Abstract 2224: DELFI as a real-time treatment response assessment for patients with cancer

Abstract Introduction: Measurement of tumor-derived DNA molecules in the plasma (ctDNA) has become a useful tool to determine the overall tumor burden in patients with cancer. The ctDNA burden may change over time, decreasing after treatment response and increasing with development of resistance to therapy. Monitoring the dynamics of ctDNA burden over the course of treatment with a rapid, non-invasive test enables physicians to make timely treatment decisions. Using the DNA evaluation of fragments for early interception (DELFI) approach, we have developed the DELFI Monitoring Score (DMS) to longitudinally assess ctDNA burden during therapy of cancer patients. Methods: We performed low coverage whole genome sequencing on 302 cfDNA libraries of 76 treatment-naive metastatic colorectal cancer (mCRC) patients with at least one blood draw prior to and after treatment initiation. Mutations in KRAS, NRAS, or BRAF were independently measured by digital droplet PCR (ddPCR) for all timepoints evaluated. We trained a Bayesian regression model with the mutant allele fraction (MAF) of KRAS, NRAS, or BRAF as response and fragmentation-related features as predictors as well as a random intercept. To avoid overfitting and assess generalizability, we used cross-validation, wherein each patient’s samples were held out of the dataset and a model was trained on the samples of all other patients to generate predictions for the held-out samples. Patients with DMS below and above a detectability threshold at the first blood draw post-treatment (between 4-12 weeks) were classified as molecular responders and non-responders, respectively. Progression-free survival outcomes, defined as time to progression by RECIST 1.1 or death, were evaluated using a Kaplan-Meier estimator in these two groups. Results: Molecular responders based on DMS assessment experienced longer progression-free survival than molecular non-responders (p < 0.01), similar to a MAF-based approach. Additionally, in the 43 patients for whom the MAF of the KRAS, NRAS, or BRAF variants was 0% at the first blood draw post-treatment, we observed further separation on progression-free survival by the DELFI Monitoring Score (p = 0.014). Conclusions: We developed a novel quantitative measure of ctDNA burden using cell-free DNA fragmentation patterns that is estimable from low coverage whole genome sequencing. The DMS appears to be useful for measuring ctDNA burden and enabling a non-invasive approach to treatment monitoring, as it distinguishes therapeutic responses in a mCRC cohort. Our ability to detect differences in progression-free survival among patients with undetectable ddPCR MAFs suggests that DMS may be more sensitive and predictive than conventional mutation-based approaches for treatment response monitoring. Citation Format: Keith Lumbard, Laurel A. Keefer, Iris van't erve, Jacob Carey, Bryan Chesnick, Denise Butler, Michael Rongione, Cornelis J. Punt, Nicholas C. Dracopoli, Remond J. Fijneman, Gerrit A. Meijer, Robert B. Scharpf, Victor E. Velculescu, Alessandro Leal. DELFI as a real-time treatment response assessment for patients with cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2224.