Neighborhood characteristics, testosterone metabolism genotypes, and prostate cancer etiology.

Fifth AACR International Conference on Frontiers in Cancer Prevention Research, Nov 12-15, 2006 B194 Neighborhood characteristics such as poverty and educational attainment have been shown to influence various health outcomes, including cancer risks and outcomes. Individual characteristics including screening, genetics, lifestyle, and exposures also contribute to cancer etiology and outcomes. However, there have been few studies of multilevel influences on cancer etiology or outcomes that consider both neighborhood and individual factors simultaneously. Thus, we evaluated the joint relationship of socioeconomic status and educational attainment of residential census tract with genetic effects on prostate cancer using 1,002 Caucasian prostate cancer cases and 387 Caucasian controls. We geocoded the addresses of each participant to the census tract level using 2000 U.S. census data, and determined median cutpoints of census tract-specific values for analysis, including low per capita income (<$25,319), high poverty (<4.9%), and low high school graduation (<88.3%) . Genotypes of the androgen receptor (AR) CAG repeats (< 20 vs. ≥ 20) and SRD5A2-V89L or SRD5A2-A49T variants were studied. Genotype-neighborhood interactions were evaluated using logistic regression models, adjusted for age and participants' educational attainment. Genotype main effects on prostate cancer occurrence were in the expected directions based upon previous reports. However, genotype effects were observed primarily in men residing in the least deteriorated census tracts. We observed a significant interactions between SRD5A2-V89L for high-stage case status between high (OR=0.57, 95% CI=0.38-0.86) and low (OR=1.27. 95% CI=0.76-2.11) per capita income census tracts (interaction p-value=0.017). In addition, we observed a significant genotype-neighborhood interaction for AR-CAG genotypes among men living in areas with a higher percent of high school graduates (OR=0.55, 95%CI=0.32-0.95) compared to men residing in tracts with a lower percent of high school graduates (OR=1.23 95%CI=0.72-2.10; interaction p-value=0.040). Our findings confirm previous associations of testosterone metabolism genes with prostate cancer, and suggest that these genetic effects may be modulated by neighborhood characteristics including poverty and educational attainment. The etiological explanation for these neighborhood effects may include exposure to stress or other unhealthy lifestyle factors, for which neighborhood characteristics are a surrogate. These results therefore suggest that studies of genotype-environment interactions of cancer etiology or outcomes may expand the definition of "environment" to include both individual and neighborhood characteristics.