Materials and Methods

Study population.
The study was based on the prospective Danish cohort Diet, Cancer and Health, which has been described in detail elsewhere (Tjønneland et al. 2007). In brief, 160,725 persons 50–64 years of age and living in one of 23 municipalities in the Copenhagen or Aarhus area were invited to participate. Of these, 57,053 persons (27,178 men and 29,875 women) accepted the invitation and were enrolled between 1993 and 1997. At enrollment, information was collected including on diet, beverages, smoking, education, medical conditions, occupations, reproductive factors, body mass index, and skin reaction to sun. The study “Diet Cancer and Health” has been approved by the relevant Scientific Committees and the Danish Data Protection Agency. Informed consent was obtained from all participants to search information from medical registers including the Danish Cancer Registry.
Since establishment of the Danish Central Population Registry in 1968, all citizens of Denmark have been given a unique personal identification number, which allows accurate linkage among Danish registers. The cohort members were followed up for cancer incidence in the population-based Danish Cancer Registry (Storm et al. 1997) from the time of enrollment until the date of first cancer diagnosis, emigration, death, or 1 August 2003, whichever came first. We included cancers of the lung, bladder, liver, kidney, prostate, female breast, and colorectum, and non-melanoma and melanoma skin cancers. Only first cancers were included, although a case of cancer was included even if it had been preceded by a non-melanoma skin cancer.
Of the 57,053 cohort members, we included 56,378 persons, who filled in the lifestyle questionnaire, reported daily intake of tap water, and had not had a cancer diagnosis before the enrollment.
Residential histories.
Using the personal identification numbers of the cohort members, we traced residential histories between 1970 and 2003 by record linkage to the Central Population Registry. With this method, we identified 202,339 unique addresses, each with a unique identification code composed of a municipality code, a road code, and a house number. The date the person had moved to and from the address was noted. The addresses were then linked to a database of all official addresses in Denmark, resulting in geographic coordinates for 198,758 (98%) of the cohort addresses. Subsequently, the addresses were mapped with the ArcGIS 9.1 geographic information system software (Environmental Systems Research Institute, Inc., Redlands, California, USA), and the proportion of addresses in each of the 271 Danish municipalities was calculated in relation to the total number of geocoded cohort addresses (Figure 1).
Figure 1
Distribution of geocoded cohort addresses (n = 198,758) in 271 Danish municipalities. The proportions are calculated as number of geocoded cohort addresses in each municipality divided by the total number of geocoded cohort addresses