Occupational exposures and head and neck cancers among Swedish construction workers

Objectives Occupational exposures in the construction industry may increase the risk of head and neck cancers, although the epidemiologic evidence is limited by problems of low study power and inadequate adjustment for tobacco use. In an attempt to address this issue, the relationship between selected occupational exposures and head and neck cancer risk was investigated using data from a large cohort of Swedish construction workers. Methods Altogether 510 squamous cell carcinomas of the head and neck (171 in the oral cavity, 112 in the pharynx, 227 in the larynx) were identified during 1971–2001 among 307 799 male workers in the Swedish construction industry. Exposure to diesel exhaust, asbestos, organic solvents, metal dust, asphalt, wood dust, stone dust, mineral wool, and cement dust was assessed using a semi-quantitative job-exposure matrix. Rate ratios (RR) and 95% confidence intervals (95% CI) were calculated for head and neck cancers in relation to occupational exposure, using Poisson regression with adjustment for age and smoking status. Results Asbestos exposure was related to an increased laryngeal cancer incidence (RR 1.9, 95% CI 1.2–3.1). Excesses of pharyngeal cancer were observed among workers exposed to cement dust (RR 1.9, 95% CI 1.2–3.1). No occupational exposures were associated with oral cavity cancer. These findings did not materially change upon additional adjustment for cigarette pack-years. Conclusions These findings offer further evidence that asbestos increases the risk of laryngeal cancer. The observation of a positive association between cement dust exposure and pharyngeal cancer warrants further investigation.

There is considerable international variation in the incidence of squamous cell carcinomas of the oral cavity, the pharynx, and the larynx ("head and neck cancers") (1). In Sweden, the incidence of these cancers is relatively low. Incidence rates for cancer of the oral cavity and pharynx from 1993 to 1997 (age-standardized to the world population) were 6.78 and 3.19 per 100 000 for Swedish men and women, respectively, while the rates for laryngeal cancer were lower (2.20 per 100 000 for men, 0.29 per 100 000 for women) (1).
Risk factors for head and neck cancers include tobacco and alcohol use (2) and probably low dietary intake of fruits and vegetables (2). Some studies have pointed to occupational risks. Occupational exposure to mists of strong inorganic acids is an established cause of laryngeal cancer (3). Other suspected risk factors for laryngeal cancer include work in the rubber industry (4,5) and exposure to cement dust (6,7) and asbestos (8)(9)(10)(11). There is little evidence of a relationship between occupational exposures and cancers of the oral cavity and pharynx; some chemicals that have been linked to these cancers include polycyclic aromatic hydrocarbons (12) and formaldehyde (13).
Construction workers may be exposed to known or suspected airborne carcinogens, including silica, asbestos, cement dust, solvents and diesel particulates, with potential exposure to the upper aerodigestive tract. Workers in the construction industry have been reported to have elevated rates of head and neck cancers in several studies (13)(14)(15)(16)(17)(18)(19), although not in all investigations (20, Purdue et al 21). However, this epidemiologic evidence is generally limited by the absence of information on possibly confounding exposures. We analyzed data from a large cohort of Swedish construction workers for which smoking data were available to investigate further the relationship between occupational exposures in the construction industry and head and neck cancers.

Study population and methods
The studied cohort has been described earlier (22,23). Briefly, in 1968 the Swedish construction industry started the Organization for Working Environment, Occupational Safety and Health (in Sweden, Bygghälsan), a program to offer nationwide health services to all employees of the Swedish construction industry. As part of this program, regular health examinations were offered to all blue-collar and white-collar workers in the building industry through regular personal invitations (at intervals of 2-3 years) and through visits or advertisements at virtually all major building sites. Although the program was voluntary, approximately 80% of the eligible workers participated at least once. Examinations were conducted at stationary or mobile clinics, typically staffed by a few nurses and a physician. A computer register includes examination data from 389 132 workers participating in Bygghälsan between 1971 and 1993.
A job-exposure matrix was developed to assign exposures (diesel exhaust, asbestos, organic solvents, metal dust, asphalt, wood dust, stone dust, mineral wool, and cement dust) for over 300 job codes in the industry (24,25), based on a survey of occupational exposures carried out by Bygghälsan from 1971 through 1976, where each occupation was studied at visits to approximately five different sites in different geographic regions of Sweden. These assessments were carried out by a team of occupational hygienists and physicians. Only the job title at each worker's first health examination was used, since the information available was insufficient for constructing lifetime occupational histories. Each exposure was graded on an ordinal scale (0-5), with level 3 corresponding to the threshold limit for the exposure in question. We categorized the exposure scales into no exposure (0), moderate exposure (0.5-1), and high exposure (2)(3)(4)(5). Most of the exposed workers had an ordinal score of 1; most of the workers classified as having high exposure had ordinal scores of 2 or 3. Workers were also categorized by smoking status (never, former, current smokers), pack-years of cigarette use [pack-years = (number of years smoked) × (number of packs [ie, 20 cigarettes] smoked per day), calculated as a time-dependent variable] and snuff use (never used, used ≤10 times per day, used >10 times per day, use unknown), based on information collected from the first health examination. When information on smoking and snuff use was not available, information from a later visit was used.
For this analysis, we linked the Bygghälsan computerized register of male participants (96% of all participants) to the Swedish National Cancer Registry through 2001 to identify first primary cancers of the oral cavity [International Classification of Diseases, seventh revision (ICD-7) codes 141-144], pharynx (ICD-7 145, 147, 148), and larynx (ICD-7 161). Less than 0.15% of the cohort members were lost to follow-up. Only cases of histologically determined squamous cell carcinoma (approximately 95% of all head and neck cancers) were included for this analysis. The person-years were computed for each cohort member from the date of the first health examination to the date of the cancer diagnosis, death, emigration or 31 December 2001, whichever occurred first. We used Poisson regression modeling with the software package EPICURE (26) to calculate rate ratios (RR) and 95% confidence intervals (95% CI) in relation to each cancer type to ever-never exposure to each workplace agent, adjusted for attained age (<50, 50-59, 60-69, ≥70 years), and smoking status (nonsmoker, former smoker, current smoker). Additional analyses of occupational exposures by exposure level (no exposure, moderate exposure, high exposure) were also performed, and models adjusting for all occupational exposures simultaneously and for pack-years of smoking were also fit. We tested for multiplicative interaction between occupational exposures and smoking variables (current smoker versus never or former smoker, <10 versus ≥10 pack-years) using the likelihood ratio test, comparing models with and without parameters specifying each interaction of interest.
We repeated all of the analyses with office workers excluded from the cohort to assess whether our findings differed for a more socioeconomically homogeneous study population.
The study was approved by the local committee of ethics at the Umeå University and by the steering committee of the register.

Results
Occupational exposure data were available for 329 665 male workers. After exclusions due to missing information on smoking habits, 307 799 workers (93%) remained eligible for the analysis. The six most common occupations were wood worker (21% of all workers), concrete worker (12% of all workers), electrician (11% of all workers), foreman (9% of all workers), plumber (8% of all workers), and painter (8% of all workers); other occupations were present in the cohort at a frequency of 5% or less. The study population contributed 6 471 211 person-years of observation and 510 incident cases of squamous-cell head and neck cancers to this study. Of these cancers, 171 occurred in the oral cavity, 112 in the pharynx and 227 in the larynx.
After adjustment for age and smoking status (table 1), asbestos exposure was associated with a 50% increase in head and neck cancer, the strongest associations being for laryngeal cancer. Mineral wool exposure was also associated with laryngeal cancer, and cement dust exposure was associated with pharyngeal cancer. No occupational exposures were significantly associated with cancer of the oral cavity. The findings for asbestos, mineral wool, and cement dust did not show an exposure-response relationship although the numbers of cases among the hightly exposed workers were small. These findings did not materially change upon adjustment for age only or upon additional adjustment for pack-years of cigarette use (data not shown).
Generally speaking, the aforementioned associations did not change upon adjustment for all other occupational exposures (data not shown). An exception was found for asbestos and mineral wool exposures, which were highly correlated (Pearson correlation coefficient 0.80). After adjusting one for the other, we found that the laryngeal cancer risk associated with asbestos exposure was only slightly attenuated (RR 1.8, 95% CI 0.8-4.3), while the mineral wool association was no longer present (RR 1.1, 95% CI 0.5-2.3).
When we repeated our analyses excluding office workers from the cohort, our findings did not change (data not shown).

Discussion
In this large cohort of male Swedish construction workers, we found an increased incidence of laryngeal cancer among asbestos-exposed workers and elevated pharyngeal cancer risks among workers exposed to cement dust.

Asbestos
Asbestos is the commercial name for naturally occurring silicate mineral fibers that were earlier added to many products used in the construction industry, including insulation and fire-proofing materials, cement, and wallboard. Asbestos inhalation causes mesothelioma and lung cancer; however, its role in laryngeal cancers is uncertain (8)(9)(10)(11). Animal models are lacking, and epidemiologic evidence is inconsistent. Altogether 14 case-control studies (6 hospital-based, 8 populationbased) have investigated the relationship between asbestos and laryngeal cancer. While nearly all of these studies found odds ratios greater than one (6,12,14,(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37), the association was statistically significant in only three studies (12,36,37). The case-control design is generally well-suited to the study of rare diseases such as laryngeal cancer; however, the low prevalence of occupational asbestos exposure in the general population limits the statistical power of such studies, particularly when the sample size is small. Most of these studies were too small to detect, with reasonable power, a statistically significant association with asbestos of moderate magnitude; of the five studies that included at least 250 cases (6,12,28,36,37), three reported a statistically significant association with asbestos. The findings from cohort studies of asbestos-exposed workers have also given inconsistent results. A pooled analysis of 27 cohort studies yielded a 30% excess risk of laryngeal cancer relative to the general population (meta-standardized mortality ratio 133, 95% CI 114-155) (38). When the authors used the mesothelioma mortality rate of each cohort as a surrogate for the severity of asbestos exposure, no dose-response relationship between exposure severity and laryngeal cancer was observed. However, misdiagnosis of mesothelioma may have attenuated a dose-response effect.
Our findings offer additional evidence supporting an association between asbestos and laryngeal cancer. The strength of our results as evidence of causation is limited by the fact that we did not observe a dose-response relationship with increasing asbestos exposure, although there were too few laryngeal cancers in the high-exposure group (N=2) to address this question adequately.

Cement dust
Cement is a powder composed mainly of lime (CaO, 60-67%), silica (SiO 2 , 17-25%), alumina (Al 2 O 3 , 2-8%), and iron oxide (Fe 2 O 3 , 0-6%) that is combined with water and sand to create concrete and mortar. Cement dust may be carcinogenic due to the presence of hexavalent chromium, an established carcinogen (39), in some cements or due to the lime content, which, upon contact with oral mucosa, induces alkaline conditions potentially leading to the formation of reactive oxygen species (6,40). Increased risks of lung, stomach, bladder, and colon cancer have been reported in some cohort studies of workers involved in the manufacture or handling of cement (41)(42)(43), and an increase in chromosomal aberrations among cement production workers has also been reported (44). In addition, in our present cohort, an increased risk for esophageal adenocarcinoma has been linked to cement exposure (45). Cement dust has also been associated with laryngeal cancer (6,7,28,46) and pharyngeal cancer (47) in case-control studies.
In the past, asbestos was included in the manufacture of some types of cement to increase durability. Given the absence of an association between asbestos and pharyngeal cancer in our data, exposure to asbestos cement is an unlikely explanation for the observed association with cement dust.

Other exposures
We observed an association between mineral wool exposure and the incidence of laryngeal cancer. However, this finding disappeared upon adjustment for asbestos. Others have reported excess risks of laryngeal cancer accompanying exposure to wood dust (34,48), diesel fumes (34), and metal dust (49). In our study no such relationships were identified. Regarding wood dust, it should be mentioned that exposure to dust in our cohort mainly came from pine, and dust from hardwood such as cedar, teak, and the like was very uncommon. Thus a negative result was expected. In addition, in our study, there were too few cases exposed to wood to detect moderate associations with reasonable statistical power.

Strengths and limitations
This is the largest cohort study investigating exposure to asbestos and cement dust and cancers of the head and neck. Furthermore, we were able to control directly for the potentially confounding effects of tobacco use (both snuff use and smoking), the major risk factor for these cancers. Use of a job-exposure matrix developed for this cohort to assess occupational exposure is another advantage, compared with previous studies in which exposure was self-reported or based on job titles alone. Nevertheless, exposure misclassification probably biased our results toward the null. Furthermore, the small numbers of workers with high-level exposure and the lack of information on duration of exposure prevented us from conducting more-detailed analyses. It should be noted, however, that most of the workers remained in the same job; among the workers who participated in repeated examinations, 74% reported the same occupation. We were not able to adjust for alcohol consumption, another established risk factor for head and neck cancer. However, it is unlikely that alcohol use would vary in relation to asbestos and cement dust exposure within this socioeconomically homogeneous cohort.
In conclusion, we identified an increased incidence of laryngeal cancer among workers likely to have been exposed to asbestos; this result represents an important contribution to the ongoing debate as to whether asbestos increases the risk of laryngeal cancer. We also found an increased incidence of pharyngeal cancer accompanying exposure to cement dust. Our finding adds to the accumulating evidence suggesting increased risks of cancers at various sites, including the pharynx, among exposed workers. The relationships between asbestos, cement dust, and head and neck cancer risk warrant further investigation.