Page 11: Guidelines for Canadian Drinking Water Quality: Guideline Technical Document – Arsenic

10.0 Classification and assessment

Arsenic is a documented human carcinogen. It has therefore been classified in Group 1 (carcinogenic to humans) both by Health Canada (as defined in Health Canada (1994)) and by the International Agency for Research on Cancer (IARC). Toxic effects other than cancer have also been observed in populations ingesting arsenic-contaminated water supplies; however, carcinogenicity is considered to be the critical effect for derivation of the guideline.

It is important to note that while animal studies have confirmed the carcinogenicity of arsenic, significant differences concerning the observed toxic effects of arsenic exist between animal species. Hence, human studies remain the most reliable sources to be used in establishing a maximum acceptable concentration (MAC).

While early studies on the southwestern Taiwanese population indicated an association between arsenic in drinking water and cancer of internal organs (Chen et al., 1985; Wu et al., 1989), this information on its own was not deemed to be sufficient for quantitative risk assessment during the development of the 1989 guideline for arsenic in drinking water. As a result, the 1989 guideline was based on the increased incidence of skin cancer observed in the southwestern Taiwanese population (Tseng et al., 1968) and a model devised by the U.S. EPA, which estimated lifetime skin cancer risks associated with the ingestion of arsenic in drinking water using a multistage model modified to take into account incidence stratified by age group. This model was quadratic as well as linear in dose and included an adjustment for the larger water consumption of southwestern Taiwanese compared with North American men. Based on this model, lifetime risks of skin cancer in the general population in Canada for ingestion of 1 µg/L of arsenic in drinking water were estimated to range from 1.3 × 10-5 (based on southwestern Taiwanese women) to 3.6 × 10-5 (based on southwestern Taiwanese men).

New data have become available that suggest that the risk of internal cancers due to ingestion of arsenic in drinking water is greater than previously believed (U.S. NRC, 1999). Chen et al. (1992) evaluated cancer potency indices in the liver, lung, bladder, and kidney for cancers induced by the ingestion of inorganic arsenic in drinking water. A comparison of observed number of deaths and mortality rate by age, sex, and arsenic level in drinking water for these various internal cancers indicated that lung and bladder cancer presented the greatest lifetime risks for development at an arsenic level of 10 µg/kg bw per day. Morales et al. (2000) calculated excess lifetime risk estimates in the same population for bladder, liver, and lung cancers resulting from exposure to arsenic in drinking water using several mathematical models (generalized linear model, multistage Weibull model, and several variations of these); results for risk estimates were sensitive to the choice of model used.

In addition, a review of the health assessment concerning the toxicity of arsenic in drinking water based on human data from southwestern Taiwan indicates a positive relationship between internal organ cancers (lung, bladder, liver, and kidney) and the ingestion of arsenic in drinking water. Similar conclusions were also reported by U.S. EPA (2001a), U.S. NRC (1999, 2001), and WHO (2003). It should also be noted that, although lacking in necessary data for risk quantification, other studies support the association of arsenic in drinking water with cancers of internal organs (lung and bladder) (Kurttio et al., 1999; Lewis et al., 1999; Ferreccio et al., 2000; Chiou et al., 2001).

The southwestern Taiwan ecological study, as reported by many authors, including Wu et al. (1989), Chen et al. (1992), and the U.S. NRC (1999), has been recommended for quantitative risk assessment (U.S. EPA, 2001a; U.S. NRC, 2001). This study population has been chosen because it presents sufficiently long-term exposure to arsenic and follow-up, extensive pathology data, homogeny between lifestyles of the population, and a large population size (approximately 40 000 people) (U.S. NRC, 2001). A statistical analysis by Morales et al. (2000) fit nine Poisson-type models and one Weibull model to this data set in estimating the risk of cancer to the bladder, liver, and lung from exposure to arsenic in drinking water. Although the U.S. EPA (2001a) concluded that model 1 from Morales et al. (2000), which did not use a comparison population, was more reliable than those models utilizing a comparison population, the U.S. NRC (2001) recommends that an external, unexposed population should be used in the dose-response analysis. The use of an external comparison population is classically used in the analysis of cohort data (Breslow and Day, 1987), since it provides a more accurate estimate of the baseline cancer rates and minimizes the impact of exposure misclassification in the low dose range within the study population. On the basis of a review of the available data, Health Canada used the increased incidence of internal organ cancers observed in the southwestern Taiwanese population for calculating the estimated unit risk of cancer due to arsenic exposure through drinking water instead of the increased incidence of skin cancer that was used in the 1989 guideline. Health Canada (2005) concluded that a Poisson model recommended by the U.S. EPA (2001a) and fit by Morales et al. (2000) with an external unexposed comparison population is the most appropriate for estimating the cancer risks associated with the ingestion of arsenic in drinking water. The population from the southwestern region of Taiwan was chosen over the entire Taiwanese population as an external comparison population since it reduces potential bias and confounding that can be associated with differences in populations (i.e., the urban national population versus the rural southwestern region). In the quantitative risk assessment, Health Canada (2005) adopted assumptions similar to those of the U.S. EPA (2001a) regarding the choice of risk metric and the use of a southwestern Taiwanese to Canadian conversion factor. The Health Canada (2005) model analysed data from Morales et al. (2000), who sourced their data from Chen et al. (1985) and Wu et al. (1989).

Overall, using a 1% increase in risk, the unit risks associated with ingestion of 1 µg/L of arsenic in drinking water are estimated to range from 3.06 × 10-6 to 3.85 × 10-5, with 95% upper bounds ranging from 6.49 × 10-6 to 4.64 × 10-5 The most sensitive endpoint for both males and females was lung cancer (Health Canada, 2005). The overall unit risk associated with the ingestion of arsenic in drinking water is reported as a range, given that lifetime exposure to arsenic results in more than one cancer endpoint in different individuals. The above unit risk range has the liver cancer unit risk (3.06 × 10-6) as its lower bound and the lung cancer unit risk (3.85 × 10-5) as its upper bound. This range also includes the estimated risks for cancers of bladder and other internal organs.

Epidemiological data are often reported with the 95% upper-bound value. This value quantifies the variability in the unit risk due to the variability in the data from the study population. Sources of variability in these data may be, for example, individual differences in arsenic metabolism, drinking rates, or body weights. The 95% upper bound is often interpreted as a reasonable conservative upper-bound estimate of the unit risk. In other words, in repeated trials of the experiment, 95% of the time, the 95% upper-bound value will be above the true value of the unit risk.

Based on this unit risk calculation, an acceptable concentration of arsenic in drinking water can be established that would present an "essentially negligible" level of risk. This target concentration, which is based solely on health considerations, is calculated as 0.3 µg/L. The upper 95% confidence interval for the lifetime cancer risk associated with this concentration in drinking water is 1.9 × 10-6 to 1.39 × 10-5, which falls within the range considered to be "essentially negligible." In the context of drinking water guidelines, Health Canada has defined the term "essentially negligible" as a range from one new cancer above background per 100 000 people to one new cancer above background per 1 million people (i.e., 10-5 to 10-6) over a lifetime.

Page details

Date modified: