Spatial Clustering of Childhood Acute Lymphoblastic Leukaemia in Hungary

The aetiology of childhood acute lymphoblastic leukaemia has been linked with spatially heterogeneous environmental exposures. The presence of spatial clustering would be consistent with geographically localized environmental exposures over long periods of time. The present study is the first to examine spatial clustering amongst children aged 0–4 years using population-based data from Hungary. The data set consisted of 134 children diagnosed with acute lymphoblastic leukaemia who were resident in part of Hungary during the period 1981–2000. Two levels of spatial aggregation were examined: counties and settlements. The Potthoff-Whittinghill and Moran I autocorrelation methods were used to test for spatial clustering. Additionally, an evaluation of the environmental changes during the study period was considered. Specifically analyses were carried out on sub-periods to investigate a possible effect of the Chernobyl catastrophe. There was statistically significant spatial clustering both at the county (estimate of extra-Poisson variation \( \left( {\hat{\beta}} \right)=0.56 \), P = 0.04) and settlement levels (estimate of extra-Poisson variation \( \left( {\hat{\beta}} \right)=0.68 \), P = 0.0003). At county level, the finding was attributable to clustering amongst female cases, but at settlement level, the finding was limited to male cases. There was significant spatial autocorrelation in the sub-periods immediately following the accident (1986–1990 & 1991–1995), but not before 1986, nor after 1995. A significant autocorrelation was observed during the 5 year period immediately following the accident (1986–1990, global Moran I = 0.1334, p = 0.005). The centre of significant excesses of ALL cases was located in the county of Baranya. Our study is consistent with an environmental aetiology for acute lymphoblastic leukaemia in children associated with constant exposure to an, as yet unknown, environmental factor in small geographical areas. Although a possible effect of the Chernobyl accident was found in the autocorrelation analysis, the role of chance cannot be excluded.