Evaluating the Risk of Noise-Induced Hearing Loss Using Different Noise Measurement Criteria

Evaluating the Risk of Noise-Induced Hearing Loss Using Different Noise Measurement Criteria Abstract Objectives This article examines whether the Occupational Safety and Health Administration’s (OSHA) average noise level (LAVG) or the National Institute of Occupational Safety and Health’s (NIOSH) equivalent continuous average (LEQ) noise measurement criteria better predict hearing loss. Methods A cohort of construction workers was followed for 10 years (2000–2010), during which time their noise exposures and hearing threshold levels (HTLs) were repeatedly assessed. Linear mixed models were constructed with HTLs as the outcome, either the OSHA (LAVG) or NIOSH (LEQ) measurement criteria as the measure of exposure, and controlling for age, gender, duration of participation, and baseline HTLs (as both a covariate or an additional repeated measure). Model fit was compared between models for HTLs at 0.5, 1, 2, 3, 4, 6, and 8 kHz using the Akaike information criterion (AIC). The 10th, 50th, and 90th percentiles of hearing outcomes predicted by these models were then compared with the hearing outcomes predicted using the ISO 1999:2013 model. Results The mixed models using the LEQ were found to have smaller AIC values than the corresponding LAVG models. However, only the 0.5, 3, and 4 kHz models were found to have an AIC difference greater than 2. When comparing the distribution of predicted hearing outcomes between the mixed models and their corresponding ISO outcomes, it was found that LEQ generally produced the smallest difference in predicted hearing outcomes. Conclusions Despite the small difference and high correlation between the LEQ and LAVG, the LEQ was consistently found to better predict hearing levels in this cohort and, based on this finding, is recommended for the assessment of noise exposure in populations with similar exposure characteristics. exposure assessment, hearing, noise Background It is estimated that about 24 million workers are exposed to hazardous levels of occupational noise each year in the USA alone (Tak et al., 2009). Prolonged exposure to hazardous noise can lead to noise-induced hearing loss (NIHL), which is estimated to affect 11.4% of the working population in the USA (Tak and Calvert, 2008). NIHL can diminish a worker’s ability to detect audible warnings and hinder communication with coworkers (Morata et al., 2005) and may also increase the risk of injury in the workplace (Moll van Charante and Mulder, 1990; Barreto et al., 1997; Choi et al., 2005; Picard et al., 2008; Cantley et al., 2015a, 2015b). Outside of the workplace, those with NIHL can feel socially isolated and have a higher prevalence of depression and anxiety compared to those without hearing loss (Melamed et al., 1992). Regulations and recommendations with regards to occupational noise exposure have changed since the first noise exposure limit was introduced in the 1950s (McIlwain et al., 2008). Before the founding of the Occupational Safety and Health Administration (OSHA), the Department of Labor (DOL) used its authority under the Walsh–Healey Public Contracts Act to propose a permissible exposure limit (PEL) for noise of 85 dBA with a time–intensity exchange rate (ER)—i.e. the amount of change in average noise level (LAVG) needed to double or halve the allowable exposure time—varying between 2 and 7 dB based on the intermittency of the noise exposure (Department of Labor, 1968). However, following input during the public rulemaking process, this standard was replaced by a PEL of 90 dBA with a simplified 5 dB ER, which was adopted by OSHA when that agency was established in 1971 and which remains in effect today (Department of Labor, 1971, 1981, 2004; Suter, 1992). In 1972, the National Institute of Occupational Safety and Health (NIOSH) released its initial ‘Criteria for a recommended standard for occupational exposure to noise’ in which NIOSH ‘reluctantly concurred with the generally acceptable 90 dBA exposure level for an 8-hour day’. However, NIOSH also recognized the need to reducing the 8-h exposure level to 85 dBA based on the evidence presented in the document (NIOSH, 1972). In this document, NIOSH did not take a position on the appropriate ER. In 1994, the American Conference of Governmental Industrial Hygienists (ACGIH) revised its threshold limit value for noise to be 85 dBA with a 3 dB ER (Sliney, 1993). NIOSH revisited the issue in 1998 when they released a revised ‘Criterion for a recommended standard occupational exposure to noise’ with a recommended exposure limit of 85 dBA and a 3 dB ER (NIOSH, 1998). The difference between the 5 dB and 3 dB ERs has a major impact on the allowable exposure durations for high levels of intermittent noise. For example, if the occupational exposure limit (OEL) for noise is 85 dBA as an 8-h time-weighted average with a 5 dB ER would allow for that worker to be exposed to 100 dBA of noise for 1 h, whereas the 3 dB ER would allow that worker to be exposed to 94 dBA for 1 h. A difference of 3 dB over 8 h is approximately a doubling of sound power (Earshen, 2000). For truly non-varying noise, the ER used makes no difference when comparing the measurements to the same OEL, as there is no variability in levels and hence no role for the time–intensity trading relationship summarized by the ER. However, as noise becomes more variable, as is commonly the case in many industries such as construction, the difference between the two ERs becomes increasingly important (Petrick et al., 1996; Neitzel et al., 1999). The current debate around the 5 and 3 dB ER The divergence between the OSHA regulation and NIOSH recommendation for occupational noise exposure has been a point of contention in the industrial hygiene profession (Dobie and Clark, 2013, 2015; Morata et al., 2015; Suter, 2015). However, much of the debate has focused on the differing ERs rather than the differing exposure limits. The 3 dB ER is based on the equal energy hypothesis, which states that an equal amount of sound energy will produce an equal amount of hearing damage regardless of the temporal distribution of the exposure over a work shift or longer period (Suter, 1992). This was supported mainly by the research done by Eldred et al. in 1955 and was further buttressed by Burns and Robinson in 1970 (Suter, 1992). Since then several studies have provided further support to the equal energy hypothesis, and field studies using the 3 dB ER have found NIHL rates that are similar to those documented in ISO 1990:1999 (now ISO 1999:2013) (Passchier-Vermeer, 1969; EPA, 1973, 1974; Johansson et al., 1973; ISO, 2013). Unlike the 3 dB ER, the 5 dB ER used by OSHA attempts to account for predictable, intermittent exposure to noise (e.g. noise exposures interrupted by regularly spaced quiet breaks) that may occur in the workplace. However, there is no formal definition in OSHA’s noise standard of what the distinction is between continuous and intermittent noise. The 5 dB ER was first suggested in a set of damage risk criteria curves published by the Committee on Hearing, Bioacoustics, and Biomechanics’ (CHABA) Intersociety Committee in its 1967 guidelines for controlling noise exposure (Van Atta et al., 1967). ACGIH also initially endorsed a 5 dB ER in 1969 (Suter, 1988). In the same year, the DOL adopted a regulation virtually identical to ACGIH’s standard (Suter, 1988). Despite the fact that most countries have adopted the 3 dB ER for regulatory standards and that much of the published literature supports using the 3 dB ER, some authors argue that there is insufficient evidence to support this presumably more protective ER (Clark, 1991; Ward, 1991; Suter, 2003; Dobie and Clark, 2013). The main argument put forth by those opposed to the 3 dB ER is that there are very few modern studies examining whether the 3 or 5 dB ER produces better exposure estimates for predicting NIHL, and some older studies found that using a 3 dB ER would lead to an overestimated risk of NIHL (Johansson et al., 1973). Because it is widely accepted that hazardous noise exposure leads to NIHL, it is unethical to conduct experimental human exposure studies. Animal studies, primarily of chinchillas (Martin, 2012), have found that the same amount of noise exposure produces a similar amount of NIHL regardless if the noise exposure occurs with breaks or continuously, suggesting that the equal energy hypothesis, and thus the 3 dB ER, is acceptable (Hamernik et al., 2007; Qiu et al., 2007). However, there is still considerable uncertainty when extrapolating these results to humans because of inter-species differences in NIHL risk and the use of noise exposures that are not characteristic of exposures in the workplace (Clark et al., 1987; Fredelius and Wersäll, 1992; Pourbakht and Yamasoba, 2003). Studies of highly exposed worker populations are challenging because of the need for long-term access to, and cooperation from, the workers. In addition, OSHA’s hearing conservation amendment in 1981 required employers to provide an effective hearing conservation program to all employees exposed >85 dBA as an 8-h time-weighted average (TWA) (Department of Labor, 1981). This resulted in a large increase in the use of hearing protection devices (Suter, 2009), which substantially complicates the estimation of personal noise exposures and subsequent study of NIHL risk. Annual audiometric evaluations are used to determine the degree of change in hearing over time, which may be the result of noise exposure during the interval between tests. According to both the OSHA noise standard and recommended practice, workers should receive a baseline audiogram before employment or being assigned to an area with hazardous noise. The test measures pure-tone hearing threshold levels (HTLs) at various audiometric test frequencies (0.5, 1, 2, 3, 4, 6, and sometimes 8 kHz) after a quiet period of at least 14 h (OSHA, 2010). The worker is then given a subsequent audiogram annually. Evaluation of within-worker changes in hearing thresholds between baseline and subsequent audiograms allows for surveillance and identification of NIHL. While large, longitudinal audiometric datasets are maintained by corporations and organizations in the USA and globally, these datasets are often not available to researchers, and the quality of the audiometric measurements (and supporting noise measurement data) contained in the datasets can be highly variable because of variations in testing procedures and environments, as well as supporting information collected at the time of the test (Rabinowitz et al., 2012; Mosites et al., 2016). To overcome these difficulties, we have re-analyzed exposure and audiometric data from a research cohort of construction apprentices that were first described in Seixas et al. (2004), and subsequently in 2012 (Seixas et al., 2012). This inception cohort was chosen because of reported infrequent use of hearing protection and the availability of high-quality baseline and annual audiometric test data accompanied by a robust set of longitudinal noise measurements (Neitzel et al., 2011; Seixas et al., 2012). Using linear mixed models, we estimated the amount of NIHL experienced by these workers at the 0.5, 1, 2, 3, 4, 6, and 8 kHz hearing frequencies when using the 3 dB ER as well as the 5 dB ER to estimate noise exposure. We then compared the models to see which best fit the observed changes in audiometric hearing thresholds. Predictions from both models were then compared to the International Organization for Standardization (ISO) model ISO 1999:2013 (ISO, 2013) for estimating NIHL. Methods The exposure and audiometric data for this analysis come from a 10-year longitudinal study of commercial construction apprentices from eight different trades described previously by Seixas et al. (2005b, 2012). The study was divided into two different phases. In Phase 1 (2000–2005), construction apprentices were recruited during their first year of apprenticeship training and were given baseline questionnaires and audiometric tests at 0.25, 0.5, 1, 2, 3, 4, 6, and 8 kHz using a Tremetrics RA 300 audiometer with TDH-39 headphones in a test van meeting OSHA’s requirements for background noise (Seixas et al., 2005b). Subjects were then given follow-up tests approximately every year for 4 years. Graduate students assumed to have non-harmful (i.e. <70 dBA) occupational exposures were recruited as control subjects (EPA, 1974). Subjects who had completed at least two tests were re-recruited for additional yearly audiometric tests for another 4 years during Phase 2 (2006–2010) (Seixas et al., 2012). Audiograms were obtained at 0.5, 1, 2, 3, 4, 6, and 8 kHz using a Grason–Stadler GSI-61 audiometer with ER-3A insert earphones (Eden Prairie, MN) in a test booth meeting the American National Standard Institute’s (ANSI) criteria for an audiometric test environment (Seixas et al., 2012). To account for the two different phases, a dummy variable was included in all statistical models to control for the phase of the study. Exposure to noise was assessed using a task-based approach as described by Neitzel et al. (2011). The task-based noise levels were calculated from 1310 full-shift noise measurements (with noise levels data logged at 1-min intervals and simultaneous recording of task involvement and timing by subjects) collected between 1997 and 2008 on commercial construction sites (Seixas et al., 2012). Information on task duration from the questionnaires was combined with task-specific noise levels and normalized to a 2000-h working year to account for the large variability in the number of hours worked across subjects (Seixas et al., 2012). Exposure metrics were calculated for each subject within the interval between audiometric tests and also cumulated over the subject’s full duration in the study. Equation 1 from Seixas et al. (2012) calculates the interval-specific LEQ—the equivalent continuous sound level using a 3 dB ER and no threshold—where Lt is the mean LAVG level for task t, which was done for H hours as reported by individual i in the subject-interval j lasting Y years, and LNC denotes non-construction hours in noisy jobs that were assigned a level of 85 dBA.  