There are calls for a national screening programme for prostate cancer: what is the evidence to justify such a national screening programme?

Screening methods

Prostate screening will normally involve serum PSA checking and DRE.

A normal serum PSA level is defined as being 4.0 ng/mL or lower by most assay manufacturers, using this value the sensitivity of PSA screening ranges from 47% to 72% and specificity from 44% to 91%. ⁵

Results from a screening trial which evaluated screening by DRE alone showed a sensitivity of 58%, specificity of 96% and a positive predictive value of only 20%. ¹⁰

Studies justifying prostate cancer screening

Two RCTs were identified which reported positive results from prostate cancer screening trials. The first, carried out between 1988 and 1996, was reported in 1999 ¹¹ and included all (n = 46,193) men aged 45–80 on the electoral rolls in the Canadian city of Quebec, who were stratified by age and residential area. This may have introduced selection bias. The study excluded any men with a previous diagnosis or those referred with a suspicion of prostate cancer, and no public announcement about the study was made. Men were assigned to the screened or unscreened groups on a 2:1 ratio to try and compensate for the predicted lower uptake of screening, the actual response to invitation was even worse than predicted at only 15.7% of those invited. For most men, screening was defined as a yearly PSA measurement and DRE. If the PSA level was greater than 3.0 ng/mL initially or on follow-up or had increased by more than 20% since the last result, a TRUS-guided prostate was performed. An abnormal DRE was also an indication for TRUS biopsy, with the only exceptions being the first 1002 men who received a TRUS regardless of the other results. The report writers argued that by choosing the PSA cut-off level to be 3.0 ng/mL rather than 4.0 (the recommended level), up to an extra 18% of the smaller cancers were identified. ¹¹ The primary end point was death from prostate cancer. As data were collected from the Death Registry the rate of diagnosis in the unscreened group was not recorded, which would have been worth comparing to that of the screened group to evaluate over diagnosis. Blinding of the assessors does not appear to have taken place, which increases the risk of bias, it was obviously not possible to blind the participants which is the case with all such screening trials. A major flaw standing out in this study is the high rate of crossover and contamination between groups and a lack of intention-to-screen analysis in the report, which is essential for screening trials. Of those invited to screening only 23.6% actually were screened. In total 7.3% of the control group were known to have been screened but this will be an underestimate. The main finding of the trial was a 67.1% decrease in the death rate of the screened group compared to the unscreened group (13.7 deaths per 100,000 man years compared to 41.6, p = 0.02). On the first screen, prostate cancer prevalence was 3.00% and then 0.52% on average for the remaining 8 years of follow-up; furthermore, during the follow-up period not one of the cancers detected had reached a metastatic stage as a result of the regular screening.

The second RCT which comes out in support of screening is the largest (n = 182,000) to date with men from seven different countries involved. Published in 2009 the European Randomised study of Screening for Prostate Cancer (ERSPC) ¹² ran from 1991 to the end of 2006. PSA measurement with a 3.0 ng/mL cutoff without DRE was used in most countries. Men were randomised from population lists and assigned to either group without any stratification; the full age range was from 50 to 75 – although most participants were aged 55 to 69. The primary outcome was the rate of death from prostate cancer, blinding of outcome assessors also took place.

Statistical analysis took place on the intention-to-screen principle. Compliance within the screening group was better than the QUEBEC trial at 82%, but had a higher contamination of the control group at 20%. The study fails to explain how contamination was evaluated so could potentially be substantially greater. The cumulative incidence of prostate cancer was 8.2% in the screening and 4.8% in the control group. As in the QUEBEC trial, cancers were detected at an earlier stage with only 27.8% in the screened group being detected at a Gleason score of 7 or more (on a scale of 2–10) compared to 45.2% in the control group. The ERSPC authors reported the number needed to screen to prevent one prostate cancer death was 1410 (95% CI 1142–1721) with an average of 1.7 screens within 9 years. It is possible that the lower frequency of follow-up encouraged more men to take part in the screening group compared to the poor numbers in QUEBEC, although it does seem to have impacted on the numbers picked up on screening compared to the QUEBEC trial. There were many (mainly minor) differences between screening protocols at each study centre leading to potential statistical bias due to the differing sizes of groups in each centre. For example, Sweden screened every 2 years and included men aged 50–69 diagnosed prostate cancer in 11.8% of their subjects whereas Italy included the older men aged 55–75 and screened every 4 years with a positive PSA cutoff for biopsy at 4.0 ng/ml diagnosed only 3.9% of their screening group.

