Serial ctDNA evaluation to predict clinical progression in patients with advanced urothelial carcinoma.

Abstract

532 Background: It is not known whether serial circulating tumor DNA (ctDNA) can augment imaging for assessing treatment response in patients (pts.) with advanced urothelial carcinoma (UC). We hypothesized that serial ctDNA measurements predict the clinical progression of advanced UC and map its evolutionary trajectories. Methods: We analyzed 182 serial ctDNA samples from 53 pts. with advanced UC sequenced by the Guardant360 ctDNA platform. The aggregate variant allele frequency (aVAF) was defined as the sum of VAFs of all genomic alterations (GAs) in a ctDNA sample. Clinical response status was categorized into progressive disease (PD) and non-PD based on imaging. Progression-free survival (PFS) was defined as the time from treatment start till PD or death. The Mann-Whitney test was used to compare radiologic response status (PD vs. non-PD) and the ctDNA aVAF collected within 4 weeks of restaging imaging. Results: Pts. with lower initial ctDNA aVAF (≤0.2) had longer overall survival (OS) (hazard ratio (HR): 0.31, 95%CI: 0.11-0.90, p = 0.03). Pts. who achieved clearance of their ctDNA aVAF at the time of any subsequent ctDNA sample (n = 19/53) had longer OS (HR: 0.26, 95%CI: 0.08-0.85, adjusted p = 0.027). Combining ctDNA aVAF values from two consecutive samples improved the performance of a clinical prognostic model based on age, sex, and liver metastasis (C-statistic improved from 0.65 to 0.84). The mean ctDNA aVAF and the median number of GAs per ctDNA sample significantly increased at the time of PD vs. non-PD (12.31 vs. 2.10, p < 0.0001, and 3 vs. 1, p = 0.0006, respectively). Delta ctDNA aVAF increases predicted radiologic PD with an area under the receiver operating characteristic curve of 0.84 (95%CI: 0.65-0.95, p < 00.1). Delta ctDNA aVAF improved the patient risk stratification, pts with both decreased delta ctDNA aVAF and non-PD had the longest overall survival (Log-rank p = 0.05). A subgroup of 20 pts. had increasing ctDNA aVAF and ≥ 1 available subsequent radiologic scans within six months. Increasing ctDNA VAF ≥1 predicted PD in 90% (18/20) of pts with a median lead time of 92 days over imaging. APOBEC3-induced mutations (A3-m) were identified in ctDNA samples of (23%) 12/53 pts. Pts. with ctDNA A3-m had longer median PFS on immune checkpoint blockade than pts. without (17 vs. 3 months, Log-rank p = 0.01). There was no significant difference in OS according to ctDNA A3m status (Log-rank p = 0.37). Serial ctDNA provided insights into the clonal dynamics of treatment resistance, including the acquisition of ERBB2 S653C resistance mutation at the time of PD on lapatinib. Conclusions: Serial ctDNA predicts clinical outcomes and provides real-time assessment of treatment effectiveness in pts. with advanced UC. This has the potential to guide future adaptive therapy paradigms.