Is Randomized Trial Data Enough to Recommend Against Routine PSA Screening? 'Point/Counterpoint' Debate in Medical Care
Newswise — Philadelphia, Pa. (March 25, 2013) - Do the results of recent randomized trials justify the recent U.S. recommendation against yearly measurement of prostate-specific antigen (PSA) as a screening test for prostate cancer? That's the topic of debate in a special "point/counterpoint" section in the April issue of Medical Care. The journal is published by Lippincott Williams & Wilkins, a part of Wolters Kluwer Health.
The recommendation against routine PSA measurement relies too heavily on randomized trial data, according to an article by Ruth Etzioni, PhD, of Fred Hutchinson Cancer Research Center, Seattle, and colleagues. They argue that modeling studies provide a truer picture of the long-term benefits of PSA screening. But Dr Joy Melnikow of University of California, Davis, and colleagues disagree, asserting that randomized trials provide a sufficient level of certainty to recommend against PSA screening.
Point: Short-Term Trials Don't Reflect Long-Term Risk
Last year, the U.S. Preventive Services Task Force recommended against routine PSA measurement to screen for prostate cancer. The recommendation was mainly based on two recent studies—one conducted in Europe and one in the United States—in which men were randomly assigned to annual PSA screening or no screening. Both studies concluded that annual screening did not reduce the risk of death from prostate cancer.
But randomized trials have important limitations as a basis for screening policies, according to Dr Etzioni and colleagues. They note that screening trials generally provide short-term results, in contrast to the long-term results generated by population-wide screening programs. They argue that taking the randomized trial data at face value "misrepresents the likely long-term population impact of PSA screening (relative to no screening) in the United States."
Dr Etzioni and coauthors discuss the results of modeling studies that give a different picture of the benefits of PSA screening. Based on those models, screening may explain 45 percent of recent declines in U.S. deaths from prostate cancer, while changes in treatment account for 33 percent. When the randomized trial data are extrapolated to the U.S. population over the long term, the absolute reduction in deaths attributed to screening appears at least five times greater than in the original trial reports.
Modeling studies also suggest a lower rate of overdiagnosis—screening detection of slow-growing prostate cancers that otherwise would have caused no harm—than reported in the trials. Dr Etzioni and colleagues conclude, "With a disease whose hallmark is a lengthy natural history, the harms of developing cancer screening policies based primarily on limited-duration screening trials may well outweigh the benefits."
Counterpoint: Trials Are Best Evidence on Screening Effects
But in their "Counterpoint" essay, by Dr Melnikow and colleagues notes that the U.S. and European trials provided 11 to 13 years' follow-up in more than 250,000 individuals. They also point out that the U.S. trial was highly representative of the population and showed no reduction in death resulting from annual PSA testing. (Dr Melnikow and colleagues were members of the U.S. Preventive Services Task Force when the recommendation was made.)
They add that, because of "competing causes of death," it becomes even less likely that a large reduction in deaths from prostate cancer will appear over long-term follow-up. The chances of overdiagnosis and potential harms from screening are also likely to increase with continued aging. Dr Melnikow and coauthors conclude, "Projections from models are subject to mistaken assumptions and investigator biases, and should not be accorded the same weight as evidence from randomized controlled trials."
In an editorial response, Dr Etzioni's group points out that modeling plays an essential role in addressing questions about the harms and benefits of screening. "While we acknowledge the centrality of screening trials in the policy process," they write, "we maintain that modeling constitutes a powerful tool for screening trial interpretation and screening policy development."
The debate is "no mere academic exercise," according to an editorial by Ronnie D. Horner, PhD, of University of Cincinnati Medical Center. With the increased emphasis on disease prevention under health care reform, it is essential to offer those services most likely to represent value—including cancer screenings. While there's no easy answer, Dr Horner writes, "I am hopeful that this Point-Counterpoint exchange will initiate a discussion among healthcare scientists that will yield greater guidance for determining whether a health care service is, indeed, value health care."