A Novel Methylation Marker NRN1 plus TERT and FGFR3 Mutation Using Urine Sediment Enables the Detection of Urothelial Bladder Carcinoma

Simple Summary In the present study, we aimed to construct a methylation diagnostic tool using urine sediment for the detection of urothelial bladder carcinoma (UBC), and improved the diagnostic performance of the model by incorporating single nucleotide polymorphism sites. In stage I, single NRN1 exhibited the highest AUC. At the best cutoff value of 5.16, single NRN1 biomarker showed a sensitivity of 0.93 and specificity of 0.97. In stage II, random forest algorithm was applied to construct the model, including NRN1, TERT C228T and FGFR3 p.S249C. The tool exhibited AUC values of 0.953, 0.946 and 0.951 in training, test and all cohort. In the external validation cohort (stage III), the model achieved an AUC of 0.935, sensitivity of 0.864 and specificity of 0.895. The model also exhibited a superior sensitivity and comparable specificity compared with conventional cytology and FISH, and may be used as a replaceable approach for the detection of UBC. Background: Aberrant DNA methylation is an early event during tumorigenesis. In the present study, we aimed to construct a methylation diagnostic tool using urine sediment for the detection of urothelial bladder carcinoma, and improved the diagnostic performance of the model by incorporating single-nucleotide polymorphism (SNP) sites. Methods: A three-stage analysis was carried out to construct the model and evaluate the diagnostic performance. In stage I, two small cohorts from Xiangya hospital were recruited to validate and identify the detailed regions of collected methylation biomarkers. In stage II, proof-of-concept study cohorts from the Hunan multicenter were recruited to construct a diagnostic tool. In stage III, a blinded cohort comprising suspicious UBC patients was recruited from Beijing single center to further test the robustness of the model. Results: In stage I, single NRN1 exhibited the highest AUC compared with six other biomarkers and the Random Forest model. At the best cutoff value of 5.16, a single NRN1 biomarker gave a diagnosis with a sensitivity of 0.93 and a specificity of 0.97. In stage II, the Random Forest algorithm was applied to construct a diagnostic tool, consisting of NRN1, TERT C228T and FGFR3 p.S249C. The tool exhibited AUC values of 0.953, 0.946 and 0.951 in training, test and all cohorts. At the best cutoff value, the model resulted in a sensitivity of 0.871 and a specificity of 0.947. In stage III, the diagnostic tool achieved a good discrimination in the external validation cohort, with an overall AUC of 0.935, sensitivity of 0.864 and specificity of 0.895. Additionally, the model exhibited a superior sensitivity and comparable specificity compared with conventional cytology and FISH. Conclusions: The diagnostic tool exhibited a highly specific and robust performance. It may be used as a replaceable approach for the detection of UBC.