Multilevel analysis of solar radiation and cancer mortality using ecological data in Japan

A preventive effect of solar radiation on cancer has been suspected. This study aimed to compare the statistical relationship between solar radiation and cancer mortality according to hierarchical models and adjustment for confounding factors, and then to demonstrate the relationship with main site-specific cancer mortalities in Japan. We examined the relationship between all-site and main site-specific cancer mortalities and global solar radiation using Poisson regression with municipal data around 2000. The models included single-level (municipality) and multilevel (municipality and prefecture) with/without potential confounding factors (lifestyle and socioeconomic variables). For all-site cancer, single-level analysis showed a significant, strong negative association with solar radiation. However, multilevel analysis showed a moderate or no association. In multilevel analysis with potential confounding factors, solar radiation was significantly negatively associated with most site-specific cancers, but not with gallbladder and liver cancer in men and stomach and breast cancer in women. Our findings support the preventive effective of solar radiation on several types of cancer. However, to show a concrete relationship, a statistical model with an appropriate hierarchy and adjustment for potential confounding factors is required.