Our efforts in this area are illustrated by our support for the NCI Cooperative Family Registries (CFR), for which Dr. Thomas chairs an expert panel charged with providing methodologic advice on proposed family studies. The registry comprises over 10,000 families, identified in various ways through either a breast or colorectal cancer proband. Family histories, questionnaires on environmental exposures, and biological specimens for genotyping on their extended families have been collected and are being maintained in a central data base. These registries provide a rich resource both for characterizing already-identified genes and their interactions with environmental factors or other genes, as well as for discovering new genes. Many designs have been considered for addressing such questions - case-control studies with unrelated or family-member controls (Witte et al, 1999), familial cohort studies, case-case studies, etc.- as reviewed at a recent NCI workshop which Dr. Thomas helped organize and edit the proceedings (Schaid et al, 1999; Thomas 1999a), but there is still little consensus as to the optimal design for particular purposes. For example, Siegmund et al (1999d) considered multistage sampling designs based on different ways of exploiting the family history information as it is accrued. The theory of Horwitz-Thompson estimation equations has provided a theoretical basis for optimizing the sampling fractions currently being used in the USC/Dartmouth component of the colorectal registry (Haile et al, 1999). Similar techniques are being studied in a somewhat different context, a nested case-control study of the interaction of radiation and the ATM gene in a cohort of survivors of a first breast cancer (J Bernstein, PI, Mt Sinai School of Medicine) in which the design requires some form of multistage sampling because of the tremendous cost of abstracting treatment information and dose estimation using phantoms. The formal machinery we have developed under our grant on "survival models in genetic epidemiology" for investigating the bias and statistical efficiency of alternative designs (Thomas, 1999a) and for correcting for ascertainment (Langholz et al, 1999c) have proved to be the key to addressing such questions. Gauderman and Faucett (1997b) and Gauderman and Thomas (2000) have described some new approaches for linkage studies to detect genes which interact with environmental agents, and shown the potential increase in power that such approaches can have compared with traditional gene hunting methods which ignore environmental covariates. Dr. Thomas is currently working with CFR investigators on the design of studies of gene-environment interactions for breast cancer and genome-screens for new genes in colorectal cancer.

In addition to such design questions, we have been active in the development of new methods of analysis, particularly using such techniques as Markov chain Monte Carlo (MCMC) and Generalized Estimating Equations (GEE). The major focus of our analysis efforts has been on the development of methods for studying gene-environment interactions, particularly for age-dependent incidence of chronic diseases. For example, Gauderman et al. have reanalyzed data on lung cancer segregation for evidence of an interaction between smoking and the putative major gene: in a first analysis (Gauderman et al, 1997d), they found that the two factors combined multiplicatively, but after allowing for a strong modifying effect of age on the main effect of the major gene, a more-than-multiplicative interaction effect became apparent (Gauderman and Morrison, 2000). More recently, our efforts have turned to problems of gene mapping. For the 10th Genetic Analysis Workshop, we presented a Bayesian approach (using MCMC methods) to linkage mapping of an unknown number of major genes, allowing for environmental covariates (Thomas et al, 1997). For the 11th Genetic Analysis Workshop, we discussed linkage- and association-based approaches to detecting linkage in the presence of G´E interactions (Gauderman et al, 1999a; Siegmund et al, 1999a) and presented a GEE approach that was able to unscramble a complex pattern of G´E and G´G interactions involving both linked markers and candidate genes, including some three-way interactions (Thomas et al, 1999b). Dr. Thomas chairs one of the organizing committees for the 12th Genetic Analysis Workshop, which will be aimed at evaluating methods for joint linkage and linkage disequilibrium mapping.

The general themes of population-based research, gene-environment interactions, and incorporation of the underlying biology into the design and analysis of genetic epidemiology studies have guided all our research in this area. Indeed, these three themes were the basis of Dr. Thomas's Presidential Address to the International Genetic Epidemiology Society (Thomas, 2000).