The developing Cancer Survival Query System (CSQS) and its application to prostate cancer

Just over a month ago we mentioned the CSQS in commenting on a Swedish paper (by Eloranta et al.) that dealt with prediction of the “crude” or “real world” probability of death from prostate cancer (as opposed to potential, competing causes of death).

The CSQS is an evolving prognostic tool based on data from the Surveillance, Epidemiology, and End Results (SEER) data registry that has been designed to provide such “real world” projections of risk of death from cancer or other causes for U.S. cancer patients who have a new or recent diagnosis of cancer. At the present time, the two cancers for which it is able to offer detailed projections are colorectal cancer and prostate cancer.

To build this model for prostate cancer patients, Feuer and his colleagues at the National Cancer Institute (NCI) and other centers had to do two quite different things:

  • They had to develop a way to project the life expectancy a prostate cancer patient would have had if he had not developed cancer.
  • They had to develop cause-specific estimates of the chance of dying of prostate cancer as a function of the stage and tumor characteristics.

We are not going to try to get into the details of the statistical model underlying the CSQS. (Interested readers would need to get their own copy of the full text of the paper by Feuer et al. in which this model is described.) However, here is basically how the model works:

  • Based on Medicare claims data linked to SEER, patients’ health conditions can be established in the year prior to their cancer diagnosis.  The authors are able to use this patient-specific information to develop a “health status adjusted age” for any patient, taking account items specific to his medical history (e.g., cardiovascular disease, diabetes, rheumatoid arthritis).  The “health status age adjustment” is the amount you would add to or subtract from a person’s chronologic age to account for good health or poor health (prior to their diagnosis with cancer) relative to the average person of the same gender and racial group in the U.S. population.
  • Then, for each patient — based on the SEER data — they take account of prostate cancer-specific findings such as clinical stage, pathological stage (for men treated surgically), Gleason score, etc. (although the system does not specifically include  PSA level if the tumor is incidentally detected for surgery to treat an enlarged prostate, or through a screening PSA test, this is partially accounted for in the staging).

They are then able to use all of these data to estimate the patient’s probability of each of the following between 1 and 10 years post-diagnosis:

  • Death from prostate cancer
  • Death from other causes
  • Still being alive

Furthermore, they can compare the data for this patient to similar data for a man  with the same tumor characteristics, but who has “average” health conditions for his age.  In an example provided in their paper, Feuer et al. show that an African American male of 74 years of age, with a history of heart disease, rheumatologic disease, and diabetes, has a health status adjusted age of 81. Then, if this man is diagnosed with clinical stage T2 prostate cancer and a Gleason score of 8 to 10, and does not have surgery:

  • His probability of death from prostate cancer within 5 years is 7.1 percent
  • His probability of death from prostate cancer within 10 years is 12.6 percent
  • His probability of death from other causes within 5 years is 39.2 percent
  • His probability of death from other causes within 10 years is 65.8 percent
  • His probability of still being alive at 5 years is 53.7 percent
  • His probability of still being alive at 10 years is 21.6 percent

Now there are clearly limits to the utility of the CSQS at this time. It can not, for example, be used to actually compare the relative outcomes of treatment by surgery as opposed to treatment by radiotherapy or management by active surveillance. The SEER data used to build the model is simply not sufficiently robust to allow this because patients are not randomized between alternative treatments, the information on stage is not comparable, and there are many strong selection biases inherent in who is healthy enough and chooses to have surgery, radiation, or active surveillance.

At present, Feuer and his colleagues are working with selected health care provider groups to determine  the clinical utility of the CSQS, and to see if their model is accurate in its predictions in real world health care settings. What is interesting to The “New” Prostate Cancer InfoLink about this work is two things:

  • Will it really have real practical utility? Most specifically, will we be able to use a model like this to help (for example) a man of 67 years of age with significant other health conditions decide how aggressive to be about the treatment of his newly diagnosed, localized prostate cancer?
  • If it can be shown to have real practical utility, then can the model be made robust enough and straightforward enough to be directly available to patients (without the need for a professional intermediary) in the way, for example, that the Kattan nomograms and the Partin tables are directly available to patients?

Clearly Dr. Feuer and his colleagues have a ways to go before the CSQS is “ready for prime time,” but this is certainly an interesting and potentially patient-oriented tool that may be able to help many men make better decisions about how to manage a new diagnosis of localized or apparently localized prostate cancer.

As we mentioned before, you can also find a recent, online video presentation about the output of this program given by one of Feuer’s colleagues at the 2012 Health Data Palooza in Washington, DC.

Editorial comment: The “New” Prostate Cancer InfoLink would like to thank Dr. Sandra Eloranta for bringing the development of the CSQS to our attention and Dr. Eric (“Rocky”) Feuer and two of his colleagues at the NCI for spending considerable time in helping us to fully appreciate both the opportunities and the current limitations of the CSQS.

2 Responses

  1. Thanks as always for finding and interpreting these interesting studies and enterprises.

    One additional piece of the mental picture for the patient, in addition to the likelihoods of death by prostate cancer and death from other causes, is the probability of levels of suffering of varying durations associated with each. For prostate cancer, my impression is that the end process is prolonged and difficult, though my own plan, hopefully no longer needed, was to use pain killers liberally. For other causes of death, such as heart attack and stroke, the end process often seems to be brief and without the need to combat serious and prolonged conscious pain. At least that is my layman’s impression without knowing the odds of swift demise free of prolonged suffering versus prolonged and painful decline for other causes.

    I’m also not envying the task of the modellers in coping with another issue: likely lengthening of survival due to the sudden advent of a handful of powerful new drugs approved for late stage disease that are already migrating toward earlier use, as in the very recent approval of Zytiga for men with chemo-naive disease having the other previously defined features in the earlier approval. Exploitation of this new, rich arsenal is only beginning, but experts like Dr. Howard Scher have already commented on prospects for trying combinations of these drugs and using them earlier in the course of the disease, boosted by liberalization of FDA approval evidence that at least is going to include radiological evidence instead of just survival. That’s a nice problem to have for all of us, and perhaps the modellers could cope with it by updating their model every year or so.

  2. Having read this I believed it was rather informative.

    I appreciate you spending some time and energy to put this short article together. I once again find myself personally spending a significant amount of time both reading and leaving comments. But so what, it was still worthwhile!

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: