Occupational exposure to chlorinated solvents and risk of head and neck cancer in men: a population-based case-control study in France

Background Few epidemiological studies have investigated the link between occupational exposure to solvents and head and neck cancer risk, and available findings are sparse and inconsistent. The objective of this study was to examine the association between occupational exposure to chlorinated solvents and head and neck cancer risk. Methods We analyzed data from 4637 men (1857 cases and 2780 controls) included in a population-based case-control study, ICARE (France). Occupational exposure to five chlorinated solvents (perchloroethylene [PCE], trichloroethylene [TCE], methylene chloride [MC], chloroform [CF], and carbon tetrachloride [CT]) was assessed through job-exposure matrices. Odds ratios (ORs) and confidence intervals (95% CI) were estimated by unconditional logistic regression, adjusted for age, tobacco smoking, alcohol consumption, asbestos exposure, and other potential confounders. Results We observed no association between chlorinated solvent exposure and head and neck cancer risk, despite a non-significant increase in risk among subjects who had the highest cumulative level of exposure to PCE, (OR = 1.81; 95% CI = 0.68 to 4.82). In subsite analysis, the risk of laryngeal cancer increased with cumulative exposure to PCE (p for trend = 0.04). The OR was 3.86 (95% CI = 1.30 to 11.48) for those exposed to the highest levels of PCE. A non-significant elevated risk of hypopharyngeal cancer was also observed in subjects exposed to the highest levels of MC (OR = 2.36; 95% CI = 0.98 to 5.85). Conclusion Our findings provide evidence that high exposure to PCE increases the risk of laryngeal cancer, and suggest an association between exposure to MC and hypopharyngeal cancer. Exposure to other chlorinated solvents was not associated with the risk of head and neck cancer. Electronic supplementary material The online version of this article (doi:10.1186/s12940-017-0286-5) contains supplementary material, which is available to authorized users.


Background
Until recently, chlorinated solvents have been widely used in practically all branches of modern industry. Their use has since decreased in response to various factors, including increasing knowledge pertaining to their toxicity and environmental impact [1][2][3][4]. Several chlorinated solvents are known or suspected carcinogens. Commonly used for metal cleaning and degreasing, trichloroethylene (TCE) was classified in 2012 as carcinogenic for humans (group 1) by the International Agency for Research on Cancer (IARC), based on sufficient evidence that it causes kidney cancer and limited evidence for liver cancer and non-Hodgkin lymphoma (NHL) [5]. The same year, perchloroethylene (PCE), widely used in the dry cleaning sector, was classified as a probable carcinogen (group 2A), based on limited evidence of an increased risk of bladder cancer [5].
More recently, methylene chloride (MC) was also classified as a probable carcinogen (group 2A) based on limited evidence that it causes biliary tract cancer and NHL [6]. Moreover, chloroform (CF) and carbon tetrachloride (CT) are considered to be possibly carcinogenic (group 2B) [7].
Few epidemiological studies have examined the link between occupational exposure to chlorinated solvents and head and neck cancer risk, and available findings are sparse and inconsistent. Cohort studies could not control for smoking and alcohol use and lacked statistical power to detect moderate increases in risk, because they were often conducted in countries where the incidence of head and neck cancer is low. However, nonsignificant increases in the risk of oral, pharyngeal, and laryngeal cancer were found in several cohorts of workers exposed to TCE or PCE [8,9]. Case-control studies included proper adjustment for confounders, but exposure to solvents was often not well characterized. Exposure to solvents in general has been found to be associated with oral and pharyngeal [10] or laryngeal and hypopharyngeal [11] cancer risk, although others did not find an association [12]. Shangina et al. found a significantly increased risk of laryngeal cancer associated with chlorinated solvent exposure [13] and Vaughan et al. found a non-significant increased risk of oral, pharyngeal, and laryngeal cancer among subjects exposed to PCE [14].
No study has yet examined the association between exposure to several specific chlorinated solvents and head and neck cancer risk, by cancer site. In a previous analysis of our study population by occupation, we observed increased risks of head and neck cancer in dry cleaners and metal workers, suggesting a possible role of exposure to chlorinated solvents [9]. Here, our objective was to investigate the association between exposure to five chlorinated solvents and head and neck cancer risk, using data from a large population-based case-control study, the ICARE study.

Data collection
Trained interviewers conducted face to face interviews using a standardized questionnaire. Data included sociodemographic characteristics, smoking and alcohol consumption history, and a detailed lifetime occupational history, which covered all jobs held for at least one month. For subjects who had difficulty answering because of sickness or tiredness, a shorter version of the questionnaire was used for either in-person interview or to interview a next-of-kin. This shorter version included mainly information on smoking, alcohol consumption, and occupational history and was used for 11% of the cases and 2% of the controls. Occupations and branches of industry were coded by trained coders, blinded to the case-control status of the subjects, according to the International Standard Classification of Occupations (ISCO) [18] and the French Nomenclature of Activities (NAF) [19].

Assessment of occupational exposure
Occupational exposure to TCE, PCE, MC, CF and CT was assessed using job-exposure matrices (JEMs), developed in the context of the Matgéné program [1]. For each combination of ISCO and NAF codes, the JEMs provided three indices of exposure: (i) the probability of exposure expressed as the percentage of exposed workers (categorized into not exposed, 1-10,11-20,21-30, up to 91-100%), (ii) the intensity of exposure (for PCE, TCE and MC: not exposed, 5-25, 26-50, 51-100, >100 ppm; for CT and CF: not exposed, very low, low, medium, and high) and (iii) the frequency of exposure (not exposed, 1-10, 11-20, 21-30, up to 91-100% of working time). In addition, a specific JEM assessed the probability of exposure to at least one of these five chlorinated solvents, together with the average level of exposure during a usual working day. Exposure indices were provided for different calendar periods (1950-1969, 1970-1984, 1985-1994, >1995) to account for changes in working practices and regulation over time.
As an example, in France, for methylene chloride, exposure limits decreased from 100 ppm in 1985 to 50 ppm in 1995, and a general improvement in working conditions was assumed to occur at the beginning of the 1970's. After linking these indices with lifetime occupational history, the following exposure variables were obtained for each subject: 'ever/never' exposed to a specific chlorinated solvent ('ever' defined as having worked in at least one job with a probability of exposure greater than zero), total duration of exposure, and cumulative exposure index (CEI). CEIs were calculated by summing the product of the exposure probability, frequency, intensity, and duration of each job period, over the entire work history, using the central value of the classes. For exposure to at least one solvent, as the JEM provides the average level and the probability of exposure, the CEI was calculated by summing the product of exposure probability, average level and duration. We then categorized these variables. The duration of exposure to solvents was categorized into four classes: 'never exposed' , and three categories according to approximate tertiles of the distribution among exposed controls, 'short' , 'intermediate' , and 'long' exposure. To examine potential effects of the highest exposures to chlorinated solvents, CEIs were categorized as follows: 'never exposed' , and three categories according to the percentiles of the distribution among exposed controls (low: < 50th; medium: 50th-90th; high: > 90th). The CEI categories "medium" and "high" for exposure to CF and CT were combined because of the small number of subjects in these groups.
We estimated the prevalence of lifelong exposure to chlorinated solvents by weighting the number of subjects exposed in each class of maximum probability by the central value of the class. Non-exposed subjects were then recalculated accordingly. Occupational exposure to asbestos was assessed through a specific JEM [20].

Statistical analysis
Multivariable unconditional logistic regression models were used to model associations between chlorinated solvents and head and neck cancer risk. Odds-ratios (ORs) and corresponding 95% confidence intervals (95% CI) were adjusted for previously described covariates. Each solvent was analyzed separately. We also estimated mutually adjusted ORs in a model including all solvents.
Tests for linear trends were performed by modelling the median of each category as a continuous variable.
Additional analyses were performed in separate models adjusting for educational level or occupational class of the longest job held. We also conducted analyses by cancer site (oral cavity, oropharynx, hypopharynx, oral cavity or pharynx not specified, and larynx) and subsites using polytomous logistic regression. Finally, to determine whether joint-exposure to several chlorinated solvents was associated with head and neck cancer risk, we examined the risk associated with combinations of chlorinated solvents with sufficient numbers (at least 10 exposed cases), using "never exposed to any chlorinated solvents" as the reference category.
For each solvent, we also assessed potential interactions with smoking, alcohol drinking and asbestos exposure by including cross-product terms in the models. None of the interactions were statistically significant, and the results were not presented. We also conducted sensitivity analyses excluding subjects with the shorter version of the questionnaire, and the results were similar (data not shown).

