Results

Demographic, dietary, occupational, and other characteristics of the cohort members are presented in Table 1.

Table 1
Demographic, lifestyle, and dietary characteristics of the cohort.
The time-weighted arsenic exposure of the cohort members calculated from 41 years of age up to date of enrollment varied between 0.05 and 25.3 μg/L, with a median concentration of 0.7 μg/L and a mean concentration of 1.2 μg/L. The exposure was generally higher among those enrolled in Aarhus than those enrolled in the Copenhagen area. Aarhus: mean = 2.3 μg/L, median = 2.1 μg/L; Copenhagen: mean = 0.7 μg/L, median = 0.6 μg/L (Table 2).


Table 2
Time-weighted average arsenic exposure from 41 years of age to date of enrollment.
Figure 3.
illustrates a weak tendency towards increasing drilling depth over the last 18 years. Drilling depth explained 4% of the variation in arsenic concentration (R2 = 0.04; n = 3,396).
Figure 3
Groundwater drilling depths as a function of time, based on 3,396 measurements from drillings used for drinking-water.
The results without adjustment for enrollment area (Table 3)
showed no significant association between exposure to arsenic and risk for any type of cancer, except for non-melanoma skin cancer, for which higher arsenic exposure was associated with lower risk. The IRR for non-melanoma skin cancer was 0.88 [95% confidence interval (CI), 0.84–0.94] per micrograms per liter increase in time-weighted average exposure. A similar pattern was seen for cumulated arsenic exposure, with an IRR of 0.95 (95% CI, 0.92–0.97) for a 5-mg increase in exposure. The risk estimates for kidney cancer and melanoma were correspondingly low for both exposure measures but insignificant. Results adjusted for enrollment area (Table 3)
showed virtually no effect for non-melanoma skin cancer, a stronger but still insignificant inverse risk association for melanoma skin cancer, and a significantly increased risk for breast cancer in association with time-weighted average exposure to arsenic (IRR = 1.05; 95% CI, 1.01–1.10).
Table 3
Incidence rate ratios for cancer in association with arsenic exposure.
Quartile-based analyses showed an IRR of 0.73 (95% CI, 0.59–0.91) for non-melanoma skin cancer for the upper quartile compared with the lower quartile of cumulated exposure from 41 years of age to date of diagnosis, but no decrease in risk was seen after adjustment for enrollment area (IRR = 1.14). The similar IRRs for melanoma skin cancer were 0.52 (95% CI, 0.28–0.98; p = 0.04) and 0.53 (95% CI, 0.32–0.88; p = 0.01) with and without adjustment for enrollment area respectively. The risk did not differ significantly between upper and lower quartile for any of the other cancers regardless of adjustment for enrollment area or not (all p > 0.12) (results not shown).
Spline and quadratic tests showed deviation from a linear dose–response relation for cancers of the breast, lung, prostate, and liver, and for melanoma and non-melanoma skin cancers. When evaluated graphically, the dose–response relation for non-melanoma skin cancer showed a systematic nonlinear pattern, with a decreasing trend that leveled off with increasing exposure (Figure 4).
The departure from linearity appeared to be random and nonbiological for the other cancers (results not shown).
Figure 4
Dose–response curve for non-melanoma skin cancer. Reference, IRR = 1 at average time-weighted arsenic exposure of 0.05 μg/L.
In the overall analyses (Table 3,
no adjustment for area), the risk estimates were affected to only a small extent by calculating time-weighted average exposure to arsenic from 1970 for all cohort members regardless of age, by introduction of a 5-year latency or by exclusion of individuals for whom the closest water utility was used as the expected source of drinking-water at the residence (results not shown). Furthermore, these results showed no significant interaction between arsenic and smoking, as the IRR estimates for never, former, and current smokers were not significantly different for cancers of the lung or bladder or non-melanoma skin cancer (all p > 0.12) (results not shown).
Table 4
shows inconsistent directions of the risk association in the two enrollment areas for non-melanoma skin cancer and a consistent direction of the risk association (inverse) for melanoma skin cancer, which was insignificant for both enrollment areas. Further, a consistent direction of the risk association for breast cancer (higher exposure was associated with higher risk) was observed, which was statistically significant in Aarhus when the time-weighted average exposure measure was applied (IRR = 1.06; 95% CI, 1.0–1.11; p = 0.02). None of the risk estimates differed significantly between the two areas (all p > 0.15).
Table 4
Incidence rate ratios for cancer in association with arsenic exposure in the two enrollment areas.