Registration Dossier

Administrative data

Description of key information

Additional information

Epidemiology Summary – focus on cancer

There is a large epidemiological dataset available for trichloroethylene and cancer. The studies almost exclusively assess occupational exposure and cancer risk/incidence. The epidemiology data have been reviewed several times over the past decade, including the review within the EU risk assessment, the review by SCOEL, the IARC assessment and the recent review by the EPA in the IRIS toxicology summary. The output of these reviews is that trichloroethylene is likely a human kidney carcinogen. However, several other cancer types have also been potentially associated with trichloroethylene exposure including liver and non-Hodgkin lymphoma (NHL) albeit with less robust associations.

With respect to the epidemiological database, almost without exception, the exposure estimation represents a challenge when interpreting the available studies. The vast majority of studies (including both Case control and Cohort studies) use job exposure matrices, self reported job histories, and attempt to assess exposure quantitatively without the use of actual measured data. As such, it is likely that exposures are not consistently assessed across studies, and there is high potential for misclassification of exposure. It is also noteworthy that in studies where associations are identified between occupational exposure to trichloroethylene and cancer, the authors are typically confident that the exposure assessment was robust and was unlikely to have introduced bias or uncertainty in the associations identified. Conversely, where no association was identified (for example in Vlaanderen et al, 2013) the authors extensively critique the tools used for exposure assessment as a key contributing factor to failing to identify an association.

In spite of the challenges with the interpretation of the epidemiology data due to the exposure assessments, IARC recently concluded that trichloroethylene could be considered as a human carcinogen, indicating that the kidney was the key cancer site. However, the risk of non-Hodgkin lymphoma and liver cancer were not consistently increased across studies and therefore there was less certainty of an association with TCE exposure. Similar conclusions regarding the relevance of kidney cancer to humans have been reached by the EU risk assessment, SCOEL and the EPA IRIS document.

The key factors that led to the IARC conclusion were the findings from a recent publication by Moore et al (2010) which appeared to be consistent with one interpretation of kidney cancer mode of action. The publication of Moore et al (2010) assessed occupational exposure to trichloroethylene, kidney cancer incidence and genotypic variations in the GSTT1 gene for a glutathione-s-transferase that is known to conjugate small, halogenated compounds. The rationale for the study was that the glutathione conjugation pathway has been identified in rats as leading to the formation of a genotoxic metabolite. Therefore, assuming the same pathway is important in humans, in those subjects with an active form of the GSTT1 gene, one would expect to observe an increased risk of kidney cancer associated with trichloroethylene exposure, since this population would more effectively metabolise trichloroethylene to the reactive metabolites.

In the study of Moore et al (2010) an association was identified between kidney cancer incidence and exposure to trichloroethylene, and also an association between trichloroethylene exposure, kidney cancer and the presence of the active GSTT1 gene. The associations (although weak) were therefore consistent with a hypothesized mode of action and therefore sufficient to support the classification of trichloroethylene by IARC as a category 1 carcinogen. However, when considering this conclusion and the hypothesis underpinning it, there are some additional considerations that should be discussed.

With respect to the relevance of the hypothesised mode of action to humans, the key consideration is the activity of the glutathione pathway in humans. This is discussed in more detail in the mode of action discussion attached to the dossier. However, in brief, there has been significant debate about the prevalence and activity of this pathway in rodents and humans as a consequence of the analytical methodologies employed by different research groups in order to measure the activity of the pathway.Trichloroethylene is first conjugated with glutathione in the liver and potentially also the kidney, and then undergoes further metabolism via the kidney beta lyase enzyme and mercapturic acid formation enzyme systems to formS-(1,2,dichlorovinyl-L-cytseine) and S-(1,2,2-trichlorovinyl- L-cysteine) (DCVG and TCVG). These metabolites are then considered to be causal agents in kidney toxicity and subsequent tumour formation in rats and potentially in humans. The activity of the pathway has been assessed in the past usinga complex two-step derivatization procedure(Reedet al., 1980)followed by ion-exchange chromatography and UV-detection(Lashet al., 1999b;Lashet al., 1998a;Lashet al., 1998b). This methodology suggests that the activity of the reductive pathway in humans exceeds that in the rat. However due to the complexity of the methodology and the formation of interfering peaks, it is likely greatly overestimating the activity of the pathway(Greenet al., 1997a). Conversely, several other studies report a far lower activity using14C-labelled TRI as a substrate, identifying DCVG and TCVG by mass spectrometry, and quantifying14C-DCVG and14C-TCVG by HPLC-separation and counting of14C-activity(Dekantet al., 1998;Dekantet al., 1990a;Dekantet al., 1987;Greenet al., 1997a;Greenet al., 1990). Due to the far higher specificity of this methodology the results are considered more reliable than those from the two-step derivatization method in that they specifically quantitate the levels of DCVG and TCVG metabolites, while the derivatization method is non-specific and may be confounded by other non-DCVG or TCVG endogenous substrates. Given that the more accurate analytical methodology shows a very low activity of the reductive metabolic pathway in humans versus rats, there is significant doubt about whether the hypothesized genotoxic mode of action is biologically plausible in humans. Therefore one must also question the nature of the associations identified in the Moore et al (2010) study.

With respect to the findings of the Moore et al (2010) study the associations between exposure and cancer were weak. Specifically, there wasincreased risk observed among subjects ever TCE exposed [odds ratio (OR) = 1.63; 95% confidence interval (95% CI), 1.042.54]. Exposure-response trends were observed among subjects above and below the median exposure [average intensity (OR = 1.38; 95% CI, 0.812.35; OR = 2.34; 95% CI, 1.055.21; Ptrend = 0.02)]. A significant association was found among TCE-exposed subjects with at least one intact GSTT1 allele (active genotype; OR = 1.88; 95% CI, 1.063.33) but not among subjects with two deleted alleles (null genotype; OR = 0.93; 95% CI, 0.352.44; Pinteraction = 0.18). Similar associations for all exposure metrics including average intensity were observed among GSTT1-active subjects (OR = 1.56; 95% CI, 0.793.10; OR = 2.77; 95% CI, 1.017.58; Ptrend = 0.02) but not among GSTT1 nulls (OR = 0.81; 95% CI, 0.24-2.72; OR = 1.16; 95% CI, 0.275.04; Ptrend = 1.00; Pinteraction = 0.34). Many of the associations, although indicating increased risk, are not significant, and contrary to the claims made by the authors about the size of the study, when one considers the population that were exposed to TCE, there were only 48 cases and 40 controls. Taken together, one can question whether the associations identified were therefore real, or a consequence of small sample size, bias or the other limitations of the study that were not discussed by the authors:

1.      There is no validation of the exposure estimates for TCE or any of the other exposures in the study. This is generally a limitation of most case control studies;

2.      This is a relatively small study for evaluating TCE exposures among cases and controls with GSTT1 active (32 cases of kidney cancer among exposed workers) since high exposure to TCE is relatively rare.;

3.      Both smoking and body mass index are risk factors for kidney cancer. However, there is no attempt to control these factors in the analysis.

4.      A categorical analysis was used for the study rather than the more conventional continuous exposure-response approach. The authors state that the reason they did this was because exposure estimates were categorical. However, they did calculate cumulative exposure to TCE. Therefore to be comprehensive, they should have presented both continuous and categorical models.

With these limitations, it is difficult to come to a firm conclusion of the findings.

Taking into consideration the biological plausibility of the findings in the Moore et al 2010 study along with the limitations identified within the study it is questionable whether this work can be considered as critical in the assessment of trichloroethylene induced kidney cancer, particularly in the light of some far larger studies, Lipworth et al (2012), Hansen et al (2013) and Vlaanderen et al (2013). These large cohort studies failed to demonstrate an association between occupational trichloroethylene exposure, kidney cancer and non-Hodgkin lymphoma and it is noteworthy that the study of Hansen et al (2013) used biological monitoring data to assess exposure potential, and as such was more robust in terms of exposure quantification than other similar studies.

Prior to the Moore et al (2010) study, a study by Charbotel et al (2006) of screw cleaning workers in France was considered as a critical study in the association of trichloroethylene exposure and kidney cancer. In particular this study plays a critical role in the EPA IRIS review, forming the basis for a dose response assessment for kidney cancer. Although it was a relatively small study of a specific population, it was considered to have several advantages of other similar studies due to its use of more robust exposure assessment methodology that provided quantitative exposure information for the individual participants. However there are several concerns with this study that should be taken into consideration when assessing its use in risk assessment and hazard characterization. For example,potential selection bias, the quality of the exposure assessment and the potential confounding due to other exposures in the work place. With respect to the potential for selection bias, no cancer registry was available for this region to identify all relevant renal cell cancer cases from the target population. Case ascertainment relied on records of local urologists and regional medical centers; therefore, selection bias may be a concern. Given the concerns of the medical community in this region regarding renal cell cancer (RCC) among screw cutting industry workers, it is likely that any cases of renal cell cancer among these workers would likely be diagnosed more accurately and earlier. It is also much more unlikely that a RCC case among these workers would be missed compared to the chance of missing an RCC case among other workers not exposed to TCE. This preference in identifying cases among screw cutting industry workers would bias findings in an upward direction. Concerning the potential for other exposures that could have contributed to the association, screw cutting industry workers used a variety of oils and other solvents. Charbotel et al. reported lower risks for TCE exposure and renal cell cancer once data were adjusted for cutting oils. In fact, they noted, “Indeed, many patient had been exposed to TCE in screw-cutting workshops, where cutting fluids are widely used, making it difficult to distinguish between cutting oil and TCE effects.” This uncertainty questions the reliability of using data from Charbotel et al. since one cannot be certain that the observed correlation between kidney caner and exposure is due to trichloroethylene.

Given the complexity of the epidemiological database for trichloroethylene several investigators have used meta-analyses to better interpret the data and identify if there are real associations between trichloroethylene exposure and cancer. In these analyses, the cancers of interest are multiple myeloma (MM), leukemia, liver, kidney and non-Hodgkin lymphoma (NHL). The meta-analyses show the following:

1.      MM and leukemia have very consistent results across studies and the meta-relative risks are close to 1.0. The authors conclude that the data do not support an association with leukemia or MM and TCE exposure (Alexander et al., 2006);

2.      The meta-analyses on NHL and TCE exposure are not consistent (Mandel et al. 2006; Scott & Jinot, 2011). Therefore, no conclusions about NHL can be offered, although the Mandel et al. conclusion “insufficient evidence to suggest a causal link” appears appropriate;

3.      The three meta-analyses on kidney cancer do not give consistent results on the heterogeneity across studies (Scott & Jinot, 2011; Karami et al. 2012; Kelsh et al. 2010). Karami et al. (2010) the most recent, concludes the reason for lack of consistency is the poor exposure classification of TCE. Assuming this is the case, there does appear to be some unresolved issues of a potential association; and

4.      The single meta-analysis on liver cancer (Scott & Jinot, 2011) report consistent finding across studies. However, the authors do mention that reporting bias is likely an issue (i.e., studies report only positive finding [increased relative risk] for rare cancers such as liver and fail to report lack of findings), and the number of liver cancers across studies make it difficult to come to a conclusion on these cancers.


In summary, there is not a lot of consistency across studies of trichloroethylene exposed workers and cancer risk. This may be a result of poor exposure characterization in these studies and in the case of liver cancer the possibility of reporting bias. Where positive associations were identified (Scott and Jinot, 2011) the observed summary relative risk estimates from the meta-analyses of kidney cancer, liver cancer, and non-Hodgkin’s lymphoma (NHL) were not sufficiently strong to be able to rule out other potential explanations such as bias due to confounding, exposure misclassification, or other factors (e.g. selection bias in case control studies).

In conclusion, in spite of the wealth of epidemiology data there is still uncertainty about the association of trichloroethylene exposure and human cancer. The strongest associations appear to exist with kidney cancer and to a lesser extent liver cancer. However, to date, the evidence is insufficient for increased cancer risk for TCE.