Polygenic risk score does not improve prediction of aggressive prostate cancer

A prostate cancer polygenic risk score did not better predict the risk for aggressive disease compared with a clinical risk predictor and may not be clinically useful, according to study results published in JAMA Internal Medicine.

Currently, prostate cancer screening instruments commonly end up with patients undergoing biopsies that find either no cancer or cancer that is clinically insignificant, Kerry R. Schaffer, MD, MSCI, an assistant professor at Vanderbilt University, and colleagues wrote.

But there is some interest in using a polygenic risk score, which can measure “the burden of risk-associated single-nucleotide polymorphisms” that a patient carries to improve the risk prediction of prostate cancer.

“Although polygenic risk scores have been associated with prevalent and incident disease, prior studies have shown that they do not improve risk prediction for aggressive prostate cancers with the potential to be fatal,” they wrote.

So, the researchers conducted a retrospective cohort study to compare the discriminative ability of a prostate cancer polygenic risk score — PRS269 — with the Prostate Biopsy Collaborative Group (PBCG) risk calculator, a clinical tool that evaluates the probabilities of finding cancer on the results of a man’s first prostate biopsy.

They used Vanderbilt University Medical Center’s private DNA biobank to identify the participants: 655 men aged 40 to 80 years who did not have prostate cancer and underwent a biopsy.

Schaffer and colleagues found that, compared with the contemporary clinical risk predictor, the polygenic risk score did not improve risk prediction of aggressive prostate cancer.

Prostate cancer was identified on the biopsy results of 52% of men, 27% of whom had high grade cancer — grade group (GG) of 2 or higher, the researchers wrote. The PRS269 risk score was associated with a finding of any cancer on biopsy with an age-adjusted odds-ratio per standard deviation increase of 2 (95% CI, 1.6-2.4).

Notably, when added to the PBCG predictor, the PRS269 improved discrimination for any cancer, but not GG of 2 or higher cancer.

“Although the PRS269 improved model discrimination for all cancers, improvement was less than has been observed for other validated prostate cancer biomarker predictors such as the Prostate Health Index,” the researchers wrote.

In a related commentary, Robert J. Klein, PhD, an associate professor at the Icahn School of Medicine at Mount Sinai, wrote that, as currently formulated, the polygenic risk score “will not enhance clinical decision making.”

“There is a need to demonstrate accuracy and utility — defined for the relevant clinical context — before rolling such scores out on a large scale,” Klein wrote.

Current polygenic risk scores, he wrote, are designed based on the genetic association studies for diagnosing any prostate cancer, but a tool that does not identify aggressive disease may do more harm than good.

“A major reason population-level prostate-specific antigen (PSA)-based screening for prostate cancer is controversial is that many prostate cancers are indolent and could remain untreated without adverse consequences. Treatment of such cancers results in needless treatment-related morbidity,” Klein wrote. “To the extent that, like PSA, a polygenic risk score can only predict diagnosis of prostate cancer, it runs the same risk of overdiagnosis as observed for PSA-based screening.”

The most useful polygenic risk score, Klein added, would be able to identify those who will have clinically significant cancer, and by this measure, “the polygenic risk score did not outperform the PBCG calculator, suggesting it may not be useful in this context.”

“Just as PSA screening has never been shown to reduce all-cause mortality and is associated with harm, a polygenic risk score that is nonspecific for aggressive disease would likely cause harm while not reducing mortality,” he wrote.

References:

• Klein RJ, et al. JAMA Intern Med. 2023;doi:10.1001/jamainternmed.2022.6782.
• Schaffer KR, et al. JAMA Intern Med. 2023;doi:10.1001/jamainternmed.2022.6795.