As indicated by , few studies reported significantly increased risks of lung or bladder cancer from exposure to < 100 μg arsenic/L. That does not suggest, however, that the lung and bladder cancer excess risks estimated by the NRC are incorrect. The sample sizes, with type 1 error = 0.05 and power = 0.80, needed to detect the NRC-estimated excess risks of bladder cancer in females are relatively large (). The sample sizes needed to detect associations with lung cancer in females or lung or bladder cancer in males would be even larger. None of the studies conducted since the NRC report had sample sizes large enough to detect the excess risks estimated by the NRC. Nevertheless, there is evidence of increased lung and bladder cancer risk from exposures < 100 μg arsenic/L in the recent studies, particularly in the studies by Chen et al. (2010a
. The monotonic dose response seen across all exposures in Chen et al. (2010b)
and the significant trend (p
< 0.001) for dose response provides strong evidence for the bladder cancer risks seen at the lower exposures.
The three case–control studies that examined the relative risks of bladder cancer from low-arsenic drinking water concentration drew cases and controls from arsenic-endemic areas (Bates et al. 2004
; Meliker et al. 2010
; Steinmaus et al. 2003
). The reason for drawing cases and controls from the same area is, of course, to minimize potential differences other than the factor under study (i.e., arsenic). Drawing cases and controls from the same area, however, may also reduce the difference in arsenic exposure, requiring a larger sample size to determine whether an excess risk exists for a given exposure. Exposure misclassification probably further reduced the difference between groups. Because the estimated exposure difference between cases and controls was minimal in each of these studies, the statistical power to detect the excess risks predicted by the NRC would also have been minimal.
Two of the case–control studies that examined bladder cancer (Karagas et al. 2004
; Michaud et al. 2004
) used toenail arsenic as the measure of exposure. According to Karagas et al. (2000)
, 1 μg arsenic/L water corresponds to 0.1 μg arsenic/g toenail, whereas a doubling of toenail arsenic concentration is associated with a 10-fold increase in water arsenic in samples with ≥ 1 μg arsenic/L. Based on this relationship, cases and controls in all three studies would have been exposed to approximately ≤ 1 μg arsenic/L. Even if this level of exposure were assumed to be lifetime, the power of the studies of Karagas et al. (383 cases, 641 controls) and Michaud et al. (280 cases, 293 controls) to detect the bladder cancer relative risk estimated by the NRC (2001)
would have been minimal. Toenail arsenic is considered a reliable indicator of arsenic exposure, with the strongest relationship being drinking water–arsenic exposure (Adair et al. 2006
). When the concentration of arsenic in water is low, however, the contribution of arsenic to toenail arsenic concentrations becomes less clear as other sources (e.g., food, air, dermal absorption) become more important (Slotnick and Nriagu 2006
Han et al. (2009)
, Lamm et al. (2004)
, and Meliker et al. (2007)
were ecologic studies. The exposure metric in all three was an average exposure measurement by county, a much larger geopolitical unit than the villages in the Taiwan study. Lamm et al. (2004)
relied on U.S. Geological Survey (USGS) data for mean and median arsenic concentrations by county (Focazio et al. 1999
). The USGS data were based on as few as five wells per county and were not restricted to wells used for drinking water. All three ecologic studies were conducted in the United States, where the population is very transient—much more transient than, for example, the southwest Taiwanese population that is the basis of the NRC risk estimates. Hansen (1998)
reported that the median duration in a residence for those in the United States is only 5.2 years. In addition, the U.S. population consumes different sources of fluids (e.g., tap water from other jurisdictions, bottled beverages) in addition to potentially contaminated well water, whereas those in southwest Taiwan likely consumed well water as their principal, if not the only, source of fluids. Furthermore, both Lamm et al. (2004)
and Meliker et al. (2007)
examined bladder cancer mortality rather than incidence. The NRC estimates are for cancer incidence. Because the 5-year survival rate for bladder cancer is approximately 80% (American Cancer Society 2010
), studies based on bladder cancer mortality would underestimate the risks described by NRC. Finally, the arsenic exposures in all three studies were very low (< 10 μg/L). Han et al. (2009)
compared incidence of cancer for different sites by low (< 2 μg/L), medium (2–9 μg/L), and high (≥ 10 μg/L) arsenic counties. The median arsenic concentrations in the counties studied by Lamm et al. (2004)
were 3–60 μg/L, with 65% of the counties and 82% of the population in the range of 3–5 μg/L. The six Michigan counties studied by Meliker et al. (2007)
had population-weighted mean and median arsenic concentrations of 11.00 μg/L and 7.58 μg/L, respectively, compared with 2.98 μg/L and 1.27 μg/L for the remainder of the state. The differences were relatively small.
Some of the recent studies suggest that the ability of an epidemiologic study of arsenic exposure to detect associations with lung or bladder cancer could be impacted by the number of smokers in the study population. Bates et al. (2004)
, Karagas et al. (2004)
, and Steinmaus et al. (2003)
all report an increased risk of bladder cancer in smokers, but not nonsmokers, exposed to relatively low concentrations of arsenic in drinking water. The ecologic study by Meliker et al. (2007)
found a significant increase in lung cancer mortality for the most urbanized of the six arsenic-endemic counties (Genesee County). Because the prevalence of smoking in the most urbanized county was expected to be higher than in the other five counties, the authors suggested the significant elevation in lung cancer mortality is due to the synergy between arsenic and smoking. An earlier study by Kurttio et al. (1999)
also presents evidence of an interaction between smoking and arsenic with respect to bladder cancer, while Chen et al. (2004)
, Ferreccio et al. (2000)
, and Mostafa et al. (2008)
suggest an interaction between smoking and arsenic with respect to lung cancer. Chen et al. (2010a)
found a synergistic effect of arsenic exposure and cigarette smoking for squamous- and small-cell carcinomas of the lung but not for adenocarcinoma. The authors also reported that a relationship between smoking, arsenic, and lung cancer was evident by the significantly elevated RRs among exposed smokers compared with exposed nonsmokers.
The study by Chen et al. (2010b)
found that drinking arsenic-contaminated water since birth had a higher urinary cancer risk than beginning to drink arsenic-contaminated water later in life. Smith et al. (2006)
, in a study of lung cancer risk in an arsenic-endemic area where drinking water concentrations were relatively high, found that early lifetime exposure may convey a greater risk for lung cancer. Bates et al. (1995
, Marshall et al. (2007)
, and Steinmaus et al. (2003)
found that the latency for arsenic-induced cancer was particularly long, indicating that a 40- to 50-year follow-up may be required to detect an excess risk.
The recent studies by Chen et al. (2010a
are noteworthy for their potential to improve the quantitative risk assessment for arsenic. The population is relatively large (8,086) and includes a large number exposed since birth. Arsenic-contaminated drinking water concentrations range from < 10 μg/L to > 300 μg/L. Individual drinking water measurements are available for most in the cohort, as opposed to the village measurements on which the NRC risk estimates are based. Data are also available on age of the individuals, sex, education, cigarette smoking, habitual alcohol consumption, age when the individuals started drinking arsenic-contaminated well water, and age when the individual stopped drinking arsenic-contaminated well water. Additional follow-up of this cohort will provide a valuable database for future risk assessments. Individual data from the study will be needed to do a quantitative risk assessment. The Health Effects of Arsenic Longitudinal Study (HEALS) of an arsenic-exposed cohort of almost 20,000 in Bangladesh (Ahsan et al. 2006
; Chen et al. 2009
) may also be of use in the future in assessing cancer risk at low arsenic exposures. The study has collected individual data on smoking, education, socioeconomic status, skin lesions, arsenic exposure (including biomarkers of exposure), and other variables.