Abstract
Background: Liver cancer rates are rising, particularly in minority populations. Risk factors for liver cancer, such as alcohol/drug use, metabolic disorders, and viral hepatitis B/C (HBV/HCV), are generally modifiable through lifestyle interventions, vaccinations, and treatments. However, resources are often limited and strategies to prioritize geographic areas most in need of cancer prevention are needed. Recommendations exist to focus prevention efforts on groups with the highest rates of liver cancer, namely Hispanics, Blacks, and individuals born 1950-1959 who are at high risk for HCV. We compare the sensitivity/specificity of this approach to a dual approach that couples liver cancer disease cluster statistics with neighborhood-level demographic data to inform liver cancer prevention.
Methods: Pennsylvania (PA) Liver Cancer Registry data from 2007-2014 were linked to the 2010 U.S. Census data via a geocode at the census tract level with ArcGIS software. Using the space-time scan statistic in SaTScan and relative risk estimates from BayesX, we also identified high-risk clusters or geographic areas with significantly elevated rates of incident liver cancer, adjusted for age, gender, and diagnosis year. Analyses were conducted using elliptical spatial windows and Poisson models. Census tracts in the top 80th percentile for percentage of Hispanics, Blacks or those born between 1950-1959 were also identified. The sensitivity, specificity, and positive predictive value (PPV) of a census tract being located in a high-risk cluster and/or testing positive or negative for at least one of three neighborhood variables (higher % Black, % Hispanic, 1950-1959 birth cohort) were calculated.
Results: There were 9,460 cases of liver cancer diagnosed in PA. Five high-risk clusters were identified (relative risks ranged from 1.83-3.73, all p<0.05). Of 3,217 census tracts in PA, 412 were located in one of the 5 high-risk clusters, whereas 1,596 were positive for at least one of 3 neighborhood demographic variables. Within high-risk clusters, 365 census tracts also had a higher percentage of the birth cohort, Blacks or Hispanics. While sensitivity was relatively high (88.66%), specificity (56.1%) and PPV (22.8%; i.e., the chance a census tract with at least one demographic variable truly was located in a high-risk cluster) were low.
Conclusions: Coupling disease cluster statistics with neighborhood demographic data refines the identification of areas that carry a greater than expected burden of liver cancer and reduces intervention targets more than neighborhood demographics alone. However, additional analyses are needed to improve the sensitivity/specificity of this combined geospatial approach. Consideration of other patient and neighborhood level socioeconomic data is also being explored and will be presented to further inform the prioritization of liver cancer prevention efforts.
This abstract is also being presented as Poster C014.
Citation Format: Shannon M. Lynch, Daniel Wiese, Kristen Sorice, Minhhuyen Nguyen, Evelyn Gonzalez, Kevin Henry. Geospatial analytics and sensitivity/specificity assessments to inform liver cancer prevention [abstract]. In: Proceedings of the Eleventh AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2018 Nov 2-5; New Orleans, LA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(6 Suppl):Abstract nr PR13.