Radiological Wheeler staging system: a retrospective cohort analysis to improve the local staging of prostate cancer with multiparametric MRI.

Abstract

The knowledge of tumor location and extension can allow a modulated radical prostatectomy in order to minimize positive surgical margins and reduce functional morbidity after surgery in patients with prostate cancer (PCa). Multiparametric (mp) magnetic resonance imaging (MRI) could allow the assessment of tumor extension and of its relationship with external structures. Aim of this study is to propose a new radiological Wheeler (rW) staging system applied to mp-MRI, based on the pathologic staging system (pW) for the local assessment of PCa. This retrospective single-center multi-reader study included consecutive patients with PCa and preoperative mp-MRI, who underwent non-nerve sparing radical prostatectomy. Three radiologists reported on all examinations and classified each selected lesion according to imaging criteria following rW. Whole-mount histological sections were used as the reference standard. An experienced pathologist classified the extent of prostatic capsular invasion of each PCa according to the pW. Each histological section was scanned for comparison with mp-MRI findings. The rate of PCa correctly classified by radiologists using the pW was assessed. To evaluate the accuracy of mp-MRI in the discrimination between T2 and T3 PCa, the AUC was computed. One-hundred and five patients with a total of 195 PCa foci were included in the study. 130/195 tumors with a clear overlap between mp-MRI and surgical specimens were selected. The sensitivity of the most experienced reader was lower than that of the other two readers (48.6% vs. 68.6% and 62.9%, P>0.09) while specificity and PPV were higher (95.8% vs. 79.0% and 57.9%, P<0.001; 81.0% vs. 54.6% and 35.5%, P<0.041; respectively). The AUC values for the most and the intermediate experienced readers in the detection of extracapsular extension were in the range 0.72-0.74. The rW staging system has low accuracy in predicting each single pW class, while accuracy was over 80% for experienced readers in the identification of organ-confined (T2 stage class) tumors and non-organ confined cases (T3 stage class).