INCIDENCE OF CHILDHOOD LEUKEMIA AND NON-HODGKINS-LYMPHOMA IN THE VICINITY OF NUCLEAR SITES IN SCOTLAND, 1968-93

Objectives-The primary aims were to investigate the incidence of leuka emia and non-Hodgkin's lymphoma in children resident near seven nuclea r sites in Scotland and to determine whether there was any evidence of a gradient in risk with distance of residence from a nuclear site. A secondary aim was to assess the power of statistical tests for increas ed risk of disease near a point source when applied in the context of census data for Scotland. Methods-The study data set comprised 1287 ca ses of leukaemia and non-Hodgkin's lymphoma diagnosed in children aged under 15 years in the period 1968-93, validated for accuracy and comp leteness. A study zone around each nuclear site was constructed from e numeration districts within 25 km. Expected numbers were calculated, a djusting for sex, age, and indices of deprivation and urban-rural resi dence. Six statistical tests were evaluated. Stone's maximum likelihoo d ratio (unconditional application) was applied as the main test for g eneral increased incidence across a study zone. The linear risk score based on enumeration districts (conditional application) was used as a secondary test for declining risk with distance hom each site. Result s-More cases were observed (O) than expected (E) in the study zones ar ound Rosyth naval base (O/E 1.02), Chapelcross electricity generating station (O/E 1.08), and Dounreay reprocessing plant (O/E 1.99). The ma ximum likelihood ratio test reached significance only for Dounreay (P = 0.030). The linear risk score test did not indicate a trend in risk with distance from any of the seven sites, including Dounreay. Conclus ions-There was no evidence of a generally increased risk of childhood leukaemia and non-Hodgkin's lymphoma around nuclear sites in Scotland, nor any evidence of a trend of decreasing risk with distance from any of the sites. There was a significant excess risk in the zone around Dounreay, which was only partially accounted for by the sociodemograph ic characteristics of the area. The statistical power of tests for loc alised increased risk of disease around a point source should be asses sed in each new setting in which they are applied.