15-year metastasis-free survival in men on active surveillance in The Netherlands

According to new data reported at the annual meeting of the European Association of Urology in London, active surveillance of men diagnosed with low-risk prostate cancer was not associated with an elevated risk for metastatic disease at 15 years of follow-up.

A report in Renal & Urology News discusses this poster (no. 804) presented by Veerbeck et al. (a group of Dutch researchers).

The research is based on data from the actual treatment of 900 men who were initially diagnosed with localized prostate cancer during the first and second rounds of screening in the Dutch cohort of patients enrolled in the European Randomised Study of Screening for Prostate Cancer (ERSPC). All 900 patients were considered to be low risk and potentially appropriate for management on active surveillance (with a Gleason score of ≤ 6 and a clinical stage of ≤ T2a), but only a quarter of them actually agreed to be managed on active surveillance.

We should be clear up front that these data do not come from a randomized, controlled trial, but the data are none the less of interest.

Here is what Verbeek et al. reported in London:

  • Of the 900 eligible patients,
    • 223 (24.8 percent) were initially placed on active surveillance.
    • 312 (34.7 percent) elected immediate radiation therapy.
    • 365 (40.6 percent) elected immediate radical prostatectomy.
  • The 15-year rates of metastasis-free survival were
    • 96.9 percent for the men initially placed on active surveillance.
    • 96.6 percent for the men treated with immediate radiation therapy.
    • 97.7 percent for the men treated with immediate radical prostatectomy.
  • The 15-year rates of prostate cancer-specific survival were
    • 97.2 percent for the men initially placed on active surveillance.
    • 97.5 percent for the men treated with immediate radiation therapy.
    • 98.5 percent for the men treated with immediate radical prostatectomy.
  • There were no statistically significant differences between the groups of patients with regard to either metastasis-free or prostate cancer-specific survival.

These outcomes data for the men on active surveillance at 15 years of follow-up are therefore “better” than the data reported last year from the ProtecT trial in the UK at 10 years of follow-up.

The senior author of this poster, Dr. Monique Roobol, explained the possible reasons for the differences between these data and the data from the ProtecT trial as follows:

  • In the ProtecT trial, the active surveillance group included not only patients with low-risk prostate cancer, but also patients with intermediate- and high-risk prostate cancer, who would most likely not be considered for active surveillance in clinical practice.
  • The active monitoring algorithm used in the ProtecT trial “is in no way comparable with contemporary AS protocols, where at regular intervals rectal exams and biopsies are used to note potential progression of disease.”

We have previously discussed the fact that the characteristics of patients enrolled into the active surveillance arm of the ProtecT trial bear only partial resemblance to the patients considered to be appropriate for active surveillance today. We have also addressed the unusual design of the ProtecT trial  itself. What is interesting about the ProtecT trial is that after 10 years, in a large, “partially randomized”, clinical trial, there was no difference in the 10-year prostate cancer-specific survival of the patients and only a relatively small difference in the 10-year metastasis-free survival.

It is arguable that even though the patients in the Dutch arm of the ERSPC trial were not randomized to different types of treatment, they may actually be more representative of a group of patients who were all really appropriate for active surveillance and were all more appropriately monitored for risk of progression. However, we will need to see more data from this study before that type of determination could reasonably be made.

8 Responses

  1. Either the data reporting is incorrect; i.e., that the data are reported to three significant figures, or the conclusion is; i.e., that “There were no statistically significant differences between the groups of patients with regard to either metastasis-free or prostate cancer-specific survival.”

    The data show that, of every 1000 patients, 13 more of those who are treated by RP survive at 15 years, and 11 more experience metastasis-free survival at 15 years. That is significant!

  2. Dear John:

    With respect, you can’t simply multiply the study numbers in each treatment group by three or four and then claim that the results are significant. That’s just statistically inappropriate. The statistical level of significance can only be based on the actual data not on a projection.

    And it also seems highly unlikely to me that someone like Dr. Roobol would have allowed a report like this to occur unless she was extremely confident about the data. She is a highly trained medical statistician and epidemiologist!

  3. > you can’t simply multiply the study numbers in each treatment group
    > by three or four

    Mike, who multiplied the numbers by three or four? I sure didn’t!

    I based my comments on the actual numbers in your article. The uncertainty in your numbers is on the order 0.1. So a difference of 1.3 between two numbers is significant when the uncertainty in each number is ~0.1.

  4. Dear John:

    You changed the baseline to numbers of patients per 1,000 treated by each form of therapy and then you claimed there was a statistical difference. You can’t do that. You can only project data out to larger numbers of patients if there was a clear statistical difference based on the actual numbers, and there isn’t.

    If you want to clarify this I suggest you take it up with Dr. Roobol and her colleagues. I am just the messenger.

  5. “Statistical significance”

    John, following up on Sitemaster’s comments, “statistical significance” is a term with specific meaning in the field of statistics. In basic terms, it has to do with the numbers of research participants involved, their scores on the measure of interest, and deviations of their scores from measures of how far each was from the norm, which can be as simple as an average or more complex as nearness to a “best fit” line.” The projection of events per 1,000 person years can have varying degrees of statistical confidence, depending on the above factors. For instance, a study of 100 men that projects to 1,000 person years would for the same study protocol would have a lot lower confidence that a study of 10,000 men that projected to 1,000 person years, per my somewhat rusty resurrection of once fluent statistical know-how.

    Wikipedia has an article about statistical significance, but it may be daunting if you have no background in statistics.

  6. Thanks for reporting these compelling figures!

    I have made up a poster of an expanded version of the article for our Us Too chapter’s active surveillance resources, using yellow highlight on the key results.

  7. Jim, thanks. I do understand (I hold a PhD in physics). My recollection of statistics is also rusty but, I do know that statistical significance is related to the uncertainty in the data vs. the size of the effect, Δx/x. If Δx is comparable to x, then the data are noisy, probably not significant. When Δx is much less than x, then the data is probably significant.

    According to the numbers Mike quoted, Δx is less than 1/10 of x, so, the change between the two groups is significant. (Dr. Chodak may be a great doctor but he doesn’t know his statistics.)

  8. Statistics again, and decision making

    Hi ieeesensors,

    What is lacking in the surface numbers in the article is the underlying data foundation. Without that, it is impossible to do your Δx/x check for “statistical significance”, a term which has a specific meaning in the field of statistics for the particular kind of statistic being employed, which is not described in the report. Dr. Chodak is not at all involved here. The lead researcher is Dr. Monique J. Roobol, PhD, MSc, with both of these degrees in epidemiology, which is heavily oriented toward statistics. No doubt she is the one who did the calculations for statistical significance.

    I suppose the root statistical question here is whether the differences in metastatic freedom and disease survival are real, or due to chance. For illustration, what if a different group of 900 patients had been studied: is there enough of a basis to conclude that you would again see differences among the groups that are the same as in this study, which is something statistical techniques help estimate; if there were enough data with results in the necessary directions, then the results would be “statistically significant.” In this case, with the differences so small and the groups rather large, the result of “no significant difference” is not surprising.

    However, even if these extremely small differences were significant statistically, the key question for patienst and for doctors should be: “Is the difference in benefit — about a 1% advantage in not having metastases or dying of prostate cancer — worth a fairly strong likelihood of having some bothersome side effects for the rest of your life?” I have some side effects from my treatments. They are quite mild and tolerable — a very small price to pay for success against a once life-threatening case, but if I had had a low-risk case and was choosing what to do aided by the statistics in this study, I really doubt I would choose immediate treatment instead of active surveillance. Also these days active surveillance is managed even better than it was back in the days of this trial, with continued improvement likely (genetics, imaging, supportive medications like metformin as a possibility, etc.), so it is likely that as time goes on, freedom from metastases and survival at 15 years with active surveillance will be even closer to 100% than they were in this trial.

    It is a little surprising to see in this thread the instances — probably the tip of the iceberg — of resistance to the strikingly clear bottom line: little difference between active surveillance and immediate treatment for these patients at the 15-year point, with quite awesome success for all three groups. Percentages are statistics, in fact simple ones, and sometimes, as in this instance, the numbers speak for themselves because they are so clear. Yet it is also clear that some of us still do not see that. I’m thinking it’s because often “believing is seeing,” and, if you want to believe that active surveillance is not worthy, then any possible rebuttal will be attractive.

    I hope anyone reading this far will consider this: the number of patients in this research — 900 — is quite large for a group that has 15-year follow-up (remarkably long) for prostate cancer research. It is also worth noting that this group was well-vetted at the start, as it was part of arguably the best managed national group in the European Randomized Study of Screening for Prostate Cancer. That’s important because it means you were not getting an unknown but possibly wide-ranging group of patients in the trial as far as the key variable of risk for prostate cancer was concerned.

    With this in mind, it amazes me that the metastasis-free and cancer-specific survival rates for all three groups were within a percent and a fraction of each other, at 15 years from the start! How wonderful!

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 )

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: