Lowdown on PSA Levels - Should Current Threshold Drop?

September 2003
Feature Story

Paul Karr
Reprinted with permission of CAP TODAY

A new study suggests that the traditional PSA threshold for recommending prostate biopsy may need to be reduced, at least in men under 60 years of age.

The study, by Rinaa S. Punglia, MD, MPH, et al, published in the July 24 issue of the New England Journal of Medicine (2003; 349: 335–342), reports that the current threshold is so high it may be missing up to 82 percent of cancerous prostate tumors in men under the age of 60 and 65 percent of cancers in men age 60 or older. The authors suggest that a lower threshold (2.6 ng/mL versus the current 4.1 ng/mL) be considered for triggering biopsy, particularly in younger men who rarely experience benign prostate enlargements that can skew prostate-specific antigen test results.

"It may be appropriate for physicians attending younger men—men who are less likely to have benign enlarged glands—to think about reducing the cutoff point for a biopsy, because that will pick up more cancers," Anthony V. D’Amico, MD, PhD, associate professor of radiation oncology at Brigham and Women's Hospital, told CAPTODAY. Dr. Punglia, who is an instructor in radiation oncology at Harvard Medical School, and Dr. D'Amico coauthored the study with William J. Catalona, MD, Kimberly A. Roehl, MPH, and Harvard School of Public Health statistician Karen M. Kuntz, ScD.

To reach their conclusion, they used statistical methods to re-examine data from 6,691 subjects who were enrolled in a PSA screening study at Washington University School of Medicine, St. Louis, between 1995 and 2001. The study was conducted under the supervision of Dr. Catalona, who is now professor of urology at Northwestern University's Feinberg School of Medicine and director of the clinical prostate cancer programs at the Robert H. Lurie Comprehensive Cancer Center and at Northwestern Memorial Hospital, Chicago. Eleven percent of the study patients (705 total) subsequently received ultrasound-guided biopsies, usually consisting of five to eight cores of prostate tissue. The biopsies were performed because the patients' PSA levels exceeded 2.6 ng/mL or digital rectal examinations were suspicious for developing tumors, or both.

The study authors believed a key problem with the existing PSA test protocol was its verification bias: Only those patients with levels high enough to receive biopsies actually received them, making the PSA test appear to be nearly perfectly sensitive. But many other patients with cancer had blood PSA levels between 2.6 ng/mL and 4.1 ng/mL and thus never received biopsies. The absence of those other cases in the sample data may paint a falsely optimistic picture of the PSA test's sensitivity.

To test that idea, and estimate the number of cancers that may have been missed by PSA testing in the larger sample, Drs. Punglia and D'Amico's team first analyzed all cases in which biopsies had been performed and cancer diagnoses determined. By subjecting the results of those biopsies to multivariate analyses—using a logistic-regression model—the team determined the importance and predictive ability of such factors as race, age, size of prostate, digital rectal examination results, and family history.

They then used those factors to calculate the likelihood that the nonbiopsied patients would have been diagnosed with cancer had a lower threshold of PSA level been used to advise biopsy. In essence, cancer results derived from the biopsies were used to extrapolate cancer probabilities in the larger population. A specialized statistical software model recalculated the area under receiver-operating-characteristic curves—a measure of the overall clinical usefulness of a test—for each diagnostic possibility: false-positive, false-negative, true positive, and true negative.

This exercise simulated subjecting every member of the group of 6,691 original subjects to biopsy, rather than taking biopsies only from patients with elevated PSA blood levels.

The authors wrote: "[T]here may be variables besides age, PSA level, results on digital rectal examination, race, and family history that both predict the chance of undergoing prostate biopsy and are related to underlying disease status. This is a problem with retrospective studies."

Nevertheless, Dr. D'Amico asserts that the analytical methods used in the study were robust.

"The statistics developed for this study are first-rate," he says. "I feel safe saying you will catch double the cancers in younger men by lowering their cutoff point for a biopsy to 2.6 ng/mL."

In an editorial accompanying the study (N Engl J Med. 2003;349: 393– 395), Fritz H. Schröder, MD, PhD, and Ries Kranse, PhD, of the Erasmus Medical Center, Rotterdam, Netherlands—who are now supervising the world’s largest PSA screening study, involving about 200,000 randomized subjects—wrote that the problem of verification bias has been recognized previously but that Dr. Punglia and colleagues address it by using advanced statistical techniques. "They found that the characteristics of the PSA test can be misleading if correction for verification bias is not performed," Drs. Schröder and Kranse said.

Still, they wrote, "These findings are difficult to apply clinically." And they concluded: "Lowering the PSA threshold for performing a biopsy will increase the rate of overdiagnosis and, potentially, overtreatment. This recommendation is not ready for routine clinical practice."

Drs. Schröder and Kranse argue that new screening recommendations should come from randomized studies designed to show whether screening "reduces mortality...without unacceptably reducing the quality of life."

Others, too, urge caution. "There are a lot more reasons not to do it [change the threshold] than to do it," says Gerald Chodak, MD, clinical professor of surgery at the University of Chicago and director of the Midwest Prostate and Urology Health Center at Weiss Memorial Hospital, Chicago.

"Lowering the threshold is just a bad idea," he says. "We don't even know if routine testing saves lives yet. All this would be is an excuse to stick more needles in the prostate. We need more study."

"I do think this study is good," says Daniel Chan, PhD, "for bringing up an awareness that by using the standard cutoff of 4 ng/mL, you miss a lot of cancers. "On the other hand," he adds, "biopsy is somewhat subjective, and there's an impact on health care—eventually we all have to pay for biopsies. So we're trying to avoid unnecessary ones as much as possible." Dr. Chan is professor of pathology, oncology, radiology, and urology at Johns Hopkins University School of Medicine and co-director of the pathology core lab and director of its biomarker discovery center.

A more sensitive test would also trigger radiation treatment, surgery, and other therapies in men who may not have prostate cancer—or don't have aggressive forms of it.

Mitchell Sokoloff, MD, assistant professor of surgery and co-director of urologic oncology at the University of Chicago, calls the paper a "rather elegant academic exercise." He says most of the cancers that a lower PSA threshold and subsequent biopsy would catch would be diagnosed by rectal examination or other means during a time when the chance for disease-free survival is still favorable.

"Using these cutoff values will not appreciably affect the global cure rates for prostate cancer," he adds. "The best policy is for patients and their physicians to educate themselves fully regarding the pros and cons of prostate cancer screening and treatment."

Dr. Catalona counters that many men with PSA test results in the 2.6 to 4.1 ng/mL range will eventually reach the 4.1 ng/mL threshold.

"When we did our cancer screening study in St. Louis, we saw PSA rising in many men, but we could not recommend biopsy until it reached 4 ng/mL because that was the protocol," Dr. Catalona says. “Then it would reach 4 ng/mL, and we would do a biopsy and find cancer and do a radical prostatectomy. Those men would become very upset because the writing was clearly on the wall earlier than that, but we could do nothing with the higher threshold."

Dr. Catalona examined his screening data in 1995 and found that in about half of the men with PSA results between 2.6 and 4.1 ng/mL, the levels rose to above 4.1 ng/mL within four years.

"Critics say we're going to be doing more biopsies with the lower threshold level," Dr. Catalona says. "But most of these men are going to be having biopsies anyway. There's a benefit to patients to do them earlier. So we're not doubling the number of biopsies with this recommendation. We’re simply doing the same biopsies a bit earlier in life."

Dr. D'Amico says the study provides a "tantalizing" answer to the question of PSA cutoff points but that it doesn't answer the question of whether the test is effective for diagnosing prostate cancer.

"This paper and this recommendation will become very important only if the screening studies come back positive and confirm that PSA testing makes a difference in overall mortality," he says. "The screening studies in process should answer that. Until we have that answer, we need to exercise caution."

PSA velocity—the change in year-to-year values—may hold greater promise than total PSA as a marker for prostate cancer, Dr. D'Amico says, since benignly enlarged prostates produce PSA at a fixed low rate while prostate cancer produces sudden jumps in PSA levels.

"I am always very concerned when I see a man of 55 years who has a PSA of, say, 0.5 ng/mL one year and the next year it is 2.6," Dr. D'Amico says. "That is much worse than going from, say, 2.1 one year to 2.6 ng/mL the next."

"Those who bet on such things," he adds, "believe that using a change in PSA over time to recommend biopsy may make a big difference."

Free PSA may also make a difference. Recent work has pointed to three forms of free PSA as potentially more promising indicators of prostate cancer than total PSA. An assay for proPSA or one of the other two known forms of free PSA (benign PSA and intact PSA) may correlate with early detection of prostate tumors more directly than total PSA.

"The PSA test is good, and it is really the best cancer-detection test in all of medicine," Dr. Catalona says. "But it still only largely detects the larger, more aggressive prostate cancers."

"Combining free PSA with a lower cutoff of total PSA of 2.6 ng/mL," Dr. Chan adds, "would be a good improvement in that it would help pick up more cancers while minimizing unnecessary biopsies. Adding a test for free PSA could help us pick up 30 percent more cancers."

Dr. Chan notes that in the long run, new markers are needed. "We need to find ways to do prostate cancer diagnosis not just a little better—but much better."

Paul Karr is a writer in North Bergen, NJ.