Risk, genomic risk, and prognosis of risk for prostate cancer metastasis and death


A recently published study in European Urology has again (partially) validated the utility of a specific type of genomic testing as a way to project risk for progressive disease in men initially diagnosed with localized prostate cancer.

The new paper by Van Den Eeden et al. provides data from a retrospective study of biopsy tissue and long-term follow-up data from a cohort of 297 men diagnosed with low-, intermediate-, and high-risk prostate cancer, all of whom were diagnosed and actually treated by radical prostatecomy between 1995 and 2010 by physicians associated with Kaiser Permanente Northern California.

Archived biopsy tissue for these patients was tested using the Genomic Health Oncotype DX  test and a Genomic Prostate Score or GPS (ranging from 0 to 100) was assigned to each set of tissue. The GPS data were then used to project risk for prostate cancer metastasis and prostate cancer-specific mortality for these patients, and the projections were compared to actual outcomes data  from the actual patients over time.

Here is a summary of the findings:

  • The average (median) follow-up period for the 279 actual patients was 9.8 years.
  • During follow-up
    • 64/279 men (23 percent) died of prostate cancer.
    • 79/279 men (28 percent) exhibited prostate cancer metastasis (presumably including all 64 of the men who died of their prostate cancer).
  • Valid, retrospective GPS results were obtained for 259/279 patients (93 percent).
  • In univariable analysis, GPS was strongly associated with
    • Time to prostate cancer-specific death (hazard ratio [HR] per 20 GPS units = 3.23)
    • Time to identification of metastasis (HR per 20 GPS units = 2.75)
  • The association between GPS and both endpoints was still significant after adjusting for risk criteria (as defined by the National Comprehensive Cancer Network, the American Urological Association, and the Cancer of the Prostate Risk Assessment or CAPRA methodology).
  • No patient with low- or intermediate-risk disease and a GPS of < 20 (i.e., 31/259 patients or 12 percent) developed metastases or died of prostate cancer during the follow-up period.

The authors conclude that:

GPS is a strong independent predictor of long-term outcomes in clinically localized [prostate cancer] in men treated [by radical prostatectomy] and may improve risk stratification for men with newly diagnosed disease.

One of the “good things” about this study is that it provides data from a relatively large cohort of “real world” patients being treated in a largely community-based setting. On the other hand, a downside to the study is that it was carried out at a time when the vast majority of low-risk patients were still being given immediate treatment of some type for localized disease (as opposed to active surveillance).

It also has to be said that we are still going to need data from long-term, prospective studies to get a clear picture of just how accurately the Oncotype DX and other similar tests can project risk. … In other words, we need data from studies in which the decision to recommend (and to actually have) active surveillance or immediate treatment is based on the results of these tests and the associated clinical information (PSA level, PSA density, Gleason score, clinical stage, etc.) and then the patients are followed prospectively for 10 or more years to see if the availability of the genomic data adds to the accuracy of prognosis compared to the clinical data alone.

7 Responses

  1. Is there any explanation in the full paper as to how the GPS is calculated?

  2. I think you can find that on the Genomic Health web site Rick.

  3. GOOD GRIEF, Morticia; you chose the right date to post this Hallowe’en horror story. Never mind the prognosis; what about the casualty rate!

    The full text is one click behind the link so we can see “The distribution of NCCN clinical risk groups was as follows: very low — 3%; low — 21%; intermediate — 67%; and high — 9%.” Of these, 23% died of prostate cancer (never mind the possibly-underdiagnosed metastases in 28%) within a median 9.8 years of follow-up (27 years maximum). That’s an unbelievable rate among patients who all received primary treatment. What the heck?

    I had a look at the original cohort which is much larger, at 6,184. This was then whittled down based on the “prespecified cohort sampling schema”.

    Luckily we find:

    “Owing to a small number of PCD [which stands for prostate cancer death] events as compared with the overall RP-treated cohort, this study employed a stratified cohort sampling design within the study eligible cohort, with strata determined by treatment year, race, and NCCN risk groups. From the full cohort, all cases with documented PCD and available tissue were selected, along with non-PCD men sampled at a target ratio of 1:2 to 1:3. Non-PCD cases with missing tissue were replaced with additional non-PCD cases with tissue, resulting in a final ratio of 1:2.6.”

    I struggle to parse those sentences but I think they mean “a small number of PCD events within the overall RP-treated cohort” when they say “a small number of PCD events as compared with the overall RP-treated cohort”. The change makes it all make sense.

    So phew; we can put Morticia back in her coffin; there were not enough deaths so they went out and built a synthetic population with by design the enormous statistics that alarmed me.

    So the real mortality rate is actually 69/6184 or 1.12%. That’s more like it!

    You might want to lead with a disclaimer to that effect; that is not how your summary initially reads!

    What the hazards are in mapping such a distilled sample back to the original 6,184 of which ex ante any of us would be part I will leave to a day where I have recovered from my shock. But the words “MONTY HALL” are rolling around and around and around in my head in the meantime. (In fairness, I see what they have done, namely keep the work-load manageable while securing a statistically-significant “control” (alive) sample set. But it’s bothering me.)

  4. Digging a little more, a couple of interesting things pop up; there is up-grading and down-grading between biopsy and post-surgical grading, with traffic being generally towards 3 + 4. But there are quite a lot of 3 + 3s around given a cohort that started from a prior of PCSM. What gives?

    Luckily we see “strata determined by treatment year, race, and NCCN risk groups”. i.e the 3 + 3s very possibly creep into the controls by virtue of being high PSA (now known “safe”). So everyone can relax.

    Also can relax intermediate-risk patients who have a GPS <20 "It was noteworthy that no patients in this study with NCCN very low–, low-, and intermediate-risk disease (n = 31) and a GPS of <20 suffered a late event (metastases or PCD)." That's actually quite important. (Worth highlighting). I thank I shall, quite unreasonably, deem myself to have had such. Fingers crossed.

  5. GOOD GRIEF, Unreasonable Man: The warning is right there in the second paragraph of the commentary — “a retrospective study of biopsy tissue and long-term follow-up data.”

    All retrospective studies like this are biased by the assumptions being made by the investigators — who, in this case, had no interest in how many people died, only in how well the OncoType DX test worked in the patients they selected to do the analysis in. This is why it says nearer the end that, “we are still going to need data from long-term, prospective studies to get a clear picture of just how accurately the Oncotype DX and other similar tests can project risk.”

  6. And furthermore … upgrading and downgrading are both common between biopsy and surgical specimens, for all sorts of reasons, although upgrading is always more likely than downgrading.

    Again, you are analyzing this paper to look for data that the paper was never about.

  7. OK, and always with the greatest love, respect, and appreciation of everything that you do for us, but I was “analyzing it” with expectations set by the title: “Risk, genomic risk, and prognosis of risk for prostate cancer metastasis and death”. I have the keenest possible interest in my “Risk …for prostate cancer metastasis and death” and the first thing I saw was 23%. So, just for future readers with similar interests, whose eye is also caught by the 23%, forget the number; not important if you read carefully.

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 )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: