Replacing surrogate measures by direct quantification of ultraviolet radiation exposure in registry-based analyses of seasonality of melanoma diagnoses

Seasonal variation in melanoma diagnoses has been observed in numerous studies that used calendar time indicators. Depending on the latitude (and altitude) of the study region, the magnitude of seasonal and year-to-year variation of ultraviolet radiation (UVR) is neglected in these studies. An alternative approach comprises the direct incorporation of UVR measurements into such analyses. The aim of this investigation is a comparative evaluation of these approaches. The population-based Bavarian cancer registry recorded 11 901 incident melanoma cases between 2003 and 2008 that were used for the analysis. UVR intensity data for the same period were available from the solar radiation station at Munich-Neuherberg. Negative binomial regression modelling yielded adjusted relative risks (RR) controlled for year of diagnosis and age in 16 subgroups defined by sex, Breslow thickness and localization. Overall, the analyses showed slightly differing yet consistent results for exposure effects in subgroups. Melanoma evolving on the extremities showed the most pronounced association with increasing level of the UV index among men [e.g. RR=1.086, 95% confidence interval (CI) 1.054-1.119, and RR=1.102, 95% CI 1.046-1.161, for thin and thick melanoma on the upper limbs, respectively] and women (e.g. RR=1.088, 95% CI 1.058-1.119, and RR=1.056, 95% CI 1.010-1.103, for thin and thick melanoma on the lower limbs, respectively). Our analysis provides a benchmark for international comparisons and synthesis of epidemiologic evidence of seasonal variability in melanoma diagnoses. Future studies should use direct UVR measures to enable pooling of risk estimates and resolve remaining inconsistencies potentially resulting from latitudinal differences in exposure between international studies.