Objectives: We investigated the impact of interchanging free prostate specific antigen (f-PSA) concentrations from 10 different assays over a reference total PSA (t-PSA) on predicting prostate histology with free-to-total PSA ratios (f/t-PSA).
Methods: Archival sera from 80 t-PSA- and age-matched pairs of histologically confirmed prostate cancer (PCA) and benign prostatic hyperplasia (BPH) patients with t-PSA levels between 2 and 25 microg/l were investigated. Serum aliquots were analyzed for t- and f-PSA using a reference method (Access, Beckmann-Coulter Hybritech) and 10 commercially available f-PSA assays. Passing Bablok and linear regression were performed to investigate the interassay agreement between f-PSA assays. To compare diagnostic performance, ROC curves for PCA detection were calculated for the 10 f/t-PSA combinations using the reference t-PSA as denominator. Sensitivities, specificities and f/t-PSA cut-offs were calculated for varying points of the ROC curve.
Results: Despite good correlation of all 10 f-PSA assays with the reference method 4 showed significantly lower mean f-PSA levels. For f/t-PSA as a predictor of prostate histology, areas under the ROC curve (AUC) varied between 0.65 and 0.71 and, if compared to the reference method (AUC=0.70), were significantly lower in three cases. Ensuring 80% specificity, sensitivities ranged between 34% and 54% (reference method: 53%) and f/t-PSA cutpoints differed considerably depending on the f-PSA assay used (range: 0.15-0.24; reference: 0.15). Similar variations were noted at 95% specificity and 80% and 95% sensitivity.
Conclusions: Arbitrary combinations of f- and t-PSA assays should not be used to calculate f/t-PSA ratios unless adequate studies have validated the diagnostic performance and cut-offs of that particular assay choice.