Benign PSA, measured in serum with a proprietary automated assay, improves prostate cancer detection when it is incorporated into an artificial neural network.
Berlin-Benign PSA (bPSA), measured in serum with a proprietary automated assay (Access BPHA; Beckman Coulter, Fullerton, CA), improves prostate cancer detection when it is incorporated into an artificial neural network, according to German researchers.
bPSA is a subform of free PSA (fPSA) that has previously been shown to be more closely associated with BPH than with prostate cancer. Its use in enhancing discrimination between patients with prostate cancer (287 patients) and controls with no evidence of malignancy (254) was investigated in a study using archived sera of men having a total PSA (tPSA) of 0 to 10 mcg/L.
The diagnostic performance of various markers alone and in combination was determined on the basis of receiver operating characteristic (ROC) curve analysis and calculations of specificity at 90% and 95% sensitivity levels.
His group's work was reported at the AUA annual meeting in Chicago and subsequently published in Urology (2009; 74:873-7).
"Previous studies evaluating bPSA as a diagnostic tool have been performed using a different assay and in smaller patient populations. Our investigation using the Access BPHA assay is the largest study of bPSA reported so far, and the results suggest this test is useful for improving the discrimination between men with benign and malignant prostatic disease," Dr. Stephan told Urology Times.
Comparisons between the prostate cancer patients and the controls showed differences between the groups in median levels of BPHA as well as tPSA, %fPSA, and BPHA/tPSA that were consistently statistically significant. In the ROC analyses of single factors, %fPSA had an AUC of 0.77 and was significantly superior to tPSA (0.59), BPHA (0.57), and prostate volume (0.69) as well as compared with BPHA/fPSA (0.51) and BPHA/tPSA (0.69).
"An artificial neural network combining tPSA, volume, and age with %fPSA had the same AUC as with %fPSA alone. However, when this multifactorial model incorporated BPHA/tPSA, the AUC was significantly higher-0.81-showing the importance of BPHA/tPSA," said Dr. Stephan.
In the specificity analyses, the artificial neural network including BPHA/tPSA with tPSA, %fPSA, age, and prostate volume had a specificity of 53.9% at 90% sensitivity and 44.5% at 95% sensitivity. The specificities at the 90% and 95% sensitivity levels were 50.0% and 40.6%, respectively, for the artificial neural network model excluding BPHA/tPSA and only 40.9% and 27.2%, respectively, for %fPSA alone. The differences compared with %fPSA alone were statistically significant for the model including BPHA/tPSA at both sensitivity levels but only at 95% sensitivity for the multivariate model without BPHA/tPSA.
Beckman Coulter provided the test kits for the assays.