TCE Epidemiology and Cancer Risk

RoseyFields evaluates the human epidemiological evidence for Trichloroethylene (TCE) as follows.

1. Study Populations and Data Sources

TCE has one of the more mature human epidemiological literatures among chlorinated solvents, built from several distinct categories of study populations spanning decades and multiple countries.

Occupational cohorts form the backbone of the TCE cancer literature and include large registry-linked cohorts of Danish, Finnish, and Swedish workers identified through national employment and industrial hygiene records; German case-control studies incorporating direct biomonitoring of urinary TCE metabolites among metal-degreasing and electronics workers; and several major U.S. cohorts of aircraft and aerospace manufacturing workers and autoworkers, many of whom had substantial historical TCE exposure from degreasing operations. A methodologically important subset of these studies used job-exposure matrices or biomarker monitoring to assign individual, TCE-specific exposure estimates rather than relying on broad “solvent-exposed” classifications, and a systematic review identified roughly two dozen such higher-quality studies as most informative for causal inference.

Community or contamination-site cohorts are unusually prominent in the TCE literature relative to many other environmental chemicals, reflecting TCE’s widespread historical use as a degreasing solvent and its frequent detection as a groundwater contaminant near industrial and military sites. Well-known contaminated-water communities studied include Woburn, Massachusetts; Tucson, Arizona; Endicott, New York; the U.S. Marine Corps Base at Camp Lejeune, North Carolina; a cluster of New Jersey towns; the area surrounding the Rocky Mountain Arsenal in Colorado; and a TCE-emissions-based residential study in Milwaukee, Wisconsin. These studies have examined a wide range of outcomes, including pregnancy loss, birth weight, congenital malformations (particularly cardiac defects), and childhood cancer, though nearly all involve co-exposure to other contaminants and comparatively small numbers of affected individuals, limitations discussed further in Section 5.

General-population biomonitoring plays a smaller role in the TCE literature than it does for more persistent chemicals. Because TCE is rapidly metabolized and eliminated from the body, with a much shorter biological half-life than bioaccumulative chemicals, a single blood or urine measurement reflects only very recent exposure, which limits the utility of one-time biomonitoring for reconstructing longer-term individual exposure history in the way that is possible for more persistent chemicals. As a result, the strongest TCE epidemiological studies rely predominantly on occupational job-exposure matrices, industrial hygiene monitoring records, or, in the community studies, environmental water or air concentration data rather than on a general-population biomonitoring program comparable to what exists for chemicals like PFOA and PFOS.

Birth cohort and reproductive-outcome studies for TCE are drawn primarily from the community contamination-site populations described above (Woburn, Tucson, Camp Lejeune, Endicott, New Jersey) rather than from dedicated prospective birth cohorts of the kind used extensively in the PFAS literature, along with a smaller number of occupational studies of solvent-exposed working women examining spontaneous abortion and other pregnancy outcomes.

As with the other chemicals in RoseyFields’ chemical profile series, using multiple, diverse populations strengthens causal inference for TCE in several specific ways. First, the consistency of elevated kidney cancer risk estimates across registry-based Nordic cohorts, German biomonitoring studies, French case-control studies with detailed exposure assessment, and U.S. aerospace and autoworker cohorts — populations with different exposure sources, different co-exposure profiles, and different national background cancer rates — substantially reduces the likelihood that the association reflects chance, a single population’s unique confounding structure, or a shared methodological artifact. Second, because occupational and community-based studies use fundamentally different exposure assessment approaches (industrial hygiene/biomarker data in the former, environmental monitoring in the latter), replication of findings such as developmental cardiac defects across both study types, even where individual studies have important limitations, provides a different and complementary form of evidentiary support than replication within a single study design would. Third, studies conducted across multiple decades — from early German case reports in the 1980s through Danish registry cohorts followed into the 2000s — allow assessment of whether associations persist as industrial hygiene practices, exposure levels, and co-exposure patterns changed over time.

2. Overview of Human Health Outcomes Investigated, by Body System

TCE’s epidemiological literature has examined an unusually broad range of outcomes across nearly every organ system, reflecting both its extensive historical occupational use and its status as a common environmental contaminant.

Outcomes with strong, consistent, quantitatively modeled evidence center on cancer. EPA’s weight-of-evidence review, following a formal Hill-criteria causal analysis and multiple published meta-analyses, concluded there is convincing evidence of a causal association between TCE exposure and kidney (renal cell) cancer, and correspondingly classified TCE as “carcinogenic to humans” by all routes of exposure — EPA’s highest-confidence cancer descriptor. Non-Hodgkin lymphoma (NHL) has strong, though somewhat less convincing, epidemiological support, and liver and biliary tract cancer has more limited but still statistically significant support in meta-analysis. These three cancer types are discussed in detail in Section 3.

Outcomes with suggestive but inconsistent evidence span multiple body systems: additional cancer types, including bladder, esophageal, prostate, cervical, and breast cancer, and childhood leukemia, each supported by a smaller number of studies with a mixed pattern of risk estimates and generally without TCE-specific exposure-response data; autoimmune disease, particularly systemic sclerosis (scleroderma), for which a formal meta-analysis of TCE-specific case-control studies found a statistically significant roughly 2.5-fold increased risk in men but no clear association in women, alongside more limited and inconsistent evidence for rheumatoid arthritis and other autoimmune conditions; developmental cardiac defects, based principally on the community contamination-site studies described in Section 1, several of which found statistically significant associations with specific defect subtypes (particularly conotruncal defects) while others found no association, and which collectively were influential enough that EPA incorporated developmental cardiac malformation as one of several co-critical effects (alongside animal data) underlying its noncancer reference dose; nonspecific kidney toxicity, based on urinary biomarker studies in occupationally exposed workers; and neurological effects, including a body of case-report and small case-control literature raising the possibility of a Parkinson’s-disease-like syndrome associated with heavy TCE exposure, an association EPA and other reviewers considered biologically plausible but not yet adequately characterized in larger, better-designed studies.

Outcomes investigated with null or largely inconsistent findings include: systemic lupus erythematosus, for which two case-control studies found no clear association with solvent exposure generally; congenital malformations other than cardiac defects and, in some studies, oral clefts, where community-study findings were inconsistent across sites; spontaneous abortion and fetal death, examined across several occupational and community studies with generally null or inconsistent results once individual study limitations are accounted for; and a range of other outcomes (hearing loss, specific immune cell population changes in adults, and additional reproductive endpoints) supported only by isolated or low-quality studies.

3. Strong-Evidence Outcomes: Key Studies, Exposure Metrics, and Effect Sizes

Kidney (renal cell) cancer. This is the cancer type with the strongest and most consistent epidemiological support for TCE, and the primary basis for EPA’s “carcinogenic to humans” classification. A formal meta-analysis of 15 cohort and case-control studies identified as having a high likelihood of individual TCE exposure and adequate epidemiologic design found a statistically significant summary relative risk of 1.27 (95% CI: 1.13, 1.43) for kidney cancer with overall TCE exposure, a result that was robust to removal of any individual study and showed no evidence of heterogeneity or publication bias. Restricting the meta-analysis to only the highest-exposure groups within these same studies yielded a higher summary relative risk of 1.58 (95% CI: 1.28, 1.96), providing support for a genuine exposure-response gradient. Individual contributing studies include several well-designed cohort and case-control investigations: a French case-control study with an unusually detailed, TCE-specific quantitative exposure assessment found relative risks around 2.2 for high cumulative exposure, with a statistically significant trend of increasing risk with increasing exposure; a large Danish registry-linked cohort study found a relative risk of 1.7 for workers with five or more years of exposure in a higher-exposure subcohort; a German case-control study using urinary biomarker-confirmed exposure found a relative risk of 2.47; a study of California aerospace workers using detailed job-history-based exposure reconstruction found relative risks around 2.0–2.4 in the highest exposure categories with a statistically significant trend; and a German case-control study using an internal referent group found a relative risk around 4.9 in the highest TCE exposure score category, also with a statistically significant trend. Across these and other contributing studies, statistical adjustment for smoking, body mass index, hypertension, and co-exposure to other occupational chemicals (including cutting and petroleum oils, which are not independently associated with kidney cancer) did not eliminate the association, and a meta-analysis of the same cohorts found no corresponding elevation in lung cancer risk — a finding inconsistent with heavy smoking as an alternative explanation for the kidney cancer signal. EPA’s formal Hill-criteria review concluded this pattern of consistency, exposure-response gradient, and resistance to confounding by known and suspected risk factors provided convincing evidence of causality.

Non-Hodgkin lymphoma (NHL). A meta-analysis of 17 cohort and case-control studies meeting similar quality and exposure-assessment criteria found a statistically significant summary relative risk of 1.23 (95% CI: 1.07, 1.42) for NHL with overall TCE exposure, and 1.43 (95% CI: 1.13, 1.82) when restricted to the highest exposure groups, again supporting an exposure-response gradient. Individual studies contributing to this finding include a Danish cohort study finding a statistically significant relative risk of 3.1, a Swedish case-control study finding a relative risk of 7.2 (though based on self-reported exposure, which introduces uncertainty about the magnitude of this specific estimate), and a U.S. case-control study finding relative risks of 3.3 to 7.9 in the highest cumulative and weekly-average exposure categories with a statistically significant trend. Some heterogeneity was observed across the full set of 17 studies, though it was not statistically significant, and appeared attributable largely to differences between cohort and case-control study designs rather than to a lack of an underlying association. EPA characterized the NHL evidence as strong, though somewhat less convincing than the kidney cancer evidence, reflecting this additional heterogeneity and the comparatively larger proportion of individual studies not reaching statistical significance on their own (even though most showed a positive, directionally consistent point estimate).

Liver and biliary tract cancer. A meta-analysis of nine cohort studies (no case-control studies were available for this outcome) found a statistically significant summary relative risk of 1.29 (95% CI: 1.07, 1.56) with overall TCE exposure. This finding, however, depends heavily on a single large Danish cohort study, which reported nearly six times more liver and gallbladder cancer cases than the next-largest contributing study and itself found a statistically significant relative risk of 1.35 (95% CI: 1.03, 1.77); the other eight contributing studies had few cases and correspondingly wide confidence intervals. Because liver cancer is comparatively rare and TCE-specific exposure-response data were limited to duration-of-employment metrics in the largest study, the exposure-response evidence for liver cancer is weaker than for kidney cancer or NHL, and restricting the meta-analysis to only the highest-exposure groups actually yielded a lower, non-statistically-significant summary estimate. EPA accordingly characterized the liver cancer evidence as more limited than the kidney cancer and NHL evidence, while still concluding it provided meaningful supporting evidence for TCE’s overall carcinogenicity.

Developmental cardiac defects. Although this outcome does not reach the same level of epidemiological consistency as the three cancer types above, it merits discussion alongside them because EPA ultimately relied on it, in combination with animal data, as one of the critical effects underlying TCE’s noncancer reference dose. A study of a Tucson, Arizona community with TCE-contaminated well water found a significantly increased prevalence of congenital heart disease among children born to mothers exposed during the first trimester of pregnancy compared with unexposed families, though the association was present only before the contaminated wells were closed and not after. A separate study using residential proximity to TCE-emission sources in Milwaukee, Wisconsin found a substantially elevated risk of congenital heart defects among children of mothers 38 years or older, though a smaller, non-independent elevation was also observed among unexposed older mothers, complicating interpretation. Community studies of TCE-contaminated water at Endicott, New York found statistically significant elevations in both total and major cardiac defects, with a particularly pronounced elevation in conotruncal heart malformations that remained significant even after excluding infants with Down syndrome; a New Jersey study of contaminated public drinking water found a smaller, non-statistically-significant elevation in major cardiac defects; and a federal investigation of families at the Camp Lejeune Marine Corps Base found suggestive but statistically imprecise elevations in birth defects including cardiac defects, though this analysis has since been flagged for reanalysis due to an identified error in the underlying exposure assessment. Because these studies vary in whether they find an association at all, in which cardiac defect subtypes are affected, and in the magnitude of effect, and because most involve co-exposure to other contaminants in the same water supplies, EPA characterized this body of evidence as having “important limitations” while still concluding that “the overall weight of evidence supports an effect of TCE on cardiac development” sufficient to help anchor its noncancer toxicity value, in combination with corroborating animal data showing cardiac malformations following gestational TCE exposure in rats.

4. Study Quality Considerations

Confounding. Kidney cancer risk factors, including smoking, obesity, and hypertension, were addressed directly in most of the case-control studies contributing to the kidney cancer meta-analysis, and the association with TCE persisted after adjustment; a parallel meta-analysis of lung cancer risk (a cancer strongly associated with smoking) in the same cohorts found no corresponding elevation, arguing against residual confounding by smoking as an alternative explanation. Confounding by co-exposure to other occupational chemicals, particularly cutting and petroleum oils common in metalworking and degreasing operations, was also considered; these oils are not themselves associated with kidney cancer in the broader occupational epidemiology literature, reducing (though not eliminating) concern that they explain the observed TCE associations. For NHL, direct confounder adjustment was less consistently available, particularly in cohort studies, though several case-control studies adjusted for smoking, age, sex, and education, and internal (rather than general-population) referent groups were used in some of the higher-quality cohort studies specifically to reduce confounding from lifestyle differences between exposed and unexposed workers. For liver cancer, established risk factors such as heavy alcohol use and viral hepatitis could not be directly examined in the available cohort-only study designs, representing a genuine limitation on confidence in this specific outcome relative to kidney cancer and NHL.

Exposure misclassification. Many of the older or lower-quality studies in the broader TCE literature relied on job title or broad “solvent-exposed” classifications rather than individual, chemical-specific exposure assessment, which introduces nondifferential misclassification that generally biases observed associations toward the null and may explain some of the studies with null or weaker findings relative to the higher-quality subset with individual job-exposure-matrix or biomarker-based assessment. Duration of employment, used as an exposure surrogate in several studies (including the large Danish liver cancer study), is a particularly weak metric because it does not capture exposure intensity, which can vary substantially between workers with similar tenure. Community contamination-site studies face a related but distinct challenge: exposure is typically assigned based on residential location or water-district boundaries rather than individual biomonitoring, and actual individual-level exposure (accounting for time spent at the residence, water consumption patterns, and behavior) is rarely well characterized, contributing to the imprecision seen in several of the developmental studies discussed in Section 3.

Co-exposure to other chemicals. This is a particularly significant issue for the community contamination-site studies: TCE-contaminated water supplies at Woburn, Camp Lejeune, New Jersey, and elsewhere typically contained multiple co-contaminants, including tetrachloroethylene, 1,1,1-trichloroethane, carbon tetrachloride, 1,2-dichloroethane, benzene, and, in some cases, chromium, making it difficult to attribute observed associations specifically to TCE rather than to the broader contaminant mixture, a limitation the underlying study authors and EPA’s review both acknowledge directly for several of the developmental and reproductive-outcome studies. In occupational settings, co-exposure to other solvents and metalworking fluids similarly complicates attribution for less-studied outcomes, though for kidney cancer specifically, as discussed above, the specific co-exposures most plausibly implicated (cutting and petroleum oils) have not themselves been independently linked to kidney cancer.

Reverse causation. Reverse causation is not a major concern for the primary cancer findings discussed in Section 3, since these rely on prospectively assembled occupational cohorts or case-control designs with exposure assessed independently of, and generally prior to, disease diagnosis. It is a more relevant consideration for the nonspecific kidney toxicity biomarker studies discussed in Section 2, where two of the four key studies were conducted in subjects with already-diagnosed kidney cancer, limiting the ability to distinguish whether observed biomarker changes reflect TCE exposure, the cancer disease process itself, or both; the two studies conducted in disease-free subjects, which found broadly consistent evidence of tubular toxicity, are accordingly more informative for characterizing TCE’s direct nephrotoxic potential independent of any cancer-related confounding.

Statistical power and study heterogeneity. As discussed in Section 3, the liver cancer literature in particular suffers from a small number of available studies and comparatively few cases per study, given the rarity of this cancer type, making the overall summary estimate heavily dependent on a single large study. The NHL literature shows more heterogeneity across studies than the kidney cancer literature, some of which may reflect genuine differences in study populations, exposure assessment approaches, or disease subtype classification (NHL comprises multiple biologically distinct subtypes that were often grouped together in the available studies) rather than an absence of a true underlying association.

5. Independent Expert Panel and Agency Weight-of-Evidence Conclusions

EPA’s IRIS program, applying its Guidelines for Carcinogen Risk Assessment and a formal Hill-criteria causal analysis, concluded that TCE is “carcinogenic to humans” by all routes of exposure — the Agency’s highest confidence cancer descriptor, reserved for cases with convincing epidemiologic evidence of a causal association. This conclusion rests specifically on the kidney cancer evidence described in Section 3, which EPA’s review states “cannot be reasonably attributed to chance, bias, or confounding,” supplemented by strong evidence for NHL and more limited evidence for liver and biliary tract cancer. This is a notably stronger classification, and one grounded more centrally in human epidemiological evidence rather than primarily in animal data, than the classifications RoseyFields has documented for several other chemicals in this series. The International Agency for Research on Cancer has likewise classified trichloroethylene as a Group 1 human carcinogen, its highest-confidence category, based substantially on the same kidney cancer epidemiology, with sufficient evidence in humans for kidney cancer and positive associations with NHL and liver cancer. For TCE’s noncancer effects, EPA’s review characterized the evidence for an immunotoxic hazard (both adult and developmental) as high confidence and the evidence for a nephrotoxic hazard as high confidence, while characterizing the evidence for an effect on cardiac development as supported by an “overall weight of evidence” notwithstanding “important limitations” in the individual underlying studies discussed in Section 3. RoseyFields is not aware of a single independent expert panel comparable to the C8 Science Panel that has issued a unified “probable link” determination across all TCE-associated outcomes; rather, the weight-of-evidence conclusions for TCE have been developed largely through EPA’s own formal, externally peer-reviewed IRIS assessment process and corroborated by IARC’s independent classification, both of which reached convergent conclusions specifically on kidney cancer causation based substantially on the same underlying epidemiological record.

6. Use of Epidemiological Evidence to Support Causation Opinions

For TCE, epidemiological evidence plays an unusually central and well-developed role in supporting general causation opinions relative to many other environmental chemicals, particularly for kidney cancer. The consistency of statistically significant, dose-related kidney cancer findings across registry-based Nordic cohorts, German biomonitoring studies, French case-control studies with detailed quantitative exposure assessment, and U.S. aerospace and autoworker cohorts — populations differing in nationality, industry, exposure assessment method, and background confounding structure — satisfies core Bradford Hill considerations (consistency, strength, exposure-response gradient, temporality, and biological plausibility, the last independently reinforced by corroborating animal bioassay and mechanistic/mutagenicity data) to a degree that supports a general causation opinion for kidney cancer with an evidentiary foundation as strong as, or stronger than, that available for any other chemical in RoseyFields’ chemical profile series. General causation opinions for NHL rest on a similar, though somewhat less uniformly consistent, evidentiary structure, while opinions addressing liver cancer, other cancer types, developmental cardiac defects, and autoimmune disease should be calibrated to the more limited and, in some cases, inconsistent state of the underlying evidence for those specific outcomes, as detailed in Sections 2 through 4.

For specific causation in an individual case, the unusually well-developed TCE kidney cancer literature — including formal meta-analytic summary relative risk estimates, exposure-response gradients, and multiple large, well-characterized occupational cohorts — provides a stronger quantitative foundation than is available for many other chemicals for evaluating whether an individual plaintiff’s documented occupational or environmental TCE exposure history falls within ranges associated with meaningfully increased risk, and for characterizing the magnitude of that increased risk in combination with individual medical history, differential diagnosis, and consideration of alternative causes. For outcomes with more limited or inconsistent epidemiological support, particularly developmental cardiac defects and autoimmune disease, specific causation analysis should candidly reflect that the underlying studies — while collectively informative enough to support a qualified general causation opinion — individually carry meaningful limitations (co-exposure to other contaminants, imprecise exposure assessment, and inconsistent findings across sites and studies) that bear directly on the strength of any individual-case causation opinion and should be addressed explicitly rather than glossed over.

Bibliography