Screening for prostate cancer: evidence and perspectives

Authors

  • F.H. Schröder,

  • M.F. Wildhagen,

  • and the Rotterdam Study Group of the ‘European Randomised Study of Screening for Prostate Cancer’ (ERSPC)


F.H. Schröder, MD, Professor and Chairman, Department of Urology, Erasmus University & Academic Hospital Rotterdam, Dr Molewaterplein 40, PO Box 1738, 3000 DR Rotterdam, the Netherlands.

A summary of the SIU Lecture 2001, presented at the Annual Meeting of the European Association of Urology, Geneva, 6 April 2001

Introduction

The early detection of prostate cancer is feasible; the use of PSA-driven screening techniques produces a ‘lead time’ and advances the diagnosis of the disease by 5–10 years [1–3]. While early detection regimens dramatically shift the stage at the time of detection toward locally confined disease, and while there is no debate that early aggressive treatment is the only way of curing prostate cancer, the application of screening in practice and healthcare policy remains controversial. This is because there are uncertainties about the degree of ‘over-diagnosis’ which most probably results from screening, and from uncertainties about the risk-benefit balance.

This paper expands on the Société Internationale d'Urologie lecture given at the Annual Meeting of the European Association of Urology in 2001. In it we address four practice-related issues of screening for prostate cancer by reviewing recent literature and using recent data obtained within the ‘European Randomised Study of Screening for Prostate Cancer’ (ERSPC) Section, Rotterdam [4].

Early detection—a practice policy?

There is increasing indirect evidence that early detection and early aggressive treatment may reduce prostate cancer mortality. Cancer-related mortality is the most important end-point in establishing the effectiveness of screening. Survival will always be better in historical or randomized comparisons because of the production of ‘lead-time’ by early detection measures. However, any judgement of the value of screening as a public health issue will be based on ‘quality-of-life adjusted gain in life years’ (QUALYs)

Evidence for the effectiveness of screening and early treatment

There is evidence of decreasing prostate cancer mortality in several areas of the world, including those where screening is not prevalent, e.g. the UK [5] and the Netherlands [6]. However, the decrease in mortality may be more pronounced in areas where screening is prevalent, as in the USA [7] and in the area of Innsbruck where a prospective study was carried out [8]. To date, the results of only one randomized study have been reported, which was conducted in the area of Quebec City; there was an improvement of 67% in the relative risk of dying from prostate cancer in a screened population [9]. However, the study was methodologically flawed and has been heavily criticised [10,11]; plans for a re-analysis are currently being established.

In the USA, prostate cancer mortality has decreased by 17.6% since 1993 [7]; Fig. 1 reproduces the data on prostate cancer mortality obtainable from the website of the Center for Disease Control in the USA [12]. The plot shows crude rates of prostate cancer mortality per 100 000 men for the USA as a whole, for the area of Washington State/Seattle where screening is more prevalent than in the rest of the country, and the regression line depicting the trend of increasing prostate cancer mortality from 1979 to 1987. From 1988 onwards there was a more pronounced increase in prostate cancer mortality, which cumulates in 1993, after which the trend reverses, so that for 1996 and 1997 there is for the first time a decrease in prostate cancer mortality below the regression line indicating the previous trend of increase. This increase followed by a decrease in prostate cancer mortality has been intensively investigated and discussed. Registration problems, e.g. the phenomenon of the ‘sticking diagnosis’, which lead to the establishment of incorrect causes of death on death certificates, are blamed for at least part of the initial increase and later decrease in the mortality rates [13]. However, it is likely that at least part of the decrease since 1993 can be explained by screening [14] and the more frequent use of potentially curative treatment, which started in the mid-1980s with the improvement in radical prostatectomy and radiotherapy [15].

Figure 1.

Prostate cancer mortality in the USA (red squares and line) and in the area of Seattle, Washington State (1979–97; (green circles and dotted line). The regression line (black) of the increase in prostate cancer mortality from 1979 to 1987 is included [12].

Over-diagnosis

Over-diagnosis is currently poorly defined, but worse is that available definitions cannot be applied to individual patients because of the lack of conclusive prognostic information at the time the treatment decisions are taken. Over-diagnosis could be defined strictly as the rate of diagnosis of prostate cancer divided by the chance of dying from the disease. Such data are indicated in Table 1 and relate to the ERSPC Rotterdam; to correct for age the rates are reported per 1000 man-years of observation. The underlying mortality in the Netherlands was calculated for those aged 55–85 years. According to this definition, the chance of being diagnosed as opposed to dying from prostate cancer was 14.6 times and 2.3 times in the screening and in the control groups, respectively. The lead-time was not taken into account in this evaluation.

