Real-world evaluation of a “hyperselection” predictive biomarker for anti-EGFR therapy benefit in patients with RAS/RAF wild-type, MMRp/MSS mCRC.

e15591 Background: The PARADIGM trial showed an OS benefit to 1L doublet chemo+anti-EGFR vs doublet+anti-VEGF treatment (tx) in patients with RAS wild-type (WT) metastatic colorectal cancer (mCRC) with left-sided primary tumors and some evidence of harm in those with right-sided tumors. Subsequent ctDNA analysis found that a “hyperselection” (HS) biomarker was predictive of anti-EGFR tx benefit, including within subgroups with left- and right-sided tumors. We evaluated the predictive value of a tissue HS biomarker for anti-EGFR benefit in a real-world setting. Methods: We performed a retrospective cohort study using the nationwide Flatiron Health-Foundation Medicine (FMI) mCRC clinico-genomic database. The de-identified data originated from ~280 US cancer clinics. Adults with RAS/RAF WT, MMRp/MSS mCRC who received 1L tx including chemo+ anti-VEGF or anti-EGFR tx from 1/2012 - 12/2024 and who underwent FMI comprehensive genomic profiling on tissue collected pre-1L tx were eligible. Patients without alterations in the HS panel ( ERBB2 amp or activating alterations, MET amp, NTRK/ROS1/RET rearrangement, PIK3CA exon 20/ PTEN inactivating/ AKT1/EGFR ECD alterations) were classified as HS-; those with any alteration were HS+. Multiple imputation was performed for missing values of pre-specified covariates. Multivariable Cox proportional hazards modeling of the association of anti-EGFR vs anti-VEGF tx with OS that included imbalanced covariates and the interaction between tx and HS status was performed in the overall population. Separate models were performed within subgroups defined by tumor sidedness and HS status with adjustment for imbalanced covariates, when possible. Results: 1,332 patients met inclusion criteria, of which 1,155 were HS- and 177 were HS+ (Table). In the multivariable model, the interaction between HS status and tx was not significant (p = 0.51). In subgroup models, there was an 5% reduction in hazard of death for anti-EGFR vs anti-VEGF tx among HS- patients (HR 0.95, 95% CI 0.80 – 1.15) and an 8% increased hazard of death among HS+ patients (HR 1.08, 95% CI 0.69 – 1.69); neither result was statistically significant. Within tumor sidedness subgroups, similar patterns were observed (Table). Conclusions: While there may be some evidence of a differential tx effect by HS status, power to demonstrate predictive effect was limited by the low frequency of alterations in the panel. Among patients with RAS/RAF WT, MMRp/MSS mCRC, tissue HS was not predictive of 1L anti-EGFR tx benefit. Subgroup multivariable models. Hyperselection Negative (n= 1,155) Hyperselection Positive (n=177) Population Anti-VEGF Anti-EGFR HR for OS (95% CI) Anti-VEGF Anti-EGFR HR for OS (95% CI) Overall 870 285 0.95 (0.80-1.15) 135 42 1.08 (0.69 – 1.69) Left-Sided 703 255 0.88 (0.72 – 1.07) 95 34 1.15 (0.68 – 1.97) Right-Sided 167 30 0.94 (0.54 – 1.63) 40 8 1.78 (0.53 – 5.92)