Cancer Burden From Arsenic in Drinking Water in Bangladesh

Abstract

We assessed the potential burden of internal cancers due to arsenic exposure in Bangladesh. We estimated excess lifetime risks of death from liver, bladder, and lung cancers using an exposure distribution, death probabilities, and cancer mortality rates from Bangladesh and dose-specific relative risk estimates from Taiwan. Results indicated at least a doubling of lifetime mortality risk from liver, bladder, and lung cancers (229.6 vs 103.5 per 100 000 population) in Bangladesh owing to arsenic in drinking water.

Groundwater contamination caused by inorganic arsenic is a massive public health hazard in Bangladesh.1–4 The millions of hand-pumped tube wells installed since the 1970s have led to 95% of the country’s 130 million residents becoming dependent on supposedly pathogen-free underground water.5 It is estimated that 25 to 57 million people in Bangladesh have suffered chronic exposure to arsenic,1,5 and because decades of exposure have already accrued, the exposed population is at an elevated risk of arsenic-induced health problems.

The principal cause of arsenic-induced mortality is cancer,6–10 but little is known regarding future cancer mortality risks attributable to arsenic exposure among the population of Bangladesh.1 The goal of the present study was to estimate excess lifetime mortality rates for the most-established arsenic-related internal cancers (i.e., lung, liver, and bladder cancers)7–10 in Bangladesh.

METHODS

Calculation of lifetime excess risks due to a particular exposure requires measures of distribution of the exposure, “background” lifetime risks, and dose-specific relative risk estimates. In the present study, these measures were estimated as follows.

The arsenic exposure distribution in Bangladesh was ascertained from a sample of 65 876 people who represented the source population of an ongoing prospective cohort study focusing on the health effects of exposure to arsenic in drinking water. Water samples from 5966 contiguous hand-pumped tube wells in a well-defined geographic area of Araihazar, Bangladesh, were collected and tested for arsenic in 2000. Well owners were interviewed to collect data on the numbers and characteristics of the 65 876 regular users.11

Gender-specific lifetime mortality risks from liver, bladder, and lung cancers among the population of Bangladesh were derived, via life table methods, from the formula ΣS(t_k)P_k. Values of S(t_k) indicate the probability of surviving to the beginning of each of the 5 (i.e., k = 1–5) age groups assessed (0–14, 15–44, 45–54, 55–64, ≥65 years). Survival estimates were based on gender- and age-specific death probabilities among the overall population of Bangladesh.12 Values of P_k indicate gender-, age-, and cancer-specific mortality rates in Bangladesh; these rates were computed by the International Agency for Research on Cancer (IARC).13–16

Gender-specific, age-adjusted relative risks of liver, bladder, and lung cancer mortality due to arsenic exposure were computed on the basis of gender- and age-specific data on arsenic exposure, cancer mortality, and at-risk population obtained from studies conducted in Taiwan (detailed data regarding a published study17 were obtained from C. J. Chen and L. Ryan, January 2002). We used Poisson regression models in calculating these risk estimates, allowing us to compare rates for different levels of arsenic exposure in an endemic area with those in the general population of Taiwan.

Finally, we estimated lifetime excess mortality risks attributable to different levels of arsenic exposure by multiplying gender-specific, age-adjusted excess relative risks from Taiwan by the corresponding category-specific lifetime risks for each cancer in Bangladesh. We weighted these estimates by the arsenic exposure distribution ascertained from our study population in Bangladesh to derive overall lifetime excess risk estimates.

RESULTS

Results showed that, among the overall population of Bangladesh, lifetime mortality risks (per 100 000 population) of cancer of the bladder, lung, and liver were 5.4, 159.1, and 9.2 for males and 0.3, 23.1, and 9.5 for females, respectively. The overall mortality risk for the 3 cancers in combination was 103.5 per 100 000 (Table 1 ▶).

TABLE 1—

Cumulative Lifetime (Background) Mortality Risks (per 100 000 Population) From Bladder, Liver, and Lung Cancers: Bangladesh, 2000

		Liver Cancer		Lung Cancer		Bladder Cancer
Gender and Age Group, y	Survival Probabilitya	Mortality Rateb	Age-Specific Risk	Mortality Rateb	Age-Specific Risk	Mortality Rateb	Age-Specific Risk
Male
0–14	1.00	0.01	0.01	0.03	0.03	0.00	0.00
15–44	0.85	0.35	0.30	2.22	1.89	0.12	0.10
45–54	0.78	1.54	1.20	28.00	21.74	0.87	0.68
55–64	0.70	4.83	3.38	79.83	55.81	2.22	1.55
≥ 65	0.54	7.99	4.34	146.49	79.60	5.70	3.10
Lifetime mortality risk (per 100 000)c			9.22		159.07		5.43
Female
0–14	1.00	0.03	0.03	0.00	0.00	0.00	0.00
15–44	0.86	0.40	0.34	0.76	0.66	0.05	0.04
45–54	0.76	4.24	3.24	6.66	5.08	0.09	0.07
55–64	0.68	3.80	2.60	10.38	7.10	0.24	0.16
≥ 65	0.55	6.00	3.28	19.41	10.60	0.00	0.00
Lifetime mortality risk (per 100 000)c			9.49		23.44		0.28

^aProbability of surviving to the beginning of year t (first year) in each age group k. The survival probability can be denoted by S(t_k) where t = 0, 15, 45, 55, and 65; k = 1–5. Probability estimates were calculated on the basis of category-specific death probabilities in Bangladesh.

^bAge-specific mortality rates (P_k) for the population of Bangladesh (per 100 000), estimated by the International Agency for Research on Cancer for the year 2000.

^cLifetime (background) mortality risks were calculated via the formula ΣS(t_k)P_k.

Lifetime excess risks (per 100 000 population) of mortality from liver, bladder, and lung cancers attributable to arsenic in drinking water were 0.9, 21.5, and 175.9 for males and 3.4, 2.1, and 48.3 for females, respectively (Table 2 ▶). Overall lifetime excess mortality risks (per 100 000) from the 3 cancers in combination were 198.3 for males and 53.8 for females, with an average across-gender lifetime risk of 126.1.

TABLE 2—

Lifetime Excess Mortality Risks (per 100 000 Population) From Bladder, Liver, and Lung Cancers: Bangladesh

Cancer Type	Arsenic Concentration in Water, μg/L	Rate Ratioa	Lifetime Mortality Riskb	Exposure Population, %c	Excess Mortality Rated
Bladder
Male	< 50	5.80	5.43	46.03	12.00
	50–99	2.29	5.43	18.16	1.27
	100–299	4.09	5.43	28.92	4.85
	300–599	9.57	5.43	6.52	3.03
	≥ 599	16.87	5.43	0.38	0.32
	Total				21.48
Female	< 50	10.75	0.28	46.03	1.26
	50–99	4.15	0.28	18.16	0.16
	100–299	6.75	0.28	28.92	0.47
	300–599	12.97	0.28	6.52	0.22
	≥ 599	25.79	0.28	0.38	0.03
	Total				2.13
Lung
Male	< 50	1.81	159.07	46.03	59.31
	50–99	1.00	159.07	18.16	0.00
	100–299	2.95	159.07	28.92	89.70
	300–599	3.41	159.07	6.52	24.99
	≥ 599	4.22	159.07	0.38	1.93
	Total				175.92
Female	< 50	2.95	23.44	46.03	21.04
	50–99	2.44	23.44	18.16	6.13
	100–299	3.27	23.44	28.92	15.39
	300–599	4.29	23.44	6.52	5.03
	≥ 599	9.00	23.44	0.38	0.71
	Total				48.29
Liver
Male	< 100e	0.91	9.22	64.19	−0.53
	100–299	1.40	9.22	28.92	1.07
	300–599	1.54	9.22	6.52	0.32
	≥ 599	2.15	9.22	0.38	0.04
	Total				0.90
Female	< 100e	0.99	9.49	64.19	−0.06
	100–299	2.16	9.49	28.92	3.18
	300–599	1.39	9.49	6.52	0.24
	≥ 599	2.54	9.49	0.38	0.05
	Total				3.42

^aAge-adjusted rate ratios estimated, via Poisson regression, through data from Taiwan, with mortality rates in southwestern region of Taiwan as the reference category.

^bBackground mortality risks in Bangladesh, estimated as lifetime risks per 100 000 population (from Table 1 ▶).

^cPercentage of the study population in Araihazar, Bangladesh (n = 65 876), using well water at each arsenic concentration for drinking and cooking purposes.

^dLifetime excess mortality rate per 100 000 population attributable to arsenic in drinking water.

^eData for finer exposure categories (< 100 μg/L) were not available for liver cancer.

DISCUSSION

Our study indicates a more than doubling of future excess mortality in Bangladesh owing to cancer of the lung, liver, and bladder resulting from exposure to arsenic in drinking water (i.e., a rate of 229.6 per 100 000 population vs the background overall risk of 103.5 per 100 000 population). Our analyses employed a straightforward method measuring excess lifetime risks on an additive scale. A similar approach has been applied to predict the cancer burden due to arsenic exposure in the United States.18

Several uncertainties involved with our estimations warrant caution in interpreting our findings. First, in generating our exposure distribution, we were unable to pool data from other large-scale surveys conducted in Bangladesh since water samples in those surveys were not collected in a systematic manner and the population distributions of the individual exposure categories were unknown. However, the extent of arsenic contamination in our study area was comparable to estimates reported in those large-scale surveys.2,5,19

Second, the dose-specific relative risk estimates we used in predicting risks were derived from Taiwan data, since no such estimates are currently available in Bangladesh. Data from Taiwan have also been used in most arsenic risk assessments for the US population.18,20–22 Given the long latency of arsenic-induced cancer and the similarity in durations of well water use between the exposed populations of Taiwan (1910s–1970s)7 and Bangladesh (since the 1940s),19 the effects of arsenic are assumed to be similar in the 2 populations.

Although our relative risk estimates did not demonstrate a strict dose–response pattern, lifetime excess risk estimates did not change appreciably when the exposure categories were grouped differently (data not shown). However, the impact of potential differences in the distributions of other risk factors related to the studied cancers between the populations of Bangladesh and Taiwan is unknown. The prevalence of cigarette smoking among Bangladeshi men (56–72%) is higher than the prevalence reported in the Taiwan study from which we generated the present relative risk estimates (32%), and nutritional deficiency is more prevalent in Bangladesh.23–25 Hepatitis B virus infection was probably more prevalent in Taiwan in the 1980s, the period during which the relative risk estimates used in the present study were derived.26,27 Whether the 2 populations are comparable in terms of arsenicrelated genetic factors is unknown.

Finally, the IARC estimated cancer mortality rates for Bangladesh on the basis of cancer incidence rates in India, age-specific cancer ratios in Bangladesh, and cancer survival rates in developing countries.13–16 Because cancer mortality data are scant in Bangladesh, it is difficult to evaluate the validity of the IARC estimates. However, given the geographic and sociocultural similarities of India and Bangladesh, and the dissimilarity of the 2 countries in regard to arsenic exposure, the IARC estimates are probably the best data available for estimating “background” lifetime risks.

In conclusion, our results suggest at least a doubling of the potential cancer burden in Bangladesh due to arsenic exposure. Measures focusing on reductions in arsenic exposure, early diagnosis, and treatment of arsenic-induced cancers are thus urgently warranted. In addition, risk estimates derived directly from individual-level data are needed for more precise risk assessments tailored to the population of Bangladesh. Prospective analyses based on our ongoing epidemiological cohort study will address this issue in the near future.