Features and preliminary results of the Dutch centre of the ERSPC (Rotterdam, the Netherlands)

Authors


M.J. Roobol, Erasmus Medical Centre, Department of Urology, Room Ba 299, PO Box 2040, 3000 CA Rotterdam, the Netherlands. e-mail: m.roobol@erasmusmc.nl

Abstract

OBJECTIVE

To describe the preliminary results of the Dutch section of a large multicentre study of screening for prostate cancer, the European Randomized study of Screening for Prostate Cancer (ERSPC), initiated in the Netherlands and Belgium in 1991.

MATERIALS AND METHODS

After a series of five pilot studies which started in 1991, full-capacity screening started in 1994 with the use of a serum prostate-specific antigen (PSA) determination, a digital rectal examination (DRE) and transrectal ultrasonography (TRUS) as screening tests. Depending on the results and the screening protocol used, men were referred for further examination by sextant biopsies (extended with a seventh biopsy if TRUS showed abnormality). The protocols used, efficiency of the different screening tests, number of cancers detected in the pilot studies, initial screening round and preliminary results of the second screening round are described.

RESULTS

After the pilot studies it became clear that a study of prostate cancer screening was feasible in the Rotterdam area. The screening protocol was workable and the recruitment rate acceptable (39.5%). An inventory of the population registries of Rotterdam and surrounding municipalities, and the known recruitment rate, made it clear that a contribution of 40 000 men (aged 55–74 years) from the Dutch centre to the ERSPC was feasible. The initial screening round started in December 1993 and lasted until December 1999 (protocol 5–10). In all, 42 376 men were randomized and 1014 cancers detected (5.1%). During this screening the protocol was simplified. After evaluating the different screening tests abnormal results of the DRE and TRUS were omitted as an indication for a sextant biopsy. Only a serum PSA level of ≥ 3.0 ng/mL is now used as the indication. The second screening round started in December 1997 and continues. To December 2002, 9920 men were screened for the second time, 4 years after their initial screening visit. To date 446 cancers have been detected (4.5%); this round will last to December 2003. Further evaluation of the screening regimen and characteristics of the cancers detected are constantly assessed within the Dutch ERSPC. Meanwhile a third screening round has also been initiated, which will last to December 2007.

CONCLUSION

A prostate cancer screening study of the projected magnitude is feasible in Rotterdam; the recruitment rate is acceptable and the screening tests well tolerated. The study has generated many scientific publications and will be of great value in determining whether prostate cancer screening should be part of general healthcare.

INTRODUCTION

In the 1980s carcinoma of the prostate was (and still is) the second leading cause of cancer deaths among men in the European community [1]. From 1979 to 1988 the number of deaths from prostate cancer in the Netherlands had increased by 2% annually [2]. During the same period the identification of PSA had increased the number of methods available (adding to DRE and TRUS) for the early detection of prostate cancer [3,4]. With the increasing incidence and mortality on one hand, and the availability of several very promising detection techniques on the other, the question arose whether early detection or screening for prostate cancer would be feasible and effective in reducing prostate cancer mortality. That a cancer can be detected earlier in its natural history is no guarantee that benefits will follow. Numerous studies (case-finding, retrospective and prospective) were already undertaken to determine the efficacy of cancer screening. However, there are three important biases (lead-time, length- and selection) pertinent to many of these studies, causing serious problems with interpreting the study results. The most elegant way to account for these biases is a randomized controlled trial with cancer-specific mortality as the main endpoint [5].

In 1990 Schröder et al.[6,7] started to pursue the idea of a randomized study of screening for prostate cancer. From the start it was clear that it would be impossible to conduct such a costly trial in one European country; international cooperation was necessary. Based on institutional investments and a grant by the European Community programme ‘Europe against Cancer’ it was possible to initiate randomized pilot studies in Belgium (Antwerp, 1992–93) and the Netherlands (Rotterdam, 1991–94) [8]. At the same time the contours of the European Randomized study of Screening for Prostate Cancer (ERSPC) emerged [6,9].

METHODS

ERSPC Rotterdam started with a series of five pilot studies in October 1991 which served to obtain an impression of the logistics involved in setting up a screening study of the projected magnitude. Full-capacity screening started in June 1994. Apart from the main endpoint (prostate cancer mortality) the Dutch centre also assessed the efficiency of the different screening tests, pathological features of the cancers detected, treatments applied and quality-of-life related issues. With the use of the MISCAN computer program the screening process is being modelled.

The first part of this paper describes the results of the pilot studies and the initial screening round of the Dutch centre located in Rotterdam. The second part describes the preliminary results of the ongoing second screening round 4 years after the initial screening, a screening interval that was chosen on the basis of the ratio between prevalence in the first pilot screening and the incidence in the general population [8].

RESULTS

THE PILOT STUDIES (1991–93)

Men aged 55–74 years selected from the population registry of Rotterdam were invited for screening. The only exclusion criterion was a previous diagnosis of prostate cancer. Men who responded by returning the intake questionnaire and who provided a signed informed consent were randomized and notified of the outcome. The screening protocol consisted of three tests, i.e. serum PSA determination, a DRE and TRUS. The findings of these three tests resulted in a re-screening visit after 1 or 4 years, or a sextant biopsy, depending on the used protocol. Men with a benign biopsy result were re-invited after 1 year.

During the first four pilot studies the logistical and screening procedures were tested and optimized to a workable and acceptable protocol, which was finally tested in the fifth pilot study. The characteristics of all protocols used are shown in Table 1. The first protocol started in October 1991; in all 1186 men were randomized in a period of 15 months. The most important characteristic in this first pilot study was the randomization after PSA testing. No further screening tests were done if the serum PSA level was ≥10.0 ng/mL; these men were directly referred to their GP. The protocol changed in January 1993; men were randomized before PSA testing, meaning that only men randomized to the screening arm were PSA tested. Protocol 2 was used until March 1993 (256 men). In March 1993 the screening procedure after blood sampling was changed. If possible the biopsy was taken directly after the DRE and TRUS. Protocol 3 lasted until May 1993 (297 men). Protocols 4 and 5 had some minor changes in logistic procedures and the indication for sextant biopsy was simplified and became partly PSA driven (Table 1).

Table 1.  Characteristics of the screening protocols 1–10
Protocol numberPeriodRecruitment rate (%)Men (N)Biopsy indication used
 110/91–01/9335.6 1186DRE and/or TRUS abnormal with lesion ≥ 8 mm. PSA in all men.
 201/93–03/9336.5  256DRE and/or TRUS abnormal with lesion ≥ 8 mm or PSA ≥ 20.0 ng/mL.
 303/93–05/9342.4  297DRE and/or TRUS abnormal with lesion ≥ 8 mm or PSA ≥ 20.0 ng/mL.
 405/93–11/9342.4  679DRE and/or TRUS abnormal or PSA ≥ 4.0 ng/mL.
 512/93–05/9440.6  450DRE and/or TRUS abnormal or PSA ≥ 4.0 ng/mL.
 606/94–11/9543.4 8642DRE and/or TRUS abnormal or PSA ≥ 4.0 ng/mL.
 711/95–01/9653.9 4147DRE and/or TRUS abnormal or PSA ≥ 4.0 ng/mL. No screening if
03/96–10/96PSA < 1.0 ng/mL.
 801/96–03/9652.8 1404DRE and/or TRUS abnormal or PSA ≥ 4.0 ng/mL. No screening if
PSA < 1.0 ng/mL.
 910/96–04/9750.7 6000DRE and/or TRUS abnormal or PSA ≥ 4.0 ng/mL. No screening if
PSA < 1.0 ng/mL
1005/97–12/9948.021 733PSA ≥ 3.0 ng/mL. No screening if PSA < 3.0 ng/mL.
TotalProtocol 5–10 42 376 

The mean recruitment rate (Table 1) over the five pilot studies was 39.5%. Recruitment procedures proved to be relevant for establishing higher participation rates. Eliminating the need for the control group to visit the study centre introduced with protocol 2 led to an increase in the recruitment rates (Table 1). The screening tests were well accepted and tolerated [10].

As noted, PSA-driven biopsies were not taken in pilot study 1 and were rare in pilot study 2 and 3 (biopsy was indicated at a PSA of ≥ 20.0 ng/mL). This resulted in a detection rate of ≈ 9% in men with a PSA of 4.0–10.0 ng/mL. In pilot study 4 and 5 (biopsy indicated if PSA was ≥ 4.0 ng/mL, irrespective of DRE and TRUS results) the detection rate increased to 24%. This finding was one reason to decide on a PSA-driven screening protocol, i.e. a PSA threshold of ≥ 4.0 ng/mL in the future. Furthermore, the rate of false-positive findings (i.e. unnecessary sextant biopsies) was extremely high in protocols 5 and 6 (76 biopsies to detect two cancers) in men with PSA levels of < 2.0 ng/mL. Further analysis showed that the proportion of false-positive TRUS (24%) was about twice that of false-positive DRE (13%) in men with PSA levels of < 4.0 ng/mL. If both tests had been omitted in the group of men with a PSA of < 2.0 ng/mL, 69.3% of the study population, only two prostate cancers would have been missed. These findings showed that the role of the three available tests in prostate cancer screening was still to be determined for their predictive value and, obviously for the final study outcome, preventing death from prostate cancer [8].

