The Göteborg prostate cancer screening trial — now that we’ve read the paper


On Thursday we provided an initial summary of media-reported data from the paper just published by Hugosson et al. in Lancet Oncology. Following is a carefully considered analysis after receipt and study of a copy of the full paper and additional information.

The Göteborg study was initiated in 1994, when 20,000 men from the local population were randomly assigned to one or other of two groups of patients: a screening group and a second group who would receive no screening (a control group). The men in this study were all born between 1930 and 1944, and were therefore between 50 and 65 years of age at the start of the study. We should point out immediately that in 1994 the use of PSA testing in Sweden was in comparative infancy, and so almost no one in this cohort of patients would have received any prior prospective testing for PSA at the time this study was started. It is also worth noting that the 20,000 men in this study comprised 61.9 percent of the total male population of Göteborg aged between 50 and 65 years at initiation of the study. It therefore represents a true mass, population-based screening trial and not just early detection.

Analysis of all data in this study is provided for both the entire set of 20,000 men and broken down by three age-related cohorts: men born between 1930 and 1934; men born between 1935 and 1939; and men born between 1940 and 1944.

The patients in the screening group were all invited to receive a PSA test once every 2 years until they reached the upper age limit of the study (which apparently ranged from 69 to 71 years of age, with a median of 69 years). Men in the screening group with elevated PSA levels were offered additional tests (e.g., a digital rectal examination and a prostate biopsy).

The authors have clearly stated that this is only the first report of data from this study. The study is ongoing, because the men who have not reached the upper age limit are still being invited for biannual PSA testing. This report provides cumulative prostate-cancer incidence and mortality data calculated only up through December 31, 2008.

It is important to note the following:

  • Extraordinary efforts were made in this study to ensure that deaths from prostate cancer really were deaths that were specifically or highly probably a consequence of the disease, and not some other associated problem.
  • Prostate cancer-specific deaths included deaths caused by the disease itself and also deaths resulting from any diagnostic or treatment-related intervention specific to the disease.

So what have Hugosson and his colleagues shown over the first 14 years of their study?

Their findings are reported as follows:

  • 48 men in the screening group and 48 men in the control group were excluded from the analysis because they died or emigrated before the randomization date, or because they already had been diagnosed with prostate cancer, leaving 9,952 men in each group who were eligible for analysis.
  • 7,578/9,952 men in the screening group (76 percent) actually received at least one PSA test.
  • 1,138/9,952 men in the screening group and 718/9,952 men in the control group had been diagnosed with prostate cancer as of the end of 2008.
    • Of the 1,138 men diagnosed with prostate cancer in the screening group, 1,046 (91.9 percent) had actually attended a screening clinic and received at least one PSA test and 92 (8.1 percent) never attended a screening clinic.
  • The cumulative incidence of prostate cancer was 12.7 percent in the screening group and 8.2 percent in the control group.
  • The total numbers of deaths in the two arms of the study were almost identical, at 1,981 in the screening group compared with 1,982 in the control group.
  • 44 men died of prostate cancer in the screening group compared to 78 men in the control group.
    • Of the 44 men who died of prostate cancer in the screening group, 27 were men diagnosed after attending a screening clinic and 17  were men who never attended a screening clinic.
    • Half of the men in the screening group who attended a screening clinic and died of prostate cancer (13/27) were diagnosed at their first visit to a screening clinic.
  • 109/1,138 men (9.6 percent) in the screening group and 54/718 men (7.5 percent) in the control group who were diagnosed with prostate cancer died of other, unrelated causes.
  • The absolute cumulative reduction in risk of death from prostate cancer at 14 years was 0.40 percent (down from 0·90 percent in the control group to 0·50 percent in the screening group).
  • The ratio of the rate of prostate cancer-specific mortality was 0.56 in the screening group overall (9,952 men) compared with the control group.
  • The ratio of the rate of prostate cancer-specific mortality was 0.44 for screening test attendees (7,578 men) compared with the control group.
  • Overall, 293 men needed to be invited for screening and 12 needed to be diagnosed to prevent one prostate cancer-specific death.

In their summary interpretation of the data from this study, Hugosson et al. state that, “prostate cancer mortality was reduced almost by half over 14 years. However, the risk of over-diagnosis is substantial.” They further note that the apparent benefit of prostate cancer screening (based on their data) compares favorably to the apparent benefit of other cancer screening programs (such as breast cancer screening).

In his editorial comments about this study, Prof. David Neal of Cambridge University makes a number of careful observations, including the following:

  • This is a relatively small study (with only 20,000 men even as originally planned).
  • The overall reduction in mortality (at 0.56) is much larger than that seen even in the ERSPC (at 0.8).
  • The outcome in the Göteborg study at 9 years was similar to that of the ERSPC study at 9 years.
  • In the Göteborg study, the median time since diagnosis is 6.7 years for men in the screened group and 4.3 years for men in the control group — which is much shorter than the equivalent data for the men in the ERSPC study.
  • The median age of the men in the Göteborg study was 4 years younger than the median age of men in the ERSPC.
  • Only 56 percent of the cancers found in the Göteborg study met the D’Amico criteria for low-risk disease.
  • A diagnosis of prostate cancer in the Göteborg study did not necessarily result in radical treatment.
    • About 40 percent of the men diagnosed with prostate cancer in the screening group were placed on active surveillance protocols.
    • About 28 percent of the men diagnosed with prostate cancer in the screening group have remained on active surveillance throughout the study period to date.

Neal concludes that this study does not imply that widespread, population-based prostate cancer screening initiatives should be introduced internationally. His primary reason for this conclusion appears to be that the population that started being screened in 1994 was “screening naïve.” As an inevitable consequence, there was a very high probability of identifying men with relatively high-risk prostate cancer. While that may still be true today in large parts of a country like Nigeria, it would no longer be true in North America or in most Western European nations.

In countries where PSA testing is already widely used, Neal concludes that, until better tests are available, “Men should be aware of the benefits of early detection of prostate cancer,” and goes on to state that, “Current programmes that raise awareness and provide balanced information about the pros and cons of [individual testing] seem to be the right way forward.”

Having been able to review the complete, published article, The “New” Prostate Cancer InfoLink has to say that the degree of detailed information provided in this report is extraordinary, and the authors are to be highly commended for the clarity of the information provided.

Whatever else people may be able to make of this study, we believe that there are some very important facts to be learned:

  • This study appears to show clearly that, in a screening-naïve population of men aged between 50 and 70 years of age, biannual PSA testing can lower the risk for prostate cancer-specific mortality by at least 40 percent.
  • In addition, the study shows that the proportion of patients diagnosed with prostate cancer and requiring hormone therapy in the screening group (103/1,138 or 9.1 percent) was much less than half that of the patients in the control group (182/718 or 25.3 percent), implying that early detection also reduced the risk for metastatic disease.
  • However … the study also shows clearly that (at 14 years of follow-up) biannual PSA screening has no impact whatsoever on the overall mortality rate in the same population.

We are therefore potentially faced with the difficult question of whether mass, population-based screening that does affect disease-specific mortality but does not affect overall mortality is justifiable based on the costs, the effort, and the potential harms to the men who are over-treated.

Please note that we are referring solely to mass, population-based screening — as carefully distinguished from early detection of prostate cancer in individuals who choose (of their own free will) to undergo testing after an appropriate discussion of the risks and benefits of such testing with an appropriately informed health professional. We wish to be very clear that the right of the individual to elect to be tested is — at least in our minds — both sacrosanct and very different from a recommendation to all men who meet certain specific criteria that they should undergo PSA testing.

The single most important fact about this study, as far as The “New” Prostate Cancer InfoLink is concerned, is that it finally has provided us with a highly structured, ongoing assessment of the potential value of mass, population-based screening for prostate cancer in a previously screening-naïve population. The study also includes full treatment information on all men diagnosed with prostate cancer over the course of the study. This means that at last we have a real baseline against which to assess the data from all other screening studies, and we can use this baseline to recognize the inherent problems of the PLCO and ERSPC studies, which include short follow-up (to date) in both studies, variation in protocols (within the ERSPC cohorts), and data adulteration resulting from PSA testing among the “unscreened” patient cohort (in the PLCO study).

The data from the Göteborg study may still not provide a convincing rationale for mass, population-based screening based on use of the PSA test, but it certainly does set the standard for what must be expected from any new test that may come along and show promise as a true screening test for prostate cancer in the future. The one regrettable fact about this study is that if it had included just one additional age cohort (of men born between 1945 and 1950), we might have been able to gain real insight over time into the benefits of even earlier detection for a period of up to 30 years.

32 Responses

  1. Consider that now better screening test are available, such as PCA3 and other genetic tests, the number of false positives and false negatives could be reduced at low cost. One should look forward as well as backward.

    More effort spent on rapid movement of improved technology into clinical practice, instead of years of effort on old technology as a prerequiste to progress would be appropriate. Look at what has happened to progress in other technology in the same period surely casts a black mark and shows that something is badly amiss.

  2. No one that I know of questions the very real need for better tests that can distinguish with accuracy between risk for clinically significant and risk for indolent prostate cancer. No such test exists as yet, however, despite literally multiple millions of dollars and 20+ years of research. But it would be wrong to think that there hadn’t been an enormous ongoing effort.

  3. Having not read the full reference and based on your comments, I wonder how many of the screened population were actually treated with intent to cure and how many with hormonal suppression. Also, how many of the control group were treated with hormone suppression after their diagnosis.

    Hopefully I’ll visit the medical library next week to get a copy of this paper. It seems like an important one to study and understand.

  4. Ralph:

    The numbers are as follows:

    — Primary prostatectomy: 468 in screening group; 241 in control group.
    — Primary radiation: 93 in screening group; 75 in control group.
    — Primary hormone therapy: 80 in screening group; 162 in control group.
    — Primary surveillance then curative Tx: 142 in screening group; 36 in control group.
    — Primary surveillance then hormones: 23 in screening group; 20 in control group.
    — Surveillance through last follow-up: 314 in screening group; 152 in control group.
    — No treatment at all: 18 in screening group; 32 in control group.

    The primary RP group of patients includes 9 patients who had cryosurgery. The primary surveillance followed by Tx group includes 2 patients who had cystoprostatectomies. The no treatment group includes 7 cases that were only detected at autopsy (which I assume to be all in the control group, although the paper does not say this explicitly).

  5. Oh, I am sorry. I had thought that there were better screening tests than PSA now. I am not sure it is necessary to be able to do more than better decide if a biopsy would provide useful information to the patient in the screening context so such a decision by the patient could be made more intelligently. Of course, it would be even better if a less invasive test could replace a biopsy and provide better diagnostic capability. I wonder if a little improvement of quality of a screening test would have a large non-linear improvement in mortality results, given the poor performance of the PSA test that has provoked such controversy and studies as the one reported.

    As I understand it, once it has been determined by biopsy to be a cancer, the Gleason score is currently they best we have for telling whether the cancer is at risk for being aggressive or indolent. Unfortunately for many cancers diagnosed now, Gleason does not provide sufficient information to avoid over-treatment. Perhaps the Memorial Sloan-Kettering six categories of Copy Number Alteration, which is suggested to be more robust than Gleason, that you reported on earlier, may help in the future

    In the reported study advances such as active monitoring were included in the treatments, but no improvement in screening method apparently occurred.

  6. Dear Tasker:

    All that I can tell you is that finding better ways to discriminate between the prostate cancers that need to be treated (based on their pathology, their genetic profile, and many other factors) and those that are indolent is a major, major priority for almost everyone in the prostate cancer research community.

    The problem is that the evolution of prostate cancer (in most cases) takes such a long time, and the tests we have available today do not seem to be able to spot any single critical factors that can discriminate the aggressive cancers from the indolent ones early on. In an ideal world — perhaps even within 10 years — we may be able to develop what I would describe as “multi-component” tests through which a single blood sample can be used to classify risk based on a combination of biological and genetic factors, but we are still going to need to make sure that such a test is not just theory but actually works better than a PSA test in clinical practice over time. The potential utility of imaging to develop a “male mammogram” is another strategy that has been much sought after, but a male mammogram would still require accurate biopsy data to be able to help us make decisions about who to treat and who not to treat.

  7. A couple of questions: What definition of “elevated PSA” was used, and does the full paper state how many additional biopsies were performed in the screened group?

  8. Dear Jim:

    1. The definition of “elevated” PSA changed over time as the study progressed. Initially the PSA threshold set for further urologic workup was 3.4 ng/ml (equivalent to a WHO corrected value of 3.0 ng/ml). In 1999 the threshold was changed to 2.9 mg/ml (equivalent to a WHO correct value of 2.5 ng/ml). As of 2005 it was changed again to 2.5 ng/ml due to a change in the assay calibrator. This is all addressed in considerable detail in the study report.

    2. With regard to the number of biopsies, of the 2,469 men who had an elevated PSA as defined by the study criteria at specific points in time, 2,298 (93%) received at least one biopsy and a total of 4,153 biopsies have been carried out to date, which means that at least 1,855 of those biopsies were given as secondary, tertiary or possibly even quaternary biopsies.

  9. The Göteborg study: 44 men died of prostate cancer in the screening group compared to 78 men in the control group.

    The Göteborg study: The total numbers of deaths in the two arms of the study were almost identical, at 1,981 in the screening group compared with 1,982 in the control group.

    Dr Willet Whitmore: Growing old is invariably fatal while prostate cancer is only sometimes so.

  10. Since the all-cause mortality was the same, it seems to me that the full advantage of PSA testing would occur when men reduce their deaths from other causes: lung cancer , heart attacks, etc.

    Any other conclusions from that aspect of the results?

    Stan

  11. Stan: I think that is Terry Herbert’s precise point in his comment above.

  12. It’s not clear how the people in the control arm diagnosed were diagnosed. Does anybody know that?

    With regard to active surveillance: It would be interesting to know how long did men stay in active surveillance before switching to treatment? What was the trigger to switch from active surveillance to treatment? How many people in the control and screening arm died while on active surveillance, i.e. from other causes?

    One argument that people opposing PSA screening have is the large number of tested men per incidence of prostate cancer death and the number of men suffering from side effects due to “over diagnosis.” This study indicates 293 men need to be tested to prevent one death. I wonder how will this number be considered.

  13. Indeed it was. It seems at times that men diagnosed with prostate cancer feel that if they can only “cure” the prostate cancer they’ll live forever, but they will not.

    Prostate cancer is still a minor cause of death to men (less than 3%): most men, diagnosed with prostatew cancer or not, die of something else, and every study underlines this.

  14. Dear Reuven:

    (1) The men who were diagnosed with prostate cancer in the control arm would almost certainly have been diagnosed as a consequence of symptoms of progressive prostate cancer, ranging from lower urinary tract symptoms (LUTS, e.g., difficulty with urination) to symptoms of metastatic disease (e.g., bone pain). This is the way men have long been diagnosed if they didn’t get regular PSA testing.

    (2) Although the numbers of men on active surveillance who died of non-prostate cancer causes who were on active surveillance at the time of death is not given in the paper, clearly such information must be available to Dr. Hugosson and his colleagues — although I am not sure I understand why you think that this is of importance.

    (3) The number of men “over-diagnosed” is of much less importance than the number of men “over-treated.” The problem is that we treat men who don’t need treatment after we diagnose them (and/or they insist on being treated), not that we diagnose them in the first place.

  15. Errr … Is this so? The men who were diagnosed with prostate cancer in the control arm would almost certainly have been diagnosed as a consequence of symptoms of progressive prostate cancer,

    Reading some of the studies like Albertsen PC, Hanley JA, Gleason DF, Barry MJ. Competing risk analysis of men aged 55 to 74 years at diagnosis managed conservatively for clinically localized prostate cancer.JAMA 1998 Sep 16;280(11):975-80 I gained the impression that the majority of men diagnosed in the pre-PSA age were diagnosed in connection with a TURP (transurethral resection of the prostate), usually related to BPH (benign prostatic hyperplasia) rather than progressive prostate cancer. Hence the men diagnosed with Gleason scores as low as 2, 3 or 4, which was “low risk” disease with an estimated a 4% to 7% chance of dying from prostate cancer within 15 years of diagnosis depending on their age at diagnosis.

  16. Terry:

    You are correct … but you are referring to a selected group of low risk patients (as established in the tiotle of the paper you refer to). The majority of men diagnosed with prostate cancer in the pre-PSA era were not diagn osed with low risk disease. They were diagnosed with intermediate-risk, high-risk, and metastatic disease.

  17. Oh foolish me! I thought the fact that some of the men in the study I mentioned had Gleason scores as high as 8 to 10 and who were said to face a 60% to 87% chance of dying from prostate cancer within 15 years of diagnosis depending on their age at diagnosis indictated that the study referred to men with low risk and high risk disease, all diagnosed pre PSA.

    Is it not true (I seem to recall a study on this very site) that the number of men diagnosed with high-risk, metastatic disease has not altered signficantly in the post-PSA period — merely that they represent a smaller percentage of a very much high higher number of men diagnosed with low-risk at this time — the ‘over-diagnosed/over-treated men? Or is that another misconception?

  18. But Terry … That is exactly the point. The proportion of the men being diagnosed with more advanced forms of disease in the control arm of Goteborg study would have been far higher than the proportion in the tested arm for exactly the reason you mention.

  19. Thanks for doing such a great job in presenting this interesting paper! Your analysis and the follow-up exchange have brought forth a number of important points, and I would like to draw attention to several issues.

    I do think the relatively short follow-up, considering survival potential with current therapy options, is still substantially masking the advantage of screening, though that advantage is already emerging at 14 years since the start of the study and 6.7 years follow-up in the screening group since diagnosis with 4.3 years in the control group. That short follow-up would explain why the survival results for this study and the ERSPC study would be so similar at just 9 years. We need to bear in mind that, at the point 5 years after diagnosis, 99.4% of white, and 95.9% of black prostate cancer patients in the US are still alive, based on a look at the period from 1996-2004. (US Statistical Abstract of the US, 2009, Library Edition, Table 173, p. 118.) We need to be mindful that even high-risk patients (like me) are doing very well at the 10-year point, with 95% surviving. (Mayo Clinic, J Urol 179 page 1354, April 2008.)

    Such survival success does not leave much room to detect differences after just 14 years since the study began, especially with much shorter average follow-ups since diagnosis, the critical statistic. One expert medical oncologist specializing in prostate cancer, Dr. Charles “Snuffy” Myers, of course well-known to most of us, said he thinks we really need about 20 years of follow-up to get clear results for screening studies. Obviously, that is not even possible at this time, but we need to be careful not to overdrive our headlights here; we need that longer follow-up!

    To me the key action point here is to avoid projections of how many patients need to be screened or diagnosed to prevent a single death. If we want to consider such statements, we should at least recognize that they are, in all likelihood, lower bounds for the real impact. (In my opinion, the ERSPC really blew that call, and I suspect the New England Journal of Medicine is still smarting for not editing the ERSPC statement before publication.)

    I would also like to see more emphasis on benefits of early detection beyond just survival. The analysis mentioned the benefit of a likely lower proportion of metastatic disease in the screening arm, which is credible. However, as an eleventh year of a challenging case treated only with intermittent triple hormonal blockade, I’m very aware of other benefits, such as consequent early detection of low levels of vitamin D (me), earlier detection of decreased bone density (me), cardiovascular issues, etc. As Dr. Myers, likes to put it, a diagnosis of prostate cancer is like the canary dying in the mine: it’s a signal of other health issues.

    That leads to one point that’s puzzling: the equivalence of overall survival at this point. If fewer men are dying of prostate cancer in the screening group, that means that more men are dying of other causes in that group. Why? I can understand that minor rates of treatment-related mortality would affect the screening group at a slightly higher rate, but I’m still puzzled.

    The success rate for staying on active surveillance, 28% staying divided by 40% trying AS, or 70%, is in line with results from the major AS programs. That’s encouraging, especially in that AS programs continue to get even better. It looks to me like Sweden is doing a good job in moving away from over-treatment of prostate cancer.

    I’ll close with another view of the effectiveness of PSA: we should view it as a great multi-purpose screening tool for prostate health, not a mediocre tool for prostate cancer. The problem is in trying to fit a round peg into a square hole, in other words, wanting to make the PSA test, which is so good at indicating that something (BPH, infection, prostate cancer) is amiss with the prostate, highly sensitive and specific only for prostate cancer. Sure, I would like to see that in one test too, but I’m highly impressed that a combination of PSA, DRE, free PSA, PCA3, other staging technologies, and active surveillance do an already admirable job of separating the indolent from the stealthy aggressive cases.

    Again, thanks for presenting this study, the analysis, and the exchange.

  20. The report is extremely valuable, in particular with its age-based report, and it strongly highlights the benefits of PSA screening for the younger age cohort.

    Indeed, it shows that for the 50-54 cohort, the number of prostate cancer-related deaths among patients diagnosed with prostate cancer was 4% for the control arm compared with 0.6% for the screened group. In the 55-59 cohort the numbers were 14% versus 1.6% and in the 60-64 cohort the numbers were 13.5% versus 6%. If these numbers are representative of the population at large, they indicate a very significant reduction in prostate cancer mortality as result of PSA screening. (BTW for the total population the numbers are: 10.8% versus 2.5%.

  21. Reuven:

    This is all well and good … But the fact remains that the overall number of deaths in the screened group and the control group was exactly the same, which means that screening was having no effect at all on overall mortality at 14 years. What do you make of this?

  22. In looking at the numbers, 66.5% of the non-screened cohort received either primary treatment with intent to cure or hormonal suppression. The number for the screened cohort is 56.3%.

    In older men as represented by the men in this study, heart disease, other cancers, stroke, respiratory disease, accidental injuries, diabetes and influenza are well-known causes of death. How can screening and treatment of prostate cancer that reduced dying from prostate cancer by almost 50% improve overall survival in this particular study?

    Once diagnosed with prostate cancer and treated with intent to cure with either surgery or radiation how can these treatments, in older men improve overall survival? It seems like mission impossible to me …

  23. Reuven,

    You say … If these numbers are representative of the population at large,

    But they are not. The majority of men who die from prostate cancer are over the age of 80, not in their early 60s. A number of sources claim that modern medicine may be able to manage a disease like prostate cancer better now that was possible in the past, but it is also worth noting that the median age of disease-specific death hasn’t changed signficantly.

    When studies like this one only look at a certain age group, and one with a lower mortality rate at that, the results cannot simply be extrapolated into the general (and older) population in my oppinion.

    But then I am not a scientist or particularly statistically enabled, so I may be talking rot.

  24. Terry,

    I think the importance of PSA testing is saving the lives of relatively young and healthy individuals, who may otherwise die of prostate cancer or require complex, quality of life affecting, non-curative treatment that will only extend their lives by some years.

    The study demonstrates that for the 50-59 cohort, PSA testing helps reduce the mortality of prostate cancer patients significantly.

  25. Ralph,

    Treating prostate cancer obviously does not cure diabetes, heart failures, etc. Not doing the utmost to eradicate or at least treat prostate cancer is the equivalent of not eradicating tuberculosis. 100 years ago, when the life expectancy was roughly 40-45 years, there were very few people dying of prostate cancer. But that was because they didn’t reach the age to develop it.

  26. Sitemaster said:

    “Reuven:

    “This is all well and good … But the fact remains that the overall number of deaths in the screened group and the control group was exactly the same, which means that screening was having no effect at all on overall >mortality at 14 years. What do you make of this?”

    The total number of deaths is almost identical, but the number of deaths caused by prostate cancer is 78 in the control group and 27 in the attendees group (members of the screening group who have attended all PSA tests). This shows that PSA screening helps reduce the PrCa caused mortality.

    Indeed, people continue to die, but that’s something that will probably never be solved. :-)

  27. “CORRECTION” TO THE ERSPC STUDY: INCREASED SCREENING IMPACT

    I just learned of a paper out of the Erasmus Medical Center (by Dr. Fritz Schröder and team) that corrects for some of the flaws in the ERSPC screening study.

    Essentially, by correcting for “nonattendance” at screening in the screening arm of the study, and by correcting for other “contamination” in the control arm, the authors calculated that the impact of screening reduced the risk of dying from prostate cancer from 20%, as originally reported, to 31% in men actually screened. My impression is that “contamination” in the control arm refers to men in that arm who actually were screened despite being randomized to the non-screening arm; however, that is not clear from the abstract, which just refers to “contamination” in the control arm.

    I’m glad the leaders in the ERSPC are working to improve their findings. This may ease a concern some of us had that a desire to hold down health care costs by the European countries was subtly influencing the researchers to spin the results of their study (premature release, misleading, grossly premature calculation of numbers needed to treat to save a life, etc.).

    However, major problems with ERSPC results still exist. The major one to me is the much too short follow-up, which will give way to longer follow-up as the years pass.

    Another leading problem is that subjects were to be screened every 4 years, which is going to dilute the impact of screening compared to annual screening by letting some undetected cancers advance and others develop before the next scheduled test.

  28. (1) This type of post hoc analysis, which was not pre-specified by the study protocol, should be assessed and interpreted with great care. It is all too easy to use analyses like this to demonstrate what the researchers (and many patients) may want to believe.

    (2) What does this have to do with the Goteborg study referred to above?

  29. Re Sitemaster’s comment (preceding post) regarding “CORRECTION” TO THE ERSPC STUDY: INCREASED SCREENING IMPACT 8/17 6:34 PM

    Your point (1) about post hoc analysis certainly fits this paper, yet, in view of ethical difficulties in implementing randomized assignments (can’t order someone to either be screened, or to avoid screening and make the order stick), we will often in similar studies be stuck with either accepting non-compliance with the randomized role by some subjects, which means accepting a degree of confounding, or we will have to do some post hoc analysis, which likely will destroy the randomization aspect and statistics that are based on randomization. Is there a way out of this dilemma?

    Regarding (2) in your post, Dr. Neal of Cambridge, quoted in your initial article, compared the .56 relative risk in the Göteborg study with the .8 relative risk for the ERSPC trial. The “reanalysis” gets the ERSPC results, adjusted to a relative risk of .69, a lot closer to the .56 RR in the Göteborg study.

    By the way, the first author of the reanalysis is MJ Roobol and the last author is A Auvinen.

  30. Dear Jim:

    Unfortunately no. There is no way out of this dilemma. In the real world, people just don’t do what you want them to. And if there ever is a national guideline stating that every man between 45 and 75 should get an annual test of some type for prostate cancer, you can be absolutely sure that somewhere north of 25% of those men just won’t.

    With regard to the Goteborg data, some (but not all) of those data constitute a unique subset of the ERSPC data set, but you can’t compare the two. And Dr. Neal also pointed out in his editorial that the Goteborg data do not provide justification for recommending mass, population-based, prostate cancer screening — largely because the data only indicate the potential of regular testing over a period of time in a previously untested cohort of men.

  31. So we don’t change the death rate but we cut way down on prostate cancer deaths. Which means to me that many of the men who died of prostate cancer would have died about the same time of something else.

    If I were in charge of a national health system, given these results, there is no way I would approve PSA testing. If the total death rate does not change, it does not do much good, unless the the deaths of those dying of prostate cancer were much more unpleasant than the alternatives.

  32. Just wanted to get news about this topic, thanks.

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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: