SEARCH

SEARCH BY CITATION

Keywords:

  • early diagnosis;
  • prostate biopsy;
  • prostate cancer;
  • prostate-specific antigen;
  • screening

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Conflict of interest
  9. References

The objective of the present review was to evaluate the effect of population-based screening on the incidence of prostate cancer, prostate cancer tumor stage and grade, prostate cancer mortality, and overall mortality. A systematic review was carried out in April 2011, searching the Medline and Web of Science databases. The records were reviewed to identify comparative and randomized controlled trials evaluating the effect of screening on prostate cancer. Eight trials were identified containing personalized data on a screened versus a non-screened cohort. Prostate-specific antigen and digital rectal examination were the main screening tools. Prostate-specific antigen threshold and screening interval was not uniform among the different trials. Screening was associated with a significant increase in prostate cancer detection (relative risk 1.55; P = 0.002), and a significant shift towards more localized (relative risk 1.81; P = 0.01) and more low-grade tumors (relative risk 2.32; P = 0.001). In overall analysis, no significant effect on prostate cancer mortality (relative risk 0.88; P = 0.18) and overall mortality (relative risk 0.90; P = 0.27) in favor of screening was observed. An adjusted analysis excluding papers with short follow up, high prostate-specific antigen contamination in the non-screening group and low participation in the screening group was able to show a significant reduction in prostate cancer mortality of 24%. The ideal screening strategy is unclear. Screening is associated with better PC detection and this in a more localized stage and of less aggressive tumors. Excluding the main shortcomings in screening studies (short follow up, high prostate-specific antigen contamination in non-screening group and low participation in screening group), screening is able to reduce prostate cancer mortality.


Abbreviations & Acronyms
abno =

abnormal

AJCC =

American Joint Committee on Cancer

CI =

confidence interval

DRE =

digital rectal examination

ERSPC =

European Randomized Study of Screening for Prostate Cancer

NA =

not available

NNT =

number needed to treat

PC =

prostate cancer

PLCO =

Prostate, Lung, Colorectal and Ovarian cancer screening trial

PSA =

prostate-specific antigen

RR =

relative risk

TNM =

tumor–node–metastasis

TRUS =

transrectal ultrasound

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Conflict of interest
  9. References

PC is currently the most frequent non-cutaneous malignancy among men and is one of the major causes of cancer-related mortality.1,2

PSA testing with or without DRE is a proposed screening strategy in order to diagnose PC in an early localized stage for which a curative treatment is possible.3 Screening for PC has become a controversial issue, as a result of the lack of clear evidence supporting screening and the possible side-effects associated with screening (unnecessary biopsies, overdiagnosis and overtreatment of indolent tumors). In many Western countries, PC is mostly diagnosed as a consequence of opportunistic screening: in some parts of the USA, up to 75% of men at risk already underwent at least one PSA test ± DRE, whereas in Europe this is currently 20–40%.4 To date, no country has implemented a population-based screening program supported and funded by the government. To implement such a population-based screening, high quality of evidence is necessary, preferably derived from systematic reviews and meta-analyses (level of evidence 1a). Recently, the results of different randomized controlled and comparative screening trials have been published. This systematic review aims to evaluate to what extent screening affects PC incidence, stage and grade, PC mortality, and overall mortality.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Conflict of interest
  9. References

Search strategy and study selection

A systematic search was carried out in the Medline (PubMed) and Web of Science databases (the last systematic search was dated 4 April 2011). The following search strategy has been used: “mass screening” (MeSH terms) OR “mass screening” (all fields) AND “prostatic neoplasms” (MeSH terms) OR “prostatic neoplasms” (all fields) OR “prostate cancer” (all fields) AND “prostate-specific antigen” (MeSH terms) OR “prostate-specific antigen” (all fields) AND “mortality” (all fields) OR “mortality” (MeSH terms). This strategy was combined in the Medline database with the following limits: humans, gender (male), language (English) and article type (randomized controlled trial OR comparative study). No limits were applied for the Web of Science database. A search was also carried out for systematic reviews and meta-analyses to identify eventually missed, but eligible, trials on the topic. All randomized controlled and comparative trials dealing with screening versus no screening for PC were eligible for further review. Screening for PC was defined as testing on asymptomatic men with PSA ± DRE. Eligible randomized controlled and comparative trials were included in the present review if they provided personalized data on one or more of the following end-points: PC incidence, PC stage or grade at diagnosis, PC mortality or overall mortality. These data were extracted by two authors (NL and BDT) and in case of disagreement, arbitration by a third author (WO) was carried out.

Other data extracted from the trials were: population number, patient's age, duration of follow up, use of screening methods, interval period between screening rounds, participation in the screening arm and contamination in the control arm (no screening). For PC and overall mortality, an adjusted analysis was carried out by excluding papers if one of the following criteria, defined by the authors, was present: (i) follow up <8 years, (ii) participation in the screening group <75%. (iii) PSA-contamination in the non-screening group >33.3%.

An extensive quality assessment of the individual studies with evaluation of potential sources of bias was recently carried out by two other systematic reviews5,6 and is no further topic of this paper. The present review will only highlight the major shortcomings of each individual study.

Statistical analysis

Statistical analysis was based on the intention-to-screen principle in all studies. Cumulative analysis was carried out using Medcalc and RevMan 5.1 software. Statistical heterogeneity was tested using χ2-test and I2-test. A P-value <0.10 was used to show heterogeneity. In the absence of statistical heterogeneity, the fixed-effects model (Mantel-Haenszel method) was used; in the case of statistical heterogeneity, the random-effects model (DerSimonian and Laird) was used. The different outcome parameters between the screened and the non-screened populations were expressed as RR with 95% CI. A value of P < 0.05 was considered statistically significant.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Conflict of interest
  9. References

Literature search

The search yielded 91 and 98 citations in the Medline and Web of Science databases, respectively. After reading the abstracts, 30 articles were retrieved in full text for formal review. Lack of personalized data was the reason for exclusion in the Tyrol screening study7 and the Seattle-Connecticut cohort study.8 Subset analyses or earlier results of the same trials were excluded with the exception of the Rotterdam-Ireland,9 the Göteborg10 and the French ERSPC11 study. Although the screened group in the Rotterdam-Ireland study was a subset of the Rotterdam section of the ERSPC,12 the large control group from Ireland wasn't. The Göteborg study was initiated independently from the ERSPC, although subsequently, a subset of participants of that trial was also included into the ERSPC. Because the French section of the ERSPC was initiated later (2003), data were not included in the analysis of the ERSPC published by Schröder et al.12 Finally, 12 manuscripts representing eight studies were included into the present review:

  • 1
    Norrköping:13,14 data on metastatic PC were extracted from an earlier report,13 all other data were derived from the latest update.14
  • 2
    Québec:15 data needed to be recalculated into an intention-to-screen analysis; specific data on screening strategy provided in an earlier report.16
  • 3
    ERSPC:12 data on PSA-contamination derived from Roobol et al.17
  • 4
    PLCO:18 data on PC incidence and PC mortality were collected at 7-year follow up. Other data are at 10-year follow up.
  • 5
    Göteborg:10,19 data on PC incidence, PC and overall mortality,10 and data on tumor grade and metastases PC19 were the subject of two papers describing the same population.
  • 6
    Rotterdam-Ireland9
  • 7
    French ERSPC11
  • 8
    Stockholm20

Study characteristics

Eight studies enrolled a total of 571 594 participants; 210 772 and 360 822 of them were allocated in a screening and non-screening group, respectively. All studies used PSA as the primary screening tool, whereas some studies used DRE or TRUS as an additional screening tool. The normal upper limit of PSA differed among the different studies and with time. There were also substantial differences concerning screening-interval and follow-up duration. The screening interval ranged from once a year to one single screening visit. Study characteristics are summarized in Table 1. Availability of data on PC incidence, tumor stage, tumor grade, PC mortality and overall mortality is shown in Table 2. Tumor stage was defined as localized in the case of stage I or II (AJCC) or T1-2N0M0 (TNM). Metastatic disease is defined as stage 4 (AJCC) or any T, any N, M+ (TNM). Tumor grade was assessed using the Gleason score.

