Genetic variation and prostate cancer incidence and recurrence: How much is due to bias

Genes are hypothesized to influence prostate cancer, yet no germline variants or genetic profiles have been established as risk factors for incidence or recurrence after treatment. Understanding the role of genetic variation in prostate carcinogenesis is important for risk prediction and prognostication for treatment decision-making in men with prostate cancer. In addition, identifying and utilizing appropriate analytic techniques in epidemiologic studies is critical for estimating unbiased associations in these studies.;Using a case-control study nested in the Johns Hopkins Prostate Cancer Recurrence Cohort, we observed associations between single nucleotide polymorphisms (SNPs) in genes involved in the immune response, production and detoxification of reactive oxygen species, and repair of oxidative DNA damage, and recurrence risk after treatment. In particular, SNPs in the IL10 promoter region known to be high anti-inflammatory interleukin-10 producers and a common haplotype were associated with recurrence independent of other prognostic factors.;We simulated cohorts of at-risk men to quantify the extent of bias in prostate cancer genome-wide association studies (GWAS) resulting from the influence of SNPs on prostate specific antigen (PSA) concentration. We observed a positive bias in the association when the influence of a SNP on PSA was >0. This source of bias can inflate association estimates to the magnitude observed in published prostate cancer GWAS.;We simulated cohorts of men with prostate cancer to assess the bias and loss of precision in estimates of association between a SNP and recurrence due to missing genetic data in nested case-control studies that use incidence density sampling of controls matched to cases on strong predictors of recurrence. Assuming the mechanism of missingness was "completely at random" or "at random", we observed losses of precision for all analytic methods. Multiple imputation performed the best overall.;We observed associations between variants in genes involved in immunity and oxidation with prostate cancer recurrence. We provided evidence that studies of genes and prostate cancer are subject to detection bias when a SNP influences PSA concentration. Finally, we quantified the detrimental effect of missing data on estimates from matched case-control studies of genes and prostate cancer when using incidence density sampling of controls.