Assessing Lymph Node Involvement in Muscle-Invasive Bladder Cancer: Proposal of a Predictive Model Using Clinical Variables

Abstract

Background: Lymph node involvement (N+) in bladder cancer indicates a poor prognosis. Current preoperative evaluations of N+ are often inaccurate. We aimed to develop a predictive model for N+ using basic clinical variables and assess the diagnostic accuracy of Computed Tomography (CT). Methods: A retrospective cohort study was conducted. We include 62 MIBC patients who underwent radical cystectomy (RC) from 2010 to 2019 in our center. We evaluated diagnostic concordance between CT and histopathology for extravesical extension (T3a≥) and N+. Univariate and multivariate logistic regressions were used to create a predictive model, with an ROC curve and nomogram developed. Results: We found 59% sensitivity and 69% specificity for CT for staging cT3≥ and a sensitivity of 22% and a specificity of 21% for N+. NLR > 2.60 (OR 6.03, p = 0.02) and lymphovascular invasion (LVInv) in the TURB sample (OR 9.26, p = 0.04) were correlated with N+. Both fundus lesions (OR 0.21, p = 0.04) and creatinine > 0.94 mg/dL (OR 0.17, p = 0.025) were associated with reduced risk. The ROC curve of the model showed 80.4% AUC. Conclusions: A predictive model with good diagnostic performance for N+ can be developed from basic clinical data. CT sensitivity and specificity for the detection of N+ patients are limited.