Acoustic neuroma is a benign, generally slow-growing tumour, which develops in the acoustic nerve Schwann’s sheath in the inner auditory canal. It accounts for 6–10% of cerebral tumours. There have been several reports of increasing incidence in recent decades.1 2 Despite its localised evolution, acoustic neuroma is a prognostically severe pathology inasmuch as it frequently causes definitive deafness on the affected side, even after surgery. Moreover, without treatment, the tumour will eventually spread to the orifice of the inner auditory meatus in the pontocerebellar angle, where it may compress neighbouring nerve structures. High incidence of acoustic neuroma (or vestibular neuroma) has been reported in several occupational settings (eg, fish hatcheries,3 filling stations and postal sorting offices4). No risk factor, however, has yet been identified, although ionising radiation has been suspected.5 Benzene has also been suggested,6 as have electromagnetic fields emitted by cell phones,7 although results have been contradictory.8^–13 A few clinical case studies have implicated loud noise,14^–16 as did an earlier epidemiological study.6 More recently, reports from the German and Swedish INTERPHONE study groups, again suggested an association between loud noise and acoustic neuroma.17 18 These results were not replicated, however, in a registry-based case-control study in Sweden where a job-exposure matrix was used to characterise noise exposure history.19

The current report is the third from teams of the INTERPHONE study group.17 18 The INTERPHONE study was set up in 1999 as a multinational collaborative study to evaluate the possible association between radiofrequency emissions from cell phones and four types of head tumour (glioma, meningioma, acoustic neuroma and parotid gland tumour). Information was also collected on several other factors, including exposure to loud noise at work or as a result of leisure activities. We present here the results of an investigation of the relationship between exposure to loud noise and risk of neuroma in French subjects.

MATERIAL AND METHODS

Study population

Recruitment was among subjects aged 30–59 years and resident in either Greater Lyon or the Ile-de-France region around Paris. The design and methods of the INTERPHONE study have been described in detail elsewhere.20

Case recruitment

All newly diagnosed cases were actively ascertained between 1 June 2000 and 31 August 2003 from the participating hospital departments (basically, neurosurgery and otorhinolaryngology) in Lyon and Paris. A radiotherapy department in Marseille (France) with a nationwide recruitment for radiation treatment of head tumour also took part in the study. All the relevant departments in the Lyon area participated; in the Paris area some major departments participated throughout the study period, others for only a part of the study period, and other small departments not at all. Thoroughness of recruitment was checked against hospital Medical Information Department records, which further enabled a posteriori recruitment of certain cases (patients who were hard to contact or failed to be contacted in time are nevertheless counted for the purpose of calculating the participation rate). Incident cases were checked for inclusion criteria (age, place of residence, primary tumour), and diagnosis was confirmed histologically in most cases; for certain tumours, managed for example by radiotherapy or under simple surveillance, only the radiological results were available, in which case diagnosis was confirmed medically before inclusion.

Control recruitment

Each case was matched to two controls, for gender, age (5 years) and place of residence at the time of the case’s diagnosis. Ten controls per case were first taken at random from the electoral roll. A letter, accompanied by a consent form and stamped return envelope, was mailed to the first of the 10 controls, explaining that an epidemiological study of environmental radiation exposure was being conducted, but without any specific reference to cell phones or to noise. A reminder was sent out if necessary 1 month later, followed by three attempts at telephone contact at various times of the day (mainly early evening). In case of non-response, refusal or mistaken address, the second of the 10 controls was contacted, and so on. The participation rate does not take account of control candidates for whom we had no telephone number and/or whose mail was returned due to a mistaken address.

The study was approved by the relevant Ethics Committees.

Data collection

Cases and controls were interviewed in person either in hospital, at home or at any other convenient location, or, exceptionally, by telephone. Interviewing began in February 2001 in the Lyon area, and in May 2001 in Ile-de-France. There were 10 interviewers, with prior training on the questionnaire. As far as possible, the same interviewer managed a given case and the matched controls.

