Human Health Effects of Trichloroethylene: Key Findings and Scientific Issues

1Trichloroethylene (TCE) is a chlorinated solvent once widely used as a metal degreaser, chemical intermediate and extractant, and component of some consumer products. Total releases to the environment reported to the U.S. Environmental Protection Agency (EPA) Toxics Release Inventory have declined from > 57 million pounds in 1988 to about 2.4 million pounds in 2010. Because it has a relatively short half-life, TCE is not commonly detected in biomonitoring surveys, and the percentage of subjects with detectable levels (> 0.1 ng/mL) has declined from about 10% to 1% between samples collected in 1988–1994 and those collected in 2003–2004. From a regulatory and environmental-cleanup perspective, TCE has been identified in soil or groundwater at > 700 of approximately 1,300 Superfund hazardous waste sites listed by the. Additionally, the U.S. EPA has identified TCE as one of the volatile organic compounds to be regulated as a group in drinking water and as one of the priority existing chemicals under review for regulatory action under the Toxic Substances Control Act. Indeed, because of TCE’s continued presence in the environment, most people are likely to have some exposure to the compound through contaminated drinking water, ambient outdoor or indoor air, or, less commonly, contaminated foods.

2The U.S. EPA’s Integrated Risk Information System (IRIS) program released an updated human health risk assessment of TCE in September 2011. This assessment was developed over a period of > 20 years and underwent many stages of both internal and external peer review. Key inputs were recommendations for additional analysis and research from a National Research Council (NRC) panel report reviewing the key scientific issues pertaining to TCE hazard and dose–response assessment. This report, together with a series of issue papers developed by U.S. EPA scientists, provided the foundation for developing an objective, scientifically rigorous human health risk assessment for TCE. The U.S. EPA’s final assessment also incorporated input from two independent peer reviews by the U.S. EPA’s Science Advisory Board, other federal agencies, and the public.

3Here we describe key findings and scientific issues addressed in the U.S. EPA’s toxicological review of TCE, covering the following topics: a) the role of metabolism in TCE toxicity, which was informed by the development and use of an updated physiologically based pharmacokinetic (PBPK) model; b) the carcinogenicity of TCE, including the development of meta-analyses of epidemiologic studies for informing causal inferences, as recommended by the, and analyses of laboratory animal mechanistic and toxicokinetic data contributing to the evaluation of biological plausibility of the epidemiologic data; and c) noncancer toxicity related to two end points—immunotoxicity and developmental cardiac toxicity—for which substantial new data have become available. Findings and issues related to other important topics not discussed here (e.g., susceptibility, mixtures/coexposures, and dose–response assessment) have been described previously (e.g.,;; ).

4A broad and complex range of relevant information for assessing human health effects of TCE is available. Previous reviews have found TCE to adversely affect the central nervous system, liver, kidney, immune system, and reproductive systems and developing embryo/fetus. As shown in Figure 1, TCE is metabolized in humans and experimental animal species by both oxidation and glutathione (GSH)-conjugation metabolic pathways, with subsequent production of numerous toxicologically active compounds. These include the oxidative metabolites chloral hydrate, trichloroacetic acid (TCA), and dichloroacetic acid, and the GSH conjugation metabolites dichlorovinyl glutathione and dichlorovinyl cysteine. This complex assortment of metabolic compounds is generated from and transported across multiple tissues, making evaluation of mechanistic data especially challenging. Liver effects of TCE are thought to result from oxidative metabolites, whereas effects on kidney are generally associated with metabolites resulting from GSH conjugation. The identity of TCE metabolites involved in the induction of other health effects of TCE is less clear, although similarities have been observed between TCE and its oxidative metabolites in the respiratory tract (e.g., ) and developmental toxicity (e.g., ).

5Tools such as PBPK models can be very useful for integrating complex toxicokinetic information on absorption, distribution, metabolism, and excretion of TCE and its metabolites. Many PBPK models for TCE have been developed to predict the relationship between external measures of exposure and internal dose measures. and updated and “harmonized” these efforts into a new model for use in the IRIS assessment.

6For example, and illustrated the importance of internal dose in investigating mechanisms of TCE toxicity, addressing the key question of whether the TCE metabolite TCA can account for mouse hepatomegaly caused by TCE. They used the TCE PBPK model to compare the hepatomegaly response after TCE administration with the response after direct administration of its metabolite TCA, using the common internal dose measure of TCA liver concentration. If TCA were the only contributor to TCE-induced hepatomegaly, this comparison would show equal changes in liver weight for equal TCA liver concentrations, regardless of whether TCA was the result of TCE metabolism or the result of direct TCA administration. However, as reported by and, TCA appears to account for no more than half of the hepatomegaly that resulted from TCE exposure, implying that effects related to TCE exposure beyond those accounted for by TCA are also operative in TCE-induced hepatomegaly.

7Evaluation of cancer epidemiology for kidney cancer, liver cancer, and non-Hodgkin lymphoma (NHL). The U.S. EPA conducted a systematic review of 76 human epidemiologic studies on TCE and cancer. Each study was evaluated with respect to explicitly identified characteristics of epidemiologic design and analysis to examine whether chance, bias, or confounding could be alternative explanations for the study’s results. A more in-depth analysis (including meta-analysis) of the epidemiologic studies was conducted for kidney cancer, liver cancer, and NHL. These end points were of a priori interest based on the results of a preliminary review of the epidemiologic data and the findings from rodent bioassays of TCE exposure.

8Meta-analysis approach and results. Meta-analyses can be used to combine underpowered studies, to evaluate effects across the set of studies, and to examine consistency (or heterogeneity) of results. The identified a number of weaknesses in previous meta-analyses of TCE carcinogenicity, such as subjective assessment of quality and lack of sensitivity analyses. Thus, the U.S. EPA conducted new meta-analyses to support evaluation of the epidemiologic data on TCE. As recommended by the, the a) established objective study inclusion criteria; b) fit the data to both fixed-effect and random-effects models; c) evaluated statistical heterogeneity across the studies; d) performed sensitivity analyses examining the influence of individual studies and of different measures of relative risk (RR) from studies presenting alternative estimates (e.g., incidence or mortality); and e) conducted tests for potential publication bias (which may occur if positive studies are more likely to be published). Figure 2 presents the meta-analysis summary effect estimates (RRm) from the random-effects models for any TCE exposure (Figure 2A) and for the highest TCE exposure groups (Figure 2B).

9Issues in the interpretation of cancer epidemiologic evidence. Two additional key issues regarding the U.S. EPA’s interpretation of the cancer epidemiologic evidence for kidney cancer, NHL, and liver cancer have been raised in peer review and public comments: the modest magnitude of the RRm estimates for the three cancer types, and the role of meta-analysis within a causality determination.

10The RRm estimates from the meta-analyses for the three cancer types were modest {e.g., with overall exposure (Figure 2A): 1.27 [95% confidence interval (CI): 1.13, 1.43] for kidney cancer; 1.23 (95% CI: 1.07, 1.42) for NHL, and 1.29 (95% CI: 1.07, 1.56) for liver cancer}, raising the possibility that the observed associations could be the result of confounding. However, a detailed examination by the U.S. EPA of potential confounding from lifestyle factors or other occupational exposures concluded that confounding was not supported as an alternative explanation for the observed excesses.

11For example, although smoking can potentially confound kidney cancer results, several kidney cancer case–control studies included in the meta-analysis reported associations with TCE exposure even after controlling for smoking in statistical analyses. In addition, if the cohort studies had been confounded by smoking, increased lung cancer risk would be expected. However, increases in lung cancer risk in individual studies were either absent or insufficient to account for the observed excess kidney cancer risk. Overall, after combining studies, RRm estimates for lung cancer were 0.96 (95% CI: 0.76, 1.21) for overall TCE exposure and 0.96 (95% CI: 0.72, 1.27) for the highest exposure groups.

12Another key issue is the role of meta-analysis in the overall evaluation of causality. Meta-analysis can provide an objective, quantitative method to increase statistical power and precision because the resultant summary effect estimate is based on multiple studies. Strengths of the meta-analyses include study identification based on a systematic and transparent review, evaluations of potential publication bias, examinations of the sensitivity of the overall effect to different inputs, and investigations of possible factors responsible for any statistical heterogeneity observed across studies. However, the U.S. EPA’s characterization of the epidemiologic evidence considered multiple aspects of the data as a whole and did not rely solely on the meta-analysis findings.

13Synthesis of epidemiologic evidence. Table 1 summarizes the epidemiologic evidence according to the key concepts proposed by. For TCE and kidney cancer, there was convincing evidence of a causal association in humans. Particularly compelling was the consistency of increased RR estimates for kidney cancer across the 15 independent epidemiologic studies of different designs and populations from different countries that met the criteria for inclusion in the meta-analysis (Figure 2). The observed increased RRm estimates for kidney cancer that were robust, not being sensitive to different study or RR inputs. The also found no evidence of heterogeneity among studies or publication bias. The observations of a greater RRm estimate with the highest exposure groups (Figure 2B) and of statistically significant trends between TCE exposure and kidney cancer in two high-quality epidemiologic studies support an exposure–response gradient. Finally, potential confounding from smoking or other occupational exposures was unlikely to explain the association of TCE exposure with kidney cancer.

14The evidence on carcinogenicity from epidemiologic studies of TCE exposure was strong for NHL, although less convincing than for kidney cancer. Of the 17 studies that met the criteria for meta-analysis inclusion, most observed increased RR estimates (Figure 2A). The increased RRm estimate observed in the meta-analysis of NHL and overall TCE exposure was robust because it was not sensitive to different study or RR inputs. However, some heterogeneity among studies was observed, although it was not statistically significant. There was also some evidence of potential publication bias. An exposure–response gradient is supported by observations of a greater RRm estimate with the highest exposure groups (Figure 2B) and of a statistically significant trend between TCE exposure and NHL in a high-quality epidemiologic study.

15The epidemiologic evidence was more limited for liver cancer, where only cohort studies with small numbers of cases were available. Of the nine studies that met the criteria for meta-analysis inclusion, most reported increased RR estimates (Figure 2A). The observed a statistically significantly increased RRm estimate in their meta-analysis of liver cancer and overall TCE exposure, but the statistical significance depended on the large study by. There was no evidence of heterogeneity or publication bias. However, the data available did not support an exposure–response gradient because the RRm estimate for the highest exposure groups was lower than that for overall exposure (Figure 2B) and because none of the available studies reported a statistically significant trend between TCE exposure and liver cancer.

16Experimental animal studies, analysis of mode of action, and toxicokinetic considerations. There is clear evidence of TCE carcinogenicity in rodents. Particularly notable is the site-concordant finding of TCE-induced kidney tumors in multiple strains and both sexes of rats exposed by inhalation or gavage. Although the increased incidences were low, they were sometimes statistically significant and were considered biologically significant in light of the very low historical incidences of renal tumors in control rats in various laboratories. There is also site concordance for liver tumors, which were reported in both Swiss and B6C3F1 mice (strains with lower and higher background rates of this tumor, respectively), and in both sexes in the latter strain. The evidence was more limited for TCE-induced lymphohematopoietic cancers in rats and mice. TCE inhalation bioassays have demonstrated a statistically significant increase in pulmonary tumors in mice but not other species [i.e., rats and hamsters]. Finally, testicular (interstitial cell and Leydig cell) tumors were significantly increased in Sprague-Dawley rats exposed via inhalation and Marshall rats exposed via gavage. In three other tested rat strains, ACI, August, and F344/N, a high (> 75%) control rate of testicular tumors limited the ability to detect a treatment effect, although a positive trend was reported in ACI rats. Overall, the rodent cancer data add substantial biological plausibility for TCE carcinogenicity in humans, particularly when combined with the mechanistic data findings.

17Table 2 summarizes hypothesized modes of action and mechanistic data informative to the evaluation of TCE’s carcinogenic mode of action for liver, kidney, and other tumors. Mode-of-action analyses can inform judgments regarding the human relevance of animal bioassay results and aid in identifying particularly susceptible populations or life stages. For kidney carcinogenicity, the concluded that a mutagenic mode of action is operative for TCE, providing further biological plausibility for the epidemiologic findings of TCE-induced kidney cancer. The identification of the mutagenic metabolites as being derived from the GSH conjugation pathway further suggests increased susceptibility in populations with greater metabolism through this pathway. Consistent with this hypothesis, found a statistically significant association among TCE-exposed persons with an active GSTT1 (glutathione-S-transferase theta-1) enzyme [odds ratio (OR) = 1.88; 95% CI: 1.06, 3.33], but not among those with no GSTT1 activity (OR = 0.93; 95% CI: 0.35, 2.44). Although data are lacking on early-life susceptibility to TCE carcinogenicity, the analysis by suggested increased susceptibility to cancer from early-life exposures, particularly for chemicals acting through a mutagenic mode of action. For other end points, there are inadequate data to support a particular hypothesized mode of action.

18The evaluation of TCE carcinogenicity also considered toxicokinetic data on TCE and metabolites, which are consistent with qualitatively similar absorption, distribution, metabolism, and excretion across species and routes of exposure. Mice, rats, and humans all metabolize TCE via the pathways illustrated in Figure 1. Thus, toxicokinetic data support the biological plausibility of TCE carcinogenicity in humans because humans and experimental animals have similar mixtures of TCE and metabolites in target tissues.

19Another issue informed by toxicokinetic data is whether TCE carcinogenicity depends on route of exposure, given that the vast majority of the available epidemiologic data are from inhalation exposures to TCE. Because TCE is systemically distributed and undergoes systemic metabolism from all routes of exposure, there is no reason to expect that cancers such as kidney cancer, NHL, or liver cancer, which originate in separate tissues, would be dependent on route of exposure. Also, TCE-induced tumors have been reported in rodents by both the oral and inhalation routes. Therefore, conclusions regarding TCE carcinogenicity would apply equally to any exposure route.

20Conclusions as to carcinogenic hazard. Supported by the analyses described above and following the U.S. EPA’s Guidelines for Carcinogen Risk Assessment, TCE is characterized as “carcinogenic to humans” by all routes of exposure. This conclusion was based primarily on convincing evidence of a causal association between TCE exposure and kidney cancer in humans. The epidemiologic evidence is strong for NHL, although less convincing than for kidney cancer. Issues increasing the uncertainty in the NHL association include study heterogeneity, potential publication bias, and less evidence for an exposure–response gradient. The epidemiologic evidence was more limited for liver cancer, where only cohort studies with small numbers of cases were available. Finally, animal bioassay, mechanistic, and toxicokinetic data provide further corroboration and biological plausibility to the epidemiologic findings, thus supporting a causal link between TCE exposure and cancer (Table 1).

21As part of its evaluation of TCE noncancer toxicity, the U.S. EPA analyzed the available experimental animal, human epidemiologic, and mechanistic studies of TCE. A summary of the relevant studies for each end point is available in Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1205879). Below we discuss the data pertaining to immunotoxicity and developmental cardiac toxicity, for which there are substantial new experimental and epidemiologic studies, and about which scientific issues have been raised by reviewers or comments. We also provide an overall summary of the hazard conclusions for noncancer toxicity.

22Immunotoxicity. As recently reviewed by and documented in the TCE assessment, the human and laboratory animal studies of TCE and immune-related effects provide strong evidence that TCE exposure increases the risk of autoimmune disease and a specific type of generalized hypersensitivity syndrome. In addition to the epidemiologic studies of specific diseases (e.g., systemic sclerosis), changes in cytokine levels reflecting an inflammatory immune response have been reported in relation to TCE exposure in occupational and residential (i.e., infants exposed to TCE in indoor air) settings. Also, many case reports have associated a severe hypersensitivity skin disorder, distinct from contact dermatitis and often accompanied by hepatitis, with occupational TCE exposure, with prevalences as high as 13% of workers in the same location.

23Human evidence for autoimmune-related effects is supported by experimental animal studies. Numerous studies have demonstrated TCE-induced progressive, accelerated autoimmune responses in autoimmune-prone mice (reviewed by ). After shorter exposure periods, changes in cytokine levels appear similar to those reported in human studies. Longer exposure periods led to more severe effects, including autoimmune hepatitis, inflammatory skin lesions, and alopecia, that differ from the “normal” expression of autoimmune effects in these mice. TCE-induced autoimmune effects have also been reported in B6C3F1 mice, which are not known to have any particular immune-related susceptibility. A treatment-related increase in delayed hypersensitivity response accompanied by hepatic damage has been observed in guinea pigs following intradermal TCE injection, and increased hypersensitivity response was reported in mice exposed via drinking water prenatally and postnatally (gestation day 0 through to 8 weeks of age).

24There is less evidence regarding a possible role of TCE exposure in immunosuppression. Immunosuppressive effects have been reported in a number of experimental studies in mice and rats [see Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1205879)]. Reported effects include reduced responses to bacterial challenge in mice and decreased numbers of antibody-forming cells in rats and developmentally exposed mice.

25Overall, the concordance of human and laboratory animal studies and the spectrum of effects (from biomarkers to frank expressions of disease) strongly support the conclusion that TCE causes immunotoxicity, particularly in the form of autoimmune disease and a specific type of severe hypersensitivity skin disorder, with more limited evidence for immunosuppression. Moreover, these findings lend additional biological plausibility to the association between TCE and NHL, as alterations in immune status are associated with increased risk of NHL.

26Developmental cardiac toxicity. The TCE data include a number of epidemiologic and animal toxicity studies that indicate TCE-induced developmental toxicity. Congenital malformations, particularly cardiac defects, have been associated with exposures to TCE and/or its metabolites in both humans and experimental animals [for example studies, see Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1205879)]. Other TCE-related developmental outcomes observed in both humans and experimental animals include embryonic or fetal mortality, prenatal growth inhibition, and neurological and immunological functional deficits. (see Supplemental Material, Table S1).

