Review of evidence concerning PSA screening for prostate cancer has limitations as basis for policy development
- Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington, USA
- Correspondence to: Ruth D Etzioni
Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N. M2-B230, PO Box 19024, Seattle, WA 98109-1024, USA;
Prostate cancer screening is controversial. This study presents an evidence review commissioned by the US Preventive Services Task Force (USPSTF) in developing their updated recommendations regarding prostate-specific antigen (PSA) screening. The USPSTF based its latest recommendation (against PSA-based screening for prostate cancer) on the evidence in this review, concluding that there was moderate to high certainty that the service had no net benefit or that the harms outweighed the benefit.
The evidence review used standard protocols to identify articles published between 2002 and 2011 that addressed benefits or harms of screening and primary treatment for localised cancers. Focus was on comparative studies, particularly clinical trials. With respect to screening, most inferences were based on the US Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial and European Randomized Study of Screening for Prostate Cancer (ERSPC) trials, which were considered to be of fair quality. With respect to treatment, the Scandinavian trial of radical prostatectomy versus watchful waiting was considered to be of good quality; in addition, several cohort studies were used to inform outcomes of surgery and radiation therapy.
The review concludes that PSA-based screening is associated with a very small benefit, if any, in terms of prostate cancer mortality over a period of 10 years. This is based on the difference in prostate cancer mortality observed at the end of the follow-up period in each trial. The review cites the non-significant result of slightly higher mortality in the intervention group in the PLCO trial after 10 years and both the relative (20% reduction in prostate cancer mortality) and absolute (0.7 lives saved per 1000 men screened) reductions in the ERSPC after median 9 years of follow-up. The conclusion appears to be based on the absolute rather than the relative benefit. Two recent meta-analyses of randomised screening trials which found no pooled effect of screening on prostate cancer deaths are also cited. The significant study from the Swedish section of the ERSPC study is reported, but considered to be atypical in the context of the broader ERSPC results. The review's evidence concerning harms focuses mostly on estimates of false-positive rates (12–13% after 3–4 screening rounds) and side effects of treatment, based on clinical trial results. Estimates of overdiagnosis (diagnosis of a disease that will not cause morbidity or mortality) from the ERSPC are reported, namely that approximately 50% of screen-detected cases are overdiagnosed and 48 additional men would need to be treated to save one life.
The review summarises the empirical evidence about screening benefits and harms reported by published clinical trials. However, it is fundamentally limited in the information it does not provide.
Limited-duration screening trials generally do not provide reliable estimates of the screening benefit that would be expected in the population over the long term. In screening trials the relative reduction in disease-specific deaths grows over time.1 This is borne out by updated results from the ERSPC, which show a mortality rate ratio of 0.62 in years 10–11 vs 0.85 in years 1–9. The time dependence of benefit estimates from screening trials is even more pronounced when measuring absolute differences in disease-specific deaths. Thus, the estimate of 0.7 lives saved per 1000 men screened will almost certainly grow markedly with further follow-up.
The review reports the findings of the ERSPC and PLCO trials as conflicting. However, the PLCO was subject to extreme contamination of the control (unscreened) arm—74% had at least one screen.2 Many had been screened before entry and the number of prostate cancer deaths was far lower than anticipated. A post-hoc calculation using three models projected that the power of the trial was on the order of 10–29%.3 Thus, the trial can neither be interpreted as a negative study nor as providing evidence that screening has no benefit. Moreover, combining trial results in a meta-analysis suggests that they are comparable, which is clearly not the case.
Short-term follow-up also distorts estimates of screening harms. The number needed to treat (more accurately, number needed to detect since not all cases were treated) of 48 has already gone down to 37 with two more years of follow-up and it is projected to be far lower over the long term.4 The estimate of 50% overdiagnosis among screen-detected cases comes from the Rotterdam section of the ERSPC. Estimates based on US population data are closer to 25%.5
In summary, screening trials are notoriously difficult to conduct and interpret. Simply reviewing the empirical data from recently published trials does not provide the information that is really needed for policy development—namely the expected long-term benefits, harms and harm–benefit trade-offs of screening in the population.