Methodology

In search of induction and latency periods: Space-time interaction accounting for residential mobility, risk factors and covariates

Geoffrey M Jacquez^1,2 , Jaymie Meliker¹ and Andy Kaufmann¹

¹BioMedware, Ann Arbor, USA

²Department of Environmental Health Sciences, The University of Michigan, Ann Arbor, USA

author email corresponding author email

International Journal of Health Geographics 2007, 6:35doi:10.1186/1476-072X-6-35

Published:	23 August 2007

Abstract

Background

Space-time interaction arises when nearby cases occur at about the same time, and may be attributable to an infectious etiology or from exposures that cause a geographically localized increase in risk. But available techniques for detecting interaction do not account for residential mobility, nor do they evaluate sensitivity to induction and latency periods. This is an important problem for cancer, where latencies of a decade or more occur.

Methods

New case-only clustering techniques are developed that account for residential mobility, latency and induction periods, relevant covariates (such as age) and risk factors (such as smoking). The statistical behavior of the methods is evaluated using simulated data to assess type I error (false positives) and statistical power. These methods are applied to 374 cases from an ongoing study of bladder cancer in 11 counties in southeastern Michigan, and the ability of the methods to localize space-time interaction at the individual-level is demonstrated.

Results

Significant interaction is found for induction periods of ~5 years and latency ~19.5 years. Data are still being collected and the observed clusters may be attributable to differential sampling in the study area.

Conclusion

Residential histories are increasingly available, raising the possibility of routine surveillance in a manner that accounts for individual mobility and that incorporates models of cancer latency and induction. These new techniques provide a mechanism for identifying those geographic locations and times associated with increases in cancer risk above and beyond that expected given covariates and risk factors in geographically mobile populations.