skidmore et al. 2022, their Fig. 3B). Zhou et al. (2021) reported such “leakiness” in human CI recipients already by increasing the PD from 25 to 50 µs, with more profound effects for longer PDs (100 to 400 µs). Furthermore, by increasing PD, the charge spread over time can lead to a reduction in synchrony of neural activity, which broadens the eCAP waveform and diminishes its amplitude (Figs. 2C, D; this is a subtle effect). Also, the “leakiness” could be amplified by demyelination (i.e., fewer layers of myelin) in deafened animals resulting from neural degeneration (Agterberg et al. 2008). This may explain the negative ∆PD effect of slope at the short IPG for the long-term deafened animals (6WD, 7WD, Fig. 5B). However, while demyelination has been observed regarding SGC soma (Agterberg et al. 2008), it has not consistently been observed for peripheral processes (Waaijer et al. 2013; Ramekers et al. 2020). In conclusion, for long IPGs, the positive effect of (further) separating the PCCAs by elongating PD is counteracted more strongly than for short IPGs.

The Influence of PD on IPG Effect

In all animals, increasing the IPG led to a higher amplitude, steeper slope, lower threshold, wider dynamic range, lower level~~50%~~ and a general increase in latency (Fig. 4). These IPG effects, and the correlations with neural survival (Fig. 7), were generally largest for a PD of 30 µs, diminishing with increasing PD and all but dissipated for a PD of 100 µs. Ramekers et al. (2014), varying PD from 20 to 50 µs, observed similar effects of PD on the IPG effect. These observations can be explained by the separation of the PCCA mechanism for the IPG and PD effect (see 4.1). Clear IPG effects have been reported in guinea pigs with a PD up to 50 µs (Ramekers et al. 2014, 2015, 2022; Schvartz-Leyzac et al. 2019, 2020b; Vink et al. 2020, 2022, 2023).

For applications of the IPG effect in CI recipients, a short PD would be preferable. Most human studies have indeed reported relatively short PDs for the application of the IPG effect in neural health assessment (25 to 50 µs in Hughes et al. 2018; Schvartz-Leyzac & Pfingst 2018; He et al. 2020a; Schvartz-Leyzac et al. 2020a, 2021; Brochier et al. 2021; Zamaninezhad et al. 2023; Sijgers et al. 2024).

Latency

The short eCAP latencies in animals with long-term deafness (6WD, 7WD) compared to those in NH and short-term deaf animals (Fig. 6) agree with previous acute and chronic studies (Ramekers et al. 2014, 2022, respectively). In the chronic study with eCAPs recorded weekly, the latency first increased up to 3 weeks after deafening, then decreased and finally stabilized around 5 weeks after deafening. A smaller cell size, which is observed during degeneration (Van Loon et al. 2013), results in faster depolarization, and the gradually increasing cell loss results in less densely packed cells, causing a lower impedance, leading to an increased excitability, also explaining a shorter latency (Skidmore et al. 2022).

The PD effect of latency was strong, and it was correlated with neural survival at both IPGs (Figs. 8 and 9; Table 4) with similar R^2^ values (~0.3) as IPG effects (Table 3). This PD effect could reflect electrophysiological changes of the individual surviving neurons because of degeneration (cell shrinkage or demyelination) apart from numerical survival, as argued by Ramekers et al. (2022) for the IPG effect. The measure of the PD effect of latency, the LIF, is not only indicative of neural health, but it also reveals the extent to which the auditory nerve responds to the first and second phases, which would be particularly interesting to examine in human CI users.

Concluding Remarks and Clinical Implications

Elongating the PD resembles the effect of increasing the IPG on the eCAP, as it temporally separates charge. However, spreading out the charge in time with longer PDs, instead of a short, strong current, leads to more ambiguous results than the clear-cut IPG effect. The observation of a nonequivalent increase in N1 latency to PD (LIF ≈ 1.5) indicated a contribution to the eCAP of both the first and second phases of the pulse. Toward application in CI users, the LIF seems to be attractive and worthwhile to be studied. Using a forward-masking paradigm, the LIF can be examined for anodic-leading and cathodic-leading pulse configurations separately. One would expect the pulse with the second phase being excitatory to be the more predictive for neural survival. The IPG effect of eCAP measures has already been considered by several groups for application in CI users, as it correlates well with SGC survival, and better than absolute eCAP measures, because it is less dependent on non-neural factors as electrode location (Schvartz-Leyzac et al. 2020b). When using the IPG effect, a short PD is advised, as strongly indicated by our current data. Complications arise when one is faced with high electrode impedances (thus low compliance), urging one to make choices regarding PD, maximum current level, and selective exclusion of high-impedance electrode contacts. For application of the LIF or the IPG effect of latency, considering their ranges with neural survival (Figs. 4F, 6A, B), measurements with at least 5 µs precision are required (thus at sampling rates ≥200 kHz), which presently not all devices allow. Last, we recommend obtaining latencies at high current levels where eCAPs are generally robust and well above the noise floor (say ~~50 µV), thus allowing the N~~1~ peak to be well identified (Glassman & Hughes 2013).

ACKNOWLEDGMENTS

The authors thank Ferry Hendriksen for histological processing and analysis, and Peter Zuithoff (Julius Center, UMC Utrecht) for support on the statistical analyses.