Detecting and monitoring bladder cancer with exfoliated cells in urine

Abstract

Current methods for the diagnosis and monitoring of bladder cancer are invasive and have suboptimal sensitivity. Liquid biopsy as a non-invasive approach has been capturing attentions recently. To explore the ability of urine-based liquid biopsy in detecting and monitoring genitourinary tumors, we developed a method based on promoter-targeted DNA methylation of urine sediment DNA. We used samples from a primary bladder cancer cohort (n=40) and a healthy cohort (n=40) to train a model and obtained an integrated area under the curve (AUC) > 0.96 in the 10-fold cross-validation, which demonstrated the ability of our method for detecting bladder cancer from the healthy. We next validated the model with samples from a recurrent cohort (n=21) and a non-recurrent cohort (n=19) and obtained an AUC > 0.91, which demonstrated the ability of our model in monitoring the progress of bladder cancer. Moreover, 80% (4/5) of samples from patients with benign urothelial diseases had been considered to be healthy sample rather than cancer sample, preliminarily demonstrating the potential of distinguishing benign urothelial diseases from cancer. Further analysis basing on multiple-time point sampling revealed that the cancer signal in 80% (4/5) patients had decreased as expected when they achieved the recurrent-free state. All the results suggested that our method is a promising approach for noninvasive detection and prognostic monitoring of bladder cancer.