Spatial association between thyroid cancer incidence and environmental carcinogen emissions in Korea

Background: Thyroid cancer is the most commonly diagnosed cancer in Korea, with a recent resurgence in incidence since 2016. Environmental and occupational exposures-including air pollutants such as nitrogen dioxide (NO₂), particulate matter (PM2.5), ozone (O₃), and industrial chemicals such as toluene and mercury-have been reported as potential risk factors. This study examined whether regional variation in thyroid cancer incidence across Korea can be explained by differences in community-level carcinogenic emissions. Methods: An ecological study was conducted across 229 districts in Korea, using thyroid cancer incidence (ICD-10: C73) from 2014-2018 as the outcome. Exposure variables were district-level annual emissions of carcinogenic substances reported in the Pollutant Release and Transfer Register (PRTR), aggregated into three latency-reflective periods (2004-2008; 2009-2013; 2004-2013). Missing emission values were spatially interpolated using Bayesian spatial assumptions. Covariates included demographic (sex ratio, proportion aged 25-44), behavioral (smoking, obesity), and socioeconomic indicators (number of physicians per 1,000 population, basic livelihood recipient rate). Bayesian hierarchical Poisson models with BYM2 spatial structure were fitted using Integrated Nested Laplace Approximation (INLA). Model fit was assessed using WAIC, and relative risks (RRs) were estimated for each exposure period. Results: Higher district-level carcinogen emissions were associated with increased thyroid cancer incidence. In baseline models, a 1-unit increase in log-transformed emissions corresponded to RRs ranging from 1.037 to 1.051 across exposure periods. Spatially interpolated emission datasets showed similar associations (RR 1.037-1.041). In adjusted models controlling for demographic, behavioral, and socioeconomic factors, associations remained directionally consistent, with RRs between 1.023 and 1.028 for raw emission data and 1.024-1.025 for interpolated data. Spatial clusters of high relative risk were identified in urbanized and industrialized regions. Conclusions: Districts with higher emissions of carcinogenic substances exhibited increased thyroid cancer incidence after accounting for spatial dependence and community-level covariates. These findings support the hypothesis that environmental pollution from industrial facilities contributes to regional disparities in thyroid cancer and that latency windows of 5-10 years are relevant. Future research should incorporate refined exposure assessments at the individual level, improved temporal matching, and linkage to targeted prevention or screening strategies for high-risk regions.