A noninvasive urine-based methylation biomarker panel to detect bladder cancer and discriminate cancer grade.

ABSTRACT

BACKGROUND: Highly sensitive and specific urinary biomarkers for the early detection of bladder cancer (BC) to improve the performance of urinary cytology are needed. OBJECTIVE: To investigate the usefulness of methylation markers in voided urine to identify BC presence and grade. DESIGN, SETTINGS, AND PARTICIPANTS: Using genome-wide methylation strategies in Toronto, Canada and Liège, Belgium, we have identified differentially methylated genes (TWIST1, RUNX3, GATA4, NID2, and FOXE1) in low-grade vs. high-grade BC tissue and urine. We accrued urine samples from 313 patients using a 21 ratio in a case-control setting from Toronto, Canada, Halifax, Canada, and Zurich, Switzerland. We studied the usefulness of these 5 methylated genes to identify BC and discriminate cancer grade in voided urine specimens. Urinary cell sediment DNA was evaluated using qPCR-based MethyLight assay. Multivariable logistic regression prediction models were created. RESULTS AND LIMITATIONS: We included 211 BC patients (180 nonmuscle invasive) and 102 controls. In univariate analyses, all methylated genes significantly predicted BC vs. no BC, and high grade vs. low grade (all P < 0.05). In multivariable analysis, NID2, TWIST1, and age were independent predictors of BC (all P < 0.05). Sensitivity of NID2 and TWIST1 to predict BC and BC grade was 76.2% and 77.6%, respectively, whereas specificity was 83.3% and 61.1%, respectively. Multivariable models predicting BC overall and discriminating between high-grade and low-grade BC reached area under the receiver operating characteristics curves of 0.89 and 0.78, respectively. CONCLUSIONS: This multi-centric study in a real life scenario (different countries, techniques, and pathologists) supports the promise of epigenetic urinary markers in noninvasively detecting BC. With sensitivities and specificities in the range of 80%, the overall performance characteristics of this panel of methylated genes probably does not allow such signature to significantly alter clinical care at this stage but is worth further studying for instance in BC surveillance or screening in high-risk populations.