Bladder cancer (BC) is diagnosed by cystoscopy, which is invasive, costly and causes considerable patient discomfort. MicroRNAs (miR) are dysregulated in BC and may serve as non-invasive urine markers for primary diagnostics and monitoring. The purpose of this study was to identify a urinary miR signature that predicts the presence of BC.
For the detection of potential urinary miR markers, expression of 384 different miRs was analyzed in 16 urine samples from BC patients and controls using a Taqman™ Human MicroRNA Array (training set). The identified candidate gene signature was subsequently validated in an independent cohort of 202 urine samples of patients with BC and controls with microscopic hematuria. The final miR signature was developed from a multivariable logistic regression model.
Analysis of the training set identified 14 candidate miRs for further analysis within the validation set. Using backward stepwise elimination, we identified a subset of 6 miRs (let-7c, miR-135a, miR-135b, miR-148a, miR-204, miR-345) that distinguished BC from controls with an area under the curve of 88.3%. The signature was most accurate in diagnosing high-grade non-muscle invasive BC (area under the curve = 92.9%), but was capable to identify both low-grade and high-grade disease as well as non-muscle and muscle-invasive BC with high accuracies.
We identified a 6-gene miR signature that can accurately predict the presence of BC from urine samples, independent of stage and grade. This signature represents a simple urine assay that may help reducing costs and morbidity associated with invasive diagnostics.