JARO

JARO 9: 264–276 (2008)

DOI: 10.1007/s10162-008-0123-1

Journal of the Association for Research in Otolaryngology

Occupational Noise, Smoking, and a High Body Mass Index

are Risk Factors for Age-related Hearing Impairment
and Moderate Alcohol Consumption is Protective: A European
Population-based Multicenter Study
ERIK FRANSEN1, VEDAT TOPSAKAL2, JAN-JAAP HENDRICKX1,2, LUT VAN LAER1, JEROEN R. HUYGHE1,
ELS VAN EYKEN1, NELE LEMKENS2, SAMULI HANNULA3, ELINA MÄKI-TORKKO3, MONA JENSEN5,
KELLY DEMEESTER2, ANKE TROPITZSCH6, AMANDA BONACONSA7, MANUELA MAZZOLI7, ANGELES ESPESO8,
KATIA VERBRUGGEN9, JOKE HUYGHE9, PATRICK L. M. HUYGEN4, SYLVIA KUNST4, MINNA MANNINEN10,
AMALIA DIAZ-LACAVA11, MICHAEL STEFFENS11, THOMAS F. WIENKER11, ILMARI PYYKKÖ10, COR W. R. J.
CREMERS4, HANNIE KREMER4, INGEBORG DHOOGE9, DAFYDD STEPHENS8, EVA ORZAN7, MARKUS PFISTER6,
MICHAEL BILLE5, AGNETE PARVING5, MARTTI SORRI3, PAUL VAN DE HEYNING2, AND GUY VAN CAMP1,
1
Department of Medical Genetics, University of Antwerp, Universiteitsplein, 2610 Antwerp, Belgium
2
Department of Otorhinolaryngology, University Hospital of Antwerp, 2650 Antwerp, Belgium
3
Department of Otorhinolaryngology, University of Oulu, 90014 Oulu, Finland
4
Department of Otorhinolaryngology, St. Radboud University Medical Center, 6500 HB Nijmegen, The Netherlands
5
Department of Audiology, Bispebjerg Hospital, 2400 NV Copenhagen, Denmark
6
Department of Otorhinolaryngology, University of Tübingen, 72074 Tübingen, Germany
7
Department of Oto-Surgery, University Hospital Padova, 35128 Padova, Italy
8
College of Medicine, Cardiff University, CF14 4XW Cardiff, UK
9
Department of Otorhinolaryngology, University Hospital of Ghent, 9000 Ghent, Belgium
10
Department of Otorhinolaryngology, University of Tampere, 33014 Tampere, Finland
11
Institute of Medical Biometry, Informatics and Epidemiology, University of Bonn, 53105 Bonn, Germany
Received: 4 December 2007; Accepted: 21 April 2008; Online publication: 10 June 2008

ABSTRACT medical history. People with a history of disease that

could affect hearing were excluded. PTAs were ad-
A multicenter study was set up to elucidate the envi- justed for age and sex and tested for association with
ronmental and medical risk factors contributing to exposure to risk factors. Noise exposure was associat-
age-related hearing impairment (ARHI). Nine sub- ed with a significant loss of hearing at high sound
samples, collected by nine audiological centers across frequencies (91 kHz). Smoking significantly increased
Europe, added up to a total of 4,083 subjects between high-frequency hearing loss, and the effect was dose-
53 and 67 years. Audiometric data (pure-tone average dependent. The effect of smoking remained signifi-
[PTA]) were collected and the participants filled out a cant when accounting for cardiovascular disease
questionnaire on environmental risk factors and events. Taller people had better hearing on average
with a more pronounced effect at low sound frequen-
Electronic supplementary material The online version of this article cies (G2 kHz). A high body mass index (BMI)
(doi:10.1007/s10162-008-0123-1) contains supplementary material, correlated with hearing loss across the frequency range
which is available to authorized users. tested. Moderate alcohol consumption was inversely
Correspondence to: Guy Van Camp &Department of Medical Genetics &
University of Antwerp & Universiteitsplein 1, B - 2610 Antwerp, Belgium. correlated with hearing loss. Significant associations
email: guy.vancamp@ua.ac.be were found in the high as well as in the low fre-
264
FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment 265

quencies. The results suggest that a healthy lifestyle normal distribution with no differences between
can protect against age-related hearing impairment. males and females and no relation to age. However,
several studies indicate that typical unscreened pop-
Keywords: age-related hearing impairment, ulations have slightly worse hearing than predicted by
presbyacusis, occupational noise, smoking, risk factors the ISO 7029 standard, whereby the former seem to
have an apparent excess aging by 10–15 years
compared to the population described by the ISO
INTRODUCTION 7029 standard (Robinson 1988; Lutman and Davis
1994; Engdahl et al. 2005).
Several factors contribute to the decline in hearing
acuity with increasing age. Apart from biological
Importance of genetics
degeneration because of aging in itself, age-related
hearing impairment (ARHI) is influenced by genetic The relative importance of genes in ARHI is age-
risk factors, exposure to noise and toxic substances, dependent. A heritability estimate by Karlsson et al.
and the occurrence of certain diseases. The relative (1997) indicated that in the age stratum 56–65, 58%
contributions of the different risk factors are difficult to of the variance in hearing thresholds was attributable
estimate, and the interactions between them remain to the influence of genes, declining to 47% in the
unclear. stratum over 65. In the Framingham cohort, Gates et
Typically, ARHI is sensorineural, bilaterally symmetri- al. (1999) found a clear familial aggregation of the
cal and more pronounced at high frequencies with males hearing thresholds. The heritability also seems to be
more severely affected than females. There is a large frequency-dependent with a higher heritability in the
variation between individuals, which is larger in males low frequencies.
than in females (Lee et al. 2005). Variability increases The advent of high-throughput methods for genetic
with age and is greater at the high frequencies. analysis provides tools for identifying genetic variants
ARHI starts slowly around the fifth decade and implicated in ARHI. A genome-wide linkage study in
worsens gradually, becoming the most common senso- the Framingham cohort identified several loci with
ry impairment in the elderly. Between the ages of 60 suggestive evidence for linkage (DeStefano et al. 2003).
and 70 years, about one third of the population has an Garringer et al. (2006) reported suggestive linkage in
average hearing loss (HL) of 25 dB or more for pure the DFNA18 region in the general population. Poly-
tones at 0.5, 1, 2, and 4 kHz. Between the age of 70 and morphisms in the N-acetyltransferase 2 (NAT2) and
80 years, the proportion of individuals with a pure-tone KCNQ4 genes were found to be associated with ARHI
average (PTA) showing over 25 dB HL increases to in two independent populations (Ünal et al. 2005; Van
50% (Davis 1994). Although this is considered mild Eyken et al. 2006, 2007).
hearing loss, it seriously affects an individual’s ability
to communicate in a noisy environment.
Environmental risk factors
Whereas research into genetic variants associated with
Z scores
ARHI is still in its infancy, there is substantial literature
When comparing ARHI in men and women of dif- about environmental risk factors contributing to ARHI.
ferent ages, one needs to correct for gender and age The influence of occupational noise is well-documented.
differences between the subjects, and this correction The permanent threshold shift caused by exposure to
is different depending on the frequencies studied. occupational noise of a given intensity over a given
The expected median hearing threshold as a function period of time is predicted by the ISO 1999 standard
of age, sex, and frequency, plus the standard deviation (International Organisation of Standardization 1984).
around this median, is given by the ISO 7029 standard This effect is most pronounced at 2, 3, 4, and 6 kHz
(International Organisation of Standardization 1984). (Dobie 2005).
In a previous paper, we developed a method to quan- However, individual noise susceptibility shows great
tify how severely a person is affected by age-related variability. This may be because of complex interac-
hearing loss, given his/her age and sex (Fransen et al. tions with nonoccupational noise, other environmen-
2004). In this method, a Z score is defined as the tal factors, or genetic predisposition (Helzner et al.
standardized difference between an individual’s ob- 2005). The most deleterious source of nonoccupa-
served hearing threshold at a given frequency and the tional noise is gunfire noise (Lutman and Spencer
age- and sex-specific median for that frequency. This 1990; Clark 1991). People exposed to both occupa-
allows the comparison of individuals of different age tional noise and gunfire noise have poorer hearing in
and sex. Ideally, in a randomly collected highly both ears compared to people exposed to occupa-
screened population, Z scores should have a standard tional noise alone (Stewart et al. 2001).
266 FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment

A more than additive effect between noise and and body mass index (BMI; either very high or very
organic solvent exposure was found in a study of low) were found to be risk factors. These effects were
plastics factory workers coexposed to noise and styrene dependent on birth weight and length (Barrenas et al.
(Sliwinska-Kowalska et al. 2003), whereas an additive 2005). In particular, overweight subjects who were
effect was observed in dockyard laborers coexposed to born light and small for their gestational age had an
noise and a mixture of solvents (Sliwinska-Kowalska et increased risk for hearing loss at conscription. In an
al. 2004). Fuente and McPherson (2007) observed adult Danish cohort, very short subjects between 50
significantly increased hearing loss in workers exposed and 55 years of age had a higher prevalence of hearing
to a mixture of organic solvents compared to unex- loss compared to very tall subjects (Burr et al. 2005).
posed controls. Their results suggested solvent-in- This study is part of a European multicenter study
duced central auditory damage. into the genetic and environmental causes of ARHI. In
There is controversy on the effect of smoking. this paper, we present data from a total of 4,083
Rosenhall et al. (1993) found an association between unrelated subjects, collected through nine audiologi-
hearing levels and smoking; Cruickshanks et al. (1998) cal centers from seven European countries. Potential
and Uchida et al. (2005) found indications of a dosage risk factors were assessed by questionnaire for associa-
effect. On the other hand, no association was found in tion with ARHI as established by audiometric criteria.
the Framingham cohort (Gates et al. 1993). Recent
studies suggest an interaction between smoking and
occupational noise, whereby the deleterious effect of METHODS
noise exposure is exacerbated by smoking (Ferrite and
Sample collection
Santana 2005; Nomura et al. 2005; Wild et al. 2005).
Data on alcohol consumption are not very clear. An The collection of subjects was a collaborative effort of
association with chronic alcohol abuse has been nine expert audiological centers from seven European
observed (Rosenhall et al. 1993), but with moderate countries: two from Belgium (Antwerp, Ghent), two
alcohol intake, the results are less clear. Brant et al. from Finland (Tampere, Oulu), one from The Nether-
(1996) found no significant effect on the thresholds at lands (Nijmegen), one from Germany (Tübingen), one
speech frequencies but high frequencies were not from Denmark (Copenhagen), one from Italy (Padua),
tested. In contrast, Helzner et al. (2005) reported a and one from the UK (Cardiff). To collect study
protective effect at high frequencies, which depended subjects, the audiological centers used three different
on race and gender. recruitment strategies: (1) A clinic-based sample,
whereby subjects are collected through the regular
influx of patients visiting an audiological or ENT clinic.
Medical risk factors
As this strategy tends to recruit an excess of people with
In addition to some environmental factors, several poor hearing, the spouses of the recruited subjects were
medical risk factors are suspected to have an influence asked to join the study. (2) A population-based sample,
on hearing. Many studies have focused on cardiovascu- whereby subjects were collected via advertisements in
lar diseases (CVD), as these are very prevalent in the local media or through local population registers and
elderly population. In the Framingham cohort, an letters of invitation. (3) A mixed strategy, whereby part
association between cardiovascular events (stroke, cor- of the samples was population-based and the remaining
onary heart disease, or intermittent claudication) and part was clinic-based. The nine sample sets collected by
low-frequency hearing loss was reported (Gates et al. the audiological centers are hereafter referred to as
1993). They also reported an inverse relationship subsamples.
between high-density lipoprotein (HDL) levels and To make each subpopulation ethnically homoge-
hearing thresholds, suggesting a protective effect of neous, we requested that at least three out of the four
HDL on hearing thresholds. Torre et al. (2005) found grandparents originated from the same region as the
a significant association between myocardial infarction study subject. An effort was made to collect an
and hearing loss in females, but not in males. Brant et approximately equal number of males and females
al. (1996) reported an association between hearing and to have a uniform age distribution. All responding
threshold and systolic blood pressure. A relationship subjects underwent clinical examination and otoscopy
between high-frequency sensorineural hearing impair- and completed a detailed questionnaire on medical
ment and diabetes mellitus has been reported by history and exposure to environmental risk factors.
several investigators (Kurien et al. 1989; Vaughan et al. The complete questionnaire is available upon request.
2006), but this association is highly controversial A list of all questions and answers used in this paper is
because of the heterogeneity in diabetic patients. In provided in Supplementary Table 4. Subjects with ear
a large cohort of Swedish conscripts, followed-up from diseases, possible monogenic forms of hearing impair-
birth till conscription (age 17–24), short adult stature ment, or other major pathologies with a possible
FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment 267

influence on hearing were excluded. The main goal exposure were used: 0 = no exposure (never fired), 1 =
was to study hearing impairment in healthy subjects less than 100 rounds, 2 = 100–1,000 rounds, 3 = 1,000–
and, therefore, persons with multiple hospitalizations 10,000 rounds, and 4 = more than 10,000 rounds. For
were excluded. The complete list of exclusion criteria the heavy weapons, the latter two levels were combined
was previously reported (Van Eyken et al. 2006). In to obtain an ordinal variable with four exposure levels.
subjects passing the medical exclusion criteria, audio- The use of hearing protection was documented, which
metric thresholds were determined for air conduction allowed us to separately count shots with protection
(0.25, 0.5, 1, 2, 3, 4, 6, and 8 kHz) and bone con- and shots without protection. Hence, every subject had
duction (0.5, 1, 2, and 4 kHz) according to current four summary values describing the exposure to
clinical standards (ISO 8253). We excluded subjects gunfire noise: protected light, unprotected light,
with asymmetrical hearing loss (between-ear differ- protected heavy, and unprotected heavy.
ence in air conduction threshold larger than 20 dB To combine the cumulative exposure from the light
for at least two frequencies out of 0.5, 1, and 2 kHz). and heavy weapons, we summed the exposure levels
In case only one of the ears showed conductive from light and heavy weapons, adding one to the
hearing loss (air–bone gap of 15 dB or more at 0.5, exposure level of heavy weapons with an upper limit of
1, and 2 kHz) and in the absence of other exclusion four. This corresponds to multiplying the number of
criteria, the other ear could be included. rounds with heavy weapons by ten and adding them to
Research was approved by the ethical committees the number of light rounds.
of the institutions connected to each research center:
University of Antwerp, University Hospital of Antwerp,
Hypertension
University of Oulu, University Medical Center Nijme-
gen, Bispebjerg Hospital Copenhagen, University of Blood pressure was measured once according to
Tübingen, University Hospital Padova, Cardiff Univer- standard procedures. These measurements were
sity, University Hospital of Ghent, University of recoded into a binary variable to test for association
Tampere, and University of Bonn. All persons gave with ARHI. Subjects with either a systolic blood pressure
their informed consent before inclusion in this study. above 140 mm Hg or a diastolic blood pressure above
90 mmHg (Chobanian et al. 2003) or who were taking
antihypertensive drugs were considered hypertensive.
Z scores
Z scores were calculated as described by Fransen et al.
Statistical analysis
(2004). In brief, for each individual, we calculated the
age- and sex-specific median threshold at each fre- To enable parametric data analysis (ANOVA or regres-
quency, based upon the ISO 7029 standard. This value sion), Box–Cox transformations of the Z scores were
was subtracted from the observed hearing thresholds carried out using the statistical package R (http://
at each frequency. The difference, which may be www.r-project.org). In each of the nine subsamples, a
negative (=better hearing than median) or positive separate transformation was carried out, and all
(=worse hearing than median), was normalized by further calculations for that given subsample were
dividing by the age-, sex-, and frequency-specific performed on the transformed outcome variable. For
standard deviation given by the ISO 7029 standard. the joint analysis, a new Box–Cox transformation on
These calculations produce frequency-specific Z the combined dataset was carried out.
scores. In this study, we used summary values for the Association between the Z scores and a categorical
high and low frequencies, respectively: The high- risk factor was tested using ANOVA. Ordinal or
frequency Z score (Zhigh) is the average of the Z scores numeric risk factors were tested via linear regression.
at 2, 4, and 8 kHz, and the low-frequency Z score Theoretically, in a highly screened population, Z
(Zlow) is the average of the Z scores at 0.25, 0.5, and scores should correct for gender effects. However, we
1 kHz. In all analyses presented in this study, we used noticed that, in several of the subsamples, there was a
Zhigh and Zlow of the better hearing ear (based upon significant difference in Z scores between males and
the average PTA at 0.5, 1, and 2 kHz). females. Therefore, gender was entered as a covariate
into the regression or ANOVA models. All models
were built in a stepwise backward way. First, a full
Exposure to gunfire noise
model was fitted including the risk factor of interest,
The subjects were asked how many rounds they had along with gender and the interaction term between
fired with either light or heavy weapons. Rifles or these. In such model, the interaction term tests
machine guns were counted as light weapons, whereas whether the effect of the risk factor is significantly
large infantry weapons and artillery were counted as different between the two sexes. If the interaction
heavy weapons. For light weapons, five levels of noise term was not significant, it was omitted from the
268 FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment

model and a new model with only the two main effects linkage) was performed using the cluster package in R.
was fitted. In case the gender term was not significant The image plots were generated using the fields
in this simplified model, it was omitted, resulting in a package in R.
model containing only the risk factor. Otherwise, in
cases for which the gender term was significant, we
kept it in the model, and the significance of the risk RESULTS
factor was tested accounting for the gender effect.
Collection of subjects
To check the appropriateness of the fitted model
and to find outlying observations, residual plots Through nine clinical (ENT/audiological medicine)
(normality of the raw residuals, predicted value vs. groups, unrelated Caucasian subjects were collected
raw residuals, studentized residuals vs. independent in seven different European countries. Table 1 lists
variable) were visually inspected. the collecting groups, the sizes of the subsamples, and
In the joint analysis, all nine subsamples were com- the recruitment strategy.
bined and a categorical variable for subsample was A total of 4,083 subjects, including 1,967 males and
added to the model. This adjusted the estimated effect 2,116 females, passed our inclusion criteria. The age
size and significance of the risk factors for differences in range was set at 53 to 67 years. The mean overall age was
mean Z scores between populations to obtain a com- 60.4 for females (SD = 3.2) and 60.9 for males (SD =
mon effect size. 3.2). The mean ages for males and females for each
To deal with the multiple testing burden, a country and center are given in Supplementary Table 1.
Bonferroni correction was applied to the results of Mean audiograms, showing the mean thresholds and
the joint analyses. A total of 86 tests were performed standard deviation at each frequency, are shown in
(Supplementary Table 3), 12 of which were part of a Supplementary Fig. 1. The mean Z scores for each
multivariable analysis and not independent from the subsample are shown in Supplementary Table 2.
univariable tests. Therefore, the number of indepen-
dent tests equals 74, and the threshold for significance
Association testing
was adjusted to 0.05/74 = 0.00068.
Most questions from the questionnaire refer to previ-
ously reported environmental and medical risk factors.
Cluster analysis
We have tested these risk factors for association with
We used the p values of 43 association tests (in Zhigh hearing loss in the high and low frequencies. To
and Zlow) to check whether the results obtained in the compare the hearing acuity in subjects with a different
subsamples with the same recruitment strategy were age and gender, audiometric measurements were
more similar to each other than to the subsamples converted to Z scores.
collected using a different strategy. A list of the 43 For each of the risk factors, we first analyzed all nine
tests, corresponding to the tests shown in Supplemen- subsamples separately. We have looked for significant
tary Table 3, is given in Supplementary Table 5. associations that replicated across multiple subsamples
For the cluster analysis, the p values of the 43 tests in with all associations having the same direction. After-
the nine subsamples were arranged in a 9 × 43 matrix. wards, a joint analysis was performed on all nine
Hence, each subsample was considered to be one 43- subsamples combined. The Bonferroni-adjusted signif-
variate observation. Hierarchical clustering (complete icance threshold for the joint analysis was 0.00068.

TABLE 1

Number of samples collected in each subsample

City Country Males Females Total Recruitment

Antwerp Belgium 556 606 1,162 Population registries

Copenhagen Denmark 212 208 420 Clinic
Cardiff United Kingdom 197 159 356 Clinic
Gent Belgium 159 179 338 Clinic+advertisement
Nijmegen Netherlands 146 138 284 Clinic
Oulu Finland 218 282 500 Population registries
Padua Italy 173 189 362 Clinic+advertisement
Tampere Finland 100 155 255 Clinic
Tübingen Germany 206 200 406 Clinic+population registries
Total 1,967 2,116 4,083
FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment 269

TABLE 2

Association test between Z score and ARHI risk factors in subsamples and joint analysis
a
Risk factor Answer Antwerp Cardiff Copenhagen Ghent Nijmegen Oulu Padua Tampere Tübingen Joint analysisb

Morphometry
Height Number
Zhigh 0.04 0.009 0.05
Zlow 5.4E−5 0.05 0.02 0.03 0.007 0.02 3.5E−7
BMI Number
Zhigh 0.03 0.08 0.08 7.7E−7 0.0004
Zlow 6.3E−8 0.06 0.07 4.0E−5 2.7E−7
Medical risk factors
Heart attack Yes/no
Zhigh – – – – – 0.05
Zlow 0.02 – – 0.03 – – – 0.002
CVD event Yes/no
Zhigh – 0.09 – – –
Zlow 0.02 – 0.01 0.048 – – – 0.0003
Noise and solvent exposure
Gunfire noise Number
Zhigh_prot 0.01
Zhigh_unpr 0.003 0.009
Occupational noise exposure Yes/no
Zhigh 0.01 0.0005 0.01 3.0E−5 0.0002 0.01 0.002 0.001 1.0E−17
Zlow 0.007 0.01 0.01 0.02 0.04 1.0E−9
Years noise exposure Number
Zhigh 0.09 0.02 0.007 0.0003 0.0005 0.002 0.002 0.005 1.0E−17
Zlow 0.07 0.06 0.02 0.06 1.1E−7
Solvent exposure Yes/no
Zhigh 0.07 0.07 0.001
Zlow 0.07
Smoking
Current or former smoking Yes/no
Zhigh 0.01 0.01 0.03 0.0009
Zlow 0.02 0.06 0.08
Smoking history Number
(packyears)
Zhigh 0.053 0.001 0.02 0.006 0.0200 0.0002 1.0E−9
Zlow 0.03 0.0004
Packyears by sex Number
(high frequency)
Male 0.02 0.01 0.08 0.03 0.0004 1.9E−7
Fem 0.02 0.02 0.0006
Packyears by sex Number
(low frequency)
Male 0.07 0.03 0.001
Female 0.04 0.07
Smoking adjusted for CVD Number
and BMI
Zhigh 0.03 0.003 0.01 0.01 0.06 0.0006 2.4E−8
Zlow 0.07 0.008
Smoking dosage effect Number
in smokers
Zhigh 0.09 0.04 0.001 0.006 3.0E−7
Zlow 0.08 0.08 0.005
Alcohol consumption
Alcohol consumption Yes/no
Zhigh 0.06 0.01 0.002 0.050 0.09 8.4E−6
Zlow 0.01 0.002 0.006 5.2E−6

Only significant (pG0.05) associations or trends toward significance (pG0.1) are shown, other cells are empty. “–” means there were not enough exposed subjects to
test the risk factor.
a
An extended table with all associations tested is provided as Supplementary Table 3. The exact questions and answers as they appeared in the questionnaire are
provided as Supplementary Table 4.
b
Values in italics remain significant after Bonferroni correction for 74 tests (pG6.8E−4).
270 FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment

Table 2 lists the associations that were significant (p G TABLE 3

0.05) or showed a trend towards significance (p G 0.1) Prevalence of common diseases and smoking in ARHI
and also showed a significant association in the joint and Framingham study
analysis. A complete overview of all associations tested
Risk factor ARHI study Framingham study
is provided in Supplementary Table 3.
Smoking
Total number of subjects 4,102 1,665
Morphometry Percentage exposed 47.3 13.5
Heart attack
Besides environmental and medical risk factors, some Total number of subjects 4,078 1,662
studies have reported associations with morphometric Percentage affected 3 10.4
characteristics. In our nine subsamples, we found Intermittent claudication
several associations between Z scores and body weight Total number of subjects 4,074 1,662
and the BMI. The associations were most pronounced Percentage affected 1 7.2
Stroke
for BMI. In all subsamples with a significant association Total number of subjects 4,075 1,664
(2/9 for Zhigh and Zlow), as well as in the two further Percentage affected 0.1 7
subsamples with a trend toward association (p G 0.1), a Total CVD
higher body weight or BMI leads to a significant Total number of subjects 4,083 1,664
increase in the mean Z score. Because a higher Z score Percentage affected 2.0 27.8
Coronary heart disease
reflects worse hearing, this means that people with a Total number of subjects 4,083 1,664
higher BMI on average have more hearing loss. Upon Percentage affected 1.1 19.4
joint analysis, BMI was highly significantly associated
with both Zhigh and Zlow, whereas body weight was only
associated with Zlow. As the effect of BMI is possibly
because of confounding with CVD and smoking, we
performed a multivariable analysis that included all Either no or only very weak evidence for association
three risk factors (Supplementary Table 3). The was found for hypertension, whiplash injuries, osteo-
significance of the association between Z scores and porosis, osteoarthritis, allergy, or diabetes (Supplemen-
BMI slightly diminished when controlling for the two tary Table 3). CVD has often been reported as a risk
other risk factors, but the joint analysis was still highly factor for hearing loss, but in most of our subsamples,
significant. the number of affected individuals was too small to
No associations were found between Z scores and eye test for association with the Z scores. We could test
color, and there was only weak evidence for an four subsamples for an association between hearing
association with sunburn susceptibility and left/right- loss and heart attack. In two of them, the association
handedness. Body length was associated with Zlow in 5/9 with Zlow was significant in a way that people having
subsamples and with Zhigh in 2/9 subsamples. Across all had a heart attack had on average an increased Zlow
significant associations, we observed that a higher body compared to the rest of the subjects. The joint analysis
length was negatively associated with the Z score. This where a larger number of subjects were available
implies that taller people on average have better hearing. showed a nominally significant association with low-
frequency hearing loss. To test whether the occur-
rence of any CVD event is a risk factor for hearing
Medical risk factors
loss, we created an indicator variable denoting the
Compared to the Framingham study (Gates et al. 1993), occurrence of either coronary heart disease, heart
the ARHI study showed a very low prevalence of com- attack, stroke, or intermittent claudication. Subjects
mon diseases (Table 3). This is probably attributable having taken medicine for any of these disorders were
to the lower age range of the ARHI study compared to also considered affected. In three subsamples (of the
the mean age of the Framingham subjects (age 73) in five we could analyze), we found an association
addition to the exclusion of subjects with multiple between this indicator variable and Zlow. This signifi-
hospitalizations in the ARHI study. cant association with Zlow was also observed upon joint
In this study, we tested whether a number of com- analysis (Table 2), reaching the Bonferroni-corrected
mon diseases lead to increased Z scores. We only tested significance threshold.
for association in a subsample if the number of pa-
tients with the disease was sufficiently large (at least
Influence of drugs
15). The data on medical history were solely based on
the questionnaire and no validation has been per- The use of some drugs may have an influence on
formed. Therefore, the association results on medical hearing ability. Only in the Tampere subsample, aspirin
risk factors have to be regarded with caution. use was associated with increased Zhigh, but the
FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment 271

association was not very strong (p = 0.03, Supplemen- colleague at 1m distance. In addition, we asked for
tary Table 3). The use of painkillers, on the other the duration of the exposure and whether ear protec-
hand, was associated with increased hearing loss in the tion was used.
low frequencies in the Antwerp and Tübingen sub- In all subsamples except the one from Antwerp, we
samples, but the joint analysis did not reach signifi- found a significant association between occupational
cance (p = 0.06, Supplementary Table 3). Only in the noise exposure and high-frequency hearing loss.
Cardiff subsample were we able to test the influence of Regardless of the duration of the exposure and of
atorvastatin, but no significant effect was found the use of protection, we consistently observed that
(Supplementary Table 3). people having worked for more than 1 year in a noisy
place had higher Z scores in the high frequencies
compared to unexposed people. The same trend, but
Gunfire noise
smaller, was observed in the lower frequencies.
To score the exposure to gunfire noise, we asked the Significance was reached in 5/9 subsamples (Table 2).
subjects how many rounds of ammunition they had Quantifying the noise exposure by calculating the
fired with either light or heavy weapons with and number of years exposed and the daily exposure time
without hearing protection. A combined exposure level also showed significant and consistent associations for
for light and heavy weapons was calculated, separately hearing loss in both high and low frequencies, but the
counting the rounds fired with and without protection, significance level was slightly decreased compared to
so that each individual had an unprotected and a the previous analysis. Accounting for the exposure to
protected exposure level. gunfire noise did not have much influence on the
To test the influence of gunfire exposure on hearing, significance of the association between Zhigh and oc-
we regressed the Z scores from the best ear on the cupational noise. In the low frequencies, a slight
protected and unprotected gunfire exposure levels. increase in significance was observed.
Only two significant associations were found for Zhigh Only 16% of exposed subjects reported using
in two separate subsamples: in the Ghent subsample, hearing protection ‘always’ or ‘most of the time’. Re-
there was a significant association with the unprotect- stricting our analysis to exposed males, only four of the
ed exposure level (p = 0.003); the other association was subsamples offered sufficient numbers to test for an
with the protected exposure level in the Antwerp effect of protection on hearing impairment. Only the
subsample (p = 0.01). Upon joint analysis, we found an Oulu subsample showed a significant protective effect
association only between Zhigh and the unprotected in the high-frequency range (p = 0.02). Upon joint
exposure level (Table 2). In a multivariable analysis analysis, we also saw a significant effect in high
that also included exposure to occupational noise, the frequencies (p = 0.04, Supplementary Table 3).
effect of gunfire noise showed no consistent effect Although the significance is marginal, this p value
across the subsamples, and the overall effect was only was reached with very low numbers: of all noise-
marginally significant in the high frequencies. exposed males only 110 (12%) wore protection most
of the time or always.
Leisure noise
Solvents and toxic chemicals
Very low numbers of subjects reported repeated
exposure to noise during their leisure time, and the Subjects were asked for occupational exposure to
time they had been exposed varied considerably. organic solvents and other toxic substances. We had
Therefore, the effect of leisure noise was not analyzed enough exposed individuals for statistical testing only
further. for organic solvents, including aromatic carbohydrates
(toluene, xylene, and styrene), trichloroethylene, and
hexane. The number of people exposed to other
Occupational noise
substances was too small to test. As only 13 females
Work histories of the subjects were collected. As the were exposed to solvents, we restricted this analysis to
subpopulations were collected without any selection males.
regarding occupational noise, there was a large varia- When organic solvent exposure was scored as a
tion in the number of jobs with noise exposure, the binary trait, none of the subsamples showed a signifi-
total number of years worked in noise, and the age of cant association with hearing levels, although two
the individuals at the time of exposure. To classify the subsamples showed a trend in the high frequencies.
subjects according to occupational noise exposure, we The joint analysis gave a nominally significant associa-
asked every subject whether (s)he had ever worked for tion (not reaching the Bonferroni-corrected signifi-
more than a year in a noisy environment and whether a cance level) in the high frequencies, which may imply
raised voice was necessary to communicate to a that the negative result in the subsamples is merely
272 FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment

because of the small number of exposed subjects and whereas in the fifth subsample, both sexes remained
that there indeed may be a small effect that only significant.
becomes significant if the numbers are high enough When restricting the analysis to smokers, we saw a
(Table 2). significant dose effect in 3/9 subsamples at the high
frequencies (Table 2, smoking dosage effect). Among
smokers, there was a significant association between
Noise–solvent interaction the number of packyears and high-frequency hearing
loss. These analyses were highly significant on joint
We tested for nonadditive effects of noise and solvent
analysis. The higher significance when testing the
exposure because synergistic effects between these
association with packyears gives support to the hy-
two risk factors have been reported. In none of the
pothesis that the effect of smoking on hearing loss is
subsamples did we find a significant interaction
dose-related.
between noise and solvents. The group of people
The association between smoking and hearing loss
exposed to both occupational noise and solvents did
remained highly significant when accounting for CVD
not show significantly increased hearing loss com-
events and BMI. There were still 5/9 subsamples
pared with unexposed subjects or subjects exposed
having a significant effect of smoking on Zhigh with
to only one of the two risk factors alone (Supple-
only minor changes in significance level. This multi-
mentary Table 3). The numbers are quite small,
variable analysis shows that the association between
however, and the lack of significance may merely
smoking and Z scores cannot be attributed to
reflect the lack of power of this study to detect small
confounding with CVD events.
interaction effects.

Noise–smoking interaction
Smoking
In the male subjects, we tested for nonadditive effects of
Subjects were asked about smoking habits by first
occupational noise and smoking on hearing. Zhigh and
inquiring as to whether they had ever smoked
Zlow were regressed on the binary variables for
regularly. In addition, we asked for the number of
smoking and occupational noise and the interaction
years they had been smoking and the number of
between them. In one of the subsamples (results not
cigarettes per day.
shown), a weakly significant interaction was found (p =
Dichotomizing the population into ever-smokers
0.04), but this effect was not found upon joint analysis
and never-smokers showed a significant association with
(p 9 0.05). Therefore, we did not find strong evidence
Zhigh in 3/9 subsamples and with Zlow in 1/9 sub-
that the effect of noise on hearing may be different
samples (Table 2). To investigate dosage effects, we
between smokers and nonsmokers. Subdividing the
estimated the number of packyears by multiplying the
population into four groups (noise exposure, smoking,
time (in years) an individual had been smoking by a
both noise and smoking, or none) did not indicate a
weighting factor for daily consumption of tobacco
more-than-additive effect of noise and smoking (Sup-
(G10 cigarettes per day = 0.5; 10–20 cigarettes per
plementary Table 3). Just as in the case of the noise–
day = 1; 920 cigarettes per day = 1.5). For nonsmokers,
solvent interaction, this study does not have high
the number of packyears was set to zero. Linear
power to detect interactions. It should be noted that
regression revealed significant associations between
the lack of significance in the interaction tests may
packyears and Zhigh in 5/9 subgroups. The effect was
merely reflect a lack of power.
less pronounced at the low frequencies with only one
significant association (in the Cardiff subsample), but
the joint analysis was highly significant for both the
Alcohol consumption
high and the low frequencies.
In none of the subsamples did we find a significant Subjects were asked if they regularly (at least once a
interaction between smoking and gender. Previous week) consumed alcohol. One glass of wine, spirit, or
studies had tested for association in males and females beer counted as one unit of consumption. Analyzing
separately. With this latter association test, the most alcohol consumption as a binary variable showed
significant associations were found in males in the high significant association in 3/9 subsamples in the high
frequencies (4/9 subsamples significant; Table 2, pack- as well as in the low frequencies (Table 2). In addition,
years by sex). Among the five subsamples that initially two further subsamples (Antwerp and Tübingen)
showed a significant main effect of packyears, three showed a trend in the high frequencies (p G 0.1). In
subsamples no longer showed significant values in all these analyses, alcohol consumption consistently
women, whereas the males remained significant. The leads to a decrease in Z scores and, thus, to improved
opposite was observed in one subsample (Ghent), hearing. Contrary to previous reports, we found no
FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment 273

evidence for the effect being modified by gender (not significant associations will be observed just by chance.
shown). A classic way to overcome this multiple testing
problem is to divide the threshold for statistical
significance by the number of independent tests. This
DISCUSSION approach (the Bonferroni method) fixes the family-
wise error rate to 5%, but is known to be conservative.
The study presented in this paper is part of a study In fact, when we apply this method to the tests in the
into the genetic, environmental, and medical causes subsamples, only a few associations (with a p value
of ARHI across seven European countries. In this below 0.05/566 = 8.8E − 5) would remain significant.
study, we have used the results of a questionnaire to The joint analyses reach higher significance levels
study the association between hearing levels and because of the larger sample size. As only 74 indepen-
several putative environmental and medical risk dent joint tests were performed, the Bonferroni-
factors. The use of questionnaire data implies the corrected significance threshold is 0.00068, and 21 of
study is retrospective. No further validation of any the association tests reached significance. The joint
questionnaire result has been performed. association tests reaching Bonferroni-corrected signif-
The subjects in our study are between 53 and 67 icance are marked in italics in Table 2 and marked in
years of age, which is relatively young to study this bold in Supplementary Table 3.
pathology. There are several reasons for choosing this The joint analysis is very powerful to detect small
age range. effects that cannot be picked up in the separate
First, this study is not only focused on the hearing subsamples. In a heterogeneous sample like the ARHI
loss because of aging alone (occurring in the absence sample, it is possible that the results of the subsamples
of risk factors), but also on the influence of risk factors differ because of their different recruitment strategies.
at earlier (preretirement) age. Although we had the Therefore, we have included a covariate for origin in the
long list of medical and otological exclusion criteria, joint analysis to avoid a spurious association because of
people with noise or solvent exposure were not confounding factors that differ between subsamples. Still,
excluded. Hence, the hearing loss in our study it is useful to consider the results of the separate
population is because of a combination of exposures subsamples. A consistent effect in many of the subsam-
and aging. Second, we are using this sample set to find ples, in addition to an overall effect, is a strong indication
genes involved in ARHI, and the relative importance of that the effect is genuine, even if only borderline
genes is higher in younger age groups. Results of the significance levels are observed in the subsamples.
genetic studies are published elsewhere (Van Laer et al. The nine subsamples were collected using three
2008). Third, collecting relatively young persons different recruitment strategies (Table 1), which may
offers the potential for follow-up and longitudinal bias the sample. We have tested whether the recruit-
data analysis. This is very relevant in view of the ment strategy has an influence on the results of the
ongoing debate about the influence of early noise analysis, but no indications of this were found. First,
exposure on hearing loss later in life. Studies in mice cluster analysis on the p values did not indicate that
indicate that early noise damage can render the the results obtained in the subsamples with the same
cochlea more vulnerable for hearing loss at later ages recruitment strategy were more similar to each other
(Kujawa and Liberman 2006), whereas studies in than to subsamples collected using a different strategy
human show that a cochlea that was previously (Supplementary Fig. 2). Second, visual inspection of
damaged by noise (i.e., showing an audiometric the significance levels of the different association tests
notch) has a different pattern of age-related hearing using an image plot (Supplementary Fig. 3) did not
loss later in life compared to an undamaged cochlea suggest any relation between recruitment strategy and
(Gates et al. 2000; Lee et al. 2005). A cross-sectional significance level.
study like the one presented in this paper has weak Exposure to occupational noise was by far the most
power to disentangle the effects of early exposures significant and consistently replicated risk factor.
and aging, but the dataset offers the potential to study Although the subjects in our sample sets were not
this in the future. selectively sampled for this type of analysis, and
For every risk factor, we first tested association in exposure levels, the duration, and the type of noise
nine subsamples separately, followed by a joint analysis (impulse noise or steady noise) were highly heteroge-
on the combined sample. Not all association tests neous, the association with high-frequency hearing
showing a p value below 0.05 represent a genuine loss was significant in all but one subsample. The
effect. We have performed a large number of inde- effect at low frequencies was also consistent, although
pendent tests (a total of 566 univariable tests in the less pronounced. It is interesting to note that the
subsamples and 74 joint analysis tests, Supplementary effect on the low frequencies becomes more signifi-
Table 2). Even if no effect is present, many nominally cant when gunfire noise is taken into account.
274 FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment

Noise exposure has been reported to act synergisti- Accounting for CVD events when testing the
cally with exposure to organic solvents. We found no association between smoking and Z scores increased,
evidence for noise–solvent interactions, but this is most rather than decreased, the significance level. This
likely because of the small number of people in the indicates that the increased occurrence of CVD events
doubly exposed group, which allows the detection of in smokers does not explain the association between
only major nonlinear effects. The interaction tests in smoking and hearing loss. This does not necessarily
our study lacks power to detect small interaction effects, mean that smoking itself is directly responsible for a
and not finding a significant interaction does not mean decrease in hearing ability. We cannot exclude the
it does not exist. For solvent exposure alone, on the possibility that the significant association is because of
other hand, the joint analysis showed a significant effect confounding with unknown factors, such as socioeco-
in the high frequencies, despite the relatively low nomic or educational factors, that are associated with
number of exposed subjects and the heterogeneity in both smoking and hearing loss. It is remarkable
exposure time and constitution of the solvent. though, to see a consistent association across so many
Our study conflicts with a previous report on the of the subsamples.
effect of smoking on hearing loss (Gates et al. 1993). In The effect of CVD events seemed to be more
this study, we observed a highly significant association pronounced in the low frequencies. In fact, three of
in more than half of our subsamples, consistently the nine subsamples showed a significant association.
indicating a deleterious effect on high-frequency In the multivariable analysis, including BMI and
hearing. It is interesting to note that the effect was smoking, the significance levels of CVD events were
most pronounced in a quantitative analysis taking into only slightly decreased compared to the univariable
account the number of packyears, rather than dichot- analysis (Supplementary Table 3).
omizing the population into people who ever smoked A weak but significant association between smoking
and people who never did. This is in line with and hearing loss in the elderly has been reported
previous reports of a dosage effect (Gates et al. previously by Rosenhall et al. (1993) in a cohort of
1993). A more specific analysis of this dosage effect, people aged between 70 and 85 years. The effect was
restricted to smokers, showed the same result with only found in males, not in females, suggesting that the
several associations with the high frequencies. effect of smoking was sex-dependent. In our dataset, we
The effect of smoking on hearing loss has given rise have tested for effect modification by gender, but did
to considerable controversy. In a study of CVD, CV risk not observe a significant interaction term in any of the
factors, and hearing loss in the Framingham cohort, subsamples, despite the fact that our study has higher
Gates et al. (1993) did not find a significant associa- power. This implies that the effect of smoking was not
tion between smoking (in packyears) and hearing significantly different between males and females.
status. In general, they found hearing loss to be Although we cannot rule out that the difference in
strongly associated with CVD events (coronary heart gender effect between the study of Rosenhall et al. and
disease, heart attack, stroke, or intermittent claudica- our study can be attributed to the difference in age
tion), whereas no significant associations with the risk range between the two studies, reanalysis of our data
factors for CVD events (such as smoking) were found. seems to indicate that it was merely because of the
They concluded that CVD events themselves, rather statistical analysis technique used by Rosenhall et al.
than their triggers (e.g., smoking), lead to hearing They performed a split analysis, testing males and
loss. Hence, they claimed that previously reported females separately for association, whereas in our study,
associations between smoking and hearing loss were we have performed a single analysis on the entire
the result of a confounding effect: smokers had on sample set, including gender as a covariate. It is
average a higher incidence of CVD events, and these interesting to note that by performing a split analysis
events lead to hearing loss, resulting in a spurious on our data and reanalysing the association between
association between smoking and hearing loss. Zhigh and packyears in males and females separately, the
As a consequence of our recruitment strategy— results were similar to the findings of Rosenhall et al. In
asking for healthy people and choosing a relatively three of the subsamples, we found a significant effect in
young age range—the prevalence of CVD in the ARHI males but not in females. The most likely explanation
population is much lower than in the Framingham for the different results between the two statistical
heart study. On the other hand, our dataset contains a analysis techniques is that the nonsignificant results
higher number of smokers. Therefore, as our study for women in the split analysis were because of a
had included a relatively large number of smoking decrease in power. Although all subsamples contained
subjects without CVD events, we have more power to similar numbers of males and females and similar
detect effects of smoking alone with less confounding numbers of smokers and nonsmokers, the proportion
by CVD events. This enabled us to disentangle the of smokers is quite different between the sexes: in all
effects of smoking and CVD events on hearing. subsamples, approximately one third of the females
FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment 275

ever smoked compared to two thirds in males. Thus, gestational age with and without obesity or height catch-up
the analysis of women may have lacked power to detect growth: A prospective longitudinal register study on birth size in
245,000 Swedish conscripts. J. Clin. Endocrinol. Metab 90:4452–
small effects in the separate subsamples. In line with 4456, 2005.
this, the joint analysis with higher sample numbers BRANT LJ, GORDON-SALANT S, PEARSON JD, KLEIN LL, MORRELL CH,
resulted in a highly significant association in both males METTER EJ, FOZARD JL. Risk factors related to age-associated
and females. Hence, including both males and females hearing loss in the speech frequencies. J. Am. Acad. Audiol
into a single analysis, accounting for gender, should be 7:152–160, 1996.
BURR H, LUND SP, SPERLING BB, KRISTENSEN TS, POULSEN OM. Smoking
robust to confounding effects of gender and has more and height as risk factors for prevalence and 5-year incidence of
power to detect small effects than a split analysis. hearing loss. A questionnaire-based follow-up study of employees
Previous reports have shown an association between in Denmark aged 18–59years exposed and unexposed to noise.
hearing impairment and chronic alcohol abuse Int. J. Audiol 44:531–539, 2005.
CHOBANIAN AV, BAKRIS GL, BLACK HR, CUSHMAN WC, GREEN LA, IZZO
(Rosenhall et al. 1993). On the other hand, a protective
JL, JR, JONES DW, MATERSON BJ, OPARIL S, WRIGHT JT, JR, ROCCELLA
effect of moderate alcohol consumption on hearing EJ. The Seventh Report of the Joint National Committee on
had been noted before in some, but not all, studies Prevention, Detection, Evaluation, and Treatment of High
(Brant et al. 1996; Popelka et al. 2000; Itoh et al. 2001; Blood Pressure: The JNC 7 report. JAMA 289:2560–2572, 2003.
Helzner et al. 2005). Our study could confirm this CLARK WW. Noise exposure from leisure activities: A review. J.
protective effect. Although our analyses only reached Acoust. Soc. Am 90:175–181, 1991.
CRUICKSHANKS KJ, KLEIN R, KLEIN BE, WILEY TL, NONDAHL DM, TWEED
significance in 3/9 subsamples, the direction of the TS. Cigarette smoking and hearing loss: The epidemiology of
association was consistent across all subsamples, in hearing loss study. JAMA 279:1715–1719, 1998.
both the high and the low frequencies. It has been DAVIS A. Prevalence of hearing impairment. In: Davis A (ed)
suggested before that the cardioprotective effect of Hearing in Adults. London, Whurr, pp. 43–321, 1994.
DESTEFANO AL, GATES GA, HEARD-COSTA N, MYERS RH, BALDWIN CT.
moderate alcohol intake extends to a decreased risk
Genomewide linkage analysis to presbycusis in the Framingham
of hearing loss. We also tested whether the effect of Heart Study. Arch. Otolaryngol. Head Neck Surg 129:285–289, 2003.
alcohol consumption was gender dependent, as effect DOBIE RA. Estimating noise-induced permanent threshold shift from
modification by gender had been noted in previous audiometric shape: The ISO-1999 model. Ear Hear 26:630–635,
studies (Helzner et al. 2005). Our joint analysis 2005.
produced no significant effect modification by gender, ENGDAHL B, TAMBS K, BORCHGREVINK HM, HOFFMAN HJ. Screened and
unscreened hearing threshold levels for the adult population:
and analysis of the separate subsamples did not show a Results from the Nord-Trondelag Hearing Loss Study. Int. J.
consistent effect modification. Audiol 44:213–230, 2005.
Hearing loss has always been considered an inevi- FERRITE S, SANTANA V. Joint effects of smoking, noise exposure and
table part of aging, but more and more studies seem age on hearing loss. Occup. Med. (Lond) 55:48–53, 2005.
to indicate that this is not necessarily true. Apparently, FRANSEN E, VAN LAER L, LEMKENS N, CAETHOVEN G, FLOTHMANN K,
GOVAERTS P, VAN DE HEYNING P, VAN CAMP G. A novel Z-score-based
a healthy lifestyle can be beneficial for hearing method to analyze candidate genes for age-related hearing
conservation at higher ages. It is commonly known impairment. Ear Hear 25:133–141, 2004.
that a healthy lifestyle protects against CVD. In this FUENTE A, MCPHERSON B. Central auditory processing effects induced
study, we have observed a remarkable parallelism by solvent exposure. Int. J. Occup. Med. Environ. Health
20:271–279, 2007.
between the risk factors for ARHI and those for
GARRINGER HJ, PANKRATZ ND, NICHOLS WC, REED T. Hearing
CVD with similar roles for smoking, high BMI, and impairment susceptibility in elderly men and the DFNA18 locus.
regular moderate alcohol consumption. Arch. Otolaryngol. Head Neck Surg 132:506–510, 2006.
GATES GA, COBB JL, D’AGOSTINO RB, WOLF PA. The relation of
hearing in the elderly to the presence of cardiovascular disease
ACKNOWLEDGEMENTS and cardiovascular risk factors. Arch. Otolaryngol. Head Neck
Surg 119:156–161, 1993.
GATES GA, COUROPMITREE NN, MYERS RH. Genetic associations in age-
The authors would like to express their sincerest gratitude related hearing thresholds. Arch. Otolaryngol. Head Neck Surg
toward all the volunteers who participated in this study. This 125:654–659, 1999.
work was supported by grants from the European Commu- GATES GA, SCHMID P, KUJAWA SG, NAM B, D’AGOSTINO R. Longitudinal
nity (5th Framework project QLRT-2001-00331), the British threshold changes in older men with audiometric notches. Hear.
Royal National Institute for Deaf and Hard of Hearing Res 141:220–228, 2000.
Individuals (RNID), the Flemish Organization for Scientific HELZNER EP, CAULEY JA, PRATT SR, WISNIEWSKI SR, ZMUDA JM, TALBOTT
Research (FWO-Vlaanderen), and of the University of EO, DE REKENEIRE N, HARRIS TB, RUBIN SM, SIMONSICK EM, TYLAVSKY
Antwerp (TOP project). FA, NEWMAN AB. Race and sex differences in age-related hearing
loss: The Health, Aging and Body Composition Study. J. Am.
Geriatr. Soc 53:2119–2127, 2005.
INTERNATIONAL ORGANISATION OF STANDARDIZATION (1984). International
REFERENCES Standard ISO 7029. Acoustics—Threshold of hearing by air conduc-
tion as a function of age and sex for otologically normal persons.
BARRENAS ML, JONSSON B, TUVEMO T, HELLSTROM PA, LUNDGREN M. ITOH A, NAKASHIMA T, ARAO H, WAKAI K, TAMAKOSHI A, KAWAMURA T,
High risk of sensorineural hearing loss in men born small for OHNO Y. Smoking and drinking habits as risk factors for hearing
276 FRANSEN ET AL.: Environmental Risk Factors for Age-related Hearing Impairment

loss in the elderly: Epidemiological study of subjects undergoing STEWART M, KONKLE DF, SIMPSON TH. The effect of recreational
routine health checks in Aichi, Japan. Public Health 115:192– gunfire noise on hearing in workers exposed to occupational
196, 2001. noise. Ear Nose Throat J 80:32–34, 2001, 36, 38–40.
KARLSSON KK, HARRIS JR, SVARTENGREN M. Description and primary TORRE P, 3RD, CRUICKSHANKS KJ, KLEIN BE, KLEIN R, NONDAHL DM. The
results from an audiometric study of male twins. Ear Hear association between cardiovascular disease and cochlear func-
18:114–120, 1997. tion in older adults. J. Speech Lang. Hear Res 48:473–481, 2005.
KUJAWA SG, LIBERMAN MC. Acceleration of age-related hearing loss by UCHIDA Y, NAKASHIMAT T, ANDO F, NIINO N, SHIMOKATA H. Is there a
early noise exposure: Evidence of a misspent youth. J. Neurosci relevant effect of noise and smoking on hearing? A population-
26:2115–2123, 2006. based aging study. Int. J. Audiol 44:86–91, 2005.
KURIEN M, THOMAS K, BHANU TS. Hearing threshold in patients with ÜNAL M, TAMER L, DOGRUER ZN, YILDIRIM H, VAYISOGLU Y, CAMDEVIREN
diabetes mellitus. J. Laryngol. Otol 103:164–168, 1989. H. N-acetyltransferase 2 gene polymorphism and presbycusis.
LEE FS, MATTHEWS LJ, DUBNO JR, MILLS JH. Longitudinal study of Laryngoscope 115:2238–2241, 2005.
pure-tone thresholds in older persons. Ear Hear 26:1–11, 2005. VAN EYKEN E, VAN LAER L, FRANSEN E, TOPSAKAL V, LEMKENS N, LAUREYS
LUTMAN ME, DAVIS AC. The distribution of hearing threshold levels W, NELISSEN N, VANDEVELDE A, WIENKER T, VAN DE HEYNING P, VAN
in the general population aged 18–30years. Audiology 33:327– CAMP G. KCNQ4: A gene for age-related hearing impairment?
350, 1994. Human Mutat 27:1007–1016, 2006.
LUTMAN ME, SPENCER HS. Occupational noise and demographic VAN EYKEN E, VAN CAMP G, FRANSEN E, TOPSAKAL V, HENDRICKX JJ, DEMEESTER
factors in hearing. Acta Otolaryngol. Suppl 476:74–84, 1990. K, VAN DE HEYNING P, MAKI-TORKKO E, HANNULA S, SORRI M, JENSEN M,
NOMURA K, NAKAO M, YANO E. Hearing loss associated with smoking PARVING A, BILLE M, BAUR M, PFISTER M, BONACONSA A, MAZZOLI M,
and occupational noise exposure in a Japanese metal working ORZAN E, ESPESO A, STEPHENS D, VERBRUGGEN K, HUYGHE J, DHOOGHE I,
company. Int. Arch. Occup. Environ. Health 78:178–184, 2005. HUYGEN P, KREMER H, CREMERS C, KUNST S, MANNINEN M, PYYKKO I, DIAZ-
POPELKA MM, CRUICKSHANKS KJ, WILEY TL, TWEED TS, KLEIN BE, KLEIN LACAVA A, STEFFENS M, WIENKER T, VAN LAER L. The N-acetyltransferase
R, NONDAHL DM. Moderate alcohol consumption and hearing 2 polymorphism NAT2*6A is a causative factor for age related
loss: A protective effect. J. Am. Geriatr. Soc 48:1273–1278, 2000. hearing impairment. J. Med. Genet 44:570–578, 2007.
ROBINSON DW. Threshold of hearing as a function of age and sex for VAN LAER L, VAN EYKEN E, FRANSEN E, HUYGHE JR, TOPSAKAL V,
the typical unscreened population. Br. J. Audiol 22:5–20, 1988. HENDRICKX JJ, HANNULA S, MAKI-TORKKO E, JENSEN M, DEMEESTER
ROSENHALL U, SIXT E, SUNDH V, SVANBORG A. Correlations between K, BAUR M, BONACONSA A, MAZZOLI M, ESPESO A, VERBRUGGEN K,
presbyacusis and extrinsic noxious factors. Audiology 32:234– HUYGHE J, HUYGEN P, KUNST S, MANNINEN M, KONINGS A, DIAZ-
243, 1993. LACAVA AN, STEFFENS M, WIENKER T, PYYKKO I, CREMERS CW, KREMER
SLIWINSKA-KOWALSKA M, ZAMYSLOWSKA-SZMYTKE E, SZYMCZAK W, KOTYLO P, H, DHOOGE I, STEPHENS D, ORZAN E, PFISTER M, BILLE M, PARVING A,
FISZER M, WESOLOWSKI W, PAWLACZYK-LUSZCZYNSKA M. Ototoxic SORRI M, VAN DE HEYNING PH, VAN CAMP G. The grainyhead like 2
effects of occupational exposure to styrene and co-exposure to gene (GRHL2), alias TFCP2L3, is associated with age-related
styrene and noise. J. Occup. Environ. Med 45:15–24, 2003. hearing impairment. Hum. Mol. Genet 17:159–169, 2008.
SLIWINSKA-KOWALSKA M, ZAMYSLOWSKA-SZMYTKE E, SZYMCZAK W, KOTYLO P, VAUGHAN N, JAMES K, MCDERMOTT D, GRIEST S, FAUSTI S. A 5-year
FISZER M, WESOLOWSKI W, PAWLACZYK-LUSZCZYNSKA M, BAK M, GAJDA- prospective study of diabetes and hearing loss in a veteran
SZADKOWSKA A. Effects of coexposure to noise and mixture of population. Otol. Neurotol 27:37–43, 2006.
organic solvents on hearing in dockyard workers. J. Occup. WILD DC, BREWSTER MJ, BANERJEE AR. Noise-induced hearing loss is
Environ. Med 46:30–38, 2004. exacerbated by long-term smoking. Clin. Otolaryngol 30:517–520, 2005.