Extended high-frequency audiometry in research and clinical practice 4 5

*This paper is part of a special issue on Noise-Induced Hearing Disorders: Clinical and 1 Investigational Tools* 2 3 Extended high-frequency audiometry in research and clinical practice 4 5 Melanie Lough, and Christopher J. Plack 1,b 6 1 Manchester Centre for Audiology and Deafness, The University of Manchester, Oxford Road, Manchester, M13 7 9PL, United Kingdom 8 9 Uploaded to manuscript submission system 15 November 2021. Revised version uploaded to 10 manuscript submission system 10 February 2022. 11 12 Suggested running title: Extended high-frequency audiometry in practice 13 14 15 16 17 18 19 20 21 22

Audiometric testing in research and in clinical settings rarely considers frequencies above 8 kHz. 23 However, the sensitivity of young healthy ears extends to 20 kHz, and there is increasing evidence 24 that testing in the extended high-frequency (EHF) region, above 8 kHz, might provide valuable 25 additional information. Basal (EHF) cochlear regions are especially sensitive to the effects of aging, 26 disease, ototoxic drugs, and possibly noise exposure. Hence, EHF loss may be an early warning of 27 damage, useful for diagnosis and for monitoring hearing health. In certain environments, speech 28 perception may rely on EHF information, and there is evidence for an association between EHF 29 loss and speech perception difficulties, although this may not be causal: EHF loss may instead be a deviation. The purple squares and green triangles show the results for two listeners with very 69 similar thresholds up to 8 kHz, but markedly different thresholds above 8 kHz (in the EHF 70 range). Data from Carcagno and Plack (2020). 71

II. MEASURING EHF THRESHOLDS 72
A problem with measuring EHF thresholds accurately is that standing wave interference 73 patterns in the ear canal, which are particularly prominent in the EHF region, lead to frequency- middle-aged, and older listeners. Data from Carcagno and Plack (2020). 117

B. Middle ear disease, dysfunction, and surgery 118
Otitis media is a disease of the middle ear that most commonly affects children. It can be 119 infective (suppurative) or non-infective (non-suppurative), acute or chronic; however, these 120 categories are interrelated (World Health Organization, 2021b). All forms of otitis media have 121 been shown to cause EHF hearing loss that persists beyond recovery of the disease (Hunter et 122 al., 1996;Margolis et al., 2000;Ryding et al., 2002). This can occur despite negligible effects on 123 hearing thresholds between 250 Hz and 8 kHz. EHF hearing tends to be worse in people with 124 more severe disease histories, as defined by the number of acute otitis media (AOM) episodes, 125 for example (Laitila et al., 1997), but even a single episode of AOM can cause lasting damage to 126 limited number of studies report EHF audiometry findings in affected patients, and fewer 159 present data exclusively for radiotherapy (as distinct from chemoradiotherapy). However, those 160 that do indicate that radiation-induced hearing loss is more severe, and occurs sooner, at higher loss is also likely to be asymmetric (Cheraghi et al., 2015). 164

D. Noise exposure 165
Overexposure to noise can damage the hair cells in the cochlea. Noise-induced hearing loss 166 (NIHL) is traditionally associated with an audiometric "notch" between 3 and 6 kHz ( participants with normal thresholds up to 8 kHz, EHF thresholds were higher for those exposed 179 to industrial noise compared to non-exposed controls. 180 However, the findings are mixed and other studies show little relation between recreational noise 181 exposure and EHF thresholds. For example, despite reporting a relation between EHF 182 thresholds and long-term personal listening device use, Le Prell et al. (2013) found little relation 183 between EHF thresholds and noise exposure due to other activities, such as bar or club 184 attendance, or attendance at loud sporting events. Wei et al. (2017) found no associations 185 between total leisure noise exposure (including use of personal listening devices) and EHF 186 thresholds, and Mishra et al. (2021) found no relation between earphone or headphone use and 187 EHF thresholds after controlling for age. 188 A possible reason for the negative findings is the difficulty of estimating lifetime noise exposure 189 reliably (Wei et al., 2017), since the estimates are largely based on self-report and depend on what 190 events are included and how noise levels are calculated (Guest et al., 2018). It is particularly 191 important to determine if EHF threshold elevation is a useful predictor of future NIHL in the 192 standard clinical range. If so, this would make EHF thresholds a valuable tool for monitoring 193 hearing health, for example, in occupational settings, and EHF testing could be used to screen 194 for individuals at risk of losing hearing ability due to recreational activities. syndrome, an endocrine disorder that is described as a 'chronic proinflammatory state' (Kucur et 204 al., 2013). 205 In all of the aforementioned diseases, the pathogenesis of EHF hearing loss is not well 206 understood, although animal models and temporal bone studies report inner ear degeneration 207 consistent with either inflammatory or ischemic mechanisms (Ruckenstein, 2004). 208

A. Sound localization 210
EHF components provide important cues for sound localization. In particular, EHF 211 information is important for determination of sound elevation and for resolving front/back 212 confusions. Peaks and notches in the EHF spectrum are introduced by the filtering effects of the 213 pinna, and these patterns are dependent on the elevation angle of the sound source relative to 214 the listener. The patterns also vary due to individual differences in pinna morphology (Otte et al.,215 13 sounds (Brungart and Simpson, 2009) and speech sounds (Best et al., 2005). Consistent with 218 these findings, older adults with an EHF hearing loss are worse than younger adults at 219 determining sound elevation (Otte et al., 2013). 220

B. Speech perception 221
EHFs between 8 and 10 kHz improve the quality of speech (Moore and Tan improved (3.2 dB lower speech reception threshold) when the masking noise was low-pass 232 filtered at 8 kHz compared to when the noise was broadband, suggesting that cochlear regions 233 tuned above 8 kHz provide useful information. Until recently, tests of speech intelligibility in a 234 multi-talker environment have used a target talker (who the listener is required to understand) 235 and competing talkers directing speech towards the listener. This is a very unusual situation. 236 Normally, competing talkers would be facing away from the listener, and directing their speech 237 to someone else. When the competing talkers are facing away, the high frequencies from the 238 competing speech are reduced in level, because high frequencies are produced with high 239 directivity from the mouth and diffract less (Monson et al., 2012a). This means that the high 240 frequencies in the target speech may be more audible relative to the low frequencies (which may 241 be obscured by the low frequencies in the competing speech). Monson et al. (2019) found that 242 when the interfering speech was directed away from the listener, people performed better (about 243 2.5 dB improvement in signal-to-noise ratio at threshold) when frequencies above 8 kHz were 244 present than when they were removed by filtering. This implies that these EHFs were 245 contributing important information. Furthermore, Monson et al. found that EHF energy helped 246 listeners to judge the orientation of the speaker, which is important when determining who is 247 talking to you. In follow-up articles, Monson and colleagues reported that both temporal and 248 spectral information may contribute to the EHF benefit in these masking situations (Trine and 249 Monson, 2020), and that the benefit of having a masker orientation away from the listener 250 decreases for people with a threshold elevation at 16 kHz (Braza et al., 2022). 251

C. EHF loss as a marker for damage in lower frequency regions 252
In the previous section, it was noted that EHF hearing loss has been shown in some studies to 253 be related to deficits in speech-in-noise perception. However, this does not imply causation. In 254 addition to having direct effects on perception, EHF loss may be a marker for sub-clinical 255 deficits (i.e. deficits that are not revealed by standard PTA) in the standard frequency range 256 (Hunter et al., 2020). If so, then EHF audiometry might have broad diagnostic utility. 257 Over the past decade, there has been considerable interest in cochlear synaptopathy; a loss of 258 synapses between inner hair cells and auditory nerve fibers that is caused by noise exposure or 259 aging in animal models (Kujawa and Liberman, 2009), and has been inferred from nerve fiber

V. USE OF EHF AUDIOMETRY IN CLINICAL TRIALS 281
The World Health Organization (WHO) defines clinical trials as prospective, interventional 282 studies involving human participants that aim to assess the impact of the respective intervention 283 on health outcomes (World Health Organization, 2021a). Therefore, clinical trials involving 284 EHF audiometry may be ones in which: i) the intervention (as a diagnostic test) is EHF 285 audiometry; or ii) EHF audiometry is employed as an outcome measure for studies in which a 286 drug, or other treatment, is the intervention. only available for one of these. Expanding the search using broader search terms only increased 292 the yield by one after the records (and associated trial protocols) were screened for relevance, 293 completeness and accessibility. The following reasons may explain this apparent lack of 294 registered clinical trials involving EHF audiometry: 295 i. Such trials have been conducted but were not registered. 296 ii. A registered clinical trial may include EHF audiometry as a subsidiary part of the 297 protocol, and therefore this test is not listed explicitly. Or, the information on the 298 registry is not detailed enough to be able to determine whether EHF audiometry is 299 included (e.g., "audiometry" is listed but test frequencies are not specified). 300 iii. In clinical trials involving audiometry, only conventional frequencies are tested. This is 301 particularly plausible given that many clinical trials utilize the Common Terminology 302 Criteria for Adverse Events (CTCAE); EHF hearing loss does not constitute an adverse 303 event, even in the latest version (v.5) of the CTCAE (National Cancer Institute, 2017). 304 Nevertheless, examples of the use of EHF audiometry in phase I, II and III clinical trials can be 305 found. In phase I and IIa clinical trials, EHF audiometry has been utilized to evaluate the safety, 306 feasibility and potential efficacy of pharmaceutical interventions (Campbell et al., 2003;Peek et 307 al., 2020;Duinkerken et al., 2021). In an ongoing phase III trial of intensity-modulated proton 308 beam therapy versus intensity-modulated radiotherapy, the TORPEdO trial (ISRCTNregistry, 309 2020), EHF audiometry has been included in the trial protocol as a means of monitoring 310 ototoxicity. This will ultimately contribute to our knowledge about multi-toxicity reduction in 311 oropharyngeal cancer and should provide insight into whether EHF audiometry is a more 312 sensitive, or useful, measure (i.e., than conventional PTA) for detecting differences in ototoxic 313 effects between two types of radiotherapy. 314

VI. CURRENT CLINICAL USE OF EHF AUDIOMETRY 315
To determine the current clinical use of EHF audiometry across the globe, professional 316 audiology societies from 55 countries (across six continents) were emailed, and asked the 317 following two questions: that EHF audiometry is routinely performed in their respective countries. However, a caveat is 334 needed here. In the majority of these countries, EHF audiometry is only routine for certain 335 groups of patients, or within certain sectors. In New Zealand, for example, EHF audiometry is 336 performed routinely on patients receiving ototoxic medical treatment, as well as in some clinics 337 that provide tinnitus counselling and management, but it is rarely done outside of these settings. 338 Similarly, in India, it is a routine procedure in the training institutions, but the situation in the 339 private sector is unknown. only alludes to EHF audiometry once, suggesting that it may be conducted for "special 355 purposes;" further direction on dealing with the unique challenges that come with testing in the 356 EHFs, such as increased inter-subject variability, is not given. 357 Ototoxicity monitoring is an example of one such "special purpose" and this appears to be the 358 field in which EHF audiometry has gained most traction to date. In 2009, the American this information hints that when EHF audiometry is performed, it is mostly for assessing the 365 effects of ototoxic treatment. 366 Other current uses of EHF audiometry, according to our international contacts, are displayed in 367 Table I. 368 Where patients report a history of noise exposure.

India Trinidad and Tobago
In cases of asymmetric hearing, and vestibular complaints.

Romania
Performed on patients whose symptoms are suggestive of unilateral vestibulopathy. In cases of sudden hearing loss. Israel When requested by parents. Australia Requests reported to be exclusively from parents of children who are being enrolled in a Tomatis sound therapy program.

A. Diagnosis 372
The preceding section shows that some countries/services are already harnessing the diagnostic 373 advantage of EHF audiometry, but to use this test to its full potential requires a more consistent For patients with pre-existing hearing loss up to 8 kHz, a similar approach to that described 390 above may also prove fruitful. However, the clinical utility of EHF audiometry for these patients 391 will likely decline with increasing hearing loss, unless audiometer output limitations in the EHFs 392 can be overcome. 393 One (as yet) unexplored area where EHF audiometry could have future clinical utility is in the 394 early detection of vestibular schwannoma (VS), for the following reasons: 395 22 i. The most common initial presenting symptom of VS is progressive hearing loss on the 396 ipsilesional side (79.5% of VS patients) (Bento et al., 2012). 397 ii. Hearing loss is of a sloping configuration (i.e., high-frequency thresholds are worse than 398 low-frequency thresholds) in 51.7% of cases (Lee et al., 2015). 399 iii. Hearing loss associated with VS can be attributed, in part, to a gradual compression of 400 the tonotopically-formed cochlear nerve. 401 It therefore seems plausible that a certain degree of asymmetry in EHF hearing thresholds could 402 prompt a referral for Magnetic Resonance Imaging (MRI) -the gold standard for VS diagnosis. 403 However, MRI is expensive, and it would be essential, firstly, to develop means of differentiating 404 other causes of EHF asymmetry (e.g., conductive EHF hearing loss) in order to prevent 405 unwarranted medical costs or patient anxiety. 406

B. Hearing health monitoring 407
Ototoxicity monitoring programs still appear to be the result of individual service initiatives. 408 However, EHF audiometry is expected to increasingly feature as a key component of future 409 programs, for three reasons:

23
iii. The ASHA cochleotoxicity criteria for threshold change can be applied to the EHFs 416 (Campbell et al., 2003;Knight et al., 2007). Audiology, 2020). The benefit of being able to alert individuals about the onset of early hearing 431 damage is that they may be encouraged to adopt more protective behaviours, such as using ear 432 defenders. However, such subtle EHF threshold changes as those reported by Liberman et al. 433 (2016) and Maccà et al. (2015), will be difficult to detect clinically until inter-subject variability 434 can be better controlled for. The benefit may also be reduced for people over 30 years of age 435 (Maccà et al., 2015). 436 24

C. Fitting hearing aids 437
Articles that demonstrate how EHF audiometry can be utilised to fit hearing aids have largely 438 been limited to the Earlens system (Arbogast et al., 2019). The Earlens system comprises a 439 behind-the-ear sound processor, a signal delivery tip (which encodes the processed sound signal 440 into a pulsed light signal), and a custom-made lens that is positioned on the eardrum (which 441 receives the light signal and directly vibrates the eardrum). The system is marketed as having a 442 relatively wide bandwidth (125 Hz -10 kHz), an attribute that is associated with better sound 443 quality ratings by people with normal hearing and -in terms of clarity -mild-to-moderate hearing 444 loss (Füllgrabe et al., 2010), as well as by Earlens wearers comparing full-bandwidth and low-445 pass-filtered speech and music (Vaisberg et al., 2021). As such, EHF audiometry is necessary for 446 generating the prescription target to which the Earlens sound processor is set. The Earlens 447 system can currently be regarded as a niche product, although it is anticipated to become more 448 universally available over time. 449 The datasheets of many contemporary conventional hearing aids list bandwidth upper 450 frequencies of 9-10 kHz. Although these values, which have been calculated using American A lack of necessary equipment for performing EHF audiometry may be purely due to financial 476 constraints or the result of a lack of perceived need for the equipment in the first place. For 477 example, just because the preferential effects of platinum-based chemotherapy on the EHFs are 478 well known, does not mean that audiologists (or oncologists) deem EHF audiometry necessary; 479 this was corroborated by 16% of respondents to Brown et al. (2021). One reason why this view 480 may be held, is that unless hearing loss occurs in the "speech frequencies," the chemotherapy 26 regimen is unlikely to be altered; none of the four most widely used cochleotoxicity classification 482 systems specifically describe how to grade EHF hearing loss (Crundwell et al., 2016). Thus, the 483 utility of EHF audiometry in monitoring chemotherapy patients is restricted to counselling them 484 on the likelihood of, "practical speech frequency hearing loss" (Dasgupta et al., 2021). Whilst this 485 is undoubtedly realistic, the role EHF testing can play in forewarning patients should not be 486 underestimated, as counselling about the ototoxic effects in the early stages of treatment has 487 been shown to be particularly important for young cancer patients (Khan et al., 2020) (beyond Belgium and the US) that answer clinicians' specific concerns, would help to provide 496 reassurance that EHF audiometry can be performed reliably. 497

VIII. CONCLUSIONS AND GAPS IN KNOWLEDGE 498
The most basal region of the cochlea is the most vulnerable to injury, and hence hearing loss in 499 the EHF range is an important "early warning" of cochlear damage; for example, damage caused 500 by ototoxic drugs, disease, and possibly noise exposure. Furthermore, EHF loss impacts sound 501 localization, and may also have direct effects on speech perception in noisy environments. For 502 these reasons, EHF audiometry has great potential for diagnosis and hearing health monitoring, 503 and for fitting hearing aids. Currently, however, EHF audiometry has only limited application 504 internationally, and there is a lack of clinical guidelines and standards. There are also several gaps 505 in knowledge that limit the application of EHF audiometry. 506 First, it is unclear the extent to which noise exposure affects EHF thresholds before causing 507 threshold elevation at lower frequencies. Reaching a firm conclusion may depend on longitudinal 508 studies where individuals are tracked over a number of years, with more reliable estimates of 509 noise exposure dose than are currently provided by retrospective self-report. 510 The mechanism (or mechanisms) for the relation between EHF hearing loss and speech 511 perception difficulties has not yet been established clearly. In particular, it is unclear the extent to 512 which EHF hearing loss may have a direct effect on speech perception in noise, or is a marker 513 for sub-clinical deficits at lower frequencies. This is important for determining the potential 514 benefits of amplification in the EHF range, and for understanding what EHF audiometry may 515 tell us about cochlear health. 516 For the most commonly available EHF headphones, there is a lack of international standard 517 reference equivalent threshold sound pressure levels (RETSPLs), pediatric calibration correction what would warrant concern? How can conductive hearing losses be adequately detected in the 527

EHFs? 528
Although it is important that these issues are resolved, it is clear that EHF audiometry has 529 clinical utility, and the development of clinical guidelines and standards should not be delayed. 530 These should be founded on the current evidence-base, and supplemented with consensus of 531 expert opinion until such gaps in the knowledge are addressed. 532

ACKNOWLEDGMENTS 533
This work was supported by the Medical Research Council UK (MR/V01272X/1) and by the 534 NIHR Manchester Biomedical Research Centre. The authors also wish to thank the following 535 representatives of professional bodies and/or hearing health practitioners who kindly responded 536 to their request for information: @April_lyons76, @FranMarie67, @KimthK9   Where patients report a history of noise exposure.

India Trinidad and Tobago
In cases of asymmetric hearing, and vestibular complaints.

Romania
Performed on patients whose symptoms are suggestive of unilateral vestibulopathy. In cases of sudden hearing loss. Israel When requested by parents. Australia Requests reported to be exclusively from parents of children who are being enrolled in a Tomatis sound therapy program.