Page 10: Guidelines for Canadian Drinking Water Quality: Guideline Technical Document – Trichloroethylene

9.0 Classification and Assessment

9.1 Cancer Risk Assessment

There are now several epidemiological studies that suggest that TCE is carcinogenic and that show consistency in terms of target tissues and tumour types. However, some fail to reach a level of statistical significance or are confounded by simultaneous exposure to other substances in drinking water or in industrial settings and therefore may be inadequate to infer a causal relationship between TCE and cancer in humans. Nevertheless, there is adequate evidence of TCE carcinogenicity in two species of rodents, although the sites and types of tumour vary with gender and species. Confidence in the relevance to humans of these findings is enhanced by concordance in target tissues between animals and humans for non-cancer and cancer endpoints and by consideration of mechanistic information in the context of species differences in metabolism. Carcinogenicity has been observed in animals exposed to TCE by both inhalation and ingestion, and responses tend to increase with dose.

Several metabolites of TCE are genotoxic, and some are established as known or likely human carcinogens. Some metabolites of TCE are suspected to be carcinogenic and likely involve non-genotoxic mechanisms of effect, such as cytotoxicity and altered cell signalling, both of which may be relevant to humans. Furthermore, some of the TCE metabolites excreted by tumour-bearing animals are similar to metabolites excreted by humans with similar cancers (Birner et al., 1993; Lash et al., 2000). There is a substantial body of evidence that several different mechanisms are responsible for the observed carcinogenicity of TCE in animals, and these appear to be related to the effect mechanisms of the TCE metabolites. It is feasible that the different tumour responses to TCE are attributable to the pharmacokinetic differences between genders and species.

The results considered most pertinent in assessing the weight of evidence of carcinogenicity of TCE in humans are principally the significant increases in kidney tumours in rats (NTP, 1983, 1990), pulmonary tumours in mice (Fukuda et al., 1983; Maltoni et al., 1986, 1988; NTP, 1988) and testicular tumours in rats (Maltoni et al., 1986, 1988; NTP, 1988). Although there is some doubt about the human relevance of pulmonary tumours in mice, it cannot be concluded that the potential tumour induction mechanism in this species does not also occur in humans exposed to TCE. In addition, TCE appears to be weakly genotoxic in in vitro and in vivo assays (IPCS, 1985).

In view of the sufficient weight of evidence of carcinogenicity in two species of experimental animals, TCE can be classified in Group II (probably carcinogenic to humans). This categorization has been confirmed by the International Agency for Research on Cancer (IARC, 1995), which now lists TCE as Group 2A, probably carcinogenic to humans.

The cancer risk assessment for TCE was based on kidney tumours, which were observed in rats of both sexes and in humans. The evidence surrounding kidney tumours is reasonable on several levels. Although the tumours were few, the finding was repeatable. Such tumours are historically rare in rats, so their appearance among dosed animals was considered biologically significant. Such tumours were also observed in Sprague-Dawley rats exposed to TCE by the inhalation route (Maltoni et al., 1986). There are similarities between sites and histopathological characteristics of the tumours observed in human patients and in rat bioassays (Vamvakas et al., 1993, 1998). The metabolites derived from the likely intermediates of bioactivation of TCE are identical in humans and in experimental animals (Dekant et al., 1986; Birner et al., 1993). Small increases in renal tumours in male rats at doses inducing renal damage cannot be dismissed as irrelevant to humans; epidemiological evidence supports the conclusion that TCE may cause kidney tumours in humans. The new evidence associating human TCE exposure with transformation (VHL gene mutations) at nucleotide 454 is important evidence specific to TCE exposure, which provides a genetic fingerprint associating kidney tumours with TCE exposure (Bruning et al., 1997a,b).

The linearized multistage (LMS) method was used (Health Canada, 2003a) to calculate unit risks for the kidney tumour types observed in rats. Use of a linear (LMS) approach is supported by the possible genotoxicity associated with some TCE metabolites, particularly DCVC and DCVG, although a non-linear approach could be argued due to a possible mixed mode of action (mutagenicity and cytogenicity) of TCE and enhanced susceptibility of the rat to nephropathy. The unit risks were calculated for the data on kidney tumours (NTP, 1988, 1990). An allometric scaling factor was applied to the final unit risks, assuming a rat weighs 0.35 kg and a human weighs 70 kg.

