Time-related Factors in Cancer Epidemiology: This grant, currently in its fourth cycle, supports research in two methodologic areas: 1. the analysis of cancer epidemiology studies involving extended exposure histories using descriptive methods and stochastic models of carcinogenesis and toxicokinetics; 2. Case-control study methodologies, including nested case-control, case-cohort, unmatched case-control designs that incorporate available exposure information into the sampling to produce more efficient studies. It also supports applications of these new methods in a wide variety of cancer epidemiology studies, particularly those involving ionizing radiation, but also applicable to other environmental studies.

Methodologic Research in Genetic Epidemiology: Four grants support our efforts in the general area of study design and analysis methods aimed at gene-environment interactions. The grant entitled "Survival models in genetic epidemiology", now in its third cycle, was originally aimed at both study design and analysis issues in both gene discovery and gene characterization. As this work has matured, we have focused the original grant on study design issues in gene characterization, and gotten two additional grants to develop the other aspects. "Computational methods in genetic epidemiology" is aimed at developing joint linkage and linkage-disequilibrium approaches to gene mapping, with particular emphasis on Generalized Estimating Equations (GEE) and Markov chain Monte Carlo (MCMC) methods; the aim is to combine the tools of population genetics such as the coalescent, with the tools of genetic epidemiology such as linkage analysis, to provide more power for discovering genes. "Statistical approaches to the study of GxE interactions" aims at developing study designs and methods of analysis for incorporating environmental interactions into the gene discovery process, as well as for characterizing such interactions once the genes have been cloned. "Informatics Support for the CFRBCCS" is a subcontract from UC Irvine (Hoda Anton-Culver, PI) to coordinate an international expert panel to provide support to investigators in the Cooperative Family Registries for Breast and Colorectal Cancer Studies on design and analysis for studies based on this resource.

---

Statistical Methods in Air Pollution Epidemiology: Several grants have supported our statistical research on design and analysis issues related to the USC study of air pollution effects on children, including methods for optimizing the design of studies involving both ecologic comparisons between centers and within-center comparisons between individuals, methods of analysis combining both types of comparisons, methods of allowing for exposure measurement error, multilevel models for longitudinal data analysis, and flexible smoothing techniques for modeling the effects of age and exposure. This work was initially funded by grants from the Health Effects Institute and the California Air Resources Board (CARB) to Dr. William Navidi, and is currently supported by a supplemental contract from the CARB as a formal part of the larger Children's Health Study. Related work has been carried out in other contexts; for example, Dr. Thomas is a member of the Oversight Committee for the National Morbidity and Mortality Air Pollution Study (Samet, PI, Johns Hopkins University) and has co-authored a paper with them (Zeger et al, 2000) on the effects of measurement error in daily mortality studies.

---

Methods for Dealing with Exposure Measurement Error: The aims of this research are to develop methods for adjusting for the effects of measurement error on dose-response relationships, to explore design issues in studies which will use validation substudies to estimate measurement errors, and to apply these methods to a variety of studies, including diet and ionizing radiation. We have been applying these methods to the cohort of Colorado plateau uranium miners, which illustrate a number of unique features: extended exposure histories with the exposure rate being the primary source of misclassification, correlated errors due to the application of the same job-exposure matrix entries to different individuals, and many entries with no measurements available. We have developed an approach to correct for measurement errors in the job-exposure matrix using a hierarchical model of radon levels in mines within geographic regions and collaborated with the NAS BEIR VI committee on further analyses using error-corrected doses. Similar approaches would be applicable to studies of air pollution (as discussed above) and other environmental exposures.

---