Water supply and arsenic measurements

Arsenic concentrations in Danish drinking-water were obtained from a database managed by the Geological Survey of Denmark and Greenland (Thomsen et al. 2004), which contains the results of chemical tests in water utilities in Denmark. Since 2001, it has been compulsory for water utilities to measure arsenic in the drinking-water and to report the results to the database. The spatial locations of the water utilities were determined by their geographic coordinates, which were also registered in the database. We calculated the average arsenic concentration for each water utility on the basis of 4,954 measurements in 2,487 water utilities reported between 1987 and 2004, with most measurements dating from 2002–2004. The average at each water utility was assumed to represent arsenic concentrations throughout the study period 1970–2003. As drilling depth might affect arsenic concentrations and might have changed over time, we collected data on the drilling depth and analyzed the correlation with the arsenic concentration in drinking-water using Spearman’s correlation coefficient.
To assess the effect of arsenic in drinking-water on the risk for cancer among cohort members, it was essential to link the arsenic concentrations at the water utilities to each address of the cohort members. Therefore, information on the size and spatial location of 94 water supply areas was collected from local authorities and water utilities in 24 municipalities, covering the vast majority of the geocoded cohort addresses. Seventy-one of the collected water supply areas were supplied by only one water utility, whereas the water in 23 areas came from more than one utility. Therefore, we also collected details of the volume of water distributed from water utilities to these 23 areas to calculate water volume-weighted average arsenic concentrations. If, for example, an area received 40% of its water from one utility and 60% from another, the arsenic concentration in the area would be calculated as 0.4 × concentration at utility1 + 0.6 × concentration at utility2. The 94 water supply areas were mapped in ArcGIS 9.1 (Figure 2) and covered 84% of the cohort addresses.

Figure 2
Ninety-four water supply areas classified according to estimated average arsenic concentration (μg/L). These areas cover 84% of the 198,758 geocoded cohort addresses.
The geocoded cohort addresses, water utilities, and water supply areas with their arsenic concentrations were mapped in ArcGIS 9.1, and arsenic concentrations were assigned to the cohort members’ addresses, with the spatial join functionality. First, the 170,403 (84%) of the cohort addresses located within one of the mapped water supply areas were assigned an estimated arsenic concentration, by the “point-in-polygon” procedure. This procedure allocates the attributes of the polygon to all points within it. Second, 28,355 (14%) of the addresses were assigned the arsenic concentrations of the nearest water utility, by application of the “point-to-point”-spatial join. By this procedure all points in one data set will be given the attributes of the points in another data set based on shortest distance. The last 3,581 (2%) of the cohort addresses had no geographic reference and were allocated a “missing value” as arsenic concentration.
Arsenic exposure.
We calculated two exposures for each cohort member. The first was a time-weighted average exposure, calculated as the arsenic concentration in drinking-water multiplied by the time lived at each address, summed for all residential addresses during the study period and divided by the total observation time, with the unit micrograms per liter (Equation 1).