Skip to main content
Premium Trial:

Request an Annual Quote

Adding Genetic Information to PSA Improves Biomarker's Performance in Prostate Cancer Diagnosis

NEW ORLEANS – Applying a genetic risk score to prostate-specific antigen (PSA) measurements could improve the biomarker's sensitivity and specificity, according to research presented Monday during the annual meeting of the American Association for Cancer Research (AACR).

PSA is one of the most widely used prostate cancer biomarkers, both for screening and diagnosis and for assessing drug efficacy in clinical trials. But the measurement is imperfect, particularly since multiple factors beyond prostate cancer — including age, infection, and the presence of other conditions — can influence PSA levels, too.

"Observed PSA value is a little bit noisy," Linda Kachuri, a postdoctoral scholar at the University of California San Francisco, explained in a presentation for the press Monday morning. "It can reflect variation in PSA due to cancer, but it can also capture other factors that influence PSA … [including] genetics."

Indeed, using PSA as a diagnostic biomarker can result in over- and under-diagnoses and inaccurate classification of a cancer's aggressive potential — so much so that the United States Preventive Services Task Force has given the biomarker a "C" grade, meaning it is not recommended as a tool for systematic population-level screening, given its poor sensitivity and specificity. But because PSA has an established foothold in the prostate cancer diagnostic space despite these shortcomings, it is worth asking how its performance might be improved rather than replaced.

This is the exact question that Kachuri and colleagues aimed to address in their research.

As a first step, the team conducted a genome-wide association study (GWAS) of PSA levels in upwards of 95,000 men — the largest of its kind, according to Kachuri — all of whom had not been diagnosed with prostate cancer. "That part is important, because we would like to capture the genetics of a constitutive PSA elevation," she said.

The goal of the GWAS, in other words, was to home in on genetic variations associated with elevated PSA levels in the absence of cancer. The researchers procured data on the men from the UK Biobank, BioVU, PLCO, and Kaiser Permanente cohorts, identifying 128 different signals for PSA across the genome, 82 genes of which were novel, that had a statistically significant association with PSA levels in men without prostate cancer.

Once they'd identified these genes via their GWAS, Kachuri and colleagues used them to develop a polygenic risk score for elevated PSA, which was a weighted sum of an individual's genotypes of the variants of interest. They subsequently validated a new and improved genetically informed biomarker — PSAG — through a retrospective analysis of two prostate cancer prevention trials, the Prostate Cancer Prevention Trial (PCPT), from which they analyzed 5,737 participants, and the Selenium and Vitamin E Cancer Prevention Trial (SELECT), from which they analyzed 22,247. In PCPT and SELECT, the PSAG explained 7.3 percent and 8.7 percent, respectively, of variation in baseline PSA. The score was not associated with prostate cancer status in either study, reaffirming that it predicts for benign PSA variation.

"Rather than using this genetic information directly, as I think a lot of polygenic risk scores do, we actually wanted to use the genetic information to adjust PSA levels," Kachuri said. "The idea is that every person has a particular value for this genetic score ... and we'd like to account for that information in their PSA value, [since] someone's adjusted PSA score could either be higher or lower based on their genetics."

To assess the adjusted biomarker's performance, the researchers took real-world data on biopsy referrals from the Kaiser Permanente cohort and retrospectively reclassified the PSA thresholds used for referrals, incorporating the genetic score. The idea here was that, by using the PSAGscore instead of PSA alone, unnecessary biopsies could be avoided. According to their reclassification, roughly one-fifth of negative biopsies could have been avoided with the PSAG score, meaning that correcting the PSA analysis using genetics improved referral accuracy.

Beyond avoiding unnecessary biopsies, the researchers wanted to see if their score might help better identify aggressive prostate cancer diagnoses, defined as those with a Gleason score of seven or above, a PSA of at least 10 ng/mL, stage T3 to T4, or cancers with distant or nodal metastases. Here, measuring baseline levels using the PSAG score, again using retrospective data from PCPT and SELECT, resulted in a better prediction of aggressive prostate cancer than PSA alone.

"The classification performance of adjusted PSA exceeds the performance of directly measured, observed PSA and actually exceeds the performance of a genetic risk score for prostate cancer," Kachuri said, adding that the best prediction performance of all came when the researchers combined the PSAG biomarker with a previously established, 269-gene prostate cancer risk score published in Nature Genetics last year and used it on the PCPT data.

Going forward, Kachuri emphasized that it will be crucial to validate this finding in additional patient populations, especially because men in the cohorts that informed this study were largely white with European ancestry.

"In our subsequent efforts, we're really trying to focus on having larger and much more diverse studies, so we can comprehensively examine PSA genetics in individuals of all ancestries to really represent our target patient population," she said.

And down the line, should researchers validate this score further and in more representative patient populations, Kachuri said she is "quite optimistic" about the potential routine applications of this biomarker, given increased adoption of genetic testing. "This is something that could potentially be easier to translate than some other applications of genetic risk scores," she said, explaining that using the score as a test would require genotyping rather than whole-genome sequencing, making it "not as expensive and easier to implement." Ultimately, because so many clinicians are already familiar with PSA — the backbone of this score — she suspected that clinician hesitancy to adopt the score could be less of a barrier than other genetic testing applications.

As researchers and drugmakers design biomarker-driven clinical trials using PSA, the score may also help improve treatment precision.

"Genetic correction of PSA could be used as an alternative way of looking at PSA as a biomarker even for trying to potentially select individuals for clinical trials," Kachuri said. "Pretty much at any point that PSA is being used, the genetic correction could potentially be very helpful."