This project will develop new approaches for evaluating cancer clusters using case-only data. To date three major deficiencies of studies of space-time clustering are that they assume the place-of-residence at time of observation (e.g. diagnosis or death) is representative of place-based exposures over the life course;do not routinely assess the sensitivity of the results to geocoding error;and do not assess sensitivity of the results to specification of cancer latency. These limitations are overcome by this project. Case-only data describing place of residence and date of diagnosis/death, gender, race, age, cancer treatment, sources of comorbidity, tumor stage, and contextual variables associated with residential location such as extent urban (e.g. Beale index), and a wealth of other data are commonly available in state cancer registries. These data will be coupled with commercially available residential histories (whose accuracy was validated in Phase I using data) to determine: (1) Whether, where and when statistically significant space-time clusters arise;(2) Whether such clusters are attributable to known covariates, cancer treatments, and contextual variables;(3) How sensitive the results are to geocoding error that may vary across the urban-rural continuum;(4) Those locations, cases, local neighborhoods and dates most sensitive to geocoding error;and (5) estimates of cancer latency most likely to explain observed space-time clustering. This approach is applied to pancreatic cancer in Michigan to determine whether space-time clustering of incident cases may be explained by Hepatitis-B. This novel analytical approach is a significant advance over currently used methods that ignore residential mobility, geocoding error and cancer latency. The major innovation is the creation of methods and software for analyzing cancer case- data to accurately identify space-time cancer clusters while accounting for residential mobility, geocoding error, and specification of cancer latency.
Principal Investigator: Jacquez, Geoffrey M. Case-only Cancer Clustering for Mobile Populations Relevance: The techniques and software from this project will provide a more concise and accurate description of space-time cancer clusters using data readily available in the nation's cancer registries via (1) coupling case-level data with commercially available residential histories in an easy to use, web-based interface;(2) the automated evaluation of the sensitivity of the results to geocoding error for the specific geography, cancer and sub-population being scrutinized by the software user;(3) the automated estimation of the range of cancer latencies most likely to explain observed space-time clusters;and (4) Novel space-time cluster statistics that appropriately account for residential mobility, know risk factors, covariates and cancer latency. To our knowledge the techniques and software from this project will be the first to address all of these factors within a single, comprehensive framework.
