Development and validation of MRI-based models to predict lymph node metastasis in bladder cancer: a multi-center study

Accurate preoperative evaluation of lymph node (LN) status is crucial for guiding precise treatment. Although several methods have been developed for preoperative evaluation of LN status, accurate evaluation remains a challenge. To develop and validate MRI-based clinical prediction models for the preoperative assessment of lymph node metastasis (LNM) in patients with very high risk non-muscle invasive bladder cancer (NMIBC) and muscle invasive bladder cancer (MIBC). A total of 468 eligible patients with very high risk NMIBC and MIBC who underwent radical cystectomy and pelvic lymph node dissection from four hospitals were enrolled (derivation cohort, n=247; internal validation cohort, n=106; external validation cohort, n=115). Preoperative clinical characteristics were reviewed to identify significant predictors for constructing a nomogram and an integer-based model. The diagnostic performance of the nomogram and the integer-based model was compared using accuracy, sensitivity, specificity, and area under the curve (AUC). Kaplan–Meier survival curves were conducted to analyze the prognosis of patients. Multivariable analysis revealed that Node-RADS score and VI-RADS score were independent predictive factors for LNM. The nomogram based on these variables achieved an AUC of 0.885 for predicting LNM. An integer-based scoring system (NV score) was established, with an AUC of 0.883. Delong’s test indicated no significant difference between two models (P=0.535). The NV score demonstrated excellent discrimination in both internal (AUC: 0.930) and external (AUC: 0.924) validation cohorts. The optimal cutoff value of the NV score for predicting LNM in all patients was determined to be >3. Patients with high-LNM risk had significantly worse overall survival (HR=3.473) and recurrence-free survival (HR=4.400). The MRI-based NV score can preoperatively assess LNM status and predict postoperative prognosis in BCa patients.