The Norrkoping prostate cancer screening trial — with 20-year follow-up

A new report just published as an open access, full-text article at is going to further confuse the issue of whether mass, population-based screening for prostate cancer is or isn’t a good idea.

The report by Sandblom and his colleagues provides data from a study of all 9,026 men in the Swedish city of Norrkoping who were between the ages of 50 and 69 in 1987.

Of these 9,026 men, 1,494 were initially screened for prostate cancer with a DRE exam only in 1987 and again in 1990. In 1993, they also received a PSA test as well as the DRE (with 4.0 ng/ml as the lower “cut-off” value for a positive PSA result, which we now know to be invalid). And then in 1996 only the men who were still 69 years of age or less received a PSA and a DRE. The remaining 7,532 men did not undergo screening and therefore served as a comparison group.

Now it will immediately be evident that there are problems with this study because some men may only have ever received one PSA test and no one will ever have received more than  two PSA tests within the structure of the clinical trial. On the other hand, this study is the first to ever provide prospective information on a population of patients who were followed for 20 years.

Data on diagnosis, tumor stage, grade, and treatment were obtained from the Swedish South East Region Prostate Cancer Register, and patients were carefully monitored for prostate cancer-specific mortality through 31 December 2008.

Here are the key data reported by Sandblom et al.:

  • Attendance at screenings between 1987 and 1996 was generally good:
    • 1,161/1,492 (78 percent) in 1987
    • 957/1363 (70 percent) in 1990
    •  895/1,210 (74 percent) in 1993
    • 446/606 (74 percent) in 1996
  • 85 cases of prostate cancer were diagnosed in the screened group (85/1,494 or 5.7 percent).
    • 43 tumors were found at a screening examination and 42 in the intervals between examinations.
  • 292 cases of prostate cancer were diagnosed in the control group (292/7,532 or 3.9 percent).
  • The percentage of men with localized tumors (T1-2, N0/Nx, M0) was significantly higher in the screened group (56.5 percent) than in the control group (26.7 percent, P<0.001).
  • The rates of non-localized tumors were 37/1,494 (2.5 percent) in the screened group and 213/7,532 (2.8 percent) in the control group, which was not a statistically significant difference.
  • The risk ratio for death from prostate cancer in the screened group was 1.16.
  • In a Cox proportional hazard analysis comparing prostate cancer-specific survival in the control group with that in the screened group, the hazard ratio for death from prostate cancer was 1.23.
  • After adjustment for age at start of the study, the hazard ratio was 1.58.

The authors conclude that, “After 20 years of follow-up the rate of death from prostate cancer did not differ significantly between men in the screening group and those in the control group.”

Dr. Sandblom has been quoted as stating that, “In the light of our findings, I would say that the benefit from screening is not sufficient to support mass screening,” but he continued as follows: “However, the study was initiated more than 20 years ago, when PSA was not available and the treatment of localized prostate cancer was not as effective as it is today. I would thus not categorically advise against PSA testing based on an individual decision from a man who feels concern about prostate cancer.”

Where does this new study seem to leave us? Probably right where we started: PSA testing should be an individual decision between each man and his doctor, based on a careful assessment of the individual patient’s known risk factors. This does not mean that every patient needs to be able to “justify” PSA testing because he has specific risk factors. Rather, it means that there are better reasons to recommend PSA testing for some men than for others.

It is (frankly) unclear whether any trial currently ongoing will be able to give us a concrete set of data about the value of screening for prostate cancer using the PSA and DRE with a 20-year follow-up. The PLCO trial and the ERSCP were both horribly flawed trials in different ways, and we know of no other large, ongoing trial. The “best” large study to date — from a methodologic point of view — has been the Göteborg trial. This trial did show a prostate cancer-specific survival benefit in a previously unscreened population (but no overall survival benefit).

14 Responses

  1. Once again what other choice do we have to test for prostate cancer? People who are worried about it are going to try something. Its all we’ve got. What do we do? Go back to the days of DRE only?

  2. Dear Chris:

    I don’t know anyone who is suggesting that people shouldn’t use the PSA test. The issue, as far as I — and I believe others — are concerned, is only about how to manage expectations from the use of that test appropriately.

  3. I am amazed by the “art of writing and interpreting” demonstrated in the article. I finally got to read it and here is what I found.

    Title: “Limited Benefit Seen in 20-Year Prostate Cancer Screening Study”

    First paragraph: “Prostate cancer screening in the general population is likely to reduce disease-specific deaths by less than a third, according to results from a 20-year Swedish study.”

    Results: “Disease-specific mortality was 35% for men diagnosed in the screening group and 45% for controls.”

    Conclusion: “… screening as practiced in the present study is unlikely to reduce mortality from prostate cancer by more than a third.”

    It seems the way one presents the results is all that matters. I don’t know about others, but a one-third reduction in prostate cancer mortality sounds like a good outcome. Isn’t that so?

  4. Dear Reuven: To which “article” are you referring? The quotations you are giving do not appear in the article by Sandblom and his colleagues. They are careful to point out that the 23% increase in the hazard ratio was not statistically significant.

  5. I have not read the article BUT I see a difference of 5.7% v 3.9% in diagnosis. By my reckoning that shows a 50% improvement in Dx from PSA testing … Seems to me that is significant and a source of social and economic cost savings.

    50% more men are treated for the disease and statistically we know 10% to 15% of the additionally diagnosed men will have advanced disease.


  6. Mike:

    Reuven’s comments are on the study in question. They appear in the full version of the Sandblom reference. The reported mortality reduction is 22.2%. Given the fact that this is ANOTHER study using “intent to screen,” the improvement in mortality is diluted. The mortality reduction in those actually screened is more like 37.8%. Man, they never give up ….

  7. Dear Ralph:

    Reuven wrote that the title of the article he was reading was, “Limited benefit seen in 20-year prostate cancer screening study.” The actual title of the Sandblom article (as clearly given in the link at the beginning of the commentary above) is “Randomised prostate cancer screening trial: 20 year follow-up.”

    Reuven wrote that the conclusion of whatever he was reading stated that, “… screening as practiced in the present study is unlikely to reduce mortality from prostate cancer by more than a third.” However, the actual conclusion of the full text of the Sandblom et al. article clearly reads as follows:

    “The risk ratio for prostate cancer specific death did not indicate significant benefit from prostate cancer screening. Although the population size in our study is not sufficient to draw definite conclusions, the power is sufficient to show major differences in prostate cancer specific survival.”

    The other two pieces of text that Reuven quotes also do not appear in the actual article. I have checked this very carefully.

    I have no idea what article you guys are reading, but it most certainly is not the full text of the actual article by Sandblom et al. that is published in the BMJ. It looks and sounds very much like another person’s opinion on that article, and not a very accurate one either.

  8. Dear Ralph:

    The title of the article, and the quotations, that Reuven refers to appear to come from a news report by Jennie Smith that originated from Elsevier Gobal Medical News. That heading and the content of that news report appear all over the Web. It is not an accurate commentary on the original, full-text article at all.

    The actual conclusion of the full text of the original article by Sandblom et al. very carefully and accurately states:

    “The risk ratio for prostate cancer specific death did not indicate significant benefit from prostate cancer screening. Although the population size in our study is not sufficient to draw definite conclusions, the power is sufficient to show major differences in prostate cancer specific survival.”

    Please read that conclusion with care. It is an utterly different conclusion that the one the media are suggesting.

    It is for exactly that reason that I started the initial report above by stating that this paper “is going to further confuse the issue” still further than it already is confused. Guess what … Apparently I was correct!

  9. Mike,
    I saw that Reuven quoted the 35% vs 45% specific mortality that appear on the 3rd page of the full reference. That represents a 22.2% reduction in mortality. In analyzing the data, they used “intent to screen” to dilute the result. Otherwise why would they try to find the result of screening on PCa specific mortality and then include men who were never screened?

    In the conclusion, they ignore the 22.2% improvement they reported and state that even small in size the study has power enough to show no benefit. Why not report results on only those that were actually screened? That would yield results similar to the Goteborg trial…and the ERSPC trial when one takes away the “intent to screen” confounding maneuver.

    Were they trying to confuse the issue?. Yes, they were by reacting to oppose the results of the Goteborg trial.

  10. Ralph:

    I am sorry. I disagree with you. All that the authors are saying is that on a statistical basis, from the trial that was designed as an “intent to screen” trial, you cannot make an absolute determination that mass, population-based screening is associated with a survival benefit. Most of that is a consequence of how the trial was structured over time, but one cannot “go backwards” in time to make determinations about what trials prove and don’t prove. It is certainly possible to analyze the data in other ways, but that would not consistute level 1 evidence.

    We have huge quantities of evidence from over the years about how re-analyses of data have been inappropriately used to “prove” things that were subsequently shown to be untrue.

    I do not believe that Sandblom et al. set out to “oppose the results of the Gorteborg trial.” They report the results of this specific trial. The interpretation of what these different trials mean in relation to each other is a whole other issue, in which the mindsets of the commentators do indeed come into play, as you yourself are demonstrating. Sandblom et al. acknowledge that they might have got different result if they had been giving PSA tests to all the patients every 3 years … but they couldn’t have done that. No one in the trial got a PSA test for the first 6 years of the study because there was no PSA test!

    My point was only that Reuven was not quoting from the original study. He was quoting from an interpretation of the original study. It is inappropriate to “put words into the mouths of others” when they didn’t actually state them or to take their words out of context. I can assure you that the people who put together the ERSCP trial were very definitely of the original opinion that it would prove the value of screening. If it hadn’t been designed as an “intent to screen” study it wouldn’t have been a screening study. It would have been a “case-finding” study. I have no doubt at all that an appropriately structured “case-finding” trial could be associated with a survival benefit, but that’s the point … In whom is it most appropriate to attempt to actually find prostate cancer? We don’t yet know!

  11. Dear Sitemaster,

    I taught logic for many years, and am a pretty clear and accurate thinker and reader. That last paragraph is simply incoherent. Not quite logically inconsistent, but somehow conceptually. Briefly, it begins with a negative conclusion and seems to end with a strong positive conclusion about the same data. I wonder what “power” means in this text.

  12. Dear George:

    “Power” (as used by Sandblom et al. in their conclusion to their article, which is the “last paragraph” to which I think you are referring) is a reference to the statistical strength of a data set to allow a definitive conclusion within the constraints of a study’s design.

    For the “power” of this particular study, I refer you to the third paragraph of the section on “Statistics” in the orginal article.

  13. Right. That’s what I thought; that the coherence of this paragraph at least partly depends on the present meaning and use of the word “power.” That’s why I qualified my statement by using the word “seems.” I’ll do my best to look at the full text, although I’m not much good at statistical reasoning.

  14. So often research is overtaken by events, particularly by advances beyond the protocol in the studies. That seems to be the case with this new “20 year” research on “screening.”

    Sitemaster’s article and the exhange of responses have already made the key points, but here’s another way of looking at the lack of PSA testing for the first 6 of the 20 years, and reduced PSA testing thereafter, mainly due to age exclusions. In effect, the lack of PSA testing for 6 of 20 years, or 30% of the period with reduced testing afterward for some patients, was like a rate of non-compliance with true (though spotty) screening of at least 30%. That’s quite substantial in my view.

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 )

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: