Article Text
Statistics from Altmetric.com
Forty years after its discovery, a reanalysis of the two largest trials to date, controversially suggests that prostate-specific antigen (PSA) screening may actually be beneficial.
Most healthcare organisations do not recommend PSA screening for prostate cancer,1 2 mainly in response to conflicting evidence about the benefits and clear evidence of harms. PSA can lead to false positive or ‘overdiagnosed’ cancer (detecting prostate cells that histologically represent cancer, but will never grow to cause a patient harm).
Evidence regarding efficacy has been based on two large randomised controlled trials: The European Randomised Study of Screening for Prostate Cancer (ERSPC)3 and Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO).4 These trials are both considered to be of high quality, but the trials came to substantially different conclusions. The ERSPC showed a significant decrease in mortality in men screened with PSA compared with those that did not receive a PSA, whereas the PLCO showed no difference in mortality between the two groups.
Various differences exist between these two studies that may contribute to this discrepancy: (1) differences in screening interval (annual in PLCO vs every 2–4 years in ERSPC); (2) PSA threshold to biopsy (PLCO: 4.0 ug/L vs 3.0 ug/L in ERSPC); (3) higher prostate cancer incidence in the USA than Europe before the trials started and (4) a varying degree of ‘contamination’ in each of the control groups: many patients in the control group—not randomised to PSA—actually received PSA testing.
In response to these differences, the Annals of Internal Medicine published an analysis of data from both the ERSPC and PLCO with statistical adjustments for the trial differences.5 The results of which showed a 16% (95%CI 4% to 27%) reduction in mortality in those screened.
Can we trust these results?
The methods to statistically adjust appear to be completely novel and therefore not validated previously. This analysis includes pooled data from the two trials and adjusts for age and trial setting as well as four ‘extended analyses’. These four ‘extended analyses’ were conducted to ‘account for variable screening and diagnostic workup’ between the two studies. All four ‘extended analyses’ and the ‘traditional analysis’ concluded that PSA screening significantly reduced mortality.
These methods have attracted a mixed response. Some, including Dr Kenneth Lin, a former medical officer at the Agency for Healthcare Research and Quality, argued that statistical models should not be considered superior to real life patient data—‘No matter how sophisticated, they (statistical models) shouldn’t trump data from real people who participated in the randomised trials’.6 While others called for the controversies surrounding prostate cancer to ‘finally (be) put to rest’.7
The central issue that should inform policy is the question of the amount unnecessary risk subjects’ are willing to accept in order to benefit or save one other person?
The authors of the Annals of Internal Medicine reanalysis of ERSPC and PLCO data report that five men will be overdiagnosed to save one man’s life from prostate cancer.5 These five men will risk urinary incontinence, impotence and further harm for no benefit. The US Preventive Services Task Force estimates that the number of men overdiagnosed to save 1–2 men’s life is closer to 50.8
Therefore, how much risk are we willing to subject patients, with no benefit, to save one other person’s life?
This article first appeared on BMJ EBM Spotlight http://blogs.bmj.com/bmjebmspotlight/2017/10/19/controversies-in-psa-screening/
Footnotes
Twitter @JackOSullivan3.
Contributors JOS conceived the ideas and wrote the article. It is entirely his work.
Funding My DPhil (PhD) is funded by the Clarendon Fund, University of Oxford.
Competing interests I receive income from Oxford University Hospitals for clinical work and hold grants from the National Institute for Health Research and the Primary Care Research Trust. My views are entirely mine and do not represent any of my employers.
Provenance and peer review Commissioned; internally peer reviewed.