DPOAEs are well known for their use as a sensitive and objective measure of cochlear non-linear gain and an index of outer hair cell (OHC) electromotility and inner ear health (Kujawa et al. 1994; Kemp 1997; Hall 2000). Because noise principally damages OHCs, it is not surprising that DPOAEs have shown high sensitivity to noise injury in animal and human studies (Korres et al. 2009; de Souza Chelminski Barreto et al. 2011; Ramma et al. 2011; Meinke et al. 2013). DPOAE reduction is strongly correlated with threshold shift (Seixas et al. 2004, 2012; Sisto et al. 2007; Helleman et al. 2010; Müller et al. 2010), although there are some discrepancies across studies (Shupak et al. 2007; Helleman and Dreschler 2012). There are multiple suggestions that early deficits in DPOAE amplitudes indicate damage to the inner ear that precedes changes in conventional audiometry. DPOAEs have been described by multiple groups as providing predictive value for later pure-tone detection threshold deficits (Lapsley Miller et al. 2004, 2006; Job et al. 2009). Thus, several completed clinical trials seeking confirmatory evidence for protection of the human inner ear have already included DPOAE tests as a metric of potential protection against noise-induced cell damage. Currently, changes in DPOAE amplitude are the most common secondary outcome for noise-induced change in function in clinical trials assessing otoprotection (Attias et al. 2004; Kopke et al. 2015; Kramer et al. 2006; Le Prell et al. 2011).

Challenges in implementing the use of DPOAEs in clinical decision-making, specifically in the context of NIHL, are the topic of a recent paper by Konrad-Martin et al. (2012b). We suggest here that DPOAE-related protection would be an encouraging positive outcome; however, protection measured using DPOAE metrics should be interpreted with caution. If protection of DPOAE responses is ultimately shown to reduce later deficits during conventional pure-tone audiometry, then DPOAE data will become more useful for identifying benefits of novel drugs. Alternatively, because DPOAEs fail to identify selective damage to IHCs and the auditory nerve, the widespread use of DPOAEs may be more limited. It is interesting that Avan and Bonfils (2005) described a subgroup of workers with NIHL, defined by a change in behavioral hearing sensitivity, without compromised DPOAE functions, suggesting potential inner hair cell or neural involvement.

EHF audiometry is used to assess pure-tone thresholds at higher frequencies than are typically monitored in a hearing conservation program or during traditional audiometric monitoring; the EHF frequency range is defined as extending from 9 to 20 kHz. Elevated EHF thresholds have been observed in individuals with a history of noise exposure (Vassallo et al. 1968; Osterhammel 1979; Erickson et al. 1980; Fausti et al. 1981a, b). Because significant EHF deficits have also been reported in patients treated with ototoxic drugs such as the antineoplastic drug cisplatin (Fausti et al. 1984b; Tange et al. 1985) and aminoglycoside antibiotics (Fausti et al. 1984a), EHF testing has been proposed for use in identifying early changes in hearing following either physical trauma (i.e., noise) or pharmacological insult. Longitudinal studies assessing EHF thresholds as a function of noise are limited. In a single longitudinal study that followed a sample of 14-year-old students over a 3-year period, the largest threshold changes were observed at the two highest frequencies tested (14 and 16 kHz) with changes being approximately 5 dB (Serra et al. 2005). The observed EHF changes were attributed to concert and discothèque attendance (Biassoni et al. 2005).

