Prediction of the “real” risk of death from prostate cancer

A while ago now we first discussed research being done by Eric Feuer and others on what was initially being called the Cancer Survival Query System or CSQS. The goal was to develop a relatively simple, nomogram-based system that would allow doctors and their patients to predict risk of death from cancer as opposed to risk of death from other causes, which could be individualized based on a variety of factors that included things as diverse as the patient’s age, race, diagnosis, treatment, and other co-morbid conditions.

Such a system would help men to understand whether their risk of dying from prostate cancer specifically was actually increasing their risk of death from all causes or whether, even though they had been diagnosed with prostate cancer, their risk of death from all causes was unlikely to be affected by a diagnosis of prostate cancer. Such knowledge would, potentially, be of great value in helping a man to decide whether he could just monitor a low- or intermediate-risk form of prostate cancer or whether he really needed to get it treated.

There has now been significant progress in this research field, and last November there were three relevant papers published in the Journal of the National Cancer Institute Monograph series:

  • A paper by Feuer et al., entitled “The Surveillance, Epidemiology, and End Results Cancer Survival Calculator SEER*CSC: validation in a managed care setting”
  • A paper by Rabin et al., entitled “Health-care utilization by prognosis profile in a managed care setting: using the Surveillance, Epidemiology and End Results Cancer Survival Calculator SEER*CSC”
  • A paper by Howlader et al., entitled “Providing clinicians and patients with actual prognosis: cancer in the context of competing causes of death”

In all three papers, prostate cancer and the ability to project risk of prostate cancer-specific mortality was specifically addressed. However, the methodologies used by the SEER*CSC team were different to those used by Howlander et al. Also, all three papers are complex because of the nature of the way the projections are constructed, and it won’t be possible, here, to get into all of the details. However, the authors have been kind enough to provide your sitemaster with full text copies of all three of these papers.

Let’s deal first with the paper by Feuer et al.

What Feuer and his colleagues set out to do was to validate a refined model of the original CSQS — now known as the SEER*CSC or the SEER Cancer Survival Calculator.  This calculator allows us to make a variety of predications about a specific patient and then compare them to one another. The predictions include all of the following:

  • A man’s “physiological” or “health status-adjusted” age or HSAA (as compared to his calendar age) taking account of his co-morbid conditions other than his cancer (e.g., whether he has had things like a prior heart attack or ongoing liver disease)
  • The probability that a man with a calendar age of 66 to 94 years and diagnosed with prostate cancer will die from his cancer, or from other causes, or still be alive at
    • 1 year after his diagnosis
    • 5 years after his diagnosis
    • 10 years after his diagnosis

Using this SEER*CSC model and data from prostate cancer patients diagnosed between 2001 and 2008 in the Kaiser Permanente Colorado managed care system, Feuer et al. were then able to demonstrate that their model actually predicted the above-mentioned risks with a relatively high degree of accuracy.

Specifically, they give the example of a married, 69-year-old, African American male, aged 69 at diagnosis with prostate cancer. He has two co-morbid conditions (cerebrovascular disease and renal disease). His HSAA is evaluated as 82 years (as compared to his calendar age of 69 years). His life expectancy without his prostate cancer is estimated at 7 years. However, he has high-risk prostate cancer and is treated by radical prostatectomy. After the surgery, he is confirmed to have a Gleason score of 8 to 10 and a pathological stage of pT2-3N1M0. Taking all of these data into account, the SEER*CSC calulator projects the following:

  • At 1 year post-diagnosis
    • 2 percent of men like this will likely have died of their prostate cancer
    • 9 percent of men like this will likely have died of other causes
    • 89 percent of men like this will still be alive
  • At 5 years post-diagnosis
    • 9 percent of men like this will likely have died of their prostate cancer
    • 41 percent of men like this will likely have died of other causes
    • 50 percent of men like this will still be alive
  • At 10 years post-diagnosis
    • 18 percent of men like this will likely have died of their prostate cancer
    • 65 percent of men like this will likely have died of other causes
    • 17 percent of men like this will still be alive

Their model does not work so well in men who are younger than 65 years of age, and so the model is probably going to need some refinement before it could be used reliably in the prognosis of men diagnosed with prostate cancer in their 50s and early 60s.

The second paper, by Rabin et al., used the SEER*CSC with somewhat different intent — to see how patient prognosis might influence patterns of health service utilization, again in cohort of patients enrolled in the Kaiser Permanente Colorado managed care system.

Looking specifically at the findings for the cohort of 1,102 prostate cancer patients — all diagnosed between 2001 and 2008, Rabin et al. found that:

  • Their average age was 67 ± 8.8 years.
  • They were followed for an average (mean) of 71 ± 35.2 months.
  • 259/1,102 (24 percent) died within the study period (of any cause)
  • 908/1,102 (82 percent) were diagnosed with AJCC Stage I disease.
  • 100/1,102 (9 percent) were diagnosed with AJCC Stage IV disease.
  • The SEER*CSC projected that
    • 53 percent of these men would probably die of causes other than prostate cancer within 5 years of diagnosis (Group A).
    • 40 percent of these men had a roughly equal probability of dying from their cancer or from other causes within 5 years (Group B).
    • 22 percent were significantly more likely to die of their cancer within 5 years (Group C).

However, what was really interesting was the following:

  • The men in Groups A, B, and C all had similar rates of urology visits within the first year after diagnosis (about 11 or 12 visits per 1,000 person-days).
  • The men in Group C (as one might expect) had far more medical oncology visits (9.1 per 1,000 person-days) within the first year post-diagnosis than the men in Groups A and B.
  • Primary care visits in the first year post-diagnosis varied much less from group to group, with
    • 8.8 visits per 1,000 person-days for men in Group A
    • 6.4 visits per 1,000 person-days for men in Group B
    • 9.5 visits per 1,000 person-days for men in Group C
  • Primary care visits varied very little from year to year for men in all three groups.

Exactly why there was little significant difference or change in primary care health resource utilization after diagnosis with prostate cancer is unclear — although this could have something to do with good coordination of care between primary care physicians and specialists within the Kaiser Permanente system.

The last paper, by Howlader et al., used a somewhat different system to evaluate risk of dying of prostate cancer from risk of dying of other causes, based on the competing effects of age, co-morbidities, and cancer stage at diagnosis. As the authors of this article are careful to observe, their model is much less individualized than the SEER*CSC model. However, it does provide us with a good deal of clarity, for prostate cancer, about the relative risk of prostate cancer-specific mortality in different groups of patients categorized by the competing characteristics defined above.

This is best indicated in Figure 3 of the paper, which shows the very high risk for prostate cancer-specific mortality within 5 years among men initially diagnosed with metastatic prostate cancer (regardless of age or co-morbidities) as compared to the risk for prostate cancer-specific mortality within 5 years for men diagnosed with earlier (localized or regional) stages of the disease. We hope that we will be forgiven for reproducing this figure below.


As yet, as far as we have been able to ascertain, the SEER*CSC nomogram system is not publicly available for use by either physicians or their patients. However, we are optimistic that, after some further refinement, it may become available — for physicians at least (and ideally for patients too) in the not too distant future. We see this tool as having great potential value in helping the newly diagnosed to gain greater insight, specifically for prostate cancer, into the relative risks of expectant management as opposed to immediate active treatment — most especially for older men with low-risk disease who are at major risk for over-treatment.

Editorial comment: The “New” Prostate Cancer InfoLink thanks Dr. Borsika Rabin and Dr. Nadia Howlader for providing us with the full texts of the three articles referred to in this commentary.


10 Responses

  1. As always thanks for your analysis. Hopefully future studies will address younger men and have more consistent methodologies.

  2. Just a question. Where do researchers get information on cause of death?

  3. That depends a lot on the country in which they are working, but usually from such sources as hospital records and death certificates.

  4. Excellent idea and its description. I don’t care when I will die, as that is up to nature and accident. But it have read quite a lot about the specific treatment I got, as it had four interesting and perhaps unusual features: the highest dose escalation for HDR brachytherapy used at that time, use of that with EBRT, use of these with 3 years of ADT, neo-adjuvant therapy preceding all this. Relevant papers are difficult. So it would be good to have reasonably simple schemes for describing mortality. These studies are good starting points.

    PS: I wrote “describing mortality,” as “prediction” seems wrong here. But I cannot say why; just a feeling.

  5. I always wondered how good the data was to support statistics on dying from prostate cancer as opposed to other causes.

    I only have one data point: a friend whose “root cause” of death was clearly colon cancer. (I will spare you the details.) But at the hospice center when I asked an attendant out of curiosity what he was entering on his laptop as the cause of death, he said renal failure. When asked why, he replied that is what the doctor told him to enter.

    This has only increased my skepticism, when I read that most men with even advanced prostate cancer, die of other causes.

  6. @John I don’t get your point.

  7. George:

    The point is that studies that show no mortality benefit of, for example, PSA testing, various treatments, etc., based on death certificates, Medicare data, or hospital records may not be completely valid if the real “root cause” of death was not correctly identified. (In a past career, I used to participate in accident investigations, where it is extremely important to identify the root cause as well as contributing causes of an accident, and know how difficult this process can be.)

  8. John and George:

    Categorizing (as opposed to being able to describe) why someone with prostate cancer died can be extremely challenging.

    Men who die of physiological causes associated with the presence of metastatic prostate cancer often die from things that are not in and of themselves prostate cancer. For example, they may die of liver failure (although the reason for failure of the liver in such cases is because the liver is riddled with metastatic prostate cancer), or they may die of pneumonia (to which they have become susceptible because their immunological system is no longer in good enough shape to fight off such an infection).

    We need to be clear that while a medical record or even a death certificate in such cases may clearly indicate that the patient died of (say) “pneumonia consequent to metastatic prostate cancer” it is not “wrong” to say that the patient died of pneumonia.

    To give an analogy that related to John’s prior occupation as an accident investigator, a man killed in a car accident may well die of “massive thoracic trauma leading to heart failure” — but it could equally well be true that the “real” reason for his death was failure to have his brakes checked in a timely manner!

    It is hard to know how to address this very complicated issue.

  9. Thanks. I agree with you. And I will probably always look at results of studies, no matter how well done, with a certain degree of skepticism.

  10. John and Sitemaster,

    I see, thanks. Making distinctions like this might well be one of the few useful things that philosophers do well. I cannot remember the terms used for this kind!

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: