Article Text

Original article
Risk of lung cancer associated with six types of chlorinated solvents: results from two case–control studies in Montreal, Canada
  1. David Vizcaya1,
  2. Krista Yorita Christensen1,
  3. Jérôme Lavoué1,
  4. Jack Siemiatycki1–3
  1. 1University of Montreal Hospital Research Center (CRCHUM), Montreal, Quebec, Canada
  2. 2Department of Social and Preventive Medicine, University of Montreal, Montreal, Quebec, Canada
  3. 3Guzzo–Cancer Research Society Chair in Environment and Cancer, School of Public Health, University of Montreal, Montreal, Quebec, Canada
  1. Correspondence to Dr Jack Siemiatycki, Environmental Epidemiology and Population Health Research Group, Research Centre of the University of Montreal Hospital Network (CRCHUM), 3875 rue Saint-Urbain, Montréal, Québec, Canada H2W 1V1; j.siemiatycki{at}


Objectives To determine whether exposure to various chlorinated solvents is associated with lung cancer.

Methods Two case–control studies of occupation and lung cancer were conducted in Montreal, and included 2016 cases and 2001 population controls. Occupational exposure to a large number of agents was evaluated using a combination of subject-reported job history and expert assessment. We examined associations between lung cancer among men and six specific chlorinated solvents and two chemical families (chlorinated alkanes and alkenes). ORs were calculated using unconditional multivariate logistic regression.

Results When the two studies were pooled, there were indications of an increased risk of lung cancer associated with occupational exposure to perchloroethylene (ORany exposure 2.5, 95% CI 1.2 to 5.6; ORsubstantial exposure 2.4, 95% CI 0.8 to 7.7) and to carbon tetrachloride (ORany exposure 1.2, 95% CI 0.8 to 2.1; ORsubstantial exposure 2.5, 95% CI 1.1 to 5.7). No other chlorinated solvents showed both statistically significant associations and dose–response relationships. ORs appeared to be higher among non-smokers. When the lung cancer cases were separated by histological type, there was a suggestion of differential effects by tumour type, but statistical imprecision and multiple testing preclude strong inferences in this regard.

Conclusions There were suggestive, albeit inconsistent, indications that exposure to perchloroethylene and carbon tetrachloride may increase the risk of lung cancer. Results for other solvents were compatible with absence of risk.

  • Lung cancer
  • Carbon tetrachloride
  • Perchloroethylene

Statistics from

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

  • Some chlorinated solvents may be carcinogenic, but the evidence from the limited number of human studies is inconsistent and of variable quality.

  • Evidence concerning lung cancer risks is particularly sparse and inconclusive.

  • We found evidence that perchloroethylene and carbon tetrachloride may possibly be lung carcinogens.

  • While there was no evidence regarding the lung carcinogenicity of other solvents, this lack of association is not definite.

  • These results contribute to understanding of the possible carcinogenicity of these important substances.


Workers in many different occupations and industries have significant exposure to chlorinated solvents.1–4 Such agents have many industrial application and are used as degreasers, paint strippers, dry-cleaning solvents, spot removers, chemical reaction intermediates, aerosol propellants and anaesthetic gas. Chlorinated solvents are short chain hydrocarbons in which hydrogen atoms are substituted by chlorine atoms. They can be further separated into two main classes: chlorinated alkanes, which are saturated molecules, and chlorinated alkenes, which are unsaturated, and in general, more reactive.

The most widely used chlorinated alkenes are trichloroethylene (TCE) and perchloroethylene (PERC), while the most widely used chlorinated alkanes are carbon tetrachloride, methylene chloride, 1,1,1-trichloroethane and chloroform.1 ,2 ,5–7 The frequency of use of these agents has changed over time in response to various factors, including advances in healthcare and awareness of environmental effects.

Experimental evidence suggests that some chlorinated solvents may be carcinogenic.8–12 However, evidence for carcinogenicity in humans is inconclusive, in particular regarding lung cancer risks.10 ,12–17 The limitations of several previous epidemiological studies included weak exposure assessment and poorly controlled potential confounding factors, notably smoking.15–20

We conducted two case–control studies in the Montreal area. Study I was designed to evaluate various cancer sites, including the lung, while study II focussed on lung cancer only. In both studies, information was collected on occupational exposure to six specific chlorinated solvents (PERC, TCE, 1,1,1-trichloroethane, carbon tetrachloride, chloroform and methylene chloride) as well as to chlorinated alkenes and chlorinated alkanes as classes. The present report describes the associations found between these chlorinated solvents and lung cancer. In an earlier report we published selected results from study I.21 The present analyses include data from both studies and also present results for study I based on updated and improved exposure assessments and improved statistical methods.


Study population

Study I was conducted from 1980 to 1986 and included men aged 35–70 years. Study II was conducted from 1995 to 2001 and included both men and women aged 35–75 years. In both studies, cases and controls were restricted to Canadian citizens who were resident in the Montreal metropolitan area. Details of subject ascertainment and data collection for these studies have been presented previously22–25; fieldwork methods, questionnaires and exposure assessment procedures were almost identical between the two studies. Written informed consent was obtained for all participants and ethics approval was obtained from each participating hospital and university.

This was a population-based case–control study. The controls were randomly selected (frequency matched by age and sex) from electoral lists. Study I included 851 male lung cancer cases (79% of eligible cases) and 533 male controls (70% of eligible controls). Study II included 430 female and 735 male lung cancer cases (86% of eligible cases) and 570 female and 898 male controls (70% of eligible controls). Interviews were conducted with proxy respondents if the subject had died or could not otherwise be interviewed. The interview elicited a wide variety of personal data including socio-demographic information and detailed smoking history (age at beginning and at cessation, and intensity and duration of the smoking habit).

Socio-demographic characteristics, the proportions of proxy responses, and smoking profiles of the cases and controls in each study have been detailed previously in table 1 in the report by Pintos et al.26 This information is provided in online supplementary table S1. Next-of-kin proxies responded for about one-third of cases and one-tenth of controls. There were proportionately more French Canadians among cases than among controls in both studies. On average, controls lived in wealthier neighbourhoods and had higher educational attainment than cases. As expected, cases were heavier smokers than controls.

Exposure assessment

The methodology of exposure assessment in study II was similar to that in study I, which has been previously described.21 ,22 ,24 A semi-structured questionnaire was used to obtain details of each job that lasted at least 6 months during the subject's working lifetime. Details included type of employer, nature of the products used or manufactured, subject's tasks and those of nearby workmates, and use of protective devices. For some occupations, job-specific questionnaires were used whenever a worker reported having been in such an occupation. A team of industrial chemists and hygienists (at any given time there were between three and five in the team) examined each subject's questionnaires and translated each job into potential exposures from a list of 294 substances, without knowledge of case or control status. The information provided by the subject was processed by the experts using their collective judgement, taking into account, as far possible, the particular features of each job description. Thus the exposure assessment was based not simply on the worker's occupation and industry, but also on the individual characteristics of the workplace and tasks as reported by the subject or as inferred by the experts from independent knowledge of the company or industry.23

If an exposure was coded, the experts also rated the exposure according to three dimensions: (i) their confidence that the exposure actually occurred (possible, probable and definite); (ii) the frequency of exposure during a normal working week (<5%, 5–30% or >30% of the time); and (iii) the relative concentration of the substance (low, medium or high) with reference to certain benchmark occupations in which the substance is found. The benchmark occupations were sometimes well documented with representative measurements from the industrial hygiene literature, but benchmark occupation concentrations were often estimated by the experts. In general, it proved impossible to reliably estimate absolute concentration values for the levels coded. Non-exposure was interpreted as exposure up to the level found in the general environment. The final exposure codes attributed to a subject were based on consensus among the coders.

There was no evidence that cases provided more complete or more valid job histories than controls, as judged by the numbers of jobs reported per subject and as judged by the interviewers’ subjective ratings of the quality of interviews. Although it is very difficult to establish the validity of retrospective exposure assessments, we have demonstrated satisfactory levels of reliability and validity in the job histories and in the expert exposure assessments.27–31

The agents assessed included the two general families, chlorinated alkanes and chlorinated alkenes, as well as specific chlorinated solvents: carbon tetrachloride, methylene chloride, 1,1,1-trichloroethane, chloroform, TCE and PERC. Assignment of exposure to a family was not always simply the sum of exposures to component agents, as exposure to the family was sometimes assigned without enough information to designate a specific agent.

A previous report on associations between chlorinated solvents and non-pulmonary tumours described the main occupations in which these exposures occurred in study I.32 These profiles were rather similar in study II; the main occupations with these exposures in the two studies combined are shown in online supplementary table S2. Metal-working occupations were the most important sector in terms of numbers exposed to chlorinated solvents overall. Additional occupations with high prevalence of exposure included mechanics and repairers, apparel and furnishings service occupations, nursing occupations, and construction trades occupations.

Statistical analysis

We included only exposures which were coded by the experts with confidence levels of probable or definite. We defined ‘substantial’ exposure to a particular agent as exposure to medium or high concentration levels lasting at least 5 years, for ≥2 h per week. Exposures that did not meet these criteria were defined as ‘not substantial’. If a subject was exposed in two or more jobs, then lifetime values of confidence, frequency and concentration were averaged, weighted by the durations of the various jobs in which exposure occurred.

For the estimation of ORs of an association between lung cancer and each exposure agent, we used unconditional logistic regressions adjusted for age, census median income (percentiles), ethnicity (French vs others), educational attainment (years), questionnaire respondent (self vs proxy), tobacco smoking (Comprehensive Smoking Index, CSI)33 and occupational lung carcinogens. The latter variable was based on exposure to any of eight International Agency for Research on Cancer (IARC)-recognised probable or definite lung carcinogens: asbestos, crystalline silica, chromium VI, arsenic compounds, diesel exhaust emissions, soot, wood dust and benzo(a)pyrene.34 We created a three-category occupational carcinogen variable as follows: never exposed to any carcinogen; ever exposed to any carcinogen but not at a substantial level; ever exposed to any carcinogen at a substantial level. An indicator for the study (I or II) was included in the pooled analyses to adjust for differing study-specific case/control ratios. For each chlorinated solvent evaluation, the reference category consisted of subjects who were never exposed to any of the six chlorinated solvents evaluated or the two generic families of chlorinated hydrocarbons. We also excluded from the reference category those subjects who were ever exposed to vinyl chloride. Although it is not considered a solvent, it belongs to the family of chlorinated alkenes. Thus, for all models, there is a single reference category, which allowed direct comparison between the estimated ORs.

We also conducted separate analyses on the associations between the solvents and each histological lung cancer subtype (squamous cell, small cell or adenocarcinoma). We conducted stratified analyses to evaluate effect modification of the effect of solvents by smoking history. For smoking habit we created two subgroups: ‘Never smokers and light smokers’ (CSI≤10th CSI percentile among ‘Ever smokers’), and ‘Medium and heavy smokers’ (CSI>10th CSI percentile among ‘Ever smokers’). We conducted stratified analyses by interview respondent (self vs proxy) to verify whether proxy respondent results distorted the overall pattern of findings.

Because of limited sample sizes and overlapping exposure to multiple agents, it was not possible to estimate the risks of lung cancer among subsets of subjects who had only ever been exposed to one chlorinated solvent at a time; thus, subjects exposed to any given solvent may also have been exposed to others.


In the pooled analysis of male subjects, 14.4% of the population controls were considered to have been ever exposed to a chlorinated solvent. The analogous exposure prevalence among females was 9.6%. Because there were fewer women than men in the study and because their exposure prevalence was lower, the ability to detect risks was much lower, if not non-existent, among women than among men. None of the estimated ORs among women was statistically suggestive of an association, and the CIs were so wide that the point estimates were quite uninformative. The remaining results and the tables presented here focus on findings among men.

Lifetime exposure prevalence was 4.4% for chlorinated alkenes and 7.9% for chlorinated alkanes. Among the alkenes, more controls had been exposed to TCE (1.4%) than to PERC (0.9%). Among the alkanes, the most prevalent exposure was to carbon tetrachloride (2.8%), followed by methylene chloride (1.9%), 1,1,1-trichloroethane (1.7%) and chloroform (0.6%). In our dataset, the exposure prevalence of most of these agents increased steadily from the 1930s to the 1970s. The main exception was carbon tetrachloride, which peaked and then decreased sharply in prevalence after 1960.

Table 1 shows the adjusted ORs for each exposure agent. We conducted a separate analysis for each study and a pooled analysis of both studies. In the pooled analysis most of the ORs were close to the null, or at least not significantly different from the null. ORs corresponding to exposure at any level to the two families of chlorinated alkenes and chlorinated alkanes were borderline significantly elevated, but in both cases the OR for substantial exposure was close to 1.0. The associations with substantial exposure to PERC (OR 2.4, 95% CI 0.8 to 7.7) and to carbon tetrachloride (OR 2.5, 95% CI 1.1 to 5.7) were the most suggestive of an increase in lung cancer risk.

Table 1

Association between lung cancer among men and exposure to chlorinated solvents in two studies from Montreal

We also analysed ORs for an association between these solvents and each of three histological subtypes of lung cancer (see online supplementary table S3). With smaller numbers of subjects, these subtype analyses were much less stable than those reported in table 1. A few of the many ORs estimated were significantly elevated at the substantial exposure level, namely those for PERC and small cell lung cancer (OR 5.4, 95% CI 1.2 to 26), carbon tetrachloride and squamous cell tumours (OR 3.3, 95% CI 1.4 to 8.1) and TCE and adenocarcinoma (OR 2.7, 95% CI 1.0 to 7.6). Such subgroup analyses which increase the numbers of statistical tests also increase the likelihood of discovering some statistically significant findings by chance.

Estimates of the ORs for an association between lung cancer and chlorinated solvents were generally higher among the ‘Non-smokers and light smokers’ group than among ‘Medium and heavy smokers’ (see online supplementary table S4). For example, the OR for lung cancer associated with any exposure to PERC was 4.5 (95% CI 0.9 to 22) among ‘Non-smokers and light smokers’ and 2.3 (95% CI 0.9 to 5.4) among ‘Medium and heavy smokers’. Similarly, ORs for any exposure to TCE were 4.1 (95% CI 1.1 to 15) and 1.4 (95% CI 0.7 to 3.0), respectively. However, the small numbers among ‘Non-smokers and light smokers’ exposed to chlorinated solvents precluded any strong inference on the possibility of effect modification.

A third series of stratified analyses focused on self-respondents only, and showed that the findings among them were generally similar to those given in table 1.


There were indications of an excess risk of lung cancer among men exposed to PERC (ORany exposure 2.5, 95% CI 1.2 to 5.6; ORsubstantial exposure 2.4, 95% CI 0.8 to 7.7) and to carbon tetrachloride (ORany exposure 1.2, 95% CI 0.8 to 2.1; ORsubstantial exposure 2.5, 95% CI 1.1 to 5.7). When analyses were restricted to non- and light smokers, the point estimates were slightly higher, although the confidence limits were much wider. The other evaluated chemicals and families of chlorinated solvents yielded null or close to null results, but with wide enough CIs that these results in themselves do not demonstrate absence of risk. When the lung cancer cases were separated by histological type, there was some evidence of differential effects by tumour type, but statistical imprecision and multiple testing preclude strong inferences in this regard.

PERC has long been considered a possible carcinogen, but at the time of the 1995 IARC report, most of the evidence for carcinogenicity was derived from animal rather than human studies, and the lung was not considered a likely target organ.6 However, the epidemiological studies available at that time had limited exposure assessment and ability to adjust for the effect of smoking.8 ,19 While not unanimous, several studies published since the 1995 IARC review have shown excess risks of lung cancer among workers purportedly exposed to PERC, with the strongest evidence derived from cohorts of drycleaners.20 ,35–37 Recent reviews by expert panels concluded that the evidence for a link between PERC and lung cancer is suggestive but not yet conclusive.8 ,17 Our results, while statistically imprecise, are compatible with an increased risk of lung cancer related to PERC exposure.

Like PERC, carbon tetrachloride was considered by the 1999 IARC review to be a possible carcinogen, based mainly on animal evidence and with no evidence of lung carcinogenicity.7 ,38–40 Our finding of excess lung cancer risks with carbon tetrachloride, and in particular for squamous cell tumours, is unique in the literature and requires confirmation.

Previous studies have not indicated an excess risk of lung cancer associated with exposure to TCE and methylene chloride.6 ,7 ,41 For methylene chloride, our results showed a null association both overall and across the various subgroup analyses. Regarding TCE, our results were ambiguous. We found a non-significant indication of excess risk among those exposed at any level, especially for adenocarcinomas, but no evidence of excess risk among those with relatively high or long exposure. For the remaining specific solvents (1,1,1-trichloroethane and chloroform) there is little or no previous evidence of lung carcinogenicity, and our findings are compatible with an absence of risk.

Previous reports from our group have discussed a number of strengths of our studies (including relatively high sample size, high participation rates, histological confirmation of diagnoses, intensive expert-based exposure assessment blinded to case/control status, and good control of potential confounding by smoking, socio-economic/demographic variables and other occupational carcinogens) as well as a number of limitations including limited statistical power, multiple inference in a ‘hypothesis-generating’ paradigm, a high proportion of proxy responses, and exposure assessment not based on local measurements and thus subject to error.24 ,25 We will not repeat those discussions here; suffice it to say that these results represent a valuable contribution to the evidence regarding the lung carcinogenicity of chlorinated solvents, but require caution in the interpretation of both null and ‘positive’ results.

In conclusion, our results are compatible with an increased risk of lung cancer in relation to occupational exposure to PERC and carbon tetrachloride. Other chlorinated solvents did not show consistent associations with lung cancer in the pooled analysis and across different smoking subgroups and histological types. While these data contribute important new information, the evidence remains inconclusive on the role of these agents on lung cancer risk.


Exposure assessment methods were expertly developed and implemented mainly by Michel Gérin, Louise Nadon, Ramzan Lakhani, Denis Bégin and Benoit Latreille. Lesley Richardson coordinated the fieldwork. We thank the many research assistants and interviewers who participated, including Marie-Claire Goulet and Jerome Asselin.


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:


  • Contributors DV drafted the manuscript. DV and KY analysed the data. JS designed the study and the analysis plan. JL managed the exposure assessment data. All co-authors contributed to the interpretation of the results and discussion.

  • Funding This study was funded by a number of agencies, including the Fonds de recherche en santé du Québec, the Institut de recherche en santé et sécurité au travail du Québec, the Canadian Institutes for Health Research and the Guzzo-Cancer Research Society Chair in Environment and Cancer.

  • Competing interests None.

  • Ethics approval This study was approved by the ethics committees at the Institut Armand-Frappier (University of Quebec), McGill University and each of the 18 hospitals in which cases were ascertained.

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