From Amsterdam yesterday: European perspectives on the PSA screening issue

A report on the Medscape Oncology web site, published late yesterday, discusses two presentations given yesterday at the European Cancer Conference 2013, currently being held in Amsterdam.

The first of these two presentations was given by Mathieu Boniol, MD, the research director at the International Prevention Research Institute (IPRI) in Lyon, France.

According to Dr. Boniol, in France today, about 55 percent of men between the ages of 55 and 69 years receive regular PSA testing (and family doctors routinely add PSA tests when ordering standard blood work). As a consequence, “More than 80 percent of men aged 65 years have received a PSA test during the past 3 years.” But PSA testing has not lowered prostate cancer mortality rates.

In France, in the 1980s, before the advent of PSA testing, the incidence of prostate cancer was 5 percent, and prostate cancer-specific mortality was 2 percent. Currently, with widespread PSA testing, the incidence of prostate cancer is 14 percent (a 180 percent increase since the 1980s) and prostate cancer-specific mortality is still 2 percent. As others have argued before, Dr. Boniol stated that “We are now finding cancers that never would have appeared in these men” from a clinical  perspective during their lifetimes.

We should be very clear, however, that Dr. Boniol and his colleagues are not suggesting that no one needs PSA tests. They apparently pointed out with great care  that “there are high-risk groups, such as men with a family history of aggressive disease, who can benefit from PSA testing.”

In his presentation, Dr, Boniol also argued that the potential harms associated with mass, population-based PSA-based screening for prostate cancer outweigh the potential benefits. Specifically, based on data from a variety of sources, Dr. Boniol and colleagues made two sets of estimates.

In a group of 1,000 men aged between 55 and 69 years of age who were not exposed to PSA screening for risk of prostate cancer, there would be

  • 116 biopsies performed
  • 60 cases of prostate cancer identified
  • 119 deaths overall
  • 5.17 deaths would be as a consequence of prostate cancer

Conversely, in a group of 1,000 men aged between 55 and 69 years of age who were exposed to PSA screening for risk of prostate cancer, there would be

  • 270 biopsies performed
  • 96 cases of prostate cancer identified
  • 191 deaths overall
  • 4.1 deaths would be as a consequence of prostate cancer

In other words, for one prostate cancer-specific death to be prevented among 1,000 men, there would have to be

  • 154 additional biopsies
  • 9 cases of hospitalization for severe adverse events
  • 0.2 deaths as a consequence of complications of biopsy
  • 35 additional prostate cancers diagnoses (32 of which would be among me with low-risk disease)
  • 12 additional cases of impotence
  • 2 additional cases of incontinence
  • 1 additional case of fecal incontinence

The authors also point out that, since 18 percent of prostate cancer–related surgery is performed on men who are > 70 years of age, physicians should be aware that in such a set of 1,000 men

  • 183 deaths (0.15 percent) occurred 60 days after prostate cancer surgery.
  • The overall risk of dying was
    • 0.11 percent for men aged 40 to 69 years within 60 days of surgery
    • 0.36 percent for men of 70 years or older within 60 days of surgery

Now The “New” Prostate Cancer InfoLink is unable to either endorse or refute the data presented by Dr. Boniol. We don’t know enough about the authors’ data sources; nor do we have any detail about the statistical assumptions built into this model. On the other hand, what we do know is that it does make assumptions about the continuing inability to discriminate with accuracy between men with indolent disease and men with clinically significant forms of prostate cancer — and the consequent risks for potentially unnecessary biopsies and subsequent over-treatment of low-risk forms of prostate cancer.

The second presentation, by Jack Cuzick, PhD, professor of epidemiology at the Wolfson Institute of Preventive Medicine at Queen Mary University of London, was focused on the importance of being able to discriminate better between men with clinically significant disease and those with indolent disease who can probably be just monitored. According to Dr. Cuzick:

One of the real challenges, and in my mind, the major challenge that has to be resolved before we can really embrace screening is to separate the indolent from aggressive cancers. There is a crucial need to identify which cancers are likely to be fatal, and new molecular markers are needed.

Also according to Cuzick, the UKCAP/ProtecT initiative now includes 450,000 men and is expected to report its findings in 2016. He apparently argued that the data from this initiative may be the important “game-changer” in helping us to better understand who needs treatment and who does not, However, he also went on to note that he doesn’t believe that mass prostate cancer screening is viable in the general population until better risk markers have been identified and established.

18 Responses


    The estimates show a difference of one life saved for 1,000 men screened. That looks suspiciously like the summary number from the 11-year median follow-up (since randomization, not since diagnosis) in the ERSPC, but I have not run all the percentages against the ERSPC results at 11 years.

    I have a hunch Dr. Boniol has simply illustrated the findings of the ERSPC at this still early point in follow-up for the ERSPC, a point at which it is reasonable to expect a substantially greater benefit from screening as follow-up matures.

  2. In a comment on a previous post on this website about Dr. Boniol’s studies, I pointed out that there are several critiques of his work that argue that his estimate of the death rate from biopsies (1 death due to biopsy for every 500 biopsies) is probably too large by a factor of 400. Therefore, there is the possibility that the 0.2 number used for deaths by biopsy per life saved due to PSA screening is too high by a factor of 400.

    Jim Waldenfels (above) also points out that it looks as if Dr. Boniol’s analysis of lives saved probably only looks out 11 years or so, which will understate the lives saved.

    In Ruth Etzioni’s simulation studies using the European data, she projects the likely long-run mortality consequences of screening, beyond what we observe for the first 11 years. She comes up with “numbers needed to treat” to prevent one death from prostate cancer of somewhere between 3 and 7 in the long-run.

    So, for every one prostate cancer death averted, we might have to have 3 to 7 additional men treated due to the screening program. If this is so, then the impotence and other harm numbers presented by Dr. Boniol are much too high.

  3. There is no way that only 3-7 men need to be treated to prevent one death, those numbers just don’t work. Also, greater care is needed by everyone arguing that while screening is not good for everyone, high-risk men still benefit. That is what we hope but not a single study has ever demonstrated a differential benefit for the high-risk men. It is still possible that men at greater risk for suffering from the disease have a few cancer cells that spread so early; PSA is not able to identify them before that occurs. Assuming that high-risk men benefit from screening is the same argument that was used from the beginning that men of any risk would benefit, something we now know to be incorrect. We need some proof that this is correct.




    First, thank you for posting your thoughts, Dr. Chodak.

    Regarding point 1, do we really need another study, or can we just think this through? Maybe I’m just too used to seeing this from a patient’s perspective, but the benefits of screening (let’s make that “smart screening”) for men who turn out to be high risk look indisputable to me. With my own case as an example, my baseline, first-ever PSA was 113 in December 1999, and subsequent evidence has indicated a PSA doubling time of 3 to 4 months. Using 4 months, that would give me an untreated PSA of 904 by the end of 12 months, 7,232 ng/ml by the end of 24 months if I were still on Earth, and a theoretical 57,856 ng/ml by the end of 36 months. My recollection is that symptoms are virtually certain when the PSA hits 1,000, and metastasis, likely wide-spread metastases, are also likely as PSA rolls into the thousands. My impression is that many men with high-risk case characteristics would also, like me, likely suffer or die if undetected and untreated. Can we not regard the likelihood of suffering and death as a “given” for high-risk patients, dispensing with any further study?

    Regarding point 2, the question then is whether detection for a substantial portion of high-risk men in these days of modern staging, decision-making, and treatment can make a difference, or whether the spread of a few deadly cells will happen so often and so early that detection is not really of much use. Many of those of us in the hormonal therapy patient community are convinced that detection makes a huge difference. Even late detection, as in my case, can be timely. Again, looking to my own case, ADT3 reduced my PSA to less than 0.01 twice, and on the third round to 0.02, with a possible androgen receptor mutation possibly responsible for the slightly elevated third round nadir. At the 14th year point, I have benefitted from fairly new staging technology, and have been able to have a hopefully curative series of TomoTherapy radiation treatments. Under continuing supportive ADT3, my PSA is continuing to decline, with 0.07 the latest value. I have heard an expert in ADT3 state that typical patients enjoy about 10 to 11 years of success in controlling cancer or indefinitely long success. Some doctors are now even attempting curative therapy with radiation for men with oligometastatic disease – five or fewer bone mets, with any nodal disease of a type that can be targeted.

    Can there be any doubt that for high-risk patients, especially patients like me, detection can prevent great suffering if not loss of life?

    (Getting back to the larger picture, I too have doubts about mass screening programs, but strongly favor smart screening.)

  5. Dear Jim:

    With all due respect, passionate belief is one thing and actual data are quite another.

    Just look at the history of medicine to remind yourself how many times the passionate beliefs of physicians and patients have been proven to be wrong when we actually tested the relevant hypotheses in neutral, well-planned, and well-implemented clinical trials.

    Do I want do believe that screening of men at supposed high-risk for prostate cancer would have a real survival benefit? Sure I do! Can I prove it through any type of “mind experiment” based on the available data. No, I can’t.

    And I am also not clear which criterion you would ever have met to be identified as a high-risk individual prior to your diagnosis.

  6. Hi Sitemaster,


    My case features included a baseline PSA of 113.6, Gleason 4 + 3 =7 (read by Dr. Epstein at Johns Hopkins), Stage III per Johns Hopkins very experienced urologist, all biopsy cores positive for cancer, most 100% cancer, and perineural invasion. After starting Lupron my PSA flared to 125 within 2 weeks before dropping. I had prognoses of 5 years — 3 good years, 2 declining, first from a highly respected urologist at the City of Hope in Duarte, CA, and then from a highly respected urologist at Johns Hopkins. I was on Lupron so fast (not from Johns Hopkins) that I never had a follow-up PSA uninfluenced by ADT except for that flare, which I’m disregarding. My PSA has doubled while on vacation from Lupron and the antiandrogen, maintained with a 5-ARI, consistently in three to four months once I hit a PSA of 5. (When it has hit about 10, I’ve gone on low-dose thalidomide with vitamin B6 to extend the vacation period.)

    I understand and accept your point about passionate belief skewing a person’s view of reality. However, I find the logic compelling.

  7. Dear Jim:

    You are missing my point, which is that I am not aware of any information suggesting that you fell into a high-risk category prior to your diagnosis. As far as I am aware you are not African American; you had no significant family history of prostate cancer; etc.

    No one is implying that you were not classifiable as a D’Amico high-risk patient at the time of diagnosis. That is self-evident. The question we are discussing here is whether PSA screening of patients known to be at high risk prior to diagnosis would actually have a survival benefit. I am not aware of any data that would have identified you as a high-risk patient based on those criteria … but maybe there is something I don’t know.

  8. Dr. Chodak makes the statement: “There is no way that only 3-7 men need to be treated to prevent one death, those numbers just don’t work”.

    Dr. Chodak has apparently not read Ruth Etzioni’s work on this topic. This provides a quite elaborate simulation model, based on the European data, that in fact comes up with numbers like that (see, for example, this paper and this abstract.

    Etzioni’s model may not be correct. It relies on data and some assumptions about the natural course of prostate cancer over time. It extrapolates the European data beyond the years included in the study. There is therefore uncertainty. But it is clearly a plausible model that is consistent with the available evidence. If a plausible model that is consistent with available data comes up with an estimate, it may be wrong, but it is not correct to say that “there is no way” that this number is true, and that “those numbers just don’t work”.

    Each of us is entitled to his or her own opinion. We’re not entitled to our own facts.

  9. But Tim … You just stated that Dr. Etizoni’s model is based on projections based on assumptions. So those are no more “facts” than is Dr. Chodak’s 40-odd years of clinical experience actually diagnosing and treating men with prostate cancer.

  10. My point is that Dr. Chodak’s statement that “There is no way that only 3-7 men need to be treated to prevent one death, those numbers just don’t work”, is inconsistent with the fact that there are well-done studies that find that these numbers DO work under a particular set of modeling assumptions and using a particular data set that yields empirical results.

    And although Etzioni’s model is based on projections and assumptions, it is also based on facts. That shouldn’t be left out of describing Etzioni’s model.

    All models/theories in science are based on some combination of facts and assumptions. A model is never proven, it is simply a model that at present can not be disproven by the data. If later on, newly discovered data (facts) disprove the model, then it is no longer valid, and we need a new model/theory/.

    It is not really much of a criticism of a model to say that it is based on assumptions. A model is only destroyed if its empirical implications are found to be inconsistent with measured empirical data.

    If Dr. Chodak has empirical data from his clinical experience that he feels disproves Etzioni’s model, he should bring forth his evidence. Ideally, he would publish his evidence in a peer-reviewed journal. If his evidence is convincing enough, Etzioni’s model will be disproven.


    I was puzzled by your question but now see your point. I was, in fact, in a high-risk group prior to diagnosis based on my father’s death from metastatic prostate cancer and his father’s death (my biological grandfather) also from metastatic prostate cancer. Unfortunately for me, our family glossed over the cause of my grandfather’s death for years, so I was unaware of the cause until after my diagnosis. Despite the fact that I listed my father’s cause of death as prostate cancer on my history form, my primary care doctor at the time of my fateful exam, at age 56, did not recommend the PSA test. In fact, I had to insist. He was embarrassed when the result came back so high.

    So I now recognize the discussion, keying on Dr. Chodak’s point, is about whether it would help to study the benefits of screening for men in one or more of the higher-risk groups before diagnosis. I’ll give that some thought, though my own experience gives me a bias and plenty of passion.

  12. Dear Tim:

    If Dr. Chodak (or anyone else) had the sort of data you ask him to provide, we wouldn’t be having this conversation.

    All that I can tell you is that, after monitoring the epidemiological and clinical research on prostate cancer for 25+ years, I also find it extraordinarily difficult to believe the proposition that treatment of between 3 and 7 men diagnosed with prostate cancer will ever result in the saving of a single life from prostate cancer-specific mortality (on average). It simply doesn’t correlate with clinical experience in any way, shape, or form.

  13. Policy should be based on scientific findings. Clinical experience should only overrule scientific findings if clinical experience provides overriding evidence that is relevant to what those scientific findings are about.

    No clinician has ever “experienced” in their clinical experience what would happen to 60,000 men randomly assigned to screening and 60,000 men randomly assigned to not be screened by PSA screening over the rest of their natural lives. Of all the articles you have read over 25+ years, I doubt very much whether more than a handful of those articles have made any serious attempt to use empirical data to assess what will be the causal effects of screening vs. non-screening over the rest of a man’s life, with that modeling attempt based on a very large random assignment experiment.

    Medical issues that involve long-term outcomes that are rare, which is the case for PSA screening, are not well suited to be decided based on articles or a more time limited and sample size limited clinical experience.

    Therefore, Dr. Chodak’s clinical experience, and whatever hundreds of articles based on research you have read over 25+ years, cannot over-rule what is found by Etzioni’s model.

    To be blunt, your intuitive feelings about what you find hard to believe do not trump simulation findings from Etzioni’s empirically-based model, because your intuitive feelings are not based on studies that are looking at the longer time periods that Etzioni is considering.

    Intuitive feelings do not trump models based on data. Intuition does not beat models for issues which it is frankly difficult if not impossible to have any real world based intuition.

    To put it more succinctly:

    Why on earth should we think that human evolution has allowed human intuition to do a good job in developing intuitions about issues that involve long time periods and rare events?

    Human intuition can be very powerful when applied to subjects that human beings deal with all the time in their daily lives. But long time periods and rare events are not exactly in that category.

  14. What an interesting perspective!

  15. Dear Sitemaster,

    In one interpretation, timbartik is correct. Let me state what I think he means. We need more pure scientific knowledge (or: models) about prostate cancer. Right now this is coming from academic institutions, not from clinical trials. They take too long, supply guiding experience, but no knowledge as to causal mechanisms.

    I am helping a research physician think about this, now. If such knowledge is available, then relevant trials are secondary to that. I am assuming that the pure knowledge is generally accepted by research scientists, not by physicians, who merely *use* the knowledge. In my case one and only one relevant trial had been completed, that of Anders Widmark. Yet my treatment and initial conditions differed but were consistent with those his group studied. It was thus an educated guess. Moreover, I received the highest dose escalation for HDR brachytherapy, and the longest duration of ADT, ever tried in this combination therapy. Here, educated scientific guesses built upon but did not copy most available clinical trial set-ups. There was at least one purely retrospective study that was relevant, but no trials of precisely this technique. What should I have done? Waited years in uncertainty at best, until a clinical trial was organised and completed? I would most likely be dead right now. I think this is one way of stating timbark’s points.

  16. Dear George:

    No one has ever suggested or implied that we don’t need “more knowledge”. However, “more knowledge” does not exclude the concept of “more wisdom.” The best scientific advances are achieved by combining data and insight — such insight, in many cases, does include the experience of clinicians, at least some of whom have extensive scientific training.

  17. Dear Sitemaster,

    I agree. The experience of at least some clinicians should be combined with insight. One problem with your formulation though: the insight need not come from completed clinical trials. It can come from personal experience of a clinician, guided by his or her retrospective view of the field, within a framework of generally accepted concepts, theories and currently available, accepted, data. As an actual illustration, I knew an historian of science who told me that astrology and Newtonian mechanics are both theories and that neither is preferable to the other. I was tempted to ask him if he would use both notions together, in explaining some types of motion. The moral is, that the notion of “insight” must be spelled out in detail, not used as a place-holder.

  18. Dear George:

    I never suggested or implied that “insight” had to come from clinical trials.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: