Center News

'Overdiagnosis' quantified

July 3, 2002
Needle and arm.

A Public Health Sciences study to be released today quantifies for the first time a concern that the medical community has had for years - that a blood test called PSA screening, in use for the past decade, diagnoses more men with prostate cancer than may ever experience its symptoms.

Photo by Clay Eals

Roughly a third of older men diagnosed with prostate cancer with a routine blood test are "overdiagnosed" with the disease, according to a Fred Hutchinson study to be released today.

The study, conducted by Dr. Ruth Etzioni of the Public Health Sciences Division, quantifies for the first time a concern that the medical community has had for years—that a blood test called PSA screening, in use for the past decade, diagnoses more men with prostate cancer than may ever experience its symptoms.

Such "overdiagnosis" can prompt unnecessary surgeries or other treatments with side effects ranging from urinary to sexual dysfunction.

The Etzioni study shows that among U.S. men between ages 60 to 84 with prostate cancers detected by the PSA screening, about 29 percent of whites and 44 percent of blacks would not have been otherwise diagnosed by clinical exam or developed symptoms during their lifetimes.

Simulated results

The results were derived using a simulation that incorporated data from Medicare records of PSA use and cancer diagnosis as well as prostate-cancer incidence data published by a national cancer registry. The predicted rates in the study were consistent with the observed prostate cancer incidence for this age group from 1988 to 1998.

Etzioni's study appears in today's edition of the Journal of the National Cancer Institute.

"These findings are an important step toward calculation of the cost-effectiveness of PSA screening," she said.

Etzioni cautioned, however, that establishment of guidelines for testing awaits results of randomized clinical trials to evaluate PSA screening's effectiveness at reducing mortality.

"What we can say is that our results do raise caution for older men who undergo testing, since considerable morbidity (unhealthiness) can be associated with treatment for prostate cancer," she said.

The PSA test, originally developed to monitor disease progression, has been widely used for early cancer detection since its introduction in the mid-1980s, but there has been no rigorous assessment of its effectiveness for such screening. The procedure measures the level of a protein—prostate-specific antigen—produced by the prostate gland which, when elevated, can signal the presence of cancer.

As use of the test became prevalent, prostate-cancer incidence increased sharply and had more than doubled by 1992 before going back down. This apparent surge in incidence prompted concern over prostate cancer overdiagnosis, which might lead to unnecessary treatment for indolent tumors.

Previous research indicates that in the absence of PSA testing about 9 percent of all American men would be diagnosed with prostate cancer. An earlier study led by Etzioni, said 36 percent of all men—regardless of whether they had been tested or whether they died from prostate cancer—would be found to have a prostate tumor upon autopsy.

Huge potential for overdiagnosis

"That means only one in four men would ever know they had prostate cancer," she said. "So the potential for overdiagnosis is huge."

However, when Etzioni's earlier autopsy data are considered along with her current study, the overdiagnosis rates due to PSA testing correspond to, at most, about 15 percent for whites and 37 percent for blacks.

"As anticipated, overdiagnosis is a problem in the sense that over the last decade, many men have had prostate cancer detected by PSA," Etzioni said. "At the same time, the picture is not as bleak as people might have suspected because while there is a huge prevalence of prostate cancer in the population, we're not picking up as many cases by PSA test as we might have feared."

Moreover, the researchers state that the majority of PSA-detected cancers diagnosed between 1988 and 1998 eventually would have presented clinical signs and that only a minority of cases found only at autopsy would have been detected by PSA screening.

Racial difference in results

Etzioni speculated that the difference in overdiagnosis rates between blacks and whites in her study might reflect that there is a larger pool of black men who would be found to have prostate tumors upon autopsy than white men. Historically, blacks have higher prostate-cancer diagnosis rates than whites, and their cancers typically have a worse prognosis.

To date, there have been no large, randomized trials to assess the effectiveness of the PSA test for cancer screening, which prompted researchers to use a statistical-modeling approach to try to predict the potential magnitude of overdiagnosis.

The study was funded by the National Cancer Institute's Cancer Intervention and Surveillance Modeling Network (CISNET), a cooperative group of grantees using statistical modeling and simulation to understand the impact of intervention strategies on U.S. cancer incidence and mortality trends.

In addition to genetic and environmental components, a number of factors influence cancer trends in a population, including screening practices, new treatments and prevention through lifestyle changes and chemoprevention.

Dr. Eric Feuer, chief statistician of the branch that funds these efforts, and a co-author of the study, was at Met East last week for a meeting of the prostate-cancer wing of CISNET.

He noted that the complex relationship between the introduction of a screening test in the population and cancer-incidence trends, along with the lack of randomized trials regarding PSA testing, make statistical modeling an effective method to address the question of overdiagnosis.

"The state-of-the-art is different for different cancers," he said. "With breast cancer, for example, there have been a number of randomized trials to evaluate chemo and hormonal therapy for the treatment of disease, as well as mammography and preventive steps that use hormonal therapy.

"But with prostate cancer, the population is ahead of the trial with respect to PSA screening, so we turn to statistical models that incorporate a U.S. population perspective."

Data-based model

Etzioni said that the power of her modeling approach is that it is based on as much observable data as possible. However, limitations loom in the available data regarding use of PSA testing.

"It's unfortunate that during the years when PSA was being adopted in the population, its use was not being systematically tracked," she said. "Through medical claims databases, however, we have been able to at least partially track its use."

Two large research trials, in which men are randomized to undergo or not undergo PSA testing, are under way in the United States and in Europe. The studies are designed to assess whether PSA screening reduces death from prostate cancer.

Other collaborators on the Etzioni study included Drs. Dante di Tommaso in PHS; David Penson, Veterans Affairs Medical Center in Seattle; Julie Legler, NCI; Rob Boer, RAND Corporation; and Peter Gann, Northwestern University.

 

Fred Hutchinson Cancer Research Center is a world leader in research to prevent, detect and treat cancer and other life-threatening diseases.