Abstract 5508: Geographic clustering of melanoma malignum with high thickness in the United States

Abstract

Abstract Melanoma malignum is the 6th most common cancer in the United States in both genders with a continuously rising incidence. Therapeutic options for melanoma depend on the stage at the diagnosis; only if discovered early is this deadly disease curable. The proportion of thick melanomas (>2mm) is reported to have remained stable in the US during the last 2 decades, which assumes these cases are homogenously distributed in the country. However, there is evidence that tumor thickness varies (Breslow) as a function of geography, possibly based on factors including county level poverty and access to healthcare. Ultimately, how melanoma incidence and tumor thickness relate to geographic location of the US, and to the characteristics of the population and the environment within any given location, has not been determined. Our goal was to test for geographic clustering and risk factors for melanoma incidence and thickness at diagnosis across US. We used a population-based cancer registry for the US consisting of 17 registries of the National Cancer Institute's Surveillance Epidemiology and End Result (SEER) program for year 2008. Data consist of county level rates of cancer incidence across 7 states, representing 464 counties. The county-level covariates accessed from the 2000 US census included % of families below poverty and median income. First, global tests of spatial autocorrelation were conducted using the Moran's I statistic. Second, clustering was examined using the LISA statistic (local indicator of spatial autocorrelation). Third, geographically weighted regression was used to identify factors associated with the incidence and thickness of melanoma in US counties. Incidence rates of melanoma varied widely across US counties in 2008, ranging from 14-105/100,000. We found evidence of moderate positive spatial autocorrelation (Moran's I=0.166, p<0.05), with clustering of counties with high incidence rates observed in localized parts of northern Iowa, northern California, northern Washington, and southern Utah; counties with low incidence rates were clustered in New Mexico and Louisiana. Melanoma incidence was positively associated with county rurality and income was negatively associated with poverty. We also found that the median tumor thickness (Breslow, mm) at diagnosis varied widely across counties ranging from 0.0-4.5 mm (mean=0.7 SD=0.36), but did not exhibit global spatial autocorrelation. However, clusters of counties with deep median tumor depth were observed in Iowa, but not in counties in the US having the highest melanoma incidence. Counties with low median tumor depth clustered in New Mexico. Factors negatively associated with tumor thickness were income and poverty. Melanoma incidence and thickness do not vary randomly across the landscape of the US, but occur in well defined specific areas as a function of identifiable risk factors. This gives guidance for improvement of early detection of melanoma. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 5508. doi:1538-7445.AM2012-5508