Hearing Aid Treatment for Patients with Mixed Hearing Loss. Part II: Speech Recognition in Comparison to Direct Acoustic Cochlear Stimulation

Abstract

Background

The conventional therapy for severe mixed hearing loss is middle ear surgery combined with a power hearing aid. However, a substantial group of patients with severe mixed hearing loss cannot be treated adequately with today's state-of-the-art (SOTA) power hearing aids, as predicted by the accompanying part I of this publication, where we compared the available maximum power output (MPO) and gain from technical specifications to requirements for optimum benefit using a common fitting rule. Here, we intended to validate the theoretical assumptions from part I experimentally in a mixed hearing loss cohort fitted with SOTA power hearing aids. Additionally, we compared the results with an implantable hearing device that circumvents the impaired middle ear, directly stimulating the cochlea, as this might be a better option.

Objectives

Speech recognition outcomes obtained from patients with severe mixed hearing loss supplied acutely with a SOTA hearing aid were studied to validate the outcome predictions as described in part I. Further, the results obtained with hearing aids were compared to those in direct acoustic cochlear implant (DACI) users.

Materials and Methods

Twenty patients (37 ears with mixed hearing loss) were provided and fitted with a SOTA power hearing aid. Before and after an acclimatization period of at least 4 weeks, word recognition scores (WRS) in quiet and in noise were studied, as well as the speech reception threshold in noise (SRT). The outcomes were compared retrospectively to a second group of 45 patients (47 ears) using the DACI device. Based on the severity of the mixed hearing loss and the available gain and MPO of the SOTA hearing aid, the hearing aid and DACI users were subdivided into groups with prediction of sufficient, partially insufficient, or very insufficient hearing aid performance.

Results

The patients with predicted adequate SOTA hearing aid performance indeed showed the best WRS in quiet and in noise when compared to patients with predicted inferior outcomes. Insufficient hearing aid performance at one or more frequencies led to a gradual decrease in hearing aid benefit, validating the criteria used here and in the accompanying paper. All DACI patients showed outcomes at the same level as the adequate hearing aid performance group, being significantly better than those of the groups with inadequate hearing aid performance. Whereas WRS in quiet and noise were sensitive to insufficient gain or output, showing significant differences between the SOTA hearing aid and DACI groups, the SRT in noise was less sensitive.

Conclusions

Limitations of outcomes in mixed hearing loss individuals due to insufficient hearing aid performance can be accurately predicted by applying a commonly used fitting rule and the 35-dB dynamic range rule on the hearing aid specifications. Evidently, when outcomes in patients with mixed hearing loss using the most powerful hearing aids are insufficient, bypassing the middle ear with a powerful active middle ear implant or direct acoustic implant can be a promising alternative treatment.

Introduction

Hearing aid provision and fitting in patients with mixed hearing loss is not straightforward. Following prescription rules, in principle, full compensation of the air-bone gap (ABG) is required, as well as a substantial proportion of the sensorineural component [Wardenga et al., 2020]. Therefore, for treatment of mixed hearing loss, hearing aids have to be powerful in terms of output and stable gain. Evaluated at the cochlear level, the available gain and output of conventional hearing aids are effectively lowered by the width of the ABG. Especially in case of a wide ABG, this significantly limits the available gain and maximum output of the hearing aids at the level of the (impaired) cochlea. Moreover, such power hearing aids have a limited bandwidth. Especially in the high frequencies, the available gain is limited, as is illustrated in the accompanying paper [Wardenga et al., 2020]: the available gain at 4 kHz of powerful state-of-the-art (SOTA) hearing aids is approximately 20 dB below that at 1 kHz (see Table 1a in Wardenga et al. [2020]). Dillon [2012] (in chapter 10) argued that for patients with severe, pure sensorineural hearing loss (the main target group for power hearing aids), decreased high-frequency amplification might be beneficial as most patients might have hearing loss that is sloping, with the poorest hearing in the high-frequency range. As a consequence, the analytical power of the cochlea (e.g., tempo-spectral processing of sound) is poor at these frequencies. Therefore, it has been argued that limiting the gain in the high-frequency range might effectively reduce poorly processed high-frequency information, and fitting rules for hearing aids have been adapted accordingly [Dillon, 2012].

Table 1.

Overview of the patient groups for HA users and DACI patients

Demographic data	HA users	DACI users
Patients/ears, n	20/37	45/47
Gender (male/female), n	9/11	16/29
Mean age ± SD (min.–max.), years	61±18 (23–85)	60±13 (25–79)
Pathology	various	otosclerosis
Surgery/no treatment, n	10/27	14/33
Mean ABG ± SD (min.–max.)	14.5±9.5 (3.8–45.0)	20.7±10.8 (0.0–41.8)a
		35.3±12.5 (7.5–57.5)

HA, hearing aid; DACI, direct acoustic cochlear implant; ABG, air-bone gap. ^a After hypothetical stapes surgery (corrected air conduction).

However, for patients with mixed hearing loss using these power hearing aids, reducing the gain in the high-frequency range is not necessarily beneficial, because although the total hearing loss might be severe, the cochlear hearing loss (or the sensorineural hearing loss component) in the high-frequency range might be less than severe, which implies that the analytical power of the cochlea in that frequency range is not necessarily as impaired as in sensorineural hearing loss.

Previously, in the accompanying publication, we studied the fraction of patients with mixed hearing loss (owing to otosclerosis) that can or cannot expect sufficient benefit from SOTA power hearing aids. There we evaluated the outcomes of 374 consecutive patients and calculated the patients' individual postoperative required gain and output for optimal clinical outcomes. These requirements were compared to the available gain and maximum power output (MPO) provided by current SOTA power hearing aids at the frequencies of 0.5, 1.0, 2.0, and 4.0 kHz. It was shown that the required gain and/or required maximum output at one or more frequencies could not be delivered by these hearing aids in approximately 25% (80 of 357 ears) before and still 15% (31 of 206 ears that needed amplification) after a stapes surgery. The study performed in part I could demonstrate that in a non-pre­selected cohort, representative of our clinic, a significant proportion of patients cannot benefit sufficiently from the combination of surgical intervention and hearing aids when generally accepted theoretical criteria for required gain and output are applied.

The primary goal of the present study was to verify the theoretical assumptions of part I experimentally with adequate real-world hearing aids in patients with mixed hearing loss. Additionally, during the past decades, new amplification options have become available for patients with mixed hearing loss, such as auditory implants with their actuator coupled directly to the cochlea or the middle ear ossicles, bypassing the impaired middle ear [Snik, 2011]. As the combination of stapes surgery and hearing aids potentially leaves a gap of insufficiently treatable patients, the present study aimed at a comparison of the conventional solution with such an implantable device; more specifically, speech recognition performance obtained with SOTA conventional power hearing aids is compared with that obtained with an auditory implant that directly stimulates the cochlea: the Cochlear^TM Codacs^TM direct acoustic cochlear implant (DACI). The implanted actuator of the DACI is connected to stapes prostheses, stimulating the cochlea directly [Busch et al., 2013; Lenarz et al., 2013]. As a consequence, the gain and maximum output of this device are not limited by the ABG. Furthermore, measurements have shown that the DACI actuator can provide a broad bandwidth and high MPO [Grossöhmichen et al., 2015].

Subjects and Methods

Patients and Demographics

Patients with severe bilateral hearing loss having an air conduction (AC) pure tone average (PTA; mean threshold of 0.5, 1.0, 2.0, and 4.0 kHz) threshold >55 dB HL and using bilateral hearing aids were invited to participate. Initially, only patients with (surgically treated or untreated) otosclerosis were invited; owing to the low numbers, patients with an ABG of different origin (surgically treated or not) were invited as well. In contrast to part I [Wardenga et al., 2020], where the selection was done based on pathology to provide a representative cohort of patients seen at our clinic, the selection of patients here was performed based on the severity of the hearing loss to provide a cohort that covers the range where conventional hearing aids reach their potential limits.

Twenty patients (9 male and 11 female) agreed to participate and were included in the study. The patients had an average age of 61 years, ranging from 23 to 85 years. Five additional subjects were excluded from the study: 2 subjects declared they lacked time, 1 subject rejected the ENZO hearing aids, 1 patient suffered from sudden hearing loss 2 days after the inclusion, and 1 patient did not understand the tasks of the audiological tests. The study was performed at the Department of Otorhinolaryngology, Hannover Medical School, between April and December 2016.

For comparison, data were used as obtained from a retrospective analysis of 45 patients (47 ears) implanted with the DACI device. All these patients suffered from otosclerosis. Most patients in the DACI group had a history of surgical interventions. The mean age of these patients was 60 years, ranging from 25 to 79 years, at the time of implantation. The data on the DACI patients were taken from routine examination 3 months after the implantation. Only patients who had all audiological tests available were included in the retrospective part of the study. The implantations were conducted between November 2009 and February 2018. An overview of the patients included in both analyzed groups is given in Table 1.

Audiological Tests

Pure tone audiometry and the sound field speech test with the walk-in (WI) devices were carried out at the first appointment. Audiometry was performed using HDA200 headphones (Sennheiser electronic GmbH and Co. KG, Wedemark, Germany) and a KLH96 bone conductor (CB-Elmec GmbH, Radeberg, Germany) coupled to a PC-based audiometer (AD2017 or AD17; Audio-DATA GmbH, Duvensee, Germany). To determine the degree of hearing loss, AC and bone conduction (BC) thresholds were measured in 5-dB steps at 0.5, 1.0, 2.0, and 4.0 kHz. In cases where the audiometer limit was exceeded, missing values were replaced by a best-case estimate using the audiometer limit plus 5 dB; the audiometer limit was 110 dB HL for AC testing at all measured frequencies. For the BC thresholds, the audiometer limit was 65, 80, 80, and 75 dB HL at 0.5, 1.0, 2.0, and 4.0 kHz, respectively. Masking was applied according to standard procedures. Ears with 3 or all 4 BC thresholds exceeding the limit of the audiometer were excluded from the analysis, as these ears might have been (at least partially) deaf; as a consequence, 3 ears had to be excluded and 37 ears were included.

The protocol for speech recognition testing was carried out as described by Kludt et al. [2016]. The tests were performed in the sound field with a loudspeaker 1 m in front of the subject. Individual ears were tested separately, since the monaural performance of the respective device was the primary goal. While both hearing aids were worn, only one hearing aid was switched on, while the ear mould of the other hearing aid blocked the untested ear. A similar procedure consisting of plugging and/or masking was used in the DACI group.

To determine the word recognition score (WRS) in quiet, the Freiburg monosyllable test was used, comprising sets of 20-item monosyllable word lists. Percentage correct word scores were collected at a fixed presentation level of 65 dB SPL. The WRS was also tested in noise, using the Hochmair-Schulz-Moser (HSM) sentence test. One test list consists of 20 everyday sentences with a total of 106 words. These measurements were performed with a fixed signal-to-noise ratio (SNR) of +10 dB; the speech level was 65 dB SPL with a 55-dB SPL noise level.

A second speech-in-noise test was performed: the German version of the International Matrix Test (also known as the Oldenburg Sentence Test (OLSA) [Kollmeier et al., 2015]), using sentences presented by a male speaker with competing steady-state, speech-shaped noise. The noise was presented at a fixed level of 65 dB SPL. The level of the sentences was increased or decreased after each presentation, depending on whether or not the sentence was properly repeated by the patient [Brand and Kollmeier, 2002]. The outcome of the adaptive procedure of the test is the SNR, which is the speech reception threshold (SRT; the level at which half of the presented material is correctly understood) minus the noise level (65 dB SPL). In both speech-in-noise tests, the speaker in front of the patient presented both the speech and the noise.

In addition to providing the speech data, the participants were asked to report on their individual hearing situation at the first appointment and after the testing period with the SOTA hearing aids. The disability-specific Abbreviated Profile of Hearing Aid Benefit (APHAB) questionnaire was used [Cox and Alexander, 1995]. The APHAB queries the amount of communication difficulties experienced in different everyday situations. It assesses four domains: ease of communication (EC), speech perception in reverberation (RV) and in background noise (BN), and aversiveness of everyday sounds (AV). The APHAB average outcomes per domain were compared with normative data, obtained from patients with moderate-to-severe sensorineural hearing loss using bilateral digital hearing aids [Johnson et al., 2010].

Fitting Protocol

To minimize variables, all patients were refitted with a new SOTA power hearing aid (the same type for all patients), namely, the GN ReSound ENZO² 9 [GN ReSound A/S, 2019a]. Fitting took place at our audiology center and was always performed by one and the same experienced audiologist (N.W.), according to the manufacturer's instructions (ReSound ENZO 3D fitting guide [GN ReSound A/S, 2019b]), and afterwards fine-tuned further. Technical verification of the fitting was carried out on a 2-cc coupler. A second, additional fitting session to further optimize the fitting procedure was completed, on request, for 6 of the 20 patients. After an acclimatization period of at least 4 weeks (mean 54 days, range 26–90), the final evaluation was performed assessing the speech recognition performance in the sound field with the SOTA device. The speech recognition outcomes in quiet (Freiburg monosyllables) and in noise (HSM test and International Matrix Test) were compared with those obtained with the patient's own hearing aids (WI device) collected prior to fitting of the SOTA devices.

Grouping of the Patients

To determine whether the hearing device could satisfy the audiological needs of an individual patient, two decision criteria were defined as introduced before by Wardenga et al. [2020]. In short, speech presented at a normal conversational level must be audible, which can be translated into a required gain for proper hearing aid fitting. The required gain should comprise full compensation of the conductive part and approximately half of the sensorineural component (for details, see part I [Wardenga et al., 2020]). As a second criterion, the MPO of the hearing aid should be high enough to cover the patient's dynamic range of hearing, with a minimum of 35 dB [Zwartenkot et al., 2014; van Barneveld et al., 2018], resulting in a required output (see Table 2 for the applied criteria).

Table 2.

Estimates for required gain and output (rGAIN and rOUTPUT) for sufficient benefit in comparison to the available maximum gain (dB) and output (dB HL)

	Required gain/output – calculated for each individual patient using:		Available gain/output – SOTA power hearing aid (i) frequency index
			0.5 kHz	1.0 kHz	2.0 kHz	4.0 kHz
rGAIN, dB	1/2 BCi + ABGi + ki	≤	75	77	70	54
rOUTPUT, dB HL	ACi + 35 dB	≤	121	131	122	121

The gain and output limits of current state-of-the-art (SOTA) power hearing aids were calculated from specifications of three different brands for the frequencies (with index i) 0.5, 1, 2, and 4 kHz. With k_i = −5 dB for 0.5 kHz and 0 dB for 1, 2, and 4 kHz.

Next, following Wardenga et al. [2020], the individual ears of the patients were divided into three subgroups. Subgroup G0 comprised all the ears with the required gain and required output within the specifications (available gain and maximum output) of the SOTA power hearing aid at all 4 frequencies, while for subgroup G1, one or both requirements could not be met at just one frequency. Subgroup G2 comprised all the other ears, thus the ears where the requirements were not met at 2 or more frequencies, with potentially the poorest results. The limits for available output and gain as determined from the specifications of the SOTA power hearing aids and criteria if requirements can be satisfied for individual patients from part I are given in Table 2.

As the DACI circumvents the middle ear and thus the ABG, its indication is independent of the amount of the conductive component and should be based on the BC threshold only. However, to create DACI subgroups comparable to those for the hearing aid trial group, we created a hypothetical DACI cohort by subtracting the average improvement from surgery (0.5 kHz: 21.9 dB; 1.0 kHz: 17.9 dB; 2.0 kHz: 12.4 dB; and 4.0 kHz: 10.7 dB) from preimplantation AC thresholds (for details, see part I, Table 2 [Wardenga et al., 2020]). Only in cases where the correction would have led to ABGs <0 were AC thresholds limited to the BC threshold. The thus corrected group can be assumed a best-case estimate for the outcome in the DACI patient group if they had undergone stapes surgery instead of DACI implantation. Since most of the patients in the DACI group had a history of surgical interventions, the assumption of average improvement from a hypothetical surgery might be a very optimistic estimate of the expected benefit. Hence, subgrouping of the DACI patients into G0–G2 was performed analogously to that of the hearing aid users, based on measured BC and AC thresholds corrected for a hypothetical stapes surgery.

Statistical Analysis

The statistical analyses were performed with IBM SPSS Statistics for Windows version 23.0 (IBM Corp., Armonk, NY, USA). The data were tested for normal distribution (Kolmogorov-Smirnov test) and homogeneity of variance (Levene's test). The post hoc analysis of power was performed with SigmaPlot 14.0 (Systat Software GmbH, Erkrath, Germany). Differences were considered significant at p < 0.05.

Results

Figure 1 presents the audiometric data collected at the first appointment for the hearing aid users and before implantation for the DACI patients. The figure shows that the BC thresholds were comparable between the conventional hearing aid users and DACI users (significant difference only at 0.5 kHz; p = 0.025; Mann-Whitney U test), while the AC thresholds were worse in the group of DACI users, at all 4 frequencies and the PTA (p < 0.001; Mann-Whitney U test) (Fig. 1).

Fig. 1.