27As noted by the, the cardiac teratogenicity of TCE has been the focus of considerable study and analysis. Only geography-based epidemiology studies have evaluated whether there is an association between maternal TCE exposure and cardiac defects in offspring [see Supplemental Material, Table S1 (http://dx.doi.org/10.1289/ehp.1205879)], with some of the studies reporting statistically significant elevations in a variety of cardiac defects [Agency for Toxic Substances and Disease Registry (ATSDR) 2006, 2008; ], and others reporting nonstatistically significant elevations in risk. Interpretation of these data has been controversial because many of the studies are limited by small numbers of cases, insufficient exposure characterization, chemical coexposures, and other methodological deficiencies. In addition, these studies aggregate a broad array of TCE-associated cardiac malformations and have inadequate statistical power to identify any particular kind(s) of defect that may be more susceptible to induction by TCE. The noted that the epidemiologic studies—although limited individually—as a whole showed relatively consistent elevations for cardiac malformations with similar relative effect sizes of 2- to 3-fold, some of which were statistically significant, associated with TCE exposure across multiple studies.

28The outcomes of studies in rodents exposed to TCE during gestation show an inconsistent pattern. Some studies identified significant treatment-related increases in the overall incidence of cardiac anomalies at environmentally relevant exposure levels (e.g.,, ), whereas others reported no excess cardiac abnormalities at much higher dose levels (e.g.,; ). Several methodological factors may contribute to differences across study outcomes, such as the route of administration, test substance purity, test species or strain, timing of dosing or fetal evaluation, procedures used in dissecting and examining fetal hearts, statistical approaches applied to data evaluation, and generally uncharacterized interlaboratory variation.

29Other available data providing evidence of TCE cardiac teratogenicity come from avian and in vitro mechanistic studies. For instance, studies in chick embryos reported consistent effects on cardiogenesis (many demonstrating septal and valvular alterations) when TCE was administered during critical stages of heart development; these findings are similar to some of the cardiac defects observed in rodent studies following in utero TCE exposures. The events of cardiac morphogenesis in birds and mammals are similar; both involving mesenchymal cells that form endocardial cushion tissue with subsequent differentiation into septa and valvular structures in the adult heart. Thus, cultured embryonic chick atrioventricular canal cushion cells have been used to examine chemically induced disruptions in cardiac morphogenesis. In this model, TCE inhibited endothelial separations and mesenchymal cell formation or adhesive properties of endocardial cells, either of which could potentially result in septal or valvular malformations. Other TCE-induced effects that may have morphologic consequences in the developing heart include disruption of endothelial oxide synthetase, which has a role in endothelial cell proliferation, and interference with proteins involved in intercellular Ca2+ regulation, which may result in altered blood flow.

30Overall, the avian and in vitro data substantially increase the biological plausibility for TCE-induced cardiac teratogenesis, and thus strongly support the more limited epidemiologic and in vivo rodent data suggesting that TCE induces cardiac teratogenicity. Moreover, mechanistic data support the possibility that multiple modes of action with different targets within the developing heart may be operant in eliciting cardiac malformations, consistent with the reported association between TCE and overall cardiac malformations in the absence of a strong association with any particular type of defect.

31Conclusions as to noncancer hazard. Table 3 summarizes the evidence for TCE noncancer toxicity across target organs and systems (for additional details, see ). In addition to the immunotoxicity and developmental cardiac toxicity discussed above, there is strong evidence for TCE-induced neurotoxicity, kidney toxicity, liver toxicity, male reproductive toxicity, and several developmental effects in addition to cardiac toxicity. More limited evidence exists for the toxicity of TCE in the respiratory tract and female reproductive system.

32TCE is carcinogenic to humans by all routes of exposure and poses a potential human health hazard for noncancer toxicity to the central nervous system, kidney, liver, immune system, male reproductive system, and the developing embryo/fetus. These conclusions are based on analyses of a broad spectrum of information from thousands of scientific studies and input from numerous scientific reviews. In the last decade, substantial new scientific data on the human health effects of TCE have become available. Moreover, methodologic advancements—such as modeling of TCE toxicokinetics, meta-analyses of epidemiologic studies, and analyses of mechanistic and noncancer hazard information—have improved the scientific rigor and transparency of data interpretation. The approaches and conclusions of the U.S. EPA’s analyses are consistent with the recommendations of the and were affirmed by independent peer review through the U.S. EPA’s Science Advisory Board. In addition, the International Agency for Research on Cancer (IARC) recently upgraded its carcinogenicity classification of TCE to “carcinogenic to humans”. Finally, studies on the health effects of TCE continue to report findings similar to those described in the U.S. EPA’s assessment, such as kidney carcinogenicity and toxicity, immunotoxicity, and developmental cardiac toxicity.