Abstract 2219: Urine cell-free DNA multi-omics to detect molecular residual disease and predict survival in bladder cancer patients

Abstract Background: Circulating cell-free tumor DNA (ctDNA) analysis for minimal residual disease (MRD) detection is transforming cancer care. However, the sensitivity of these approaches remains subpar and there is significant risk of false negative results. We analyzed the most proximal biofluid (urine) in localized muscle-invasive bladder cancer (MIBC) patients, and performed two orthogonal ctDNA analytical techniques, one focusing on driver mutation detection, and another focusing on genome-wide copy number alterations. The primary objective of this combinatorial approach was to sensitively detect ctDNA MRD, predict pathologic complete response (pCR), and improve patient risk-stratification. Methods: Urine samples from 74 localized bladder cancer patients were collected preoperatively on the day of curative-intent radical cystectomy (RC) to assess urine tumor DNA (utDNA). We performed ultra-low pass whole genome sequencing (ULP-WGS) of urine cfDNA from all 74 patients as well as 15 healthy adults. Tumor fraction (TFx) level based on genome-wide copy number alterations was estimated using ichorCNA. Variant allele frequency (VAF) based on single-nucleotide variants (SNVs) was estimated by uCAPP-Seq. We also noninvasively inferred tumor mutation burden (iTMB). pCR was determined by surgical pathology. A random forest (RF) model with leave-one-out cross-validation (LOOCV) was utilized to predict disease status. Kaplan-Meier (KM) and Cox proportional hazards model survival analyses were performed to assess overall survival (OS) and progression-free survival (PFS). Results: Our study cohort consisted of 74 patients, of which 58 (78%) harbored localized MIBC, and 16 (22%) harbored treatment-refractory high-risk localized NMIBC. Among MIBC patients, 64% (37/58) received neo-adjuvant chemotherapy. VAF, iTMB and TFx levels significantly predicted pCR status. Our RF model incorporating these three utDNA parameters achieved a sensitivity of 87%, a negative predictive value of 77%, and a positive predictive value of 65% for predicting residual disease. AUC for the model was 0.80 (p<0.0001). KM analysis revealed that both PFS (HR = 3.00, P = 0.01) and OS (HR = 4.81, P = 0.009) were significantly worse for patients predicted by the model to have residual disease. Univariate and multivariate Cox proportional hazards models confirmed the significance of our MRD predictions. Survival analyses performed on MIBC, neoadjuvant chemotherapy, and held-out validation subgroups corroborated these findings. Conclusions: Integration of ULP-WGS with uCAPP-Seq enabled robust detection of residual urine tumor DNA with high sensitivity and predicted survival in localized bladder cancer patients. In the future, this type of multimodal urine-based genomic analysis may lead to more precise risk stratification and nonoperative clinical decision-making for bladder cancer patients. Citation Format: Pradeep Singh Chauhan, Alexander L. Shiang, Irfan Alahi, R. Taylor Sundby, Wenjia Feng, Bilge Gungoren, Cayce Nawaf, Kevin Chen, Ramandeep K. Babbra, Peter K. Harris, Faridi Qaium, Casey Hatscher, Anna Antiporda, Lindsey Brunt, Lindsey R. Mayer, Jack F. Shern, Brian C. Baumann, Eric H. Kim, Melissa A. Reimers, Zachary L. Smith, Aadel A. Chaudhuri. Urine cell-free DNA multi-omics to detect molecular residual disease and predict survival in bladder cancer patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2219.