Table 1.  Study characteristics
StudyScreenedControlAge (range)Screening procedureIndication for prostate biopsyScreening-interval (years)Follow up (years)Compliance for screeningCompliance for prostate biopsyContamination in control group
Norrköping14947 53250–69DRE initial; PSA + DRE from 1993PSA > 4 ng/mL or abno.DRE32078%98%Low
Québec31 13315 35345–80PSA + DRE; TRUS if abno.PSA > 3 ng/mL or abno.DRE or hypoechogenic nodus on TRUSPersonalized1124%NA7.30%
ERSPC72 89089 35355–69PSA in all centers; DRE as ancillary test (some centers)PSA > 2.5–10 ng/mL (median 3 ng/mL) or abno. DRE2–7 (median 4)982.20%85.80%30.9%
PLCO38 34338 35055–74PSA + DREPSA > 4 ng/mL or abno. DRE1785%30–40%52%
Göteborg99529 95250–64PSAPSA > 3 ng/mL (until 1998); PSA > 2.5 ng/mL (from 1999)21476%93%Low
Rotterdam-Ireland11 970133 28755–74PSA + DRE + TRUSPSA > 4 ng/mL (until October 1997), PSA > 3 ng/mL (after October 1997), abno. DRE or abno. TRUS48.594.20%NA6%
French ERSPC42 59042 19150–74PSAPSA > 3 ng/mL4427.70%45.90%NA
Stockholm240024 80455–70PSA + DRE + TRUSPSA > 7 ng/mL and abno. TRUS or PSA > 10 ng/mL or abno.DRE or abno. TRUSOnce12.974%NANA
Table 2.  Data availability in different screening studies
StudyPC incidencePC mortalityOverall mortalityTumor stageTumor grade
Norrköping++++
Québec+
ERSPC++++
PLCO+++++
Göteborg+++Only metastatic disease+
Rotterdam-Ireland+++Only metastatic disease+
French ERSPC+++
Stockholm+++

The main shortcomings of the individual studies are the following:

  • 1
    Norrköping:14 During the first two screening rounds, only DRE was used as a screening tool. Fine needle aspiration was used during prostate biopsy, which has a lower accuracy compared with true cut biopsy. A low number of patients, especially in the screening group (n = 1494), leading to a low number of PC in this group (n = 85). In the screening group, 48 localized and thus curable tumors were diagnosed, of which only 21 were treated with curative intention.
  • 2
    Québec:15 Participation in the screening group was just 24%. The only data available was on PC mortality.
  • 3
    ERSPC:12 Substantial heterogeneity in screening interval, screening strategy and PSA threshold in the different centers of the study. PSA contamination in the non-screening group was estimated to be 30.9%.17
  • 4
    PLCO:18 Short median follow up of 7 years. High rate of PSA contamination (52%) in the non-screening group. In both groups, PSA testing was carried out in 44% of included men within the year before randomization. Although the participation rate (85%) in the screening group was good, the compliance rate to undergo prostate biopsy in the case of a positive screening test was just 30–40%
  • 5
    Göteborg:10 Relative low number of patients in both groups (n = 9952). Some of the patients were already included into the ERSPC trial.
  • 6
    Rotterdam-Ireland:9 Not a prospective randomized clinical trial, but a comparison between a screened population (part of the Rotterdam section of the ERSPC) and a population where screening is not routinely carried out (Ireland).
  • 7
    French ERSPC:11 Follow up of just 4 years. Participation rate in the screening group was just 27.7%. Low compliance (45.9%) for subsequent prostate biopsy in the case of a positive screening test in the screening group.
  • 8
    Stockholm:20 Screening included only one PSA-test with DRE and TRUS. High PSA threshold (10 ng/mL or 7 ng/mL and abnormal TRUS). Low number of patients in the screening group (n = 2400). Insufficient treatment of curable prostate cancer. Reconstruction of the no-screening group.