LEQijTB2000=10log10[12000 × Yij((∑t=1THijt10Lt/10)+(HNCij× 10LNC10))]. (1) We used Equation 2 to calculate the interval-specific task-based LAVG, which is the average sound level using a 5 dB ER and an 80 dBA threshold, normalized to a 2000-h working year.  LAVGijTB2000=16.61log10[12000 × Yij((∑t=1THijt10Lt/16.61)+(HNCij× 10LNC16.61))]. (2) Cumulative exposure was calculated for each individual total number of years between their first and last audiometric tests. The resulting cumulative exposure estimates are presented in units of dB-years (e.g. the product of an exposure intensity and duration). Controls were assigned an exposure of 70 dBA because noise exposure at this level will not cause any measurable hearing loss (EPA, 1974). Pearson’s correlation was calculated to measure the correlation between the LEQ and LAVG for each subject over each study interval and cumulatively for the study duration. The ratio of the LMAX to LEQ was calculated, using logarithmic averaging to account for the fact that decibels are log-scale measurements, to determine the ‘peakiness’ of the exposure. We previously developed and applied this ratio to assess the presence of intense exposures to noise throughout the day from personal dosimetry data using information from two available and robust exposure metrics. An alternative metric for peak exposure, the peak level (LPeak), is known to be susceptible to artifacts from contact with clothing or other objects (Neitzel et al., 1999, 2009; Seixas et al., 2005a). An indicator variable was used to identify exposures with high peakiness (LMAX/LEQ ratio >50). Linear mixed models were developed to predict HTLs in each ear over time at 0.5, 1, 2, 3, 4, 6, and 8 kHz; these are the audiometric test frequencies recommended as part of a comprehensive hearing loss prevention program (NIOSH, 1998). Additionally, linear mixed models were developed to predict the average HTLs at the 2, 3, and 4 kHz frequency, which is used by OSHA to determine whether a standard threshold shift (STS) has occurred (OSHA, 2010). Noise exposure metrics were transformed by subtracting 70, thus giving an ‘unexposed’ level of 0 dBA. Models were run using either the LAVG or LEQ exposure metric. The models were run using the combined data from Phases 1 and 2 so that our results could be compared to those of Seixas et al. (2012). The models were adjusted for the baseline covariates, age (<30 years or ≥30 years, based on the distribution of ages in the cohort) and gender. The models included random intercepts for subjects (b0i), dominant ears nested within subjects (boi•l), and a random slope for years since baseline at the subject level (b1i•l). An additional set of models was developed using the exposure metrics described previously, but which included the baseline hearing thresholds as an additional covariate. This was done to compare the model results to what was found by Seixas et al. (2012). The general equations for the linear mixed models are presented in Equation 3 where i indexes the subject i1,...,i316, l the ear (dominant or non-dominant hand side) l1,…,l617, and t indexes visit time since baseline t1,…,t9 (Seixas et al., 2012). The term Tit indexes the number of years for a subject since baseline at time t, Xit is the subject’s cumulative noise exposure at time t, and Zit∙l represents the other fixed effect covariates for ear l nested within subject i at time t. By including the number of years since baseline and the cumulative noise exposure, it was possible for the model to account for the effect of ageing in addition to noise exposure on HTLs.  Yit⋅l=β0+(b0i+b0i⋅l)+(β1+b1i⋅l)Tit+β2Xit+β3Tit+γZit⋅l+εit⋅l. (3) All models were run in STATA 14 (College Station, TX) using restricted maximum likelihood estimates and an unstructured covariance. This was done to minimize the bias in the variance component while providing the best model fit and to be consistent with the previous analysis by Seixas et al. (Corbeil and Searle, 1976; Littell et al., 2000; Seixas et al., 2012). The fit of the four LEQ and LAVG models (LEQ controlling for baseline versus baseline as an additional repeated measure and LAVG controlling for baseline versus baseline as an additional repeated measure) was compared by using the Akaike information criterion (AIC), a goodness of fit statistic that penalizes complex models (Akaike, 1974). Models with lower AIC values were deemed to better fit the data. The difference in AIC scores between the LEQ and LAVG models was calculated. A difference of 0–2 indicates that there is substantial evidence that both models fit the data, a difference of 4–7 indicates that one model fits the data considerably better, and a difference >10 indicates that one model does not fit the data (Burnham and Anderson, 2002). The 10th, 50th, and 90th percentiles of hearing outcomes from the four models were compared with the 10th, 50th, and 90th percentiles estimated levels of hearing loss associated with age and noise (HTLAN) predicted using the LEQ and LAVG exposure metrics in the model proposed in ISO 1999:2013 (ISO, 2013). Briefly, this was done by first calculating the median level of predicted noise-induced permanent threshold shift (NIPTS) at the 0.5, 1, 2, 3, 4, and 6 kHz hearing frequency for each worker using Equation 4 (from ISO 1999:2013) (ISO, 2013) for both the LEQ and LAVG where N50 is the predicted median NIPTS, μ and v represent frequency-dependent correction factors (see Supplementary Table S1, available at the Annals of Work Exposures and Health online), t represents the length of exposure, t0 represents 1 year, LEX,8h represents noise exposure for an 8-h working day (either LEQ or LAVG), and L0 represents the frequency-dependent sound level at which effect on hearing is negligible (ISO, 2013). For participants who had an exposure duration less than 10 years, N was extrapolated using Equation 5 where N50, t<10 represents the median NIPTS for exposures less than 10 years, t represents the exposure time (in years), and N50, t=10 represents the estimated NIPTS at 10 years of exposure. Assuming a Gaussian (normal) distribution, the ISO model provides multiplier values that can be used with adjustment factors to calculate the 10th and 90th percentiles of the NIPTS distribution.  N50=[μ+v ×log(tt0)]×(LEX,8h−L0)2, (4)  N50,t<10=log(t+1)log(11)×N50,t=10. (5) HTLs as a function of age are discussed in detail and can be obtained from ISO 7029:2017 (ISO, 2017). HTLs as a function of age were calculated for the same audiometric frequencies as NIPTS using Equation 6 (from ISO 1999:2013), where Hmd, y is the median hearing threshold due to age, a is the gender and frequency adjustment factor, y is the person’s age, and Hmd;18 is the median hearing threshold of an ontologically normal person, that is 18 years old. Because the equation centers the age at 18, the Hmd;18 term is taken as 0. Different percentiles can be calculated for each frequency using the provided multiplier and adjustment factors. The HTL associated with age and noise was calculated at the 10th, 50th, and 90th percentiles using Equation 7 where H′ is the hearing threshold associated with age and noise exposure, H is the hearing threshold associated with age, and N is the permanent threshold shift caused by noise exposure for the respective frequency and percentile.  Hmd,y=a(y−18)2+Hmd;18, (6)  H′=H+N− H×N120. (7) Results Figure 1 presents scatter plots of the LEQ and LAVG for each worker at each interval (Fig. 1a) at which their noise exposure was estimated, as well as for their cumulative exposures (Fig. 1b). The LEQ measurements were on average 3–4 dB higher than their associated LAVG measurements. For both interval-specific and cumulative exposures, the LEQ and LAVG measurements were highly and significantly correlated (r = 0.968 and r = 0.974, respectively). The number of subjects available at each follow-up is displayed in Table 1. Table 1. The number of subjects at each follow-up. Time point  Number of subjects  1  316  2  308  3  308  4  259  5  203  6  132  7  110  8  86  9  41  Time point  Number of subjects  1  316  2  308  3  308  4  259  5  203  6  132  7  110  8  86  9  41  View Large Figure 1. View largeDownload slide (A) Interval-specific exposures’ scatter plot and correlation of LEQ and LAVG exposures. (B) Cumulative exposures’ scatter plot and correlation of LEQ and LAVG exposures. Figure 1. View largeDownload slide (A) Interval-specific exposures’ scatter plot and correlation of LEQ and LAVG exposures. (B) Cumulative exposures’ scatter plot and correlation of LEQ and LAVG exposures. Table 2 compares the AIC values for both the LEQ and LAVG models with and without the baseline HTL covariate at the 0.5, 1, 2, 3, 4, 6, and 8 kHz audiometric test frequencies. When the baseline HTLs were included, the LEQ models fit the data better than LAVG models at each test frequency. However, only the 0.5, 3, and 4 kHz test frequencies were found to have an AIC difference >2, and only the 4-kHz frequency had a difference >4. When the baseline HTLs were not included as a covariate and instead treated as additional repeated measurements, the LEQ models better fit the data at all the test frequencies except for 2 kHz. In addition, the difference between the LEQ and LAVG models’ AIC decreased at all the test frequencies except for the 3 and 4 kHz frequencies, where the differences increased by about 1–2. The AIC for the mixed model using the LEQ for the average of the 2, 3, and 4 kHz outcome was found to be 7.42 points lower than the LAVG model (i.e. 19171.54–19178.96) suggesting that the LEQ model better fits the audiometric data at the frequencies used by OSHA to determine whether a STS has occurred. Table 2. Comparison of AIC values for the LEQ and LAVG models at 0.5, 1, 2, 3, 4, 6, and 8 kHz audiometric frequencies. Audiometric frequency (kHz)  Models with baseline HTLs  Models without baseline HTLs  LEQ  LAVG  Difference (LEQ−LAVG)  LEQ  LAVG  Difference (LEQ−LAVG)  0.5  16,858.24  16,860.76  −2.52  20,010.75  20,013.18  −2.43  1  16,410.74  16,412.56  −1.82  19,459.59  19,461.00  −1.41  2  17,098.14  17,098.55  −0.41  20,226.13  20,226.05  0.08  3  17,468.53  17,471.47  −2.94  20,872.27  20,877.10  −4.83  4  18,538.37  18,542.41  −4.04  22,383.88  22,389.32  −5.44  6  19,394.87  19,395.82  −0.95  23,475.21  23,475.85  −0.64  8  19,928.21  19,928.87  −0.66  23,756.35  23,756.39  −0.04  Audiometric frequency (kHz)  Models with baseline HTLs  Models without baseline HTLs  LEQ  LAVG  Difference (LEQ−LAVG)  LEQ  LAVG  Difference (LEQ−LAVG)  0.5  16,858.24  16,860.76  −2.52  20,010.75  20,013.18  −2.43  1  16,410.74  16,412.56  −1.82  19,459.59  19,461.00  −1.41  2  17,098.14  17,098.55  −0.41  20,226.13  20,226.05  0.08  3  17,468.53  17,471.47  −2.94  20,872.27  20,877.10  −4.83  4  18,538.37  18,542.41  −4.04  22,383.88  22,389.32  −5.44  6  19,394.87  19,395.82  −0.95  23,475.21  23,475.85  −0.64  8  19,928.21  19,928.87  −0.66  23,756.35  23,756.39  −0.04  View Large The fixed and random effects from the LEQ and LAVG models with the baseline HTLs for the 4-kHz test frequency are presented in Supplementary Table S3, available at the Annals of Work Exposures and Health online. The coefficients associated with each covariate were generally similar between the LEQ and LAVG 4-kHz models. Those workers with higher baseline hearing levels were found to suffer worse hearing loss because of noise during the study than those in the baseline group in both models. Cumulative noise exposure had a small, but significant effect on hearing levels. This trend was consistent at these three frequencies that had an AIC difference >2, except at the 0.5 kHz frequency where cumulative exposure was found to be a significant predictor of hearing loss in the LEQ model, but not in the LAVG model. Table 3 presents the fixed and random effects for the LEQ and LAVG models with the additional repeated measurements. The coefficients for each covariate were very similar except for the number of years since baseline, which was found to not be associated with changes in the HTLs in the LEQ nor the LAVG models. Table 3. Fixed and random effects for the LEQ and LAVG models without baseline HTLs covariate for the 4-kHz hearing frequency. Fixed effects  4 kHz LEQ  4 kHz LAVG  Coefficient  SE  P value  Coefficient  SE  P value  Intercept  3.21  1.70  0.06  3.23  1.70  0.058  Phase 2  2.42  0.51  <0.001  2.49  0.51  <0.001  Age (>30)  7.55  1.57  <0.001  7.50  1.57  <0.001  Gender (male)  7.41  1.82  <0.001  7.39  1.82  <0.001  Years since baseline  −0.06  0.14  0.680  0.11  0.12  0.367  Noise exposure × years  0.02  0.01  0.002  0.02  0.01  0.049  Random effects  Estimate  SE  Estimate  SE  Subject: random intercept SD  10.89  0.57  10.89  0.57  Subject: random slope SD  0.82  0.06  0.82  0.06  Subject intercept–slope correlation  0.08  0.09  0.10  0.09  Ear: random intercept SD  7.67  0.33  7.67  0.33  Residual SD  3.97  0.05  3.98  0.05  Fixed effects  4 kHz LEQ  4 kHz LAVG  Coefficient  SE  P value  Coefficient  SE  P value  Intercept  3.21  1.70  0.06  3.23  1.70  0.058  Phase 2  2.42  0.51  <0.001  2.49  0.51  <0.001  Age (>30)  7.55  1.57  <0.001  7.50  1.57  <0.001  Gender (male)  7.41  1.82  <0.001  7.39  1.82  <0.001  Years since baseline  −0.06  0.14  0.680  0.11  0.12  0.367  Noise exposure × years  0.02  0.01  0.002  0.02  0.01  0.049  Random effects  Estimate  SE  Estimate  SE  Subject: random intercept SD  10.89  0.57  10.89  0.57  Subject: random slope SD  0.82  0.06  0.82  0.06  Subject intercept–slope correlation  0.08  0.09  0.10  0.09  Ear: random intercept SD  7.67  0.33  7.67  0.33  Residual SD  3.97  0.05  3.98  0.05  SE = standard error; SD = standard deviation. View Large The comparison between the 10th, 50th, and 90th percentiles of hearing loss at the 0.5, 1, 2, 3, 4, and 6 kHz audiometric frequencies from the ISO 1999:2013 hearing loss model using both the LAVG and LEQ exposure metric to the 10th, 50th, and 90th percentiles of hearing loss at the same frequencies predicted by the mixed models with the baseline HTLs are presented in Supplementary Table S4, available at the Annals of Work Exposures and Health online. The predictions produced by the ISO model and mixed model were similar for both the LEQ and LAVG exposure metrics. However, in 14 out of the 18 comparisons (77.7%), the mixed model using the LEQ exposure metric more closely matched the estimated hearing loss that was calculated by their corresponding ISO model, suggesting that the LEQ performs slightly better than the LAVG in predicting hearing loss in this cohort (see Supplementary Figure S1, available at the Annals of Work Exposures and Health online). Table 4 presents the differences between the 10th, 50th, and 90th percentiles of hearing loss at the same frequencies between the mixed models without the additional repeated measurements and the ISO 1999:2013 hearing loss model using both the LEQ and LAVG exposure metrics. Table 4. Comparison of estimated hearing loss using the LEQ and LAVG exposure metrics in the ISO hearing loss and mixed models without the baseline HTLs covariate. Frequency (kHz)  LEQ  LAVG  Smallest difference  Model  ISO  Difference (Model−ISO)  Model  ISO  Difference (Model−ISO)  10th percentile  0.5  1.62  −5.76  7.38  1.62  −5.76  7.38  Same  1  0.63  −5.61  6.24  0.64  −5.69  6.33  LEQ  2  0.93  −6.34  7.27  0.91  −6.12  7.03  LAVG  3  0.47  −6.53  7.00  0.47  −6.6  7.07  LEQ  4  1.19  −5.88  7.07  1.19  −6.76  7.95  LEQ  6  4.67  −8.56  13.23  4.68  −8.39  13.07  LAVG  50th percentile  0.5  6.11  0.85  5.26  6.12  0.85  5.27  LEQ  1  5.39  1.69  3.70  5.4  0.97  4.43  LEQ  2  6.5  2.18  4.32  6.47  1.66  4.81  LEQ  3  7.25  4.83  2.42  7.26  2.66  4.60  LEQ  4  8.25  6.66  1.59  8.23  3.64  4.59  LEQ  6  13.33  5.99  7.34  13.34  4.12  9.22  LEQ  90th percentile  0.5  13.39  9.19  4.20  13.37  9.17  4.20  Same  1  17.36  9.47  7.89  14.01  9.37  4.64  LAVG  2  18.26  13.21  5.05  24.06  11.65  12.41  LEQ  3  22.45  13.99  8.46  22.45  12.57  9.88  LEQ  4  33.34  15.70  17.64  33.35  13.74  19.61  LEQ  6  36.12  16.42  19.70  36.11  14.59  21.52  LEQ  Frequency (kHz)  LEQ  LAVG  Smallest difference  Model  ISO  Difference (Model−ISO)  Model  ISO  Difference (Model−ISO)  10th percentile  0.5  1.62  −5.76  7.38  1.62  −5.76  7.38  Same  1  0.63  −5.61  6.24  0.64  −5.69  6.33  LEQ  2  0.93  −6.34  7.27  0.91  −6.12  7.03  LAVG  3  0.47  −6.53  7.00  0.47  −6.6  7.07  LEQ  4  1.19  −5.88  7.07  1.19  −6.76  7.95  LEQ  6  4.67  −8.56  13.23  4.68  −8.39  13.07  LAVG  50th percentile  0.5  6.11  0.85  5.26  6.12  0.85  5.27  LEQ  1  5.39  1.69  3.70  5.4  0.97  4.43  LEQ  2  6.5  2.18  4.32  6.47  1.66  4.81  LEQ  3  7.25  4.83  2.42  7.26  2.66  4.60  LEQ  4  8.25  6.66  1.59  8.23  3.64  4.59  LEQ  6  13.33  5.99  7.34  13.34  4.12  9.22  LEQ  90th percentile  0.5  13.39  9.19  4.20  13.37  9.17  4.20  Same  1  17.36  9.47  7.89  14.01  9.37  4.64  LAVG  2  18.26  13.21  5.05  24.06  11.65  12.41  LEQ  3  22.45  13.99  8.46  22.45  12.57  9.88  LEQ  4  33.34  15.70  17.64  33.35  13.74  19.61  LEQ  6  36.12  16.42  19.70  36.11  14.59  21.52  LEQ  View Large Figure 2 shows that for 13 out of 18 comparisons (72.2%), the LEQ exposure metrics more closely matched the estimated hearing loss that was calculated by the ISO model. The mixed models using the LEQ with the additional measurements were found to produce a better agreement with the ISO model than the mixed models with the baseline HTLs except for the 50th percentile of the 6-kHz test frequency and the 90th percentile of the 1- and 6-kHz test frequencies. Similarly, the mixed models using the LAVG without the additional measurements were found to produce a better agreement with the ISO model than the mixed models including the baseline HTLs expect for the 50th percentile of the 6-kHz test frequencies and the 90th percentiles at the 2- and 6-kHz test frequencies. Figure 2. View largeDownload slide Difference between model (without baseline HTL covariate) and ISO predictions of hearing loss at the 10th, 50th, and 90th percentiles for the LEQ and LAVG metrics. Figure 2. View largeDownload slide Difference between model (without baseline HTL covariate) and ISO predictions of hearing loss at the 10th, 50th, and 90th percentiles for the LEQ and LAVG metrics. Discussion The debate on whether the LEQ or LAVG exposure metric is more predictive of NIHL risk is a controversial subject, and no single study will be able to conclusively settle this debate. However, this study suggests that the LEQ is the more appropriate metric for predicting NIHL and provides a better foundation for developing exposure response relationships and providing guidance for the development of regulations and standards. One of the main strengths of our study is that it used a cohort of noise-exposed workers that were followed for approximately 10 years. This represents an exposure duration sufficient for NIHL to occur; in fact, the majority of loss expected over the course of a working lifetime in noise is predicted to occur within the first 10 years of exposure (ISO, 2013). The first set of mixed models used in this analysis did not include a covariate for baseline HTLs. Instead, the baseline HTLs were considered as additional measurements in the model. This was done because audiometric tests have inherent variability and measurement error, as demonstrated by the statistically significant effect of the study phase on HTLs because of changes in equipment and operators (Karlsmose et al., 1998). In addition, the causal relationship between noise exposure and hearing loss results in the inclusion of baseline HTLs in the model biasing the results toward the mean (Glymour et al., 2005). The additional repeated measurements in the models without the baseline adjustment still allow for us to account for an individual subject’s change in HTLs over time without biasing the relationship between the exposure and hearing outcomes. Because of this, we believe the models with the additional measurements are more appropriate than the models that control for baseline HTLs; however, we presented those models here to allow for comparison with the findings of Seixas et al. (2012). The second set of mixed models used in this analysis allowed us to control several covariates including age, hearing levels at baseline, and the number of years exposed to noise during the study, all of which can impact HTLs. When comparing the mixed model using the LEQ to the mixed model using the LAVG, we found that the LEQ model produced a lower AIC compared with the LAVG model in at all test frequencies, indicating the LEQ model had a better fit. However, the difference between the two models was generally small, and only three of seven test frequencies were found to have an AIC difference >2, i.e. a difference indicative of meaningfully different performance between the LAVG and LEQ models. It is worth noting that the 3- and 4-kHz test frequencies (along with 6 kHz) have been found to be most susceptible to NIHL (Royster et al., 2003). In both LEQ models, cumulative exposure was found to be a significant predictor of hearing loss at the 0.5-kHz frequency. This was not the case for either of the LAVG models. Occupational noise exposure has not traditionally been associated with low-frequency hearing loss (e.g. at 0.5 kHz), and this may be a spurious finding. These results may also reflect potential issues with mobile audiometric van testing and possible masking effects associated with low-frequency background noise in the test environment. When the 10th, 50th, and 90th percentiles of predicted hearing loss using the ISO model were compared to the same percentiles of predicted hearing loss from our mixed models with baseline HTLs, the LEQ models showed better agreement. However, we found that in all cases, our mixed models predicted greater NIHL than the ISO models. This is likely due to the fact that a subset of workers in this cohort had already experienced hearing loss before enrollment. These workers tended to have worse and more variable hearing outcomes compared with those who enrolled in the study with less or minimal hearing loss. The ISO model provides no way for preexisting hearing loss to be factored into the NIHL predictions based on age and known noise exposure (ISO, 2013). When we compared the 10th, 50th, and 90th percentile of predicted hearing loss from the mixed models with the additional measurements, we again found that the LEQ models showed better agreement with the ISO model than the LAVG model, but overall the models without baseline HTLs had better agreement than the models with the baseline HTLs. Recently there has been an increased interest in the impact of non-Gaussian noise—complex noise consisting of varying, intermittent, and interrupted exposures—on the risk and severity of NIHL. A recent contract report to NIOSH summarized the peer-reviewed literature and came to the conclusion that an exposure metric modified by a measure of kurtosis could provide a more accurate predictor of NIHL than simply the LEQ or LAVG alone (Suter, 2016). To evaluate this possibility, we compared the AICs of our LEQ and LAVG mixed models with an added variable for peakiness, using metrics previously developed and evaluated by Seixas et al. on the same cohort of construction workers (Seixas et al., 2005a). Following the inclusion of the peakiness metric, the LEQ model still demonstrated generally lower AIC values compared with its equivalent LAVG model, but the difference between AICs was reduced to <2 for all models (see Supplementary Table S2, available at the Annals of Work Exposures and Health online). The LEQ was still a better fit in our model, but our finding that the inclusion of a measure of peakiness and resulting improvement in model fit suggests that a combination of the LEQ and some sort of measure of kurtosis may further improve the model. Further research is needed to investigate the impact of including a measure of kurtosis on NIHL predictions. Our study only examined the effects of noise exposures in construction workers, who are exposed to intermittent noise, so these results may not be generalizable to occupational groups that are exposed to truly continuous noise or who have regular, scheduled breaks from exposure. There is limited evidence to support the notion that most occupations have such breaks, consistent with the rationale behind the LAVG (Neitzel et al., 1999; Suter, 2002). The high correlation between the LEQ and LAVG exposure measurements made on construction workers resulted in similar levels of model fit and predicted hearing outcomes. This was further complicated by the fact that many of the construction workers evaluated here had preexisting hearing loss. One set of mixed models controlled for this situation through the use of a categorical variable for baseline hearing level. The other set of mixed models instead used the baseline HTLs as additional measurements and excluded the fixed effect for baseline HTLs. Regardless, it is not possible to account for baseline hearing levels in the ISO model (ISO, 2013). This is likely the reason that our mixed models consistently predicted higher hearing thresholds than the ISO model and highlights an important weakness in the ISO model. Despite these drawbacks, we believe that the results of this study add to the growing body of evidence supporting the use of the LEQ over the LAVG. Supplementary Data Supplementary data are available at Annals of Work Exposures and Health online. Conflict of Interest Declaration The authors have no conflict of interest to disclose. However, in the spirit of full disclosure, the authors would like to make the editor aware that one of the authors (B.R.) is currently employed at Cardno ChemRisk, which is a scientific consulting firm involved with litigation. The research for this article was completed when the author was a doctoral student at the University of Michigan, and his current employment has no impact on this article. References Akaike H. ( 1974) A new look at the statistical model identification. IEEE Trans Autom Control ; 19: 716– 23. Google Scholar CrossRef Search ADS   Barreto SM, Swerdlow AJ, Smith PGet al.   ( 1997) Risk of death from motor-vehicle injury in Brazilian steelworkers: a nested case-control study. Int J Epidemiol ; 26: 814– 21. Google Scholar CrossRef Search ADS PubMed  Burnham KP, Anderson DR. ( 2002) Model selection and multimodel inference: a practical information-theoretic approach . Vol. 172. 2nd edn. New York: Springer-Verlag. Cantley LF, Galusha D, Cullen MRet al.   ( 2015a) Association between ambient noise exposure, hearing acuity, and risk of acute occupational injury. Scand J Work Environ Health ; 41: 75– 83. Google Scholar CrossRef Search ADS   Cantley LF, Galusha D, Cullen MRet al.   ( 2015b) Does tinnitus, hearing asymmetry, or hearing loss predispose to occupational injury risk? Int J Audiol ; 54 (Suppl 1): S30– 6. Google Scholar CrossRef Search ADS   Choi SW, Peek-Asa C, Sprince NLet al.   ( 2005) Hearing loss as a risk factor for agricultural injuries. Am J Ind Med ; 48: 293– 301. Google Scholar CrossRef Search ADS PubMed  Clark WW. ( 1991) Recent studies of temporary threshold shift (TTS) and permanent threshold shift (PTS) in animals. J Acoust Soc Am ; 90: 155– 63. Google Scholar CrossRef Search ADS PubMed  Clark WW, Bohne BA, Boettcher FA. ( 1987) Effect of periodic rest on hearing loss and cochlear damage following exposure to noise. J Acoust Soc Am ; 82: 1253– 64. Google Scholar CrossRef Search ADS PubMed  Corbeil RR, Searle SR. ( 1976) Restricted maximum likelihood (REML) estimation of variance components in the mixed model. Technometrics ; 18: 31. Google Scholar CrossRef Search ADS   Department of Labor. ( 1968) Safety and health standards for federal supply contracts . Vol. 33. Washington, DC. pp. 1258– 14271. Department of Labor. ( 1971) Guidelines to The Department of Labor’s occupational noise standards: bulletin 334 . Washington, DC. Department of Labor. ( 1981) Occupational noise exposure; proposed hearing conservation amendment . Washington, DC. pp. 42622– 39. Department of Labor. ( 2004) Occupational safety and health act of 1970 . United States. Dobie RA, Clark WW. ( 2013) Exchange rates for intermittent and fluctuating occupational noise: a systematic review of studies of human permanent threshold shift. Ear Hear ; 35: 86– 96. Google Scholar CrossRef Search ADS   Dobie RA, Clark WW. ( 2015) Response to Suter and NIOSH. Ear Hear ; 36: 492– 5. Google Scholar CrossRef Search ADS PubMed  Earshen JJ. ( 2000) Sound measurement: instrumentation and noise descriptors. In Berger EH, Royster LH, Royster JD, Driscoll DP, Layne M, editors. The noise manual . 5th ed. Fairfax, VA: American Industrial Hygiene Association. pp. 41– 101. EPA. ( 1973) Public health and welfare criteria for noise . Washington, DC. EPA. ( 1974) Information on levels of environmental noise requisite to protect public health and welfare with an adequate margin of safety . Washington, DC. pp. 1– 242. Fredelius L, Wersäll J. ( 1992) Hair cell damage after continuous and interrupted pure tone overstimulation: a scanning electron microscopic study in the guinea pig. Hear Res ; 62: 194– 8. Google Scholar CrossRef Search ADS PubMed  Glymour MM, Weuve J, Berkman LFet al.   ( 2005) When is baseline adjustment useful in analyses of change? An example with education and cognitive change. Am J Epidemiol ; 162: 267– 78. Google Scholar CrossRef Search ADS PubMed  Hamernik RP, Qiu W, Davis B. ( 2007) Hearing loss from interrupted, intermittent, and time varying non-Gaussian noise exposure: The applicability of the equal energy hypothesis. J Acoust Soc Am ; 122: 2245– 54. Google Scholar CrossRef Search ADS PubMed  ISO. ( 2013) Acoustics—estimation of noise-induced hearing loss . Geneva, Switzerland: International Organization for Standardization. https://www.iso.org/standard/45103.html. Accessed May 2017. ISO. ( 2017) Acoustics—statistical distribution of hearing thresholds related to age and gender . International Organization for Standardization. Geneva, Switzerland. https://www.iso.org/standard/42916.html. Accessed May 2017. Johansson B, Kylin B, Reopstorff S. ( 1973) Proceedings of the 10th International Congress on Noise as a Public Health Problem (ICBEN). In W. Dixon Ward, editor. Evaluation of the hearing damage risk from intermittent noise according to the ISO recommendations . Dubrovink, Yugoslavia: U.S. Envrionmental Protection Agency. pp. 201– 10. Karlsmose B, Pedersen HB, Lauritzen Tet al.   ( 1998) Audiometry in general practice: validation of a pragmatic pure-tone audiometry method. Scand Audiol ; 27: 137– 42. Google Scholar CrossRef Search ADS PubMed  Littell RC, Pendergast J, Natarajan R. ( 2000) Modelling covariance structure in the analysis of repeated measures data. Stat Med ; 19: 1793– 819. Google Scholar CrossRef Search ADS PubMed  Martin L. ( 2012) Chinchillas as experimental models. In Suckow MA, Stevens KA, Wilson RP, editors. The laboratory rabbit, guinea pig, hamster, and other rodents (Internet) . 1st edn. Elsevier Inc. pp. 1009– 28. Available from http://dx.doi.org/10.1016/B978-0-12-380920-9.00043-2. Google Scholar CrossRef Search ADS   McIlwain DS, Gates K, Ciliax D. ( 2008) Heritage of army audiology and the road ahead: the Army Hearing Program. Am J Public Health ; 98: 2167– 72. Google Scholar CrossRef Search ADS PubMed  Melamed S, Luz J, Green MS. ( 1992) Noise exposure, noise annoyance and their relation to psychological distress, accident and sickness absence among blue-collar workers–the Cordis Study. Isr J Med Sci ; 28: 629– 35. Google Scholar PubMed  Moll van Charante AW, Mulder PG. ( 1990) Perceptual acuity and the risk of industrial accidents. Am J Epidemiol ; 131: 652– 63. Google Scholar CrossRef Search ADS PubMed  Morata TC, Themann CL, Byrne DCet al.   ( 2015) Letter to the editor : Scientific rigor required for a re-examination of exchange rate for occupational noise measurements Re : Dobie, RA, Clark, WW. (2014) Exchange rates for Intermittent and Fluctuating Occupational Noise : a systematic review. Ear Hear  36(4): 488–491. Morata TC, Themann CL, Randolph RFet al.   ( 2005) Working in noise with a hearing loss: perceptions from workers, supervisors, and hearing conservation program managers. Ear Hear ; 26: 529– 45. Google Scholar CrossRef Search ADS PubMed  Mosites E, Neitzel R, Galusha Det al.  ( 2016) A comparison of an audiometric screening survey with an in-depth research questionnaire for hearing loss and hearing loss risk factors. Int J Audiol ; 55: 782– 6. Google Scholar CrossRef Search ADS PubMed  Neitzel R, Daniell W, Sheppard Let al.  ( 2009) Comparison of perceived and quantitative measures of occupational noise exposure. Ann Occup Hyg ; 53: 41– 54. Google Scholar PubMed  Neitzel R, Seixas NS, Camp Jet al.  ( 1999) An assessment of occupational noise exposures in four construction trades. Am Ind Hyg Assoc J ; 60: 807– 17. Google Scholar CrossRef Search ADS PubMed  Neitzel R, Seixas NS, Camp Jet al.   ( 1999) An assessment of occupational noise exposures in four construction trades. Am Ind Hyg Assoc J ; 60: 807– 17. Google Scholar CrossRef Search ADS PubMed  Neitzel RL, Stover B, Seixas NS. ( 2011) Longitudinal assessment of noise exposure in a cohort of construction workers. Ann Occup Hyg ; 55: 906– 16. Google Scholar PubMed  NIOSH. ( 1972) Criteria for a recommended standard occupational exposure to noise . Cincinnati, OH: National Institutes of Occupational Safety and Health. NIOSH. ( 1998) Criteria for a recommended standard occupational noise exposure revised criteria 1998 . Cincinnati, OH: National Institutes of Occupational Safety and Health. OSHA. ( 2010) Occupational noise exposure . 1910.95. Washington, DC: Occupational Safety and Health Administration. pp. 211– 24. PubMed PubMed  Passchier-Vermeer W. ( 1969) Hearing loss due to exposure to steady-state broadband noise . Delft, The Netherlands: Research Institute for Public Health Engineering. Petrick ME, Royster LH, Royster JDet al.  ( 1996) Comparison of daily noise exposures in one workplace based on noise criteria recommended by ACGIH and OSHA. Am Ind Hyg Assoc J ; 57: 924– 8. Google Scholar CrossRef Search ADS PubMed  Picard M, Girard SA, Simard Met al.   ( 2008) Association of work-related accidents with noise exposure in the workplace and noise-induced hearing loss based on the experience of some 240,000 person-years of observation. Accid Anal Prev ; 40: 1644– 52. Google Scholar CrossRef Search ADS PubMed  Pourbakht A, Yamasoba T. ( 2003) Cochlear damage caused by continuous and intermittent noise exposure. Hear Res ; 178: 70– 8. Google Scholar CrossRef Search ADS PubMed  Qiu W, Davis B, Hamernik RP. ( 2007) Hearing loss from interrupted, intermittent, and time varying Gaussian noise exposures: the applicability of the equal energy hypothesis. J Acoust Soc Am ; 121: 1613– 20. Google Scholar CrossRef Search ADS PubMed  Rabinowitz PM, Galusha D, McTague MFet al.   ( 2012) Tracking occupational hearing loss across global industries: a comparative analysis of metrics. Noise Health ; 14: 21– 7. Google Scholar CrossRef Search ADS PubMed  Royster L, Royster J, Dobie RA. ( 2003) Prediction and analysis of the hearing characteristics of noise-exposed populations or individuals. In Berger EH, editor. The noise manual . 5th edn. Fairfax, VA: AIHA Press; 2003. pp. 669– 88. Google Scholar CrossRef Search ADS   Seixas N, Neitzel R, Sheppard Let al.   ( 2005a) Alternative metrics for noise exposure among construction workers. Ann Occup Hyg ; 49: 493– 502. Google Scholar CrossRef Search ADS   Seixas NS, Goldman B, Sheppard Let al.   ( 2005b) Prospective noise induced changes to hearing among construction industry apprentices. Occup Environ Med ; 62: 309– 17. Google Scholar CrossRef Search ADS   Seixas NS, Kujawa SG, Norton Set al.   ( 2004) Predictors of hearing threshold levels and distortion product otoacoustic emissions among noise exposed young adults. Occup Environ Med ; 61: 899– 907. Google Scholar CrossRef Search ADS PubMed  Seixas NS, Neitzel R, Stover Bet al.   ( 2012) 10-Year prospective study of noise exposure and hearing damage among construction workers. Occup Environ Med ; 69: 643– 50. Google Scholar CrossRef Search ADS PubMed  Sliney D. ( 1993) Review of the threshold limit value for noise. Appl Occup Environ Hyg ; 8: 618– 23. Google Scholar CrossRef Search ADS   Suter A. ( 1992) The relationship of the exchange rate to noise-induced hearing loss . Cincinnati, OH: Institute of Noise Control Engineering. Suter A. ( 2003) Standards and regulations. In Berger EH, Royster LH, Royster JD, Driscoll DP, Layne M, editors. The noise manual . 5th edn. Fairfax, VA: American Industrial Hygiene Association. Suter A. ( 2017) Occupational Hearing Loss from Non-Gaussian Noise . Semin Hear; 38: 225–62. doi:10.1055/s-0037-1603726. Suter AH. ( 1988) The development of federal noise standards and damage risk criteria. In: Hearing conservation in industry, schools, and the military . Boston, MA: Pearson Allyn & Bacon. pp. 45– 66. Suter AH. ( 2002) Construction noise: exposure, effects, and the potential for remediation; a review and analysis. AIHA J (Fairfax, Va) ; 63: 768– 89. Google Scholar CrossRef Search ADS PubMed  Suter AH. ( 2009) The hearing conservation amendment: 25 years later. Noise Health ; 11: 2– 7. Google Scholar CrossRef Search ADS PubMed  Suter AH. ( 2015) Letter to the editor : an alternative interpretation of issues surrounding the exchange rates. Ear and Hearing . Jul-Aug;36(4):485–487 Tak S, Calvert GM. ( 2008) Hearing difficulty attributable to employment by industry and occupation: an analysis of the National Health Interview Survey–United States, 1997 to 2003. J Occup Environ Med ; 50: 46– 56. Google Scholar CrossRef Search ADS PubMed  Tak S, Davis RR, Calvert GM. ( 2009) Exposure to hazardous workplace noise and use of hearing protection devices among US workers–NHANES, 1999-2004. Am J Ind Med ; 52: 358– 71. Google Scholar CrossRef Search ADS PubMed  Van Atta F, Cohen A, Glorig Aet al.   ( 1967) Guidelines for noise exposure control. Am Ind Hyg Assocation J .; 28: 418– 24. Google Scholar CrossRef Search ADS   Ward WD. ( 1991) The role of intermittence in PTS. J Acoust Soc Am ; 90: 164– 9. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of the British Occupational Hygiene Society. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Annals of Work Exposures and Health (formerly Annals Of Occupational Hygiene) Oxford University Press

Evaluating the Risk of Noise-Induced Hearing Loss Using Different Noise Measurement Criteria

Loading next page...
