Study Concludes That Moderate PSA Levels Are Unrelated to Prostate Cancer Outcomes
Brian Vastag

Washington - Clinicians need to brace themselves for another round in the protracted debate over the value of routine prostate-specific antigen (PSA) screening. A new study from Stanford University School of Medicine contends that PSA scores between 2 ng/mL and 9 ng/mL are "clinically useless" in predicting the size and aggressiveness of prostate tumors. The group, led by Thomas Stamey, MD, also concludes that PSA values in that middle range "have a limited relationship with [surgical] cure rates" (J Urology. 2002;167:103-111).

The report has triggered intense responses from both sides of a deep clinical rift.

"I don't want to use the word dangerous, but my concern about this article is the bottom line message of not worrying about PSA until it is in the 7 to 9 range," said William Catalona, MD, a urologic surgeon at Washington University School of Medicine, St Louis, who performs biopsies on patients with PSA levels as low as 2.5 ng/mL.

On the other hand, "The article suggests that we end up diagnosing a lot of cancers because PSA is secreted not by the cancer but by benign tissue," said Otis Brawley, MD, assistant director for cancer control at Emory University School of Medicine, Atlanta, and a former researcher at the National Cancer Institute.

These two viewpoints define the PSA fault line. Surgeons, who deal with prostate cancer on a case-by-case basis, tend to prefer routine testing. Epidemiologists and others who view the oncological landscape from a distance, in contrast, remain unconvinced about the benefits of such screening.


SEEKING PREDICTIVE VALUE

To reach their conclusions, the Stanford team examined the prostates of 875 men who had them removed between 1984 and 1997. The sample included men with cancer confined to the prostate that was discovered by digital rectal examination or PSA testing followed by biopsy. The team then correlated preoperative PSA values with tumor volume and other variables that can predict potential for a surgical cure, including Gleason score and width of surgical margins.

A trend immediately caught Stamey's attention. As PSA values increased from 2 to 9 ng/mL, the other measures remained relatively flat, showing little correlation. For the entire sample, cancer volume explained just 27% of the variation in PSA scores. However, much of that already low correlation was driven by PSA levels above 22 ng/mL. After excluding those cases, which Stamey said are usually incurable, cancer volume accounted for just 14% of the PSA variation. "It is random," said Stamey. "I could get 15% by flipping a coin."

In contrast, earlier studies have found that total prostate volume predicts PSA value nicely, shoring up benign prostatic hyperplasia as the driving force behind most moderate-range PSA values. The largest such study, directed by Claus Roehrborn, MD, of the University of Texas Southwestern Medical Center, Dallas, pooled data from 4600 men without cancer with PSA levels below 10 ng/mL. Total prostate volume accounted for more than 75% of the variation in PSA values (Urology. 1999;53:581-589).

After reviewing this data, Stamey wanted to know whether PSA values predicted cure potential. His team reviewed a subset of 406 prostatectomy patients with preoperative PSA values below 22 and at least 3 years of follow-up. The cure rate did not decline significantly until the PSA level climbed into the 7 to 9 range. "There's no way I can tell [a patient] that surgery is more likely to cure him at a [PSA level of] 2 to 3 than at 8 to 9," said Stamey.

Proponents of widespread screening questioned the length of follow-up in the study. Catalona said that a longer study would have revealed a different trend. If the data were "really mature," he said, with a minimum of 7 to 9 years of follow-up, "fewer in the 2 to 4 range would have had recurrence." Stamey agreed that longer studies would have been more desirable but declined to speculate on what such data would show.


SHIFTING THE BALANCE?

While the Stanford study is far from conclusive, when added to other evidence it appears to shift the balance of the debate toward a conservative approach.

For Stamey, the research culminates a radical shift from his being an early advocate of PSA screening to an outspoken denouncer. In the late 1980s, he published a number of influential articles, including the first large study linking PSA level to prostate cancer (N Engl J Med. 1987;317:909-916). In another article, Stamey concluded that "prostate-specific antigen (PSA) has been shown to be proportional to the volume of prostate cancer"the very finding he now refutes (Cancer. 1993;71:2031-2040).

But as he accumulated extensive data he had pathologist John McNeal, MD, analyze every prostate removed at Stanfordhis viewpoint changed. "I removed a few hundred prostates that I wish I hadn't," he says now.

Much population-based research appears to support Stamey's change of conviction. Epidemiologists assert that widespread PSA testing leads to many false-positives and unnecessary treatment while doing little to lower the death rate. For support they point to long-term trends. In the United States, for instance, the incidence rate of prostate cancer increased sharply in the early 1990s after widespread adoption of PSA testing. In Great Britain, where PSA testing is relatively uncommon, the incidence rate increased only slightly; it is now less than half the US rate. Mortality rates in both countries have closely tracked each other for 30 years (JNCI. 2001;93:1109-1110).

Autopsy data also buttress a conservative approach. Studies have consistently found that up to one third of men who died from other causes have largevisible with the naked eyeprostate cancers (Clin Chim Acta. 2002;315:71-97). "Men tend to die with prostate cancer, not from it," goes a saying spawned by this research.

Still, the American Cancer Society estimates that 30 200 deaths from prostate cancer will occur in the United States in 2002. Some men with the disease will undoubtedly die despite surgery or other treatments that might have helped if used earlier. "Those [against screening] say, 'We aren't 100% sure that surgery saves lives, but we're 100% sure that it causes incontinence and impotence,'" said Catalona, who has had vigorous discussions on the topic with Stamey, both in the literature and at conferences. "Stamey talks about the advantages of waiting [for treatment] until PSA is high, that it may allow a man to enjoy his prostate for a few more years," said Catalona. "But what does that matter if he is dying from an aggressive tumor?"

With that rhetorical question hanging, Catalona said that he recommends biopsies for men with PSA levels above 2.5, marking him as perhaps the most aggressive advocate of early treatment. "If you wait until PSA is in the 7 to 9 range, about 30% of men will have cancer," he said, citing a study he published in JAMA (1997;277:1452-1455). "I operate on these men every day and wish I had gotten to them earlier." Of every 10 prostatectomy patients he sees, Catalona says he finds two or three with metastatic cancers. "A lot of these men had a history of PSAs between 2 and 8 and it [cancer] wasn't picked up until PSA was quite high."


PERSONALIZED PROSTATE LEVELS

But that's the rubof all malignancies, the variations found in prostate cancers are perhaps the widest. Whereas a certain PSA score signals danger for one man, it may not for another. Without a method to distinguish the two, clinicians are left to sort out a confounding array of recommendations (see sidebar). McGill University Faculty of Medicine epidemiologist Olli Miettinen, PhD, neatly summarized the quandary. "The risk for cancer is individual," he said, "but the degree to which screening affects mortality is independent." In other words, positions along the PSA screening spectrum depend as much on whether one is an epidemiologist or a surgeon as they do on the state of the slowly maturing science. Until screening tests proven to reduce death rates appear, that situation is unlikely to change.


Variation in Recommendations

(Return to text.)