Introduction:

Repeat transurethral resection of bladder tumor (reTURBT) is recommended for high-risk non-muscle invasive bladder cancer (NMIBC) given the prevalence of residual disease as well as the potential for pathologic upstaging. However, upwards of 50% of patients lack disease on reTURBT. Urinary cell-free DNA (ucfDNA) based assays may help identify the presence of disease following index TURBT and identify optimal candidates for reTURBT. Given the higher sensitivity afforded by tumor informed approaches, we sought to explore the genomic concordance of index and reTURBT specimens. In addition, we evaluated the utility of low-pass whole genome sequencing (LPWGS) to predict residual disease at reTURBT, which is less resource-intensive when compared to whole exome sequencing. 

Methods:

Fifty-one patients undergoing repeat TURBT within 12 weeks of an initial diagnosis of high-risk NMIBC were prospectively enrolled at Moffitt Cancer Center. Tumor tissue was collected from index and repeat TURBT. Peripheral blood for germline analysis and urine was obtained prior to reTURBT. Tumor tissue and blood buffy coat were subject to whole-exome sequencing at ~20,000X with PredicineWES+ and ucfDNA underwent LPWGS (PredicineSCORE 1-3X). Non-synonymous mutations and significant gene copy-number variation (CNVs) were identified and compared between index and reTURBT specimens with matched germline control. Bespoke urinary tumor fraction (TF) based on tumor variants was calculated using Predicine DeepSea pipeline algorithm. 

Results:

Thirty patients had residual disease on reTURBT and 21 did not. Overall average variant concordance between index and repeat tumors was 74%. 29 out of 30 patients with disease detected at reTURBT had novel tumor variants in the reTUR tissue. Index tumors had a median 161.5 mutations (range 1-1384) and 9 CNVs (range 1-73). ReTUR tissue had a median of 43 mutations (range 2-1113) and 3 CNVs (range 1-21). Mean urinary TF from LPWGS was significantly higher in patients that were disease positivie at reTURBT (12.2% vs. 4.9% Wilcoxon, p=0.002). At a threshold TF of 2.2%, we had 81% sensitivity at 71% specificity to predict residual disease at reTUR in this cohort (AUC: 0.76).  

Conclusion:

Index and reTURBT tumor tissue have high variant concordance, demonstrating the rationale in utilizing tumor informed ucfDNA assays to identify residual disease prior to repeat transurethral resection. Novel mutations also arise in almost all patients highlighting the potential utility of personalized MRD tracking prior to reTURBT. We demonstrate that urinary TF derived from LPWGS of ucfDNA currently provides moderate sensitivity to detect disease prior to reTURBT. With further development, these two methodologies have strong potential to select optimal candidates for reTURBT in high risk NMIBC. 

Funding: N/A