Screening for Prostate Cancer: An Update of the Evidence for the U.S. Preventive Services Task Force

The American Cancer Society estimates that 189 000 men will receive a diagnosis of prostate cancer in 2002 and that 30 200 men will die of the disease (1). Many more men receive a diagnosis of prostate cancer than die of it (lifetime risk, about 1 in 6 vs. about 1 in 29). Among types of cancer, only lung cancer kills more men each year. The cause of prostate cancer is unknown, and the best-documented risk factors (age, ethnicity, and family history) are not modifiable. The burden of prostate cancer falls disproportionately on men who are older or black. The median age at diagnosis is approximately 71 years, and the median age at death is 78 years (2). More than 75% of all cases of prostate cancer are diagnosed in men older than 65 years of age, and 90% of deaths occur in this age group (2, 3). Incidence is approximately 60% higher and mortality rate is twofold higher in black men than in white men (2). Asian-American men and Hispanic men have lower incidence rates than non-Hispanic white persons (3). Although approaches to primary prevention of prostate cancer are being tested, to date none are known to be effective. The most common strategy for reducing the burden of prostate cancer is screening, but screening remains controversial. Many studies on this topic have been published since 1996, when the U.S. Preventive Services Task Force (USPSTF) last examined prostate screening (4). To assist the USPSTF in updating its recommendation, the Research Triangle InstituteUniversity of North Carolina Evidence-based Practice Center performed a systematic review of the evidence on screening for prostate cancer. Methods Using USPSTF methods (5), we developed an analytic framework and eight key questions to guide our literature search. Because we found no direct evidence connecting screening and reduced mortality, we searched for indirect evidence on the yield of screening, the efficacy and harms of various forms of treatment for early prostate cancer, and the costs and cost-effectiveness of screening. We developed eligibility criteria for selecting relevant evidence to answer the key questions (Table 1). We examined the critical literature from the 1996 USPSTF review and used search terms consistent with the eligibility criteria to search the MEDLINE database and Cochrane Library for English-language reviews and relevant studies published between 1 January 1994 and 15 September 2002. Table 1. Key Questions, Inclusion Criteria, and Articles Meeting Criteria The first author and at least one trained assistant reviewed abstracts and articles to find those that met the eligibility criteria. For these studies, the two reviewers abstracted relevant information using standardized abstraction forms. We graded the quality of all included articles according to USPSTF criteria (5). The authors worked closely with two members of the USPSTF throughout the review and periodically presented reports to the full USPSTF. We distributed a draft of the systematic evidence review to experts in the field and relevant professional organizations and federal agencies for broad-based external peer review and made revisions based on the feedback. We then revised the full systematic evidence review into this manuscript. A more complete account of the methods of this review can be found in the Appendix. The complete systematic evidence review is available on the Web site of the Agency for Healthcare Research and Quality (www.ahrq.gov) (6). This evidence report was funded through a contract to the Research Triangle InstituteUniversity of North Carolina Evidence-based Practice Center from the Agency for Healthcare Research and Quality. Staff of the funding agency and members of the USPSTF contributed to the study design, reviewed draft and final manuscripts, and made editing suggestions. Results Direct Evidence That Screening Reduces Mortality Randomized, Controlled Trials Labrie and colleagues (7) completed the first randomized, controlled trial (RCT) of prostate cancer screening with more than 46 000 men. At the end of 8 years of follow-up, approximately 23% of the invited group and 6.5% of the not-invited group had been screened with prostate-specific antigen (PSA) testing and digital rectal examination (DRE). Prostate cancer death rates did not differ between groups (4.6 vs. 4.8 deaths per 1000 persons, respectively). Two other RCTs of prostate cancer screening, both initiated in 1994, are ongoing: the U.S. National Cancer Institute Prostate, Lung, Colorectal, and Ovary Trial and the European Randomized Study of Screening for Prostate Cancer. Neither study will have data on mortality for several more years. CaseControl Studies Three well-conducted, nested casecontrol studies (two since 1994) examined the relationship between chart review documentation of DRE and advanced prostate cancer or death from prostate cancer. Two studies found no relationship (8, 9). The third study found that men who died of prostate cancer had fewer DREs in the years before diagnosis (odds ratio indicating a protective effect of DRE, 0.51 [95% CI, 0.31 to 0.84]) (10). Why results from these otherwise similar studies differ is not clear. The three studies depended on large databases and on individual medical records. They defined cases slightly differently and used different approaches to differentiate screening DRE from diagnostic DRE. Because such studies are complex in design, we were not able to determine whether one method was more accurate than another (11). All three studies were small, and all were consistent with a reduction in prostate cancer mortality of up to 50% with DRE. We found no casecontrol studies of PSA screening. This can be explained, at least in part, by the fact that insufficient time has elapsed since the introduction of PSA as a screening test in the late 1980s. Such studies are under way (12). Ecologic Studies Around 1987, use of PSA screening began to increase rapidly in the United States. Important trends in prostate cancer incidence and mortality also occurred at that time. Although incidence rates had been slowly increasing for some years before 1987, data from the U.S. Surveillance, Epidemiology, and End Results program showed a dramatic increase in age-adjusted prostate cancer incidence20% per yearfrom 1989 to 1992. The rates then decreased at 10.8% per year (13), stabilizing after 1994 (14). Most of the increase in incidence was seen in localized or regional disease. Incidence of distant-stage disease at diagnosis showed little initial increase and then began to decline; annual decline for white men was 17.9% after 1991 (15). Disease-specific mortality rates paralleled trends in prostate cancer incidence (15, 16). In the late 1980s, the average annual percentage increase rose from 0.7% to 3.1% for white men and from 1.6% to 3.2% for black men. In 1991, prostate cancer mortality rates for white men began to decline (21.6% decrease from 1991 to 1999); in 1993, rates for black men followed suit (16.0% decrease from 1993 to 1999) (14). Mortality rates decreased in all age groups at about the same time. Analyses of trends in prostate cancer incidence and mortality in Olmsted County, Minnesota (17), and in Canada (18, 19) have shown similar results. Ecologic evidence is difficult to interpret. Although screening probably explains trends in incidence of prostate cancer (20), trends in mortality are more difficult to understand. Some aspects of the trends (for example, a decline in distant-stage disease) are consistent with screening, but other aspects (for example, the short time between increased screening and decreased mortality) (21) are not as consistent with our current view of the natural history of prostate cancer. The argument that the decline in mortality can be attributed to PSA screening would be stronger if it could be shown that the decline was largest in areas with more screening. To date, data on this issue are conflicting (19, 22-27). Other possible explanations for decreased mortality include attribution bias and improved treatment. Attribution bias suggests that some deaths are mistakenly attributed to prostate cancer. If the percentage of deaths so attributed is stable, then the prostate cancer mortality rate would be expected to increase and decrease in close approximation with the incidence of prostate cancer in the population (16). Changes in prostate cancer treatment during the late 1980s and early 1990s included higher rates of radical prostatectomy, development of luteinizing hormonereleasing hormone (LHRH) agonists (allowing improved androgen deprivation therapy without castration), and refinements in radiation therapy. Such changes may explain the reduction in prostate cancer mortality. A recent study by Bartsch and coworkers (27), for example, documented a greater reduction in prostate cancer mortality in the Austrian state of Tyrol, which had instituted a free PSA screening program, compared with the rest of Austria. This finding could be a consequence of the screening program, changes in treatment that accompanied the screening program, misattribution of cause of death, or some combination of the three. Accuracy of Screening Three problems complicate any attempt to determine the accuracy of screening tests for prostate cancer. First, research has yet to clarify which tumors screening should target. Second, the reference standard (prostate biopsy) for diagnosing prostate cancer after positive results on a screening test is imperfect. Third, few studies perform biopsy on men with negative results on screening tests. Prostate cancer is a heterogeneous tumor. Different cases of prostate cancer have widely varying growth rates and potential for causing death. Ideally, prostate cancer screening would target only tumors that would cause clinically important disease. Currently available prognostic markers can distinguish a small number of men with excellent prognosis for long-term survival and a small number of men with po