MP59-01: Urine microRNAs as liquid biopsy assay for predicting response to intravesical BCG therapy

Anirban P Mitra*, Sharada Mokkapati, H Barton Grossman, Ashish M Kamat, Houston, TX, David J McConkey, Baltimore, MD, Colin P N Dinney, Houston, TX

Introduction: Intravesical bacillus Calmette-Guérin (BCG) with transurethral resection (TUR) is first-line bladder-preserving therapy for patients with intermediate- and high-risk non-muscle-invasive bladder cancer (NMIBC). Despite treatment, 30%-50% of patients fail to respond, and 15% progress to muscle-invasive disease. Molecular predictors of BCG response have been sparse, and recent studies have profiled primary tumors where only limited tissue may be available. We identify microRNAs (miRNA) in urine of patients with NMIBC prior to BCG therapy in an effort to identify predictive signatures.

Methods: Total RNA was isolated from frozen banked pretreatment urine specimens from patients undergoing TUR for NMIBC at our institution between 2005 and 2010 who subsequently received intravesical BCG. Patients were categorized as responders or nonresponders; the latter included those who were refractory to or relapsed after adequate BCG therapy. The amplification-free nCounter platform was used to profile miRNA, and samples that passed quality metrics were used for downstream analysis. 80 patients formed a discovery set (52 responders, 28 nonresponders), and 40 patients were used as a validation set (26 responders, 14 nonresponders). Random forest models with internal cross-validation were used to construct optimal classifiers in the discovery set that were blindly applied to the validation set.

Results: Median follow up was 9.1 years. There were no differences in demographic or clinicopathologic parameters between the discovery and validation sets (all, p = 0.50). Variance filtering was used to identify differentially expressed miRNAs in the discovery set. A 233-feature random forest classifier was constructed (discovery; sensitivity 81%, specificity 64%, negative predictive value, NPV, 64%) that was used to blindly interrogate a validation set (sensitivity 78%, specificity 53%, NPV, 64%). 35 highest-ranked features were used to construct another concise classifier (discovery set; sensitivity 85%, specificity 68%, NPV, 70%) that was similarly confirmed on the validation set (sensitivity 78%, specificity 46%, NPV, 71%).

Conclusions: We report identification and blinded validation of pretreatment urine-based miRNA signatures that predict BCG response in NMIBC. Pending large-scale prospective validation, such assays may serve as noninvasive liquid biopsy tests to identify NMIBC patients who may benefit from early consideration for alternate bladder-sparing therapies or more aggressive management.

Source of Funding: UT MD Anderson Cancer Center SPORE in Genitourinary Cancers.