Evaluating all procedures related to recruitment of participants, to applying the screening tests and to data collection during the pilot studies resulted in an infrastructure as shown in Fig. 1a. An inventory of the population registries of Rotterdam and surrounding municipalities and a known recruitment rate of ≈ 40% made it clear that a contribution of 40 000 men (aged 55–74 years) from the Dutch centre to ERSPC was feasible.

Figure 1.

The ERSPC, section Rotterdam. Infrastructure of the screening process.

THE INITIAL SCREENING ROUND (1993–1999)

In June 1994 protocol 6 started (Table 1) which was used until November 1995 and consisted of 8642 men. During this period again the value of PSA, DRE and TRUS was assessed to identify possible improvements in the accuracy of the screening procedures. Bangma et al.[11,12] studied the use of the ratio between free and total PSA (f/tPSA), age-specific reference ranges and PSA density, and the possibility of improving the specificity of tPSA, DRE and TRUS in prostate cancer screening. These studies resulted in positive findings for all three items. Using these detection techniques within the screening protocol could reduce the number of biopsies by ≈ 35%, with a reduction in cancer detection of 11%. The most cost-effective screening protocol however was pre-screening with tPSA and exclusion of DRE and TRUS at PSA values of ≤ 2.0 ng/mL.

Rietbergen et al.[13] also found that tPSA was the most powerful tool for predicting biopsy outcome in a group of 3963 men (981 sextant biopsies taken and 172 cancers found). In that study it became clear that most cancers were found within the PSA range of ≥ 4.0 ng/mL and that within this PSA range 3.6 biopsies had to be taken to detect one case of prostate cancer. At PSA levels of < 4.0 ng/mL, 14.2 biopsies were necessary to find one cancer. In the PSA range 0.0–1.0 ng/mL, 43 sextant biopsies were taken to detect one case of prostate cancer. Furthermore, it became clear that the cancers that were found through a positive DRE or TRUS findings alone (at PSA levels of < 4.0 ng/mL) only amounted to, respectively, 8.1% and 7% of the cancers detected. If PSA had not been used as a screening test, DRE would have detected 47.1% and TRUS 45.3% of all cancers. Calculations on the use of a PSA threshold value to avoid unnecessary prostate biopsies resulted in an optimal PSA threshold of 1.7 ng/mL. When using this value, 33.9% of the biopsies would have been avoided at the expense of 5.3% of the prostate cancers detected. As there were relatively few cancers detected at PSA levels of < 4.0 ng/mL, on which these calculations were based, it was decided to change the protocol to a pre-screen PSA threshold of 1.0 ng/mL, meaning that men with a PSA level of < 1.0 ng/mL were not screened further and directly scheduled for their next screening visit after 4 years.

Another reason for changing the protocol was related to the fact that men invited for ERSPC at that time lived in surrounding municipalities. To maintain a good recruitment rate it was necessary to reduce the number of visits to the University Hospital. Therefore it was decided that blood samples should be taken in the municipality and, if necessary, subsequent screening tests (DRE, TRUS and sextant biopsy) at the University Hospital in Rotterdam. Doing so, only ≈ 20% of the participants had to travel to the University Hospital. Protocols 7–9 were conducted accordingly, with some minor changes in the logistic procedures after PSA testing. Protocols 7–9 were used from November 1995 until April 1997 (a total of 11 551 men). This change in the screening procedure resulted in a considerable increase in the recruitment rate (Table 1).

During protocol 6–8 a study was conducted to evaluate the value of a 1-year re-screening after a benign biopsy result. Rietbergen et al.[14] found that biopsy at repeat screening diagnosed prostate cancer in 11% of the men biopsied (442 men biopsied, 49 cancers detected). Of the 984 men who were eligible for repeat screening after 1 year only 442 men were actually biopsied. A large proportion of the men (42.2%) had PSA levels of < 4.0 ng/mL and abnormalities found at initial screening could not be reproduced, or serum PSA was below the threshold of 1.0 ng/mL. Furthermore, the clinical characteristics of the tumours detected after repeat screening were more favourable, with an increased proportion of stage T1C tumours than those detected at initial screening.

In October 1996, having available more results of a re-screening after 1 year, it was decided to stop the re-screening procedure after 1 year. With the start of Protocol 9 this change was implemented (Table 1). Men with a benign biopsy result were re-invited after 4 years. Since deciding to take a sextant biopsy in participants with PSA values of ≥ 4.0 ng/mL (from pilot 4) the DRE was the mainstay of early diagnosis at lower PSA ranges. As noted [13] the DRE as a screening test at low PSA levels (i.e. ≤ 4.0 ng/mL) performed very poor (only 8.7% of all cancers detected were detected by DRE), so a further evaluation was indicated. Schröder et al.[15] studied the value of the DRE as a screening tool at low PSA ranges in a screening population consisting of 10 523 men. The data confirmed that the DRE has a low predictive value in men with low PSA levels. When PSA levels were < 3.0 ng/mL, 11 biopsies were necessary to detect one cancer.

Beemsterboer et al.[16] used a logistic regression model to predict the number of cancers for a PSA of ≤ 4.0 ng/mL if all men were biopsied (so-called ‘predictive index’) [17]. The effects of a change in PSA threshold on the outcomes of screening were explored in a group of 8600 men. Applying a DRE and TRUS only in the PSA range 1.5–3.9 ng/mL and 2.0–3.9 ng/mL to indicate that a biopsy was required would result in a decrease of biopsies by 29–36%, respectively, and a decrease in cancers detected of 5–8%, respectively. In addition, DRE and TRUS are difficult to reproduce because they are very investigator-dependent. A protocol with only a PSA of ≥ 3.0 ng/mL as a direct biopsy indication resulted in a decrease in biopsies by 12% and a decrease of detected cancers by 7.6%, and above all a much simpler screening procedure.

A protocol change which permits the missing of otherwise detectable tumours could in theory result in missing those cancers with the largest potential to contribute to the reduction in disease-specific mortality. In a study by Hoedemaeker et al.[18] the group of tumours detected by DRE and/or TRUS below a PSA level of 4.0 ng/mL was examined. This group had low pathological stages with a considerable fraction (43%) meeting the criteria for a ‘minimal tumour’ (one < 0.5 mL, lacking Gleason pattern 4 or 5 and being confined to the prostate); 86% had a tumour volume of < 0.5 mL. Men with low PSA levels are therefore most likely to harbour clinically insignificant tumours. These findings were confirmed by Vis et al.[19], who also showed that the DRE as screening test at low PSA values (i.e. ≤3.0 ng/mL) was inefficient; 289 DREs were needed to find one case of clinically significant disease, and 96 DREs were needed to diagnose a prostate cancer of any size, grade or stage. These data indicated that a change in protocol where DRE and TRUS are omitted in the lower PSA ranges was likely not to result in significant loss in potential mortality reduction, providing that re-screening was used after an adequate interval. In addition, the sextant biopsy procedure is bothersome and not without danger for the participants. On the basis of the data noted, a final protocol change took place in May 1997 (protocol 10, Table 1). From this point all men with a PSA of < 3.0 ng/mL were not screened further and men with a PSA of ≥ 3.0 ng/mL were invited for further examination (DRE, TRUS and sextant biopsies, Fig. 1b).

During the period of screening according to protocol 10 (Table 1) there was a validation study of the effects of the change in protocol [20]. In this study the cancer detection rates and tumour characteristics of the cancers detected in the ‘old protocols’ (protocol 6–9), and the ‘new protocol (protocol 10)’ were compared. The cancer detection rates were similar in the two screening regimens, because there were many more prostate cancer cases per biopsy in the PSA range 3.0–3.9 ng/mL in protocol 10. The positive predictive value of the PSA range 3.0–3.9 ng/mL was 6.4% in the ‘old protocols’ and 18.0% in the ‘new protocol’.

Prostate cancers detected with the new screening regimen had a similar distribution of Gleason scores but a larger proportion of confined disease. The final conclusion of this validation study was that the overall characteristics of the cases detected at PSA threshold of 3.0 ng/mL differed very little from those detected with the regimen based on PSA, DRE and TRUS.

Recruitment and randomization lasted until December 1999. Figure 2 shows the final data (number randomized, actually screened, biopsied and cancers detected) of protocols 5–10. These data became part of the central dataset administered at the central database of ERSPC in Edinburgh.

Figure 2.

Results of the initial screening round ERSPC, Rotterdam section, from November 1993 (pilot 5) through December 1999 (protocol 10). Numbers are not definite.

THE SECOND SCREENING ROUND (1997–2002)

For the second screening round we address three important questions: what will be the compliance and the detection rate at the subsequent screening round, and what has happened during the chosen 4-year interval to interval cancers?

In December 1997 the second screening round started (according to protocol 10) for men initially screened in protocol 5–10. Until December 2002, 14 695 men were eligible for a second screening visit. Men who had reached the age of 75 years or died within the 4-year interval (2455 men, 16.7%) were not invited for re-screening. This resulted in 12 240 men who were actually invited; 9920 responded to the invitation and were invited for blood sampling. The response in the second screening round so far is thus high, at 81%.

The main reasons for no response were bad health (5.4%]) moving out of the region (5.8%) and unknown in 7.8% of the men invited. Figure 3 shows the preliminary results (number of men screened, number of men biopsied and cancers detected) in the second screening round.

Figure 3.

Preliminary results of the second screening round ERSPC, Rotterdam section, from November 1997 through December 2002. Numbers are not definite.

Through linkage with the Cancer Registry (Fig. 1) and active medical record follow-up, the Dutch centre is also able to assess the number and characteristics of the clinically diagnosed cancers found elsewhere in the population randomized. The number of these ‘interval cancers’ gives an indication on the effectiveness of the screening tests used and of the correctness of the chosen screening interval. At present the definitions of interval cancers are being established and data on interval cases obtained through linkage with the Cancer Registry are being evaluated [21]. Preliminary results show very few interval cases so far.

Through linkage, not only can the number of interval cancers be assessed but also the number of clinically detected cancers in the control arm during the screening period. Comparing the tumour characteristics of screen-detected and a clinically detected cancer is an important intermediate endpoint of the ERSPC. Preliminary results show a favourable prognostic shift for the screen-detected cancers [I.W. van der Cruijsen, in preparation].

Using a second screening round also enables the validation of more predictors of biopsy outcome, e.g. PSA velocity, defined as the difference between the PSA value at initial screening and the PSA value at subsequent screening divided by the number of interval years, and new serum markers such as pro-PSA and hK2. Data from the initial screening round already showed that a considerable proportion of the cancers that can be detected by screening are in men with low PSA levels [22], and that a part of these cancers have potentially aggressive characteristics, are organ-confined, and thus suitable for treatment. As noted, DRE and TRUS were not very efficient as screening tools in these low PSA ranges. To investigate the possibilities for effective screening at low PSA ranges, two side-studies were initiated during the second screening round.

The first side study (Fig. 3, side study 1) ordered after a review of our protocol by one of the review committees, was set up to evaluate the value of PSA velocity as a predictor of biopsy outcome at low PSA levels. During the period December 1997 to April 2001, men with a PSA level of 1.0–3.0 ng/mL and a doubling of their PSA level within the 4-year interval were also biopsied. In this side-study, 214 men were biopsied and 34 cancers detected; the final evaluation of this study is in progress [23].

The second side study (Fig. 3, side study 2) focused on the value of fPSA and hK2 as predictors of biopsy outcome at low PSA levels (2.0–4.0 ng/mL). However, a more important goal of this study is not only to establish their value as predictor for biopsy outcome but also their value as predictors of the aggressiveness of screen-detected tumours. This side-study lasted from April 2001 until October 2002. In the PSA range 2.0–4.0 ng/mL, 734 men were biopsied and serum fPSA and hK2 determined. Together with the data from the radical prostatectomy specimen of the cancers detected, the value of fPSA and hK2 as screening tests was determined [24].

Another point of interest at a subsequent screening round are those men with a negative biopsy result at initial screening and a biopsy indication, based on an elevated PSA level, at the second screening round. As the PSA level is also strongly related to BPH, it is possible that men with persistently high PSA levels will be biopsied at every screening round. The cancer detection rate, positive predictive value and tumour characteristics of cancers detected in men with elevated PSA levels biopsied at the second screening round are currently being assessed, with a further evaluation of predictors for biopsy outcome in this particular group of men [M.J. Roobol, in preparation].

The second screening round will last to December 2003. Further evaluation of the screening regimen and characteristics of the cancers detected are the subjects of constant attention within the Dutch centre of ERSPC. In the meantime a third screening round was also initiated which will last to December 2007.

CONCLUSION

Setting up a prostate cancer screening study of the projected magnitude was feasible in Rotterdam. International cooperation was established and a central database containing all data of ERSPC set up. The recruitment rate is acceptable and the screening tests used well tolerated. Data of this study have so far led to at least 93 scientific publications from the Dutch centre and ≈ 175 scientific publications from all ERSPC centres together. This study will certainly be of great value in determining whether prostate cancer screening should be part of general healthcare.

Abbreviations
ERSPC

European Randomized study of Screening for Prostate Cancer.

Ancillary