Prostate Cancer: To Screen or Not to Screen?

In a lively session featured at the 32nd Annual Meeting of the American Society of Clinical Oncology (ASCO), Jerome P. Richie, md, Brigham and Women's Hospital, Boston, and Steven H. Woolf, md, mph, Medical College of Virginia, Fairfax, debated the merits of screening for prostate cancer. Last month, Dr. Ritchie made the case for routine screening. In this issue, Dr. Woolf, a family physician, argues that currently available data do not support the value of routine screening.

Primary-care physicians see a different patient population than do oncologists or urologists and are concerned with preventing a broader range of health problems, Dr. Woolf said. That is one reason why primary-care practitioners have a different "take" on the value of routine screening for prostate cancer.

Official groups in North America and around the world also have contrasting viewpoints on this issue, he noted. The American Cancer Society, American Urological Association, and American College of Radiology all agree on the need to do regular PSA screening beginning at age 50. However, the American College of Physicians, American Academy of Family Physicians, US Preventive Services Task Force, latest Patient Outcomes Research Team (PORT) of the Agency for Health Care Policy and Research, and several groups in Canada, Europe, and Australia all do not recommend routine screening.[1]

Why is there so much uncertainty and debate about this issue? Before any screening test, particularly a cancer screening test, is generally accepted, some basic criteria must be met: Is the condition serious? Is screening accurate in detecting the disease? Does early detection improve outcome? Is screening or treatment harmful? And does screening do more good than harm? With regard to prostate cancer screening, the answers to some of these questions are still open, Dr. Woolf asserted.

Is the Condition Serious?

As for whether prostate cancer is a serious condition, evidence from the United States and other countries clearly shows that this cancer causes suffering and/or death in a large number of men. However, these men represent only a subset of the total population with prostate cancer.

"We know from autopsy studies that there is a significant portion of the population who have latent prostate cancer that progresses slowly and produces few, if any, clinical symptoms," Dr. Woolf said. Of these patients, only a subset progress to more aggressive disease, with its potential for significant morbidity and mortality.

The actual prevalence of latent prostate cancer is debatable. Autopsy studies suggest that 30% of men over age 50 may have latent cancer,[2] although the generalizability of these data is questionable. Clearly, however, a "significant" number of prostate cancers are indolent in nature.

Is Screening Accurate?

With regard to the accuracy of PSA screening, Dr. Woolf noted that the sensitivity of the test is relatively good. In studies of men with prostate cancer, 80% have positive PSA results.[3] The positive predictive value of PSA screening in asymptomatic men is roughly one-third, which means that one out of three men with a positive PSA actually have prostate cancer.

"The flip side of that is that two-thirds of men with a positive PSA will not have prostate cancer," Dr. Woolf said. This becomes important when one considers the potential morbidity associated with further testing and treatment of this population.

Undoubtedly, the increased effort at screening will lead to the detection of many latent prostate cancers. Whether these PSA-detected cancers are clinically significant has been the subject of much debate, however. Recent evidence indicates that the types of cancers detected by PSA screening may be more aggressive than those detected on autopsy studies.[4]

Such features as advanced tumor grade and pathologic extension beyond the prostate capsule are more common in cancers that ultimately produce clinically significant disease. However, Dr. Woolf made the distinction between the actual development of clinically significant disease and the detection of histopathologic characteristics that increase patients' risk of developing clinically significant disease. Cancers detected by PSA are more likely to have these features, but whether these patients will actually develop symptoms is unknown.

Does Early Detection Improve Outcome?

If one accepts, for the purpose of discussion, that the PSA test is accurate in detecting early prostate cancer, the critical issue, as shown in Figure 1, is whether treating early cancer improves outcome, not just whether screening tests can detect early-stage disease. With respect to breast, colon, or cervical cancer, a large body of evidence from randomized controlled trials and other types of controlled outcome studies indicates that early detection and treatment improves outcomes.

What about the evidence that early detection and treatment of prostate cancer improves outcomes? Currently, there is no direct evidence, Dr. Woolf said. He noted that the ongoing randomized trials focusing on this issue will not be completed for many years, and therefore, clinicians must rely on the indirect evidence available right now. This indirect evidence suggests: (1) a higher survival rate for patients with early-stage disease; (2) the types of tumors detected by PSA are more likely to be early-stage disease; and (3) PSA-detected tumors seem to have features associated with increased risk for progression.

Lead-Time and Length Bias

Although these available studies suggest that screening may be beneficial, Dr. Woolf argued that they suffer from numerous methodologic problems. Two problems that are of particular importance are lead-time bias and length bias.

As shown in Figure 2, for a patient who develops symptoms at a particular point in life and who will die at a subsequent point, the goal of screening is to detect the disease before symptoms develop, initiate early treatment, and thereby prolong the patient's life.

"In the phenomenon of lead-time bias, we can end up artificially lengthening the survival of the patient without actually changing the outcome," Dr. Woolf said. For example, as shown in Figure 3, a hypothetical patient who develops symptoms at age 55 and dies at age 60 without cancer screening has a survival of 5 years. If that patient's cancer is detected at age 50 and early treatment is initiated but he still dies at age 60, survival will appear to have doubled from 5 to 10 years.

This phenomenon is not necessarily occurring with prostate cancer, but the extent to which lead-time bias distorts current data is unclear. Thus, said Dr. Woolf, only controlled studies that collect meaningful outcomes data, as opposed to stage shifts and 5-year survival rates, can resolve this uncertainty.

Some screening trials have found that the percentage of patients with metastatic prostate cancer has decreased over time since the advent of routine screening. This has been mistakenly interpreted as evidence that early detection and treatment are "eliminating" metastatic prostate cancer. Rather, Dr. Woolf argued, such a finding may only indicate that there is less evidence of metastases at diagnosis but does not address patients' ultimate outcome. "Are we simply moving the time of diagnosis backward...as opposed to actually changing the course of the disease?"

Another phenomenon that occurs commonly with cancer screening is length bias (Figure 4). If one screens for cancer at any point in time, one will preferentially detect the slowly growing tumors and catch only a minority of the aggressive tumors, because they exist in the screened population for less time.

"The result of that is that your survival rate is artificially inflated, so [that] you end up with higher survival rates, and a sense that the types of cancers you're detecting are having a much more favorable outcome than they actually would have in real life."