Studies finding no evidence for prostate screening

Again two RCTs have been selected which have reported no positive effect of prostate cancer screening. The Prostate, Lung, Colorectal and Ovarian (PLCO) cancer screening trial ¹³ reported in March 2009 that the difference in death rate between the screened and unscreened arm of their trial did not significantly differ. Between 1993 and 2001, 76,693 men at 10 American centres were randomised 1:1, stratified according to centre and age, into a screening or usual care (control) group and have so far been followed up for a median of 11 years with annual serum PSA tests and DRE as well as screening for lung and colon cancer. Entry to this trial was stricter than others as nobody with a history of any PLCO cancers was allowed to take part and from 1995 onwards men were excluded if they had a PSA test done within the past 3 years. When a positive DRE or PSA (>4.0 ng/mL) result was recorded men were advised to seek further diagnostic evaluation from their own doctor who would perform a TRUS or other investigations as necessary and give subsequent treatment, this ensured that the care received by each patient in the screened and control arms was the same. Contamination was closely monitored using yearly surveys for 1% of the screening group, a new 1% was surveyed each year giving an estimated contamination of up to 52% by the sixth year although at all points the screening rate was higher in the screened group. This high contamination rate could be one of the reasons for the negative findings, for several reasons in America PSA screening is more widespread than Europe with one trial reporting 49% of general male population aged 50–79 having had a screen within the past 2 years. ¹⁴ Subjects were surveyed annually for this information and chased up if no reply was received and additionally the American National Death Index and death certificates were used, although 10% had been lost to follow-up after 10 years. Reviewers, blinded to group assignment, recorded and reviewed deaths using an end point adjudication process to minimise bias. At 10 years when 67% had completed their full follow-up (PSA for 6 years and DRE for 4) 9.00% (n = 3452) of the screened group had been diagnosed with cancer compared to 7.75% (n = 2974) from the control group (rate ratio, 1.17; 95% CI 1.11–1.22), a 22% increase in diagnosis. Possible reasons for this include the higher frequency of screening, contamination or selection bias due to baseline screening, which could account for the higher rate in the control group.

A further RCT, with no evidence of beneficial effects from screening, was also published in 2009. ¹⁵ Based on the men in the catchment of Stockholm South Hospital (n = 27,204), it is the longest running trial considered in this article – running for 15 years between 1988 and 2003. Screening was a one-off DRE, PSA and TRUS with a biopsy for abnormal DRE or TRUS results and a repeat TRUS for PSA results of more than 7 ng/mL. Although by far the smallest study considered it had a very thorough initial cancer detection screen. With a median follow-up of 12.9 years, 2.23% (n = 53) of the screened group and 2.04% (n = 506) in the control group died from prostate cancer. Interestingly, the screening invited non-attendee group incurred a higher incidence of prostate cancer compared to both the screened and uninvited control group, as well as a higher mortality rate overall. This observation may suggest that those who are at most need of screening are the ones who will not volunteer, and those with greater comorbidities – the overall survival rate of screening attendees was 89% higher than non-attendees. The results of this study not only add to the argument against prostate cancer screening for reasons of harm and inaccuracy as in the PLCO trial, but should also be remembered when considering other studies that do not report on the difference between those who do and do not attend after invited for screening.