Send to

Choose Destination
See comment in PubMed Commons below
J Urol. 2006 Apr;175(4):1275-80.

New insights into the diagnostic accuracy of complexed and total prostate specific antigen using discordance analysis characteristics.

Author information

  • 1Department of Urology, University Hospital Charité, Berlin, Bayer Vital GmbH, Leverkusen, and ACOMED Statistics, Leipzig, Germany.



Comparison of diagnostic performances of total and complexed prostate specific antigen (tPSA, cPSA) using ROC curve analysis resulted in controversial interpretations. To overcome the biases due to analysis of subgroups selected by concentration ranges of only 1 marker the novel approach named discordance analysis characteristics is presented.


DAC method includes only patients who are discordantly categorized by the tests under investigation using cutoffs with identical sensitivity. Within those patients parameters describing diagnostic accuracy are calculated. The systematic nature of the results is confirmed by varying the cutoffs. DAC method is demonstrated considering the example of cPSA and tPSA data from 2 prostate cancer studies with a total of 1,624 patients (study A and B, of 283 and 565 with as well as 417 and 359 men without prostate cancer, respectively). PSA concentrations were measured using the Bayer assays.


The DAC method proved that cPSA outperformed tPSA regarding the criteria diagnostic specificity and positive predictive value. Among discordantly tested patients cPSA detected patients with an about 2-fold higher risk to have prostate cancer and with a 3.6 to 5.5-fold better specificity than tPSA. Using cPSA instead of tPSA more than 10% of unnecessary biopsies could be avoided in the tPSA range of 3 to 5 ng/ml.


The superior diagnostic performance of cPSA in comparison with tPSA warrants the recommendation to use cPSA as initial test in prostate cancer diagnostics. The DAC method is generally recommended to replace comparative ROC analyses of subgroups to evaluate the diagnostic usefulness of markers.

[PubMed - indexed for MEDLINE]
PubMed Commons home

PubMed Commons

How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for Elsevier Science
    Loading ...
    Write to the Help Desk