Article Text

Original article
Exposure to chlorinated solvents and lung cancer: results of the ICARE study
  1. Francesca Mattei1,2,
  2. Florence Guida1,2,
  3. Mireille Matrat3,4,
  4. Sylvie Cenée1,2,
  5. Diane Cyr5,6,
  6. Marie Sanchez1,2,
  7. Loredana Radoi7,8,
  8. Gwenn Menvielle9,10,
  9. Fatima Jellouli7,11,
  10. Matthieu Carton5,6,
  11. Simona Bara12,
  12. Emilie Marrer13,
  13. Danièle Luce14,15,
  14. Isabelle Stücker1,2
  1. 1INSERM, UMRS 1018, CESP Centre for Research in Epidemiology and Population Health, Environmental Epidemiology of Cancer, Villejuif, France
  2. 2University of Paris Sud 11, UMRS 1018, Villejuif, France
  3. 3INSERM, U955, Créteil Cedex, France
  4. 4Centre Hospitalier Intercommunal, Créteil Cedex, France
  5. 5INSERM, Epidemiologic Cohorts Unit—UMS 011 INSERM-UVSQ, Villejuif, France
  6. 6University of Versailles St-Quentin, UMS 011, Villejuif, France
  7. 7INSERM, UMRS 1018, CESP Centre for Research in Epidemiology and Population Health, Epidemiology of Occupational and Social Determinants of Health, Villejuif, France
  8. 8Faculty of Dental Surgery, University Paris Descartes, Paris, France
  9. 9INSERM, UMR_S 1136, Pierre Louis Institute of Epidemiology and Public Health, Villejuif, France
  10. 10Sorbonne University, UPMC University of Paris 06, UMR_S 1136, Pierre Louis Institute of Epidemiology and Public Health, Villejuif, France
  11. 11University of Versailles St-Quentin, UMRS 1018, Villejuif, France
  12. 12Registre des Cancers de la Manche, Cherbourg-Octeville, France
  13. 13Registre des Cancers du Haut-Rhin, Mulhouse, France
  14. 14INSERM, U 1085_IRSET, Campus de Fouillole—BP 145, Pointe-à-Pitre, Guadeloupe, French West Indies
  15. 15University of Rennes 1, Campus de Fouillole—BP 145, Pointe-à-Pitre, Guadeloupe, French West Indies
  1. Correspondence to Dr Isabelle Stücker, CESP Centre for Research in Epidemiology and Population Health, 16 avenue Paul Vaillant Couturier, Villejuif, Cedex 94807, France; isabelle.stucker{at}inserm.fr

Abstract

Objective To investigate the role of occupational exposure to chlorinated solvents in lung cancer aetiology.

Methods ICARE (Investigation of occupational and environmental CAuses of REspiratory cancers) is a French, multicentre, population-based, case–control study. Information on the lifelong work history of 2926 cases and 3555 controls was collected using standardised questionnaires. Occupational exposures were assessed using job-exposure matrices for five chlorinated solvents. Solvents were studied separately and in combinations. ORs were computed using unconditional logistic regression models adjusted for classic risk factors, including a history of cigarette smoking and exposure to asbestos. Adjustment for socioeconomic status (SES) was also made.

Results After adjustment for exposure to asbestos, we observed a positive, statistically significant association with lung cancer for men and women exposed to a combination of perchloroethylene (PCE), trichloroethylene and dichloromethane (DCM). Further adjustment for SES slightly decreased this association. In contrast, no statistically significant associations were found for other solvent combinations.

Conclusions These results suggest that exposure to PCE may constitute a risk factor for lung cancer, especially among women, who seem to have a higher prevalence of exposure than men.

  • Lung Cancer
  • Chlorinated Solvents
  • Occupational Exposure
  • Case-Control Study

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What this paper adds

  • In 2012, IARC classified trichloroethylene as carcinogenic for humans (group 1), but no epidemiological evidence has shown an association with lung cancer.

  • Only one study has investigated the role of exposure to chlorinated solvents in lung cancer and this suggested that exposure to perchloroethylene (PCE) and carbon tetrachloride may increase the risk of lung cancer.

  • Our findings suggest that the exposure to PCE may be a risk factor for lung cancer.

  • Unlike other solvents, exposure to PCE may also be more frequent in some sectors, such as dry-cleaning for women.

  • No positive association has been found for other chlorinated solvents.

  • These results contribute to improving prevention of lung cancer.

Introduction

In the 1950s, chlorinated solvents were used extensively in industry, especially in sectors where degreasing was required. During the 1970s, regulations were introduced, leading to reduction of their use.1 Today, chlorinated solvents are still used in a variety of work places and industries because of their good solvency and low flammability.2

Trichloroethylene (TCE) is one of the most used chlorinated solvents. In 1990, 95% of TCE production in Europe was for metal cleaning or degreasing. In 2012, the International Agency for Research on Cancer (IARC) classified TCE as carcinogenic for humans (group 1) because of convincing evidence for an association with renal-cell carcinoma.3–6 Less consistent epidemiological evidence was found for associations with liver cancer and non-Hodgkin's lymphoma (NHL).4

Perchloroethylene (PCE), another widely used chlorinated solvent, especially in the dry-cleaning sector, was classified as a probable carcinogen (group 2A) in 2012.4 Epidemiological evidence across studies showed consistent patterns of association between PCE and bladder cancer, and some positive associations have also been reported for cervical, kidney, oesophageal cancer and NHL.4 ,7 A critical review summarising the results of various cohort studies, mainly focusing on the dry-cleaning sector, showed weak associations with lung cancer.7 Subsequently, a cohort study among dry-cleaners and laundry workers showed an increased incidence of lung cancer, but no data on smoking habits were given.8 To our knowledge, only one case–control study from Canada has investigated the role of chlorinated solvent exposure in lung cancer. That study suggested that exposure to PCE and carbon tetrachloride (CT) may increase the risk of lung cancer.9

Among the other chlorinated solvents, dichloromethane (DCM), CT and chloroform (CF) are considered to be possibly carcinogenic (group 2B).10 ,11

A large case–control study, ICARE (Investigation of occupational and environmental CAuses of REspiratory cancers), was conducted in France to investigate the association between main occupational exposures and respiratory tract cancers (lung and upper aerodigestive tract). Using ICARE's data, we focused in our study on the risk of lung cancer for subjects occupationally exposed to chlorinated solvents.

Methods

Study design and population

The ICARE study is a French multicentre, population-based, case–control study, conducted between 2001 and 2007.12 It was set up in 10 administrative departments, including a cancer registry. Case recruitment was performed in collaboration with the French network of cancer registries (FRANCIM) in almost all of the healthcare establishments in each department. All new patients with histologically confirmed lung cancer,13 aged 18–75 years, who were diagnosed during the study period were eligible for the study. Population-based controls were selected by incidence density sampling. In each department, controls were frequency-matched to cases by gender and age (<40, 40–54, 55–64, ≥65 years). A further stratification was performed to make the socioeconomic distribution comparable to that of the general population living in the departments.

Of the 4865 eligible cases identified, 489 (10%) could not be located, 781 (16%) died before any contact and 238 (5%) could not be interviewed because of their health status. Among the 3357 remaining subjects, 2926 (87%) agreed to participate. Of the 4673 eligible controls, 4411 (94%) were contacted and 3555 (81%) agreed to participate. The study thus included 6481 subjects: 2926 cases and 3555 controls. More details about the subjects’ selection are available elsewhere.14–18 Ethics approval was obtained from the institutional review board of the French National Institute of Health and Medical Research (IRB-Inserm, No 01-036 and CNIL No 90120).

Data collection

Data were collected by trained interviewers during face-to-face interviews using standardised questionnaires. Information included sociodemographic characteristics, smoking history and lifetime occupational history (covering all jobs held for at least 1 month). For each job, we collected information about the company, the subjects’ tasks and specific exposures of interest. TCE was the only chlorinated solvent specifically listed, but subjects could report other agents not specified; PCE was one of the agents that was self-reported.

Occupations and branches of industry were coded blinded to case–control status according to the International Standard Classification of Occupation (ISCO)19 and the French Nomenclature of Activities (NAF).20

A short-form questionnaire, which was used for proxy respondents and subjects too sick or too tired (5% of men and 3% of women), was used to collect mainly smoking data and occupational history, but did not include detailed questions on each job held.

Exposure assessment

Occupational exposure to TCE, DCM, PCE, CF, CT and at least one chlorinated solvent was assessed using job-exposure matrices (JEMs).21 For each combination of ISCO and NAF codes, JEM assigned three indices of exposure: (i) a probability of exposure expressed as the percentage of exposed workers (categorising into <1, 1–10, 11–20, 21–30 or up to 91–100%) (ii) an intensity of exposure (<5, 5–25, 26–50, 51–100, >100 ppm for PCE, TCE and DCM or not exposed, very low, low, medium and strong for CT and CF); and (iii) a frequency of exposure (<1, 1–10, 11–20, 21–30, up to 91–100% of working time). For exposure to at least one solvent, the frequency and intensity of exposure were not assessed separately and the JEM provided an average level of exposure during a usual working day. To account for changes in exposure over time, indices were provided for different calendar periods from 1947 to 2007. Exposure information of 1947 was used for the proportion of jobs held before this date (3.8%).

In all analyses, subjects never exposed to any chlorinated solvents were used as the reference group. The following exposure variables were computed: ever exposed to a specific solvent (subjects having at least one job with probability of exposure greater than zero); maximum probability of exposure; maximum frequency of exposure; maximum intensity of exposure; cumulative duration; and Cumulative Exposure Index (CEI). For exposure to at least one solvent, the level of exposure was used rather than the frequency and intensity of exposure. CEI was obtained by summing the product of the exposure probability, frequency, intensity and duration of each job period, over the entire work history, using the median value of the classes. The CEIs were transformed into categorical variables according to deciles of the distribution among the controls.

Because exposure to one solvent did not preclude exposure to the others, subjects were categorised into mutually exclusive exposure groups according to various combinations of specific solvents. Only groups with the largest number of subjects were investigated. Distributions of jobs with the highest prevalence of exposure were shown graphically.

We estimated lifelong exposure prevalence to chlorinated solvents by weighting the number of subjects exposed in each class of maximum probability by the median values of the classes (ie, 0.05 for subjects with maximum probability between 1% and 10%; 0.15 for 10–20%, etc). Non-exposed subjects were then recalculated accordingly.

Statistical analysis

Multivariable unconditional logistic regression models were applied to estimate ORs of lung cancer and their corresponding 95% CIs for all exposure variables. Analyses were stratified by gender and systematically adjusted for age at interview (<50, 50–60, 61–70, >70 years), departments, number of jobs held (1, 2, 3, 4, >5) and cigarette smoking history, which was summarised using the Comprehensive Smoking Index (CSI).22 This index combines duration of smoking, time since cessation and intensity. It is zero in never smokers. Previous analyses have reported that the CSI varies linearly with lung cancer risk; thus, it was included as a continuous variable.14 ,16 ,18

The CEI of each solvent was categorised according to the median, tertiles or quartiles of the distribution of the controls.

Additional adjustments were made for exposure to asbestos (CEI in quartiles) and socioeconomic status (SES), assessed by the longest held occupation, according to the standard classification of the French National Institute of Statistics (INSEE). Exposure to asbestos had been previously assessed using a specific JEM.14 ,18 When not specified, ORs reported in result section are adjusted for age at interview, department, CSI and number of jobs held during working life.

Sensitivity analyses were performed among subjects never exposed to asbestos (1828 men and 1096 women) and separately for subjects who answered the short-form (262 men, 54 women) or complete questionnaire (4794 men, 1371 women). Additionally, we repeated the analyses using lag periods of 5 and 10 years between exposure and cancer diagnosis.

We also analysed the association between exposure to TCE and PCE and lung cancer according to subjects’ answers in the questionnaire.

Dose–response relationships were tested with maximum likelihood estimates based on the categorical variables, after testing the linearity of the relationship. Multinomial logistic regression models were applied to investigate the main histological types of lung cancer.

Statistical analyses were performed using SAS software (SAS Institute Inc; North Carolina, USA; V.9.3). All p values were two-sided and a p value ≤0.05 was the threshold for statistical significance.

Results

The main sociodemographic characteristics of the study population are reported in table 1. Male cases were slightly older than male controls, while the opposite was found among women owing to the unique control group for the two groups of cancer cases (lung and upper aerodigestive tract). An increased risk of lung cancer was found among men and women who had spent the longest period of their working life as self-employed workers, office and sales employees and blue-collar workers.

Table 1

Sociodemographic characteristics among cases and controls and distribution of histological subtypes of lung cancer cases, separately for men and women, France, 2001–2007

Among women, more than half of the cases had adenocarcinoma (54%) while in men squamous cell carcinoma and adenocarcinoma were found in similar proportion (35%).

The prevalence of lifetime exposure to chlorinated solvents among controls was 8.5% for men and 2.1% for women. TCE was the solvent with the highest prevalence of exposure (7.6% for men and 1.1% for women). Among men it was followed by DCM (1.1%), PCE (0.3%), CF (0.2%) and CT (0.1%) and among women by PCE (0.9%), DCM (0.3%), CF (0.1%) and CT (0.1%).

Table 2 shows the ORs of lung cancer for exposure to each chlorinated solvent (ever exposure and CEI). The results for maximum probability, maximum intensity and cumulative duration of exposure are reported in online supplementary material (eTable 1). Among both men and women, ORs for ever exposure to at least one solvent showed an increased risk of lung cancer (OR=1.42, 95% CI 1.23 to 1.64; OR=1.52, 95% CI 1.03 to 2.23, respectively). Among men, positive associations emerged between lung cancer and ever exposure to TCE, DCM and PCE and for some of the levels of CEI, intensity and duration of exposure. However, these associations were no longer found after adjustment for exposure to asbestos, except for the association with PCE, whose ever exposed OR was borderline significant (OR=1.41, 95% CI 0.98 to 2.05). Adjustment for SES further slightly decreased the ORs, leaving PCE as the only solvent with an OR for ever exposed >1.00, although not statistically significant. Among women, borderline significant positive associations were found for subjects exposed to TCE, PCE and DCM. Although adjustment for occupational exposure to asbestos and SES slightly decreased the ORs, dose–response relationships were still found. No statistically significant associations emerged for CF or CT exposure, but some ORs were >1.00. Analyses restricted to subjects never exposed to asbestos were based on numbers too small to obtain precise estimates, except for TCE, where no association was found for men or women.

Table 2

Association between lung cancer and exposure to five chlorinated solvents, separately for men and women, France

Results according to exposure to particular combinations of chlorinated solvents are presented in table 3. TCE was the only solvent we could isolate for a sufficient number of subjects (see online supplementary eTable 4). Among men, we found no association with lung cancer after adjustment for exposure to asbestos, a result that was also confirmed in men never exposed to asbestos (40 cases and 69 controls exposed; OR=1.01, 95% CI 0.62 to 1.67), even for those with CEI levels higher than average (16 cases and 19 controls exposed; OR=1.27, 95% CI 0.59 to 2.92). Among women, we observed a weak association with minimal confounding by asbestos exposure. However, no association was found among women never exposed to asbestos (17 cases and 16 controls exposed; OR=0.88, 95% CI 0.40 to 1.96).

Table 3

Association between lung cancer and exclusive exposure to combination of chlorinated solvents, separately for men and women, France

Because almost all subjects exposed to DCM were also exposed to TCE, the results for exposure to TCE and DCM, among men, reflected those for exposure to TCE only. Among women, no associations were found.

Indications for a positive association were found for men and women exposed to TCE and DCM and PCE (OR=1.81, 95% CI 1.18 to 2.76; OR=5.39, 95% CI 1.37 to 21.16, respectively). Adjustment for asbestos exposure and SES reduced the associations; however, they were still significant in women and borderline significant in men. No association was found for men or women exposed to the combination of DCM and CT and CF.

Figure 1 compares the distribution of jobs in which men and women are exposed to (a) TCE, (b) and PCE. We have shown only those jobs where at least 5% of exposed workers were employed. Men exposed to TCE were mainly employed in jobs related to metal works, known to be characterised also by the presence of asbestos—for example, they were machinery fitters and assemblers, plumbers, welders, blacksmiths and toolmakers. In contrast, women's jobs were distributed across several job categories in addition to the metals sector, such as laundry and dry cleaning and the shoe and leather industry. For exposure to PCE, most jobs were in laundry and dry-cleaning for women and printing for men, both unlikely to entail exposure to asbestos.

Figure 1

Comparison of the distribution of men and women in the most common occupations with exposure to (a) trichloroethylene and (b) perchloroethylene; according to International Standard Classification of Occupation (ISCO) Classification, France.

Because of the sample size, analyses stratified by complete or short-form questionnaire were achievable only for the group of subjects who answered the complete questionnaire, and the results remained unchanged.

When 5- and 10-years lag periods were taken into account, the exposure status was modified for a small number of subjects and the results of statistical analyses remained unchanged (data not shown).

Self-reported results for exposure to TCE and PCE are presented in table 4. After adjustment for exposure to asbestos, exposure to TCE was not associated with lung cancer risk in men or women. Although very few subjects reported exposure to PCE, findings suggested increased lung cancer risk and the strength of the association was similar for men and women (OR=3.25, 95% CI 1.23 to 8.59; OR=3.12, 95% CI 0.50 to 19.28, respectively) and was not affected by adjusting asbestos exposure.

Table 4

Association between lung cancer and exposure to trichloroethylene and perchloroethylene, according to subjects’ declaration in the generic questionnaire, separately for men and women, France

Analyses by histological subtypes were only conducted for exposure to TCE, PCE and at least one solvent. All exposure indices were studied. The results reflected those reported in table 2 but were less precise (see online supplementary eTables 2 and 3). No specific relationship with any histology was found.

Discussion

This analysis was conducted using data from one of the largest population-based, case–control studies investigating occupational risk factors in respiratory cancer, the ICARE study.

Misclassification inevitably occurs when retrospectively assessing occupational exposures. Using specific JEMs based on subjects’ lifelong occupational history allows assessment of exposure to be made automatically and reproducibly.23 Additionally, occupational coding was assigned blinded to case and control status, which also prevents differential misclassification.24 However, non-differential misclassification could result in an average bias toward the null. Previous findings using asbestos or silica dust JEMs on the ICARE dataset were consistent with the literature,18 and allowed us to be confident that the coding of jobs was accurate. Furthermore, the prevalence of lifelong exposure to at least one solvent estimated among controls was similar to that found by the National Institute of Health Surveillance (InVS; 8.5% vs 9.3% for men; 2.1% vs 1.7% for women).1

When solvents were investigated separately, the results showed weak associations often confounded by asbestos, except for PCE. To overcome the limitation of overlap between solvent exposures, analyses were performed on mutually exclusive exposures. TCE was the most prevalent chlorinated solvent used among men and women. More than 40% of men had a probability of exposure to TCE greater than zero and among them 75% were never exposed to any other type of solvent. Analyses performed on subjects exposed solely to TCE (see online supplementary eTable 4) showed associations with lung cancer that were no longer found after adjustment for asbestos exposure, even for the highest class of each index of exposure. To determine whether these results might be due to overadjustment for asbestos exposure, we restricted the analysis to subjects never exposed to asbestos and again found no association with TCE in men or women. However, while it is better to analyse the role of a substance by isolating subjects exposed to that substance alone, we found that these subjects had the lowest mean CEI for exposure to TCE compared with those exposed to other combinations of solvents (ie, 98.9 (SD=129.7), 71.2 (SD=108.7), 27.0 (SD=30.3) for men exposed to TCE, PCE and DCM; to TCE and PCE; and to TCE only; and 70.1 (SD=84.8), 42.7 (SD=105.3), 15.24 (SD=24.5) for women, respectively). For this reason, we investigated the associations for men with CEI for exposure to TCE >99 (ppm×years) and >70 (ppm×years) for women, finding again, after adjustment for asbestos, no association with lung cancer (see online supplementary eTable 4). Finally, no evidence of differential effects of tumour histology emerged from multinomial logistic regression analyses. The number of subjects was large enough to detect a minimum OR of 1.3, considering a lifelong exposure prevalence of 9% and statistical power of 80%, suggesting that failure to observe an association between lung cancer and TCE is not due to low statistical power. Increased incidence of lung cancer in association with TCE occurred only in studies of animals.25 Some evidence has emerged about the potential TCE carcinogenicity in human populations for the kidney,3 ,26–28 liver29 and NHL.30 ,31 The 2012 IARC evaluation of TCE carcinogenicity concluded that there was convincing evidence of a positive association with kidney cancer and limited evidence for an association with liver cancer and NHL; thus, the substance was classified into group 1.4 Our findings do not suggest any association between TCE and lung cancer.

Because it was not possible to isolate sufficient subjects exposed to DCM only, it is difficult to identify its role. Eighty-seven per cent of the subjects exposed to DCM had also been exposed to TCE, and the results of analyses on subjects who had ever been exposed to DCM in at least one job also exactly reflected the results for exposure to TCE.

Only five subjects had been exposed to PCE exclusively and therefore the effects could not be studied. Fifty-one subjects had been exposed to PCE and TCE only and 153 to PCE, TCE and DCM only. For these combinations, we found positive associations with lung cancer both among men and women, which remained after adjustment for asbestos exposure, although statistical significance was reached (with a wide CI) only for women. However, because we grouped TCE, PCE and DCM together it is more difficult to identify which of the three is responsible for the observed association. No association was found for exposure to TCE or exposure to TCE and DCM, leading us to believe that PCE is probably responsible for the observed increased risk of lung cancer.

We found that the associations were slightly stronger among women than among men. However, before reaching any conclusion, it is important to keep in mind that the number of women exposed was much smaller than that of men, leading to wide CIs. In addition, asbestos was not a confounding factor for the association between lung cancer and PCE exposure in women, whereas asbestos adjustment among men slightly decreased the observed associations.

The analysis based on self-reported exposure to specific agents in the questionnaire confirmed the association between lung cancer and exposure to PCE, despite the reduced sample size, while the analysis of self-reported TCE again did not show any association. Furthermore, it is interesting to note that the magnitude of the association was similar in men and women. It might be suspected that cases tend to over-report their occupational exposure in comparison with controls, but we believe that the lack of information and knowledge on the effects of exposure to PCE on lung cancer exclude this possibility almost entirely. The percentage of controls declaring exposure to PCE reflected the lifelong exposure prevalence expected according to InVS estimation. This is probably because only those subjects who were used to using this substance knew that they were exposed to it, making this self-report fairly reliable.

Various cancer sites associated with exposure to PCE have been identified in the literature—for example, bladder, oesophagus, kidney, cervix and NHL, but only the findings for bladder cancer were consistent across studies.4 ,7 IARC classified PCE as carcinogenic to animals with sufficient evidence and possibly carcinogenic to humans (group 2A).4 The lung was not considered a likely target organ.32 Interestingly, one study recently published also suggested an association with lung cancer risk for ‘substantial’ levels of exposure to PCE, although the number of subjects was small.9 Because PCE is used extensively in the dry-cleaning sector, several cohort studies were conducted on subjects employed in dry-cleaning as a proxy for PCE exposure,8 ,33 ,34 and reported a moderately increased risk of lung cancer. However, according to a critical review on occupational exposure to PCE and cancer risk, the controlling of confounding factors, such as smoking habits and asbestos exposure, was often inadequate.7 In our study, we also found that PCE is a solvent typically found in sectors such as dry-cleaning or printing. However, we were unable to assess the effects of exclusive exposure to PCE because almost all subjects were simultaneously exposed to other solvents.

