Issue 7 / 4 March 2013

EACH year in Australia around 20 000 new cases of prostate cancer are diagnosed and close to 3000 men die — more than the number of women who die from breast cancer.

It is the most common cancer and second to lung cancer as the most common cause of cancer deaths in Australian men.

With early detection, when prostate cancer is generally asymptomatic, the odds of survival are relatively high. It is by any objective standard a major public health problem but one that is potentially amenable to cure by screening and early detection.

The most common method used for early detection, other than physical examination, is the prostate specific antigen (PSA) test. The PSA test is not perfect but it is one of the few non-invasive tools available for early detection. And it’s better than a digital rectal examination (DRE) in detecting asymptomatic prostate cancer.

A meta-analysis reported the pooled sensitivity, specificity and positive predictive value (PPV) for screening PSA were 72.1%, 93.2% and 25.1%, respectively, significantly higher than those for screening DRE, which were 53.2%, 83.6% and 17.8% respectively.

The PPV for PSA in detecting prostatic cancer is greater than the PPV of screening mammography for the detection of breast cancer.

Yet in May 2012 the US Preventive Services Task Force issued advice that the benefits of PSA screening do not outweigh its potential harm, while continuing to urge women to pursue screening for breast cancer.

The warning against PSA screening for prostate cancer was based primarily on results from the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial. It sought to determine if certain cancer screening tests reduced the risk of dying of those cancers. It includes nearly 76 693 people aged from 55 to 74 years with no previous relevant history.

Participants were randomly assigned to either PSA screening or non-screening. However, 52% of the participants in the non-screened arm had a PSA test, diluting any potential statistical difference between the groups.

Despite the significant contamination of the control group in the PLCO study, the preventive task force accepted the study’s conclusion that PSA screening tests had no impact on the prostate cancer mortality rates.

The PLCO trial findings also conflict with those of the European Randomized Study of Screening for Prostate Cancer (ERSPC), which did not suffer from similar cross-contamination. The findings in the ERSPC study revealed a statistically and clinically significant 29% decrease in mortality in PSA screened men.

Despite the disparity, the task force recommended against PSA screenings. Given the fundamental flaw with the PLCO study, one is compelled to ask: why?

Dr Michael Gliksman is an occupational physician based in Sydney and chairman of the Professional Issues Committee of the AMA NSW. On Twitter @MGliksmanMDPhD

Posted 4 March 2013

14 thoughts on “Michael Gliksman: The PSA paradox

  1. Michael Tam says:

    The important outcome in any cancer screening program is not whether it detects more cancer. Rather, it is whether we reduce cancer related mortality and death.

    Furthermore, the USPSTF did not primarily use data from the PLCO trial, but considered the evidence from both it and the ERSPC. It is acknowledged that in the ERSPC that screened men had a lowered rate of prostate cancer death, but this is at the very substantial cost of overdiagnosis and overtreatment.

    It is also beneficial to describe the absolute reduction in mortality, rather than given relative risk reductions. Even in the ERSPC which gives the most supportive evidence for prostate cancer screening, the number needed to treat is around 10,000 men (per year). I’ve calculated this elsewhere, but to contextualise this for Australian readers, even if the average GP screened every male patient aged 50-70 for prostate cancer with PSA, the odds are that they would not prevent a single prostate cancer death over the course of a 40-year career!

    The USPSTF recommendations, which are concordant with the RACGP guidelines, recognises that the potential clinical benefit from prostate cancer screening does not outweigh potential harms from population screening. There is no conspiracy against men. Rather, the best interpretation of the empiric evidence is simply that PSA screening likely does more harm than good and as such, should not be offered routinely.

  2. Diane Moller says:

    It was refreshing and pleasing to read Dr Gliksman’s article. Until there is a more definitive and better test to detect prostate cancer, this is all we have. Why is it so difficult for doctors to add a PSA test to routine blood tests on an annual basis. While PSA testing remains such a controversial issue and men are advised to “speak to their doctor about testing”, we will continue to have men presenting initially with metastatic disease.

  3. Michael Gliksman says:

    Thank you to Dr Tam for his thoughtful analysis. The problem I (and I’d argue, other men) have is not with the accuracy of numbers needed to screen to save a life etc, but the gendered selectiveness in perception of net merit applied to prostate cancer screening vis a vis breast cancer screening.
    The study into ‘Absolute numbers of lives saved and overdiagnosis in breast cancer screening, from a randomized trial and from the Breast Screening Programme in England’ (J Med Screen. 2010 March; 17(1): 25–30) reported that:A substantial and significant reduction in breast cancer mortality was associated with screening in both the Two-County Trial (P < 0.001) and the screening programme in England (P < 0.001). The absolute benefits were estimated as 8.8 and 5.7 breast cancer deaths prevented per 1000 women screened for 20 years starting at age 50 from the Two-County Trial and screening programme in England, respectively.
    ie: 20,000 person years of reasonably targeted screening to save between 5-9 lives, and this before any account is taken of the impact of false positives on women’s lives.
    Is there any GP out there who has practiced for 20, let alone 40 years, who does PSA testing and follow ups to check for change over time, who has *not* detected a potentially curable cancer undetected by other means?

  4. Mark Frydenberg says:

    The issue with USPSTF is whether one looks at this from a societal viewpoint or from an individual’s viewpoint; they are not the same, and as a urologist we are confronted by patients who see it from an individual’s viewpoint. USPSTF looked at all-cause survival over 10 years; when one considers a cancer that predominantly grows slowly then they were never going to conclude a benefit for screening; it would have been appropriate to have also looked at prostate cancer-specific survival. It also included flawed studies – PLCO which included men who were pre-screened before entering hence reducing power of the trial, 52% of men who were not supposed to be tested were in fact tested; and if an abnormal result was found the only requirement was to inform the GP, which was done but only about 1/3 of men actually underwent evaluation or treatment for their elevated PSA. It is also included another flawed trial, from Norkopping trial from Sweden, which had only 1700 men, and for the first half of the trial PSA was not even used, diagnosis was made by fine needle cytology and not by biopsy, most men only attended once due to illness or age; these are not high quality studies to have included in a major review. In addition, psychological and physical risks of treating localised disease were overestimated, risks of symptoms from advanced prostate cancer for those who die from cancer was underestimated; so one could easily argue re: the validity of the class D recommendation, and suggest a class C recommendation instead – where the test could be offered with appropriate discussion and counselling of risks and benefits.

    Studies have also shown that the main driver for a GP ordering PSA tests is if the patient requests it – which assumes that the patient knows what to ask for and that they are educated re prostate cancer which most are not. Studies of patients have also shown that patients have no idea if they are at risk, older men who do not need to be tested are coming along and being tested, men with family histories who are at risk did not perceive that they were at risk, and studies have shown that the main determinants of a patient asking for a PSA test are if they viewed they were at risk, if they had a friend with prostate cancer or if they were affluent; hardly good reasons to be requesting a test.

    Surely it is about time that patients are carefully evaluated if they may benefit from the test or not by their GPs, and if young and healthy enough to benefit, a proactive discussion takes place between GP (who should know the data and be able to provide advice) and patient, and then an informed decision can take place if PSA testing should take place. For the many thousands of men who were never going to benefit, then that discussion needs to occur as well, as currently 60-70% of men over 75 years of age with no symptoms are being tested with PSA when the vast majority would never benefit. Conversely only 25% of men aged 50-60 years are being tested and arguably these are the men that may benefit the most from testing. We need to be doing better than this .

  5. Henry Woo says:

    I suspect that the quoted NNT of 10,000 men per year to be treated to save one life by Dr Tam is a typo given that this is not the finding of the ERSPC study or any other study that I am aware of. The number needed to screen to prevent one death was about 1400 and the number needed to treat to prevent one death was 48 – it is essential that readers understand that the ERSPC was at 9 years follow up which is hardly sufficient to accurately characterise NNT for men with many life years at risk. If we look at the prostate cancer specific mortality curves, we can extrapolate that by 12 years, we will see the NNT figure fall to around 18. We all know that relatively few prostate cancers will kill otherwise well men within 9 years although the anti-prostate cancer screening lobby latches onto this data as well as the largely discredited PLCO data. We can adopt a head in the sand approach and say that because the data is not yet there we should not talk about it unless the patient raises the matter (not far off the current RACGP suggested approach) – but the answer lies in-between where we have to get smarter about who we screen, who we biopsy and who we treat. 3200 deaths due to prostate cancer in Australia each year is not an insignificant disease.

  6. Guy Hibbins says:

    This is what the latest Cochrane Review on screening for prostate cancer has to say:
    Cochrane Database Syst Rev. 2013 Jan 31;1:CD004720 Screening for prostate cancer.
    “Prostate cancer screening did not significantly decrease prostate cancer-specific mortality in a combined meta-analysis of five RCTs. Only one study (ERSPC) reported a 21% significant reduction of prostate cancer-specific mortality in a pre-specified subgroup of men aged 55 to 69 years. Pooled data currently demonstrates no significant reduction in prostate cancer-specific and overall mortality. Harms associated with PSA-based screening and subsequent diagnostic evaluations are frequent, and moderate in severity.
    Overdiagnosis and overtreatment are common and are associated with treatment-related harms. Men should be informed of this and the demonstrated adverse effects when they are deciding whether or not to undertake screening for prostate cancer. Any reduction in prostate cancer-specific mortality may take up to 10 years to accrue; therefore, men who have a life expectancy less than 10 to 15 years should be informed that screening for prostate cancer is unlikely to be beneficial. No studies examined the independent role of screening by DRE.”
    If there were no risk of overtreatment, then I suspect that there would be a lot less controversy about screening. Maybe this is where the prostate cancer debate should be centred.

  7. Ken Sikaris says:

    Dr Gliksman states that PSA is not a perfect test (which is true of all tests) and this is true for PSA despite the massive improvements in understanding and measurement of this natural protein.
    All tests need to be used in a clinically appropriate manner and it has been proven that PSA, as it was used in PLCO, was largely unproductive and therefore inappropriate. There are many men that will not benefit from PSA testing particularly those who are elderly and/or more likely to die of other causes.
    However prostate cancer as a significant cause of death will not go away, unless we apply the lessons we have learnt from this research and focus on the 20% or more of men that can be saved from a prostate cancer death (as was proven in ERSCP).
    The heated and often emotional debate regarding prostate cancer and PSA testing is slowly making way for an objective view of the evidence and the lessons that may benefit patients.

  8. Michael Tam says:

    No, this was not a typo. The number needed to screen was 1400 (i.e., the NNT to screen as opposed to not-screening) over the nine-year period.

    The recent update of the ERSPC at 11 years gives very similar numbers.

    That is, the NNT (PER ANNUM) to prevent one prostate cancer death is around 10,000 men using the ERSPC data (i.e., NNT is of around 10,000 patient-years). This is the better figure to use as it standardises the time period allows for comparisons between data sets.

    Of those DIAGNOSED, the NNT to prevent a death is the aforementioned 48 (if screened over 9 years). If we use the more favourable Göteborg subset of the data, the NNT of those DIAGNOSED is still around 12 – i.e., the most supportive data suggests that 12 men will have to be diagnosed and treated for prostate cancer to have made a difference to a single man for prostate cancer death.

    The other important result that has not be raised so far is that there is absolutely no evidence that prostate cancer screening reduces OVERALL mortality – this was not demonstrated by the ERSPC. We should not assume that a reduction in prostate cancer mortality necessarily implies a reduction in overall mortality.

    There is a basic philosophical problem with the PSA debate that often does not get addressed, and I suspect, have different groups arguing past each other.

    Firstly, there is an assumption that disease screening is necessarily a good idea – i.e., the aphorism that “prevention is always better than cure”. We see this in the suggestion that even a “flawed” test is better than none. We should acknowledge that this position is an ideological one. Prevention is OFTEN better than cure, but the important thing to focus on is “better” – we need to be guided by outcomes.

    Secondly, there is the assumption that screening tests cannot harm, i.e., the idea that surely it is “better to know” then decide. Screening tests should be considered a type of medical intervention. Population screening tests by definition are interventions in people who are otherwise “well”. I would argue that we need to have good evidence that the test is actually beneficial in a meaningful way before its broad implementation. People are harmed by screening programs – by overdiagnosis and overtreatment. Yes, there needs to a consideration of tradeoffs as we can never have a perfect intervention. However, I think the data is reasonably clear on this so far – the majority of diagnoses in PSA screening are likely to be overdiagnoses.

    Thirdly, there is the suggestion that by “not screening” we are ignoring the severity of the disease. This is an unfair position. No one is suggesting that prostate cancer is not a major cause of morbidity and death in men. However, we need to be guided by an empiric approach to management. The question we should be asking should be “is PSA screening an effective strategy in reducing morbidity and mortality?” and “what are the harms to PSA screening?”. The answer from the evidence is that PSA screening appears to yield small absolute benefits to prostate cancer mortality, no benefits to all-cause mortality, and results in not insubstantial harm.

    I completely agree that we need to investigate new prostate can screening methods. However, I disagree with some of the commentators in this thread that we can be “smarter” in the ways we offer this to patients, at this time. To be blunt, all the suggested “smarter” methods of screening are effectively untested. They might work, but just the same, they may not and all the while, patients can be harmed. To engage in this type of management, there is an ethical duty to inform patients that it is guided by your best judgement, but not necessarily evidence.

    The onus of evidence has been somewhat reversed in debates on prostate cancer screening. Proponents seem to be asking for evidence that it is NOT effective, or that it is harmful. In almost every other setting when discussing interventions, it is usually the other way – that we need to demonstrate that an intervention IS beneficial and not harmful before widespread recommendation!

    We should reflect the historical context in which the debate takes place – that we have been widely offering prostate cancer screening in the absence of empiric evidence. The uncomfortable reality we now face is that the evidence does not support our beliefs on the “goodness” of the test. Dr Gliksman’s identification of the issue with breast cancer screening highlights a very similar problem. Breast cancer screening is likely to be much less effective than we believed/hoped, and certainly much less effective than what the general public believe. The issue of overdiagnosis is a major one – somewhere between a quarter to half of breast cancers detected on routine screening mammography are probably overdiagnoses. There are major ethical considerations here (e.g., informed consent to screening) that are still yet answered.

  9. Bruce Campbell says:

    I am retired pathologist and the editor of Lab Tests Online, an independent website funded by the Department of Health and Ageing to provide information about pathology testing for the Australian general public. Given the amount of controversy about prostate cancer screening and breast cancer screening amongst medical practitioners, imagine how difficult it is for us to explain this to lay persons. We have however, tried to do this and a couple of news items relating to these issues might be useful resources for practitioners to refer their patients to. The most recent entitled “Cancer Screening – It’s not as easy as you might think” relates to breast cancer screening and an earlier item entitled “Evidence-based medicine and PSA testing” was in response to the USPSTF announcement. Both of these can be found on the site at and either using the search box or looking back through the news items.

  10. Troy Gianduzzo says:

    I would be grateful if someone would explain the derivation of 10,000. The screening interval in the ERSPC was 4 years. Therefore I feel that to discuss the number need to screen in annual terms is simplistic. It appears as though 1400 (or thereabouts) is multiplied by 9 years. The correct answer is 1400 men over 9 years need to be screened. That is all that the paper said and that is the study conclusion. Interpretation of the figures beyond this relies on assumptions that are fraught with error. If anything one would DIVIDE 1400 by 9 (ie. approx 155) but again this is overly simplistic.
    Either way, when compared to breast cancer screening, the results are still favourable. The Cochrane review on mammographic screening identified that 2000 women needed to be screened over 10 years. The ERSPC of 1400 over 9 years is, at worst, comparable. This was the thrust of Glicksman’s article.
    At worst the number needed to treat is 48 for prostate cancer. At best, when corrected for co-morbidity, the number needed to treat is 5 as demonstrated by Crawford et al J Clin Oncol 29:355-361. In this analysis PLCO data was examined and the effect of co-morbidity considered. For men with minimal co-morbidity the number needed to screen was 723 while the number needed to screen was 5. This is what is meant when Henry has said we need to be “smarter”.
    Ultimately this does come down to the definition of “screening” and the definition of “number needed to treat”. Testing every man and his dog irrespective of competing health risks is not sensible. However select and judicious use of PSA in men likely to benefit is helpful. Is this then PSA “screening” or is this case finding?
    Also what is meant by “number needed to treat”? Not all men receive radical treatment. Active surveillance is increasingly used. In the Goteborg study just under 1/3 of those diagnosed did not receive active primary treatment. So the number needed to treat of 12 is misleading and does not imply that 12 men need to have radical prostatectomies to save one life. Rather 12 men need to be DIAGNOSED and MANAGED, of which 1/3 had surveillance. Therefore one can argue that the number need to DIAGNOSE in the Goteborg study was 12 while the number needed to treat actively was around 8.

  11. Rose says:

    The ERSCP study confirmed a significant decrease in mortality for PSA screened men-that is good enough for me to continue screening. What about morbidity-surely men with prostate cancer want a reduction in morbidity, such as bone pain, and lytic fractures?

  12. Dr Horst Herb says:

    What the article fails to mention is that while we have very effective and proven treatment to offer for breast cancer, the benefit for treatment of asymptomatic prostate cancers detected through PSA screening has yet to be demonstrated. Treatment causes significant morbidity (impotence, incontinence) and often dramatically reduced quality of life as I often witness it in my patients.

    How many of the 3,000 men who die of prostate cancer in Australia have been
    a) screened
    b) treated?
    These figures seem hard to come by, and yet the relevance of the screening would very much hinge on them

  13. Marilyn McMurchie says:

    My concern about PSA screening is not the statistics above but the potential negative financial and emotional risks to men if their test is in the mid range. The usual GP response will be to refer to a specialist urologist. I know of only two urologists working in the public sector in Sydney; one only sees men with prostate cancer and the waiting list for the other was approaching two years last time I checked. Urologists in private practice in general seem to charge AMA fees. In my view, any screening should be supported by a system of follow-up and I don’t see that here.

  14. Ben says:

    In the discussion above, and in some of the research papers, there is mention of disease specific mortality. This is an accounting trick and should be discouraged – if your screening program saves lives, then mortality should be lower overall. If it is not lower, then the extra mortality (the gap between disease specific and total mortality) in the screened group had to come from screening and treatment program (unless the study was badly designed). From the ERSPC 11 year follow up: Overall mortality was similar in the two study groups, with 18.2 deaths per 1000 person-years in the screening group and 18.5 per 1000 personyears in the control group (rate ratio, 0.99; 95% CI, 0.97 to 1.01). Finally, the list of conflicts amongst the authors is disturbing – as well as the familiar “lecture fees”, one Author holds a PSA assay patent, and another has been a board member of some big pharmaceutical companies. We should keep research and advertising separate.

Leave a Reply

Your email address will not be published.