Occupational Noise Exposure and Vestibular Schwannoma: A Case-Control Study in Sweden Free

Abstract

It has been suggested that the association between self-reported occupational noise exposure and vestibular schwannoma (VS), found in several studies, represents recall bias. Therefore, we aimed to study the relationship in a large case-control study using occupational noise measurements. We performed a case-control study using data from Sweden for 1,913 VS cases diagnosed in 1961–2009 and 9,566 age- and sex-matched population controls. We defined occupational history by linkage to national censuses from 1960, 1970, 1980, and 1990. We estimated occupational noise exposure for each case and control using a job-exposure matrix. There was no association between occupational noise exposure and VS. Among subjects assessed as ever exposed to occupational noise levels of ≥85 dB (214 cases and 1,142 controls), the odds ratio for VS per 5 years of exposure was 1.02 (95% confidence interval: 0.90, 1.17). Workers with noise levels of ≥85 dB for at least 15 years (5-year latency period), showed no increased risk of VS (odds ratio = 0.98, 95% confidence interval: 0.73, 1.31) compared with those who had never been exposed to noise levels of 75 dB or higher. In summary, our large study does not support an association between occupational noise exposure and VS.

Abbreviations

 
• CI

  confidence interval

 
• JEM

  job-exposure matrix

 
• NOCCA

  Nordic Occupational Cancer Study

 
• VS

  vestibular schwannoma

Vestibular schwannoma (VS) is a benign tumor that arises from the nerve sheath of the vestibular branch of the vestibulocochlear nerve. The reported incidence rate has increased, probably due to increased longevity and availability of magnetic resonance imaging (1). A Danish study from 2014 showed a crude rate of 30.7 diagnosed cases per million inhabitants per year (2). More than 90% of the tumors are unilateral and sporadic (3), whereas bilateral tumors are pathognomonic for neurofibromatosis type 2 (4). Both types are associated with inactivation of the tumor suppressor neurofibromatosis type 2 gene (5, 6).

The role of environmental factors in the etiology of sporadic VS is not well defined (7). As to occupational noise, which is one of the most common occupational hazardous exposures (8), systematic reviews find that studies are inconsistent (9–11). While studies based on self-reported exposure data show an association (7, 12–16), 2 of the 3 studies using occupational noise measurements to develop a job-exposure matrix (JEM) (14, 15, 17) do not show such a link (15, 17). It has been suggested that the association in self-report studies represents recall bias; hearing problems in VS cases might influence awareness and reporting of noise exposure, while underreporting of exposure might be common among the controls. On the other hand, the 2 JEM-based studies with no indication of an association (15, 17) included only 30 and 67 VS cases, respectively, who were assigned a noise level of ≥85 dB, which is associated with an increased risk of inner ear damage and noise-induced hearing loss (18). Therefore it has been questioned whether the null findings could be due to few cases with high enough exposures.

To assess the link between occupational noise exposure and VS, there is a need for the present large study in which the exposure is classified objectively independent of the case-control status.

METHODS

Study design and population

We used a case-control design developed within the Nordic Occupational Cancer Study (NOCCA) (19). In the present study we had access only to the Swedish individual records. There were 2,536 VS cases diagnosed between 1961 and 2009 and registered in the nationwide Swedish Cancer Registry (International Classification of Diseases, Seventh Revision, code 1930 and histopathological code 451). Among cases, the date of tumor diagnosis was defined as the “index date.” We selected 5 age- and sex-matched controls for each case from the Swedish records of those who were alive and without a registered VS diagnosis in the year of case diagnosis (12,680 controls), resulting in 15,216 subjects. We excluded subjects born before 1920 (n = 3,618); we had information from national censuses in 1960 and later, and they would be older than 40 years at the time of the first occupational information. We assumed that possible occupational exposure would start at approximately age 20 years, or in the case of change of occupation, possibly later. The cohort included participants born before 1960.

Of the remaining 11,598 subjects, we excluded people who were too young at the index date to have been sufficiently noise-exposed to be at risk of VS. Using a 5-year latency period, subjects who were under 25 years of age at index date (n = 119) were excluded from the analyses. For the 10-year latency period, we excluded 390 subjects. Accordingly, the final sample for the 5-year latency analyses included 11,479 subjects, and for the 10-year latency analyses 11,208 subjects.

Occupational noise exposure

We used a Swedish noise JEM developed by Sjöström et al. (20), a group of experienced occupational hygienists. The JEM contains 321 occupations with information regarding occupational noise from 1970–2004. The occupational noise information derives from measurements collected from different workplaces using stationary and person-applied dosimeters. There were several measurements for each occupational group. For each occupation, the estimated 8-hour average (time-weighted average) noise level in decibels (dB(A)) for every 5-year period was coded as either <75 dB(A), 75–84 dB(A), or ≥85 dB(A). The validity of the JEM was tested in the study, using nonparametric methods, based on classification consensus between 2 separate groups of occupational hygienists. The validation results showed 80% agreement and no systematic differences in classification between the 2 different groups of occupational hygienists classifying the occupational noise exposure (20).

Occupational codes.

The Swedish VS data were linked to occupational census data in the NOCCA cohort data set, in which occupations were registered with Population and Housing Census 1980 (FoB80) occupational codes (21). In the JEM, however, the estimated decibel levels were assigned to the Swedish Nordic Standard Classification (NYK83) occupational codes (22) based on the International Standard Classification of Occupations (ISCO)-58. This required the FoB80 codes to be translated into NYK83 codes. Most codes overlapped 1:1 between FoB80 and NYK83, except for 55 FoB80 codes in the NOCCA data set, which were translated into several NYK83 codes in the JEM. To select the most likely translation of a FoB80 code, the number of persons working in each of the corresponding NYK83 codes were obtained from the 1990 Swedish census data. The NYK83 code with the highest proportion of workers was selected for the FoB80 code.

Duration of noise exposure.

We considered 20–65 years of age as the working age range. Occupational codes were registered in censuses in 1960, 1970, 1980, and 1990. For individuals who lacked a census registration, no exposure was applied for that census time period. Individuals with different occupational codes in different census records were assumed to have changed occupation in the middle of the period between the census years. The occupational code registered in the 1960 census was applied for the estimated decibel levels from 1940–1964, using the decibel estimate for noise measured between 1970 and 1974; the occupational code of the 1970 census was applied for decibel levels from 1965–1974, also using the decibel estimate for 1970–1974; the occupational code of the 1980 census was applied for decibel levels from 1975–1984, using the decibel estimates for 1975–1979 for the 5-year period 1975–1979 and the decibel estimate for 1980–1984 for the 5-year period 1980–1984; and finally, the occupational code of the 1990 census was applied for decibel levels from 1985–2004, using the decibel estimates for 1985 until 2004.

Reference group.

The unexposed category (6,757 and 6,715 subjects for the 5- and 10-year latency assumptions, respectively) included participants who had never been exposed to occupational noise of ≥75 dB before the index date, as well as those whose first noise exposure of ≥75 dB occurred during the latency period (112 and 249 subjects for the 5- and 10-year latency periods, respectively), similar to the practice in the study by Edwards et al. (17).

Statistical analyses

We used conditional logistic regression, using Stata, version 15.1 (StataCorp LLC, College Station, Texas), to estimate the association between occupational noise exposure and VS. We defined 5 noise-exposure categories based on various combinations of levels and durations of exposure (ever ≥75 dB; ≥75 dB for ≥10 or ≥15 years; and ≥85 dB for ≥10 or ≥15 years). Each exposure category was compared with the same reference group (never exposed to ≥75 dB)—that is, we considered 5 different dichotomous exposure variables. We used a 95% confidence interval.

Among subjects assessed with an exposure level of ≥85 dB, we also estimated the association between number of years with noise exposure (per 5 years, continuously scored) and VS. In all analyses, we evaluated both a 5-year and 10-year latency period. We performed all analyses stratified by sex. Because the occupational code registered in the 1960 census was applied for the estimated decibel levels from 1940–1964, we performed sensitivity analyses with exclusion of the subcohort born before 1935.

RESULTS

Table 1 shows the descriptive statistics of the final study population. The sample for the analyses using a 5-year latency period included 1,913 cases and 9,566 controls. There were 5,355 men and 6,124 women. Mean age at index date was 54 years (range 25–82 years) and mean birth year was 1938 (range 1920–1960). Mean duration of occupational noise exposure of ≥75 dB before the 5-year latency period was 7.5 (standard deviation, 11.4; range 0–45) years.

Table 2 shows the results from the regression analyses in which we estimated the association between the defined noise exposure categories and VS. There was no association between any exposure category and VS. Even the highest-exposure group, which was assessed with occupational noise levels of 85 dB or higher for at least 15 years (81 cases and 427 controls, using the 5-year latency period), showed no increased risk of VS, compared with those who had never been exposed to occupational noise levels of 75 dB or higher (odds ratio = 0.98, 95% confidence interval (CI): 0.73, 1.31).

Table 3 shows the results from the regression analyses among subjects assessed with an exposure level of ≥85 dB (1,356 and 1,282 subjects using a 5-year and 10-year latency period, respectively). Among these subjects, we estimated the association between number of years with noise exposure (continuously scored) and VS. The odds ratios for VS per 5 years of noise exposure were 1.02 (95% CI: 0.90, 1.17) and 1.01 (95% CI: 0.87, 1.17), using a 5-year and a 10-year latency period, respectively. Noise exposure was more common among men than among women, resulting in few exposed cases, wide confidence intervals, and less precise results for women. According to the point estimate, the risk of VS for women increased by 37% per 5 additional years of exposure level of ≥85 dB (odds ratio = 1.37, 95% CI: 0.45, 4.13); however, the confidence interval was very wide. The results from the sensitivity analyses, excluding the subcohort born before 1935, showed no statistically significant association between occupational noise and VS.

DISCUSSION

Our large case-control study, using JEM-based noise exposure information, showed no association between occupational noise exposure and VS, even for prolonged high-intensity noise exposure.

A strength of our study is its large sample size, the cohort design and the objective noise information, in which exposure was classified independently of case-control status (JEM based on occupational noise measurements). The JEM has been validated and showed approximately 80% agreement in classification of occupational noise exposure, without systematic differences between 2 groups of occupational hygienists (20). The authors concluded that the JEM gives a good general estimate for the occupational noise levels in Sweden for different occupations during 1970–2004. Nevertheless, it cannot be excluded that our JEM-based study might be subject to nondifferential misclassification due to broad job codes, which might have led to underestimation of the effect estimates. We did not have information on the use of hearing protection, and so we could not adjust for this. Edwards et al. (12) showed an association between self-reported occupational noise and VS—but not among noise-exposed workers who used hearing protection most of the time. In earlier decades, most workers did not use hearing protection; thus, we believe that information on use of hearing protection would not have influenced the results.

The previously observed link between occupational noise and VS might also reflect detection bias. In people who are under medical surveillance due to occupational noise exposure, the detection of unilateral hearing loss might be followed by MRI and (potentially) a VS diagnosis, which would overestimate the results. It has also been argued that the relationship between self-reported occupational noise exposure and VS (7, 12, 13, 15, 16) might represent recall bias. Hearing problems, such as tinnitus or hearing loss, in VS cases and among noise-exposed individuals might influence awareness and reporting of noise exposure and would be expected to inflate odds ratios. Similarly, underreporting by controls, who might be less aware of noise exposure, would also tend to inflate odds ratios for all exposure metrics. The latter is supported by validation studies, which describe cases being more likely than controls to report their noise exposure similar to the JEM or the expert evaluation, while there was underreporting by controls (23). We assessed information on noise exposure objectively, based on data registered before the diagnosis of the VS, to avoid the risk of recall bias.

Recall bias might also explain the previously reported association between self-reported leisure noise and VS (12, 13, 15). Our study did not consider nonoccupational noise exposure. If a true association between noise and VS exists, then nonoccupational noise would contribute to the risk burden and to a possible misclassification in the present study. However, in the general population, occupational noise exposure is considered to have a higher impact on hearing compared with leisure noise (24).

Of the 3 studies using occupational noise measurements to develop a JEM (14, 15, 17), only one showed a relationship between occupational noise and VS (14). Preston-Martin et al. (14) used a JEM based on job titles, as well as individual job description and information about the employer; noise exposure was determined by a blinded review of job histories and linkage to the US National Occupational Hazard Survey. Responses to interviews were compared for 86 VS patients and 86 neighborhood controls. During the period 10 or more years before the year of case diagnosis, more cases than controls had a job involving exposure to extremely loud noise (odds ratio = 2.2, 95% CI: 1.12, 4.67) (14).

Like prior studies of noise exposure and VS, we also used a latency period, assuming that noise exposure during this period before the index date has no effect on tumor development. The results for 5- and 10-year latency periods were similar: no association between occupational noise and VS. With our 1,913 cases, of which 214 were assigned noise levels of ≥85 dB (5-year latency period), our study adds more information than the 2 prior, negative JEM-based studies (15, 17), which included only 30 and 67 cases, respectively, who were assigned noise levels of ≥85 dB (5-year latency period).

In summary, our large study, using JEM-based noise exposure based on information collected and registered before the index date, does not support an association between occupational noise exposure and VS. Currently, case-control studies using self-reported exposure have shown correlations, while 3 of 4 studies using registry-based noise information show no correlations. In the future, a prospective, longitudinal study with complete occupational history combined with better exposure assessment could make contributions to causal inference.

ACKNOWLEDGMENTS

Author affiliations: National Institute of Occupational Health, Department of Occupational Medicine and Epidemiology, Oslo, Norway (Lisa Aarhus, Øivind Skare, Ingrid Sivesind Mehlum); Cancer Registry of Norway, Institute of Population-Based Cancer Research, Oslo, Norway (Kristina Kjærheim, Jan Ivar Martinsen); Finnish Cancer Registry, Institute for Statistical and Epidemiological Cancer Research, Helsinki, Finland (Sanna Heikkinen, Eero Pukkala); Faculty of Social Sciences, Tampere University, Tampere, Finland (Eero Pukkala); Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden (Jenny Selander, Mattias Sjöström); Boston College, Chestnut Hill, Massachusetts (Kurt Straif); and ISGlobal, Barcelona, Spain (Kurt Straif).

We thank the Nordic Cancer Union for funding of the Nordic Occupational Cancer Study (NOCCA) project.

Conflict of interest: none declared.

REFERENCES