Circulating tumor DNA as a predictive biomarker for pathologic response after treatment with neoadjuvant immunotherapy for localized dMMR/MSI-H colorectal cancer.

Abstract

3612 Background: Neoadjuvant immunotherapy (IO) has shown promise as a potential non-surgical, curative therapy in pts with localized dMMR/MSI-H CRC. However, response assessment to IO is challenging due to difficulty measuring luminal tumors, the invasive nature of endoscopy, and known discordance between these assessment modalities. ctDNA may serve as an important, novel marker of response in this population. We evaluated pre-treatment ctDNA and ctDNA clearance and their association with pathologic complete response (pCR) in pts with localized dMMR/MSI-H CRC treated with IO. Methods: We performed a 70-gene, tumor-agnostic, NGS-based ctDNA assay (LB-70) to detect somatic mutations in a CLIA setting on 49 plasma samples collected serially from 13 pts with localized dMMR/MSI-H CRC treated with neoadjuvant pembrolizumab (mean and median cycles per pt was 8; range 1 - 16) prior to definitive surgery as part of a prospective phase 2 trial (NCT04082572). NGS was also performed on each pt’s tumor tissue to confirm mutations identified in ctDNA were tumor-specific. We defined “ctDNA clearance” as a reduction in mutational variant allele frequency (VAF) to a threshold below LB-70’s limit of detection (<0.3% VAF) after neoadjuvant IO in mutations that were present in both plasma and tumor tissue. We conducted a descriptive analysis to evaluate the association between the presence of pre-treatment (“baseline”) ctDNA, ctDNA clearance, and pCR. Results: On a median number of 3 samples per pt, mutations in baseline ctDNA were detected in 7/13 pts and were confirmed orthogonally in all pts with matched tumor tissue. The mean and median VAF was 8.5% and 0.7% (range 0.3% - 26.9%). Among 10/13 pts with a pCR, none recurred clinically after median follow-up 29.7 mo. Three pts did not experience a pCR, 1 of whom had disease recurrence after surgery. pCR was observed in 6/6 pts without detectable ctDNA at baseline, while only 4/7 pts with detectable ctDNA at baseline had a pCR (100% vs 57%). Of the 7 pts with detectable ctDNA at baseline, 5 cleared their ctDNA in the neoadjuvant period, among whom, 4 had a pCR (80% likelihood of pCR with ctDNA clearance). Notably, the 1 pt who did not have a pCR despite clearing ctDNA had a near-pCR (carcinoma identified in 1/91 lymph nodes, none in cecum or colon). The mean and median number of cycles to ctDNA clearance was 2.8 and 2 (range 1 - 8). For the 2 pts who did not clear their ctDNA, neither (0%) had a pCR. Conclusions: Clearance of ctDNA following neoadjuvant IO appears to predict pCR in this pilot analysis of pts with localized dMMR/MSI-H CRC, even after a median of 2 cycles. The absence of detectable ctDNA at baseline, perhaps a surrogate for lower tumor burden at initial diagnosis, may also predict a higher likelihood for pCR. Our work suggests utility of ctDNA as a biomarker for improving risk-stratification and response assessment sensitivity.