Assessing Prostate Cancer Risk: Can Genomics Help?


Prostate cancer screening has been the subject of much debate, given the recent recommendation by the U.S. Preventive Services Task Force against utilizing prostate-specific antigen for screening asymptomatic healthy males. Other organizations, including the American Cancer Society, recommend that men be offered PSA screening after an informed decision of the pros and cons of its use as a screening test.

About one in six men will be diagnosed with prostate cancer during their lifetime. For many, the disease will remain indolent. Identifying men who are at high risk for an aggressive form of the cancer or those who will develop it at an earlier age remains a challenge.

Dr. Peter Hulick

Clearly, we need to improve screening options and our ability to accurately identify patients whose risk for aggressive prostate cancer is high. This seems like an ideal situation for genetics and genomics to help stratify risk and to guide treatment or screening interventions based on an individual’s risk for developing an aggressive tumor.

The identification of high-risk genes for prostate cancer has proved difficult. We do not have highly penetrant and relatively common genes for prostate cancer, similar to, for instance, the BRCA1/2 genes among families at high risk for breast and ovarian cancers. Some discoveries have been achieved, including the identification of a rare mutation in HOXB13. But such discoveries have so far provided answers for only a minority of families and patients (N. Engl. J. Med. 2012;366:141-9).

Given the challenge of identifying single genes that confer a high risk for developing disease, prostate cancer research has focused on detecting weaker genetic markers that in aggregate could potentially help explain why certain men face a higher risk for developing prostate cancer, or a more aggressive subtype of cancer.

Genomewide association studies (GWAS) have been conducted and single nucleotide polymorphisms (SNPs) "panels" have been designed to help predict the risk of prostate cancer. These panels have led to commercial testing that is either physician directed or geared directly to the consumer. Most of the SNPs on these panels have low odds ratios (less than 2.0), so individually, they are not helpful for predicting significant disease risk or the likelihood of having aggressive disease. However, if multiple SNPs are aggregated on a panel and tested, a clinically useful picture could – in theory – be created.

In July, as part of the EGAPP (Evaluation of Genomic Applications in Practice and Prevention) project, the Agency for Healthcare Research and Quality issued a final report on the current evidence of the "validity and utility of using SNP panels in the detection, diagnosis, and clinical management of prostate cancer." The extensive review identified 15 distinct SNP-based prostate cancer risk panels, including those marketed by Proactive Genomics LLC and deCode Genetics.

How well did these SNP panels perform in stratifying future risk or screening for current disease? Screening performance is often reported by generating an ROC (receiver operating characteristic) curve and measuring the AUC (area under the curve). Typically an AUC of at least 75% is necessary for the test to be considered clinically useful. AUCs for a 5-SNP panel ranged from 58% to 73%. The conclusion: There was little incremental gain over non-SNP–based models of prediction, and therefore there was little evidence that they improved risk stratification.

Could the SNP panels discriminate between clinically significant and indolent prostate cancer? Four panels ranging in size from 5 to 35 SNPs were evaluated, and none of the panels was able to reliably distinguish between more- or less-aggressive tumors.

As for prognosis prediction, a 5-SNP panel (with and without inclusion of family history), a 6-SNP panel, and a 16-SNP panel were used to predict mortality in men who had prostate cancer. Follow-up periods ranged from 3.7 to 10 years, depending on the study. Again, there was no evidence that the SNP panels improved the prediction of mortality – and thus prognosis – even when the information gained from the panel was added to models that included conventional prognostic factors (age, PSA, Gleason score, and tumor stage).

Given the limitations of PSA screening for detecting and determining the aggressiveness of prostate cancer, physicians hoping for a better screening tool may be tempted to utilize a genomics-based risk profile test such as an SNP panel. To date, unfortunately, the SNP-based risk models for prostate cancer risk assessment have not helped us to reliably distinguish between aggressive and nonaggressive disease, nor have they identified high-risk patients. Thus, such testing should not be utilized in clinical settings outside of research protocols.

Next Article: