Incomplete Followup and the Role of Chance

Given the long lead times (10 to 15 years) associated with PSA-based prostate cancer screening, the findings of the PLCO and ERSPC trials may change with additional followup. The observed mortality benefit in the ERSPC trial began to emerge around the same time that the median length of study followup was reached. As several centers have not yet provided data for the time period at which the effect starts to occur, it is possible that the observed mortality benefit could increase or disappear with additional followup. Although the point estimates for the effect on prostate cancer mortality are qualitatively different for the two trials, the confidence intervals overlap (0.75–1.70 for PLCO; 0.73–1.00 for ERSPC).^6,7 As such, both could potentially be consistent with either a small mortality benefit or no effect. The confidence intervals from the Djulbegovic and Cochrane meta-analyses are consistent with this conclusion (0.71–1.09 and 0.85–1.07, respectively).^14,15

Differences in the Proportion of Men With Prior PSA Testing

Neither trial excluded men who had a history of PSA testing (starting in 1995, PLCO excluded men with more than one PSA test in the previous 3 years). A relatively high rate of prescreening among participants, as documented in the PLCO trial, would have reduced the number of prevalent tumors that could be detected within the confines of the trial, thus lowering its power to detect a modest mortality benefit. As with contamination, however, this does not provide an explanation for the trend toward increased risk of prostate cancer mortality that was observed in men invited to screening in the PLCO trial compared with men assigned to usual care.

Effects of Screening in Controls and Noncompliance in the Intervention Arm

Opportunistic screening in control groups may result in underestimates of screening efficacy. In the ERSPC trial, contamination rates were extrapolated to be about 20% across all centers. Eighty-two percent of participants in the screening arm received at least one screening test. ERSPC study investigators published a separate, post-hoc statistical analysis to adjust for noncompliance and contamination, and concluded that the true effect of PSA-based screening on prostate cancer mortality was a 30% relative reduction in deaths.²³ In the PLCO trial, approximately half of the men in the control arm received a PSA test at least once during the course of the study, and 85% of participants in the screening arm were compliant with at least one PSA test.⁶ However, adjusting for compliance and contamination in this trial further increases the prostate cancer mortality rate ratio in screened men (from 1.13 to 1.47 or 1.72, depending on the level of contamination), making it less likely that a substantial mortality benefit was missed due to these factors. (See Appendix 5 for details about the model utilized.)

Differences in PSA Cut-Off Points, Screening Intervals, and Treatment Choices

A lower PSA cut-off point will lead to the detection of more cases of prostate cancer.^27,28 As such, generally lower PSA cut-off points utilized in the ERSPC trial (≥2.5 to 4.0 ng/mL)—compared with the PLCO cut-off point of ≥4.0 ng/mL—may have led to the detection of more cases of prostate cancer potentially amenable to curative interventions. On the other hand, more frequent screening also tends to increase the detection of prostate cancer cases, and while most ERSPC study centers screened men every 4 years, the PLCO trial performed annual screenings.

There were also different treatment preferences in the ERSPC and PLCO trials. The proportion of men in the ERSPC trial that initially chose active surveillance or expectant management instead of curative treatment was higher than in the PLCO trial (18.6% vs. 10%). More conservative management in the ERSPC trial may have reduced treatment-associated morbidity and mortality, which could be important within the context of overdiagnosis. Conversely, the shorter screening interval used in the PLCO trial, coupled with a high frequency of immediate intervention, may provide some explanation for the trend toward increased harm in the screen-invited population.

Limitations

Evidence from the two highest-quality discrete trials is inconsistent regarding the efficacy of PSA-based screening. Additionally, information for both trials is currently limited to interim results. We restricted the search on potential harms of PSA-based screening to information available from randomized efficacy trials; new information from other study designs may have become available since the previous evidence review.

Future Research

Longer-term followup of both the PLCO and ERSPC trials is important to assess the potential effects of PSA-based screening on prostate cancer mortality beyond 10 years. Given the risk of overdiagnosis and subsequent overtreatment observed within trials of prostate cancer screening, new methods that would allow for the distinction between indolent disease and disease that is likely to clinically progress are critically needed. Research that clarifies the long-term benefits and harms of immediate treatment versus initial conservative management in men with screen-detected prostate cancer would be of great importance. Results from the ongoing U.S. Prostate Cancer Intervention Versus Observation Trial²⁹ and the U.K. Prostate Testing for Cancer and Treatment Trial³⁰ should shed some light on the groups of men who might benefit most from treatment of PSA-detected prostate cancer.

Conclusions

Five randomized controlled trials (two fair- and three poor-quality) and two meta-analyses have evaluated the impact of PSA-based screening on prostate cancer mortality. After about 10 years, PSA-based screening is associated with the detection of additional cases of prostate cancer, but small to no reduction in prostate cancer-specific mortality.