Background: Clinical validation studies have shown that early dynamic changes in circulating tumor DNA (ctDNA) tumor fraction (TF) can predict clinical outcomes. Yet few studies have evaluated the clinical value of longitudinal monitoring throughout the course of treatment. Here we evaluate longitudinal changes in ctDNA TF and clinical outcomes in an advanced real-world pan-cancer cohort of patients (pts) treated with immune checkpoint inhibitors (ICIs).

Methods: The cohort included deidentified pts from the Tempus clinicogenomic database with stage IIIB or IV solid tumors who underwent ctDNA NGS and were treated with an FDA-approved ICI +/- chemotherapy (CT). Pts had a pre-treatment baseline liquid biopsy (T0) and ≥1 on-treatment sample Ti within 21-180 days of ICI. xM for treatment response monitoring estimates TF via an ensemble algorithm that incorporates variant and copy number information. Response status was determined at each on-treatment timepoint Ti relative to T0. Pts were classified as a Molecular Responder (MR) if TF < 1% at T0 and Ti or if TF decreased by ≥ 50% from T0 to Ti; otherwise, they were classified as a Molecular Non-Responder (nMR). Pts with TF < 0.09% at T0 and Ti were classified as ctDNA not detected. The longitudinal cohort included pts with ≥2 on-treatment timepoints, further classified based on the most frequent classification across each Ti as longitudinal MR, longitudinal nMR and longitudinal ctDNA not detected. If no classification was dominant, the most recent classification
was used. Real-world overall survival (rwOS) was defined from T1 to death and assessed by log-rank test.

Results: The full cohort of 84 pts with > 10 solid tumors included 34.5% (n = 29) NSCLC and 16.7% (n = 14) SCLC. The majority of pts (59.5%, n = 50) received ICI+CT, (64.0% first-line [1L]), and 40.5% (n = 34) ICI-only (35.3% 1L). rwOS was longer for 53 MRs vs. 18 nMRs (median not reached vs. 7.0 months, P < 0.005); 13 pts had no ctDNA detected. The longitudinal subcohort of 35 pts was 31.4% (n = 11) NSCLC, 20.0% (n = 7) prostate cancer, and 14.3% (n = 5) SCLC; 45.7% (n = 16) ICI+CT (56.3% 1L), 54.3% (n = 19) ICI-only (31.6% 1L). On average, pts had 3 on-treatment liquid biopsies, with a median time between on-treatment samples of 91 days. Twelve pts were longitudinal nMRs (10 MR or ctDNA not detected at T1), 18 longitudinal MRs, and 5 longitudinal no ctDNA detected. There were 6 death events in the longitudinal nMRs, 3 in the longitudinal MRs (all > 18 months after T1), and no death events in the longitudinal no ctDNA detected group. Longitudinal MRs had longer rwOS than longitudinal nMRs (median 37.4 months vs 8.7 months, P < 0.005).

Conclusions: Longitudinal nMR was associated with worse survival compared to longitudinal MR. Longitudinal, molecular biomarker dynamics may be a useful clinical treatment decision tool for monitoring treatment response to ICI therapy.

VIEW THE PUBLICATION

VIEW THE POSTER