Table 1.  Incidence/mortality ratios; the cumulative incidence of prostate cancer and underlying mortality, in men aged ≥ 55 years. The assumptions are that the mortality rate in the ERSPC Rotterdam was the same as in the Netherlands in 1997 and that mortality is not influenced by the screening programme. The threshold date for follow-up in the ERSPC is 31 December 1999
VariableScreenedControl
No. of men21 15621 132
No. of person years55 286.557 133.2
No. with prostate cancer1190189
Crude incidence  (/1000 person years)21.523.31
Mortality rate  (/1000 person years,  the Netherlands, 1997)1.471.47
Ratio, screened/control 6.51
Ratio, incidence/mortality14.632.25

Another way of assessing over-diagnosis is by not only considering death from prostate cancer but also the morbidity of clinically progressive disease. This might be approximated by dividing the rate of men diagnosed in the screening arm (Table 1) by the rate of the clinically diagnosed cases in the control arm. With this definition the resulting rate of over-diagnosis would differ by a factor of 6.5. The incidence/mortality ratio is very large when compared with the most frequent values from the USA, where the estimated numbers of diagnosis and deaths from prostate cancer are 198 100 and 31 500, giving a ratio of 6.28 [7]. This discrepancy is not surprising; not every man in the age group at risk in the USA is screened. In this sense the situation in a randomized controlled study is closer to the reality of over-diagnosis by screening. Almost 95% of men randomized to screening are indeed screened and biopsied if indicated. The incidence in the control group reflects regional healthcare routines. The incidence of prostate cancer in the Netherlands has increased by 43% in recent years [16], a modest rise compared with that in the USA.

Quality of life

Even if an advantage of screening and early treatment of prostate cancer could be shown conclusively, the advantage in mortality must be balanced against changes in the quality of life (QoL) related to treatment. QUALYs and cost-effectiveness (cost/QUALY gained) will be the variables by which to judge the value of screening. Related data are only just emerging. A recent short-term evaluation of QoL after radical prostatectomy and radiotherapy showed no relevant differences between screen-detected and clinically diagnosed cases [17]. Litwin et al. describe further details related to this approach [18].

What is practice policy?

It is anticipated that ongoing randomized studies (the ERSPC in Europe and the Prostate, Lung, Colon and Ovary randomised screening study of the National Cancer Institute in the USA and their associated QoL studies) will resolve the pending problems or at least contribute in a major way. However, conclusive data will not be available before 2005 and more probably 2008. The problem is what can be done in the interim. While it is uncertain whether screening a population will contribute to reducing prostate cancer mortality, there is little doubt that early prostate cancer can be eradicated and individual men ‘cured’. The uncertainty arises from the impossibility of predicting whether a given individual benefits from an aggressive approach, in terms of decreasing his chance of premature death from prostate cancer, with an acceptable QoL. As shown later, there is worldwide consensus that testing cannot be refused to well-informed men who have decided to accept the risks involved and who feel that the benefit will outweigh the risks.

There are three decision points; the decision to be tested, the decision to undergo a biopsy and the decision to be treated if the biopsy is positive. Men with very low PSA levels (< 1.0 and probably < 2.0 ng/mL) should not undergo further diagnostic evaluation [19,20]; age must also be a factor in the decision. The appropriate indication for taking a biopsy and the numbers of biopsies to be taken are subject to intense discussion. While it is uncertain on the basis of available prognostic factors which men are not at risk of dying from prostate cancer, the best available epidemiological data show that men with a locally confined prostate cancer and a well differentiated tumour (by Gleason score) have only a 9% chance of dying from prostate cancer within 15 years [21,22]. Obviously, little can be gained by aggressively diagnosing and treating this group of men; the problem is in their accurate identification by pretreatment variables [23]. Thus men with a low PSA level and minimal findings of well differentiated cancer on biopsy are the best candidates for active surveillance in watchful-waiting regimens. (Currently, watchful waiting is not an established form of management of locally confined disease, because there are no comparative studies. The threshold for initiating active treatment is, among other features, poorly defined.)

In summary, it is currently uncertain whether screening lowers prostate cancer mortality. There is likely to be a high rate of over-diagnosis if PSA-driven screening policies are applied to asymptomatic men. Uncertainty also persists about the best methods of screening and treatment, and whether there is more benefit than damage. Finally, the request by well-informed patients to be screened must be fulfilled. The development of validated information tools is desirable and a clear definition of what information should be provided to men who wish to be screened [24]. Fortunately, an increasing volume of recent reports address these issues.

How best to detect prostate cancer?

The most commonly used method of detecting prostate cancer is based on the serum PSA level, with geqslant R: gt-or-equal, slanted 4 ng/mL indicating the need for a biopsy. If the PSA level is leqslant R: less-than-or-eq, slant 4 ng/mL an abnormal or suspicious DRE also indicates the need for a biopsy. The positive predictive value (PPV, the number of cancers found in those who have a positive test) also depends on age but is only ≈ 10%, 25% and 50–60% in the PSA ranges leqslant R: less-than-or-eq, slant 4, 4–10 and > 10 ng/mL, respectively [25]. To express the accuracy of a test, the ‘sensitivity’ is used, defined as the number of true-positive tests in all those who have cancer. As the denominator (‘all those who have cancer’) is unknown, the term ‘sensitivity’ cannot be applied strictly. One approximation is to use the term ‘relative sensitivity’, where the total number of cancers detected in a given regimen (the number of positive biopsies) is used as the denominator and considered the ‘gold standard’. By extrapolation or logistic regression analysis it is possible to determine the number of cancers that can be found in a given population if every participant is biopsied. Such calculations have been carried out for those aged 55–75 years, and show that in men with PSA values of leqslant R: less-than-or-eq, slant 4 ng/mL, two-thirds of the cancers are missed that could have been detected if every man were biopsied [26]. The crucial issue is whether those cancers that are missed would contribute to the decrease in prostate cancer mortality; are they life-threatening to their carriers? In cancers detected in men with a PSA of leqslant R: less-than-or-eq, slant 4.0 ng/mL by a DRE there is a direct relationship between PSA level on one side and T category and Gleason score on the other. Half of the cancers detected in men with a PSA of 2–3.9 ng/mL and who have subsequently undergone radical prostatectomy have a Gleason score of geqslant R: gt-or-equal, slanted 7, and are organ-confined. At the other end of the spectrum, ≈ 30% of cancers are too advanced for cure and these are usually in men with PSA values of > 8 ng/mL.

What then is the right ‘window of opportunity’ to diagnose prostate cancers which will then contribute to decreased prostate cancer mortality? Large ongoing randomized studies are a constant source of valuable information in relation to test procedures. Within the ERSPC Rotterdam, a DRE in men with PSA values of leqslant R: less-than-or-eq, slant 1.0 ng/mL had a PPV of only 2.2%, i.e. almost 50 biopsies were necessary to detect one cancer [19]. All the cancers were classified as ‘minimal’[27]. It was then decided not to use a DRE in men with a PSA value of leqslant R: less-than-or-eq, slant 1.0 ng/mL, i.e. 37% of the total population of men aged 55–75 years. In those with a PSA of 3–3.9 ng/mL, a DRE had a PPV of 6.3%, while logistic regression predicted a PPV of ≈ 12% if every patient in this PSA range was biopsied [28,29]. A validation study was carried out in almost 8000 men to verify this calculation and to study the variables associated with aggressiveness in cancers found in this PSA range; the PPV was then 18%, even higher than expected. In those men who underwent radical prostatectomy and had cancers detected by PSA alone in the validation protocol, the distribution of variables of aggressiveness in cancers detected by DRE was virtually identical. About half had organ-confined disease and concomitant Gleason scores of geqslant R: gt-or-equal, slanted 7 [30]; a mean of 4.7 biopsies were needed to detect one prostate cancer in this PSA range.

As detection rates remained virtually identical on including routine biopsy in the PSA range 3–3.9 ng/mL and by omitting the DRE in men with a PSA of < 3.0 ng/mL, and because the characteristics of most cancers in those with a PSA of < 3.0 ng/mL suggested the possibility of detection in the second screening round while the disease was still curable, for the purpose of the ERSPC protocol it was decided to omit the DRE completely and to biopsy every man with a PSA of geqslant R: gt-or-equal, slanted3.0 ng/mL [28,30].

Hence, the best way of detecting prostate cancer remains unknown; obviously, it is desirable to detect all cancers that threaten the life of their carriers, and to avoid diagnosing cancers that will not cause morbidity or lead to death. Such regimens would have to consider cancer-related and general health-related prognostic factors. While it will probably be impossible to reach this goal, an approximation is needed and will result from further study, especially in a randomized setting or in carefully designed case-control studies. The effort should concentrate on improving test specificity (avoiding unnecessary biopsies) in the lower PSA ranges. The available variables of tumour aggressiveness must be correlated with outcome. In a retrospective analysis of variables at the time of diagnosis and of natural PSA histories it will probably be possible to improve test procedures. Improving specificity may permit the detection of more relevant, confined cancers with fewer biopsies. Many cancers are missed by not taking biopsies in groups of men at risk, and possibly by not taking enough biopsies. Eskew et al.[31] were the first to report that increasing the number of biopsies to an average of 13 and changing their location increased the cancer-detection rate by ≈ 36%. However, just by taking the standard sextant biopsies in a more lateral position the detection rate could also be improved by 31%. It will be crucial to study as many modifications of the testing procedure as possible in a randomized controlled setting; fortunately, such initiatives have been taken.

Watchful-waiting and immediate treatment

Surveillance or watchful-waiting as a treatment option is often used when managing locally confined disease. Table 2 summarizes the types of treatment applied in the ERSPC Rotterdam for 756 prostate cancers detected during the first screening round, or 4 years later during the second screen. The increase in the choice of watchful-waiting from 10.5% to 26.6% can be partly explained by the more favourable prognostic factors at the time of diagnosis. Clearly there is also a trend towards the more frequent use of watchful-waiting in the Rotterdam region. In the USA the use of watchful waiting has been relatively constant with time, at ≈ 30%, and (contrary to the screening study) included men with T1a and T1b tumours detected at the time of prostatectomy for benign disease [15]. In a population of 4458 men included in the CaPSURE database, 329 (7.4%) were treated by initial surveillance. Comparing the characteristics of the surveillance group with 4129 patients treated otherwise, an age > 75 years, being white, clinical stage leqslant R: less-than-or-eq, slant T2, a PSA of < 10 ng/mL and a Gleason score of 2–6 were significantly more common in the surveillance group (P = 0.001). Just over half (52%) of the men initially treated by surveillance eventually chose to be treated; the characteristics of this group were that they were younger, had a higher cT category and higher PSA values. The increasing PSA level was the main factor determining therapeutic decisions. Within a 5-year period, 23 patients (7%) of those who chose surveillance had died; prostate cancer was considered the cause of death in only three of these men [32].

Table 2.  The initial therapy for 756 screen-detected prostate cancers detected at Visit 1 or 4 years later at Visit 2 (ERSPC Rotterdam 2001)
TreatmentVisit 1Visit 2
Radical prostatectomy217 (40.9)40 (35.4)
Radiotherapy242 (45.6)42 (37.2)
Watchful waiting56 (10.5)30 (26.6)
Endocrine therapy15 (2.8)1 (0.9)
Pending10102
Total541215

There is strong evidence that progression rates are not only very slow in men with low PSA values, T1c cancers and Gleason scores of < 6 who are considered to have non-aggressive tumours, but also in cancers which must be classified as aggressive by current standards. The delayed-treatment control arm of an EORTC protocol (30846) gave an opportunity to study progression rates in lymph node-positive men who had not undergone radical prostatectomy [33]. With a mean follow-up of 41 months there was no progression in 21 of 57 men. Another study [34] reported the results of watchful-waiting in 162 men classified as having T1b-T2cN0M0 disease, a Gleason score of leqslant R: less-than-or-eq, slant7 and a PSA of leqslant R: less-than-or-eq, slant 15 ng/mL. The PSA doubling time (PSADT) was evaluated during the follow-up. The study recruited patients from 1995 and 113 patients were followed for at least 1 year. The median (range) PSA at entry was 6.2 (0.5–14.6) ng/mL and the median follow-up 21 months. The distribution of PSADT was: leqslant R: less-than-or-eq, slant2 years, 20 patients; 2–3 years, 12; 3–4 years, 14; 4–5 years, nine; 5–10 years, 19; 10–20 years, nine; 20–50 years, five; and geqslant R: gt-or-equal, slanted50 years, 25 patients. In 35% of these men the PSADT was >10 years. While the previous study [35] suggested that a rise in PSA of geqslant R: gt-or-equal, slanted20% was the most relevant predictor, the PSADT is another tool that needs further evaluation.

The results cited may give a general framework and several landmarks for individual decision-making. Until the proper thresholds for initiating a treatment that would not allow progression to incurable disease have been defined, watchful-waiting remains, to some extent, a risk. However, the slow progression rates of many cancers with favourable characteristics at the time of diagnosis suggest that surveillance is an option in selected groups of patients for limited periods.

In summary, the indications for watchful-waiting, follow-up procedures, and thresholds for initiating treatment are poorly defined. A PSA determination, DRE and possibly TRUS every 3–6 months seem to be standard procedures. The PSADT, percentage increase in PSA or PSA velocity are likely to be useful variables in deciding future therapy during surveillance. Prognostic factors at entry, at least in one study, show a poor correlation with PSADT in patients with progressive disease after radical prostatectomy [36].

Screening policies worldwide

Screening for prostate cancer has clearly become a political issue; in the USA, the Medicare system pays the costs of screening. Even in the UK, where there are critical attitudes about the diagnosis of early prostate cancer, the Secretary of Health stated in an interview with ‘The Times’ on 25 March 2001, ‘Tackling prostate cancer has become a top priority’. He proposed a scheme of informed choice; if, after careful information, a man still wishes to be tested, a PSA test must be carried out and paid for by the NHS. The speedy processing of abnormal tests and a biopsy within 2 weeks is provided and paid for by the NHS. In Germany, where screening by DRE has been part of a preventive package for many years, there are also preparations to promote PSA-driven prostate-cancer screening as a healthcare policy.

A summary of policy statements issued in the recent past is given in Table 3. The AUA and the American Cancer Society both issued statements in 1992 in which screening by PSA testing and DRE was recommended for men aged geqslant R: gt-or-equal, slanted 50 years and for risk groups when aged geqslant R: gt-or-equal, slanted 40 years [35]. The recommendation included the advice to ‘take a personal decision after consultation’. The guidelines of the American Cancer Society have been revised in 2000 and published in 2001 [37]. The recommendation has been enforced by re-phrasing the guideline and further specified in the subsequent ‘narrative’. Further screening is discouraged if the PSA values are leqslant R: less-than-or-eq, slant 1.0 ng/mL. The provision of proper information before screening is included. The American Medical Association concluded, in an extensive statement, that ‘mass screening is premature’ but gives space for ‘well-informed decision taking’[38]. The recommendations of the American Task Force for Preventive Disease and the American College of Physicians are similar [39,40]. The latter considered that routine PSA screening is ‘inappropriate’. Men requesting screening should be counselled and guided to make a well-informed decision. The committee of cancer screening of the European Union Programme ‘Europe against Cancer’ published their recommendations [41]. The introduction of screening as healthcare policy is considered premature. The recommendation is to complete the ongoing randomized studies and to provide conclusive information about potential risks and benefits of screening to men who wish to be tested.

Table 3.  Policies of prostate cancer screening
GroupPolicy statementRecommendations
AUA [35]Annual screening
 recommended from
 age 50 years
Take personal decision
 after consultation
ACS [39]Screening should be
 offered annually
 from age 50 years
Provide proper information
 of risk and potential benefit
AMA [38]Mass screening is
 premature
Allow ‘well-informed
 decision’
ACP [40]Routine PSA assayCounsel patient, informed
 ‘inappropriate’ decision
EU [41]Introduction as
 policy premature
Complete randomised
 studies, provide
 information on potential
 risk and benefit if
 screening is requested

The common denominator of all policy statements seems to be that if proper information is provided, testing cannot be refused. The crucial issue is then what constitutes ‘proper information’. As noted previously, there are ever more reports on this issue and some of the items are in the process of being validated. Appropriate informative videotapes should be established and field-tested; Chan and Sulmasy [24] provide further reading on this issue.

Conclusions

Early and aggressive treatment provides the only chance to cure prostate cancer. PSA-based screening advances the diagnosis by 6–10 years and leads to a significant stage reduction. Whether prostate cancer screening will reduce prostate cancer mortality currently remains unknown. The eventual judgement must include measurements of QoL and costs. Over-diagnosis is a significant problem; screening tests must become more selective for those men at risk of morbidity and death. Testing after an informed request cannot be refused and should then be carried out aggressively. The evidence that screening is effective, as produced by currently available data, is insufficient to introduce prostate cancer screening as a general healthcare policy. Ongoing randomized studies should be completed and may be expected to produce the information needed for healthcare policy making.

Acknowledgements

This paper contains data relating to the ERSPC Section Rotterdam. This project is a co-operative venture between the Department of Urology (C.H. Bangma, I.W. van der Cruijsen-Koeter, W.J. Kirkels, M.J. Roobol-Bouts, F.H. Schröder and M.F. Wildhagen), Department of Pathology (Th.H. van der Kwast, A.N. Vis), Department of Clinical Chemistry (B.G. Blijenberg), The Comprehensive Cancer Center (R. Damhuis) and the Department of Public Health (G. Draisma, H.J. de Koning, P.J. van der Maas, S. Otto).

Ancillary