Screening for Prostate Cancer With PSA Testing: Current Status and Future Directions

The ultimate utility of the serum prostate specific antigen (PSA) assay as a screening test for reducing prostate cancer mortality has been an area of intense controversy since its introduction. PSA testing was not initially envisioned as a screening tool, but as a way to evaluate treatment responses in men with prostate cancer. Far in advance of evidence from randomized trials, the rapid and widespread uptake of PSA screening into US practice was initially driven by the intuitively logical assumption that the earlier one detects a malignancy, the more likely treatment is to be curative while minimizing associated harms. However, a growing body of observational evidence began to point to a substantial burden of associated overdiagnosis and overtreatment triggered by PSA testing. The interim results of several randomized clinical trials specifically designed to evaluate the impact of PSA testing on prostate cancer mortality have recently become available, but their incongruent results seem to have added fuel to the debate. This article presents a review of the literature on screening for prostate cancer with PSA testing; we include a detailed discussion of potential explanations for the contradictory results of the two largest randomized trials as well as reflections on the future of prostate cancer screening.

Introduction

Prostate cancer is the most common non-skin cancer diagnosed in the United States (an estimated 217,730 cases in 2010); it is also the second leading cause of cancer death in men, accounting for approximately 32,000 deaths in 2010.[1] Screening has been advocated as a means of improving these statistics, since early-stage prostate cancer is more easily amenable to curative interventions. Early detection methods have included the digital rectal examination, serum prostate-specific antigen (PSA) concentrations, and, less commonly, transrectal ultrasound (TRUS). The digital rectal exam had for many years been the modality traditionally employed for prostate cancer screening, although rigorous evaluation of the test was largely lacking. Since the introduction of PSA screening assays into clinical practice in the late 1980s, the digital rectal exam is no longer commonly used alone. The PSA assay has been demonstrated to have a superior sensitivity, specificity, and positive predictive value compared to digital rectal exam.[2] However, serum PSA testing has important limitations, and it remains unclear whether any benefits of screening outweigh its harms.[3]

First, increased PSA levels are not specific to prostate cancer but can also be caused by prostatic infection and inflammation, as well as by benign prostatic hyperplasia (BPH). More importantly, there is no clear cutpoint value at which a man can be assured that he does not harbor prostate cancer: a continuum of risk exists at all PSA values. The Prostate Cancer Prevention Trial (which primarily examined whether finasteride was effective for the prevention of prostate cancer) was unique in that a subset of participants in the control group with negative PSA test results-all less than 4.0 ng/mL-and normal digital rectal exams (2,950 out of 9,459 men) underwent prostatic biopsy at the end of the trial. Study investigators demonstrated wide variance in the sensitivity and specificity depending on the value utilized: PSA cutoff points of 1.1, 2.1, 3.1, and 4.1 ng/mL yielded sensitivities and specificities of 83%/39%, 53%/73%, 32%/87%, and 21%/94%, respectively.[4] There was no PSA value at which the prevalence of prostate cancer dropped to zero, even at the lowest ranges; it was 7% with a PSA value ≤ 0.5 ng/mL, 10% with values between 0.6 and 1.0 ng /mL, and 17% with values between 1.1 and 2.0 ng/mL.[5]

PSA testing was not initially envisioned as a screening strategy; rather, its first use was for the evaluation of treatment responses in men with prostate cancer. It is now one of the most common cancer screening activities in the United States. According to 2008 Behavioral Risk Factor Surveillance System (BRFSS) data, the estimated prevalence of men in the United States aged 40 years or older who had a PSA test during the preceding 2 years ranged from 38.9% to 70.1%, depending on locale (median: 56.2%).[6] Despite its widespread use, the ultimate utility of the PSA assay as a screening test has been questioned since its first introduction into practice. A large body of observational evidence has been amassed in regard to the question; interim results of several randomized, clinical trials specifically designed to evaluate the impact of PSA testing on prostate cancer mortality have also recently become available. This article will present a detailed review of this body of evidence, with reflections on the resulting future of prostate cancer screening.

Observational Evidence of the Efficacy of PSA Screening for Prostate Cancer

Ecologic and case-control studies have addressed the association between screening and prostate cancer mortality, although they are weak study designs for the determination of screening efficacy. Farkas et al examined the US Surveillance, Epidemiology, and End Results (SEER) database for the years 1973 to 1994 and noted a marked increase in the overall proportion of clinically localized, moderate- to well-differentiated tumors diagnosed in the population after the widespread introduction of PSA testing; this suggested that the test was effective at detecting earlier-stage disease.[7] Several other ecologic studies demonstrated greater decreases in prostate cancer mortality in areas where screening was commonly performed, compared with those regions where it was not. For example, Collin et al compared age-specific and age-adjusted prostate cancer mortality rates in the United States (where PSA screening had become widespread) and in the United Kingdom (where routine PSA screening remained less common) between 1975 and 2004. After 1994, they found a four-fold difference in the rate of decline of prostate cancer mortality between the United States (4.17% annual decrease [95% confidence interval {CI}, 3.99-4.34]) and the UK (1.14% annual decrease [95% CI, 0.84-1.44]); rates were most divergent for men aged 75 years and older. The stage shift toward predominately localized disease in the United States, as noted by Farkas, was also observed by Collin et al; this phenomenon was not paralleled in the United Kingdom.[8] A retrospective analysis of age-adjusted community prostate cancer–specific mortality rates in Olmsted County, Minnesota before and after the introduction of PSA screening showed a decline from a baseline rate of 25.8 per 100,000 men between 1980 and 1984 to 19.4 per 100,000 men between 1993 and 1997.[9] Finally, an ecologic study compared prostate cancer mortality rates in the Austrian state of Tyrol, where routine screening was introduced in 1993, to rates in the rest of the country, where screening was not freely available; the study found that while mortality rates decreased throughout Austria, the rate of decline was statistically significantly greater for Tyrol between 1993 and 1999 than for other regions.[10] In these studies, however, concomitant changes in treatment strategies may be important confounding factors that make interpretation of changes in mortality rates difficult.

In addition, not all observational evidence has supported the efficacy of PSA screening in reducing prostate cancer mortality rates. A case-control study of New Jersey men between 1989 and 2000, which compared prostate cancer mortality rates in those who ever received screening with rates in those who did not, found no difference between the two groups.[11] A comparison of men in the Seattle–Puget Sound area between 1987 and 2001-where screening uptake was rapid with the introduction of PSA testing-and Connecticut-where uptake was slower-revealed, despite higher resulting rates of radical prostatectomy and radiotherapy in Seattle, that the adjusted rate ratio of prostate cancer mortality between the two regions was 1.02 (95% CI, 0.96-1.09).[12] Perron et al examined 15 birth cohorts of men aged 50 years and older in Quebec, Canada to determine the relationship of changes in prostate cancer incidence between 1989 and 1993 to subsequent prostate cancer mortality between 1995 and 1999 (after the introduction of PSA screening). The study found that although most birth cohorts did show both an increase in prostate cancer incidence and a decrease in prostate cancer deaths, there was no statistically significant correlation between the two variables (Pearson’s r = 0.33, P = 0.89). The authors repeated the analysis, dividing the men into 15 regional populations, and again found no association between the size of the increase in prostate cancer incidence due to screening and the size of reductions in mortality (Pearson’s r = 0.13, P = 0.68).[13] Some of the “positive” observational studies have also demonstrated important limitations to PSA testing: for example, in the Collin study, although the comparative annual rate of decline in prostate cancer mortality in the US population (screened) vs the UK population (unscreened) appears large, the absolute difference in death rates for all ages combined between the two countries is about 5 men per 100,000 person-years or less, depending on year.[8]

Perhaps most critically, a growing body of observational evidence points to the existence of overdiagnosis and overtreatment of prostate cancer triggered by PSA testing. Cancer is a heterogeneous disease, resulting from a lengthy process of cellular and molecular events that do not always follow a predictable course. PSA testing, by detecting asymptomatic disease, identifies lesions whose biological behavior may not be the same as that of clinically detected tumors. Overdiagnosis occurs when a screening test detects a tumor so indolent it would not impact the life span of the person harboring it, either because it would progress so slowly that the person would die of another cause first, or because it is essentially benign despite its histological appearance. Autopsy studies of men who have died of unrelated causes have shown that there is a large reservoir of histologically detectable but clinically silent prostate cancer in the population.[14-16] For example, in an autopsy study of men aged 40 to 59 years, occult prostate cancer was identified in approximately one-third of subjects.[17] A recent review-based on a series of previously published autopsy studies-estimated the reservoir of potentially screen-detectable prostate cancer in men over the age of 60 years to be in the range of 30% to 70%.[18]

FIGURE

Ecologic Evidence of Overdiagnosis in Prostate Cancer

Internal server error