Questionnaire

The questionnaire was computer assisted (Computer Assisted Personal Interviewing (CAPI), software in Blaise),21 enabling direct data entry during the interview itself. The questionnaire comprised sections on sociodemographic data, smoking, medical history of subject and family, use of cell phones and other radiocommunication systems, radiation exposure in medical and work settings, and work history. Questions related to past exposure to loud noise were also asked, using a “noise thermometer” (as reference) to assist in the characterisation of volume and type of noise.

Exposure to noise at work

Subjects listed jobs involving noise exposure, with starting and ending dates of the exposure, source(s) and type(s) of noise (continuous, as with a textile loom, intermittent as with a pneumatic drill, or explosive), frequency of exposure (a few hours per month, less than 10 h per week, 10 to 20 h per week, or more than 20 h per week), and whether protection was provided and, if so, how often it was used (never, sometimes, usually).

Exposure to noise in private life

Subjects reported leisure activities that exposed them to noise on a weekly basis over a period of years, specifying whether they had protection during these activities and, if so, how often it was used (never, sometimes, usually), and whether they had lived for at least 10 years in a noisy environment (eg, near an airport, railway or highway junction).

Data analysis

To ensure comparability between case and control exposure estimates, each triplet (case +2 controls) was given a reference date, which was set to the date of diagnosis of the case in each matched set. A second step in analysis took account of onset of neuroma symptoms such as hearing loss or tinnitus, which may develop long before the diagnosis. When cases mentioned such symptoms occurring before the date of diagnosis, a second reference date for the triplet was set to that at which the symptom first occurred on the same side as the tumour. When symptoms were bilateral or contralateral to the tumour, the date of diagnosis served as sole reference. Similarly, the diagnosis date served as reference for cases experiencing no hearing problem, or only after neuroma diagnosis.

Several variables were analysed, considering only exposure at least 1 year before the reference date:

• Noise exposure setting (work and/or leisure and/or environmental).

• Type of leisure exposure (music, shooting, do-it-yourself work, etc) and leisure exposure duration.

• Work exposure: type of noise (continuous, intermittent, or explosive) and duration. Duration was first analysed as total duration in all jobs, then weighted by the percentage weekly exposure time per job.

These variables were treated as both categorical (by control-group quartile) and continuous. The effect of frequency of use of hearing protection was assessed.

The reference category for noise-related variables was that of subjects claiming never to have been exposed, whether at work or at leisure. Subjects claiming to have been exposed but unable to specify the duration were classified in the shortest duration category. Analysis was repeated for exposure durations excluding the 5 years preceding the reference date because acoustic neuroma is a slow-growing type of tumour.

The association between noise exposure and head tumour was estimated by odds ratio (OR) calculated from conditional logistic regression taking account the matching. Adjustment variables were selected by univariate analysis of all variables concerning sociodemographic data, individual and family medical history, radiocommunication system use, medical radiation exposure and cell phone use. Those variables that proved significant at 15% were used in subsequent analyses. The smoking-related variable was adopted regardless of its significance in univariate analysis, in view of the high related cancer risk. A stepwise approach was used to drop variables from the model which were no longer significant in the multivariate analyses. Variables related to noise exposure, whether in work or leisure settings, were introduced separately into the model.

Analyses were undertaken using SAS Version 9.1 software.

RESULTS

Participation and population characteristics

A total of 140 neuroma cases and 290 control subjects were identified in the Paris and Lyon areas. Cases of neuromas were exhaustively reported by all participating departments over their respective periods of participation. Nine patients residing in Paris were recruited via the Marseille radiotherapy department. Participation was 77.1% for neuroma cases. The refusal rate was 13.6% (19 patients). Two patients agreed but could not be interviewed in time, and seven (5%) could not be contacted (address and phone number unknown). Five cases were excluded from analysis: one was a trigeminal neuroma; controls could not be found for the four other cases. Four cases could be matched to only one control subject. Participation for control subjects was 74% (66 refusals, 6 impossible to contact, 3 non-French-speaking). One control interview was lost due to computer problems. The final analysis included 108 neuroma cases and 212 controls. Excluded cases showed no significant difference in gender or age at diagnosis (not shown).

Most interviews (62.0% for cases and 49.5% for controls) were conducted in the subject’s home; 8.3% of the cases were interviewed in the hospital and 14.2% of controls at work. For 6.5% of cases and 9.9% of controls, the interview was by phone. The median interval between diagnosis and interview was 5.6 months, and 4.9 months between case and control interviews. Surgery was performed after the interview in 12% of cases; 21.3% of cases were under simple surveillance. In 56.5% of cases, the tumour was on the left side of the head. Eighty-one per cent of cases presented unilateral hearing loss compared with 16% of controls, and 58% experienced tinnitus compared with 9% of controls. Hearing loss, when reported, was bilateral in only 9.3% of cases compared with 11.8% for controls. Hearing loss was the initial symptom for 31 and tinnitus for six of the cases; 19 reported experiencing both symptoms around the same time.

Neuromas were evenly distributed by gender (49.1% male) and 41.7% of cases were in the executive or professional socioeconomic categories compared with 61.4% of controls (p<0.001; see table 1).

Noise exposure

Exposure to loud noise of whatever source at least 1 year before the reference date (date of symptom onset) was reported by 42.6% of cases and 25.9% of controls (p<0.01) (table 2). After adjusting for socioeconomic status (SES), mean daily cigarette consumption and duration of cell phone use in an urban environment, the risk of neuroma in relation to loud noise was 2.55 (95% CI 1.35 to 4.82). Using diagnostic date of the case as the reference date, the OR did not change much (adjusted OR (adjOR) = 2.48; 95% CI 1.42 to 4.33).Therefore the findings are presented using the date of onset of symptoms as the reference date. Only two cases and four controls were classified as not exposed to loud noise once periods when subjects reported “usually” wearing protection were excluded (adjOR = 2.55; 95% CI 1.35 to 4.81). Having lived more than 10 years in a noisy environment was not associated with any increased risk of neuroma.

Exposure was most often exclusively at work (table 2). Both work and leisure exposure were, however, significantly associated with occurrence of neuroma. Focusing exclusively on those who did not regularly wear protection made little difference to the OR (adjOR for work = 2.26; 95% CI 1.08 to 4.72; adjOR for leisure = 4.94; 95% CI 1.32 to 18.48). Exposure to loud music was more frequent among cases than controls (adjOR = 3.88; 95% CI 1.48 to 10.17). Few subjects reported leisure exposure to other types of noise. By not taking into account the period 5 years before the reference date, the OR for work and leisure noise remained high (respectively adjOR = 2.35; 95% CI 1.11 to 4.99 and adjOR = 4.23; 95% CI 1.22 to 14.71).

Three types of noise exposure at work, continuous, intermittent and explosive, were analysed separately; both continuous and explosive noises were significantly associated with neuroma risk (table 2). The OR for continuous noise restricted to past exposure (at least 5 years previously) was higher (adjOR = 3.96; 95% CI 1.47 to 10.64).

The number of years of exposure to noise at work, whatever the weekly exposure frequency, was not significantly related to neuroma risk, despite a high OR after 18 years’ exposure. However, when the duration was weighted for weekly frequency, the risk proved significant for the most exposed category (adjOR = 3.72; 95% CI 1.45 to 9.59) (table 3). When analysing the time since first exposure at work, a significant excess appeared for the period from 20 to 30 years before the reference date.

Duration of leisure noise exposure for 6 years and more was also significantly associated with increased neuroma risk (table 4), while there was little evidence of a difference in risk by time since exposure (greater or less than 20 years, based on small numbers of cases).

DISCUSSION

The risk of developing neuroma was observed to increase with exposure to loud noise, whether in a work or a leisure setting (particularly high-volume music). The risk was especially apparent in relation to prolonged exposure. In the work setting, explosive and continuous noise appeared particularly linked to elevated neuroma risk.

This study thus confirms previous reports: Preston-Martin found a twofold risk increase in subjects exposed to loud noise at work in the USA6; in a Swedish INTERPHONE study,17 increased acoustic neuroma risk (OR = 1.55) was associated with loud noise exposure at work or in leisure activities. This increase was observed when analysing the exposure to noisy machines or loud music.22 Exposure dating back more than 13 years was associated with a risk greater than 2. Similar results were reported from the German INTERPHONE study.18

The lack of a relationship with time since exposure (for leisure or for work) is difficult to interpret due to the small number of subjects. It is possible that risk may be higher for more recent (less than 20 years) exposure or that subjects who are older — and tend to turn the volume down when they listen to music — remember less well their past exposure to loud noise. Alternatively, neuroma-related hearing loss may pass undetected in the presence of presbyacousis, which is age related.23

We have chosen to present results using as reference date the date of symptom onset as the results did not change much when using the date of diagnosis. As neuromas are slow-growing tumours and as several months or years can separate the first symptom date from the diagnostic date, it is more appropriate to use the onset date, in spite of the fact that this approach is not traditional.

Study limitations

The gender distribution among cases was as expected.2 All case subjects had their medical files and pathology reports systematically checked with respect to inclusion criteria and diagnosis. Contact was attempted with all cases ascertained during the participation period of all hospital departments; the refusal rate among cases was less than 14% (the other cause of non-participation being failed contact due to delayed identification). Participating and non-participating cases did not differ in gender or age. Identification was exhaustive in Lyon, but not in Ile-de-France, where it was estimated at 50% over the study period as a whole. Although departments with a good reputation in neuroma management participated, some of them, including large neurosurgery departments, participated either not at all or only for a part of the recruitment period. However, this probably did not entail any selection bias, since recruitment time and choice of department appear to be unrelated to noise exposure. On the other hand, the incompleteness of recruitment in the Paris region did impair the power of the study. While few cases managed by simple surveillance may also have failed to be identified despite a posteriori checking of medical information records, there is no reason to consider that they would be different in exposure to noise.

The distribution of SES among controls differs from that of the general population, suggesting a possible selection bias. It is noteworthy that the controls for glioma and meningioma cases in the French INTERPHONE study, which were selected in the same fashion, do not show such a difference in SES with the general population. As the researchers recruiting controls from electoral rolls followed a pre-established procedure, it is highly unlikely that their approach should have introduced a bias only for neuroma controls. The difference therefore most likely reflects a difference in participation rates, related to the particular age and sex characteristics of the study subjects. Indeed, when adjusting for age and gender, no significant difference in SES distribution is observed between the three control groups. There was, however, a preponderance of executives among respondent controls (compared with non-respondents), which we took into account by adjusting for this parameter; however, this adjustment could be incomplete. That noise was associated with increased risk from leisure activities (involving music in particular) as well as at work rules out any effect due to socioeconomic selection bias.

Limiting the study to the 30–59-year-old age group, which was chosen to optimise the power of INTERPHONE to study risks related to mobile phone use, did entail loss of power, since acoustic neuroma mainly affects subjects over the age of 45 years and by no means stops at the age of 60 years2; it also deprived the study of information on very long histories of noise exposure but it introduced no selection bias, inasmuch as controls were matched for age and gender.

In terms of control recruitment, the use of electoral rolls impairs exhaustiveness with respect to young subjects; our controls, however, were by definition over 30 years of age, and would therefore tend to be registered on the electoral rolls. For the two major cities (Paris and Lyon), the electoral lists were obtained through computer files which were regularly updated. In the suburban municipalities, we benefited from a complete updating of the lists due to a presidential election at the mid-point of the study period. In spite of this, some randomly selected persons could not be contacted due to a recent change of address or deficient contact details. This may have introduced some selection bias, inasmuch as more mobile subjects could have socioeconomic (and hence noise exposure) specificities.

Our control refusal rate appears lower than in other published studies,9 10 but takes account only of controls for whom initial contact was possible.

Case and control interviews followed the same protocol, but the interviewer was not blind to the subject’s status. This was one of the reasons (the other being cross-country reproducibility) for adopting a closed-question strategy, to minimise variation in question presentation.

Our information regarding noise was based on self-reports, prompted by examples of typical noise levels and types. It might be possible to apply a noise job-exposure matrix to occupation codes, but the most commonly used matrix, FINJEM,24 produced by the Finnish Occupational Health Institute remains incomplete for noise and is not very sensitive.25 This could explain at least part of the lack of association seen in the second Swedish study.19 There is little reason to think that cases and controls differed in their way of assessing past noise exposure, as loud noise is not a commonly perceived risk factor for neuromas, unless tumour-related discomfort affected cases’ perception of past noisy environments. On the other hand, loud noise could be perceived as a risk factor for hearing loss, which is observed in 80% of the cases, and could be responsible for a recall bias, but the ORs are much too high to be attributed exclusively to a recall bias. In addition, one might expect a similar trend in relation to living in a noisy environment with cases reporting more than controls. However, this is not the case.

Symptom onset was used as reference date, since acoustic neuroma is a slow-growing tumour: the interval between initial symptoms and diagnosis can be very long. Initial symptom onset (tinnitus and/or unilateral hearing-loss ipsilateral to the tumour) could be precisely dated by 52% of cases; for the others, the diagnosis date was adopted as default reference, but this did not induce any difference in results.

Biological mechanisms

Chronic trauma of the inner ear due to noise is well known, and could affect both types of hair cells26 27: those of the Corti organ and those of the vestibular labyrinth. Several pathological pathways could be involved. Both balance and auditory receptors share the membranous labyrinth, and the common arterial blood supply of the cochlea and vestibular end organs via the same end artery: noise can damage the stria vascularis of the system and can lead to an intermixing of cochlear fluids by the alteration of the tight cell junction of the reticular lamina.28 On the one hand, these fluids with a high level but different composition of electrolytes bathe the hairy cells of both organs. These electrolytes are very important for the functioning of the neurotransmission to the nervous fibres of the eighth nerve (notably, calcium ion transportation could be highly perturbed in the junction between hair cells and the nervous efferent fibres). A chronic aggression by the electrolytes disequilibrium could be at the origin of the degeneration of the nervous fibre, and could lead to an alteration of the Schwann cell protecting the fibre and helping it to regenerate. On the other hand, an immunoreactivity to inducible nitric oxide synthase has been found29 in the supported and sensory cells of the sensory epithelium as well as in the dark cells area and in vestibular ganglion cells in animals exposed to loud acoustic stimulation, which suggests that free radicals are generated. Those free radicals could be at the origin of DNA damage, not only in cochlear cells but also in other tissue (such as brain), as observed in one experimental study of rodents30 and could be responsible for the development of an acoustic neuroma. The fact that continuous noise is more dangerous for the inner ear than intermittent noise31 32 is supported by our findings where the OR for exposure to continuous noise is higher than for intermittent noise. It is known that the right cochlea is less sensitive to noise than the left cochlea,33 which is coherent with our findings, where neuroma is more often observed in the left side.

Leisure noise, specially from music, is characterised by very high-level sounds34 35 in the lowest frequency bands which could be more dangerous for the vestibule35 than higher frequencies, more prevalent in industrial noises. If the vestibule is more damaged by lower frequencies, it could explain the controversial results concerning the relation between noise at work and neuroma.

CONCLUSION

The present study found an association between loud noise exposure, whether in a leisure (notably in relation to loud music) or a work setting. The association was especially strong in subjects exposed over long periods. Continuous and explosive types of noise were particularly associated with increased risk.