The unit risks calculated (Health Canada, 2003a) for pooled combined tubular cell adenomas and adenocarcinomas of the kidneys in rats (ACI, Augusta, Marshall and Osborne-Mendel strains) following oral exposure to TCE for 103 weeks (NTP, 1988, 1990) were 8.11 × 10-4 (mg/kg bw per day)-1 in males and 5.82 × 10-4 (mg/kg bw per day)-1 in females, while the unit risks for renal tubular adenocarcinomas in rats following inhalation exposure for 104 weeks (Maltoni et al., 1986) were 1.20 × 10-4 (mg/m³)-1 in males and 8.1 × 10-5 (mg/m³)-1 in females. The unit risk value of 8.11 × 10-4 (mg/kg bw per day)-1 for pooled combined tubular cell adenomas and adenocarcinomas of the kidneys in male rats (oral study) was chosen among the above values. This corresponds to the highest unit risk and therefore the most conservative value.

For the cancer risk assessment, assuming a "de minimus" (essentially negligible) cancer risk level of 10-6, the maximum acceptable concentration (MAC) for TCE in drinking water can be calculated as follows:

MAC = [70 kg × 10-6] / [8.11 × 10-4 (mg/kg bw per day)-1 × 4.0 Leq/day] 0.022 mg/L (22 µg/L)


  • 70 kg is the average body weight of an adult
  • 10-6 is the de minimis level of theoretical lifetime excess individual cancer risk
  • 8.11 × 10-4(mg/kg bw per day)-1 is the unit risk calculated using the LMS modelFootnote 3
  • 4.0 Leq/day is the daily volume of water consumed by an adult, accounting for multi-route exposure (see "Exposure" section).

Unit risk values were similarly calculated using the LMS method for the various pertinent tumour types (including liver, testis and lymphomas) observed in the rodent carcinogenicity studies with TCE. These unit risk values were used to estimate guideline values, which were then compared with the value obtained using the reproductive-developmental endpoint below. Overall, even with the use of the probably more conservative LMS method, the guideline values based on carcinogenicity were above that determined for the reproductive-developmental endpoint.

9.2 Non-cancer Risk Assessment

For effects other than cancer, a tolerable daily intake (TDI) can be derived by considering all studies and selecting the critical effect that occurs at the lowest dose, selecting a dose (or point of departure) at which the critical effect either is not observed or would occur at a relatively low incidence (e.g., 10%) and reducing this dose by an uncertainty factor to reflect the differences between study conditions and conditions of human environmental exposure.

Choice of the developmental toxicity study (Dawson et al., 1993) for non-cancer risk assessment was based on the appropriateness of the vehicle used (drinking water), the low dose at which the effects were observed, which coincides with the lowest adverse effect level in all animal studies reviewed, the severity of the endpoint (heart malformations) and the presence of evidence for similar effects (e.g., cardiac anomalies) from epidemiological studies (Lagakos et al., 1986; Goldberg et al., 1990; MDPH, 1994; Bove et al., 1995), as well as the observation of similar malformations in studies of TCE metabolites (Smith et al., 1989, 1992; Epstein et al., 1992, 1993; Johnson et al., 1998a,b). Although it is recognized that the Dawson et al. (1993) study is not the ideal key study to use in a risk assessment because of its inherent methodological limitations, it was chosen for the guideline derivation because it was considered the best available study that used a drinking water vehicle and studied the most sensitive (i.e., reproductive) endpoint. Furthermore, the same cardiac anomalies reported in Dawson et al. (1993) were corroborated by Johnson et al (2003). Although the Johnson et al. (2003) study could be used in the risk assessment, the Dawson et al. (1993) study was deemed more appropriate as the key study, because it showed a clearer dose-response relationship. Finally, the choice of a key study investigating reproductive effects was made in recognition of advancing research into the developmental health effects of TCE and to exercise the precautionary principle -- in other words, to protect against the potential for reproductive effects even if the cause-and-effect relationship has not been fully established scientifically.

As only a LOAEL was identified in the critical study, the benchmark dose (BMD) approach was used to estimate the NOAEL. This approach has recently gained acceptance for the risk assessment of non-cancer effects (Haag-Gronlund et al., 1995; U.S. EPA, 1995) due to its many advantages over the NOAEL/LOAEL/uncertainty factor methodology. For example, the BMD is derived on the basis of data from the entire dose-response curve for the critical effect rather than from the single dose group at the NOAEL, and it can be calculated from data sets in which a NOAEL was not determined (as in this case), thus eliminating the need to apply an additional uncertainty factor to the LOAEL (IPCS, 1994; Barton and Das, 1996; Clewell et al., 2000). A lower confidence limit of the benchmark dose (BMDL) has been suggested as an appropriate replacement of the NOAEL (Crump, 1984; Barton and Das, 1996). More specifically, a suitable BMDL is defined as a lower 95% confidence limit estimate of dose corresponding to a 1-10% level of risk over background levels (Barton and Das, 1996). Definition of the BMD as a lower confidence limit accounts for the statistical power and quality of the data (IPCS, 1994).

The BMD method was therefore used (Health Canada, 2003b) to estimate a dose at which the critical effect either would not be observed or would occur at a relatively low incidence, based on the teratogenicity data of the critical study by Dawson et al. (1993). Although these are developmental toxicology data, standard bioassay techniques were used, since individual pup-by-dam data were not available. Typically, developmental toxicology data contain extra-binomial variation due to the "litter effect"; that is, pups from the same dam are more similar than pups from other dams. Due to a lack of data, this variability could not be accounted for in this analysis. The key dosing scenario was the one in which dams were exposed both prior to and during pregnancy, since this most closely mimics what would be expected in the human population. Specifically, the incidence of heart abnormalities among pups was 7/238 (2.9%), 23/257 (8.2%) and 40/346 (9.2%) at doses of 0, 1.5 mg/L and 1100 mg/L (0, 0.18 and 132 mg/kg bw per day).

Using the data from this dosing regimen, the BMD and its lower 95% confidence limit (BMDL) corresponding to a 1%, 5% and 10% increase in extra risk of fetal heart malformations over background were calculated using the THRESH (Howe, 1995) software. A chi-square lack of fit test was performed for the model fit, yielding a significant p-value of <0.0001. The fitted model provided BMDL01, BMDL05 and BMDL10 values of 0.014, 0.071 and 0.146 mg/kg bw per day, respectively (Health Canada, 2003b).

The BMDL10 was chosen as a default value, as has been proposed and used elsewhere (Haag-Gronlund et al., 1995; Barton and Das, 1996). This value remains an uncertain estimate of the NOAEL due to the following: (1) the data do not elucidate the shape of the dose-response curve in the range of the BMDL10; (2) only two dose groups were used to estimate the BMDL10, since the top group was removed to eliminate lack of fit; and (3) it is not known with certainty which BMDL level best represents the NOAEL. However, Haag-Grondlund et al. (1995), applying the same method for non-cancer risk assessment for TCE, found all no-observed-effect levels (NOELs) to be higher than the BMD corresponding to 1% extra risk and 42% of the NOELs and 93% of the lowest-observed-effect levels (LOELs) to be higher than the BMD corresponding to 10% extra risk. Therefore, the BMDL10 of 0.146 mg/kg bw per day was chosen to best represent the NOAEL.

The TDI for TCE can be calculated as follows:

TDI = [0.146 mg/kg bw per day]/100 = 0.00146 mg/kg bw per day (1.46 µg/kg bw per day)


  • 0.146 mg/kg bw per day is the BMDL10, derived as described above
  • 100 is the uncertainty factor (×10 for interspecies variation, ×10 for intraspecies variation).

Using the TDI derived with the BMD method, the MAC can be calculated as follows:

MAC = [0.00146 mg/kg bw per day × 70 kg × 0.2] / 4.0 Leq/day = 0.00511 mg/L (5.11 µg/L)


  • 0.00146 mg/kg bw per day is the TDI, as derived above
  • 70 kg is the average body weight of an adult
  • 0.2 is the default allocation factor for drinking water
  • 4.0 Leq/day is the daily volume of water consumed by an adult, accounting for multi-route exposure (see "Exposure" section).

The alternative, more traditional approach would have been to derive a value using the concentrations in the 1993 Dawson et al. study (0, 1.5 and 1100 ppm) and converting them to doses which reflect actual consumptions (Tardif, 2004). Using the reported quantities ingested (TCE in µL/day), an average body weight of 306 g per rat (average weight gain of 112 g during treatment) and the TCE density of 1.44 g/ml, the concentrations of 0, 1.5 and 1100 ppm would then correspond to doses of 0, 1.18 and 70 mg/kg bw per day, respectively. Using a LOAEL of 1.18 mg/kg day, and an appropriate uncertainty factor of 1000, would yield a value of 4.13 µg/L. The use of this approach would provide a value that is consistent with the recommended MAC of 5 µg/L.

Page details

Date modified: