Stacy Loeb, 550 1st Ave, VZ30 6th floor (#612), New York, NY 10016, USA. e-mail: email@example.com
Study Type – Therapy (cohort)
Level of Evidence 2b
What's known on the subject? and What does the study add?
Radical prostatectomy was previously shown to improve long-term outcomes among men with clinically-detected prostate cancer. Our data suggests that radical prostatectomy is also associated with improved outcomes in men with screen-detected prostate cancer.
• To examine the long-term outcomes of radical prostatectomy (RP) among men diagnosed with prostate cancer from the screening and control arms of the Rotterdam section of the European Randomized Study of Screening for Prostate Cancer (ERSPC).
PATIENTS AND METHODS
• Among 42 376 men randomised during the period of the first round of the trial (1993–1999), 1151 and 210 in the screening and control arms were diagnosed with prostate cancer, respectively.
• Of these men, 420 (36.5%) screen-detected and 54 (25.7%) controls underwent RP with long-term follow-up data (median follow-up 9.9 years).
• Progression-free (PFS), metastasis-free (MFS) and cancer-specific survival (CSS) rates were examined, and multivariable Cox proportional hazards models were used to determine whether screen-detected (vs control) was associated with RP outcomes after adjusting for standard predictors.
• RP cases from the screening and control arms had statistically similar clinical stage and biopsy Gleason score, although screen-detected cases had significantly lower prostate-specific antigen (PSA) levels at diagnosis.
• Men from the screening arm had a significantly higher PFS (P= 0.003), MFS (P < 0.001) and CSS (P= 0.048).
• In multivariable models adjusting for age, PSA level, clinical stage, and biopsy Gleason score, the screening group had a significantly lower risk of biochemical recurrence (hazard ratio [HR] 0.43, 95% confidence interval [CI] 0.23–0.83, P= 0.011) and metastasis (HR 0.18, 95% CI 0.06–0.59, P= 0.005).
• Additionally adjusting for tumour volume and other RP pathology features, there was no longer a significant difference in biochemical recurrence between the screening and control arms.
• Limitations of the present study include lead-time bias and non-randomised treatment selection.
• After RP, screen-detected cases had significantly improved PFS, MFS and CSS compared with controls within the available follow-up time.
• The screening arm remained significantly associated with lower rates of biochemical recurrence and metastasis after adjusting for other preoperative variables.
• However, considering also RP pathology, the improved outcomes in the screening group appeared to be mediated by a significantly lower tumour volume.
the European Randomized Study of Screening for Prostate Cancer
the Prostate, Lung, Colorectal and Ovarian (cancer screening trial)
PSA screening is controversial despite evidence that it leads to a significant stage migration with reductions in metastatic disease and prostate cancer-specific mortality [1,2]. In the Prostate, Lung, Colorectal and Ovarian (PLCO) cancer screening trial, the screening group underwent annual PSA screening for 6 years and most of the controls received opportunistic screening . The lack of a mortality difference between the groups has generated questions about the comparative efficacy of organised and contemporary opportunistic screening in the USA.
A related issue is whether organised screening influences treatment outcomes. In 2006, Roehl et al.  compared tumour features and radical prostatectomy (RP) outcomes between 464 men diagnosed with prostate cancer as part of a different organised USA screening programme vs 2713 cases who were not. The screening group had more favourable prognostic features at diagnosis, including lower PSA levels and less high-grade disease. Additionally, they reported a significantly higher 7-year progression-free survival (PFS) rate after RP in the screening vs the referred population (83 vs 77%, P < 0.001). However, as with the PLCO, there were high rates of screening in the ‘referred’ group, of which 57% had clinical stage T1c disease.
The European Randomized Study of Screening for Prostate Cancer (ERSPC) was initiated in 1993 to study whether PSA testing reduces prostate cancer mortality. At the beginning of the trial, opportunistic PSA screening was less common in Europe than in the USA . Van der Cruijsen-Koeter et al.  previously reported that men randomised to the screening arm of the Rotterdam ERSPC from 1993 to 1999 had significantly more favourable prognostic features at prostate cancer diagnosis than men from the control arm. This included a significantly lower PSA level and Gleason grade at diagnosis, as well as lower rates of locally advanced disease (T3/T4), and distant metastases. However, the outcomes of treatment were not reported in that study. The objective of the present study was therefore to compare pathological tumour features and long-term RP outcomes between men from the screening and control arms diagnosed during the same time interval. Due to concern about lead-time bias affecting the overall results, we also examined outcomes in relation to numerous clinical and pathological prognostic factors.
PATIENTS AND METHODS
The Rotterdam section of the ERSPC was initiated in 1993, as previously described [1,7]. Lateralised sextant biopsy was recommended for abnormal DRE/TRUS or a PSA level of ≥4 ng/mL (until May 1997), and thereafter for a PSA level of ≥3 ng/mL . During the period from 1993 to October 1996, men with a negative biopsy were re-screened 1 year later. The study protocol received approval from the local Ethics Committee and the Minister of Health of the Netherlands. The ERSPC trial is registered under the ISRCTN number 49127736.
From 1993 to 1999, 42 376 men were randomised. In the Netherlands, randomisation was performed after informed consent. During this period, prostate cancer was detected in 1151 men in the screening arm (either at initial screening or repeat screening 1 year later). During the same period, 210 men in the control arm were diagnosed with prostate cancer (by routine regional healthcare providers). Cancer incidence in the control arm was assured by linkage with the regional cancer registry. Of the men with prostate cancer, 420 (36.5%) screen-detected and 54 (25.7%) controls underwent RP (1993–2000) with long-term follow-up data. The study population therefore consisted of all men (n= 474) diagnosed with prostate cancer from 1993 to 1999 from the screening and control arms who subsequently underwent RP; whereas the remaining 887 men diagnosed with prostate cancer who did not undergo RP were excluded. Treatment decisions were based upon patient and physician preference (not part of the study protocol). There was no minimum follow-up period required for inclusion in the study. The median follow-up was 9.9 years (9.3 years in screening arm and 10.0 years in control arm) and follow-up was complete up to December 31, 2008.
Demographics and tumour features were prospectively recorded. RP specimens were reviewed by pathologists at the treating hospital. Organ-confined disease was defined as pathological stage T2 with negative lymph nodes. The follow-up protocol after RP consisted of PSA measurements every 3 months for 1 year, every 6 months for the second year, and then annually. The criterion for biochemical recurrence was a postoperative PSA level of >0.2 ng/mL. Chart review was performed every 6 months for all men with prostate cancer by a team of data managers to assess possible progression of the disease. Causes of death in cancers were evaluated by an independent causes of death committee according to an algorithm that is used in all ERSPC centres .
Comparisons between RP cases from the screening and control arms were made using the t-test, Wilcoxon rank-sum, chi-square, and Fisher's exact tests. The Kaplan–Meier method was used to examine the three main endpoints in the study: PFS, metastasis-free survival (MFS) and cancer-specific survival (CSS). Comparisons between the screening and control arms were made using the log-rank test.
In addition, multivariable Cox proportional hazards models were used to determine whether screen-detected (vs control) prostate cancer was associated with each of the three endpoints after adjusting for standard predictors. The ‘preoperative’ model included screening arm (vs control), PSA level, clinical stage, biopsy Gleason score and age. Due to the small sample size with clinical stage T3, we categorised clinical stage as impalpable (T1) vs palpable (≥T2) for this analysis. ‘Postoperative’ models were applied incorporating screening arm (vs control), pathological stage and RP Gleason grade. Separate models additionally adjusting for tumour volume were also applied in the subset (n= 369) with tumour volume data.
Table 1 compares the clinical and pathological features between cases diagnosed in the screening and control arms. Patients from the control arm were slightly older and had a significantly higher median PSA level at diagnosis. Although clinical stage and biopsy Gleason scores were not significantly different between the groups, a significantly greater proportion of controls were classified in the D'Amico intermediate- and high-risk categories . At RP, men from the control arm were significantly more likely to have positive surgical margins and seminal vesicle invasion. Additionally, tumour volume was significantly higher in RP specimens in men from the control arm.
Table 1. Comparison of clinical and pathological tumour features between men from the screening and control arms of the Rotterdam ERSPC treated by RP
Number of patients
Mean age at diagnosis, years
Median (mean, range) PSA level at diagnosis, ng/mL
With a median follow-up of 9.9 years, biochemical recurrence occurred in 69 (14.6%) and 15 (3.2%) developed metastatic disease. In all, 105 men (22.2%) died during follow-up, including 12 (2.5%) from prostate cancer. Kaplan–Meier survival curves are shown in Figure 1, stratified by screening and control arms. After RP, men from the screening arm had a significantly higher 10-year PFS (88 vs 72%, P= 0.003), MFS (98 vs 86%, P < 0.001), and CSS (98 vs 88%, P= 0.048) than men from the control arm. Overall survival was also significantly better in the screening vs control group (P= 0.04 log-rank); however, this appears to be accounted for by the higher rates of prostate cancer deaths in the control group. Excluding the prostate cancer deaths there was no difference (P= 0.15) between the screening and control groups in overall mortality (Figure 1d).
Table 2 shows the Cox proportional hazards models for each endpoint including preoperative features. After adjusting for PSA level, clinical stage, biopsy Gleason score, and age, men from the screening arm had significantly better PFS (hazard ratio [HR] 0.43, 95% CI 0.23–0.83, P= 0.011) and MFS (HR 0.18, 95% CI 0.06–0.59, P= 0.005). The difference in prostate CSS among screened men did not reach statistical significance in the multivariable model (HR 0.59, 95% CI 0.11–3.06, P= 0.531), although this analysis was limited by the few events.
Table 2. Multivariable models to predict time to biochemical recurrence, metastasis, and cancer-specific mortality based upon preoperative features
HR (95% CI); P
Prostate cancer mortality
Screening group (vs control)
0.43 (0.23–0.83); 0.011
0.18 (0.06–0.59); 0.005
0.59 (0.11–3.06); 0.531
PSA level (continuous)
1.04 (1.02–1.06); <0.001
1.00 (0.95–1.06); 0.953
1.01 (0.96–1.06); 0.662
Clinical stage (T1 vs ≥T2)
2.20 (1.34–3.61); 0.002
1.38 (0.45–4.22); 0.577
0.46 (0.10–2.16); 0.325
Biopsy Gleason score
1.96 (0.89–4.33); 0.097
1.99 (0.84–4.70); 0.116
1.04 (0.98–1.10); 0.196
1.03 (0.92–1.16); 0.607
1.09 (0.95–1.27); 0.227
In multivariable models adjusting for the stage and grade at RP (Table 3a), men from the screening group again had significantly better PFS (HR 0.43, 95% CI 0.23–0.81, P= 0.009) and MFS (HR 0.24, 95% CI 0.07–0.81, P= 0.021) than men from the control arm. Table 3b shows separate postoperative multivariable models in the subset with tumour volume measurements. After additional adjustment for tumour volume along with stage and grade, there was no longer a statistically significant association between the screening arm with PFS and MFS.
Table 3. Multivariable models to predict time to biochemical recurrence, metastasis, and cancer-specific mortality based upon postoperative features incorporating stage and grade at RP (a), and additionally considering tumour volume (b)
HR (95% CI); P
Prostate cancer mortality
Screening group (vs control)
0.43 (0.23–0.81); 0.009
0.24 (0.07–0.81); 0.021
0.44 (0.09–2.09); 0.302
0.50 (0.30–0.85); 0.01
0.48 (0.15–1.53); 0.213
0.43 (0.12–1.56); 0.199
3.1 (2.08–4.72); <0.001
5.41 (2.17–13.48); <0.001
4.30 (1.60–11.58); 0.004
Screening group (vs control)
0.65 (0.27–1.56); 0.338
0.71 (0.08–6.52); 0.763
0.47 (0.05–4.41); 0.509
0.74 (0.38–1.43); 0.365
0.44 (0.10–2.02); 0.293
0.42 (0.07–2.52); 0.342
2.56 (1.59–4.12); <0.001
4.98 (1.70–14.64); 0.003
5.33 (1.55–18.31); 0.008
Tumour volume (continuous)
1.31 (1.16–1.49); <0.001
1.12 (0.90–1.39); 0.297
1.15 (0.90–1.46); 0.262
In the present study, we showed that men from the screening arm of the Rotterdam ERSPC had significantly improved outcomes after RP, as compared with men from the control arm.
Previously, the Swedish Prostate Cancer Group reported a significant reduction in metastasis and cancer-specific mortality in men randomised to RP compared with watchful waiting . The survival advantage persisted in all pathological subgroups, including low-risk disease. Nevertheless, most patients in that study were diagnosed with prostate cancer clinically.
There is less data on the long-term outcomes of RP in screen-detected cases. The PIVOT trial performed at Veterans Affairs hospitals in the USA (∼50% T1c) suggested that the benefits of RP were confined to higher risk cases .
The present study only included men who had a RP from the screening and control arms of the Rotterdam ERSPC. Although this analysis therefore does not provide data on what the outcomes would have been without treatment, the results do suggest that screening was useful to diagnose higher-risk patients within the window of curability. By contrast, controls were diagnosed with higher-risk disease and larger tumour volumes, resulting in a reduced likelihood of surgical cure during follow-up.
Further stratification by Gleason score showed that there was a PFS advantage with RP for men in the screening arm compared with controls with Gleason 7–10 disease; whereas, Gleason 6 cases in both arms had better oncological outcomes with no statistically significant difference (data not shown).
One of the most intriguing findings in the present study was that tumour volume appears to be a critical determinant of treatment outcome. This was surprising considering previous findings from our group suggesting that tumour volume was associated with RP outcomes on univariate analysis, but was no longer significant after adjusting for other clinicopathological variables among men with screen-detected prostate cancer . However, the present study included a larger sample size from both the screening and control arms, including a wider range of tumour volumes and longer follow-up for survival endpoints.
Indeed, the present results suggest that one of the ways that screening improves survival outcomes is through a reduction in the volume and therefore burden of disease at diagnosis. It is noteworthy that many of the published criteria for ‘insignificant’ disease incorporate measurements of tumour volume, which are frequently used in management decisions [13,14]. The robust relationship between this variable with long-term outcomes suggests that its inclusion in risk stratification tools is appropriate. In the future, it is possible that advances in MRI or other markers may further aid in the assessment of tumour volume before definitive therapy . As in previous studies , Gleason score was also a robust predictor of RP outcomes.
Several limitations of the present study warrant discussion. First, RP specimens were examined at the treating hospital and central pathological review was not performed. In addition, of the men who were randomised during 1993 to 1999, fewer men in the control arm underwent RP. However, this might be explained, at least in part, by the significantly worse stage distribution in the control arm. For example, in the overall ERSPC trial from which this population was drawn, at 9 years the screening arm had a 41% relative reduction in metastases at diagnosis . As RP is only indicated for clinically localised disease, fewer men are candidates for this type of curative therapy in the absence of screening. Nevertheless, treatment was not randomly assigned and some of the differences between groups may have already been dampened in the process of surgical selection. For example, it is also possible that some patients with higher risk localised disease underwent radiation therapy rather than RP. Although the final sample size was limited with a relatively small number of events, it was sufficient to obtain statistically significant results for survival endpoints.
Lead-time bias can be considered to have affected the results, as screening at 4-year intervals has been shown to advance the diagnosis of prostate cancer by ≈11 years for the control arm . Methodology to adjust for lead-time bias in the present situation is not available, and it is unclear to what extent this bias can be decreased or even eliminated by reporting outcomes in relation to clinical and pathological prognostic factors. Accordingly, the overall Kaplan–Meier plots must be considered as less reliable. Because men in the present study were followed for a median of 9.9 years after RP, additional follow-up will be essential for the evaluation of long-term survival endpoints. Finally, we do not have data on the type of RP or surgeon case load, which has been shown to influence outcomes after RP [18,19].
In conclusion, among men treated by RP from the Rotterdam section of the ERSPC, screen-detected cases had significantly better PFS, MFS and CSS than controls within the available follow-up time (median 9.9 years). The reduction in biochemical progression and metastases in cases from the screening arm persisted after adjusting for other preoperative features (PSA level, clinical stage, Gleason score and age) or RP features (pathological stage and grade). However, subgroup analysis showed that the improved outcomes in the screening group appeared to be mediated by a significantly lower tumour volume. This suggests that a reduction in tumour burden at diagnosis is a mechanism through which PSA screening improves treatment outcomes.
S.L. was supported by the Louis Feil Charitable Lead Trust and the Society of Women in Urology Elisabeth Pickett Research Award. The ERSPC Rotterdam is supported by: The Dutch Cancer Society (KWF 94-869, 98-1657, 2002-277, 2006-3518, 2010-4800); The Netherlands Organisation for Health Research and Development (ZonMW-002822820, 22000106, 50-50110-98-311, 62300035).