Of tigers and pussycats and the ability to discriminate between them


One of the things that we have been emphasizing on this site for years is the desperate need for much better initial tests for risk of clinically significant prostate cancer that can distinguish with high accuracy between low- and very low-risk and higher-risk forms of this disease.

Progress in this area has been slow. The tests that are commercially available at this time that are based on simple blood or urine samples are still of limited accuracy, i.e., the various forms of the PSA test, the Prostate Health Index or phi test, and the 4KScore test (all based on blood samples) and the PCA3 test (based on a urine sample after a vigorous prostate massage/rectal exam).

However, we are beginning to see the early stages of development of a whole series of newer tests that may come closer to being able to distinguish with a higher level of accuracy between the “tigers” (the aggressive forms of prostate cancer than need immediate treatment) and the “pussycats” (the low-risk form of prostate cancer that can be monitored for extended periods of time before any invasive treatment actually becomes necessary).

Here are some examples of newer tests that are in development:

  • At the recent Genitourinary Cancers Symposium in Orlando, the group at the Cleveland Clinic presented data (Klein et al., abstract no. 20) from a highly sophisticated new form of PSA testing which they refer to as IsoPSA and which they believe will be able to discriminate with greater accuracy between lower- and higher-risk forms of prostate cancer, but it is clear that this test will still need to be validated in larger data sets before it is “ready for prime time”.
  • A recent study by Trabulsi et al. in BJU International discusses a new type of urine-based test that may be able to differentiate with high accuracy between risk for prostate cancer and risk for benign prostatic hyperplasia (BPH). Such a test would definitely be helpful when it comes to saving men with pure BPH from unnecessary biopsies (but it still won’t help us to differentiate between tigers and the pussycats of prostate cancer).
  • An article by Luca et al. that is to be published this morning in the new journal European Urology Focus discusses the use of a novel system for classification of prostate cancers that is based on complex mathematical models that consider information from transcriptome data sets and the outcomes of differing patients treated surgically for prostate cancer. The authors use these data to define a category of prostate cancer called DESNT cancers that are associated with relatively poor outcomes compared to other categories of prostate cancer. However, it is unclear, as yet, whether we could ever identify patients who fall into the DESNT category based on a simple blood or urine test, thus allowing for avoidance of unnecessary biopsies of men with truly low-risk disease.

The real problem, however, is the heterogeneous nature of prostate cancer (i.e., the diversity of and dissimilarities between the many different forms of the disease). We still have no really clear understanding of exactly what makes one form of prostate cancer “aggressive” and another form “benign” at a molecular level. We know that things like Gleason score and PSA level and clinical stage and genetic profiles are ways to help to discriminate between aggressive and less benign forms of the disease, but as yet we have been unable to identify any one critical biological factor that can be used to definitively discriminate between lower-risk and high-risk disease.

Ideally, we are still looking for those two or three biological markers that can tell us with 90 percent certainty (or higher) — based on a simple blood test or urine test — whether a man is or is not at high risk for any form of prostate cancer that necessitates a biopsy and that might lead to metastatic disease and prostate cancer-specific mortality.

5 Responses

  1. I can attest to the heterogeneous nature of prostate cancer. My older brother succumbed to metastatic Gleason 8 disease. He was treated by hormonal therapy and three chemotherapy agents. His cancer in turn became resistant to each modality. In 2004, I was diagnosed with localized Gleason 9 cancer. I received external radiation, plus hormonal therapy for 2.5 years. At no time was there any evidence of resistance developing; and I am obviously still around to post this comment.

  2. Great news, tugboat.

    I just finished an article with my mentor, Dr Donald Wheeler, an expert on the use of statistical process control (SPC) to continuously improve the quality of manufactured products that extends this tradition into healthcare applications. This is based on the recognition that cancer is, at the very core, a routine biological process run amok, but one which can still be studied on its own terms by using PSA test results as surrogate measures of prostate cancer tumor growth patterns. Much has been written on the use of “PSA kinetics” for this purpose. The key to using PSA results for this purpose is to be able to separate routine variation attributed to factors other than the growth of the cancer, from those signals present in the data which indicate whether or not an exponential pattern of change, characteristic of how cancer reproduces when it is in its “progression” stage.

    Analyzing data this way is called “listening to the voice of the process” within the SPC community. Don and I have illustrated how it is possible to combine the tradition of “celeration charting” developed in the field of applied behavior analysis by Dr Og Lindsley, with the use of process behavior charts to help provide evidence obtained by examining the time series data, displayed on a semi-log chart, where the vertical axis uses a base 10 log scale to plot the PSA test result, while time [usually displayed in months], is displayed using an interval scale for the horizontal axis.

    In this manner, it is possible to clearly see “changes in the rate of change” across the natural history of an individual’s prostate cancer and the slope of the “celeration line” observed on the graph when contiguous points are connected, provides direct evidence about how “aggressively” his cancer is growing. Doubling times before and after treatment can be directly calculated from the graph by a method of interpolation we illustrate in our article, thereby obviating the need to use complicated mathematical formulas, like those embedded in the MSKCC doubling time nomogram, which are of questionable use when contrasting doubling times gathered for different groups of subjects.

    ‘Pussycats” are prostate cancers whose PSA outputs display only random variation about a fixed mean, or very slow relative growth rates according to procedures we describe in the article.
    “Tigers” are those prostate cancers associated with an exponential rate of growth of PSA values graphed in the semi-log format describes above. The actual slope of the line connecting these points can be calculated to provide a “specific growth rate”, which also provides a quantitative answer to the question “How aggressively is my prostate cancer growing?” Issues related to the “trustworthiness” of PSA kinetics as a reflection of growth characteristics of the underlying cancer are discussed in the article.

    Cancer grows during a long period of dormancy before it can be clinically detected, and some cancers can be lulled back into a period of dormancy, when treatments are able to achieve a durable remission, as indicated both by imaging studies which directly measure the tumor, or indirectly by tracking and comparing rates in the progression of PSA test results before and after treatment.

  3. Dear pfadtag:

    Apparently you and you colleague are not aware that some of the most aggressive forms of prostate cancer are found in men with extremely low PSA levels at diagnosis whose PSA kinetics do not conform in any way to the “norms” that you describe above. The use of the types of mathematical analysis you describe may have some relevance to “average” types of prostate cancer but the “tigers” tend to come in a set of very different subspecies that have much more to do with genetics and the biological microenvironments in which they develop than the amount of PSA they are producing.

    We already know of something like 30 different genetic subtypes of prostate cancer each of which can behave in a variety of differing ways depending on its human host. Things like PSA can be helpful in tracking the progression and risks associated with prostate cancer in a lot of patients … but it can be utterly useless in others. I know of men diagnosed with metastatic disease and PSA levels of > 7,000 who have lived for > 10 years post-diagnosis. I also know of men who died of metastatic prostate cancer within 18 months of diagnosis with a PSA of < 0.2 and whose PSA never reached 500 ng/ml.

    Applying things like process control to human disease is an intellectually interesting concept but often utterly fails to take account of the complexity of the processes within the system to which it is being applied.

  4. Ouch.

    I appreciate the honest feedback but must beg to differ with your conclusion.

    You are entirely correct that there is no hard relationship between the absolute level of a man’s PSA level and the risk that he will be diagnosed with cancer. My cancer started while I had a very low PSA, below 4 ng/mL according to the available evidence.

    You are wrong about SPC being inconsistent with a deep understanding of cancer as a biological system. Actually it is the conventional view of cancer as a for sign entity that needs to be vanquished which is simplistic and misleading, as Don and I point out in our article.

    This discussion would be much more productive if you could read the article in question so you could determine what part of the debate of using PSA kinetics as a surrogate measure of tumor growth and proliferation I “fail to understand” versus “approach from a different angle”.

    Limitations in using PSA kinetics to guide clinical decision making derive from a fundamental misunderstanding in the prostate cancer community about the nature of variability in scientific measurements that your response to my post illustrates.

    Rather than debate you in this forum, I hope we can have a discussion about this this weekend at the Florida conference.

    I planned on canceling my travel plans because of some complications regarding my treatment, but I will stay with my original plans if you can make 30 minutes or so available to discuss the article with me this weekend. I will be available on Sunday after the Conference is over as well as Friday evening before it begins.

    I look forward to a productive discussion.

  5. Dear pfadtag:

    I am sure we can find 30 minutes sometime over the next 3 days. However, I think you are entirely missing the point that I am trying to make, which is that we actually have almost no insight whatsoever in the multiple different “processes” that can cause one man to have a specific and highly aggressive form of prostate cancer and another to not get prostate cancer at all (with all of the possible variations in between). If one has no understanding of the causative processes, it is a little difficult for me to appreciate how to apply process control to such an undefined (or at least very ill-defined) system.

    Also … Almost no one with a sophisticated appreciation of the development and progression of cancer has thought about this as “a foreign entity that needs to be vanquished” for decades. That perception was dead by the mid-1960s or earlier.

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: