Case-only Cancer Clustering for Mobile Populations
R44 Phase II (End date 06/30/2011)
As stated in the original proposal, the following four aims will be accomplished in Phase II:
- Build and test a complete software package based on results of the prototype, to be implemented as a module in BioMedware’s Space Time Intelligence System (STIS), providing methods for the visualization and analysis of space-time clustering in cancer cases accounting for residential mobility and providing for the estimation of cancer latency and induction periods.
- Further refine the methodology to include Bayesian techniques for estimating cancer latency and induction periods.
- Conduct case studies of pancreatic cancer in Michigan to evaluate hypotheses regarding the role of Hepatitis B infection.
- Disseminate knowledge of the new approach by presentations at scientific meetings, and by disseminating the materials to organizations engaged in community-based participatory research.
Studies and Results
Great progress was made in the first year towards accomplishing Aims 1-4, as indicated by achievement and/or progress on the following milestones.
- Development and preparation of a manuscript (to be submitted, attached) evaluating the accuracy of residential history data obtained from ChoicePoint (a commercial vendor of individual-level data) with data collected by survey within a case-control study of bladder cancer in southeastern Michigan. Residential mobility histories are required for the methods developed in this research project, yet are difficult to obtain in the United States. This contrasts with the situation in Scandinavian countries (such as Denmark), which routinely records residential histories as part of the population registry. This milestone evaluated the accuracy of residential histories in the US obtained from commercially available data.
- Publication of a geocoding error sensitivity analysis of cancer clusters using simulated errors representative of those routinely encountered in human health studies (Jacquez and Rommel 2009). The propagation and assessment of positional and temporal error in residential mobility data is a significant advance for this project since such methods are required to assess uncertainties in the space-time interaction statistics that account for residential mobility. Notice that the current “state-of-the-art” rountinely ignores geocoding positional error.
- Publication of a book chapter on disease clustering for mobile individuals (Jacquez and Meliker 2009). This further advances the dissemination Aims planned for Phase II.
- Implementation of a new module within the TerraSeer STIS software for the analysis of spacetime interaction in disease cases for mobile individuals. These are the Vesta and Janus statistics (Jacquez, Meliker et al. 2007) developed in Phase 1 and implemented in Phase II.
- Obtained Human Subjects approvals from Western IRB and the Michigan Department of Community Health needed to conduct the Phase II study of pancreatic cancer and Hep-B infection status in southeastern Michigan. Prepared and submitted the data request to the Michigan Cancer Registry. The data obtained from this request will be used in project Aim 3.
- Development and use of a 2 hour lecture and 2 hour lab module on the methods that was used in the course “NRE 543 Advanced Space-Time Data Analysis for Health and Environment” that was taught by Dr. Jacquez in Fall 2009 at The University of Michigan, Ann Arbor.
- Presentation of research results at 3 conferences including an invited presentation at the GeoMed conference in Charleston SC (Nov 2009); the keynote address at the Centers for Disease Control and Prevention GIS-Week (Atlanta, GA, Nov 2009), and the Association of American Geographer’s 2010 annual meeting (Washington DC, April 2010). This further advances the dissemination objective of project Aim 4.
The substantial benefit of this research is its utility in accessing, mapping and analyzing local populations defined by excess risk using case-only data while accounting for residential mobility, cancer latency, and geocoding error; and by linking these local populations with diverse environmental, local and regional data such as business histories, pollution maps, and demographics. The ability to make sound, scientific inferences from case-only data while accounting for residential mobility, geocoding error and latency is a significant advance in the geographic analysis of cancer. Tools from this project ultimately will allow cancer control professional to accurately quantify space-time cluster metrics needed to target interventions and to assess achievement of cancer control objectives.
The successful completion of Research Aims 1 through 4 will be accomplished through:
- Additional software development effort in project year 2 that will continue to validate, test and undertake software quality assurance analysis on the case only clustering module in STIS.
- Implementation of the Bayesian model of cancer induction and latency period that is to be completed in project year 2 by collaborator Dr. Andrew Lawson.
- Analysis of the pancreatic cancer incidence and Hep-B dataset that is being prepared by the Michigan Department of Community Health. This analysis will evaluate the hypothesis that prior infection by Hepatitis-B may be implicated as a risk factor for pancreatic cancer later in life.
- Dissemination activities will continue in 2010 and 2011 will include presentations at scientific meetings, publication of the attached draft manuscript, and preparation and publication of the Bayesian models of disease latency currently under development.