Background: Plasma cell-free DNA (pcfDNA) has shown great promise for non-invasive, multi-cancer early detection (MCED), but has lower sensitivity for early-stage urological cancers due to low tumor fraction in plasma. Urine cfDNA (ucfDNA) has the potential to improve detection and monitoring of early-stage urological cancers due to its proximity to the affected organs and ease of collection. We conducted an exploratory study to assess the utility of methylation patterns in ucfDNA to detect BC in patients with suspicious bladder lesions, and compare to detection using matched pcfDNA. Methods: Urine and blood were collected from patients with suspicion of new (N=17) or recurrent (N=20) non-muscle invasive BC (NMIBC), and from non-cancer (NC) patients with urological conditions (N=16). Patients with suspicion of NMIBC were diagnosed and staged by transurethral resection of bladder tumor (TURBT) and conventional imaging. Tumor allele fraction (TAF) estimates from ucfDNA were inferred using a method trained on methylation patterns enriched in BC tissue (N=49) relative to an external reference dataset of NC ucfDNA (N=176). We set a detection threshold, using a maximum TAF value from a separate set of NC urine samples (N=50), to determine ucfDNA sensitivity for detecting BC in our study. Sensitivity in pcfDNA was determined using a validated MCED test classifier at 99% specificity. Results: Of 17 patients with suspicion of new NMIBC, 12 were diagnosed with BC after TURBT (Stage 0: N=6, I: N=5, II: N=1), and 10/12 were high grade (HG). Among patients with confirmed BC, ucfDNA sensitivity was 91.7% overall (11/12; 95% CI 61.5-99.8%) and 90% for HG (9/10). Whereas, pcfDNA sensitivity was 16.7% overall (2/12) and 10.0% for HG (1/10). Of 20 patients with suspicion of recurrent NMIBC, 14 were confirmed as BC (Stage 0: N=10, I: N=2, II: N=2) and 11/14 were HG. Sensitivity of ucfDNA for recurrence detection was 78.6% overall (11/14; 95% CI 49.2-95.3%), and 100% for HG (11/11), while pcfDNA sensitivity was 14.3% (2/14) overall and 18.2% (2/11) for HG. Notably, TAF in urine from NC patients (N=16) and patients with suspicion of new NMIBC found to be benign by TURBT (N=5) were all below the detection threshold. Among patients with suspicion of recurrent NMIBC but not found to have BC by TURBT, TAF estimates for 4/6 (66.7%) were above the detection threshold. Conclusions: We observed increased sensitivity in urine compared to matched plasma in patients with NMIBC, consistent with local shedding of bladder tumors into stored urine. A urine-based cfDNA assay with high sensitivity at high specificity, combined with non-invasive sampling, could be an ideal tool to use alongside the standard of care (e.g., cystoscopy) for clinical diagnosis and monitoring of BC. Further studies are needed to validate these findings and determine the clinical utility of ucfDNA in the diagnosis and surveillance of BC.