European Urology

1. Introduction

Screening for prostate cancer (PCa) remains controversial, even after the European Randomized Study of Screening for Prostate Cancer (ERSPC) [1] and [2] reported a 21% relative reduction of deaths from PCa at 11 yr of follow-up and the Gothenburg (Sweden) trial, one of the ERSPC centers, showed 44% reduction at 14 yr [3]. The main reason for controversy concerns the harms associated with PCa screening, such as the high risk of overdiagnosis and overtreatment [4].

The occurrence of metastatic disease is a very important contributor to the suffering caused by PCa. Its prevention or delay is a legitimate secondary end point of screening. Earlier studies have reported that early diagnosis by prostate-specific antigen (PSA) testing results in a reduction of the number of men with metastatic disease [5], [6], and [7]. The ERSPC study has shown a 41% decrease of metastatic disease in the screening arm at the time of diagnosis [1]; however, progression to metastatic disease during follow-up after initial treatment that was not evaluated in the ERSPC study [1] should also be influenced by screening.

The goal of this paper is to compare the incidence of metastatic disease, not only at diagnosis but also during subsequent follow-up, in the screening and control arms of four major centers of the ERSPC study group.

2. Materials and methods

The ERSPC is a randomized trial comparing an intervention arm of men to whom regular PSA screening is offered with a control arm to which such screening is not offered. Details of the methodology have been described previously for all centers in the ERSPC study [1] and for the Gothenburg trial by Hugosson et al. [3]. In all centers, known cases of PCa were excluded before randomization. The Swedish center utilized a 2-yr screening interval, whereas a 4-yr interval was applied at the other centers. In Finland, randomization was carried out at a ratio of 1:1.5 between the screening group and the control group. In addition, men with PSA 3.0–3.9 ng/ml were not referred directly to further diagnostic examination but were offered an ancillary test, a digital rectal examination (in 1996–1998) or a free/total PSA ratio (with cut-off of 0.16), from 1999 onward.

As in the previous reports, the focus of the paper is on the core age group of 55–69 yr. Follow-up of cancer cases in the control arm was by 6-mo chart review. Diagnostic and treatment decisions were in the hands of the regional care providers.

The current evaluation is limited to those centers for which information on metastatic status is available in both the screening and control arms of the study and was collected during the entire period of postdiagnosis follow-up. No common follow-up scheme was established in the ERSPC protocol, but patients with progressive disease were typically checked every 3–6 mo for PSA and clinical status. Bone scans were generally performed in men with bone pain or suspicion of M1 as well as in men with PSA values of ≥20 ng/ml and/or high Gleason scores ≥7. Similar procedures were followed for patients in both arms within all centers.

The M+ status was assigned in the presence of one of two findings. M1 disease status was defined in line with the TNM definition by an obviously positive bone scan or x-ray. In doubtful situations, computed tomography or magnetic resonance imaging studies were used to confirm or exclude metastatic disease. Patients with a PSA value >100 ng/ml were considered to have metastatic disease if imaging studies were not reported. An analysis of the rate ratios (RRs) for M1 disease at imaging and for M+ disease overall, at diagnosis, and during follow-up was carried out to investigate the correctness of this assumption. We categorized men who were diagnosed with M+ disease in spite of screening as missed at screen (biopsied according to protocol, no PCa detected), nonattendees, above the upper age cut-off (age 69–74 yr), and rapid progressors (no biopsy indication at last screen).

For purpose of risk stratification, low, intermediate, and advanced clinical risk groups at the time of detection were defined, respectively, in line with definitions used by Schröder et al. [1] and [2]: (1) clinical stage T1, T2 and Gleason score ≤6; (2) clinical stage T1, T2 with Gleason score 7, or T3 with Gleason score ≤7; and (3) clinical stage T1, T2, or T3 with Gleason score 8–10 or clinical stage T4 with any Gleason score. To allow better comparison with current guidelines, we also analyzed the effect of the inclusion of PSA in line with the European Association of Urology guidelines [8].

3. Results

The study population is shown in Figure 1, which also shows the part of the total ERSPC population that is included in the present report. In addition, absolute numbers of participants, cancers and M+ cases diagnosed, and deaths per arm are shown. The numbers of participants in the screening and control arms, the numbers of men with M+ disease per center and per study arm, and the relative risks of M+ at diagnosis and during follow-up are specified in Table 1. The follow-up periods differed slightly between centers with an overall median of 12.0 yr (Netherlands: 12.0 yr; Sweden: 14.9 yr; Finland: 12.9 yr; Switzerland: 9.1 yr). The RRs of screening versus control of M+ disease diagnosed solely by imaging (proven M1) were compared with the M+ data based on both imaging and PSA >100. For M+ disease, overall RRs of 0.66 and 0.70 resulted for screening versus control, respectively; for M+ at diagnosis, RRs were 0.54 and 0.50; and for M+ during follow-up, RRs of 1.00 and 1.16 were found. The closeness of these RRs justifies the use of PSA >100 as a surrogate for M1 disease.

Fig. 1 Flow diagram of the study.ERSPC = European Randomized Study of Screening for Prostate Cancer; PCa = prostate cancer.*Comprising data of four of the eight ERSPC centers: The Netherlands, Sweden, Finland (Tampere), and Switzerland.^$ Percentage of men with M+ PCa that died.

The Nelson-Aalen plot showed divergence of the cumulative hazard for M+ disease overall (ie, at time of diagnosis and during follow-up combined) starting at years 4 and 5, with a lower risk in the screening arm (Fig. 2a). At 12 yr of follow-up, the RR of M+ disease overall in the screening group compared with the control group was 0.70 (95% confidence interval [CI], 0.60–0.82; p = 0.001). The absolute risk reduction was 3.1 (95% CI, 1.8–4.4) metastatic PCa cases per 1000 in men randomized (0.31%). The excess incidence of PCa (all cases) in the screening arm amounted to 37.6 cases per 1000 men. This number translates into a 55.6% higher incidence rate in the screening arm compared with the control arm, indicating the potential amount of overdiagnosis. The resulting NNI for screening to avoid 1 case of metastatic disease was 328, and the NND was 12. If the analysis was restricted to attendees only, this resulted in a relative risk reduction of 42% compared with the control group (RR: 0.58; 95% CI, 0.45–0.74; p < 0.001). Figures 2b and 2c show the Nelson-Aalen curves for M+ PCa found at diagnosis and during follow-up. An analysis of the numbers of M+ cases detected at diagnosis per year after randomization is shown in Figure 3. The effect seen at diagnosis represents the main effect of screening in preventing M+ disease, but this effect decreases during the later years of follow-up due to increasing numbers of M+ starting with year 7 in the screening arm.

Fig. 2 Nelson-Aalen cumulative hazard estimates of M+ prostate cancer (a) overall, (b) at time of diagnosis, and (c) during follow-up.PCa = prostate cancer.

Fig. 3 Rate per 10 000 person-years of M+ at diagnosis per study arm and by years after randomization.

4. Discussion

Our results show an absolute risk reduction of 3 M+ PCa cases per 1000 men, corresponding to a relative risk reduction of 30% (p = 0.001) in favor of screening. This reduction in M+ disease must be balanced against the downside of screening, mainly the rate of overdiagnosis in the screening arm, estimated at ≥50% in the setting of the ERSPC [4]; in the current analysis, a potential rate of overdiagnosis of 55.6% was shown. This report is next to that of Kilpeläinen et al. [10], which is limited to one center and a shorter mean follow-up of only 9 yr; it is the first report analyzing the effect of screening on metastatic PCa in a prospective randomized setting. In addition, a few nonrandomized studies have addressed the issue [4], [5], and [6]. The data confirm absolute and/or relative reductions of metastatic disease within the three different populations studied. The absolute risk reduction of 3 metastatic cases per 1000 men seen in our study, which is equivalent to a relative risk reduction of 30%, is modest but clinically relevant, especially if one considers the relative risk reduction of 42% seen after correction for nonattendance in men actually screened.

To provide a better balance between benefits and harms of screening for PCa, approaches to decrease overdiagnosis and other negative quality-of-life effects remain an important task. The use of available risk-stratifying instruments to decrease “unnecessary” diagnostic measures is a realistic option [11]. The relative risk reduction of 30% is lower than the relative risk reduction of 41% reported in 2009 for the whole of ERSPC with a 9-yr follow-up [1]. The reason for this is that M+ at diagnosis does not reflect the complete picture with respect to the relative decrease of M+ disease by screening. This detailed analysis shows that an almost equal number of PCa cases in both arms progress to M+ disease during follow-up, reducing the effect on M+ at diagnosis from 50% to 30% overall.

Next to establishing the effect of screening on the risk of developing M+ PCa, the exploration of factors that may influence the reduction of M+ disease is of importance. We show an increase of M+ disease overall in the screening arm as of year 7. This is explained in large part by nonparticipation and ending screening at age 70. The findings given in Figure 3 confirm that most metastatic progression occurs after the sixth year of follow-up (Table 2) in cancers diagnosed at the first screen. The data show, however, that very few (only 9 of 59) M+ cases might have been avoided by more aggressive screening. This late increase of M+ cases at diagnosis in the screening arm has the potential to decrease the overall effect of screening.

The finding that the rate of PCa cases progressing to M+ disease during follow-up is equal in both arms is new and is relevant for the understanding of screening studies and of screening for PCa in general. A potential explanation for the late, unexpected high rate of M+ disease in screen-detected cases is lead-time bias and the observation that, in the screening arm, many of the cases diagnosed at the first screen are already far advanced at the time of diagnosis. Unfortunately, our data cannot fully explain the lack of an effect of screening on M+ disease occurring during follow-up. Follow-up may, however, still be too short, and our conclusion may change in the future if the trend seen in Figure 2c after year 12 continues. Differences in potentially curative treatments between nonprogressive and progressive cancers in the screening arm may contribute to the reported observations.

5. Conclusions

Our study shows relative reductions of 30% and 42% in M+ disease by screening for the whole population and for men who are screened, respectively, after 12 yr of follow-up. This benefit must be balanced against potential harms of screening.

Author contributions: Fritz H. Schröder had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Schröder, Hugosson, Carlsson, Tammela, Määttänen, Auvinen, Kwiatkowski, Recker, Roobol.

Acquisition of data: Schröder, Hugosson, Carlsson, Tammela, Määttänen, Auvinen, Kwiatkowski, Recker, Roobol.

Analysis and interpretation of data: Roobol, Schröder.

Drafting of the manuscript: Schröder.

Critical revision of the manuscript for important intellectual content: Schröder, Hugosson, Carlsson, Tammela, Määttänen, Auvinen, Kwiatkowski, Recker, Roobol.

Statistical analysis: Roobol.

Obtaining funding: Schröder, Hugosson, Carlsson, Tammela, Määttänen, Auvinen, Kwiatkowski, Recker, Roobol.

Administrative, technical, or material support: Schröder, Hugosson, Carlsson, Tammela, Määttänen, Auvinen, Kwiatkowski, Recker, Roobol.

Supervision: None.

Other (specify): None.

Financial disclosures: Fritz H. Schröder certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.

Funding/Support and role of the sponsor: None.