Evaluation of preoperative prostate MRI for the detection of extraprostatic extension and prediction of positive surgical margins following laparoscopic radical prostatectomy

The aim of this retrospective study is to evaluate the ability of preoperative prostate magnetic resonance imaging (MRI) findings to predict pathological extraprostatic extension (EPE) and positive surgical margins (PSM) in prostate cancer patients who underwent laparoscopic radical prostatectomy (LRP), to assess the diagnostic performance of grading or staging systems used for EPE prediction, to investigate the role of MRI-detectable features in predicting PSM, and to analyze the relationship between PSM and other parameters such as lesion location and pelvic anatomical measurements. After obtaining local ethics committee approval, 357 of 623 patients who underwent LRP between December 2017 and December 2024 were included in the study based on the following criteria: having a preoperative prostate MRI obtained at our center within six months before surgery, meeting the minimum technical standards of PI-RADS v2, and no history of prostate-targeted therapy prior to MRI or surgery. Radiological assessments were performed by a fourth-year radiology resident trained in prostate MRI and a radiologist with 27 years of experience in uroradiology, both blinded to histopathological results, through consensus reading. Clinical data included serum PSA levels and ISUP grade determined by biopsy, while imaging measurements included prostate volume and PSA density. Based on index lesions, parameters such as tumor-capsule contact length (TCCL), index tumor volume (ITV), capsular bulging, capsular irregularity, neurovascular bundle (NVB) involvement, rectoprostatic and angle obliteration (RPA), and direct signs of EPE were analyzed. Cutoff values for PSA, PSA density, TCCL, and ITV were determined through ROC analysis, and these parameters were categorized based on the cutoffs for risk analysis using odds ratios (OR) in predicting pathological EPE and PSM. Sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) were calculated for each finding. The diagnostic performances of the EPE grading system and the T staging system were evaluated in EPE prediction. The findings of the T staging system were initially grouped as early or late features, and then classified into low- and high-risk subgroups based on a scoring system. The relationships of these four subgroups with pathological EPE were analyzed. The median values for the 357 patients included were: age 62 (range, 51–78), prostate volume 44.06 cc (range, 13–201), serum PSA 11.62 ng/ml (range, 1.26–96), PSA density 0.36 ng/ml/cc (range, 0.03–2.8), ITV 2.06 cc (range, 0.03–35.95), and TCCL 15.00 mm (range, 0–77). The most common biopsy ISUP group was ISUP 2 (47.9%), followed by ISUP 3 (17%), ISUP 1 (19%), ISUP 4 (6.1%), and ISUP 5 (9.8%). Pathological examination revealed EPE in 199 patients (55.7%) and PSM in 59 patients (16.5%). MRI findings showed capsular bulging in 206 patients (57.7%), capsular irregularity in 171 (47.8%), RPA obliteration in 97 (27.1%), VPA obliteration in 77 (21.5%), and direct EPE in 59 (16.5%). According to the EPE grading system, 89 patients (24.9%) were classified as grade 0, 84 (23.5%) as grade 1, 125 (35%) as grade 2, and 59 (16.5%) as grade 3. Lesion locations with potentially significant PSM risk included the distal apex (23.8%), posterolateral region (53.5%), anterior fibromuscular stroma (11.2%), and the central zone–seminal vesicle junction (22.1%). The area under the curve (AUC) values in ROC analysis were 0.801 and 0.814 for early and late T-staging signs, respectively, and 0.829 and 0.870 for the EPE grading system alone and combined with clinical data, respectively, showing good performance in predicting EPE. Median PSA and PSA density were significantly higher in patients with pathological EPE and PSM. For EPE prediction, the AUC for PSA was 0.669 (cutoff 9.28 ng/ml; sensitivity 52.8%, specificity 73.42%) and for PSA density 0.719 (cutoff 0.14 ng/ml/cc; sensitivity 73.87%, specificity 61.15%). For PSM, the AUC for PSA was 0.587 (cutoff 19.72 ng/ml; sensitivity 27.12%, specificity 91.61%) and for PSA density 0.621 (cutoff 0.264 ng/ml/cc; sensitivity 44.1%, specificity 76.8%). ITV was significantly higher in patients with EPE and PSM; the AUC for EPE prediction was 0.758 (cutoff 1.512 cc; sensitivity 67.3%, specificity 70.9%), and for PSM prediction it was 0.621 (cutoff 2.058 cc; sensitivity 61%, specificity 62%). Median TCL was 20 mm in patients with EPE versus 11 mm in those without, and 19 mm in those with PSM versus 12 mm in those without. The AUC for TCCL in predicting EPE was 0.781 (cutoff 13 mm; sensitivity 79.9%, specificity 63.3%) and for PSM it was 0.642 (cutoff 20 mm; sensitivity 49.15%, specificity 69.8%). For EPE prediction, capsular bulging showed 71% sensitivity, 67% specificity, 74% PPV, and 70% NPV; capsular irregularity 67%, 76%, 78%, and 65% respectively; direct EPE 28%, 97%, 93%, and 52%; RPA obliteration 37%, 84%, 75%, and 51%; NVB involvement 21%, 97%, 91%, and 50%. For PSM prediction, capsular bulging showed 77.9% sensitivity, 46.3% specificity, 22.3% PPV, and 91.4% NPV; capsular irregularity 72.9%, 57.0%, 25.1%, and 91.4%; direct EPE 27.1%, 85.6%, 27.1%, and 85.6%; NVB involvement 22%, 88.9%, 28.3%, and 85.2%; RPA obliteration was not significantly associated with PSM. In logistic regression, risks for EPE were significantly higher above the cutoff values: PSA (OR 2.621; 95% CI: 1.654–4.715), PSA density (OR 4.591; 95% CI: 2.708–7.782), ITV (OR 6.303; 95% CI: 3.654–10.873), and TCL (OR 6.704; 95% CI: 4.135–10.870). MRI findings associated with increased EPE risk included capsular bulging (OR 7.363), capsular irregularity (OR 8.464), direct EPE (OR 25.942), RPA obliteration (OR 2.633), and NVB involvement (OR 11.888). For PSM prediction, increased risks were found for PSA (OR 4.484), PSA density (OR 2.462), ITV (OR 2.31), TCCL (OR 2.035), capsular bulging (OR 1.836), capsular irregularity (OR 1.923), direct EPE (OR 2.207), and NVB involvement (OR 2.269), while RPA obliteration was not significant. Among lesion locations, distal apex was the highest risk site for PSM, followed by the AFMS region and the CZ–SV junction; posterolateral location was not associated with increased risk. In distal apex lesions, apex depth (AD) and the apex depth/symphysis pubis height (AD/SPH) ratio correlated with PSM rates. In conclusion, although none of the individual MRI parameters evaluated for EPE and PSM prediction are independently decisive, their combined interpretation may provide substantial clinical benefit. The NCI EPE grading system demonstrates good predictive performance, and all T-staging features were statistically significant for EPE prediction. Subgroup analysis xxi based on early and late T-staging findings and a scoring system offers a clinically applicable and diagnostically strong method. The fact that many EPE-predictive MRI parameters were also associated with PSM, though to a lesser extent, highlights their relevance in PSM prediction. Additionally, anatomical measurements such as AD and AD/SPH ratio may offer supplementary value in predicting PSM. Keywords: Prostate Cancer, Extraprostatic Extension, Positive Surgical Margin, Prostate MRI