Arsenic in Drinking Water and Risk of Urinary Tract Cancer: A Follow-up Study from Northeastern Taiwan

In several epidemiologic studies, arsenic exposure via drinking water has been linked to increased urinary cancer risk, including bladder and kidney (1-8). In 2001, the U.S. Environmental Protection Agency lowered the standard for arsenic in drinking water from 50 to 10 μg/L, mainly based on the extrapolation from data from southwestern Taiwan, where very high concentrations of arsenic in drinking water are found (9). Strongest and most consistent evidence linking arsenic exposure via drinking water to increased urinary cancer risk comes from studies of populations in relatively high exposure areas (>100 μg/L) including Taiwan, Argentina, and Chile. However, the studies from lower arsenic concentration exposure areas (<100 μg/L; refs. 10-17) reported inconsistent results. Most of them reported no overall elevated bladder cancer risk (15-17), and some reported increased risk only among smokers (10-14).

The residents from an arseniasis-endemic area in northeastern Taiwan were exposed to arsenic via drinking water at relatively lower concentrations than those in other areas of southwestern Taiwan (1). Instead of sharing a couple of wells among the whole village in southwestern endemic areas, the residents in northeastern endemic areas usually have a well in their own backyard and drank from the well water for a long period until the implementation of tap water in the early 1990s. The study using data from the northeastern cohort reported an increased risk of urinary cancer, especially urothelial carcinoma in a dose-response fashion (1). For exposure groups between 10 and 100 μg/L, the relative risk (RR) of urinary cancer and urothelial carcinoma was elevated but did not reach statistical significance due to the small number (n = 18). In addition, whether other arsenic exposure metrics including duration, recency, and latency of drinking arsenic well water are associated with urinary cancer remains unknown. We conducted current analysis to elucidate the dose-response relationship with the incidence of urinary cancer in this northeastern Taiwan cohort. Furthermore, with the information supplemented from the questionnaire, we investigated the influence of duration, recency, and latency of drinking arsenic well water containing >10 μg/L arsenic on the risk of urinary cancer. The dose-response relationship of arsenic concentration was then stratified from these characteristics of well water consumption to evaluate whether this dose-response relationship differed in different well water consumption conditions.

This study recruited participants from Tungshan, Chuangwei, Chiaohsi, and Wuchieh in the Lanyang Basin located on the northeastern coast of Taiwan. Residents here have consumed water from shallow wells (<40 m in depth) that contain inorganic arsenic since late 1940s until tap water was implemented in early 1990s. Detailed descriptions of the study areas and enrollment procedures were described previously (1, 18). Briefly, a total of 8,102 residents ages ≥40 years from 4,586 households of 18 villages in the study areas participated in the baseline home interview from 1991 to 1994. Because the national identification numbers were used to link to the national cancer registry profile, 16 participants with missing or redundant numbers were excluded from the analysis, resulting in 8,086 participants in the final analysis. Information collected via the questionnaire included demographic characteristics, cigarette smoking and habitual alcohol consumption status, and residential and well water consumption history. Four well-trained local public health nurses conducted personal interviews using standardized questionnaire. All participants had written informed consents to participate in this study and the data collection procedures had been reviewed and approved by the Institutional Review Board of College of Public Health, National Taiwan University.

Arsenic concentrations were estimated using 3,901 water samples collected from 4,584 households (85.1%). Because the wells of 685 households no longer existed, the arsenic exposure of 1,136 residents was classified as “unknown.” The concentration of arsenic was determined by hydride generation combined with flam atomic absorption spectrophotometry. The range of arsenic concentration in the water samples was between undetectable (<0.15 μg/L) and >3,000 μg/L. We were unable to determine the arsenic concentration of well water samples for 62 participants, resulting in a total of 1,198 residents with “unknown” arsenic concentration. The water samples were collected at their residence at the time of interview, whereas no information on the arsenic concentration of the well water at their prior residences was obtained. In addition to the arsenic concentrations in individual wells, several other measurements of arsenic exposure were evaluated. These included duration of exposure, age starting (latency) and ending of drinking well water, if residents still consumed well water at enrollment (recency), and cumulative exposure status (concentration times duration). All these variables were derived from a questionnaire containing detailed history of residential addresses and well water consumption information including age when they started drinking well water as well as when they stopped drinking well water and changed to tap water. For those who reported to drink well water since birth, they were defined as “started drinking well water from birth” regardless of the arsenic concentration of the well water they consumed. The duration of drinking well water was based on information regarding well water drinking history that included age started and ended and type of water (well water, artesian well water, or tap water). The cumulative arsenic exposure was calculated as the products of arsenic concentration (μg/L) and years of drinking water from the specific well for the participants' lifetime until recruitment. However, for those who have moved, because we have no information on arsenic concentration in every well they ever drank water from, we used the arsenic concentration from water sample we collected to estimate their cumulative arsenic exposure.

All incidences of urinary cancer were identified through linkage with the computerized national cancer registry profiles in Taiwan using subjects' unique national identification numbers. This cancer registry system was implemented in 1978 in Taiwan and is considered as a nationwide cancer registry system with updated, accurate, and complete information. Urinary cancer cases are defined as newly diagnosed cancers of the urinary tract including bladder [International Classification of Diseases for Oncology, Ninth Edition (ICD-9) code 188], kidney (ICD-9 code 189.0), and others (ICD-9 code 189.1-189.9). The histologic type of urothelial carcinoma (formally called transitional cell carcinoma) was determined by the ICD-9 codes.

Person-years for each participant were calculated from the date of questionnaire interview to the date of cancer diagnosis, death, or December 31, 2006, whichever came first. Arsenic concentration was arbitrarily divided into <10, 10 to 49.9, 50 to 99.9, 100 to 299.9, and ≥300 μg/L to put emphasis on lower exposure levels. Cumulative arsenic exposure was categorized using 400 μg/L years as reference group because the average well water consumption duration was 40 years in this cohort and 10 μg/L was the current standard for drinking water. RRs and 95% confidence intervals (95% CI) were estimated by Cox proportional hazards regression models. The adjustment variables in the final model included age (continuous), gender (male, female), schooling years (0, 1-6, >6 years), cigarette smoking status (never, past, current), and habitual alcohol consumption (no, yes). Because cumulative exposure increases with duration of exposure that naturally increases with age, age adjustment is critically important. Adjustment of the effect of age by inserting age as a continuous variable assumes a functional form of risk with age; therefore, we incorporated age as indicator variables for each 10 years for adjustment of age in the model with cumulative arsenic exposure. In addition to the adjustment variables, exposure from birth was added as an indicator variable in the models that included cumulative arsenic exposure and arsenic concentration. Trends across levels of categorical variables were assessed by testing the statistical significance of a single trend variable coded as the category of exposure. Additional test of dose-response trend using the median value in each stratum as the marker for each was provided. The corresponding median values were 0.95, 25.47, 74.87, 139.23, and 535.56 μg/L. All above analyses were done by Stata statistical software (version 8.2; Stata).

The average (SD) follow-up period was 11.6 (3.6) years and a total of 45 newly diagnosed urinary cancer cases were identified. There were 23 bladder cancers and another 23 belonged to other urinary tract cancers. One case was diagnosed with both bladder cancer (ICD-9 code 188.2, urothelial carcinoma) and kidney cancer (ICD-9 code 189.0, renal cell carcinoma); therefore, the total number used for analysis as urinary cancer was 45. Among 23 nonbladder urinary cancer cases, 8 were in “kidney except pelvis” (ICD-9 code 189.0), 9 were in renal pelvis (ICD-9 code 189.1), 5 were in ureter (ICD-9 code 189.2), and 1 was unknown (ICD-9 code 189.9). Of these 45 patients, 36 were categorized as urothelial carcinoma, and the remaining were mostly renal cell carcinoma (n = 5), followed by squamous cell carcinoma (n = 2) and then carcinoma not otherwise specified (n = 1), whereas one had unknown histologic type. The age- and sex-adjusted RRs and corresponding 95% CIs of different measurements of arsenic exposure in relation to urinary cancer risk are shown in Table 1. There was a significant dose-response trend of increasing arsenic concentration associated with increasing risk of urinary cancer, and the age- and sex-adjusted RR (95% CI) was 7.73 (2.69-22.3) for ≥300 μg/L compared with <10 μg/L. Those who reported to be still drinking well water containing ≥10 μg/L arsenic at enrollment were at a significantly increased risk of urinary cancer (RR, 3.54; 95% CI, 1.35-9.32) when compared with those with arsenic level <10 μg/L. Among these people, most belonged to the categories “started from birth and still drank well water at enrollment without any interruption” (63.4%) and “not started from birth and still drank well water at enrollment” (34.3%), and only 68 (2.3%) reported “interrupted but resumed and continued drinking well water at enrollment.” Residents who started drinking well water from birth were associated with ∼4-fold increased risk of urinary cancer (RR, 3.64; 95% CI, 1.29-10.2). Among 1,555 (19.2%) residents who stopped drinking well water and changed to tap water before enrollment, stopping at younger ages was associated with a higher risk of urinary cancer (RR, 4.77; 95% CI, 1.22-18.6 for stopping at age ≤50 years) than those who stopped at older ages. Relative to those who were exposed to accumulative arsenic of <400 μg/L years, the RR became significant at ≥5,000 μg/L years and the dose response trend was statistically significant.

Table 2 showed multivariate-adjusted RRs of different categories of well water arsenic concentration in relation to “all urinary cancer,” histologically confirmed “urothelial carcinoma” and “nonurothelial carcinoma” as well as “urinary cancer in renal pelvis.” Using well water arsenic concentration of <10 μg/L as reference, the RR started to increase at exposure level of 10 to 50 μg/L but did not reach statistical significance until arsenic concentration was >100 μg/L for all urinary cancer (RR, 4.13; 95% CI, 1.32-12.9 for concentration 100 to <300 μg/L and RR, 7.80; 95% CI, 2.64-23.1 for concentration ≥300 μg/L) and urothelial carcinoma (RR, 5.50; 95% CI, 1.39-21.8 for concentration 100 to <300 μg/L and RR, 10.8; 95% CI, 2.90-40.3 for concentration ≥300 μg/L). The dose-response trends were significant for “all urinary cancer” and “urothelial carcinoma” but not in the “nonurothelial carcinoma” disease group. Exposure to the highest arsenic concentration level had a 9-fold increased risk of urinary cancer in renal pelvis (RR, 9.00; 95% CI, 0.90-90.1).

We further investigated the association between the characteristics of well water consumption and risk of urinary cancer and urothelial carcinomas, respectively, after adjustment of age, gender, education years, cigarette smoking, and habitual alcohol consumption status at enrollment (Table 3). Drinking well water with concentration ≥10 μg/L from birth was associated with a 4-fold increased risk of urinary cancer (RR, 3.69; 95% CI, 1.31-10.4) and this risk was even higher in urothelial carcinoma (RR, 4.39; 95% CI, 1.18-16.3). Residents who reported to be still drinking well water at enrollment were also at a significantly higher risk of urinary cancer (RR, 3.50; 95% CI, 1.33-9.22) as well as urothelial carcinoma (RR, 4.55; 95% CI, 1.34-15.5). Those who stopped drinking well water at age <50 years were also at increased risk (RR, 5.11; 95% CI, 1.30-20.0 for urinary cancer and RR, 7.77; 95% CI, 1.47-41.1 for urothelial carcinoma). Among residents who reported to be still drinking well water at enrollment, those who started drinking well water from birth and were still drinking at enrollment without any interruption were at increased risk of urinary cancer (RR, 4.37; 95% CI, 1.46-13.0) and urothelial carcinoma (RR, 4.99; 95% CI, 1.24-20.0), respectively. For cumulative arsenic exposure, the dose-response trends were both significant for urinary cancer and urothelial carcinoma.

The association of arsenic concentration and risk of urothelial carcinoma was then stratified by well water consumption duration (≤50, >50 years) and whether they started drinking well water at birth (no, yes). Among those who drank well water for <50 years, the dose-response trends were statistically significant, with the highest RR (95% CI) of 8.37 (1.63-43.0) when comparing the highest concentration category (≥300 μg/L) with the reference (<10 μg/L; Table 4). The RRs reached ∼5-fold among those who were exposed to arsenic concentration of >50 μg/L for >50 years. The highest RR (95% CI) was 17.8 (1.97-159.8) for concentration ≥300 μg/L compared with <10 μg/L. Residents who started drinking well water from birth were at a 6-fold increased risk of urothelial carcinoma at the concentration level of 50 to <100 μg/L (RR, 6.21; 95% CI, 0.56-69.4; Table 4). The RR (95% CI) for those who were exposed to high concentration of arsenic (>300 μg/L) but did not start drinking well water from birth was 8.77 (1.85-41.6).

Compared with our previous analysis (1), after 10 more years of follow-up, there were only 28 new urinary cancer cases and most of these additional cases were of urothelial carcinoma (25 new urothelial carcinoma cases). Significant dose-response trend was found for urinary cancer, especially the urothelial carcinoma histologic type, and a 10-fold increased risk of urothelial carcinoma was found for the highest arsenic concentration (≥300 μg/L). This confirmed our previous findings as well as studies from other areas that the RRs of urinary cancer were quite high with the exposure to high concentration of arsenic via drinking water (1-8). However, in low arsenic concentration (<100 μg/L), we found similar magnitude of increased risk associated with arsenic concentration of 10 to <50 and 50 to <100 μg/L, but none were statistically significant. This indicated that a longer follow-up period with more cases was warranted to detect a significant effect in this lower exposure group (<100 μg/L) to obtain adequate statistical power.

In addition to individual well water arsenic concentration, we were able to investigate the effect of other characteristics of well water consumption including duration, recency, and latency as well as the combination of concentration and these characteristics. We found that those who started drinking well water from birth, still drank at enrollment, and drank it for long period were at a significantly increased risk of urinary cancer, especially urothelial carcinoma. Evidence from experimental studies (19-22) showed that early-life exposure to arsenic may increase the health risks during early childhood and later in life. Up to date, only one ecologic study (23) conducted in Chile showed that exposure to arsenic in drinking water during early childhood or in utero was associated with an increased mortality in young adults from both malignant and nonmalignant lung diseases. Our study is the first epidemiologic study based on individual data that suggested exposure to arsenic from birth may increase urinary cancer risk much later in life. Even those who stopped drinking well water at enrollment had a 3- to 4-fold increased risk for urinary cancer, especially urothelial carcinoma. The finding of a long latency period for ingested arsenic exposure in developing urinary cancer was consistent with most other epidemiologic studies (8) from either high or low exposure areas throughout the world.

Using cumulative arsenic exposure can provide further information on the relationship between arsenic exposure and urinary cancer because it combines both duration and concentration. For urinary cancer and urothelial carcinoma, the dose-response trends were statistically significant when using cumulative exposure of <400 μg/L years as reference. We further investigated the association between arsenic concentration and urinary cancer risk stratified from duration and latency. Higher urothelial carcinoma risks were found among those who drank higher concentrations of arsenic well water for longer periods. For those who started drinking well water from birth, the RR started to increase significantly at arsenic concentration of >50 μg/L and dropped at ≥300 μg/L, whereas among those who did not start from birth the only significantly increased risk was for the exposure level ≥300 μg/L. The interaction terms between categories of arsenic concentrations and the indicator variables of whether started from birth and durations (≤50 and >50 years) were not statistically significant (P = 0.816 and 0.467, respectively). The pattern of the association between arsenic concentration and risk of urothelial carcinoma appeared similar among those who started drinking well water from birth and those who started later in life. This may suggest that not only high concentrations but also persistent exposure for long durations may also increase a person's risk of developing urothelial carcinoma.

This is the only long-term follow-up study with information on arsenic concentration from individual well supplemented with other important characteristics of well water consumption to thoroughly investigate the complete risk profile of arsenic exposure via drinking water on urinary cancer. The addition of these characteristics of well water consumption could complement the limitation of only one measurement of arsenic concentration. Because of the relatively secluded location of Ilan county in Taiwan, the majority of the residents have lived in this area for several generations and cohort members were similar with regard to socioeconomic status (mostly farmers) and years when tap water system was implemented. All incidences of cancer cases were identified through data linkage; thus, the possibility of selection bias was minimized.

The arsenic concentration was unknown for ∼15% of the study participants and they were at significantly increased risk of urinary cancer. There was no difference in age and gender distribution between the resident with unknown concentration and residents with known concentration. One possibility was that the arsenic concentration of these wells that no longer existed was mostly >10 μg/L. Another limitation was that the exposure level for each participant was based on only one testing on one well, and this well may not be the only source of their water intake. Misclassification of exposure at the high exposure end should not be of major concern because if the well in their backyard contained very high concentrations of arsenic, they were probably exposed. Therefore, this study provided further important information on the hazardous effect of exposure to high arsenic levels on urinary tract cancer. However, the possibility of misclassification in lower exposure group, especially <100 μg/L, cannot be ruled out. Participants who reported “started drinking well water from birth” were assigned to the arsenic concentration of the residences at the time of enrollment, whereas some of them were drinking water from the wells of their previous residences in which we did not have information on the arsenic concentration. However, this cohort was quite stable with about half (4,428, 54.8%) of the participants reported to be living in the same residence for their whole life and another 39.7% (3,208 residents) reported moving only once. In addition, among 2,904 residents who reported to have started drinking well water from birth, 1,643 (56.6%) reported never moving to other places up until enrollment. When we restricted the analysis to these 1,643 residents, although the numbers were very small, their risk of urinary cancer was increased to 5-fold (RR, 5.45; 95% CI, 0.88-33.7; P = 0.068) with arsenic concentration of 50 to <100 μg/L, and the risk pattern was similar to the results in Table 4. Compared with those whose well water contained arsenic with concentration <10 μg/L, the RR (95% CI) of urinary cancer were 3.11 (1.19-8.10), 4.16 (1.24-13.9), and 3.35 (1.05-10.7) for those who did not drink well water from birth, drank well water from birth and have moved, and drank well water from birth and never moved, respectively. Although the misclassification of low exposure groups and other variables representing characteristics of well consumption cannot be ruled out, we have no reason to believe that these misclassifications would be dependent on their arsenic exposure. Thus, this nondifferential misclassification would pull the association estimates toward the null and underestimate the true relationship. The small number of urinary cancer cases limited the statistical power in some parts of these analyses, rendering wide 95% CIs in some risk estimates. However, the risk estimates were consistently high across all categories and all kinds of arsenic exposure measurements, indicating reasonable validity in these study results.

Up to date, there are still millions of people consuming water with arsenic level exceeding 10 μg/L or even 50 μg/L in many parts of the world (24, 25). Because most arsenic in water supplies are from natural sources and with diminishing clean water sources on earth, the complete avoidance of drinking arsenic contaminated water is near impossible in some areas. Our study provided evidence that exposure to arsenic from birth may increase urinary cancer risk much later in life and suggested that, at relatively low concentration levels, long-term and persistent exposure increased risk of developing urothelial carcinoma. This may add more evidence to the full scope of the health consequences from exposure to arsenic via drinking water.

The authors declare they have no competing financial interests involved in this study.

Grant Support: National Science Council grant NSC-83-0412-B002-231 and Department of Health, Executive Yuan, Taiwan grant DOH85-HR-503PL from.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.