Quantitative baseline ctDNA as a prognostic biomarker in metastatic urothelial cancer.

Abstract

e16572 Background: Detectable circulating tumor DNA (ctDNA) in patients with metastatic urothelial cancer (mUC) is associated with poor outcomes. Whether quantitative ctDNA values confer independent prognostic information beyond conventional parameters is unclear. Methods: We retrospectively analyzed patients with mUC (urothelial carcinoma [UC], variant, or mixed histology) treated at our institution who had ctDNA measured using the tumor-informed Signatera assay within 21 days prior to initiating a line of systemic therapy. Patients could contribute multiple treatment lines. ctDNA levels were reported as mean tumor molecules per milliliter (MTM/ml). The primary endpoint was overall survival (OS) from therapy initiation. Cox proportional hazards models evaluated baseline ctDNA as: (1) a quantitative continuous variable (log10-transformed), and (2) a qualitative binary variable (detectable vs. undetectable). Univariate and multivariable models estimated hazard ratios (HRs) with 95% confidence intervals (95% CIs). Multivariable models were adjusted for line of therapy, therapy class, Karnofsky Performance Status (KPS) <80% vs. ≥80%, and visceral metastases. Model comparison used Akaike Information Criterion (AIC) and likelihood ratio testing to assess incremental prognostic value of quantitative vs. qualitative ctDNA. Histologic subtype effect modification was assessed via interaction terms. Results: Among 38 unique patients with mUC, we evaluated 48 pre-treatment ctDNA timepoints (UC: n=29; variant/mixed histology: n=9). Baseline ctDNA was detectable in 89.6% of observations (43/48), with quantitative values ranging from 0 to 5879.2 MTM/ml (median 13.1 MTM/ml, interquartile range 2.5-83.6 MTM/ml). In Cox models, higher quantitative ctDNA was independently associated with worse OS adjusted for line of therapy, therapy class, KPS, and visceral metastases (Table). Quantitative ctDNA provided superior prognostic discrimination versus binary detectability thresholds, with improved AIC (ΔAIC -10.0) and likelihood ratio testing (p=0.002). No histologic subtype effect modification was observed (interaction p=0.4). Conclusions: Quantitative baseline ctDNA levels provide superior prognostic risk stratification versus qualitative detectability thresholds across therapy lines and histologic subtypes in mUC, supporting its use for prognostic assessment and as a potential stratification factor for clinical trials. Association of baseline quantitative ctDNA with OS using Cox proportional hazards models (log10[ctDNA + 1]). HR >1 indicates increased risk of death. Model HR (95% CI) P-value Unadjusted 2.6 (1.5–4.4) <0.001 Adjusted for line of therapy 2.6 (1.5–4.3) <0.001 Adjusted for KPS 2.6 (1.5–4.5) 0.001 Adjusted for visceral metastases 2.7 (1.5–4.9) <0.001 Adjusted for therapy class 2.4 (1.4–4.2) 0.001 Fully adjusted 3.0 (1.5–5.8) 0.002