ERG rearrangement as a novel marker for predicting the extra-prostatic extension of clinically localised prostate cancer.

Currently, there are no well-established preoperative clinicopathological parameters for predicting extra-prostatic extension (EPE) in patients with clinically localised prostate cancer (PCa). The transmembrane protease serine 2 (TMPRSS2)-ETS-related gene (ERG) fusion gene is a specific biomarker of PCa and is considered a prognostic predictor. The aim of the present study was to assess the value of this marker for predicting EPE in patients with clinically localised PCa. In total, 306 PCa patients with clinically localised disease, including 220 patients (71.9%) with organ-confined disease and 86 EPE cases (28.1%), were included in the study. Receiver operating characteristic curves and logistic regression were employed to establish the optimal cut-off value and to investigate whether ERG rearrangement was an independent predictor for the EPE of clinically localised PCa. A leave-one-out cross-validation (LOOCV) model was implemented to validate the predictive power of ERG rearrangement. An increase in ERG rearrangements was identified to be associate'd with EPE, and the optimal cut-off for predicting EPE was determined to be 2.25%, with a sensitivity of 70.24% [95% confidence interval (CI), 62.6-78.9%], a specificity of 80.43% (95% CI, 75.4-85.1%), and an area under the curve (AUC) of 0.781 (95% CI, 0.730-0.826). In the LOOCV model, ERG rearrangement also demonstrated good performance for predicting EPE (sensitivity, 76.923%; specificity, 71.429%; 95% CI for AUC, 0.724-0.958). In addition, a high Gleason score (≥7) and a cT2c classification upon biopsy were independent factors for EPE.