UriBLAD - a urine-based gene expression assay for non-invasive detection of bladder cancer

Bladder cancer is the most common urinary system neoplasm, with approximately 550,000 new cases per year worldwide. Current methods for diagnosis and monitoring of bladder cancer are often invasive and/or lack sensitivity and specificity. In this study, we aimed to develop an accurate, noninvasive urine-based gene expression assay for the detection of bladder cancer. Urine specimens were collected at five Chinese hospitals from bladder cancer patients, and from healthy and other control subjects. The expression levels of 70 genes were characterized by quantitative reverse transcription-PCR in a training cohort of 211 samples. Machine learning approaches were used to identify a 32-gene signature to classify cancer status. The performance of 32-gene signature was further validated in a multicenter, prospective cohort of 317 samples. In the blind validation set, the 32-gene signature achieved encouraging performance: 90% accuracy, 83% sensitivity and 95% specificity. The area under the receiver operating characteristic curve reached 0.93. Importantly, the 32-gene signature performed well in the detection of non-muscle invasive tumor and low-grade tumor with the sensitivity of 81.6% and 81%, respectively. In conclusion, we present a novel gene expression assay that could accurately discriminate patients with bladder cancer from controls with urine samples. Our results might prompt the further development of this gene expression assay into an In-Vitro Diagnostic test amenable to routine clinical practice.