Adjustment for SES in occupational cancer studies is debated because SES is strongly associated with occupational risk factors.35 ,36 According to the rules proposed by Richiardi et al,35 based on directed acyclic graphs, all results not adjusted and adjusted for SES are presented in the tables. SES is a complex index that comprises different elements—for example, education, diet, smoking habits and jobs, and consequently also occupational exposures including asbestos. While we had previously carefully adjusted our results for smoking habits using the CSI index, which accounts for the three main smoking characteristics that affect lung cancer risk (ie, intensity, duration and time since quitting) and for asbestos exposure, using a lifelong CEI, all the associations decreased after adjustment for SES. This indicates that the possibility of confounding by SES cannot be excluded even after adjustment for smoking habits and occupational exposure to asbestos. Adjustment for SES should therefore be considered. On the other hand, the complex link between SES, exposure to an agent widely used in the working environment such as asbestos and the job may lead to overadjustment when SES is considered. For this reason, results with and without adjustment have been presented.

Overall, adjustment for SES affected the analyses among women less than those among men. The association between TCE, PCE and DCM and lung cancer was clearly visible in women (model not adjusted for SES: OR=5.16, 95% CI 1.30 to 20.54; model adjusted for SES: OR=4.57, 95% CI 1.14 to 18.34). Among men, the association decreased from 1.45 (95% CI 0.94 to 2.24) to 1.28 (95% CI 0.83 to 1.98). Although this modest result might be explained by residual confounding among men, this is unlikely among women given the strength of the association. Overall, these two independent results suggest that PCE alone or in combination with other solvents may increase the risk of lung cancer.

The large number of analyses raises concern about multiple testing. This could also apply to the multiple hypotheses explored in the ICARE study. Whether or not adjustment for multiple comparisons is required in epidemiological studies has been extensively debated, although no obvious solution has been found.37 We preferred to focus on the consistency of the results rather than their statistical significance. For exposure to PCE, we observed indications for an association with lung cancer (i) for ever exposure, as assessed by a JEM and (ii) for self-reported exposure; (iii) we identified a dose–response relationship with cumulative duration of exposure and (iv) when investigating exposure to a combination of substances, the group of solvents including PCE almost always resulted in association. On the other hand, although we followed the same statistical analysis strategy, we almost never found that exposure to TCE was associated with lung cancer, despite all the analyses performed.

In conclusion, our findings suggest that the exposure to PCE may be a risk factor for lung cancer. Exposure to PCE was less common than exposure to TCE and was more typical of sectors such as dry-cleaning for women and printing for men. Nevertheless, further investigations are necessary to replicate these results in a larger exposed population.

Acknowledgments

The authors thank Ms Joëlle Fevotte for designing occupational questionnaires and all members of the MatGéné working group from the Institut de Veille Sanitaire and, in particular,r Ms Brigitte Dananché for providing job exposure matrices.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Contributors IS and DL are the co-principal investigators of the ICARE (Investigation of occupational and environmental CAuses of REspiratory cancers) Study. They designed the study, directed its implementation and oversaw all aspect of the study, including recruitment of patients and controls, funding and quality control of data. FM carried out the statistical analyses, interpreted the results and wrote the manuscript. FG contributed to the statistical analysis and interpretation of the results. SC, DC, MS managed the data, applied job exposure matrices and prepared datasets for statistical analyses. MM, LR, GM, FJ, MC conceived the variables included in the analysis and the strategy of the analysis. They participated in writing the manuscript. SB and EM are responsible for two cancer registries and coded the histology of the lung cancer cases.

  • Funding This analysis was supported by La Fondation ARC pour la Recherche sur le Cancer. The ICARE study was funded by French National Research Agency (ANR); French National Cancer Institute (INCA); French Agency for Food, Environmental and Occupational Health and Safety (ANSES); French Institute for Public Health Surveillance (InVS); Fondation pour la Recherche Médicale (FRM); Fondation de France; La Fondation ARC pour la Recherche sur le Cancer; Ministry of Labour (Direction Générale du Travail); Ministry of Health (Direction Générale de la Santé).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Institutional review board of the French National Institute of Health and Medical Research (IRB-Inserm, No 01-036 and CNIL No 90120).

  • Provenance and peer review Not commissioned; externally peer reviewed.