Outcome parameters

PC diagnosis

All trials, except the Québec trial,15 contained information about PC incidence. PC was detected in 12 447 out of 179 639 (6.9%) screened participants compared with 14 413 out of 345 469 (4.2%) non-screened participants. Screened patients had a significantly higher probability of being diagnosed with prostate cancer with a RR of 1.55 (CI 1.17–2.06; P = 0.002; Fig. 1). All but the Québec15 and Stockholm trials20 provided information about tumor stage and grade. Concerning tumor stage, the Rotterdam-Ireland9 and Göteborg10 trials only provided information on metastatic disease. Subgroup analysis showed that 8832 out of 155 317 (5.7%) screened participants versus 5850 out of 177 426 (3.3%) non-screened participants were diagnosed with localized PC, resulting in a RR of 1.81 in favor of screening (P = 0.01; Fig. 2a). Metastatic PC was diagnosed in 281 (0.2%) and 1360 (0.4%) out of 177 259 screened and 320 686 non-screened participants, respectively, resulting in a RR of 0.63 (P = 0.079; Fig. 2b).

image

Figure 1. Forest plot for PC incidence.

Download figure to PowerPoint

image

Figure 2. Forest plot for (a) localized PC and (b) metastatic PC.

Download figure to PowerPoint

Low-grade PC (Gleason ≤ 6) was found in 7682 out of 177 259 (4.3%) screened participants versus 5601 out of 320 686 (1.7%) non-screened participants. Thus, a screened man had a higher probability of having low-grade PC (RR = 2.32, P = 0.001; Fig. 3a). High-grade PC (Gleason ≥ 8) was detected in 820 out of 177 259 (0.5%) screened participants versus 1863 out of 320 686 (0.6%) non-screened participants, resulting in a RR of 0.91 (P = 0.42; Fig. 3b). In all these analyses, there was significant heterogeneity and the random effects model was used for evaluation.

image

Figure 3. Forest plot for (a) PC Gleason ≤6 and (b) PC Gleason ≥8.

Download figure to PowerPoint

PC-specific and overall mortality

Data on PC mortality were available in all trials, except the French ERSPC.11 The Québec trial,15 the ERSPC12 and the French ERSPC11 did not provide data on overall mortality. The Norrköping trial only reported on overall mortality in patients diagnosed with PC, but no data on overall mortality for the whole population were available.14 For the adjusted analysis, short follow-up (<8 years) was the reason for exclusion for the PLCO18 and French ERSPC.11 A low participation rate (<75%) was the reason for exclusion for the Québec trial,15 Stockholm trial20 and French ERSPC,11 and a high PSA-contamination rate in the non-screened group (>33.3%) was the reason for exclusion in the PLCO.18 This adjusted analysis thus only included data derived from the Norrköping,14 ERSPC12, Göteborg10 and Rotterdam-Ireland trials9 for PC mortality, and derived from the Göteborg10 and Rotterdam-Ireland trials9 for overall mortality.

PC mortality was 579 out of 168 182 (0.34%) participants for screening versus 1786 out of 318 631 (0.56%) participants for no screening, resulting in a RR of 0.88 (P = 0.18; Fig. 4a). In the adjusted analysis, PC mortality was 323 out of 96 306 (0.34%) screened versus 1161 out of 240 124 (0.48%) non-screened participants, resulting in a significant relative reduction of PC mortality of 24% (RR = 0.76, P = 0.04; Fig. 4b). The RR for overall mortality was 0.90 (P = 0.27; Fig. 5a) based on 8596 and 43 451 deaths by any cause in 62.665 screened and 206.393 non-screened participants, respectively. In the adjusted analysis, overall mortality was 3657 out of 21 922 (16.7%) screened versus 29 065 out of 143 239 (20.3%) non-screened participants (RR = 0.83, P = 0.32; Fig. 5b).

image

Figure 4. Forest plot for (a) PC mortality and (b) after adjustment with exclusion of studies with follow up <8 years, PSA-contamination in the non-screening group >33.3% and participation in the screening group <75%.

Download figure to PowerPoint

image

Figure 5. Forest plot for (a) overall mortality and (b) after adjustment with exclusion of studies with follow up <8 years, PSA-contamination in the non-screening group >33.3% and participation in the screening group <75%.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Conflict of interest
  9. References

To evaluate the true impact of mass-screening, several trials were started comparing a screened population with a non-screened population. One of the major drawbacks of evaluating the different screening trials is the large heterogeneity in screening protocol among these trials, making it difficult to pool data into a meta-analysis. Only the age group for inclusion is comparable. This criticism was also highlighted in another recent review.6 The current meta-analysis failed to detect a significant difference in PC and overall mortality in favor of mass screening and there is considerable overdiagnosis of the disease.

The question of whether PSA testing in an informed patient can decrease his chance of dying from PC is clearly different from the question whether mass screening, organized and paid by the government, can decrease the PC mortality in that region. This was the reason to carry out an “adjusted” analysis with data comparing those patients who received PSA testing with those who were not PSA tested. Therefore, we excluded studies with substantial PSA-contamination (more than 1 out of 3 patients undergoing a PSA test in the non-screening group). This was one of the reasons why the PLCO was excluded. This PSA-contamination could lead to earlier PC detection and could subsequently reduce PC mortality in the non-screening group. This adjusted analysis also excluded the PLCO18 and the French ERSPC11 because of a follow up of <8 years. The reason for this is the observation in the ERSPC and Göteborg trial (the only two randomized controlled trials that showed a significant reduction in PC mortality), the curves only started to diverge after >7 years follow up. Also, in a trial comparing radical prostatectomy versus watchful waiting, PC mortality only started to become significant after 8 years of follow up.21 This shows that 8 years is the minimum follow up when carrying out studies evaluating PC mortality and that screening is not likely to be beneficial for men with a life expectancy of <8 years. Although participation in the PLCO was good, compliance for a subsequent prostate biopsy in the case of a positive screening test was low, which could also have reduced the benefit of screening. Low compliance for prostate biopsy was also observed in the French ERSPC.11 A third reason for exclusion in the adjusted analysis was low participation in the screening group, which was the case for the Québec trial,15 Stockholm trial and French ERSPC.11 A low participation will dilute the effect of true screening, making it more difficult to attain significant differences.

The screening trials with a proven positive effect on PC mortality (ERSPC,12 Göteborg10 and Rotterdam-Ireland9 trials) all had a good compliance rate (>75%) for screening and excellent compliance rates (if available) for subsequent prostate biopsy. Participation and subsequent compliance for prostate biopsy in the case of a positive screening test is an important issue to take into account when starting up a screening program. One cannot obligate people to undergo screening or prostate biopsy, and thus non-participation and non-compliance will always be a factor reducing the benefit of mass screening. Motivating people with campaigns in the media could ameliorate participation rates. For the Québec15 and ERSPC17 trials, data could be adjusted for non-compliance (in the screening group) and PSA-contamination (in the control group). A significant relative risk reduction of 64% and 31% in favor of “true” screening was calculated for the Québec and ERSPC trials, respectively.

