Purpose: Childhood cancers are rare and little is known about their etiology. Potential risk factors include environmental exposures that might implicate spatial variation of cancer risk. Previous studies of spatial clustering have mainly focused on childhood leukemia. We investigated spatial clustering of different childhood cancers in Switzerland using exact geocodes of place of residence.
Methods: We included 6,034 cancer cases diagnosed at age 0-15 years during 1985-2015 from the Swiss Childhood Cancer Registry. Age and sex-matched controls (10 per case) were randomly sampled from the national censuses (1990, 2000, 2010). Geocodes of place of residence were available at birth and diagnosis for both cases and controls. We used the difference in k-functions and Cuzick-Edwards test to assess global clustering and Kulldorff's circular scan to detect individual clusters. We also carefully adjusted for multiple testing.
Results: After adjusting for multiple testing, we found no evidence of spatial clustering of childhood cancers neither at birth (p = 0.43) nor diagnosis (p = 0.13). Disregarding multiple testing, results of individual tests indicated spatial clustering of all childhood cancers combined (p < 0.01), childhood lymphoma (p = 0.01), due to Hodgkin lymphoma (HL) (p = 0.02) at diagnosis, and embryonal tumors of the central nervous system (CNS) at birth and diagnosis, respectively (p = 0.05 and p = 0.02).
Conclusions: This study provides weak evidence of spatial clustering of childhood cancers. Evidence was strongest for HL and embryonal CNS tumors, adding to the current literature that these cancers cluster in space.
Keywords: Cancer clusters; Cancer registry; Hodgkin lymphoma; Medulloblastoma; Primitive neuroectodermal tumors.