Influence of low-coverage whole-genome sequencing–based urinary tumor fraction and chromosomal instability on recurrence risk stratification in systemic immunotherapy-treated non–muscle–invasive bladder cancer.

4603 Background: Urinary tumor DNA (utDNA) is increasingly explored for prognostic assessment in bladder cancer, but most studies rely on a simple positive/negative classification, and its prognostic value in patients receiving systemic immune checkpoint inhibitors (ICIs) remains unclear. Low-coverage whole-genome sequencing (LC-WGS) of urine enables low-cost quantitative assessment of tumor fraction (TF) and chromosomal instability (CIN), providing richer molecular information than conventional binary utDNA status. We evaluated whether LC-WGS–derived TF and CIN improve prognostic stratification in ICI–treated high- and very high-risk non–muscle–invasive bladder cancer (NMIBC). Methods: Patients with high- or very high-risk NMIBC were enrolled in a prospective cohort treated with ICI tislelizumab plus low-dose nab-paclitaxel. Pretreatment urine samples were collected and analyzed using LC-WGS to derive TF and CIN. Clinicopathologic variables were collected at baseline. The primary endpoint was 2-year recurrence-free survival (RFS). Prediction models were constructed using clinicopathologic features (tumor size, number, concomitant carcinoma in situ, lymphovascular invasion) alone and in combination with TF and CIN. Results: A total of 35 patients were included. After a median follow-up of 27.3 months, 11 patients experienced disease recurrence. Models based on clinicopathologic features alone demonstrated limited discrimination for 2-year RFS (AUC = 0.61). Higher urinary TF and CIN were significantly associated with increased recurrence risk (TF: HR 2.18, 95% CI 1.29–3.97, p = 0.004; CIN: HR 2.74, 95% CI 1.41–5.32, p = 0.002). Incorporation of TF improved discrimination (AUC = 0.71), and further addition of CIN achieved the highest performance (AUC = 0.80; ΔAUC = 0.19 vs clinical model, p = 0.01). Patients classified as high molecular risk based on combined TF and CIN had significantly lower 2-year RFS than those at low molecular risk (HR 3.92, 95% CI 1.62–9.48, p = 0.002). Notably, the molecular-only model incorporating TF and CIN demonstrated discrimination comparable to the clinicopathologic-integrated model (AUC = 0.79 vs 0.80, p = 0.68), suggesting that urinary genomic features alone may provide sufficient information for recurrence risk stratification in this setting. Conclusions: Urinary TF and CIN derived from cost-effective LC-WGS substantially improved recurrence risk stratification beyond conventional tumor characteristics in NMIBC patients treated with systemic immunotherapy. These findings suggest that urine-based molecular profiling may offer a practical and scalable complement to conventional clinicopathologic risk assessment for baseline risk stratification and post-therapy management.