Research Article: Incorporating Known Genetic Variants Does Not Improve the Accuracy of PSA Testing to Identify High Risk Prostate Cancer on Biopsy

Date Published: October 2, 2015

Publisher: Public Library of Science

Author(s): Rebecca Gilbert, Richard M. Martin, David M. Evans, Kate Tilling, George Davey Smith, John P. Kemp, J. Athene Lane, Freddie C. Hamdy, David E. Neal, Jenny L. Donovan, Chris Metcalfe, Neal Shore.


Prostate-specific antigen (PSA) testing is a widely accepted screening method for prostate cancer, but with low specificity at thresholds giving good sensitivity. Previous research identified four single nucleotide polymorphisms (SNPs) principally associated with circulating PSA levels rather than with prostate cancer risk (TERT rs2736098, FGFR2 rs10788160, TBX3 rs11067228, KLK3 rs17632542). Removing the genetic contribution to PSA levels may improve the ability of the remaining biologically-determined variation in PSA to discriminate between high and low risk of progression within men with identified prostate cancer. We investigate whether incorporating information on the PSA-SNPs improves the discrimination achieved by a single PSA threshold in men with raised PSA levels.

Men with PSA between 3-10ng/mL and histologically-confirmed prostate cancer were categorised as high or low risk of progression (Low risk: Gleason score≤6 and stage T1-T2a; High risk: Gleason score 7–10 or stage T2C). We used the combined genetic effect of the four PSA-SNPs to calculate a genetically corrected PSA risk score. We calculated the Area under the Curve (AUC) to determine how well genetically corrected PSA risk scores distinguished men at high risk of progression from low risk men.

The analysis includes 868 men with prostate cancer (Low risk: 684 (78.8%); High risk: 184 (21.2%)). Receiver operating characteristic (ROC) curves indicate that including the 4 PSA-SNPs does not improve the performance of measured PSA as a screening tool for high/low risk prostate cancer (measured PSA level AU C = 59.5% (95% CI: 54.7,64.2) vs additionally including information from the 4 PSA-SNPs AUC = 59.8% (95% CI: 55.2,64.5) (p-value = 0.40)).

We demonstrate that genetically correcting PSA for the combined genetic effect of four PSA-SNPs, did not improve discrimination between high and low risk prostate cancer in men with raised PSA levels (3-10ng/mL). Replication and gaining more accurate estimates of the effects of the 4 PSA-SNPs and additional variants associated with PSA levels and not prostate cancer could be obtained from subsequent GWAS from larger prospective studies.

Partial Text

Prostate cancer is the second most commonly diagnosed cancer and the 5th most common cause of cancer death worldwide for males [1]. Prostate-specific antigen (PSA) testing, followed by a prostate biopsy if the PSA level is raised (typically PSA 3-4ng/mL), is a widely accepted screening method for the disease. Early diagnosis and treatment may impact on survival in some men, but the majority of screen-detected prostate cancers are at low risk of progression, with potential harm caused by unnecessary diagnosis and treatment [2,3].

Our study of 868 men with raised PSA (3-10ng/mL) and histologically-confirmed prostate cancer has demonstrated that genetically correcting PSA for the combined genetic effect of four PSA-SNPs, did not improve upon the traditional single PSA threshold for biopsy when distinguishing high and low risk disease.