Another “Holy Grail” being touted for diagnosis of high-risk prostate cancer: an update


We have now been able to track down a link to detailed information about the article by Yu et al. referred to in yesterday’s commentary under the same heading.

The article is focused on the use of “copy number variation” (CNV) among the genomes in specific prostate cancer tumors.

The authors sought to discover whether such CNV  could be used to predict biochemical relapse based on analysis of benign tissues adjacent to the tumor or the blood of patients actually treated for prostate cancer. Most specifically, they were interested in whether testing of non-cancerous sample specimens from blood and from non-cancerous prostate tissues could be used to predict risk for aggressive forms of prostate cancer

They were able to use 241 samples from patients actually treated (104 actual tumor samples; 49 samples of matched, benign tissues adjacent to tumor; 85 matched blood samples, and  three cell lines).

According to the available information, Yu et al. were able to show that:

  • Based on the use of gene-specific CNV from actual tumor samples, their model correctly predicted
    • Biochemical relapse in 73 percent of actual cases
    • A short PSA doubling time of < 4 months in 75 percent of actual cases
  • Based on the use of gene-specific CNV from benign tissues adjacent to the prostate, their model correctly predicted
    • Biochemical relapse in 67 percent of actual cases
    • A short PSA doubling time of < 4 months in 77 percent of actual cases
  • Based on the use of median-sized CNV from patients’ blood samples, their model correctly predicted
    • Biochemical relapse in 81 percent of actual cases
    • A short PSA doubling time of < 4 months in 69 percent of actual cases
  • Based on the use of median-sized CNV from actual tumor samples, their model correctly predicted
    • Biochemical relapse in 75 percent of actual cases
    • A short PSA doubling time of < 4 months in 80 percent of actual cases

In their conclusion, the authors state that, “For the first time, our analysis indicates that genomic abnormalities in either benign or malignant tissues are predictive of the clinical outcome of a malignancy.” This certainly appears to be the case.

These study data are distinctly interesting, but this is a small study; the predictive accuracy is “only” of the order of 75 percent (i.e., it was wrong about a quarter of the time); and it will need to be replicated by others to see if the results are actually even this good when carried out at another center by other researchers.

Does this meet “Holy Grail” standards? Well, not in our opinion. “Holy Grails” are meant to imply some form of perfection … and a 25 percent inaccuracy rate is nowhere near perfection in our humble estimation. On the other hand … the fact that a blood sample was able to project risk for recurrence of an aggressive form of prostate cancer opens all sorts of doors to refinement of this research (if this initial finding does prove to be replicable).

2 Responses

  1. From reading this, I’m not clear on the sensitivity vs. specificity issue. They seem to be saying that 75% of the cancers they predicted would relapse actually did. That would be an improvement over having no idea. What about the cancers they predicted would not relapse? Did any of those actually relapse? If 25% of those actually did relapse, that would throw cold water on using this to make a treat vs watch decision. However, as a screening blood test to find those at higher risk for aggressive cancer in the first place, it might be fantastic. Perhaps, in a manner similar to certain tests for a genetic predisposition to breast or colon cancer, this could be done in early adulthood and appropriate precautions taken for those coming up positive. (Assuming it is verified to work)

  2. Doug:

    I am no wiser than you are about this. You have seen the same data that I have. I think the only real way to look at this at the moment is as “a potentially interesting work in progress.”

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: