Domestic radon exposure and childhood cancer risk by site and sex in 727 counties in the United States, 2001-2018.

Background: Childhood cancer has few established risk factors and environmental influences are underexplored. This ecologic study investigated the association between domestic radon exposure and childhood cancer risk in a large sample of United States (U.S.) counties.

Methods: Monthly ZIP code-level basement radon estimates from a geographic machine learning model were aggregated annually to counties, analyzed as continuous and dichotomized (cut point: 74 Bq/cubic meter (Bq/m) or 2.0 picocuries/L (pCi/L)) versions, and lagged by one year. Annual county-level counts of sex- and site-specific (all, leukemia, brain and central nervous system [CNS], and other sites) incident cancer diagnoses among those 0-19 years from 2001 to 2018 were obtained from the National Cancer Institute's Surveillance, Epidemiology, and End Results Program database. Sex- and site-specific counts were modeled as zero-inflated Poisson distributions in a Bayesian spatiotemporal framework and sequentially adjusted for random and fixed confounder effects.

Results: In 727 counties across 14 states, the average population aged 0-19 years was 41,599 people at baseline. Results from fully adjusted spatiotemporal statistical models indicated 1.05 (95% credible interval, CrI: 1.00, 1.09) times higher relative risks (RRs) of leukemia among both sexes and a RR of 1.06 (95%CrI: 1.00, 1.12) in males from a 50 Bq/m (1.35 pCi/L) increase in radon concentration the year prior. For radon exposures ≥74 Bq/m (2.00 pCi/L) the year prior, RRs were 1.08 (95%CrI: 1.02, 1.15) for both sexes and 1.12 (95%CrI: 1.04, 1.22) for females. No associations were found with other cancer sites or sexes from prior year radon exposures.

Conclusions: County-level childhood leukemia risk in both sexes were associated with average radon levels below U.S. Environmental Protection Agency guidelines recommending mitigation (148 Bq/m or 4.00 pCi/L). These findings warrant further investigation using population-based and individual-level study designs.