In contrast to longitudinal studies, cross-sectional studies are more common. Our own data, as well as those from other labs, also suggest potential EHF loss with long-term use of personal music player devices, although the deficits tend to be limited (Meyer-Bisch 1996; Peng et al. 2007; Kim et al. 2009; Figueiredo et al. 2011; Le Prell et al. 2013), and some groups report no threshold deficits (Wong et al. 1990; Mostafapour et al. 1998; Kumar et al. 2009; Shah et al. 2009). Because none of those data are longitudinal, it remains possible that other unreported noise exposures contributed to the reported group differences. The most robust support for noise-induced deficits at EHF frequencies comes from a cross-sectional analysis of hearing thresholds in adult male factory workers, with varied noise exposure (Ahmed et al. 2001). In a group of workers with thresholds that were 20-dB HL or better from 250 Hz to 8 kHz, the subset of workers that were routinely exposed to noise had deficits at frequencies from 12 to 20 kHz, whereas workers that were not assigned to noisy areas did not have these EHF threshold deficits. Those data have been interpreted to suggest that EHF threshold deficits precede hearing loss at lower frequencies after noise, as observed following ototoxic drug therapies. Changes at EHF frequencies may ultimately prove to be a useful tool for identifying individuals with increased vulnerability to noise insult (see Osterhammel 1979) or those likely to develop NIHL at conventional test frequencies (250 Hz to 8 kHz). However, longitudinal studies, incorporating serial monitoring, are essential for determining whether workers with EHF deficits go on to develop hearing loss at lower frequencies over time.

If EHF tests provide a reliable sensitive metric for early effects of noise on the inner ear, then EHF testing might have some utility for hearing conservation monitoring. However, we are reluctant to advocate EHF testing for use as a primary outcome in clinical trials assessing otoprotective agents based on a variety of studies in which noise-induced EHF deficits have not been reliably observed. For example, when 18–21-year-old male soldiers not yet exposed to military weapons noise were compared to young soldiers seen after acute acoustic trauma, there were no threshold differences from 12.5 to 20 kHz; the deficits in the noise-exposed personnel were greatest at 4–8 kHz and extended only to 11.2 kHz (Balatsouras et al. 2005). When conventional and EHF thresholds measured in Finnish Air Force Military Personnel (19–48 years old; 50 male, 1 female) were compared to Finnish normative data, there were no hearing deficits for either conventional or EHF stimuli (Kuronen et al. 2003). Data from musicians are mixed. Minimal EHF threshold deficits (at 12.5 and 14 kHz) in conjunction with deficits at conventional test frequencies (3–8 kHz) were present in one group of musicians (Schmuziger et al. 2006), and EHF deficits developed slowly in a group of swedish pop musicians followed over time (Axelsson and Lindgren 1978; Axelsson et al. 1995), but deficits were absent in other groups (Johnson et al. 1985, 1986).

EHF deficits have not been any more reliable in temporary threshold shift (TTS) studies. Comparison of pre- and postflight hearing tests revealed small but statistically significant TTS at both conventional and EHF frequencies (approximately 1–3 dB) in military personnel, suggesting no additional benefit was obtained by supplementing a conventional hearing test paradigm with EHF testing (Kuronen et al. 2003). In a study on TTS after music player use, EHF deficits did not accompany TTS measured at lower frequencies (Le Prell et al. 2012). When TTS after music rehearsal was evaluated, TTS was detected at frequencies at and below 8 kHz but not at or above 9 kHz (Schmuziger et al. 2007). In summary, current clinical and industrial practice does not include routine monitoring for NIHL at frequencies beyond 8 kHz, and EHF testing has not reliably produced significant sensitivity above conventional testing. Thus, if otoprotective efficacy is limited to EHF frequencies, protection should be interpreted with caution until reduction in EHF deficits can be shown to predict hearing retention at conventional frequencies.

The use of word recognition, and more specifically, words or speech-in-noise tests, may be useful in otoprotection studies, as it has long been shown that the typical listener with high-frequency hearing loss, including noise-induced, has disproportionate difficulty in noise (Quist-Hanssen et al. 1979). The American Academy of Otolaryngology (AAO) has recently recommended that word recognition scores should be included in all clinical trials that assess auditory function (Gurgel et al. 2012). To the best of our knowledge, speech-based functional tasks have not been used in any of the completed clinical trials assessing otoprotective agents. Interest and enthusiasm for these tasks appear to be increasing given the repeated suggestions that loss of neural connections from inner hair cells that result in decreased ABR amplitude in the absence of overt permanent threshold shift may also result in poorer speech-in-noise discrimination (Kujawa and Liberman 2009, 2015; Lin et al. 2011; Makary et al. 2011). Given these recent recommendations, several tests that might be considered for use in clinical trials are discussed below.

The following section focuses on speech-in-noise tests rather than speech-in-quiet tests based on recommendations that speech-in-noise tests be considered a “stress test” for auditory function (Wilson 2011). As per the review by Wilson, it was Carhart (1951) who first recognized that the ability to understand speech in noise was an important measure of function as some patients reported disproportionate difficulty hearing in noise. Later, Carhart and Tillman (1970) explicitly advocated that audiologic evaluations should include a measure of the ability of patients to understand speech in competing background noise in addition to the traditional pure-tone audiogram and speech discrimination in quiet.

Following the suggestion by Carhart and Tillman (1970), new tests began to emerge, including, for example, the Speech Perception in Noise (SPIN) Test (Kalikow et al. 1977). The test items in the SPIN test are sentences read by a male talker presented with a simultaneous 12-voice multi-talker babble, and the listener’s task is to identify the last word in the sentence, which is always a monosyllabic noun. The SPIN test has a total of 250 words divided across ten lists, with each list containing 12–13 alternate types of sentences, including a low-predictability condition where the final word is difficult to predict based on sentence context and a high-predictability condition where the final key word is somewhat predictable using sentence context. The background babble is presented at the same level as the sentences in the original description (Kalikow et al. 1977). Later revisions of the test (Revised Spin Test, R-SPIN) implemented a protocol with the speech track presentation level set at 50 dB above the estimated threshold and the signal-to-babble ratio set at 8 dB (Bilger et al. 1984). The R-SPIN was composed of only 200 target words, with the target words divided into four list pairs, each consisting of two 50-sentence lists. The paired lists had the same words, presented in the opposite context conditions (high context or low context). One of the important outcomes from a study by Bilger et al. (1984) was the finding that the ten different word lists were not equivalent, with large differences in mean performance on the low-context sentences within each of the ten word lists suggesting the different lists could not be used interchangeably.

The Connected Speech Test (CST) emerged shortly after the R-SPIN (Cox et al. 1987, 1988). Intended primarily to quantify hearing aid benefit, the test was composed of sentences spoken by a female speaker presented at six signal-to-babble ratios ranging from −3 to −8 dB. The original set of 72 sentences was ultimately reduced to 48 test sentences and 9 practice sentences, with all sentences composed of basic vocabulary and syntactically simple sentences, and the scoring criteria selected such that most normal-hearing listeners achieved test scores of approximately 50 % correct (Cox et al. 1987). The lists were divided into two sets of 24 sentences to facilitate test–retest comparisons. When the test was applied to listeners with hearing loss, there were a subset of patients that had disproportionate difficulty in noise (Cox et al. 1988). Several other tests followed, such as the Speech Intelligibility Rating (SIR) test (McDaniel and Cox 1992; Beck and Speaks 1993) and the Revised Speech Intelligibility Rating (RSIR) test (Speaks et al. 1994).

The Hearing in Noise Test (HINT) was the next to emerge (Nilsson et al. 1994). During the development of the HINT, the Bamford–Kowal–Bench (BKB) sentences described by Bench et al. (1979) were modified for American English, as the British words and sentences were not commonplace for American listeners. The HINT includes a large set of sentence materials (250 sentences, spoken by a male speaker) with the sentences selected to have relatively uniform length. The sentences are divided into 25 phonemically balanced lists, with ten sentences per list, used for measuring sentence-based speech reception thresholds (sSRTs). Sentences are presented both in quiet and in spectrally matched masking noise. The HINT remains one of the more common tests used clinically and across research studies. Another relatively well-defined test is the Quick Speech-In-Noise (QuickSIN) test (Killion et al. 2004). The QuickSIN is a nonadaptive test of speech perception in four-talker babble. The test consists of sentence lists (six sentences per list) with five target words per sentence (yielding 30 target words per list). Sentences are syntactically correct but contain few semantic or contextual clues. Sentences are presented at a fixed level, and the SNR is sequentially decreased. Participants repeat the sentences and their SNR loss is calculated based on the number of target words correctly recalled. Further details are available in the QuickSIN User’s Manual. There is an anecdotal report that patients believe the QuickSIN better estimates their actual difficulty with listening in noise (Anderson et al. 2012).

The Words-in-Noise (WIN) test is perhaps now among the best developed tests; the WIN test measures word recognition performance for words presented in multi-talker babble at seven signal-to-noise ratios with the 50 % correct point (in dB SNR) used as the primary performance metric (for review of data across studies, see Wilson 2011). In brief, the WIN was initially designed as a 70-word instrument; there were ten unique words to be presented at seven signal-to-noise (S/N) ratios decreasing from 24 (easiest) to 0 (most difficult) in 4-dB decrements (Wilson et al. 2003). This 70-word list was subsequently divided into two 35-word lists with equivalent recognition performance established for the two lists (Wilson and Burks 2005). A third list (WIN List 3) has been developed for use as a practice list to familiarize participants with the task of listening to words presented in background babble (Wilson and Watts 2012). The multi-talker babble has six female voices. The WIN has been validated against the Speech Recognition in Noise Test (SPRINT) (Wilson and Cates 2008). We note here that the SPRINT was developed by and is used by the US Army to assess communication ability and fitness for duty. As described by Wilson and Cates (2008), the WIN and the SPRINT are similar in that both use Northwestern University Auditory Test No. 6 (NU-6) words. They differ in that the WIN uses a smaller number of words and a larger number of signal-to-noise ratios (S/N). Whereas the WIN has seven discrete S/N rations in 4-dB steps of increasing difficulty (Wilson and McArdle 2007; Wilson and Cates 2008), the SPRINT includes multi-talker background babble presented at a single (9 dB) S/N ratio. Directions (written by Cord et al., undated) and sound files are available online at http://​militaryaudiolog​y.​org/​site/​2009/​01/​sprint-test/​.

In summary, the early observations by Carhart (1951) that some patients have disproportionate difficulty in noise has been repeatedly confirmed, with multiple investigators reporting that participants with similar audiometric thresholds vary significantly with respect to speech recognition performance in background noise (Cox et al. 1988; Smoorenburg 1992; Vermiglio et al. 2012; Grant et al. 2013). Vermiglio et al. (2012) reported that the ability to recognize speech in steady-state noise could not be predicted from the audiogram and suggested a new classification scheme for hearing impairment requires both the audiogram and speech-in-noise thresholds. It is highly relevant that there are now proposals that the HINT be used as a diagnostic tool for employment decisions for jobs where communication performance is critical (Giguère et al. 2008). It is difficult to provide good guidance on the selection of speech-in-noise tasks for clinical trials as there have not been many studies that provide empirical data directly comparing performance of participants across speech-in-noise tests. There are some notable examples of studies that do offer back-to-back comparisons, such as Wilson et al. (2007) and Grant and Walden (2013), and based on these data, we suggest that either the WIN or the QuickSIN would appear to be reasonable choices.

Interest in ABR tests as a primary metric in human research has been increasing given that several groups have reported loss of neural connections from inner hair cells, with corresponding decrease in ABR amplitude, after noise exposures that produce a robust threshold shift (i.e., approximately 40-dB TTS, measured 24-h post-noise) (Kujawa and Liberman 2009; Lin et al. 2011; Wang and Ren 2012). Although primary spiral ganglion degeneration, in the absence of sensory cell loss, has been reported in human temporal bones, noise history was only known for three ears (Makary et al. 2011). Physiological data from humans supportive of these claims are also limited. When ABR amplitude and latency data from a large population of veterans with known noise exposure were compared to data collected in other nonveteran populations, there were no qualitative differences (Konrad-Martin et al. 2012a). The effects of noise were explored as a secondary analysis, however, with age effects being the primary variable of interest. When the potential for noise-induced ABR deficits was specifically assessed in professional pop/rock musicians, no deficits in ABR amplitude were detected (Samelli et al. 2012). However, that study only included a limited number of subjects (16 musicians and 16 controls) over a broad range of ages (21–43 years old). ABR input–output functions are a “gold standard” metric for assessing lasting effects of noise on the neural population. However, there have not been any large-scale studies that evaluate whether behaviors that result in TTS are correlated with reduced suprathreshold ABR amplitude in humans. If human ABR amplitude varies with noise history, in the absence of threshold deficits, we may ultimately gain insight into the “critical boundary” at which noise becomes hazardous to synapses in human ears, a boundary that is not currently known for either humans or laboratory animals (for discussion of critical boundary, see Le Prell et al. 2012; Spankovich et al. 2013). For the purpose of clinical trial guidance, we note here that the sensitivity of ABR recordings is improved using electrodes that utilize the ear canal as a recording site (Gaddam and Ferraro 2008); the amplitude of ABR Wave I was significantly larger and easier to identify when the ear canal was used as one of the recording sites in comparison to more conventional scalp (mastoid) recordings. Amplitude of Wave I may not be the only evoked potential of interest. Recently, Ruggles et al. (2011, 2012) identified variability in the auditory brainstem frequency-following response in normal-hearing young adults as a key factor for difficulty discriminating sounds in noise. Other groups are independently arriving at similar conclusions; Hopkins and Moore (2011) also reported that reduced sensitivity to temporal fine structure cues may underlie speech perception difficulties. Stamper and Johnson (2015) reported smaller ABR Wave I amplitudes in humans for data collected using mastoid electrodes, but not tympanic membrane electrodes, when amplitude was assessed as a function of noise history. Unfortunately the authors did not assess the potential role of gender, which is an important confound. Males typically have smaller ABR amplitudes and longer ABR latencies than females (see Hall 1992). If they also report more sources of noise exposure, the purported effect of noise may be related to gender, rather than noise history.

Patient-reported observations (PRO) are the standard for determining the incidence, loudness, and severity of subjective tinnitus. A visual analogue Tinnitus Loudness Rating Scale has been used to determine the incidence, loudness, and severity of subjective tinnitus reported by study subjects (as used in NCT00808470; Le Prell et al. 2012; Spankovich et al. 2013). At study onset, participants complete a brief questionnaire to determine the extent to which they normally experience tinnitus in either or both ears, with tinnitus operationally defined as a hissing, ringing, or other auditory sensation in the absence of an external sound. For clinical trials, we would advocate that participants indicating that tinnitus is experienced frequently (at least one incidence daily) or with significant severity (defined as severe enough to distract their attention from and/or disrupt ongoing activity) be asked to complete the Tinnitus Functional Index (TFI). The TFI is the most recent tinnitus-related screening tool; it has been responsive to treatment effects, and data suggest a 13-pt change is clinically significant (Meikle et al. 2012). Other well-validated scales include the Tinnitus Handicap Questionnaire (Kuk et al. 1990), the Tinnitus Reaction Questionnaire (Wilson et al. 1991), and the Tinnitus Severity Scale (Sweetow and Levy 1990). Each has been well-validated for use in patients who experience severe tinnitus. However, it is difficult to predict whether these questionnaires will be sensitive to subtle changes in the frequency or severity of tinnitus associated with study participation, as might be predicted with long-term daily exposure to occupational noise. Moreover, it is not clear which of these scales, if any, will be sensitive to changes in noise-induced tinnitus percepts in clinical trials. Participants reporting tinnitus at the time of audiometric testing should be asked to participate in pitch-matching and loudness-matching tasks to more precisely characterize the nature of the noise-induced tinnitus if possible.

Studies assessing otoprotective benefits in both animals and humans have primarily focused on the ability of these agents to reduce or prevent threshold shift. However, there is an emerging body of evidence that suggests that in addition to threshold shift or lack thereof, there may be important changes in hearing at suprathreshold levels. Provocative data from animal studies suggest that some noise exposures that do not result in threshold shift can show evidence of damage to synapses on auditory nerve fibers. Suprathreshold deficits including speech discrimination and speech discrimination in noise are common, but there is little agreement whether to use the WIN, QuickSIN, HINT, SPRINT, or other tests. Noise-induced ABR amplitude changes are a “hot topic” in animal models, but such changes have not been established in humans. Data collection paradigms in use in recently completed studies that are not yet described in the literature and other ongoing clinical trials on the prevention of NIHL are summarized in Table 9.2.