Unaided pure tone average (PTA) and hearing thresholds at 4 separate frequencies (0.5, 1, 2, and 4 kHz). Upper panel: bone conduction (BC) thresholds of the conventional hearing aid users (white) and direct acoustic cochlear implant (DACI) users (light gray). Lower panel: air conduction (AC) thresholds of the conventional hearing aid users (white) and DACI users (light gray), as well as after correction by mean improvement with a theoretical stapes surgery (dark gray). The medians are depicted by bold black horizontal lines and the boxes mark the interquartile ranges (i.e., 25th and 75th percentiles). After exclusion of outliers (circles), the error bars indicate the minimum-to-maximum range. Statistically significant differences (top: Mann-Whitney U test; bottom: Wilcoxon signed-rank test) are indicated by brackets.

This is not a surprise, because with the DACI available, the surgeon might decide not to reconstruct the ossicular chain (by stapedotomy, to optimize hearing for proper conventional hearing aid fitting) but instead to use a DACI for implantation. As a consequence, prior to implantation, DACI patients might have a significant ABG, as was found in this study. In accordance with the AC thresholds, the unaided speech-in-quiet baseline in terms of SRT was worse among the DACI patients (preoperative SRT: >110 dB SPL, 104.5/110.0 [median, 25th/75th percentile]) than in the hearing aid group (SRT: >98.2 dB SPL, 86.7/110.0 [median, 25th/75th percentile]).

When we computed hypothetical AC thresholds for the DACI group by subtracting the average improvement from stapes surgery from preimplantation AC thresholds and compared them to the AC data for the hearing aid users, these hypothetical AC thresholds of the DACI group showed no statistically significant differences (Wilcoxon signed-rank test), indicating similar average AC and BC thresholds for the hearing aid and the DACI cohort (Fig. 1).

Grouping

The ears were grouped according to the criteria for required gain and required output. Table 3 shows which of the cases belonged to subgroups G0, G1, and G2. The BC thresholds and PTA in the hearing aid and the DACI cohort, divided into subgroups G0–G2, were similar, with a minor difference (<10 dB) at 0.5 kHz (Fig. 1). The AC thresholds showed significant differences, which disappeared when the DACI results were corrected for average improvement with a hypothetical stapes surgery.

Table 3.

Grouping of ears according to the requirements for gain (rGAIN) and power output (rOUTPUT)

Subgroup	Requirements for rGAIN and rOUTPUT	SOTA HA, n (%)	DACI, n (%)a	DACI, n (%)
G0	Satisfied at all frequencies	13 (35.1)	13 (27.7)	7 (14.9)
G1	Not satisfied at 1 frequency	16 (43.2)	20 (42.6)	14 (29.8)
G2	Not satisfied at ≥2 frequencies	8 (21.6)	14 (29.8)	26 (55.3)
Sum		37 (100.0)	47 (100.0)	47 (100)

Allocation to subgroups G0–G2 was performed using the individual air conduction (AC) and bone conduction (BC) thresholds in the SOTA power hearing aid cohort, and the BC and corrected and raw preimplantation AC thresholds in the DACI cohort. In subgroup G0, both conditions were satisfied at 0.5, 1.0, 2.0, and 4 kHz, and in subgroup G1, at all except 1 frequency; in subgroup G2, the requirements were not satisfied at 2 or more frequencies. SOTA, state-of-the-art; HA, hearing aid; DACI, direct acoustic cochlear implant. ^a After hypothetical stapes surgery.

Hearing Performance with the WI Device and the SOTA Device

First of all, the outcomes with the SOTA power hearing aid were compared to those obtained with the WI device. Figure 2 presents the WRS obtained with either hearing aid. A data point above the diagonal represents a better score with the SOTA device. To determine whether a data point was significantly below the diagonal, the binomial model was used [Thornton and Raffin, 1978] with the significance level set at 5%.

Fig. 2.

Comparison of the word recognition score (WRS) obtained with the walk-in (WI) hearing aid (HA) (x axis) to the WRS obtained with the newly fitted state-of-the-art (SOTA) HA (y axis). a WRS in quiet, obtained with the Freiburg monosyllable test. b WRS in noise, obtained with the Hochmair-Schulz-Moser test. Symbols indicate individual results within the three different subgroups G0 (circles), G1 (triangles), and G2 (crosses). The dashed lines indicate limits for 95% critical differences, calculated according to Thornton and Raffin [1978].

Regarding the WRS in quiet (Fig. 2a), for 16 ears, performance with the SOTA hearing aid was significantly better than with the WI hearing aid – in contrast to 1 data point showing that performance was significantly better with the WI hearing aid. A similar picture was seen for the WRS in noise, where 17 subjects showed a better performance with the SOTA hearing aid versus 6 with the WI hearing aid. Because of the latter observation, the analyses were performed twice, once with the SOTA device and once with “the best hearing aid.” In “the best hearing aid” case, the better of the scores obtained with the SOTA device or with the WI device was used.

Speech Perception with the SOTA Device and DACI

Comparing the WRS in quiet, the subgroups G0, G1, and G2 of hearing aid users (Fig. 3a) revealed an obvious decrease, with the poorest WRS in quiet in the G2 subgroup. Although not statistically significant in subgroup G1, the decline was statistically significant in subgroup G2 (G0 vs. G2: p = 0.008; G1 vs. G2: p = 0.021; Kruskal-Wallis test with subsequent Bonferroni correction; power ≥0.829, 2-sided, α = 0.05). In contrast, the result for the DACI patients was constant across subgroups (no significant differences). Moreover, the SOTA hearing aid subgroup G2 was statistically significantly worse than all DACI subgroups, including the corresponding G2 subgroup (G0: p = 0.003; G1: p = 0.006; G2: p = 0.006; Kruskal-Wallis test with subsequent Bonferroni correction; power ≥0.990, 2-sided, α = 0.05).

Fig. 3.

Aided monaural speech intelligibility in the sound field for patient subgroups G0, G1, and G2 of hearing aid users (state-of-the-art device; white) and direct acoustic cochlear implant (DACI) users (gray). a Word recognition score (WRS) (Freiburg monosyllable test) in quiet. b WRS (Hochmair-Schulz-Moser test) at a fixed signal-to-noise ratio (SNR). c The SNR determined with the International Matrix Test after exclusion of outliers with an SNR >15 dB (not shown; 1 hearing aid user in G2 and 1 DACI user in G1). The median is depicted by black horizontal lines and the boxes mark the interquartile ranges (i.e., 25th and 75th percentiles). The error bars indicate the minimum-to-maximum range. Significance levels were determined using the Kruskal-Wallis test and subsequent Bonferroni correction.

Figure 3b presents data from the HSM speech-in-noise test, a test with a relatively low noise level (55 dB SPL) and a relatively high SNR (+10 dB). Similar trends are seen when comparing Figure 3a and b. The decrease in WRS in noise was less pronounced, partly owing to a larger spread in the results. Regarding the G2 subgroup, a significantly worse result was found for the hearing aid users in noise compared to both the DACI users in subgroup G1 and the DACI users in subgroup G2 (G1: p = 0.03; G2: p = 0.03; Kruskal-Wallis test with subsequent Bonferroni correction; G1: power ≥0.827, G2: power ≥0.922, 2-sided, α = 0.05). The median value was 54% lower among the hearing aid users. Comparing the subgroups of hearing aid users, the median WRS in noise dropped from 70% in subgroup G0 to 24–30% in subgroups G1 and G2.

Comparison of the hearing aid results to the overall DACI results without subdividing DACI users into subgroups (data not shown) showed that the WRS results in quiet and in noise of the hearing aid users with insufficient power (G1 and G2) were significantly decreased in comparison to the DACI results (in quiet: G2 vs. DACI, p < 0.001; in noise: G1 vs. DACI, p = 0.005, G2 vs. DACI, p = 0.005; Kruskal-Wallis test with subsequent Bonferroni correction).

In summary, DACI users demonstrated better speech recognition in quiet and in noise in subgroup G2 than those with conventional hearing aids (Fig. 3a, b). In contrast, significant differences in SNR were not found (Fig. 3c).

When using “the best hearing aid” (either SOTA or WI device) – i.e., the device with the best WRS – the results (data not shown) were rather similar to those shown in Figure 3, with the WRS in quiet of DACI patients being significantly better than that in the “best hearing aid” G2 subgroup (p = 0.009), and the WRS in noise being significantly better than in the “best hearing aid” G1 and G2 subgroups (G1: p = 0.022; G2: p = 0.031).

Abbreviated Profile of Hearing Aid Benefit

APHAB scores with the conventional WI and SOTA devices were obtained at the first and the last visit, respectively. When the APHAB results were compared between WI and SOTA devices, this revealed a significant improvement (Wilcoxon signed-rank test) with the SOTA device in subcategory EC (p = 0.007) and in the global score (the average of EC, RV, and BN; p = 0.022). For the SOTA devices, the mean domain scores (±SD) were: 23 ± 16% for EC, 48 ± 25% for RV, 47 ± 28% for BN, and 37 ± 24% for AV. A comparison to published norms (EC: 27 ± 21%; RV: 38 ± 19%; BN: 39 ± 21%; AV: 43 ± 27% [Johnson et al., 2010]) indicated nonsignificant differences in all subcategories except in subcategory RV (p < 0.05; two-tailed t test).

Discussion

In our previous paper [Wardenga et al., 2020], we calculated that conventional hearing aids might not suffice in approximately 25% of the cases before and 15% of the cases after stapes surgery on ears with mixed hearing loss. The present study validated the results and assumptions of part I, where the success rate of stapes surgery and hearing aid provision was determined in a clinically representative cohort. By investigating the theoretical assumptions of part I based on technical specifications in a cohort of patients with a broad range of mixed hearing loss, we could verify the appropriateness of the earlier approach experimentally.

A group of conventional hearing aid users were supplied and fitted with SOTA power hearing aids (ENZO² 9 device). According to the APHAB questionnaire, the patients' scores were better than those achieved with their WI devices and comparable to the norm, suggesting better and adequate fitting with SOTA hearing devices. Regarding speech recognition, again better scores were found with the SOTA device than with the patients' WI device in the majority of cases. Next, we compared conventional amplification with a relatively new option that directly stimulates the cochlea acoustically (the implantable DACI device) to explore the potential of this and similar technologies to fill the gap in available treatment options.

The audiological baseline values of the experimental hearing aid group and the DACI controls in terms of BC thresholds were comparable, with the exception of 0.5 kHz, at which frequency a small (<10 dB) difference was found. In terms of unaided AC thresholds, as expected the baseline values were significantly different between the hearing aid and the DACI cohort. However, correcting the AC thresholds of the DACI group with the frequency-specific average improvement expected from hypothetical stapes surgery led to AC thresholds similar to those found in the hearing aid cohort, without statistically significant differences. Although the expected benefit from surgical intervention presumably overestimates the improvement, as many of the DACI patients had a history of earlier interventions, the corrected hypothetical AC created a control group of similar baseline and average BC thresholds and ABGs.

Figure 3 presents the speech recognition scores obtained with the SOTA and DACI devices for each of the subgroups G0, G1, and G2. The WRS in quiet with conventional hearing aids declined statistically significantly across the subgroups, in contrast to the DACI results (Fig. 3a), demonstrating that the performance of SOTA hearing aids is decreasing and below its optimum when the gain and output requirements chosen here and in part I [Wardenga et al., 2020] are not satisfied. This in turn supports the conclusions drawn for a clinically representative cohort in part I by providing evidence that the theoretical criteria used there do reflect the limits of (even powerful) conventional hearing aids in a real-world setting.

Regarding the hearing aid users, the highest WRS in quiet was found for subgroup G0; the scores of subgroups G1 and G2 were significantly worse (p = 0.008 and p = 0.021, respectively). The median score of subgroup G2 was 35% below that of subgroup G0. Within subgroups, a comparison between hearing aid users and DACI users demonstrated significantly better hearing in quiet in all DACI subgroups compared to hearing aid subgroup G2 (Fig. 3a), when the conventional hearing aids reach their technical limits. The WRS-in-noise results (Fig. 3b) support this finding, most pronouncedly for subgroup G2, where the hearing aid results were inferior to those of the G1 and the corresponding DACI subgroup (both p = 0.03). This clearly demonstrates that hearing aids are suboptimal when limits are exceeded at 2 or more frequencies (subgroup G2). Although the aided APHAB results indicate an adequate hearing aid benefit in the present study, it should be noted that the investigated group was selected to cover the range where conventional hearing aids reach their limits. Hence, it cannot be excluded that the results obtained with conventional hearing aids in subgroup G0 already were close to the technical limit. However, the results of subgroups G1 and G2, reflecting borderline conditions for conventional hearing aids, are not subject to this bias.

As shown, for example, by Amlani et al. [2002], speech-recognition-in-quiet scores evidently depend on speech audibility. Figure 3 suggests that conventional hearing aids are less adequate than DACIs regarding speech recognition in quiet; the score of subgroup G2 in particular was poor, which is indicative of insufficient amplification/output. Specifically the majority of the subjects in subgroup G2 suffered from a limited MPO. Hence, a reduced MPO and output dynamic range available to patients constitutes a main factor for insufficient benefit. As expected, according to the inclusion criteria, the required gain and/or MPO provided by conventional hearing aids was not adequate for the patients in subgroup G2.

In contrast to the speech-recognition-in-quiet score, the SNR does not primarily depend on audibility but rather assesses “cochlear distortion” [Plomp, 1978], on the assumption that the noise is audible. Cochlear distortion refers to proper cochlear processing of speech cues. In case of poor sound quality of a hearing device, cochlear processing will be affected, resulting in a poorer SNR. However, as the SNR was comparable for all groups of patients (Fig. 3c), the sound quality of the amplified speech of either type seems to be comparable. Figure 3b presents data from the speech-in-noise test with a fixed noise level and a mild SNR (+10 dB). Here, the WRS in noise showed a pattern comparable to that of the WRS in quiet (Fig. 3a), albeit with a larger spread. Most importantly, a significant difference between SOTA and DACI devices was observed for the WRS in quiet as well as for the WRS in noise within subgroup G2, emphasizing the role of limits in MPO and gain for sufficient benefit.

Moreover, the results found in this study are in general agreement with our earlier findings comparing the WRS in quiet of DACI users with that of hearing aid users [Maier et al., 2018]. Although this earlier study was based on PTA hearing loss only, it demonstrates that DACI outcomes are significantly better than hearing aid outcomes in borderline patients with the same pronounced sensorineural hearing loss.

Bypassing the middle ear and ABG with a powerful active middle ear implant yields better results if the available conventional hearing aid technology reaches its limits, using a technology that is potentially less limited in its output to the cochlea [Zwartenkot et al., 2014]. We could demonstrate that applying common criteria for hearing aid fitting realistically predicts where a conventional approach consisting of surgery and hearing aid provision fails. Specifically, an insufficient MPO and reduced output dynamic range leads to poor outcomes. Whether the specific mode of coupling of an active middle ear implant makes a difference (coupled to the incus, stapes, or round window, or with stapes prostheses) is unknown and was not the subject of this study. However, despite significantly better average in-noise speech recognition results with the DACI, we had a few DACI patients (approximately 9%) with a measured SNR >10 dB, which may be indicative of inaudibility of the noise in a few cases [Wardenga et al., 2015] and needs to be investigated separately.

In summary, providing sufficient acoustic input to the cochlea with an implantable device that covers the audiological requirements can overcome the technical limits of the application of powerful conventional hearing aids in cases of severe mixed hearing loss. In our study, we had the advantage of having a sufficiently large group of DACI patients available to demonstrate the significance of MPO and available gain. As this specific device is not available on the market anymore, our results also point out the importance of investigating MPO and available gain if current or future devices are evaluated. Sufficiently powerful devices would provide a patient benefit beyond the possibilities of current conventional high-power hearing aids.

As some patients performed better with their own hearing aid (Fig. 2b), an additional analysis was performed using the better device. Whenever instead of the SOTA device the “best hearing aid” was used in the analysis (SOTA or WI device), the conclusions remained unchanged.

Conclusions

Conventional power hearing aids do not meet the needs for gain and maximum output of some patients with severe mixed hearing loss, resulting in non-optimal speech recognition. The limits of conventional hearing aids can be estimated by applying simple rules for required gain and output. In contrast, bypassing the middle ear and ABG with a powerful active middle ear implant yields better results. For such patients, direct acoustic stimulation of the cochlea is the more promising hearing solution; improvements in WRS of up to 30% might be expected.

Statement of Ethics

The study was performed according to the Declaration of Helsinki at the Department of Otorhinolaryngology, Hannover Medical School. All participants were informed orally and in writing about the study and gave their written informed consent. Prior to enrolment of subjects and the retrospective analysis of DACI patient data, approval from the local ethics committee (EC Hannover No. 2761-2015 and 2925-2015) had been obtained.

Disclosure Statement

This work is part of the doctoral thesis of N.W. and was supported by a project grant from Cochlear Ltd. and the DFG Cluster of Excellence EXC1077/1 “Hearing4all.” All authors received travel support from Cochlear Ltd. for meetings. B. Waldmann is an employee of Cochlear Ltd., which provided the declared support.