The shortcomings of prostate-specific antigen (PSA) as a screening test for prostate cancer have long been obvious to scientists, to clinicians, and most of all to men who have had elevated levels and faced difficult decisions about what to do next. A considerable scientific enterprise has been devoted to improving on or replacing the PSA test, which detects prostate disease but not necessarily prostate cancer. Even as studies continue and controversy roils around how to make the most of the PSA test, a large contingent of researchers is seeking better biomarkers for prostate cancer that take advantage of genomic technologies. However, even as some genomic tests are making their way into the marketplace, leading researchers have sounded a note of caution.

"It's a very daunting task, introducing a biomarker that's going to impact clinical care. That's really the important issue at hand," said Mark Rubin, MD, a professor of pathology and laboratory medicine at Weill Cornell Medical College in New York City. "In developing biomarkers for prostate cancer, a major clinical question to be addressed is distinguishing between aggressive versus indolent disease. Conceptually, this is easy and the terms indolent and aggressive are easily spoken, but are much more difficult to define."

Rubin spoke recently at the annual meeting of the American Association for Cancer Research (AACR) as part of a panel exploring prostate cancer screening now and in the future. He went on to add that next-generation sequencing has the potential to address the many demands placed on biomarkers for prostate cancer, from diagnosis all the way to personalizing drug therapies of the future (See Table, right). However, he also warned against being beguiled by the novelty of this technology if it doesn't decisively answer key clinical questions.

PSA's History

The U.S. Food and Drug Administration (FDA) first cleared the PSA test in 1986 to monitor treatment in men already diagnosed with prostate cancer and extended approval in 1994 for detecting cancer in symptomatic men. By that time, however, the PSA horse was out of the proverbial barn, and the test was being used widely, particularly in the U.S., for screening asymptomatic men.

Off-label use of PSA as a general screening test has been at the center of the controversy, for many reasons. One is that PSA, a glycoprotein produced by the epithelial cells of the prostate, can be elevated for reasons other than prostate cancer, including from prostatitis, trauma, or benign prostatic hyperplasia. As a consequence, the test is not very specific for detecting prostate cancer. At modest elevations its specificity is as low as 20%, but rises to as much as 50% at levels ≥10 µg/L.

This perhaps would not be such an issue if the prostate cancer field was not weighed down by over-diagnosis and over-treatment. American men have a 16% lifetime risk of developing the disease, but only about 3% die from it. As many as four-out-of-five prostate biopsies on men with elevated PSA levels turn out to be negative for cancer, and up to half the true-positives are unlikely to pose problems for men during their lifetimes. At the same time, both prostate biopsies and treatments pose risks: biopsies have infectious complication rates ranging from 0.6–4.1%, while radical prostatectomy leads to sexual function issues in an estimated 20–70% of men and urinary problems in 15–50%.

Despite these cautionary statistics, once men hear the word cancer, many still pursue treatment. The bias toward treatment in the U.S. contrasts sharply with other countries. For example, an estimated 20–30% of European men choose active surveillance compared with only about 5–10% of American men. This screen-and-treat mentality is beyond frustrating to Otis Brawley, MD, chief medical officer of the American Cancer Society (ACS), who also spoke at the AACR conference. "One of the problems I've had with this frenzy of prostate cancer screening and treatment over the last 25 years is that we've actually impeded scientific development of the new markers we desperately need to figure out the cancers that kill versus the cancers that don't kill," he contended.

Which Cancers Need to Be Treated?

Knowing which men have indolent disease that merely deserves active surveillance and which have aggressive cancer that must be acted on, has, in fact, been at the heart of the PSA debate. Quite a number of different strategies for using PSA and its various forms have been proposed to address this issue, including: PSA density, PSA velocity, PSA doubling time, free PSA (fPSA), proPSA, and complexed PSA. All of these have been found to be complementary to PSA in certain circumstances, and the FDA has cleared fPSA for use in men with total PSA in the diagnostic gray zone between 4–10 µg/L.

Different cutoffs also have been proposed to fine-tune PSA's sensitivity and specificity. A cutoff of 4 µg/L has been commonly adopted and used in major screening studies, and an ACS literature review found this threshold to have an estimated sensitivity of 21% for detecting any prostate cancer and 51% for detecting high-grade cancer. Lowering the cutoff to 3 µg/L boosts these sensitivities to 32% and 68%, respectively, but at the cost of specificity, which drops from 91% to 85%. However, lower thresholds have not been endorsed formally by professional organizations. In fact, the National Academy of Clinical Biochemistry Laboratory Medicine Practice Guideline on the use of tumor markers recommends not using age-specific reference intervals and notes that the reported benefits from lowering the clinical decision limit below 4 µg/L are too uncertain to make a blanket recommendation (Clin Chem 2008:54:e11–79).

A Lack of Clarity

Researchers and clinicians who hoped clarity about the role of PSA in screening would come from two major clinical trials—the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO), and the European Randomized Study of Screening for Prostate Cancer (ERSPC)—were disappointed. PLCO, a study of 77,000 men at 10 U.S. centers, did not identify a significant reduction in prostate cancer-specific mortality. However, this study has been criticized because half of the men assigned to the control group also underwent screening by virtue of receiving usual care. "PLCO does not provide an answer to value of screening—it compares screening to the current U.S. practice of screening," contended AACR panelist, Fritz Schröder, MD, PhD, a professor of urology at Erasmus University in Rotterdam, the Netherlands and one of the lead ERSPC investigators.

The initial report from ERSPC involving 182,000 subjects from seven countries, found the death rate of prostate cancer was 20% lower in the screening versus control arm, but 1,410 men would need to be screened and 48 cases of prostate cancer detected to prevent one death from the disease. The second report, published in March 2012 and reflecting 2 more years of follow-up, consolidated the authors' prior findings that PSA significantly reduced prostate cancer-related mortality—a 29% relative risk reduction—but did not affect all-cause mortality, the crucial criterion for population-based screening.

More recently, ERSPC researchers examined quality of life issues around prostate cancer screening and concluded that long-term effects after diagnosis and treatment diminished the benefits of screening but that more data was needed to make universal screening recommendations. "There are significant harms from screening, and this is the agreement in general around the world," said Schröder. "The most important message is that the harms do not exceed the benefits; they adjust the benefits in a reasonable way."

Based on the controversial and still confused picture of PSA's role, Schröder and Brawley agreed that individual informed decision-making for now remains the best prostate cancer screening strategy. "The present evidence does not justify the introduction of population-based screening programs," said Schröder. "However, well-informed men should not be denied PSA-driven testing." Brawley drove the point home more directly. "Mass screening for prostate cancer is commonly done in the U.S., and I've been on a 22-year mission to drive it out of the picture and drive screening into the doctor-patient relationship."

Molecular Biomarkers to the Rescue?

With PSA's flaws, research to come up with something better has been intense, particularly in the genomic realm. Molecular-based biomarkers hold the promise of answering crucial clinical questions, such as who should be screened, who has prostate cancer, and who needs treatment (See Table, p. 7). PSA currently plays a more or less unsatisfying role in addressing these concerns.

Already, at least 80 single nucleotide polymorphisms (SNP) have been linked to prostate cancer, and multigene panel tests that detect or predict the disease are available commercially. A recent New York Times article indicated that more than one dozen companies have already or intend to introduce such tests, some intended to be used as adjuncts to PSA results, others independently. Several are aimed at teasing out from biopsy samples how aggressive a cancer is, including Genomic Health's Oncotype DX Prostate Cancer Test. Others strive to guide treatment choice or provide prognostic information after surgery, like Myriad Genetics' Prolaris test. Still others propose to forestall unnecessary second biopsies in men whose PSA levels stay elevated after an initial negative biopsy. One such example is the PCA3 test cleared by the FDA in 2012, which detects a long noncoding RNA elevated in >90% of prostate cancer tissue but not normal tissue.

In the midst of all this test development, the Agency for Healthcare Research and Quality reviewed 15 multigene panels and concluded that all have "poor discriminative ability" for predicting risk of prostate cancer. Likewise, leading researchers have been underwhelmed with some proposed tests, and AACR panelist William Nelson, MD, PhD, explained the high bar assay developers face.

"We're very interested in sensitivity and specificity, but when we think about employing these tests for the purposes of medical decision-making we need to think a lot more about predictive value," he said. "As many as 15 to 20 percent of all men if biopsied would be found to have cancer in their prostates. So a test with a negative predictive value of 80 percent is not giving me any more information than I had just by walking around anyway. You have to beat this threshold to have a useful test; otherwise, it's not adding to the clinical picture." Nelson is professor of oncology and director of the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University in Baltimore.

One of the issues in developing clinically useful multigene panels is that most genome-wide association studies have identified relatively more common SNPs. That means that men both with and without prostate cancer carry at least some of the risk markers detected in these panels, making it challenging to set cutoffs for high- and low-risk groups, and affecting the positive predictive power of the test.

Demands for Prostate Cancer Biomarkers

Biomarkers are needed to address prevalent clinical issues in all stages of prostate cancer.
Prostate Cancer State Medical Need Estimated Prevalence
Screening/early detection Risk stratification, detection, and diagnosis. 15–20% of asymptomatic men.
Localized prostate cancer

Treatment stratification, treatment selection, and recurrence monitoring.

Active surveillance monitoring.

Untreated: 33% of men will suffer with metastatic cancer and 21% will die from it.

Treated: 22% of men will suffer with metastatic cancer and 15% will die from it.

20–40% of men on active surveillance will progress to need intervention over 5–10 years.
Recurrent/advanced prostate cancer Risk stratification, treatment monitoring, and progression detection. Rising serum PSA: 30% progress to metastases over 8 years.
Castration-resistant prostate cancer (metastatic and non-metastatic)

Treatment stratification and treatment monitoring.

Treatment selection.

Rising serum PSA: 33% progress to metastases over 2 years.

Significant responses to abiraterone and enzalutamide in 40–60%.
Courtesy of William G. Nelson, MD, PhD

Promising Genetic Signatures

Some of the genetic signatures that have been studied the most and show the most promise in prostate cancer include PTEN, TMPRSS2, and SPOP. As an example, loss of PTEN inhibits androgen receptor signaling and causes resistance to androgen receptor-based therapies. PTEN deletion also has been associated with poor outcomes, suggesting it may be of both prognostic and predictive importance. Similarly, gene fusions between TMPRSS2 and ERG—present in about half of all prostate cancer cases—especially in conjunction with PTEN deletion, may be a biomarker of aggressive disease.

Rubin also discussed an intriguing line of investigation involving chromothripsis, a dramatic rearrangement of DNA in localized chromosomal regions with usually one-to-two copy number states across the rearranged region. Cells not only survive this catastrophic event, but through their new genomic make-up seem to confer selective advantage to the clone, setting the stage for cancer. First reported in 2010 in the genome of chronic lymphocytic leukemia, bone cancer, and small cell lung cancer, it now has been found in the genome of other cancers, including prostate cancer.

"The lack of stability of the genome could potentially be a biomarker," he explained. "From the first study looking at this, investigators found these rearrangements often incorporated known oncogenes. This suggests that this pattern of rearrangement is a type of biomarker where there may be a selection for alterations over-expressing an oncogene or knocking out a tumor-suppressor gene."

Rubin was quick to add that the chromothripsis phenomenon represents not only the excitement and promise but also the hard work ahead in changing the prostate cancer diagnostic landscape. "This is very complex biology that's going to take a while to understand, but my point is it represents biomarkers that a few years ago would never have been thought about."

Molecular Biomarkers in Prostate Cancer

Molecular-based biomarkers hold the promise of answering crucial clinical questions from screening to treatment monitoring in metastatic disease.
Prostate Cancer State Medical Need Clinical Application Biomarkers Existing and in Development
Screening/
early detection

Risk stratification

Detection/ diagnosis

Who should be screened?

Who has prostate cancer?

PSA, prostate cancer genetic susceptibility loci.

DNA, RNA, protein biomarkers.
Localized prostate cancer

Risk stratification

Treatment stratification

Treatment monitoring

Who needs treatment?

What treatment is needed?

Is treatment effective?

DNA, RNA, protein, circulating cells, grade/stage, PSA.

Currently based on risk.

PSA
Recurrent/
advanced prostate cancer

Risk stratification

Treatment monitoring

When should treatment begin?

Is treatment effective?

PSA, previous grade/stage.

 

PSA
Castration-
resistant prostate cancer (metastatic and non-metastatic)

Treatment stratification

Treatment monitoring

Which treatment is most likely to be of benefit?

Is treatment effective?

?

 

PSA
Courtesy of William G. Nelson, MD, PhD

Mutations and Methylation

Both Rubin and Nelson noted the importance of two studies published in 2012 involving whole genome sequencing in sizable prostate cancer cohorts. One involved 112 cancer tumors at diagnosis prior to treatment, and the other, 50 treatment-resistant, ultimately fatal cases. Mutation rates in the two types of disease were fairly similar, but the treatment-resistant tumors had markedly increased copy-number alterations. Overlap in mutations in both types of tumors also suggest that dysregulation of the androgen receptor pathway could be an early event in prostate cancer development.

Nelson's own work and that of other research teams involves exploring DNA methylation alterations. He reported that at least 5,000 of these epigenetic changes have been found in prostate cancer, and that the changes tend to be stable in individuals but vary between individuals. "What's becoming surprising is just how many of these look like they're drivers of the neoplastic phenotype. We're working hard to figure out how that works, which are drivers, which are passengers," he said.

If this work seems very far removed from helping a man and his physician make an informed decision about prostate cancer screening and treatment, both Rubin and Nelson suggested otherwise. "By looking at advanced disease, hopefully it will provide insight in early detection as to which tumors will go on to develop into those aggressive cancers. There are a number of activities going on with dream teams dedicated to this issue," said Rubin.

Nelson reflected on how quickly the field moved in 2004/2005 to begin using the alpha-methylacyl-CoA racemase (AMACR) monoclonal antibody, P504S, in immunohistochemistry analysis for prostate biopsy specimens that are difficult to diagnose by morphology alone. From the time several research teams discovered that AMACR was very commonly over-expressed in prostate cancer but not in normal prostate tissue, he recalled that the field rallied rapidly to make use of this finding. "It moved incredibly quickly, using an enabling technology that had appeared on the scene at that time—tissue microarrays—and a collaborative spirit among prostate pathologists. They got together very quickly, looked at a very large number of prostate samples and were able to introduce the antibody staining for this literally from the first report of it to wide-spread community adoption improving cancer care within two years," he said. "So we are ready as a field for a good biomarker to deliver into the community. I'm very optimistic about what we can accomplish."