A diagnostic model for non-invasive urothelial cancer early detection based on methylation of urinary tumor DNA.

Background: Diagnostic methods for urothelial cancer (UC) are often invasive, while urinary cytology, a non-invasive alternative, suffers from limited sensitivity. This study aimed to identify differentially methylated markers in urinary tumor DNA and develop a diagnostic method to enhance the sensitivity of non-invasive UC detection.

Methods: Whole-genome bisulfite sequencing and deep methylation sequencing were employed to identify significantly hypermethylated UC-associated genes in clinical samples and public UC datasets. Further screening was conducted using tumor biopsies and urine samples from patients, leading to the selection of three hypermethylated UC markers. A diagnostic model based on these markers was constructed and validated in a cohort (N = 432) comprising patients with UC, other cancers, benign lesions, and non-UC urinary tract diseases.

Results: Validation in a cohort of 432 subjects demonstrated that the UC diagnostic model, incorporating three hypermethylated markers (VIM, TMEM220, and PPM1N), achieved an overall sensitivity of 94.44% in 108 UC patients. Specificities were 96.34%, 90.76%, and 87.72% in 191 non-neoplastic individuals, 76 patients with benign lesions, and 57 patients with other cancers, respectively, resulting in an overall specificity of 93.52%. Methylation level analysis revealed significantly higher methylation (P < 0.001) for three markers in UC samples compared to non-UC samples. Furthermore, the model exhibited sensitivities of 80% and 88.57% for detecting stage 0a/0is and stage I UC, respectively.

Conclusions: The UC diagnostic model demonstrates excellent diagnostic performance, particularly in the early detection of UC. This non-invasive approach, characterized by high sensitivity and specificity, holds significant potential for further clinical evaluation and development as a reliable tool for UC diagnosis using urine samples.