Can phi density accurately predict risk for clinically significant prostate cancer?

A new article in BJU International has suggested the possibility that Prostate Health Index density (i.e., a patient’s phi score divided by his prostate volume) may be able to predict the probability of a finding of clinically significant prostate cancer on biopsy.

Tosoian and colleagues at Johns Hopkins in Baltimore report data from a cohort of 118 patients who all had a PSA level of > 2 ng/ml, a negative digital rectal exam, and who were given both a phi test and a standard, systematic prostate biopsy at Johns Hopkins in 2015. The phi score was calculated in the standard manner [([-2]proPSA/free PSA) × (PSA)(½) ] and prostate volume was determined from the transrectal ultrasound data.

The authors defined “clinically significant” prostate cancer as a cancer exhibiting any one of the following characteristics:

  • Any amount of cancer with a Gleason score of 7 or higher
  • Cancer with a Gleason score of 6 and cancer in > 2 biopsy cores
  • Cancer with a Gleason score of 6 and cancer in > 50 percent of any positive core

Here is a summary of the key findings:

  • 47/118 patients (39.8 percent) had clinically significant prostate cancer (as defined above).
  • The average (median) phi density score was
    • 0.70 in all 118 patients
    • 0.53 in patients who had a negative biopsy or clinically insignificant prostate cancer
    • 1.21 in patients with clinically signifciant prostate cancer
  • Clinically significant prostate cancer was detected in
    • 3.6 percent of men in the lowest quartile of phi density (i.e., < 0.43)
    • 36.7 percent of men in the interquartile range of phi density (i.e., 0.43 to 1.21)
    • 80 percent of men in the highest quartile of phi density (i.e, > 1.21)
  • Based on a phi density threshold value of 0.43, phi density had
    • 97.9 percent sensitivity for clinically significant prostate cancer
    • 38.0 percent specificity for clinically significant prostate cancer
    • 100 percent sensitivity for cancer with a Gleason score of ≥ 7

Now it is important to recognize that, for quite a while, the group at Johns Hopkins has been using a PSA density of 0.15 ng/ml/cm3 or higher as a potential indicator of risk for clinically significant prostate cancer among their otherwise low-risk patients, so how does phi density stack up against PSA density and other possible indicators of risk?

Tosoian et al. report that phi density “demonstrated the highest discriminative ability” for clinically significant prostate cancer (again, as defined above) as compared to five other possible indcators of risk:

  • Total PSA alone (AUC = 0.52)
  • PSA density (AUC = 0.70)
  • %Free PSA (AUC = 0.75)
  • %Free PSA × prostate volume (AUC = 0.79)
  • phi score alone (AUC = 0.76)
  • phi density (AUC = 0.84)

The AUC value (which stands for “area under the curve”) is a well-established statistical tool used to assess the quality of a particular form of measurement. The closer the AUC value is to 1.00, the higher the accuracy of the form of measurement.

The authors conclude, cautiously, that

Based on this prospective single-center experience, phi density could be used to avoid 38 percent of unnecessary biopsies while failing to detect only 2 percent of clinically-significant cancers.

The phi test doesn’t seem to have been widely adopted by the urology community to date (as far as we can tell). Very few patients report phi scores when they log into our social network or post data to systems like the Prostate Problems Mailing List (PPML). It is also not clear from the abstract of thsi paper whether the research team at Johns Hopkins used the commercial form of the phi test or simply calculated the phi values themsleves using the equation shown above. However, …

If the phi density level really is that accurate at being able to predict clinically significant prostate cancer, this could be a “watershed moment”. On the other hand, the relatively low specificity of the test (at just 38 percent) is still a relative drawback.

It will be interesting to see if other prospective and perhaps larger studies can replicate these results. It would also be interesting to see if similar results could be achieved with slightly different definitions of clinically significant prostate cancer, such as this one:

  • Any amount of cancer with a Gleason score of 4 + 3 = 7 or higher
  • More than 2 mm of Gleason pattern 4 tissue and cancer with a Gleason score of 3 + 4 = 7
  • Cancer with a Gleason score of 6 and cancer in > 3 biopsy cores
  • Cancer with a Gleason score of 6 and cancer in > 50 percent of any positive core

We still need to build greater consensus about how to define clinically significant and clinically insignificant prostate cancer, while allowing for some recognition of the need for individual variation of acceptance of risk for metastasis after 10 or 15 years.


2 Responses

  1. It is well known that systematic biopsy misses a certain percentage (maybe 25%?) of clinically significant cancers. I have to wonder how many of the false positives of the phi density test are actually cancers missed by the biopsy. I hope that the researchers will do mpMRI follow-up on the false positives of this cohort to see how many can be reclassified as true positives.

  2. The way the formula for phi is presented above is rather hard to understand. The first reason is the presence of the constant [-2] which creates the impression that the following number should be multiplied by -2. (In fact, “[-2]proPSA” is just the way that this variant of proPSA has been named.) The other reason is that the exponent of ½ is presented inline as if it were a factor to be multiplied. It would be better to show it as a superscript. I don’t know if your text editor has an “Insert HTML” option. If it does you can use it to enter “½” to get the superscript. If not, I would suggest “PSA^(½)” which is another common way of representing exponents.

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 )

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: