Predicting Recurrence Using a Tumor-Uninformed ctDNA Assay Detecting MRD in Patients With Resected Stage II or III Colorectal Cancer: Subset Analysis From the GALAXY Study in CIRCULATE-Japan

ASCO Gastrointestinal Cancers Symposium 2024 Presentation
Authors Yoshiaki Nakamura, Kristiyana Kaneva, Christine Lo, Daniel Neems, Jonathan Freaney, Hala Boulos, Seung Won Hyun, Eiji Oki, Daisuke Kotani, Jun Watanabe, Masahito Kotaka, Hideaki Bando, Kentaro Yamazaki, Ichiro Takemasa, Takeshi Kato, Chithra Sangli, Rick Blidner, Kate Sasser, Halla Nimeiri, and Takayuki Yoshino

Background: The presence of circulating tumor DNA (ctDNA) after curative-intent surgery can identify patients with minimal residual disease (MRD) who are at risk of recurrence and aid with postsurgical risk stratification. ~80% of stage II and only 50% of stage III colorectal cancer (CRC) patients are cured by surgery alone. A tumor-uninformed plasma-only approach for MRD assessment accelerates the turnaround time, enabling rapid adjuvant chemotherapy (ACT) treatment decision making, optimizing the impact on patient outcomes.

Methods: A subset of samples from GALAXY, an observational arm of the CIRCULATE-Japan study (UMIN000039205), were analyzed for postsurgical ctDNA detection at a landmark time point (LMT) defined as 4 weeks after curative surgery in pathological stage II or III CRC patients. In a prospective lead-in to clinical validation, 80 patients meeting pre-specified eligibility criteria were randomly selected with enrichment for recurrence to 50% while maintaining the stage II:III recurrent (R)/nonrecurrent (NR) ratio to that observed in GALAXY. Residual plasma samples were analyzed with the Tempus xM assay (xM), a tumor-uninformed ctDNA MRD assay that integrates methylation and genomic variant MRD classifiers to deliver a binary MRD call. All calls were blinded to clinical outcomes. The methylation-basedclassifier detects fragments with CRC methylation signatures in differentially methylated regions trainedby sequencing CRC and presumed-healthy samples on a 6Mbp panel. The variant-based MRD classifier detects highly prevalent CRC variants. Artifacts, CHIP, and germline variant filtering algorithms were also applied.

Results: Of the 80 patients (pts), 70 [Stage II n= 29 (41%), Stage III n= 41 (59%)] were evaluable (36 R and 34 NR). At the LMT and with at least 1-year follow-up, 18/36 R pts had detectable ctDNA (xM MRD+) and 30/34 NR pts had undetectable ctDNA (xM MRD-), providing a sensitivity of 50% and specificity of 88%. Additionally, xM MRD+/- status strongly correlated with disease-free survival (DFS) with a hazard ratio (HR) of 5.09, adjusted to account for a 24% recurrence rate in GALAXY. In an updated, ongoing longitudinal analysis of the 4 false positive pts, 2 were Stage III, MRD+ from both the methylation and genomic classifiers, received ACT, and have likely cleared their ctDNA without any sign of clinical recurrence. An adjusted sensitivity and specificity analysis resulted in xM clinical performance of 53% (20/38 R) and 94% (30/32 NR), respectively.

Conclusions: xM is a novel tumor-uninformed assay that demonstrated remarkable performance to detect clinical recurrence at a LMT, and a clinically meaningful correlation with DFS. A larger clinical validation study including longitudinal analysis, correlation with ACT, and outcomes from the GALAXY study is currently underway.