Can we accurately project risk for over-diagnosis and over-treatment?


There is an interesting new article just published in the Journal of the National Cancer Institute describing a nomogram designed to estimate risk for over-diagnosis of men diagnosed with prostate cancer. The abstract of this article by Gulati et al. is freely available on line, but we have had the opportunity to review the full text of the article, as well as an accompanying editorial.

The authors state their purpose as follows:

… to use age, Gleason score, and prostate-specific antigen (PSA) level to help inform patients with screen-detected prostate cancers about the chances their cancers were overdiagnosed.

It is important to understand, in the context of this article, what the authors mean by an “over-diagnosed” case of prostate cancer. Specifically, they state

An overdiagnosed cancer is one that would never have become symptomatic or clinically apparent [during the patient’s lifetime] in the absence of screening.

The implication is that the risk of diagnosis of a prostate cancer of this type is associated with the sequential risk that interventional treatment of some type may be administered (as opposed to some form of very simple monitoring) for a cancer that never needed to be diagnosed in the first place. It does not mean that that diagnosis is inherently inappropriate. It is the potential for inappropriate intervention (including even aggressive forms of active surveillance) that is the problem.

Now this is a complex challenge. As Gulati et al. point out themselves, “whether a tumor has been overdiagnosed is not observable.” The full text of the paper offers a thorough discussion of how the authors went about developing the model on which their nomogram is based, and your sitemaster is certainly not competent to make judgments about the validity of the model or its projections.

According to Gulati and his colleagues their model “reasonably approximates age-, year-, stage-, and grade-specific incidence patterns, and projects that

  • 33 percent of men have prostate cancer onset in their lifetimes.
  • 38 percent of these men would be diagnosed in the absence of screening.
  • The average time from disease onset to diagnosis is 14 years.

They also note that these results are comparable to those projected by other models for the development of prostate cancer.

Other key points made in the paper, based on this model, are as follows:

  • Age is the single most important predictor of risk for over-diagnosis.
  • Among men with a Gleason score of ≤ 6 and a PSA level between 4.0 and 4.9 ng/ml, for example, the risk of over-diagnosis increases as follows:
    • 11.6 percent for men aged between 50 and 54 years
    • 59.9 percent for men aged between 70 and 74 years
    • 83.4 percent for men aged between 80 and 84 years
  • Lower PSA levels are associated with a substantially greater risk of over-diagnosis relative to higher levels (e.g., the predicted risk of over-diagnosis for men with PSA levels just greater than 4 ng/ml can be twice  as high as those for men with PSA levels close to 10 ng/ml).
  • For a given age and PSA level, the risk of over-diagnosis for men with a Gleason score of ≥ 7 is only slightly  lower compared to that for a man with a Gleason score of ≤ 6.
  • For patients with a PSA level ≥ 10 ng/ml, the risk of over-diagnosis for men with a Gleason score of ≥ 7 is 31.8 percent  lower compared to that for a man with a Gleason score of ≤ 6.

The actual nomogram developed by Gulati et al. is, as the authors intended, based on just three  predictor variables: age, Gleason score, and PSA level at diagnosis. (For those able to obtain a copy of the full text article, see Figure 1 and Table 3 of the actual paper, on page 4 of the article.)

Now one comes to the critical question, which is whether this nomogram is of sufficient accuracy for actual use in clinical practice today. Here it appears that there are some differences of opinion about methodology (that we are certainly not competent to comment on, beyond acknowledging their existence).

In a media release about the study issued by the Fred Hutchinson Cancer Research Center, Dr. Daniel Lin, a urologic oncologist, is quoted as follows:

This nomogram will provide clinically relevant information for both the provider and patient on which to base practical decision-making for newly diagnosed prostate cancer. We fully realize that many men do not need immediate treatment for their cancer and can be safely placed on an active surveillance protocol, and this important research will further clarify those who may avoid treatment versus those men who may benefit from attempted curative therapy.

However, the editorial by Freidlin and Korn in the same issue of the Journal of the National Cancer Institute, is entitled “A model too far”. In their editorial, Freidlin and Korn express concerns about

  • The methods used by Gulati et al. to derive the estimates on which their nomogram is based, and
  • The clinical relevance to a prostate cancer patient of knowing his chance of over-diagnosis when making treatment decisions

The “New” Prostate Cancer InfoLink is of the opinion that, for men diagnosed with low-risk prostate cancer in particular, especially as patients’ ages increase, the clinical relevance to patients of understanding risk of over-diagnosis, and the value of avoiding over-treatment, is actually extremely high … if the tools available to project that risk are sufficiently accurate.

The questions that The “New” Prostate Cancer InfoLink would raise are these:

  • Is the nomogram developed by Gulati et al. testable in a prospective manner?
  • Are there existing data sets of any type against which to validate their model?
  • Is the nomogram sufficiently granular? For example, in terms of Gleason scores, it groups patients into just two sets: those with a Gleason score of 7 and higher and those with a Gleason score of 6 and lower. Is that really good enough? It is certainly arguable that many older men diagnosed with small amounts of Gleason 3 + 4 = 7 disease also also being over-treated.
  • Are the data good enough to actually limit the need for even the commonly used forms of active surveillance to only a subset of the diagnosed? (We do know that repetitive prostate biopsies, in and of themselves, come with their own risks.)
  • Do Dr. Gulati and his colleagues intend to make this nomogram available on line in the same way as the PCLO prostate cancer risk calculator, the Kattan nomograms, and the Partin tables are available on line?

There seems little doubt to The “New” Prostate Cancer InfoLink that a nomogram that can accurately project risk for over-treatment (based on risk of over-daignosis) is of great value … but is this that nomogram? It is worth reminding readers about the evolving Cancer Survival Query System (CSQS) that we have mentioned previously, and that is being worked on by staff at the National Institutes of Health and other partners. The CSQS will also be highly relevant to the whole “over-diagnosis, over-treatment” problem, particularly among older, sicker, and otherwise low-risk patients.

We see the proposed nomogram developed by Gulati et al. as a step toward a really valuable set of accurate prognostic tools for newly diagnosed prostate cancer patients. How big and how safe that step is may need a little more supporting information.

Editorial note: The “New” Prostate Cancer InfoLink would like to thank Dr. Roman Gulati for providing us promptly with access to a full text copy of the primary article for our review, as well as to the accompanying editorial by Freidlin and Korn.

4 Responses

  1. In the environment you have taken care to foster of presenting useful and timely information in a careful, fair, well-reasoned, unbiased way, this article stands out particularly as exemplifying all those qualities.

    Very thoughtfully written — thank you.

  2. I am an 89-year-old prostate cancer patient, diagnosed by biopsy at age 87 (PSA 14 ng/ml; Gleason score 8-10; clinical stage T2NxM1).

    I am currently being treated with Zytiga (and have been for the past 14 months). However, my PSA is rising slowly. My initial bone scan (from 2 years ago) showed metastasis to the thoracic spinal region with significant involvement of T8 going into the anterior LIP and multiple other lesions noted. I have not experienced any pain to date.

    It has been suggested that I have radium-223 administered, but I am reluctant to start this because of quality of life concerns and because this is a very expensive program over 6 months. My feeling is that I should refuse this procedure (at least for the time being). The literature for radium-223 indicates that it is for treatment of pain. I have no pain.

    I am getting another bone scan this week. My health is otherwise good for an 89-year-old.

    Any thoughts?

  3. Sidney:

    (1) You are correct that the primary indication for radium-223 is for the treatment of pain associated with metastatic prostate cancer.

    (2) You may have more insight into what to do when you get the results of the new bone scan. This will show you to what extent your cancer may have been put into remission by the prior treatments.

    (3) A key question is probably what your PSA level is now and how fast it is rising (the PSA doubling time). Do you know these?

  4. It is an interesting intellectual enterprise to consider the possibility of creating a meaningful predictor out of three easily accessibly, important-to-diagnosis, (relatively) inexpensive variables in assessment of prostate cancer.

    However, there’s no way around the reality that PSA is not a measure specific to prostate cancer and Gleason scores are subjective interpretations that are most subjective on the “benign” end of the interpretation scale. I remain unconvinced that age is reliably a truly independent predictor of risk. While it seems pretty clear that younger men with higher Gleason scores have a high need for intervention, it’s not as clear that younger men with low Gleason scores (6 and lower) and low PSAs are not systematically subjected to over-treatment.

    The PSA is a problem — a huge, huge problem. And Gleason scores need to stop being inflated to “6” when analysis suggests a lower score, indicating less disease and less dangerous disease is what is present. When we can have an honest discussion about the quality of measurement currently available and used, is when we’ll get somewhere in regard to over-diagnosis, and thus, treatment. And not until.

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: