Retrospective cohort analysis shows mortality benefit in prostate cancer (but there’s a big “however”)


A new report in European Urology may offer some insights into the value of large, retrospective data analyses as compared to multi-center, randomized clinical trials when it comes to the assessment of the effects of different first-line treatments on the management of clinically localized prostate cancer.

This new study by Abdollah et al. was designed to evaluate the effects of radical prostatectomy and “observation” (i.e., watchful waiting) on the prostate cancer-specific mortality and the “other cause” mortality of US Medicare patients (i.e., men of 65 years and older diagnosed with T1 or T2 prostate cancer). The patients were all identified by use of the the Surveillance Epidemiology and End Results Medicare-linked database and had been treated between 1992 and 2005.

Here is what Abdollah and his colleagues showed:

  • The database included 44,694 men aged ≥ 65 years and diagnosed with localized prostate cancer.
    • 22,244 patients (49.8 percent) were treated with radical prostatectomy (RP).
    • 22,450 patients (50.2 percent) were managed by observation.
  • It was possible to categorize 23,338 patients into 11,669 matched pairs.
  • These 11,669 matched pairs were randomly divided into two cohorts
    • A “development” cohort (used to develop a predictive nomogram of the potential benefit of RP compared to observation)
    • A “validation” cohort (used to assess the accuracy of this nomogram in an independent group of patients)
  • In the development cohort
    • The 10-year rate of prostate cancer-specific mortality was estimated to be 2.8 percent for men treated by RP.
    • The 10-year rate of prostate cancer-specific mortality was estimated to be 5.8 percent for men managed by observation.
    • The absolute rate of risk reduction (RR) for men treated by RP compared to observation was 3.0 percent.
    • The relative rate of RR for men treated by RP compared to observation was about half (51.7 percent).
  • In the validation cohort
    • Prostate cancer-specific survival could be predicted with 73 percent accuracy.
    • Other cause mortality could be predicted with 69 percent accuracy.

Abdollah et al. note that, “The nature of observational data may have introduced a selection bias” in the conduct of their analysis, but they conclude that, “On average RP reduces the risk of [prostate cancer-specific survival] by half” in this cohort of patients relative to observation and that  “The individualized protective effect of RP relative to observation may be quantified with our nomogram.”

However (and it is a big “however”), the data from the PIVOT study, reported earlier this year at the annual meeting of the American Urological Association, has clearly shown us that:

  • “Surgery did not reduce mortality more than observation in men with low PSA or low-risk prostate cancer” and
  • Compared to observation, surgery offered “reductions in all-cause and prostate cancer mortality that were not significant and less than 3 percent in absolute terms over 12 years.”

Men enrolled into PIVOT were all 75 years of age or younger. We have not yet seen a detailed publication giving the average age of the patients in the PIVOT trial at the time of treatment, but we would expect this to be of the order of 65 years of age.

In other words, the only large, randomized clinical trial completed to date in a comparable population of US males diagnosed with localized prostate cancer showed nothing like a 50 percent reduction in risk for prostate cancer-specific mortality between men managed by radical prostatectomy compared to observation.

How does one account for such a discrepancy? Simple. As Abdollah and his colleagues carefully noted, “The nature of observational data may have introduced a selection bias” in the conduct of their analysis.

Retrospective data analysis of large cohorts can introduce all sorts of issues that can profoundly affect the results of such studies. This is why The “New” Prostate Cancer InfoLink is always so careful to indicate that data from such studies must be interpreted with great caution. It commonly turns out that the results of such studies are completely contradicted by studies (such as randomized, multi-center, controlled clinical trials) that meet category 1 levels of evidence. In this particular case, it is also worth noting that the Scandinavian randomized, multi-center study (albeit in a somewhat more advanced group of patients diagnosed with clinically evident prostate cancer) has also shown no significant difference in prostate cancer-specific mortality between immediate radical prostatectomy and watchful waiting after 12 years of follow-up in men of 65 years and older at the time of diagnosis and randomization.

5 Responses

  1. I love it. The more I read about the great debates on screening and, in this particular post, on treatment vs. active surveillance, the more I get convinced that one can always find a study to justify one point of view.

  2. Reuven:

    Tell me a point of view you want to justify and I am near to 99.99% certain that I will be able to find at least two studies that can be used to support it. (One just has to ignore the data that don’t!)

    :O)

  3. I don’t care if RP only offers a one half of 1 percent reduction in death from prostate cancer. I bet it mattered to the 0.5% who died. I’ll take every percentage point that I can get to reduce my chances of dying from prostate cancer. To me there is no such thing as “statistically insignificant.”

  4. Mike,

    You said: “In other words, the only large, randomized clinical trial completed to date in a comparable population of US males diagnosed with localized prostate cancer showed nothing like a 50 percent reduction in risk for prostate cancer-specific mortality between men managed by radical prostatectomy compared to observation.”

    Randomized yes, large no. Is a population of 713 large? The PIVOT trial went on for years and it could never recruit enough men to agree to be randomized. At the end of 12 years it showed a benefit for RP, but only when they separated high risk and low risk. Are these two trials comparable? I don’t think so …

  5. Ralph:

    “Large” is a relative term. Anything over 500 patients is generally considered “large” when one considers randomized clinical trials. And it wasn’t the fault of the PIVOT trial coordinators that they couldn’t randomize men to this trial. It was a consequence of human nature. The other study was not a trial at all. It was a retrospective data analysis with literally dozens of inbuilt assumptions. Such studies always make me suspicious when they claim statistically valid survival benefits.

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: