PSA density vs. PSA level and the prediction of prostate cancer risk


A newly published study in the journal Urology (the “Gold Journal”) has suggested that — at least for patients with a PSA level between 4 and 10 ng/ml — PSA density may be better than PSA level in the determination of which of these patients need to go on to have a biopsy.

We should be clear that this is not a new topic of discussion in the urology community. However, what is different in the new paper by Jue et al. is that they have used data from a large, well-conducted clinical trial that had involved > 1,200 men to evaluate the relative value of PSA density as compared to PSA density as predictors of risk, and correlated the prediction of risk to the actual finding of prostate cancer (or not) among these patients.

Basically, Jue et al. used the data from the prospective, multi-institutional trial that had originally been designed to evaluate the utility of the 4KScore test as a prognostic test for risk of prostate cancer. This trial involved patients tested at 26 different centers in the USA, all of whom had been referred for a prostate biopsy based on suspicion of prostate cancer by the referring physician. Immediately prior to their biopsies, all 1,290 patients had a blood sample drawn, and that blood sample was tested using a standard PSA test as well as the 4KScore test at a single laboratory. The patient all had a rectal exam and then a TRUS-guided template biopsy with not less than 10 cores being sampled. Prostate volumes were also measured on all patients. As a consequence, Jue et al. had all of the relevant information from a highly defined group of men at perceived risk for prostate cancer available for the conduct of their analysis.

So here is what they report:

  • 1,290 patients actually participated in all elements of the trial
  • 438/1,280 patients (34 percent) had a PSA level of < 4 ng/ml, and of those men
    • 144 were diagnosed with prostate cancer of any type.
    • 44 were diagnosed with clinically significant prostate cancer.
  • 725/1,290 patients (56 percent) had a PSA level of 4 to 10 ng/ml, and of those men
    • 361 were diagnosed with prostate cancer of any type.
    • 177 were diagnosed with clinically significant prostate cancer.
  • 127/1,290 patients (10 percent) had a PSA level of > 10 ng/ml, and of those men
    • 80 were diagnosed with prostate cancer of any type.
    • 63 were diagnosed with clinically significant prostate cancer.
  • PSA density was a much more accurate predictor of risk that PSA alone for all men with a PSA level of 4 ng/ml or higher.
  • Among men with a PSA level of < 4 ng/ml, PSA density was no better than PSA at predicting actual prostate cancer (of any type or of clinical significance).
  • Among men with a PSA level between 4 and 10 ng/ml
    • PSA density was better at predicting prostate cancer of any type (AUC = 0.70 vs. 0.53; P < 0.0001).
    • PSA density was better at predicting clinically significant prostate cancer (AUC = 0.72 vs. 0.57; P < 0.0001).
  • Among men with a PSA level of > 10 ng/ml
    • PSA density was better at predicting prostate cancer of any type (AUC = 0.84 vs. 0.65; P < 0.0001).
    • PSA density was better at predicting clinically significant prostate cancer (AUC = 0.82 vs. 0.68; P < 0.0001).
  • Differing PSA density cut-points had differing impacts on the ability to predict the probability of cancer.
    • At a PSA density cut-point of ≥ 0.05
      • 537/585 cases of prostate cancer of any type (92 percent) would be found on biopsy.
      • 273/284 cases of clinically significant prostate cancer (96 percent) would be found on biopsy.
    • At a PSA density cut-point of ≥ 0.15
      • 256/585 cases of prostate cancer of any type (44 percent) would be found on biopsy.
      • 167/284 cases of clinically significant prostate cancer (58 percent) would be found on biopsy.
  • The incremental gain in performance between PSA density and PSA alone was highest in men who had had a prior, negative biopsy.

It will be interesting to see how the urology community reacts to these data. The authors themselves state that “the best improvement in cancer detection” could be observed in the subset of men who had already had a prior, negative prostate biopsy, and this may have particular implications for men with larger prostates (who are often likley to have higher than average PSA levels and therefore are more likely to be referred for biopsy more often).

We should be clear that what this study also shows is that PSA density alone (at any cut-point) is not a perfect marker for deciding whether an individual patient should or should not have a prostate biopsy, but it may certainly be very helpful in combination with other data.

Editorial note: We thank Joshua Jue and Dr. Sanoj Punnen of the University of Miami Miller School of Medicine for providing us with a copy of the full text of this article for our review.

7 Responses

  1. Yet another screening strategy unproven to save lives. Sad.

  2. I assume there are typos in the discussion of cut points:

    “At a PSA density cut point of…..”

    The percentage of cancers found between the cut points of 0.05 and 0.15, about 50% of those found, is surprising as most active surveillance programs suggest triggering a biopsy at PSA densities of > 0.15, a few at > 0.10.

    If the cut point data are accurately reported above (cut points are not discussed in the study abstract), then the PSA density standards currently used in active surveillance programs could miss a significant number (about 39%) of clinically significant cancers. This seems inconsistent with published results of active surveillance programs.

  3. Dear Jon:

    The data are accurately reported from the full report by Jue et al., which is exactly why I wanted to see the entire paper before reporting on it. And you are correct, this does raise questions about using a PSA density of 0.15 as a cut point for acceptance into active surveillance programs. However, I should point out that active surveillance is not intended to be a curative form of prostate cancer care. It is a strategy designed to defer treatment until such treatment is actually necessary (which may never happen for many men). Given that perspective, a cut point of 0.15 may still be an entirely appropriate cut point for entry into active surveillance.

    Please note that I did insert the missing word “density” into the text above and removed the parentheses from your comment so that everything matched up. Thanks for noticing that.

  4. The entry criteria in past active surveillance programs has been to estimate that the candidate has not yet had “clinically significant” cancer. If the cut point for PSA density is 0.05 to indicate “clinically significant” cancer, then a man with a “normal” 25 cm3 prostate would have to have a PSA less than 1.25 ng/ml to be considered for active surveillance. This would exclude a large number of men who have successfully deferred treatment with active surveillance. Without further evidence, it is difficult to imagine an active surveillance program using a PSA density of 0.05 ng/ml as a cut point for accepting candidates.

  5. Dear Jon,

    I don’t know that it’s true that “most active surveillance programs suggest triggering a biopsy at PSA densities of > 0.15, a few at > 0.10.” The Epstein criteria, which include a PSA density < 0.15, are still used in some very restrictive programs as an admittance condition. There is a lot of controversy about how, if at all, PSA should be used in monitoring. With mpMRIs so prevalent now, changes in imaging are increasingly used to trigger biopsies.

    I think the study discussed here is more concerned with how PSA density can be used in the initial detection of significant prostate cancer. As the Sitemaster pointed out, it’s better than PSA alone, but it’s not very good (30% of most men will get false predictions). Remember that with AS, the issue isn’t preliminary detection (it’s already been detected); the issue is progression. So it is not a big concern that low PSA density misses detecting the cancer, it is only the change in PSA density that would set off a red flag for AS. That remains to be validated. PSA changes in men on dutasteride (which shrinks the prostate) have not proved to be prognostic for progression on AS.

    The other kallikrein-based markers (phi, 4KScore, and isoPSA) are a bit more specific and do only a slightly better job than PSA density. There is an interesting opportunity for them to play an increased role in AS monitoring, especially phi because it is relatively inexpensive and covered by insurance. But such use remains to be validated.

    By the way, Sitemaster, I think there is a typo here; the abstract reads:”PSA density performed better than PSA in detecting any prostate cancer within a PSA between 4-10 ng/ml (AUC: 0.70 vs. 0.53, P<0.0001)”, while the above reads “AUC = 0.84 vs. 0.57” for that same group.

  6. Dear John:

    First, most active surveillance programs have never used PSA density cut points as entry criteria at all. The major exception was the Johns Hopkins protocol, which was well understood to have one of the most stringent sents of entry criteria. I completely agree with you that there is no definitive evidence as yet to set a cut point of 0.05 as an entry criterion for active surveillance (although I am sure that there are some urologists who would like that).

    Second, I think you are underestimating the size of the “average” prostate in men of 50+ years of age, which I think is more like 45 cm3.

  7. Allen:

    Actually there was more than a typo. Two whole lines of data had vanished on me! They have now been accurately replaced. Mea culpa.

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 )

w

Connecting to %s

%d bloggers like this: