Assessment of ecological regression in the study of colon, breast, ovary, non-Hodgkin's lymphoma, or prostate cancer and residential UV

Abstract

Recent ecological studies have suggested a possible association between exposure to ultraviolet-B (UVB) radiation and reduction in the risk of various cancers; however, ecological studies are known to be subject to bias. The objective of this study was to demonstrate difficulties with the ecological approach. We conducted a multicountry ecological study using cancer incidence rates, residential UV levels, dietary intake, and different sociodemographic variables for 38 locations spanning 33 countries worldwide. The effect of residential UV exposure on cancer incidence was assessed using multiple linear regression models. The results of our multivariate analyses show no indication of an inverse association between residential UV levels and the risk of colon, non-Hodgkin's lymphoma (NHL), ovarian, prostate, or breast cancer in women. For colon cancer and NHL, a significant positive association was calculated. The rates of melanoma, which were used to examine the methods of this study, showed a strong and significant (P<0.01) association with solar radiation. Our results provide no evidence to support previous ecological results that UV exposure may reduce the risk of NHL, colon, breast, ovary, or prostate cancer. The study demonstrates the high sensitivity of ecological studies to adjustments for various confounders, and casts doubts on results of ecological analyses in this field.