Real-world outcomes in gastrointestinal cancer patients with targetable genomic alterations identified on serial liquid biopsy.

4238 Background: Precision oncology in gastrointestinal (GI) cancers relies on identifying targetable genomic alterations (TGA) to guide therapy. Liquid biopsy (LB) using circulating tumor DNA (ctDNA) enables minimally invasive molecular profiling and serial monitoring. As targeted therapies for GI cancers expand, repeat LB is increasingly used to detect emergent alterations. However, the clinical relevance of TGA not previously identified, particularly when detected at low variant allele fraction (VAF) where clonal burden is difficult to infer, remains unclear. Methods: GI cancer patients (pts) with >1 LB were identified from the Guardant Health InfinityAI Data Library, a real-world database of insurance claims and de-identified clinical Guardant360 records. Pts without TGA on initial LB, but gained TGA on subsequent LB were included. TGA were defined as biomarkers with cancer-type specific or pan-tumor matched targeted therapy (MTT) approvals: KRAS G12C, ERBB2 amplification (amp), BRAF V600E, IDH1, MSI-H, and NTRK/RET/FGFR2 fusions or with select approved MTT in non-GI cancers: PIK3CA and ERBB2 mutation (mut). When available, methylation-derived tumor fraction (TF), a measure of overall ctDNA burden independent of variant-specific effects, was compared with emergent TGA VAFs, excluding MSI-H and ERBB2 amp. Pts with multiple cancers or prior MTT were excluded. Cox proportional hazards model assessed associations between MTT and real-world overall survival (rwOS), adjusting for age, gender, cancer type, MSI-H status, and number of gained TGA. Results: Among 16,435 GI cancer pts with >1 LB and no initial TGA, 2,072 (12.6%) gained TGA on serial LB and were included in the overall cohort. For the survival analysis, 804 pts with gained TGA had treatment and survival data; 13.4% (108/804) received MTT. Colorectal cancer pts had the highest rate of emergent TGA with approved MTT (17%). The most common gained TGA across the overall cohort were PIK3CA mut (62%), ERBB2 mut (11.7%), BRAF V600E (9.5%), KRAS G12C (8.6%), ERBB2 amp (7%), MSI-H (4.4%), FGFR fusions (3%), and NTRK fusions (1.9%). Gained TGA were often detected at low VAF (median 0.2%; 0.01% - 41.4%), and below the TF, suggesting potential subclonal emergence. Cholangiocarcinoma and gastroesophageal cancers had the highest MTT rates (14.2% and 10%, respectively), whereas no anal cancer pts received MTT. Including pts who gained MSI-H status, 12-and 60-month rwOS from LB-detected TGA was 67% and 25% without MTT versus 78% and 44% with MTT, p<10-4; excluding MSI-H, rwOS with MTT was 82% and 34%, p=0.00025. Conclusions: Repeat LB identified TGA in a subset of GI cancer pts lacking baseline TGA, many detected at low VAF. MTT receipt for emergent TGA was associated with improved rwOS, supporting serial LB in later-line decision making. Further studies are needed to assess outcomes by TGA type, therapy, clonality, and GI cancer.