Genome-wide association studies (GWAS) have identified over two dozen loci associated with risk of prostate cancer in men of European ancestry. Although much work remains to be done to identify the causal alleles at these loci and understand their function, the current set of GWAS-identified marker alleles could be used to predict individualized risks of prostate cancer diagnosis, which might then be used to inform individuals’ screening and treatment choices. However, despite initial optimism about their clinical utility, genetic risk profiles for prostate cancer have had low discriminatory and predictive power.

We evaluated the performance of various risk models based on an expanded set of 24 SNPs known to alter prostate cancer risk in 9,109 prostate cancer cases and 9,045 matched controls from the NCI Breast and Prostate Cancer Cohort Consortium (BPC3). As we found no evidence of pairwise gene-gene interactions, we considered two models for the log odds of disease which assume additivity across loci: a simple sum of risk alleles carried and a step function by deciles of the weighted allele sum.

Using the allele sum model, the area under the ROC curve (AUC) was 0.598. Adding information about family history increased this estimate to 0.612. We observed no difference in AUC estimates when stratifying for age of onset (AUCs: ≤60 years: 0.603, 60-70 years: 0.604, 70-80 years: 0.593, >80 years: 0.552).

Per-allele increases in the odds of diagnosis with stage C or D cancer were similar to those for stages A and B (p=0.96), but the multiplicative increase in odds for high grade cancer (Gleason > 7) for each risk allele was attenuated by a factor of 0.97 relative to low grade cancer (p=0.03). We observed no interaction between family history of prostate cancer and number of risk alleles carried (P=0.83)

Compared to the 10th percentile, men in the highest 90th percentile are at a 3.6-fold risk (95% CI: 3.2-4.2) of developing prostate cancer assuming an allele sum model and at a 3.8-fold risk (95% CI: 3.3-4.3) assuming a weighted allele sum model.

These scores can be used to estimate absolute risks of prostate cancer. For example, the risk that a 40-year old European-ancestry man in the highest decile of genetic risk and a family history of prostate cancer will be diagnosed with prostate cancer by age 50 0.67%, or about 2.5 times the average. These risks could be used by individual men when weighing the benefits and risks of prostate cancer screening, although we caution that our predicted risks require external calibration.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 4726.