Prostate Cancer Screening: What We Know and What We Need To Know

History and Development of the Prostate-Specific Antigen Assay

The discovery and development of PSA have been described in detail [4, 5]; PSA was initially found in seminal fluid in 1971 by Hara and coworkers [6], who called it -seminoprotein. This protein was characterized in 1978 by Sensabaugh [7] and was given the abbreviated name p30. Because p30 was semen specific, Graves and colleagues [8] found it to be an excellent marker for semen identification and useful in identifying rape victims. In 1979, Wang and coworkers [9] in Chu's laboratory isolated an antigen from prostatic tissue that was immunologically identical and biochemically similar to that isolated from seminal fluid.

Prostate-specific antigen is a serine protease of unknown physiologic function found in epithelial tissue in the healthy prostate, in benign prostatic hyperplastic tissue, and in primary and metastatic prostate cancer tissue. It is secreted into the lumina of the prostatic ducts and is present in high concentrations in seminal plasma. Androgens regulate expression of the PSA gene [10, 11]. In early studies, PSA was not found in any nonprostatic tissues. However, recent preliminary information suggests that PSA also appears in high concentrations in the saliva of both men and women and that PSA-positive cells can be found in the parotid glands [12].

Papsidero and coworkers [13] identified PSA in human serum and verified its molecular identity with the molecule purified from prostatic tissue. Kuriyama and coworkers [14] measured the mean concentrations of PSA in healthy men, and Wang and coworkers [15] found increased PSA serum levels in men with benign prostatic hyperplasia as well as with all stages of prostate cancer. Nadji and colleagues [16] showed that PSA is a useful immunohistologic marker for prostate cancer. Thus, increased serum PSA levels were shown to be a potential indicator of prostatic disease. Early studies were equivocal as to the relation between prostatic volume and the serum PSA level. A study by Kane and coworkers [17] indicates that it may be possible to standardize serum PSA values against volume as determined by transrectal ultrasound measurements. Vallancien and coworkers [18], although confirming a rough correlation between PSA levels and prostatic volume, noted wide variations in PSA levels on a per-gram basis in men with enlarged prostates.

Prostate-specific antigen assays that are commercially available include Tandem-R PSA (Hybritech, Inc., San Diego, California), Pros-Check PSA (Yang Laboratories, Bellevue, Washington), Tandem-E PSA, IRMA-Count PSA (Diagnostic Products Corp, Los Angeles, California), and Abbott IMX PSA (Abbott Park, Illinois). The Tandem-R and the Abbott IMX assays have been approved by the Food and Drug Administration for tumor burden monitoring in diagnosed cases. No PSA assay is approved by the Food and Drug Administration for screening or diagnosis.

Is the Prostate-Apecific Antigen Assay Useful for Cancer Screening?

Experience to Date

No randomized trial has tested the efficacy of screening for the early detection of prostate cancer. Nevertheless, information exists on the cancer detection capabilities of the digital rectal examination compared with the serum PSA assay [2, 19]. First, PSA assays detect more prostate cancers than do digital rectal examinations. On a prevalence screen, the Hybritech assay (with 4 ng/mL of PSA as the upper limit of normal) detects prostate cancer in about 2% to 2.5% of men older than 50 years compared with a rate of about 1.5% using a digital rectal examination alone [19]. The men in these studies had responded to direct mail advertising [20] or to press releases [21] inviting them to screening programs in two university hospital settings. Some studies have reported higher cancer detection rates in the setting of urologic practices, but a substantial proportion of patients in such practices are symptomatic, and the tests are, therefore, used in a diagnostic rather than a screening mode. Second, the PSA assay has been shown to detect about one third of diagnosed cancers in asymptomatic men that are missed by the digital examination, whereas the digital examination detects about 20% of those missed by using a PSA assay [19]. Some investigators have, therefore, concluded that neither test is sufficiently sensitive or specific as a solitary screening test [18, 19].

The PSA assay is not specific for prostate cancer. Serum levels can be increased with benign disease and are often normal with malignant disease. The overlap in PSA serum concentrations in men with benign prostatic hyperplasia alone compared with men with organ-confined prostate cancer is substantial [18]. Table 1 shows estimates of the sensitivity, specificity, and positive predictive value of the PSA assay and the digital rectal examination in the screening of 2425 men calculated from the American Cancer Society prostate cancer screening project, a multi-institutional demonstration project in which men of ages 55 to 70 years (mean age, 63 years), without known or suspected prostate cancer, were invited to receive annual screening [22]. As shown in Table 1, investigation of men who have either an abnormal PSA level or an abnormal digital rectal examination can result in increased sensitivity for prostate cancer but at a cost of decreased specificity [21]. Similarly, investigation of men only if both the PSA level and the rectal examination are abnormal, results in improved specificity and positive predictive value but at a cost of considerably lower sensitivity. Further, in current practice it is unlikely that physicians would demand that both screening tests be abnormal before testing for cancer.

However, men in the prostate cancer screening project with negative tests were not fully investigated for a diagnosis of prostate cancer, sufficient follow-up to determine all false-negative results was not possible, and not all men with abnormal tests had biopsies. Therefore, there was no gold standard for diagnosis against which the PSA assay or the digital rectal examination could be compared. Sensitivity is the number of men with prostate cancer who have abnormal tests (true-positive results) divided by the total number of men with prostate cancer as determined by a gold standard. Specificity is the number of men with a normal test who are proven not to have cancer (true-negative results) divided by the number of men who do not have the disease when fully investigated by the same gold standard. Therefore, accurate calculations of sensitivity and specificity are not possible from the American Cancer Society project (or any other report to date). They are likely to be lower than the calculations suggest. Because an unknown fraction of the men included in the demonstration project had symptoms, accurate determination of positive predictive values under true screening conditions are also impossible from the reported data. This follows from Bayes' theorem [23]: for a fixed sensitivity and specificity of any diagnostic test, the post-test probability of having the disease of interest (positive predictive value) varies with the pretest probability (prevalence) of disease. Because at least some prostatic symptoms are due to prostate cancer, the prevalence of cancer in a symptomatic group of patients is higher than in an asymptomatic screening group. A substantial number of symptomatic men would inflate calculations of the predictive value of an early detection test.

Because PSA levels tend to increase as prostate cancer progresses, some investigators have explored ways to improve the test characteristics of the PSA assay through the use of serial measurements. Carter and coworkers [24] measured PSA levels in serum that had been stored during a longitudinal study in aging men. They found increased specificity, with little loss in sensitivity. Several caveats exist, however, due to the retrospective nature of the analysis: interassay variability was kept at a minimum, because all stored samples were done simultaneously rather than at periodic intervals, as would occur in a true screening setting. Also, the sample size was small and might not have been fully representative (18 patients with cancer). Recent preliminary reports suggest problems. In a small prospective study of 224 consecutive men with digital rectal examinations that were not suspicious, a serial change in PSA levels during a 12- to 36-month period before prostatic biopsy did not discriminate benign from malignant disease any better than did single determinations [25]. Of additional concern is a retrospective analysis by Riehmann and coworkers [26] of an unexpected increase in PSA levels in consecutive samples drawn from 13% of 129 men with no underlying malignancy. The authors of the study concluded that it is doubtful whether a serial increase in PSA levels indicates the need for invasive testing for prostate cancer.

Uncertainties about the Efficacy of the Prostate-Specific Antigen Assay as a Screening Tool for Cancer

Five main criteria are necessary for a test for mass screening of asymptomatic people to be useful [27]. First, the disease in question should represent a substantial burden at the public health level and should have a prevalent, asymptomatic nonmetastatic phase. This is certainly true in the case of prostate cancer. There are nearly 160 000 patients with prostate cancer reported each year in the United States, making it the most common nonskin cancer in men. It is the second most common cause of cancer death in U.S. men, and about 35 000 deaths per year (nationwide) are attributed to prostate cancer [28]. At autopsy in men older than 50 years who died of causes unrelated to prostate cancer, latent cancers of the prostate are found in nearly 30%. About two thirds of men older than 60 years who are examined by autopsy have subclinical prostate cancer discovered at that time [29-33].

Second, the asymptomatic, nonmetastatic phase should be recognizable. This is also true without doubt for prostate cancer. Many cases are incidental findings at transurethral resection done for benign prostatic hypertrophy, and many are found after routine rectal examination of asymptomatic men.

Third, a good screening test should be available that has reasonable values of sensitivity, specificity, and predictive value, is of low risk and low cost, and is acceptable to both the screener and the person screened. High specificity is particularly important in screening patients in whom no clinical suspicion of cancer exists, to avoid large numbers of false-positive results. As stated above, the sensitivity, specificity, and predictive value of the PSA assay are unknown in the screening setting. Because it involves only the drawing of blood followed by a laboratory assay, the test itself appears to be acceptable and of low risk. On the other hand, procedures that follow a positive test result may carry substantial risk for morbidity or mortality or both (see below), and false-positive results incur unnecessary attendant anxiety. The cost of the initial screening test ranges from $25 to $60, which is reasonable compared with accepted screening tests for other diseases such as breast cancer.

Fourth, curative potential should be substantially better in early compared with advanced stages of disease. In the case of clinical prostate cancer, this criterion is often assumed to be true, because men with disease confined to the prostate have 10-year relative survival rates of 75% compared with 55% in those with regional extension beyond the prostatic capsule or about 15% in those with distant metastases (Surveillance, Epidemiology, and End Results [SEER] program. Unpublished data). Favorable survival is, therefore, typically restricted to those persons diagnosed at early stages of disease. The contribution of therapy to this favorable survival in early stages is not known, however. A recent Swedish population-based study [34] showed a 10-year survival rate of 87% in 223 men with early stage, predominantly well-differentiated clinical prostate cancer who had no initial treatment at all. The survival rate was also 87% after 10 years in the subset of 58 men with clinical stage B disease (palpable but organ confined) who would have been candidates for radical prostatectomy but who were, nevertheless, followed without initial surgery. In another study [35], 50 men with well- and moderately well-differentiated cancer that had invaded through the prostate capsule but who had no clinically detectable metastases were followed with no initial therapy, often because of patient preference. The cancer-specific survival rate was 88% and 70% after 5 and 9 years, respectively. Because this was not a population-based series, more selection factors existed than in the Swedish series above. Nevertheless, it does show that even men who have prostate cancer with extracapsular invasion may have an indolent course. It is also important to point out that all reported treatment series of prostate cancer, such as radical prostatectomy, are also highly selected when compared with population-based series. A recent structured literature review [36] of treatment for localized prostate cancer concluded that all published series were affected by selection factors and that these factors prevented definitive statements about the relative efficacy of radical prostatectomy, definitive radiation, and deferred treatment until disease progression. Clearly, some men have highly aggressive disease although most do not, and we do not have tests that can reliably differentiate the two subsets prospectively [37]. Less clear is whether treatment of men with highly aggressive disease changes the natural history, because this has never been definitively tested in a randomized, controlled study (is cure possible in those for whom it is necessary? [1].

The fifth criterion of utility for a screening test is that treatment of screen-detected patients should improve outcome as measured by cause-specific mortality. This last criterion is the most important as well as the most difficult to fulfill and has not been reported to date. The best evidence would come from randomized trials. Any other study design can be affected by three major biases: selection bias, lead-time bias, and length bias [38].

Selection bias can occur whenever the group actually screened differs from the universe of potential screened subjects (Figure 1). The actual screened patients may have better survival rates regardless of screening. Lead-time bias occurs when a screening test advances the time of diagnosis by detecting disease before the onset of symptoms but does not affect the natural history of the disease and does not change the time of death (Figure 2). In essence, the patient merely spends a longer time aware of the disease. With length bias, disease is detected between scheduled screening tests because of the onset of symptoms (Figure 3). These symptomatic cancers are typically biologically more aggressive than screening-detected asymptomatic cancers and, therefore, are associated with worse survival. Conversely, the screening-detected, more slowly growing cancers would have better outcome even in the absence of screening. Comparison of survival between the two groups of patients is, therefore, biased in favor of screening. These biases are not simply theoretical; they have been documented in lung as well as in breast cancer screening studies [39, 40]. Length bias has also been suggested in a prospective cohort study of screening for prostate cancer using the digital rectal examination [41]. In that study, the 10-year prostate-specific survival rate was superior for men diagnosed at the initial screening rectal examination compared with those diagnosed after subsequent examinations.

Potential exists in prostate cancer screening for a special type of length bias known as over-diagnosis (Figure 4 )[38]. This occurs when prostate cancer is detected through screening, is treated, and is apparently cured, and the patient dies of another cause. If the cancer had never been detected, the patient would still have died of the same unrelated cause at the same time. The unusually high prevalence of incidental and indolent carcinomas of the prostate makes over-diagnosis a real possibility. These generally small tumors probably have such a slow rate of growth that they do not become clinically important within a man's lifetime, persisting undetected for years [37]. This possibility is strengthened by the Swedish population-based study of untreated patients noted above [34], in which after 10 years of follow up, only 8.5% of 223 patients with clinically confined disease had died of their prostate cancer.

When indolent tumors are detected by PSA screening or by other means, the treatment using radical surgery or radiation may give the false impression of a cure. Whitmore and coworkers [42] noted that the rapid increase in the incidence of clinical disease with age, concurrent with a progressive risk for death from other causes makes quality survival (with disease) a legitimate alternative to cure in selected patients with prostatic cancer. Growing older is invariably fatal; prostatic cancer is only sometimes so!

Potential Harm of Mass Screening Assays for Prostate-Specific Antigen

Currently available tests cannot reliably distinguish tumors that are destined to be lethal but are still curable [37]. For this reason, many men must be treated in an attempt to benefit those who have potentially lethal disease, and therapy may have negative as well as positive outcomes. The 1987 consensus development conference from the National Institutes of Health on the management of early-stage prostate cancer stated that either radiation or radical prostatectomy had equivalent efficacy in the treatment of tumors clinically confined to the prostate [43]. Several complications of these treatments were noted, including rectal injury and the need for colostomy, impotence, incontinence, urethral stricture, and therapy-related mortality. Optenberg and Thompson [44] recently reviewed and summarized the pertinent literature on treatment complications for clinically nonmetastatic prostate cancer. Their findings are summarized in Table 2. The 25% impotence rate associated with radical prostatectomy may be overly conservative, because it presumes the use of a nerve-sparing procedure. Many urologists do not do this procedure, and the impotence rate for standard radical prostatectomy is far higher.

We estimated the post-surgical mortality after radical prostatectomy using the SEER program data from the National Cancer Institute [45]. The SEER program consists of a series of nine geographically distinct, population-based tumor registries, covering approximately 10% of the U.S. population, that collect clinical information about cancer patients in their respective areas. The SEER data are less subject to selection bias inherent in reports from individual referral centers and are routinely cited as the definitive source for estimates of national cancer incidence and survival [28]. For men ages 50 to 74 years, we calculated a 4-month, post-surgical mortality rate of 1.2%. Based on inspection of the SEER program data, a 30-day period for estimation of post-surgical mortality did not appear to adequately capture the total mortality burden after surgery. The figure we estimated is consistent with other published reports of between 1% and 2% [44, 46]. By way of comparison, the national 30-day postoperative mortality rate after radical prostatectomy reported by the Health Care Financing Administration is 0.9% [47]. No published national mortality figures exist for prostate radiation therapy.

We estimated the number of complications and treatment-related deaths in the first year if 21.2 million U.S. men, ages 50 to 74 years, were screened using the digital rectal examination and the assay for PSA. This is less than the total of 23.8 million men in this age group, because we conservatively excluded men with high cardiovascular risk (about 11%) [48]. Using the weighted average detection rates in the Appendix, about 3.1% of asymptomatic men in this age group would have cancer [19]. Of these, roughly 95% should have clinically locally confined tumors [21, 49]. The use of radical prostatectomy among men with clinically confined tumors who are younger than 70 years has increased from 20% in 1983 to 60% in 1989, based on SEER program data [50]. We estimated that roughly two thirds of men in this age group would have radical surgery, because the trend toward surgery was still increasing in 1989. Using the estimates of Optenberg and Thompson [44] for complication rates (see Table 2), this would imply more than 250 000 instances of complications severe enough to require medical attention. Based on the treatment-related mortality of 1.2% after surgery and 0.5% after radiation, 5510 excess deaths would occur in the first year. (Using somewhat less conservative assumptions, Optenberg and Thompson estimated that 26 197 treatment-related deaths would occur in the first year if all men ages 50 to 70 years were screened ). We calculated the effect of various PSA cutoff values and the age of screened men on our estimates of treatment-related deaths (Table 3) as follows: 1) a screening program using an upper limit of normal for PSA of 3 ng/mL, as suggested by Labrie and coworkers [51]; 2) using an upper limit of normal for PSA of 10 ng/mL; and 3) a screening program in men ages 50 to 64 years rather than 50 to 74 years. The estimated number of excess treatment-related deaths would be 6241, 2087, and 2293, respectively. Although limiting the age of screened men to 50 to 64 years would decrease the number of treatment-related deaths, only about 53% of the cases detected in men ages 50 to 74 years would be detected in the more restricted age range, thus limiting the potential benefits as well.

Using our baseline assumptions, screening would have to save at least 5510 lives to exert a net mortality benefit. Complicating any risk:benefit assessment is the fact that the benefits of screening in terms of lives saved would be delayed by some unknown but substantial interval because screening introduces considerable lead time and the median survival of men with prostate cancer in the absence of mass screening is at least 5 years (SEER. Unpublished data).

In the absence of over-diagnosis, potential exists to avert far more than 5510 deaths from prostate cancer in the screened cohort. A 10-year, prostate-specific death rate of 8.5% among untreated patients [34] applied to the 592 000 additional patients detected in the first year of screening would yield an estimate of more than 50 000 deaths. Presuming that treatment is effective, the potential deaths averted by a prostate cancer screening program could thus more than offset the initial treatment-related mortality (although obviously not from the perspective of the thousands of persons who die of a treatment-related cause).

Because a definitive trial of the efficacy of immediate compared with deferred therapy for prostate cancer has yet to be done, some investigators have used decision analysis methods to weigh the risks and benefits of therapy for clinically organ-confined prostate cancer compared with watchful waiting with symptomatic care. Even when favorable assumptions were made about therapy, including a 100% efficacy rate for the cure of disease confined to the prostate, benefits in the model reported by Mold and coworkers [52] were modest. The gain in absolute life expectancy averaged 1.1 months. Additionally, when results were presented in terms of quality-adjusted life years, a net loss of 3.5 months occurred for the average man having definitive therapy compared with observation. Fleming and coworkers [53] used a decision analysis model to compare radical prostatectomy, external-beam radiation therapy, and watchful waiting with delayed hormonal therapy if metastatic disease developed [53]. The potential benefits of immediate therapy were small enough to be sensitive to many of the model assumptions. The authors concluded that watchful waiting is a reasonable alternative to surgery or radiation for many men. Two other preliminary decision models have been reported in abstract form, arriving at similar results. Schiffman and coworkers [54] used Monte-Carlo simulation models to estimate the net risk:benefit ratio of screening men ages 50 to 74 years with digital rectal examinations and assays for PSA compared with the strategy of no screening at all. For more than 90% of the iterative simulations, neither rectal examinations nor PSA assays was the preferred strategy using quality-adjusted life days as the outcome variable. Similarly, Krahn and coworkers [55] estimated from their decision analysis model that a screening program using the PSA assay as the early diagnostic test would change the average life expectancy of a 60-year-old man from 19.76 to 19.77 years; however, a net loss would occur in quality-adjusted life years. Conclusions were the same even if the sensitivity and specificity of the PSA assay were both set at 100%. One contributing factor to the marginal gains in life expectancy even using perfect sensitivity and specificity characteristics for the PSA assay is the fact that, other than nonmelanoma skin cancer, prostate cancer has less impact on average years of life lost per person dying of the disease than any other malignancy [56].

Predictions from models are inherently much less reliable than results from prospective randomized trials. Until mortality reduction benefits of the PSA assay have been determined definitively, questions about net benefit versus harm cannot be answered, and its use for screening of asymptomatic men is merely investigational. In the interim, men should be fully apprised of the potential downstream risks and benefits that could be incurred by the screening test and of the lack of definitive evidence that it favorably affects mortality or that resulting treatment improves quality-adjusted survival. This educational process should take place before the test is done, rather than after it is found to be abnormal.

Cost Implications

In the absence of a controlled clinical trial with prospective ascertainment of costs at the individual level, it is impossible to determine the net cost of a mass screening program for prostate cancer. Uncertainty exists about treatment cost savings that might be associated with shifts to earlier stage disease and the potential cost increases from over-diagnosis. It is, however, possible to provide a descriptive account of what level of costs would be implied in the first year of a mass screening program. Under the assumptions listed in the Appendix, we have estimated the first-year cost of treating the extra cases of prostate cancer that would be detected in a hypothetical mass screening program for all men ages 50 to 74 years (excluding those with preexisting cardiovascular disease) to be about $11.9 billion (Table 4). Again, using the three additional screening assumptions used in our calculations of treatment-related deaths, we calculated a range of total costs from $5.2 billion to $14.1 billion (Table 3).

Optenberg and Thompson [44] have independently estimated the costs of the first-year mass screening effort for prostate cancer using PSA and have arrived at an even higher estimate of $27.9 billion. The higher estimate they obtained is due mainly to their assumption that screening with PSA assays would result in treatment for prostate cancer in 9.3% of all men screened, about double the rate we assumed.

Based on SEER incidence estimates, prostate cancer would be detected in 62 476 men in this cohort in the absence of a mass screening program, resulting in treatment costs of about $1.35 billion. Therefore, the additional cases of prostate cancer detected in the first year of the mass screening program would result in increased costs for treating this age group by about ten times. In the absence of over-diagnosis, the extra first-year costs should eventually be balanced by lower costs in the later years, because fewer cancers would be diagnosed subsequently in the screened patients. Notwithstanding, in the American Cancer Society program a substantial number of cases of cancer were detected in years 2 and 3 [57]. So the cost of the first 3 years of the program would total considerably more than $11.9 billion, perhaps an additional 50% or more. To recoup this cost the detection rate would have to decrease to nearly zero for the next decade, or more, of the program. Given that the ratio of the first-year detection rate to the observed SEER incidence rate of prostate cancer is in the range of 10 to 20, the detection rate would be expected to decrease markedly within the first few years of the program in the absence of over-diagnosis. If prostate cancer detection rates were to continue at substantial levels beyond the first few years of the program, the total lifetime cumulative probability of the occurrence of clinical prostate cancer would soon be surpassed; and program costs would continue to increase without compensating savings in the out-years.

What Do Professional Groups Recommend?

Professional disagreement exists about the value of the PSA assay for the early detection of prostate cancer. The American Urologic Association (Executive Committee Report, January, 1992); the American Cancer Society (Board of Directors Meeting, November, 1992); and the American College of Radiology (Resolution #36, approved October, 1991) recommend the use of the PSA assay for screening. On the other hand, the U.S. Preventive Services Task Force report [58] states that serum markers, including PSA, are not recommended for routine screening. The Canadian Task Force on the Periodic Health Examination recommends [59] not using the assay, stating that there is fair evidence to exclude (the PSA assay) from the periodic health examination of asymptomatic men older than 40 years. The National Cancer Institute [60] does not recommend using the assay for routine screening until more information becomes available.

Controlled Evaluation of the Prostate-Specific Antigen Assay for Screening

Because of the uncertainty surrounding the efficacy of screening for prostate as well as a number of other cancers, the National Cancer Institute has launched a major screening trial, called the Prostate, Lung, Colon-rectum, and Ovary cancer screening trial to evaluate early detection techniques for these four diseases. In this randomized, controlled trial, the PSA assay and the digital rectal examination are being evaluated together to determine whether screening decreases prostate cancer mortality.

The prostate portion of the Prostate, Lung, Colon-rectum, and Ovary trial evolved out of a recognized need to evaluate methods of screening for this disease. Thirty-seven thousand men are to be screened annually four times and followed to determine if screening decreases mortality. The digital rectal examination and the PSA assay will be done at entry followed by three additional annual tests. An equal number of men will receive usual medical care. Morbidity and mortality associated with screening and treatment, costs, as well as test sensitivity, specificity, and positive predictive value will be determined. All men in the trial will be followed for at least 10 years from entry into the trial to ascertain incident cases plus date and cause of death for every fatality that occurs. The sample size provides slightly more than 90% power to detect a 20% decrease in prostate cancer mortality using a one-sided test with a type I error (false declaration of efficacy) probability of 0.05. The power would, of course, be decreased in proportion to the number of men in the control group who seek screening outside of the study. For this reason, all men will be fully apprised of the uncertainties, pros, and cons, of prostate cancer screening as part of the informed consent process. Further, compliance in the study group and contamination in the control group will be carefully monitored during an initial 2-year pilot phase of the study. If necessary, sample size could be increased to compensate for unexpectedly high levels of crossover.

Results come slowly in cancer screening trials, because it takes years to recruit and randomize enough patients and to conduct the screening. Also, the mortality trends in control and screened groups must be followed for an adequate period to permit confident statistical comparison. It usually takes at least 4 and perhaps 7, 10, or even more years to determine if early detection leads to decreased death rates. No short cuts exist.

Given the importance of the issues and the debate surrounding the risk:benefit ratio, the time and effort devoted to such a trial is appropriate. Attempting to resolve the issue of whether screening results in a benefit, and whether the benefit outweighs the harm, by any means other than a randomized trial would involve the use of an observational study design at best, or no design at all in the worst case. Observational screening studies are known to involve biases that render conclusions suspect [38], prolonging the debate and the process of deciding whether an intervention is of value. If at worst no design at all is used but instead large segments of society are encouraged to participate in PSA testing, a net harm, should it occur, may go undetected. If the history of medicine has taught us nothing else, it has taught us that interventions that seem reasonable based on the current medical paradigms may ultimately prove to be worthless or even harmful. Whenever a new medication is formulated for the treatment of people with disease, meticulous testing is required before it is approved for widespread use. The process can take years, but the price is worth the benefit of avoiding previous medical debacles. We should demand no less of new screening tests for use in healthy persons. Until the answers are in, we must be content to live with uncertainty rather than assume efficacy. As Daniel Boorstin, historian and former director of the Library of Congress, has pointed out: illusions of knowledge are the obstacles to discovery [61].

Abbreviation

PSA: prostate-specific antigen

SEER: Surveillance, Epidemiology, and End Results program