 
/lp/ou_press/evaluating-the-risk-of-noise-induced-hearing-loss-using-different-OyAtLBOEKs
Publisher
British Occupational Hygiene Society
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the British Occupational Hygiene Society.
ISSN
2398-7308
eISSN
2398-7316
D.O.I.
10.1093/annweh/wxy001
Publisher site
See Article on Publisher Site

Abstract

Abstract Objectives This article examines whether the Occupational Safety and Health Administration’s (OSHA) average noise level (LAVG) or the National Institute of Occupational Safety and Health’s (NIOSH) equivalent continuous average (LEQ) noise measurement criteria better predict hearing loss. Methods A cohort of construction workers was followed for 10 years (2000–2010), during which time their noise exposures and hearing threshold levels (HTLs) were repeatedly assessed. Linear mixed models were constructed with HTLs as the outcome, either the OSHA (LAVG) or NIOSH (LEQ) measurement criteria as the measure of exposure, and controlling for age, gender, duration of participation, and baseline HTLs (as both a covariate or an additional repeated measure). Model fit was compared between models for HTLs at 0.5, 1, 2, 3, 4, 6, and 8 kHz using the Akaike information criterion (AIC). The 10th, 50th, and 90th percentiles of hearing outcomes predicted by these models were then compared with the hearing outcomes predicted using the ISO 1999:2013 model. Results The mixed models using the LEQ were found to have smaller AIC values than the corresponding LAVG models. However, only the 0.5, 3, and 4 kHz models were found to have an AIC difference greater than 2. When comparing the distribution of predicted hearing outcomes between the mixed models and their corresponding ISO outcomes, it was found that LEQ generally produced the smallest difference in predicted hearing outcomes. Conclusions Despite the small difference and high correlation between the LEQ and LAVG, the LEQ was consistently found to better predict hearing levels in this cohort and, based on this finding, is recommended for the assessment of noise exposure in populations with similar exposure characteristics. exposure assessment, hearing, noise Background It is estimated that about 24 million workers are exposed to hazardous levels of occupational noise each year in the USA alone (Tak et al., 2009). Prolonged exposure to hazardous noise can lead to noise-induced hearing loss (NIHL), which is estimated to affect 11.4% of the working population in the USA (Tak and Calvert, 2008). NIHL can diminish a worker’s ability to detect audible warnings and hinder communication with coworkers (Morata et al., 2005) and may also increase the risk of injury in the workplace (Moll van Charante and Mulder, 1990; Barreto et al., 1997; Choi et al., 2005; Picard et al., 2008; Cantley et al., 2015a, 2015b). Outside of the workplace, those with NIHL can feel socially isolated and have a higher prevalence of depression and anxiety compared to those without hearing loss (Melamed et al., 1992). Regulations and recommendations with regards to occupational noise exposure have changed since the first noise exposure limit was introduced in the 1950s (McIlwain et al., 2008). Before the founding of the Occupational Safety and Health Administration (OSHA), the Department of Labor (DOL) used its authority under the Walsh–Healey Public Contracts Act to propose a permissible exposure limit (PEL) for noise of 85 dBA with a time–intensity exchange rate (ER)—i.e. the amount of change in average noise level (LAVG) needed to double or halve the allowable exposure time—varying between 2 and 7 dB based on the intermittency of the noise exposure (Department of Labor, 1968). However, following input during the public rulemaking process, this standard was replaced by a PEL of 90 dBA with a simplified 5 dB ER, which was adopted by OSHA when that agency was established in 1971 and which remains in effect today (Department of Labor, 1971, 1981, 2004; Suter, 1992). In 1972, the National Institute of Occupational Safety and Health (NIOSH) released its initial ‘Criteria for a recommended standard for occupational exposure to noise’ in which NIOSH ‘reluctantly concurred with the generally acceptable 90 dBA exposure level for an 8-hour day’. However, NIOSH also recognized the need to reducing the 8-h exposure level to 85 dBA based on the evidence presented in the document (NIOSH, 1972). In this document, NIOSH did not take a position on the appropriate ER. In 1994, the American Conference of Governmental Industrial Hygienists (ACGIH) revised its threshold limit value for noise to be 85 dBA with a 3 dB ER (Sliney, 1993). NIOSH revisited the issue in 1998 when they released a revised ‘Criterion for a recommended standard occupational exposure to noise’ with a recommended exposure limit of 85 dBA and a 3 dB ER (NIOSH, 1998). The difference between the 5 dB and 3 dB ERs has a major impact on the allowable exposure durations for high levels of intermittent noise. For example, if the occupational exposure limit (OEL) for noise is 85 dBA as an 8-h time-weighted average with a 5 dB ER would allow for that worker to be exposed to 100 dBA of noise for 1 h, whereas the 3 dB ER would allow that worker to be exposed to 94 dBA for 1 h. A difference of 3 dB over 8 h is approximately a doubling of sound power (Earshen, 2000). For truly non-varying noise, the ER used makes no difference when comparing the measurements to the same OEL, as there is no variability in levels and hence no role for the time–intensity trading relationship summarized by the ER. However, as noise becomes more variable, as is commonly the case in many industries such as construction, the difference between the two ERs becomes increasingly important (Petrick et al., 1996; Neitzel et al., 1999). The current debate around the 5 and 3 dB ER The divergence between the OSHA regulation and NIOSH recommendation for occupational noise exposure has been a point of contention in the industrial hygiene profession (Dobie and Clark, 2013, 2015; Morata et al., 2015; Suter, 2015). However, much of the debate has focused on the differing ERs rather than the differing exposure limits. The 3 dB ER is based on the equal energy hypothesis, which states that an equal amount of sound energy will produce an equal amount of hearing damage regardless of the temporal distribution of the exposure over a work shift or longer period (Suter, 1992). This was supported mainly by the research done by Eldred et al. in 1955 and was further buttressed by Burns and Robinson in 1970 (Suter, 1992). Since then several studies have provided further support to the equal energy hypothesis, and field studies using the 3 dB ER have found NIHL rates that are similar to those documented in ISO 1990:1999 (now ISO 1999:2013) (Passchier-Vermeer, 1969; EPA, 1973, 1974; Johansson et al., 1973; ISO, 2013). Unlike the 3 dB ER, the 5 dB ER used by OSHA attempts to account for predictable, intermittent exposure to noise (e.g. noise exposures interrupted by regularly spaced quiet breaks) that may occur in the workplace. However, there is no formal definition in OSHA’s noise standard of what the distinction is between continuous and intermittent noise. The 5 dB ER was first suggested in a set of damage risk criteria curves published by the Committee on Hearing, Bioacoustics, and Biomechanics’ (CHABA) Intersociety Committee in its 1967 guidelines for controlling noise exposure (Van Atta et al., 1967). ACGIH also initially endorsed a 5 dB ER in 1969 (Suter, 1988). In the same year, the DOL adopted a regulation virtually identical to ACGIH’s standard (Suter, 1988). Despite the fact that most countries have adopted the 3 dB ER for regulatory standards and that much of the published literature supports using the 3 dB ER, some authors argue that there is insufficient evidence to support this presumably more protective ER (Clark, 1991; Ward, 1991; Suter, 2003; Dobie and Clark, 2013). The main argument put forth by those opposed to the 3 dB ER is that there are very few modern studies examining whether the 3 or 5 dB ER produces better exposure estimates for predicting NIHL, and some older studies found that using a 3 dB ER would lead to an overestimated risk of NIHL (Johansson et al., 1973). Because it is widely accepted that hazardous noise exposure leads to NIHL, it is unethical to conduct experimental human exposure studies. Animal studies, primarily of chinchillas (Martin, 2012), have found that the same amount of noise exposure produces a similar amount of NIHL regardless if the noise exposure occurs with breaks or continuously, suggesting that the equal energy hypothesis, and thus the 3 dB ER, is acceptable (Hamernik et al., 2007; Qiu et al., 2007). However, there is still considerable uncertainty when extrapolating these results to humans because of inter-species differences in NIHL risk and the use of noise exposures that are not characteristic of exposures in the workplace (Clark et al., 1987; Fredelius and Wersäll, 1992; Pourbakht and Yamasoba, 2003). Studies of highly exposed worker populations are challenging because of the need for long-term access to, and cooperation from, the workers. In addition, OSHA’s hearing conservation amendment in 1981 required employers to provide an effective hearing conservation program to all employees exposed >85 dBA as an 8-h time-weighted average (TWA) (Department of Labor, 1981). This resulted in a large increase in the use of hearing protection devices (Suter, 2009), which substantially complicates the estimation of personal noise exposures and subsequent study of NIHL risk. Annual audiometric evaluations are used to determine the degree of change in hearing over time, which may be the result of noise exposure during the interval between tests. According to both the OSHA noise standard and recommended practice, workers should receive a baseline audiogram before employment or being assigned to an area with hazardous noise. The test measures pure-tone hearing threshold levels (HTLs) at various audiometric test frequencies (0.5, 1, 2, 3, 4, 6, and sometimes 8 kHz) after a quiet period of at least 14 h (OSHA, 2010). The worker is then given a subsequent audiogram annually. Evaluation of within-worker changes in hearing thresholds between baseline and subsequent audiograms allows for surveillance and identification of NIHL. While large, longitudinal audiometric datasets are maintained by corporations and organizations in the USA and globally, these datasets are often not available to researchers, and the quality of the audiometric measurements (and supporting noise measurement data) contained in the datasets can be highly variable because of variations in testing procedures and environments, as well as supporting information collected at the time of the test (Rabinowitz et al., 2012; Mosites et al., 2016). To overcome these difficulties, we have re-analyzed exposure and audiometric data from a research cohort of construction apprentices that were first described in Seixas et al. (2004), and subsequently in 2012 (Seixas et al., 2012). This inception cohort was chosen because of reported infrequent use of hearing protection and the availability of high-quality baseline and annual audiometric test data accompanied by a robust set of longitudinal noise measurements (Neitzel et al., 2011; Seixas et al., 2012). Using linear mixed models, we estimated the amount of NIHL experienced by these workers at the 0.5, 1, 2, 3, 4, 6, and 8 kHz hearing frequencies when using the 3 dB ER as well as the 5 dB ER to estimate noise exposure. We then compared the models to see which best fit the observed changes in audiometric hearing thresholds. Predictions from both models were then compared to the International Organization for Standardization (ISO) model ISO 1999:2013 (ISO, 2013) for estimating NIHL. Methods The exposure and audiometric data for this analysis come from a 10-year longitudinal study of commercial construction apprentices from eight different trades described previously by Seixas et al. (2005b, 2012). The study was divided into two different phases. In Phase 1 (2000–2005), construction apprentices were recruited during their first year of apprenticeship training and were given baseline questionnaires and audiometric tests at 0.25, 0.5, 1, 2, 3, 4, 6, and 8 kHz using a Tremetrics RA 300 audiometer with TDH-39 headphones in a test van meeting OSHA’s requirements for background noise (Seixas et al., 2005b). Subjects were then given follow-up tests approximately every year for 4 years. Graduate students assumed to have non-harmful (i.e. <70 dBA) occupational exposures were recruited as control subjects (EPA, 1974). Subjects who had completed at least two tests were re-recruited for additional yearly audiometric tests for another 4 years during Phase 2 (2006–2010) (Seixas et al., 2012). Audiograms were obtained at 0.5, 1, 2, 3, 4, 6, and 8 kHz using a Grason–Stadler GSI-61 audiometer with ER-3A insert earphones (Eden Prairie, MN) in a test booth meeting the American National Standard Institute’s (ANSI) criteria for an audiometric test environment (Seixas et al., 2012). To account for the two different phases, a dummy variable was included in all statistical models to control for the phase of the study. Exposure to noise was assessed using a task-based approach as described by Neitzel et al. (2011). The task-based noise levels were calculated from 1310 full-shift noise measurements (with noise levels data logged at 1-min intervals and simultaneous recording of task involvement and timing by subjects) collected between 1997 and 2008 on commercial construction sites (Seixas et al., 2012). Information on task duration from the questionnaires was combined with task-specific noise levels and normalized to a 2000-h working year to account for the large variability in the number of hours worked across subjects (Seixas et al., 2012). Exposure metrics were calculated for each subject within the interval between audiometric tests and also cumulated over the subject’s full duration in the study. Equation 1 from Seixas et al. (2012) calculates the interval-specific LEQ—the equivalent continuous sound level using a 3 dB ER and no threshold—where Lt is the mean LAVG level for task t, which was done for H hours as reported by individual i in the subject-interval j lasting Y years, and LNC denotes non-construction hours in noisy jobs that were assigned a level of 85 dBA.  LEQijTB2000=10log10[12000 × Yij((∑t=1THijt10Lt/10)+(HNCij× 10LNC10))]. (1) We used Equation 2 to calculate the interval-specific task-based LAVG, which is the average sound level using a 5 dB ER and an 80 dBA threshold, normalized to a 2000-h working year.  LAVGijTB2000=16.61log10[12000 × Yij((∑t=1THijt10Lt/16.61)+(HNCij× 10LNC16.61))]. (2) Cumulative exposure was calculated for each individual total number of years between their first and last audiometric tests. The resulting cumulative exposure estimates are presented in units of dB-years (e.g. the product of an exposure intensity and duration). Controls were assigned an exposure of 70 dBA because noise exposure at this level will not cause any measurable hearing loss (EPA, 1974). Pearson’s correlation was calculated to measure the correlation between the LEQ and LAVG for each subject over each study interval and cumulatively for the study duration. The ratio of the LMAX to LEQ was calculated, using logarithmic averaging to account for the fact that decibels are log-scale measurements, to determine the ‘peakiness’ of the exposure. We previously developed and applied this ratio to assess the presence of intense exposures to noise throughout the day from personal dosimetry data using information from two available and robust exposure metrics. An alternative metric for peak exposure, the peak level (LPeak), is known to be susceptible to artifacts from contact with clothing or other objects (Neitzel et al., 1999, 2009; Seixas et al., 2005a). An indicator variable was used to identify exposures with high peakiness (LMAX/LEQ ratio >50). Linear mixed models were developed to predict HTLs in each ear over time at 0.5, 1, 2, 3, 4, 6, and 8 kHz; these are the audiometric test frequencies recommended as part of a comprehensive hearing loss prevention program (NIOSH, 1998). Additionally, linear mixed models were developed to predict the average HTLs at the 2, 3, and 4 kHz frequency, which is used by OSHA to determine whether a standard threshold shift (STS) has occurred (OSHA, 2010). Noise exposure metrics were transformed by subtracting 70, thus giving an ‘unexposed’ level of 0 dBA. Models were run using either the LAVG or LEQ exposure metric. The models were run using the combined data from Phases 1 and 2 so that our results could be compared to those of Seixas et al. (2012). The models were adjusted for the baseline covariates, age (<30 years or ≥30 years, based on the distribution of ages in the cohort) and gender. The models included random intercepts for subjects (b0i), dominant ears nested within subjects (boi•l), and a random slope for years since baseline at the subject level (b1i•l). An additional set of models was developed using the exposure metrics described previously, but which included the baseline hearing thresholds as an additional covariate. This was done to compare the model results to what was found by Seixas et al. (2012). The general equations for the linear mixed models are presented in Equation 3 where i indexes the subject i1,...,i316, l the ear (dominant or non-dominant hand side) l1,…,l617, and t indexes visit time since baseline t1,…,t9 (Seixas et al., 2012). The term Tit indexes the number of years for a subject since baseline at time t, Xit is the subject’s cumulative noise exposure at time t, and Zit∙l represents the other fixed effect covariates for ear l nested within subject i at time t. By including the number of years since baseline and the cumulative noise exposure, it was possible for the model to account for the effect of ageing in addition to noise exposure on HTLs.  Yit⋅l=β0+(b0i+b0i⋅l)+(β1+b1i⋅l)Tit+β2Xit+β3Tit+γZit⋅l+εit⋅l. (3) All models were run in STATA 14 (College Station, TX) using restricted maximum likelihood estimates and an unstructured covariance. This was done to minimize the bias in the variance component while providing the best model fit and to be consistent with the previous analysis by Seixas et al. (Corbeil and Searle, 1976; Littell et al., 2000; Seixas et al., 2012). The fit of the four LEQ and LAVG models (LEQ controlling for baseline versus baseline as an additional repeated measure and LAVG controlling for baseline versus baseline as an additional repeated measure) was compared by using the Akaike information criterion (AIC), a goodness of fit statistic that penalizes complex models (Akaike, 1974). Models with lower AIC values were deemed to better fit the data. The difference in AIC scores between the LEQ and LAVG models was calculated. A difference of 0–2 indicates that there is substantial evidence that both models fit the data, a difference of 4–7 indicates that one model fits the data considerably better, and a difference >10 indicates that one model does not fit the data (Burnham and Anderson, 2002). The 10th, 50th, and 90th percentiles of hearing outcomes from the four models were compared with the 10th, 50th, and 90th percentiles estimated levels of hearing loss associated with age and noise (HTLAN) predicted using the LEQ and LAVG exposure metrics in the model proposed in ISO 1999:2013 (ISO, 2013). Briefly, this was done by first calculating the median level of predicted noise-induced permanent threshold shift (NIPTS) at the 0.5, 1, 2, 3, 4, and 6 kHz hearing frequency for each worker using Equation 4 (from ISO 1999:2013) (ISO, 2013) for both the LEQ and LAVG where N50 is the predicted median NIPTS, μ and v represent frequency-dependent correction factors (see Supplementary Table S1, available at the Annals of Work Exposures and Health online), t represents the length of exposure, t0 represents 1 year, LEX,8h represents noise exposure for an 8-h working day (either LEQ or LAVG), and L0 represents the frequency-dependent sound level at which effect on hearing is negligible (ISO, 2013). For participants who had an exposure duration less than 10 years, N was extrapolated using Equation 5 where N50, t<10 represents the median NIPTS for exposures less than 10 years, t represents the exposure time (in years), and N50, t=10 represents the estimated NIPTS at 10 years of exposure. Assuming a Gaussian (normal) distribution, the ISO model provides multiplier values that can be used with adjustment factors to calculate the 10th and 90th percentiles of the NIPTS distribution.  N50=[μ+v ×log(tt0)]×(LEX,8h−L0)2, (4)  N50,t<10=log(t+1)log(11)×N50,t=10. (5) HTLs as a function of age are discussed in detail and can be obtained from ISO 7029:2017 (ISO, 2017). HTLs as a function of age were calculated for the same audiometric frequencies as NIPTS using Equation 6 (from ISO 1999:2013), where Hmd, y is the median hearing threshold due to age, a is the gender and frequency adjustment factor, y is the person’s age, and Hmd;18 is the median hearing threshold of an ontologically normal person, that is 18 years old. Because the equation centers the age at 18, the Hmd;18 term is taken as 0. Different percentiles can be calculated for each frequency using the provided multiplier and adjustment factors. The HTL associated with age and noise was calculated at the 10th, 50th, and 90th percentiles using Equation 7 where H′ is the hearing threshold associated with age and noise exposure, H is the hearing threshold associated with age, and N is the permanent threshold shift caused by noise exposure for the respective frequency and percentile.  Hmd,y=a(y−18)2+Hmd;18, (6)  H′=H+N− H×N120. (7) Results Figure 1 presents scatter plots of the LEQ and LAVG for each worker at each interval (Fig. 1a) at which their noise exposure was estimated, as well as for their cumulative exposures (Fig. 1b). The LEQ measurements were on average 3–4 dB higher than their associated LAVG measurements. For both interval-specific and cumulative exposures, the LEQ and LAVG measurements were highly and significantly correlated (r = 0.968 and r = 0.974, respectively). The number of subjects available at each follow-up is displayed in Table 1. Table 1. The number of subjects at each follow-up. Time point  Number of subjects  1  316  2  308  3  308  4  259  5  203  6  132  7  110  8  86  9  41  Time point  Number of subjects  1  316  2  308  3  308  4  259  5  203  6  132  7  110  8  86  9  41  View Large Figure 1. View largeDownload slide (A) Interval-specific exposures’ scatter plot and correlation of LEQ and LAVG exposures. (B) Cumulative exposures’ scatter plot and correlation of LEQ and LAVG exposures. Figure 1. View largeDownload slide (A) Interval-specific exposures’ scatter plot and correlation of LEQ and LAVG exposures. (B) Cumulative exposures’ scatter plot and correlation of LEQ and LAVG exposures. Table 2 compares the AIC values for both the LEQ and LAVG models with and without the baseline HTL covariate at the 0.5, 1, 2, 3, 4, 6, and 8 kHz audiometric test frequencies. When the baseline HTLs were included, the LEQ models fit the data better than LAVG models at each test frequency. However, only the 0.5, 3, and 4 kHz test frequencies were found to have an AIC difference >2, and only the 4-kHz frequency had a difference >4. When the baseline HTLs were not included as a covariate and instead treated as additional repeated measurements, the LEQ models better fit the data at all the test frequencies except for 2 kHz. In addition, the difference between the LEQ and LAVG models’ AIC decreased at all the test frequencies except for the 3 and 4 kHz frequencies, where the differences increased by about 1–2. The AIC for the mixed model using the LEQ for the average of the 2, 3, and 4 kHz outcome was found to be 7.42 points lower than the LAVG model (i.e. 19171.54–19178.96) suggesting that the LEQ model better fits the audiometric data at the frequencies used by OSHA to determine whether a STS has occurred. Table 2. Comparison of AIC values for the LEQ and LAVG models at 0.5, 1, 2, 3, 4, 6, and 8 kHz audiometric frequencies. Audiometric frequency (kHz)  Models with baseline HTLs  Models without baseline HTLs  LEQ  LAVG  Difference (LEQ−LAVG)  LEQ  LAVG  Difference (LEQ−LAVG)  0.5  16,858.24  16,860.76  −2.52  20,010.75  20,013.18  −2.43  1  16,410.74  16,412.56  −1.82  19,459.59  19,461.00  −1.41  2  17,098.14  17,098.55  −0.41  20,226.13  20,226.05  0.08  3  17,468.53  17,471.47  −2.94  20,872.27  20,877.10  −4.83  4  18,538.37  18,542.41  −4.04  22,383.88  22,389.32  −5.44  6  19,394.87  19,395.82  −0.95  23,475.21  23,475.85  −0.64  8  19,928.21  19,928.87  −0.66  23,756.35  23,756.39  −0.04  Audiometric frequency (kHz)  Models with baseline HTLs  Models without baseline HTLs  LEQ  LAVG  Difference (LEQ−LAVG)  LEQ  LAVG  Difference (LEQ−LAVG)  0.5  16,858.24  16,860.76  −2.52  20,010.75  20,013.18  −2.43  1  16,410.74  16,412.56  −1.82  19,459.59  19,461.00  −1.41  2  17,098.14  17,098.55  −0.41  20,226.13  20,226.05  0.08  3  17,468.53  17,471.47  −2.94  20,872.27  20,877.10  −4.83  4  18,538.37  18,542.41  −4.04  22,383.88  22,389.32  −5.44  6  19,394.87  19,395.82  −0.95  23,475.21  23,475.85  −0.64  8  19,928.21  19,928.87  −0.66  23,756.35  23,756.39  −0.04  View Large The fixed and random effects from the LEQ and LAVG models with the baseline HTLs for the 4-kHz test frequency are presented in Supplementary Table S3, available at the Annals of Work Exposures and Health online. The coefficients associated with each covariate were generally similar between the LEQ and LAVG 4-kHz models. Those workers with higher baseline hearing levels were found to suffer worse hearing loss because of noise during the study than those in the baseline group in both models. Cumulative noise exposure had a small, but significant effect on hearing levels. This trend was consistent at these three frequencies that had an AIC difference >2, except at the 0.5 kHz frequency where cumulative exposure was found to be a significant predictor of hearing loss in the LEQ model, but not in the LAVG model. Table 3 presents the fixed and random effects for the LEQ and LAVG models with the additional repeated measurements. The coefficients for each covariate were very similar except for the number of years since baseline, which was found to not be associated with changes in the HTLs in the LEQ nor the LAVG models. Table 3. Fixed and random effects for the LEQ and LAVG models without baseline HTLs covariate for the 4-kHz hearing frequency. Fixed effects  4 kHz LEQ  4 kHz LAVG  Coefficient  SE  P value  Coefficient  SE  P value  Intercept  3.21  1.70  0.06  3.23  1.70  0.058  Phase 2  2.42  0.51  <0.001  2.49  0.51  <0.001  Age (>30)  7.55  1.57  <0.001  7.50  1.57  <0.001  Gender (male)  7.41  1.82  <0.001  7.39  1.82  <0.001  Years since baseline  −0.06  0.14  0.680  0.11  0.12  0.367  Noise exposure × years  0.02  0.01  0.002  0.02  0.01  0.049  Random effects  Estimate  SE  Estimate  SE  Subject: random intercept SD  10.89  0.57  10.89  0.57  Subject: random slope SD  0.82  0.06  0.82  0.06  Subject intercept–slope correlation  0.08  0.09  0.10  0.09  Ear: random intercept SD  7.67  0.33  7.67  0.33  Residual SD  3.97  0.05  3.98  0.05  Fixed effects  4 kHz LEQ  4 kHz LAVG  Coefficient  SE  P value  Coefficient  SE  P value  Intercept  3.21  1.70  0.06  3.23  1.70  0.058  Phase 2  2.42  0.51  <0.001  2.49  0.51  <0.001  Age (>30)  7.55  1.57  <0.001  7.50  1.57  <0.001  Gender (male)  7.41  1.82  <0.001  7.39  1.82  <0.001  Years since baseline  −0.06  0.14  0.680  0.11  0.12  0.367  Noise exposure × years  0.02  0.01  0.002  0.02  0.01  0.049  Random effects  Estimate  SE  Estimate  SE  Subject: random intercept SD  10.89  0.57  10.89  0.57  Subject: random slope SD  0.82  0.06  0.82  0.06  Subject intercept–slope correlation  0.08  0.09  0.10  0.09  Ear: random intercept SD  7.67  0.33  7.67  0.33  Residual SD  3.97  0.05  3.98  0.05  SE = standard error; SD = standard deviation. View Large The comparison between the 10th, 50th, and 90th percentiles of hearing loss at the 0.5, 1, 2, 3, 4, and 6 kHz audiometric frequencies from the ISO 1999:2013 hearing loss model using both the LAVG and LEQ exposure metric to the 10th, 50th, and 90th percentiles of hearing loss at the same frequencies predicted by the mixed models with the baseline HTLs are presented in Supplementary Table S4, available at the Annals of Work Exposures and Health online. The predictions produced by the ISO model and mixed model were similar for both the LEQ and LAVG exposure metrics. However, in 14 out of the 18 comparisons (77.7%), the mixed model using the LEQ exposure metric more closely matched the estimated hearing loss that was calculated by their corresponding ISO model, suggesting that the LEQ performs slightly better than the LAVG in predicting hearing loss in this cohort (see Supplementary Figure S1, available at the Annals of Work Exposures and Health online). Table 4 presents the differences between the 10th, 50th, and 90th percentiles of hearing loss at the same frequencies between the mixed models without the additional repeated measurements and the ISO 1999:2013 hearing loss model using both the LEQ and LAVG exposure metrics. Table 4. Comparison of estimated hearing loss using the LEQ and LAVG exposure metrics in the ISO hearing loss and mixed models without the baseline HTLs covariate. Frequency (kHz)  LEQ  LAVG  Smallest difference  Model  ISO  Difference (Model−ISO)  Model  ISO  Difference (Model−ISO)  10th percentile  0.5  1.62  −5.76  7.38  1.62  −5.76  7.38  Same  1  0.63  −5.61  6.24  0.64  −5.69  6.33  LEQ  2  0.93  −6.34  7.27  0.91  −6.12  7.03  LAVG  3  0.47  −6.53  7.00  0.47  −6.6  7.07  LEQ  4  1.19  −5.88  7.07  1.19  −6.76  7.95  LEQ  6  4.67  −8.56  13.23  4.68  −8.39  13.07  LAVG  50th percentile  0.5  6.11  0.85  5.26  6.12  0.85  5.27  LEQ  1  5.39  1.69  3.70  5.4  0.97  4.43  LEQ  2  6.5  2.18  4.32  6.47  1.66  4.81  LEQ  3  7.25  4.83  2.42  7.26  2.66  4.60  LEQ  4  8.25  6.66  1.59  8.23  3.64  4.59  LEQ  6  13.33  5.99  7.34  13.34  4.12  9.22  LEQ  90th percentile  0.5  13.39  9.19  4.20  13.37  9.17  4.20  Same  1  17.36  9.47  7.89  14.01  9.37  4.64  LAVG  2  18.26  13.21  5.05  24.06  11.65  12.41  LEQ  3  22.45  13.99  8.46  22.45  12.57  9.88  LEQ  4  33.34  15.70  17.64  33.35  13.74  19.61  LEQ  6  36.12  16.42  19.70  36.11  14.59  21.52  LEQ  Frequency (kHz)  LEQ  LAVG  Smallest difference  Model  ISO  Difference (Model−ISO)  Model  ISO  Difference (Model−ISO)  10th percentile  0.5  1.62  −5.76  7.38  1.62  −5.76  7.38  Same  1  0.63  −5.61  6.24  0.64  −5.69  6.33  LEQ  2  0.93  −6.34  7.27  0.91  −6.12  7.03  LAVG  3  0.47  −6.53  7.00  0.47  −6.6  7.07  LEQ  4  1.19  −5.88  7.07  1.19  −6.76  7.95  LEQ  6  4.67  −8.56  13.23  4.68  −8.39  13.07  LAVG  50th percentile  0.5  6.11  0.85  5.26  6.12  0.85  5.27  LEQ  1  5.39  1.69  3.70  5.4  0.97  4.43  LEQ  2  6.5  2.18  4.32  6.47  1.66  4.81  LEQ  3  7.25  4.83  2.42  7.26  2.66  4.60  LEQ  4  8.25  6.66  1.59  8.23  3.64  4.59  LEQ  6  13.33  5.99  7.34  13.34  4.12  9.22  LEQ  90th percentile  0.5  13.39  9.19  4.20  13.37  9.17  4.20  Same  1  17.36  9.47  7.89  14.01  9.37  4.64  LAVG  2  18.26  13.21  5.05  24.06  11.65  12.41  LEQ  3  22.45  13.99  8.46  22.45  12.57  9.88  LEQ  4  33.34  15.70  17.64  33.35  13.74  19.61  LEQ  6  36.12  16.42  19.70  36.11  14.59  21.52  LEQ  View Large Figure 2 shows that for 13 out of 18 comparisons (72.2%), the LEQ exposure metrics more closely matched the estimated hearing loss that was calculated by the ISO model. The mixed models using the LEQ with the additional measurements were found to produce a better agreement with the ISO model than the mixed models with the baseline HTLs except for the 50th percentile of the 6-kHz test frequency and the 90th percentile of the 1- and 6-kHz test frequencies. Similarly, the mixed models using the LAVG without the additional measurements were found to produce a better agreement with the ISO model than the mixed models including the baseline HTLs expect for the 50th percentile of the 6-kHz test frequencies and the 90th percentiles at the 2- and 6-kHz test frequencies. Figure 2. View largeDownload slide Difference between model (without baseline HTL covariate) and ISO predictions of hearing loss at the 10th, 50th, and 90th percentiles for the LEQ and LAVG metrics. Figure 2. View largeDownload slide Difference between model (without baseline HTL covariate) and ISO predictions of hearing loss at the 10th, 50th, and 90th percentiles for the LEQ and LAVG metrics. Discussion The debate on whether the LEQ or LAVG exposure metric is more predictive of NIHL risk is a controversial subject, and no single study will be able to conclusively settle this debate. However, this study suggests that the LEQ is the more appropriate metric for predicting NIHL and provides a better foundation for developing exposure response relationships and providing guidance for the development of regulations and standards. One of the main strengths of our study is that it used a cohort of noise-exposed workers that were followed for approximately 10 years. This represents an exposure duration sufficient for NIHL to occur; in fact, the majority of loss expected over the course of a working lifetime in noise is predicted to occur within the first 10 years of exposure (ISO, 2013). The first set of mixed models used in this analysis did not include a covariate for baseline HTLs. Instead, the baseline HTLs were considered as additional measurements in the model. This was done because audiometric tests have inherent variability and measurement error, as demonstrated by the statistically significant effect of the study phase on HTLs because of changes in equipment and operators (Karlsmose et al., 1998). In addition, the causal relationship between noise exposure and hearing loss results in the inclusion of baseline HTLs in the model biasing the results toward the mean (Glymour et al., 2005). The additional repeated measurements in the models without the baseline adjustment still allow for us to account for an individual subject’s change in HTLs over time without biasing the relationship between the exposure and hearing outcomes. Because of this, we believe the models with the additional measurements are more appropriate than the models that control for baseline HTLs; however, we presented those models here to allow for comparison with the findings of Seixas et al. (2012). The second set of mixed models used in this analysis allowed us to control several covariates including age, hearing levels at baseline, and the number of years exposed to noise during the study, all of which can impact HTLs. When comparing the mixed model using the LEQ to the mixed model using the LAVG, we found that the LEQ model produced a lower AIC compared with the LAVG model in at all test frequencies, indicating the LEQ model had a better fit. However, the difference between the two models was generally small, and only three of seven test frequencies were found to have an AIC difference >2, i.e. a difference indicative of meaningfully different performance between the LAVG and LEQ models. It is worth noting that the 3- and 4-kHz test frequencies (along with 6 kHz) have been found to be most susceptible to NIHL (Royster et al., 2003). In both LEQ models, cumulative exposure was found to be a significant predictor of hearing loss at the 0.5-kHz frequency. This was not the case for either of the LAVG models. Occupational noise exposure has not traditionally been associated with low-frequency hearing loss (e.g. at 0.5 kHz), and this may be a spurious finding. These results may also reflect potential issues with mobile audiometric van testing and possible masking effects associated with low-frequency background noise in the test environment. When the 10th, 50th, and 90th percentiles of predicted hearing loss using the ISO model were compared to the same percentiles of predicted hearing loss from our mixed models with baseline HTLs, the LEQ models showed better agreement. However, we found that in all cases, our mixed models predicted greater NIHL than the ISO models. This is likely due to the fact that a subset of workers in this cohort had already experienced hearing loss before enrollment. These workers tended to have worse and more variable hearing outcomes compared with those who enrolled in the study with less or minimal hearing loss. The ISO model provides no way for preexisting hearing loss to be factored into the NIHL predictions based on age and known noise exposure (ISO, 2013). When we compared the 10th, 50th, and 90th percentile of predicted hearing loss from the mixed models with the additional measurements, we again found that the LEQ models showed better agreement with the ISO model than the LAVG model, but overall the models without baseline HTLs had better agreement than the models with the baseline HTLs. Recently there has been an increased interest in the impact of non-Gaussian noise—complex noise consisting of varying, intermittent, and interrupted exposures—on the risk and severity of NIHL. A recent contract report to NIOSH summarized the peer-reviewed literature and came to the conclusion that an exposure metric modified by a measure of kurtosis could provide a more accurate predictor of NIHL than simply the LEQ or LAVG alone (Suter, 2016). To evaluate this possibility, we compared the AICs of our LEQ and LAVG mixed models with an added variable for peakiness, using metrics previously developed and evaluated by Seixas et al. on the same cohort of construction workers (Seixas et al., 2005a). Following the inclusion of the peakiness metric, the LEQ model still demonstrated generally lower AIC values compared with its equivalent LAVG model, but the difference between AICs was reduced to <2 for all models (see Supplementary Table S2, available at the Annals of Work Exposures and Health online). The LEQ was still a better fit in our model, but our finding that the inclusion of a measure of peakiness and resulting improvement in model fit suggests that a combination of the LEQ and some sort of measure of kurtosis may further improve the model. Further research is needed to investigate the impact of including a measure of kurtosis on NIHL predictions. Our study only examined the effects of noise exposures in construction workers, who are exposed to intermittent noise, so these results may not be generalizable to occupational groups that are exposed to truly continuous noise or who have regular, scheduled breaks from exposure. There is limited evidence to support the notion that most occupations have such breaks, consistent with the rationale behind the LAVG (Neitzel et al., 1999; Suter, 2002). The high correlation between the LEQ and LAVG exposure measurements made on construction workers resulted in similar levels of model fit and predicted hearing outcomes. This was further complicated by the fact that many of the construction workers evaluated here had preexisting hearing loss. One set of mixed models controlled for this situation through the use of a categorical variable for baseline hearing level. The other set of mixed models instead used the baseline HTLs as additional measurements and excluded the fixed effect for baseline HTLs. Regardless, it is not possible to account for baseline hearing levels in the ISO model (ISO, 2013). This is likely the reason that our mixed models consistently predicted higher hearing thresholds than the ISO model and highlights an important weakness in the ISO model. Despite these drawbacks, we believe that the results of this study add to the growing body of evidence supporting the use of the LEQ over the LAVG. Supplementary Data Supplementary data are available at Annals of Work Exposures and Health online. Conflict of Interest Declaration The authors have no conflict of interest to disclose. However, in the spirit of full disclosure, the authors would like to make the editor aware that one of the authors (B.R.) is currently employed at Cardno ChemRisk, which is a scientific consulting firm involved with litigation. The research for this article was completed when the author was a doctoral student at the University of Michigan, and his current employment has no impact on this article. References Akaike H. ( 1974) A new look at the statistical model identification. IEEE Trans Autom Control ; 19: 716– 23. Google Scholar CrossRef Search ADS   Barreto SM, Swerdlow AJ, Smith PGet al.   ( 1997) Risk of death from motor-vehicle injury in Brazilian steelworkers: a nested case-control study. Int J Epidemiol ; 26: 814– 21. Google Scholar CrossRef Search ADS PubMed  Burnham KP, Anderson DR. ( 2002) Model selection and multimodel inference: a practical information-theoretic approach . Vol. 172. 2nd edn. New York: Springer-Verlag. Cantley LF, Galusha D, Cullen MRet al.   ( 2015a) Association between ambient noise exposure, hearing acuity, and risk of acute occupational injury. Scand J Work Environ Health ; 41: 75– 83. Google Scholar CrossRef Search ADS   Cantley LF, Galusha D, Cullen MRet al.   ( 2015b) Does tinnitus, hearing asymmetry, or hearing loss predispose to occupational injury risk? Int J Audiol ; 54 (Suppl 1): S30– 6. Google Scholar CrossRef Search ADS   Choi SW, Peek-Asa C, Sprince NLet al.   ( 2005) Hearing loss as a risk factor for agricultural injuries. Am J Ind Med ; 48: 293– 301. Google Scholar CrossRef Search ADS PubMed  Clark WW. ( 1991) Recent studies of temporary threshold shift (TTS) and permanent threshold shift (PTS) in animals. J Acoust Soc Am ; 90: 155– 63. Google Scholar CrossRef Search ADS PubMed  Clark WW, Bohne BA, Boettcher FA. ( 1987) Effect of periodic rest on hearing loss and cochlear damage following exposure to noise. J Acoust Soc Am ; 82: 1253– 64. Google Scholar CrossRef Search ADS PubMed  Corbeil RR, Searle SR. ( 1976) Restricted maximum likelihood (REML) estimation of variance components in the mixed model. Technometrics ; 18: 31. Google Scholar CrossRef Search ADS   Department of Labor. ( 1968) Safety and health standards for federal supply contracts . Vol. 33. Washington, DC. pp. 1258– 14271. Department of Labor. ( 1971) Guidelines to The Department of Labor’s occupational noise standards: bulletin 334 . Washington, DC. Department of Labor. ( 1981) Occupational noise exposure; proposed hearing conservation amendment . Washington, DC. pp. 42622– 39. Department of Labor. ( 2004) Occupational safety and health act of 1970 . United States. Dobie RA, Clark WW. ( 2013) Exchange rates for intermittent and fluctuating occupational noise: a systematic review of studies of human permanent threshold shift. Ear Hear ; 35: 86– 96. Google Scholar CrossRef Search ADS   Dobie RA, Clark WW. ( 2015) Response to Suter and NIOSH. Ear Hear ; 36: 492– 5. Google Scholar CrossRef Search ADS PubMed  Earshen JJ. ( 2000) Sound measurement: instrumentation and noise descriptors. In Berger EH, Royster LH, Royster JD, Driscoll DP, Layne M, editors. The noise manual . 5th ed. Fairfax, VA: American Industrial Hygiene Association. pp. 41– 101. EPA. ( 1973) Public health and welfare criteria for noise . Washington, DC. EPA. ( 1974) Information on levels of environmental noise requisite to protect public health and welfare with an adequate margin of safety . Washington, DC. pp. 1– 242. Fredelius L, Wersäll J. ( 1992) Hair cell damage after continuous and interrupted pure tone overstimulation: a scanning electron microscopic study in the guinea pig. Hear Res ; 62: 194– 8. Google Scholar CrossRef Search ADS PubMed  Glymour MM, Weuve J, Berkman LFet al.   ( 2005) When is baseline adjustment useful in analyses of change? An example with education and cognitive change. Am J Epidemiol ; 162: 267– 78. Google Scholar CrossRef Search ADS PubMed  Hamernik RP, Qiu W, Davis B. ( 2007) Hearing loss from interrupted, intermittent, and time varying non-Gaussian noise exposure: The applicability of the equal energy hypothesis. J Acoust Soc Am ; 122: 2245– 54. Google Scholar CrossRef Search ADS PubMed  ISO. ( 2013) Acoustics—estimation of noise-induced hearing loss . Geneva, Switzerland: International Organization for Standardization. https://www.iso.org/standard/45103.html. Accessed May 2017. ISO. ( 2017) Acoustics—statistical distribution of hearing thresholds related to age and gender . International Organization for Standardization. Geneva, Switzerland. https://www.iso.org/standard/42916.html. Accessed May 2017. Johansson B, Kylin B, Reopstorff S. ( 1973) Proceedings of the 10th International Congress on Noise as a Public Health Problem (ICBEN). In W. Dixon Ward, editor. Evaluation of the hearing damage risk from intermittent noise according to the ISO recommendations . Dubrovink, Yugoslavia: U.S. Envrionmental Protection Agency. pp. 201– 10. Karlsmose B, Pedersen HB, Lauritzen Tet al.   ( 1998) Audiometry in general practice: validation of a pragmatic pure-tone audiometry method. Scand Audiol ; 27: 137– 42. Google Scholar CrossRef Search ADS PubMed  Littell RC, Pendergast J, Natarajan R. ( 2000) Modelling covariance structure in the analysis of repeated measures data. Stat Med ; 19: 1793– 819. Google Scholar CrossRef Search ADS PubMed  Martin L. ( 2012) Chinchillas as experimental models. In Suckow MA, Stevens KA, Wilson RP, editors. The laboratory rabbit, guinea pig, hamster, and other rodents (Internet) . 1st edn. Elsevier Inc. pp. 1009– 28. Available from http://dx.doi.org/10.1016/B978-0-12-380920-9.00043-2. Google Scholar CrossRef Search ADS   McIlwain DS, Gates K, Ciliax D. ( 2008) Heritage of army audiology and the road ahead: the Army Hearing Program. Am J Public Health ; 98: 2167– 72. Google Scholar CrossRef Search ADS PubMed  Melamed S, Luz J, Green MS. ( 1992) Noise exposure, noise annoyance and their relation to psychological distress, accident and sickness absence among blue-collar workers–the Cordis Study. Isr J Med Sci ; 28: 629– 35. Google Scholar PubMed  Moll van Charante AW, Mulder PG. ( 1990) Perceptual acuity and the risk of industrial accidents. Am J Epidemiol ; 131: 652– 63. Google Scholar CrossRef Search ADS PubMed  Morata TC, Themann CL, Byrne DCet al.   ( 2015) Letter to the editor : Scientific rigor required for a re-examination of exchange rate for occupational noise measurements Re : Dobie, RA, Clark, WW. (2014) Exchange rates for Intermittent and Fluctuating Occupational Noise : a systematic review. Ear Hear  36(4): 488–491. Morata TC, Themann CL, Randolph RFet al.   ( 2005) Working in noise with a hearing loss: perceptions from workers, supervisors, and hearing conservation program managers. Ear Hear ; 26: 529– 45. Google Scholar CrossRef Search ADS PubMed  Mosites E, Neitzel R, Galusha Det al.  ( 2016) A comparison of an audiometric screening survey with an in-depth research questionnaire for hearing loss and hearing loss risk factors. Int J Audiol ; 55: 782– 6. Google Scholar CrossRef Search ADS PubMed  Neitzel R, Daniell W, Sheppard Let al.  ( 2009) Comparison of perceived and quantitative measures of occupational noise exposure. Ann Occup Hyg ; 53: 41– 54. Google Scholar PubMed  Neitzel R, Seixas NS, Camp Jet al.  ( 1999) An assessment of occupational noise exposures in four construction trades. Am Ind Hyg Assoc J ; 60: 807– 17. Google Scholar CrossRef Search ADS PubMed  Neitzel R, Seixas NS, Camp Jet al.   ( 1999) An assessment of occupational noise exposures in four construction trades. Am Ind Hyg Assoc J ; 60: 807– 17. Google Scholar CrossRef Search ADS PubMed  Neitzel RL, Stover B, Seixas NS. ( 2011) Longitudinal assessment of noise exposure in a cohort of construction workers. Ann Occup Hyg ; 55: 906– 16. Google Scholar PubMed  NIOSH. ( 1972) Criteria for a recommended standard occupational exposure to noise . Cincinnati, OH: National Institutes of Occupational Safety and Health. NIOSH. ( 1998) Criteria for a recommended standard occupational noise exposure revised criteria 1998 . Cincinnati, OH: National Institutes of Occupational Safety and Health. OSHA. ( 2010) Occupational noise exposure . 1910.95. Washington, DC: Occupational Safety and Health Administration. pp. 211– 24. PubMed PubMed  Passchier-Vermeer W. ( 1969) Hearing loss due to exposure to steady-state broadband noise . Delft, The Netherlands: Research Institute for Public Health Engineering. Petrick ME, Royster LH, Royster JDet al.  ( 1996) Comparison of daily noise exposures in one workplace based on noise criteria recommended by ACGIH and OSHA. Am Ind Hyg Assoc J ; 57: 924– 8. Google Scholar CrossRef Search ADS PubMed  Picard M, Girard SA, Simard Met al.   ( 2008) Association of work-related accidents with noise exposure in the workplace and noise-induced hearing loss based on the experience of some 240,000 person-years of observation. Accid Anal Prev ; 40: 1644– 52. Google Scholar CrossRef Search ADS PubMed  Pourbakht A, Yamasoba T. ( 2003) Cochlear damage caused by continuous and intermittent noise exposure. Hear Res ; 178: 70– 8. Google Scholar CrossRef Search ADS PubMed  Qiu W, Davis B, Hamernik RP. ( 2007) Hearing loss from interrupted, intermittent, and time varying Gaussian noise exposures: the applicability of the equal energy hypothesis. J Acoust Soc Am ; 121: 1613– 20. Google Scholar CrossRef Search ADS PubMed  Rabinowitz PM, Galusha D, McTague MFet al.   ( 2012) Tracking occupational hearing loss across global industries: a comparative analysis of metrics. Noise Health ; 14: 21– 7. Google Scholar CrossRef Search ADS PubMed  Royster L, Royster J, Dobie RA. ( 2003) Prediction and analysis of the hearing characteristics of noise-exposed populations or individuals. In Berger EH, editor. The noise manual . 5th edn. Fairfax, VA: AIHA Press; 2003. pp. 669– 88. Google Scholar CrossRef Search ADS   Seixas N, Neitzel R, Sheppard Let al.   ( 2005a) Alternative metrics for noise exposure among construction workers. Ann Occup Hyg ; 49: 493– 502. Google Scholar CrossRef Search ADS   Seixas NS, Goldman B, Sheppard Let al.   ( 2005b) Prospective noise induced changes to hearing among construction industry apprentices. Occup Environ Med ; 62: 309– 17. Google Scholar CrossRef Search ADS   Seixas NS, Kujawa SG, Norton Set al.   ( 2004) Predictors of hearing threshold levels and distortion product otoacoustic emissions among noise exposed young adults. Occup Environ Med ; 61: 899– 907. Google Scholar CrossRef Search ADS PubMed  Seixas NS, Neitzel R, Stover Bet al.   ( 2012) 10-Year prospective study of noise exposure and hearing damage among construction workers. Occup Environ Med ; 69: 643– 50. Google Scholar CrossRef Search ADS PubMed  Sliney D. ( 1993) Review of the threshold limit value for noise. Appl Occup Environ Hyg ; 8: 618– 23. Google Scholar CrossRef Search ADS   Suter A. ( 1992) The relationship of the exchange rate to noise-induced hearing loss . Cincinnati, OH: Institute of Noise Control Engineering. Suter A. ( 2003) Standards and regulations. In Berger EH, Royster LH, Royster JD, Driscoll DP, Layne M, editors. The noise manual . 5th edn. Fairfax, VA: American Industrial Hygiene Association. Suter A. ( 2017) Occupational Hearing Loss from Non-Gaussian Noise . Semin Hear; 38: 225–62. doi:10.1055/s-0037-1603726. Suter AH. ( 1988) The development of federal noise standards and damage risk criteria. In: Hearing conservation in industry, schools, and the military . Boston, MA: Pearson Allyn & Bacon. pp. 45– 66. Suter AH. ( 2002) Construction noise: exposure, effects, and the potential for remediation; a review and analysis. AIHA J (Fairfax, Va) ; 63: 768– 89. Google Scholar CrossRef Search ADS PubMed  Suter AH. ( 2009) The hearing conservation amendment: 25 years later. Noise Health ; 11: 2– 7. Google Scholar CrossRef Search ADS PubMed  Suter AH. ( 2015) Letter to the editor : an alternative interpretation of issues surrounding the exchange rates. Ear and Hearing . Jul-Aug;36(4):485–487 Tak S, Calvert GM. ( 2008) Hearing difficulty attributable to employment by industry and occupation: an analysis of the National Health Interview Survey–United States, 1997 to 2003. J Occup Environ Med ; 50: 46– 56. Google Scholar CrossRef Search ADS PubMed  Tak S, Davis RR, Calvert GM. ( 2009) Exposure to hazardous workplace noise and use of hearing protection devices among US workers–NHANES, 1999-2004. Am J Ind Med ; 52: 358– 71. Google Scholar CrossRef Search ADS PubMed  Van Atta F, Cohen A, Glorig Aet al.   ( 1967) Guidelines for noise exposure control. Am Ind Hyg Assocation J .; 28: 418– 24. Google Scholar CrossRef Search ADS   Ward WD. ( 1991) The role of intermittence in PTS. J Acoust Soc Am ; 90: 164– 9. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of the British Occupational Hygiene Society.

Journal

Annals of Work Exposures and Health (formerly Annals Of Occupational Hygiene)Oxford University Press

Published: Apr 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off