Childhood Leukaemia incidence around the Belgian nuclear sites: Surrogate exposure modelling.

Background: Health eects among populations living in the vicinity of nuclear installations have been a major area of concern for several decadesalready. The main focus is on childhood leukaemia. The dominant approach is an ecological study using residential proximity to the nuclear site, however complexradio-active discharge models have also been used. Data: The Belgian Cancer Registry provided data on Childhood leukaemia incidence by commune, age and sex. For the Flemish region this data spans theincidence years 2000-08; for the Walloon and Brussels-Capital Region the data span 2004-08. The study considers four nuclear sites. For each site, the FederalAgency for Nuclear Control used survey stations to measure wind direction and velocity for the period 2003-08.Methods: This paper describes an ecological study on leukaemia incidence in children living in the vicinity of nuclear installations in Belgium. Each nuclear sitewas treated as a point-source and single-site focussed hypothesis tests were used to test for a gradient in childhood leukaemia incidence with residential proximityto the site. In addition, two other surrogate exposures were used: prevailing wind direction and simulated radioactive discharges. The hypothesis tests were com-plemented by estimating the shape of the exposure-response relationship using Generalized Additive Models.Results: For the nuclear power plants of Doel and Tihange, no evidence of a gradient in childhood leukaemia cancer incidence related to any surrogate exposurewas found. For the nuclear research- and industry sites of Mol-Dessel and Fleurus, individual tests were less uniform. Only for Mol-Dessel it can be concludedthat the evidence may be suggestive of a potential association with the site.Conclusions: The use of multiple measures of surrogate exposures and multiple statistical methods has added value when investigating a priori dened point-sources. Each surrogate has implicit assumptions and precision and each statistical method relies on its own assumptions. By combining and comparing theresults of multiple approaches, new insights can be gained. However, it should be noted that the results are still subject to the limitations of the ecological design.