Personalized PSA levels could improve prostate cancer screening

The most common screening test for prostate cancer — a measure of prostate-specific antigen, or PSA, levels — so often suggests cancer where there is none that clinical guidelines no longer recommend the test for men over 70 and leave the decision up to younger patients.  

Scientists at Stanford Medicine and their collaborators aim to make PSA screening more accurate — by calibrating PSA levels to each man’s genetics. Applying this type of personalization could significantly reduce overdiagnosis and better predict aggressive disease. Their research was published June 1 in Nature Medicine.

In addition to the regular blood-based PSA test, such personalized screening would require a germline genetic test, typically done on saliva, blood or cheek swab samples, to look for inherited genetic variants that affect PSA levels.

Elevated PSA levels can be a sign of a cancerous prostate tumor, but also can be caused by factors unrelated to cancer, such as inflammation, infection, an enlarged prostate or simply older age.

“Some men have higher PSA levels due to their genetics,” said John Witte, PhD, a professor of epidemiology and population health and of biomedical data sciences and the senior author of the study. “They don’t have cancer, but the higher PSA level leads to a cascade of unnecessary medical interventions like biopsy.”

By one estimate, less than one-third of men with elevated PSA levels were confirmed by a biopsy to have prostate cancer. Moreover, 15% of men with normal PSA levels were later found to have prostate cancer.

Turning down the noise

The trouble with current PSA screening can be compared to a signal-to-noise problem in engineering, in which the desired output is mixed with background noise, the researchers said.

“To improve the signal, which is the variation in PSA levels caused by a prostate tumor, we subtract out the noise, which in this case comes from genetics,” said Linda Kachuri, PhD, an assistant professor of epidemiology and population health and the lead author of the study.

To better understand the genetics behind normal variation in PSA levels, the researchers looked at the genomes and PSA levels of 95,768 men without prostate cancer. The data had been collected as part of earlier studies and included mostly men of European ancestry. 

Through their analysis, the researchers estimated that 30% to 40% of variation in PSA levels is in fact “noise” — that is, determined by genetics unrelated to cancer.

“Specifically, what we’re trying to capture are the genetic determinants of normal PSA variation,” Kachuri said. By separating out normal variations, the researchers hope to more accurately determine when a PSA test indicates prostate cancer.

“This is different from our usual research deciphering the genetic basis of cancer,” Witte said. “We want to remove the non-cancer-related part that’s making PSA a less specific biomarker.”

Personalized PSA

The researchers identified 128 sites in the genome that can affect a person’s inherent PSA level. They developed a way to calculate PSA that accounts for an individual’s normal genetic variations at these sites — known as a PSA polygenic score.

“A polygenic score is a quantitative way of summarizing someone’s genetic predisposition for a trait in a single value,” Kachuri said. In this case, the trait is a higher baseline PSA level.

The researchers then evaluated the PSA polygenic score on data from a separate group of nearly 32,000 men without prostate cancer. They found that the score could predict close to 10% of variation in PSA levels, though it was much more effective among men of European ancestry than among men of East Asian or African ancestry.

When the researchers applied their score to data from a group that included men with and without prostate cancer, as confirmed by biopsy, they found that roughly 30% of men could have been spared a biopsy.

The adjusted PSA levels particularly improved detection of the more aggressive forms of prostate cancer, although the benefit was noticeable only in men of European ancestry.

“What we’re really worried about are those aggressive cases, so the fact that we’re able to show that genetically adjusted PSA is more predictive of aggressive disease is really promising,” Kachuri said.

Improving the score

On the flip side, the adjusted PSA levels would have missed approximately 9% of positive biopsies. The majority of these overlooked cases were slow-growing tumors, which may not require treatment, but the misclassifications point to room for improving the score.

Because the polygenic score was developed using data predominantly from men of European ancestry, the team is working on a larger study in collaboration with the Million Veteran Program, which will include more men from diverse ancestral populations.

“Ideally, we want to come up with a single score that works well for everybody, across the spectrum of ancestry,” Kachuri said.

With a disease as prevalent as prostate cancer — one in nine men in the U.S. will be diagnosed with prostate cancer, and one in 40 will die from it — even a small improvement in screening could save lives.

“Prostate cancer is the most common cancer among men in the U.S.,” Witte said. “Because it’s so common, even though prostate cancer has a high survival rate, it’s the second leading cause of cancer death among men in this country.”

Researchers from UC San Francisco, the National Cancer Institute, Vanderbilt University Medical Center, the Fred Hutchinson Cancer Research Center, CHRISTUS Santa Rosa Medical Center Hospital, Memorial Sloan Kettering Cancer Center, Skåne University Hospital in Sweden, Kaiser Permanente Northern California, the University of Southern California and the Icahn School of Medicine at Mount Sinai also contributed to the study.

The study was funded by the National Institutes of Health/National Cancer Institute (grant R01CA24141002).