Prostate Health Index (PHI)is more accurate than PCA3 assay in the prediction of prostate biopsy (PBx) in patients who have undergone initial and repeated prostatic biopsies
Vincenzo Scattoni, Massimo Lazzeri, Stefano De Luca*, Enrico Bollito*, Donato Randone*, Firas Abdollah, Carmen Maccagnano, Giovanni Lughezzani, Alessandro Larcher, Giuliana Lista, GiulioMaria Gadda, Patrizio Rigatti and Giorgio Guazzoni.
Department of Urology - University Vita-Salute,Scientific Institute H San Raffaele, Milan, Italy
*Department of Urology, OspedaleGradenigo, Torino, ITALY
Introduction
Beckman Coulter PHI and PCA3-Score have shown to improve the prediction of prostate cancer (PCa) in men undergoing first and/or repeated extended PBx. However, a head-to-head comparison of the PCa prediction accuracy of these two molecules is still lacking.
Material & Methods: The performance characteristics of Beckman PHI [(p2PSA/fPSA) x √tPSA)] and PCA3 across different PSA ranges was evaluated in a combined dataset of 210 European men (median age 67 yrs, median PSA: 7.2 ng/ml±7.1) undergoing first (n=116 cases) or (multiple) repeat PBx (n=94 cases) (12-22 cores) in two large academic centers. Both tests were performed prior to DRE and PBx. The blood samples were processed using UniCel DxI800 Immunoassay System analyzer (Beckman & Coulter, Brea, CA, USA). The PCA3 score was determined by the PROGENSA ™ PCA3 Assay. PHI and PCA3 were also compared to age, total PSA, prostate volume, DRE, PSA density (PSAD), and %fPSA. Statistical analysis encompassed descriptive statistics, nonparametric analysis of differences in PBx+ vs. PBx- men. A basic multivariable model (BM) was developed, and consisted of age, PSA, %fPSA, and DRE. The area-under-the-curve (AUC) was used to quantify the discrimination of the basic model, and the Hosmer-Lemeshow statistic was used to measure its calibration. The same benchmarks were used to measure the improvement of the prediction accuracy (discrimination and calibration) of the basic model by adding PCA3 and PHI separately. Finally, we relied on decision curve analysis (DCA) to compare the various models directly in a head-to-head fashion.
Results: Cancer was detected in 70/210 patients (33.3 %) in the whole group (35.1% and 30.9% in the initial and repeat group, respectively). In the entire cohort, PHI had the highest AUC value (0.69), which was significantly higher than PCA3 (0.58) and %FPSA (0.59). Likewise, in the initial setting, AUC was highest for PHI (0.69) followed by %fPSA (0.67), PCA3 (0.58).Similarly, in the repeat setting, AUC was highest for PHI (0.72) compared to PCA3 (0.63), %fPSA (0.52).Adding the PHI index and PCA3 to the basic model (AUC 0.77) improved its predictive accuracy (AUC 82 and 78, respectively),but this increase reacheda full statistical significance only for PHI. Adding PCA3 to the basic model + PHI did not increase the predictive accuracy in the whole population, as well as in the two settings.
In DCA, the highest net-benefit was observed when PHI was added to the basic model. This holds true in the initial as well as in the re-biopsy setting. Adding PCA3 to the basic model + PHI did not result in a further improvement of the net-benefit.
Conclusions: PHI seems to be the strongest parameter to predict PBx outcome in the initial as well as in the repeated biopsy. Adding PCA3 to the basic model +PHI did not improve the prediction accuracy in both settings. PHI may therefore be useful in guiding biopsy decisions.