Results
The age distribution differed slightly between cases and controls, but the mean age was similar. Cases had a lower level of education than controls and were more often blue collar workers. As expected, cases were more often smokers than controls and had a higher level of alcohol consumption (Table 1).
Concerning lifetime exposure to chlorinated solvents, the most prevalent exposure was to TCE with 8.76% of cases and 7.53% of controls exposed followed by MC with 1.43% of cases and 1.14% of controls exposed. The prevalence of exposure to PCE, CF, and CT was lower with 0.44%, 0.09%, and 0.06% of cases, and 0.28%, 0.16%, and 0.09% of controls, respectively. Exposures to solvents were correlated, and were also correlated with exposure to asbestos (Additional File 1: Fig. S1).
Individual analysis of the solvents ( Table 2) showed no significant association between ever exposure, duration of exposure, or cumulative exposure to PCE, TCE, MC, CF, or CT and head and neck cancer risk. However, we observed a non-significant increased risk among subjects who had the highest cumulative levels of exposure to PCE or MC relative to never exposed subjects (OR = Further adjustment for the level of education or occupational class did not substantially modify the results (see Additional file 1: Tables S2 and S3).
The results remained globally unchanged when we included all solvents in the model ( Table 3).
Analysis of specific head and neck cancer sites (Table 4) showed a significant dose-response relationship between cumulative exposure to PCE and laryngeal cancer risk (p for trend = 0.04), and subjects who had the highest cumulative levels of exposure to PCE had a significantly higher risk of laryngeal cancer (OR = 3.86; 95% CI = [1.30 to 11.48]). The OR was lower than 1 in the medium exposure category, but the confidence intervals around each OR were wide. We used restricted cubic splines (4 knots) to verify the linearity assumption, and no significant departure from linearity was found (p = 0.61 We found no significant association between combinations of chlorinated solvents with at least 10 exposed cases and head and neck cancer risk, or specific subsites (Table 5). However, although we observed no increase in laryngeal cancer risk for subjects who were exposed to only TCE (OR = 0.84; 95% CI [0.63 to 1.12]), or the combination of TCE and MC (OR = 0.95; 95% CI [0.59 to 1.53]), the risk of laryngeal cancer was nonsignificantly elevated for subjects who were exposed to the combination of TCE, PCE, and MC (OR = 1.32; 95% CI [0.68 to 2.55]).
Finally, we investigated the association between PCE exposure and the risk of cancer of subsites of the larynx (glottis/subglottis, supraglottis, and other or nonspecified). The OR associated with the highest cumulative level of exposure to PCE was higher for cancer of the glottis/subglottis (OR = 5.95; 95%CI [1.73 to 20.53]; five exposed cases) than for cancer of the supraglottis (OR = 3.96; 95%CI [0.72 to 21.78]; two exposed cases) or of other/non-specified subsites (OR = 1.76; 95%CI [0.19 to 16.12]; one exposed case), but the ORs were not significantly different.

Discussion
In the present study, exposure to chlorinated solvents was not associated with an overall risk of head and neck cancer. However, high exposure to PCE or MC was associated with higher risks of laryngeal and hypopharyngeal cancers, respectively. These results suggest that these commonly used chlorinated solvents may have a deleterious impact on the upper respiratory tract.
Few epidemiological studies have examined the effects of solvents on head and neck cancers, and even fewer the effects of chlorinated solvents on these cancers.
Previous epidemiological studies have most examine exposure to non-specific solvents. In a case-control study, after adjustment for occupational agents and potential non-occupational confounders, a higher risk of both hypopharyngeal and laryngeal cancer was observed for subjects 'ever exposed' to solvents relative to those who were not exposed [11]. Another study which considered 'all chlorinated solvents' as exposure, also reported an increased risk of laryngeal cancer, although there was no dose-response relationship [13]. To a lesser extent, various other studies have shown a nonsignificant increased risk of head and neck cancers when solvents were considered overall [8,10,21,22]. However, direct comparison with our results is difficult as the solvent classes were not clearly specified in the previous studies. Indeed, few studies have focused on the specific effect of a single chlorinated solvent on the head and neck.
Among the few studies that have investigated TCE exposure, Blair et al. found a standardized mortality ratio close to the null value for oral cavity and pharyngeal cancers [23], whereas Raaschou-Nielsen et al. found a non-significant increase in the incidence of oral cavity, pharyngeal, and laryngeal cancer among blue-collar workers exposed to TCE [24]. A pooled analysis of three Nordic cohort studies, also showed a non-significant increase in the risk of oral, pharyngeal, and laryngeal cancers [25]. Similar findings were obtained in another study, which showed a non-significantly higher hazards ratio for death by cancer of the oral cavity and pharynx among male workers exposed to TCE [26]. Moreover, a non-significant excess of mortality from oral, pharyngeal and laryngeal cancer linked to TCE exposure has also been reported [27]. No information on smoking or alcohol consumption was provided by these studies. In our study, we did not find any association between exposure to TCE and cancer of the oral cavity, pharynx, or larynx after adjustment for smoking, drinking, and exposure to asbestos. An association between exposure to TCE and head and neck cancer risk was reported among women in the ICARE study [17]. In women, confounding by asbestos was minimal and adjustment for asbestos did not modify the estimates related to exposure to TCE and head and neck cancer risk. In contrast, more than 90% of the men exposed to TCE were also exposed to asbestos, and inclusion of asbestos in the models may have led to overadjustment. We repeated the analyses among subjects never exposed to asbestos (479 cases and 1159 controls), and found no association with exposure to TCE (OR = 1.06; 95%CI [0.66 to 1.70]; 55 exposed cases), even if the risk of head and neck cancer increased for those with the highest levels of cumulative exposure (OR = 1.40; 95%CI [0.48 to 4.06]; eight exposed cases). Among men never exposed to asbestos, those exposed to only TCE had a non-significantly elevated risk of head and neck cancer (OR = 1.30; 95%CI [0.74 to 2.27]; 36 exposed cases).
The rare studies that examined PCE exposure in relation with head and neck cancer yielded inconsistent results. Cohort studies reported no increase [28] or a nonsignificant increase in the frequency of head and neck cancers [29,30], but were based on a small number of cases, and were not adjusted for tobacco and alcohol consumptions. A small population based case-control study on dry cleaners that examined PCE exposure showed a higher risk of oral cavity or laryngeal cancers    in subjects exposed to PCE than those who were not, with some evidence of dose-response and durationresponse relationships for laryngeal cancer [14]. Although these results were not significant due to the small number of exposed cases (seven for oral cavity, four for laryngeal cancer), they support our findings, particularly for the association between occupational exposure to PCE and an increased risk of laryngeal cancer. Moreover, the ICARE study has recently shown that exposure to PCE was associated with an increased risk of laryngeal cancer risk in women [17], and an increased risk of lung cancer [16], further supporting a carcinogenic effect of PCE on the respiratory tract. However, there were no studies to suggest mechanisms underlying these effects. Several potential genotoxic and non-genotoxic mechanisms of liver carcinogenesis for perchlorethylene have been identified, but are unlikely to be relevant for respiratory cancer [5]. However, the upper airways are in direct contact with inhaled toxicants, and chronic irritation and inflammation may contribute to the promotion or progression of otherwise initiated lesions. PCE exposure may also facilitate the penetration of other carcinogens in the mucosa, although the lack of statistical interaction in our data between PCE exposure and other risk factors does not support this hypothesis.
Our findings also provide limited evidence of an association between exposure to MC and hypopharyngeal cancer. No other study has specifically examined the risk of head and neck cancer associated with MC exposure [31]. Only two cohort studies of workers exposed to MC reported results for head and neck cancer by subsite, and found either no [32], or a non-significant increased risk for oral cancer [33].
Our study has limitations that may affect the interpretation of the results. First, occupational exposure was retrospectively assessed, and the use of job-exposure matrices generates systematic misclassification, which is likely to be independent of case-control status. Such non differential misclassification could result in an average bias toward the null [34]. Second, the number of exposed cases was small for some chlorinated solvents resulting in large confidence intervals. Investigation of combinations of chlorinated solvents was limited for the same reason and it was not possible to investigate the association between exclusive occupational exposure to PCE and head and neck cancer risk.
Our study also has several strengths. The large number of subjects provided sufficient statistical power to detect moderate associations, and allowed assessment of cancer sites and subsites. Detailed information on lifelong occupational histories was available, allowing us to  assess the level of exposure to chlorinated solvents and investigate dose-response relationships. We adjusted for major confounders, such as tobacco and alcohol consumption, as well as exposure to asbestos. Additional adjustments for educational level, socioeconomic status, and exposure to other chlorinated solvents showed that confounding due to these variables was minimal. Selection bias, although always possible, was probably not an important issue in this study. The distribution of the included cases by age and cancer site was similar to that generally observed in France [35]. The lifetime prevalence of exposure to chlorinated solvents among our controls was close to that observed in a representative sample of the French male population [1].

Conclusions
In summary, our findings suggest that high occupational exposure to PCE may increase the risk of laryngeal cancer. Data from the ICARE study has recently suggested that occupational exposure to PCE also increases the risk of lung cancer [16]. Our results extend these findings by suggesting a likely carcinogenic effect of PCE on other parts of the respiratory tract.

Additional file
Additional file 1: Figure S1. Image plot for Spearman's correlations coefficient among chlorinated solvents and asbestos' CEI. Table S1.
Associations between exposure to chlorinated solvents and head and neck cancer, without adjustment for asbestos exposure. Table S2.
Association between head and neck cancer and exposure to chlorinated solvents, with adjustment for educational level.