Receiver-operating characteristic as a tool for evaluating the diagnostic performance of prostate-specific antigen...

Receiver-operating characteristic (ROC) analysis is often applied as evaluation tool to compare the diagnostic validity of laboratory tests. The aim of this study was to draw attention to preconditions which should be taken into account when ROC analysis is used to assess the diagnostic performance of total prostate-specific antigen (tPSA) and its molecular forms in differential diagnosis between prostate cancer and benign prostatic hyperplasia (BPH).

Using a standard software (GraphROC for Windows), ROC analyses were performed and the areas under the curves were calculated for four hypothetical pairs of groups. Every group included 40 patients with prostate cancer and with BPH showing different tPSA concentrations (range of 2-10 microg, but similar free-to-total PSA ratios (fPSA%).

The area under the fPSA% ROC curve showed the highest value, whereas the areas under the tPSA ROC curves were dependent on the distributions of tPSA concentrations in the patients. The ability of fPSA% to improve the differential diagnosis between prostate cancer and BPH in comparison to tPSA was not furthermore evident, if the prostate cancer group included more patients with higher tPSA concentrations than the BPH group.

When the diagnostic performance of tPSA and its derivatives like molecular forms in patients with prostate cancer and BPH should be compared by ROC analysis, a matching procedure is recommended prior to ROC analysis to compensate the effect of possible unequal tPSA distributions in both groups. Each BPH (or PCa) patient should be matched with a PCa (or BPH) patient with nearest tPSA concentration so that an optimum of overlapping tPSA concentrations in both groups can be achieved.