Discussion

A preliminary goal of this study was to characterize latencies of EABR wave IIIe, wave Ve, and interval IIIe-Ve with regard to stimulation site so as to study the influence of the auditory pathway anatomy on latency. The study also aimed to describe how EABR latency and the degree of auditory deprivation before implantation relate. To achieve these goals, EABRs were elicited on all the contacts along the electrode array at comfortably loud intensities. The most comfortable levels at first fitting of the cochlear implant, the audiometric pure tones thresholds (for 500, 1000, 2000, and 4000 Hz) prior to implantation and the duration of deafness were taken into account to assess the degree of auditory loss and investigate its influence on EABR latency.

The protocol applied in this study allowed examination of tonotopic elements of the auditory pathway with some specificity. It was demonstrated previously that bipolar excitation of the human cochlea produces place-specific neural excitation in humans (Lim et al., 1989; Abbas and Purdy, 1990). The stimulated segment on the cochlea can indeed be very small (Black et al., 1983; O’Leary et al, 1985). Animal studies showed a high correlation between the response and the frequency-place excitation parameters in the inferior colliculus (Black and Clark, 1980; van den Honert and Stypulkowski, 1987). However, it is worth noting that using monopolar stimulation and high stimulus levels for the EABRs in our protocol may have some effect on the place of initiation of neural action potentials along the length of the axon and peripheral process of the neuron, which could affect latency functions. Indeed, monopolar stimulation initiates action potentials at more central sites than bipolar stimulation does as shown by ECAP latency functions (Miller et al., 2003). The place of stimulation may then be more central when there is more neural spread of excitation around the electrode sites, resulting in shorter latencies. Higher intensities of stimulation could also activate more central places as latency was shown to decrease with higher stimulus levels (Firszt et al., 2002) and this effect may interact with the effect due to monopolar stimulation as well. However, it seemed important to record the EABRs with the same monopolar configuration as used for M levels measurement in the present study. This mode of stimulation is also usually used when EABRs are recorded for clinical purpose and in implant users’ everyday programs. Moreover, several arguments enable us to suppose that there was some specificity in the place activated. On the first hand, pitch discrimination is better with monopolar stimulation than with bipolar stimulation (Pfingst et al., 1997) and pitch discrimination is better at higher intensities (e.g., Pfingst et al., 1999). Morris & Pfingst (1999) even suggested that the current level adjustments necessary to achieve comparable loudness for the various configurations (broader versus narrower configurations) may significantly counter any effects of electrode configuration on the size of the activated neural population. On the other hand, the fact that differences in latency and M levels were observed for each stimulation site in this study for most subjects supports the theory that regions of excitation are generally constrained and that different populations of distal fiber processes are stimulated. It was assumed thus that this protocol was well placed to study the effects of the anatomy of the auditory pathway periphery (i.e., the place of stimulation along the electrode array in the cochlea) and the state of the auditory nerve fibers on EABR latencies even if due to lack of histological data the discussion of possible anatomical origins for the electrophysiologic findings are speculative.

It was possible to record EABRs for all electrodes from all subjects as shown by Figures 2 and 3. Wave Ve was the most robust and was present on all electrodes, unlike wave IIIe. Wave I could not be identified due to stimulus artefact, while waves IIe and IVe were seen in some subjects only. Latencies of waves IIIe and Ve were in agreement with the values obtained by other researchers. As in most previous studies, the latencies of waves IIIe and Ve were shorter than the latencies of acoustically evoked potentials, which may be explained by enhanced neural synchrony with electrical stimulation (Firszt et al., 2002) and the fact that the active biomechanics of the cochlea are shunted in cochlear implant users (Starr & Brackmann, 1979; Van den Honert & Stypulkowski, 1986). Interpeak IIIe-Ve latencies were found to be the same as for ABR (1.8 ms) accordingly to Waring (1992, 1995). The influence of stimulation site on EABR latency is shown in Figure 4. These results verify what had been found by Shallop et al. (1990), Abbas & Brown (1991), and Miller et al. (1993). They demonstrate a shift in EABR latency for different sites of stimulation in the cochlea for wave Ve with latencies 0.43 ms higher at the base than at the apex as Allum et al. (1990) and Firszt et al. (2002) had previously found. A significant decreasing baso-apical latency gradient was also observed for wave IIIe, while there is no effect of stimulus site on IIIe-Ve suggesting that effects on waves IIIe and Ve are due to differences in areas more peripheral to IIIe.

The anatomical factors which explain the latency increase toward the base may be understood in terms of auditory fibers length. Decreasing baso-apical fiber-length gradients were found in the basal turn of animal cochleae (Fernandez, 1952; Liberman & Oliver, 1984). In humans, the auditory nerve from the cochlea to the cochlear nucleus has been shown to be shorter at the apex (Moore, 1987) and the peripheral axons of the human cochlea are 0.3 mm longer in the lower basal region than in the upper basal turn (Spoendlin & Schrott, 1989). Moreover, shorter latencies at the apical part of the array may be explained by the fact that the apex of the electrode array lies at about 1.5 cochlear turns where the number of nerve fibers per mm of cochlea length is higher in animals (Spoendlin, 1972) and humans (Hinojosa et al., 1985; Spoendlin & Schrott, 1990) compared to the base. A relationship between the number of neural fibers and conduction velocity was indeed shown in the auditory pathway by Rattay (1987), as well as in the visual pathway (Cavalcanti do Egito Vasconcelos et al., 2003) and the motor pathway (Morgan & Proske, 2000). Due to their number, if some of them are degenerated because of deafness, it is also likely that more fibers of higher quality remain in this area. The difference in the length of the auditory nerve, the possibility that more rapid synaptic transmission occurs at the apex and that neural fibers are more numerous may be a physiological way to compensate for the delay the acoustic wave takes to reach the cochlea apex in a non-implanted ear. Indeed, in the normal ear, ABR latencies may be longer for low pitched tones because of the delay in these tones reaching the apex of the cochlea (Allen, 1980; Don et al, 1998).

However, the distribution of EABR latencies along the electrode array is not the same for all subjects as illustrated by Figure 5. Indeed, latencies of Subjects 2, 6 and 8, who had more than 9 years of duration of deafness showed no baso-apical gradient corroborating Propst et al. (2006) findings of similar latencies in the base and apex in subjects with congenital deafness. An explanation can be raised from the observation that higher intensities were needed to generate a comfortably loud sound at the base in Subject 2 (see Figure 5). The base would be in such a bad state that relatively few auditory cells would remain (Otte et al., 1978, Schmidt, 1985) and that current would need to be high enough to stimulate other cochlea turns resulting in the same latency over the array (Frijns et al., 1996). This would explain why Cohen et al. (1996) found that tone function is variable across subjects. It also provides an explanation for the fact that pitch is more poorly discriminated in subjects with higher behavioural thresholds (McDermott & McKay, 1994) and poorer speech perception (Busby et al., 1993; Zwolan et al., 1997; Henry et al., 2000; Donaldson and Nelson, 2000), as well as for the fact that the width of neural excitation profile increases for subjects with higher M levels (Cohen et al., 2003).

The M levels at first fitting were assumed to reflect the auditory fiber state of degeneration since it was believed that higher intensities of stimulation would be necessary for the subjects to have a comfortable perception when nerve fibers are more damaged. M levels significant increase at the base as shown by Figure 6 thus implies that pathological factors such as demyelination or degeneration may also explain the shift in EABR latency. Given that the perception of high pitch sounds is generally lost first, there could be a shift in the EABR latency due to a baso-apical gradient of neural degeneration resulting in greater degradation of distal processes of primary afferent fibres toward the base (Abbas and Brown, 1991). Auditory fibers were indeed shown to be less intact at the base in animals (Ylikoski et al., 1974; Leake and Hradek, 1988; Miller, 1992) but since in humans no histological data can corroborate our assumptions, relationships between EABR latency and various pathological factors were investigated.

If indeed the M levels reflect the state of degeneration of auditory fibers, it is reasonable to believe that latencies would increase with M levels: damaged neural fibers would require more intensity for the subject to perceive a comfortable sound and the neural transmission speed would be reduced because of damage to the myelin sheath for instance. Prolonged EABR latencies were indeed found in poorly myelinated mice (Zhou et al., 1995). As expected, stabilized latencies of waves IIIe and Ve were observed to increase significantly with higher M levels as is illustrated by Figure 7. It is however worth noting that latencies for electrodes with very low M levels (below 89 CU) are higher than would be expected. This had previously been described in the guinea pig (Nicolas-Puel et al., 1996). The authors showed that less damaged cochlear areas required less current to be stimulated and that it resulted in longer latencies, which may be due to the fact that the place of stimulation at low levels may be more peripheral. In addition, Figure 8 shows that EABR latency and M level for basal electrodes tend to be higher than values of the rest of the array. Hence, fibers at the base are indeed more damaged, while less damaged fibers (of M level group 1) may be mixed with moderately damaged ones (M levels groups 2 and 3) and scattered along the cochlea except for the base. The main tendency is however that auditory fibers in good state convey more rapidly the neural influx than damaged fibers since latencies for better audiometric thresholds were found to be significantly shorter as illustrated by Figure 9.

In addition, the multiple linear regression analysis shows that the more basal the stimulation site and the longer the duration of deafness, the longer the wave IIIe latencies are. No linear relationship was found between the M levels and latency, which is consistent with the fact that latencies for the lowest M levels was found to be higher than for higher M levels. This regression analysis also shows that wave Ve latencies reflect wave IIIe latencies. The significant interactions of the duration of deafness with the audiometric data and the M levels for waves IIIe and Ve latencies found with ANOVA analysis also show that deafness has an impact on latency from a certain duration only. Hence, for short duration of deafness, waves IIIe and Ve latency would be mostly influenced by the anatomy of the auditory pathway, while for longer deprivation, deafness would have a larger impact on the auditory fibers below the cochlear nucleus in addition to the anatomical effects (shown to have no interaction with duration of deafness).

In conclusion, this study describes EABR waves IIIe and Ve latencies, and interpeak IIIe-Ve characteristics across stimulation site for eight cochlear implant users. Comparison with previously published observations on electrically evoked potentials was made. The possible explanations for the decreasing baso-apical latency gradient observed are discussed in the light of anatomical parameters and the degree of auditory deprivation prior to cochlear implantation. Interesting correlations were found with the audiogram before cochlear implantation, the M levels at first fitting of the cochlear implant, and the duration of deafness showing that EABR latencies reflect well the state of degeneration of auditory fibers. However, the number of subjects participating to this study was small and it would be necessary to enlarge our sample to better distinguish the effects of duration of deafness versus residual hearing and assess the effects of age at onset of deafness and etiology. There are also limitations to the study such as the fact that the whole cochlea cannot be studied since only a part of the cochlea is addressed by the electrode array and that the interpretation of results is limited by the lack of corroborating histological evidence. This study needs therefore to be deepened. Further investigation should take into account EABR amplitude as this was shown to be correlated to the number of spiral ganglion cells (Smith & Simmons, 1983; Hall, 1990). It would also involve taking into account the position of the electrode array within the cochlea, since this influences EABR amplitude (Firszt et al., 2003), by comparing subjects with and without positioner. This study could also be deepened by studying further the neural spread of excitation around the electrodes by comparing bipolar and monopolar type of stimulation effects on EABR at various intensities and stimulation sites and correlating electrophysiological findings with pitch perception.