The ideal screening strategy (screening tools, PSA threshold, screening interval) is still unclear. In all screening trials, the main screening tool was PSA, with DRE as an additional screening tool in most studies. Both PSA and DRE have limitations as a screening tool, producing a substantial amount of false-positive and false-negative results. Even the combination of PSA and DRE has a relative low performance rate in prostate cancer detection (area under the curve of 0.68).22 The PSA threshold for prostate biopsy was 4 ng/mL on average among the different trials, although the Stockholm trial20 (7 ng/mL) and the Belgian section of the ERSPC12 (initially 10 ng/mL) used higher thresholds. Lowering the PSA threshold would lead to a better sensitivity with fewer tumors missed, with a possible effect on PC mortality. Nevertheless, there is no “safe” threshold under which no prostate cancer will be detected. Even if 0.5 ng/mL would be used as the threshold, some PC will be missed.23 This yields the need for better PC screening tools. Prostate Cancer Gene 3 (PCA3)24 and magnetic resonance imaging (supplemented with spectroscopy) of the prostate25–28 are promising and currently under investigation in the prebiopsy setting.

The screening interval might also influence the PC mortality and this interval differed substantially among the screening trials. One could argue that longer screening intervals will be associated with detection of more interval cancers with less favorable prognostic factors, as suggested in the Norrköping trial, in which the interval time was 3 years.13 In the PLCO,18 screening was carried out annually and no effect in PC mortality was observed. In the screening trials where a significant reduction in PC mortality was observed (ERSPC,12 Göteborg,10 Rotterdam-Ireland9) the screening interval ranged between 2 and 7 years, and was 4 years on average. In the ERSPC, the Swedish arm with a screening interval of 2 years was compared with the Dutch arm, where a 4-year screening interval was used. The 2-year screening interval indeed showed a better detection rate of PC compared with the 4-year screening interval, but was not able to lower the rate of interval or aggressive PC.29

The present meta-analysis clearly showed that screening was associated with a significant increase in PC detection. If PC was diagnosed, the chances of it being localized and less aggressive were clearly in favor of screening. This opens the possibility to offer more patients a treatment with curative intent when a tumor is detected by screening, and might reduce the need for adjuvant and expensive treatments, such as androgen suppression. The major concern associated with localized and low-grade PC is the observation that these tumors are rarely lethal, even without any treatment.30,31 Overdiagnosis of insignificant PC will lead to overtreatment with associated anxiety, complications and side-effects for the patients, and elevated costs for the community.32 This risk of overtreatment was also shown in the ERSPC12 and Rotterdam-Ireland9 trials, where a NNT of 48 and 37 was calculated, respectively. However, in the Göteborg trial,10 the NNT dropped to 12, and 30% of patients were offered active surveillance. Offering active surveillance to these localized and low-grade tumors is a beneficial strategy to deal with the problem of overtreatment.33

Although not significant, there was a clear trend towards the detection of fewer metastatic PC in favor of screening. Metastatic symptoms (e.g. bone pain or pathological fractures) have a negative impact on patients' quality of life. Standard treatment is androgen deprivation therapy and systemic treatment in case of castration-refractory PC. These treatments all have substantial side-effects and are very expensive for the community.

The present meta-analysis highlights the challenges for further screening trials: the need for a strict screening protocol with the definition of the best screening interval, better screening tools and identification of subpopulations that might benefit for screening.

The present meta-analysis had several limitations. The criteria in the adjusted analysis were defined by the authors, based on observations in the literature and clinical common sense, and therefore might be subject to serious selection bias. For all analyses, significant heterogeneity was observed and therefore, the random-effects model was used. However, one cannot be completely certain that this model fully fits to these data.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Conflict of interest
  9. References

PSA testing is currently the major screening tool for PC. Screening strategy differed substantially among screening trials. PC screening is associated with a significant increase of PC diagnosis and diagnosis in a more localized stage and of less aggressive tumors. An adjusted analysis excluding papers with short follow up, high PSA-contamination in the non-screening group and low participation in the screening group was able to show a significant reduction in PC mortality of 24%.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Conflict of interest
  9. References