Predicting Prostate Cancer’s Moves

New tests could refine therapy decisions

When a man’s physician diagnoses prostate cancer, difficult decisions about the patient’s treatment course lie ahead. Surgery or radiation therapy could either extend life or needlessly impair the man’s quality of life. Two decades ago, a diagnosis of prostate cancer was tantamount to a death sentence. Physicians detected the cancer mainly through a physical exam of the rectum, which usually identified abnormal growths too late for surgery to stop the cancer from spreading throughout the body.

PRESENT DANGER. Orange staining marks the presence of the protein EZH2 in prostate cancer cells on the left of this composite image. Red squares on the DNA microarray at right identify genes that are active in metastatic prostate cancer. Green squares indicate genes that are repressed. V. Mahavishnu / Univ. of Michigan

That’s no longer the case. Prostate cancer diagnosis changed dramatically with the discovery in 1979 of a protein known as prostate-specific antigen (PSA) and the introduction in the late 1980s of a test that measures its concentration in the blood. Today, urologists consider a high concentration of PSA to reflect abnormal prostate growth, which is often a sign of cancer in that gland. Although the test doesn’t pick up all cases, many men today are diagnosed with early-stage prostate cancer that has yet to cause symptoms or spread beyond the gland.

Early diagnosis, however, is a mixed blessing. Among the approximately 200,000 cases of actual prostate cancer detected each year, as many as 70,000 are slow-growing cancers unlikely to cause serious disease in the man’s lifetime.

“A great deal of therapy might be administered to this group of patients unnecessarily,” says William R. Sellers of the Dana-Farber Cancer Institute in Boston. PSA tests alone don’t reveal which tumors represent mortal danger. Predicting whether a given cancer will recur after prostate surgery largely boils down to guessing whether the tumor has disseminated malignant seeds throughout the body, a process known as metastasis.

Unnecessary treatment for prostate cancer means more than inflated expenses and needless suffering and anxiety for men and their families. Many men who have their prostate glands removed experience sexual dysfunction thereafter (SN: 1/29/00, p. 77: Available to subscribers at Impotence high after prostate removal). Other problems associated with various therapies include urinary incontinence and rectal bleeding. If a tumor appears to be contained within the prostate, localized surgery or merely frequent tests to monitor the tumor may be all that’s needed.

Treating prostate cancer too passively, on the other hand, could prove fatal. If microscopic clumps of cancerous cells spread to lymph nodes and other tissues, cancer can show up outside the prostate if prostate surgery isn’t followed by effective hormone therapy. “The problem we have is making the distinction between good tumors and bad tumors,” says Sellers. Medical researchers are seeking better ways to infer whether newly diagnosed cancers have metastasized.

Clouded crystal ball

Currently, prognostic tools for prostate cancer use several clinical variables. PSA concentration in the blood serves as an indicator of the aggressiveness of a prostate tumor. Other well-established predictive markers include the tumor’s grade and stage. The former is a value called the Gleason score determined by a pathologist, who considers the appearance of biopsied tumor tissue. The latter is rated according to the tumor’s size and how and where it’s identifiable.

Algorithms can combine these three markers, and sometimes other variables, to predict the probability that a given therapy will cure a patient’s cancer.

One popular set of algorithms is the Kattan nomograms, developed in the 1990s by biostatistician Michael W. Kattan of Memorial Sloan-Kettering Cancer Center in New York. Researchers use the algorithms to rank cancer severity in a group of patients. They find that, for example, one Kattan nomogram is correct about 80 percent of the time in predicting which of two patients will fare better after undergoing prostate surgery.

That’s significantly better than a coin toss but it leaves room for improvement.

A technology called gene-expression profiling may provide a step in that direction. The technique, using DNA microarrays or so-called gene chips, simultaneously determines which of thousands of genes are active in a tissue sample. Each spot where a gene has been embedded on the chip lights up if that gene is being expressed.

Gene expression can cause a cell to assemble proteins that affect the cell’s behavior, including turning it cancerous. Profiling that expression, therefore, can divulge clues to a cell’s cancerous potential.

Scientists have accomplished this for a few cancers, such as lymphomas (SN: 1/12/02, p. 21: Available to subscribers at Genes make potential target in lymph cancer) and leukemias (SN: 9/14/02, p. 171: Targeted Therapies), but researchers haven’t yet settled on gene-activity signals that can predict prostate cancer.

EZH2 does it

Several research groups have recently identified markers in gene-expression profiles of prostate cancers. One set of studies has centered on the gene for the protein called enhancer of zeste homolog 2 (EZH2). With his colleagues at the University of Michigan in Ann Arbor, Mark A. Rubin discovered that excess expression of the gene hints at the presence of metastatic prostate cancer. Expression of the gene better predicted whether the cancer would recur after surgery than did clinical measurements such as Gleason score and tumor stage, the researchers reported in the Oct. 10, 2002 Nature.

In subsequent analyses, Rubin, who is now at Harvard University, and his coinvestigators reported in the May Journal of the National Cancer Institute that relatively high expression of the gene for EZH2, coupled with relatively low production of a molecule called E-cadherin, was more predictive of a recurrence of cancer after prostate surgery than was any other pair of the 14 markers of gene expression that the researchers were considering.

EZH2 is part of the molecular machinery that enables a cell to read its own DNA. Abnormal EZH2 expression may lead to a cell “forgetting its initial identity” and turning malignant, says Arul Chinnaiyan of the University of Michigan in Ann Arbor, who has collaborated with Rubin. E-cadherin makes cells adhere to one another. Without proper adhesion, cancer cells may become detached from a tumor and move through the body, Chinnaiyan explains.

EZH2 and E-cadherin are on a growing list of gene products that appear in preliminary studies to predict how aggressively a tumor will act.

At a meeting of the American Association for Cancer Research in Washington, D.C., in July, pathologist William L. Gerald of Memorial Sloan-Kettering Cancer Center, Kattan, and his colleagues described a combination of up to nine gene-expression markers, including EZH2, that, along with PSA score and tumor stage and grade, better indicate tumor aggressiveness than the Kattan nomogram does on its own.

Using data from other volunteers, Sellers and his colleagues had described in the March 2002 Cancer Cell a separate set of genes that strongly correlates with cancer cells’ appearance as measured by Gleason score. Gene-expression data on their own predicted with 90 percent accuracy which cancers would recur within 4 years after removal of the prostate.

So far, Gerald says, “there is a long list of potential prognostic markers in prostate cancer but none that have been validated” by tests on different groups of patients. If any of these gene-expression profiles truly predict cancer behavior, they should pop up in study after study, he notes.

Unfortunately, doing such validation is difficult.

Getting tissue samples is time-consuming, and researchers use different mathematical approaches.

Multiple labs are now collaborating to bring their diverse protocols to bear on a single set of samples and a short list of candidate biomarkers.

Scanning for clues

In the hunt for better predictors of prostate cancer severity, researchers have many leads, not all of which relate to how genes are expressed in the prostate. Other new prognostic markers may come from examining genetic differences among individuals, assessing genetic damage sustained by prostate tissues over time, and measuring blood concentrations of a molecule that is a biological precursor to PSA.

One of the most promising, recently proposed predictive markers comes from a new twist on an existing prognostic tool, Rubin suggests. Surgeons who remove cancerous prostates often take tissue from nearby lymph nodes at the same time, since these nodes are likely to be the first place to which a cancer metastasizes. A cancer-positive lymph node indicates that follow-up therapy is needed to keep cancer from recurring.

Magnetic resonance imaging, or MRI, is a noninvasive but imprecise method that’s used to detect lymph node metastases. To see whether they could improve the accuracy of MRI, Ralph Weissleder of Massachusetts General Hospital in Boston and his colleagues injected tiny magnetic particles into the veins of prostate cancer patients before they had surgery. The particles were designed to migrate to lymph nodes and show up on MRI scans.

The researchers scanned each volunteer twice, once before and once after he received the magnetic particles. When they surgically removed lymph node tissue, they found evidence of metastatic disease in 33 of the 80 volunteers. The MRI scans before the magnetic particles were injected accurately revealed whether metastasis had occurred in 65 percent of the volunteers. By contrast, MRI after particle injection had 97.5 percent accuracy, the researchers report in the June 19 New England Journal of Medicine.

The new technique shows an “impressively high” accuracy in identifying nonmetastatic prostate cancer, radiologist Janet Husband and her colleagues at Royal Marsden Hospital in Sutton, England, comment in the same journal. Such information could guide decisions about whether to go ahead with surgery or instead use hormone therapy or irradiate the entire pelvis.

Rubin also lauds the finding and says that he and Weissleder have plans to collaborate on future studies that incorporate their teams’ respective approaches into a single model for predicting how each prostate cancer will behave. In the end, the best method for anticipating and blocking a tumor’s moves may draw from every corner of cancer research.


If you have a comment on this article that you would like considered for publication in Science News, send it to Please include your name and location.

To subscribe to Science News (print), go to

To sign up for the free weekly e-LETTER from Science News, go to

More Stories from Science News on Health & Medicine