A Clinically Practical Model for the Preoperative Prediction of Lymph Node Metastasis in Bladder Cancer Based on 716 Patients

Abstract

Background: Lymph node (LN) metastasis of bladder cancer is the main prognostic factor affecting the therapeutic effect. The preoperative diagnosis of LN metastasis relies on CT and MRI, but the sensitivities are only 36.9% and 76.0%. Currently, there are a few gene models predicting LN metastasis that have not been applied to the clinic. The aim of this study was to construct a clinically practical model to precisely predict LN metastasis in bladder cancer patients.

Methods: Three independent cohorts (TCGA_BLCA, GSE13507, and GSE31684) were included. The cohorts were randomized into a training cohort and a validation cohort. Differential analysis, least absolute shrinkage and selection operator (LASSO) regression, and multivariate logistic regression were applied to establish the LN metastasis prediction model. Based on the coefficient, the model was then simplified to a Fast LN Scoring System and further validated in the SYSMH cohort with 110 samples.

Findings: A total of 606 patients with 172 (28.4%) LN-positive patients were included to develop a prediction model. After multistep gene selection, the LN metastasis prediction model was constructed with 5 genes: MESP1, EFEMP1, KRT23, CALML3, and PIGZ. The model can accurately predict LN metastasis with an AUC of 0.781. Specifically, high-risk patients were more likely to develop LN metastasis in muscle-invasive bladder cancer patients (46.6% vs. 22.0%) and M0 patients (31.4% vs. 6.7%). For clinically practical use, we transformed the model into a Fast LN Scoring System using the SYSMH cohort. The LN score achieved an accuracy of 0.909, sensitivity of 0.800, and specificity of 0.926, which was superior to that of CT/MRI (accuracy 0.773, sensitivity 0.667, and specificity 0.789). High LN score patients exhibited a 61.2% LN metastasis rate, while low LN score patients showed a 3.3% LN metastasis rate.

Interpretation: In this research, we developed a 5-gene model to precisely, rapidly, and conveniently predict LN status based on biopsy specimens. The clinically practical LN score has superior diagnostic efficiency to traditional CR/MRI imaging, which will assist preoperative diagnosis for LN metastasis and guide individual therapy.

Funding: This study was supported by the National Key Research and Development Program of China (Grant No. 2022YFC2408300), the National Natural Science Foundation of China (Grant No. 82273421, 81825016, 82072827, U21A20383), Guangdong Basic and Applied Basic Research Foundation (Grant No. 2021B1515020009, 2020A1515010888), Key Research and Development Program of Guangdong (Grant No. 2018B010109006), Science and Technology Program of Guangzhou (Grant No. 202102010002), Guangdong Provincial Clinical Research Center for Urological Diseases (2020B1111170006), and Guangdong Science and Technology Department (2020B1212060018, 2018B030317001, 2017B030314026).

Declaration of Interest: The authors declare no competing interests.

Ethical Approval: The ethical consent of this study was approved by Sun Yat‐sen University Committees for Ethical Review of Research involving Human Subjects. All human tissue samples were obtained from patients with written informed consent.

Keywords: Bladder cancer, lymph node metastasis, prediction model, scoring system, imaging diagnosis