Artificial intelligence-assisted immunohistochemical (IHC) evaluation of tumor amphiregulin (AREG) and epiregulin (EREG) expression as a combined predictive biomarker for panitumumab (Pan) therapy benefit in RAS wild-type (wt) metastatic colorectal cancer (mCRC): Analysis within the phase III PICCOLO trial

111 Background: High tumor mRNA levels of the EGFR ligands, AREG and EREG are associated with anti-EGFR agent response in patients (pts) with RAS-wt mCRC, regardless of tumor location. However, ligand RNA assays have not been adopted into routine clinical practice due to issues with analytical precision and practicality. Here we test whether AREG and EREG expression assessed by IHC can predict benefit from Pan. Methods: A retrospective biomarker study within the PICCOLO trial (NCT00389870; irinotecan [Ir] ± Pan in fluoropyrimidine-resistant RAS-wt mCRC). AREG and EREG positive tumor cells were assessed by IHC in all RAS-wt patients with available tumor tissue. Pathologists annotated tumor areas on digital images of glass slides. Artificial intelligence (AI) algorithms calculated the percentage of tumor cells staining positive for AREG and EREG within the tumor areas. More than 50% AREG and/or EREG tumor cell positivity was regarded as high ligand expression. The primary endpoint was progression-free survival (PFS) and secondary endpoints were RECIST response rate (RR) and overall survival (OS). Results: 274 RAS-wt pts had available tumor tissue. High ligand expression (n = 132) was associated with significant PFS benefit from IrPan compared with Ir (8.0 vs 3.2 months; HR 0.54 [0.37-0.79]; p = 0.001); whereas low ligand expression (n = 142) was not (3.4 vs 4.4 months; HR 1.05 [95% CI, 0.74-1.49]; p = 0.78). The ligand-treatment interaction was significant (p = 0.02) and independent of BRAF-mutation status and primary tumor location. Likewise RR was significantly improved in pts with high ligand expression (IrPan vs Ir: 48% vs 6%; risk ratio, 7.8 [2.90-20.69]; p < 0.0001) but not those with low ligand expression (IrPan vs Ir: 25% vs 14%; risk ratio, 1.8 [95% CI, 0.89-3.65]; p = 0.10) (interaction p = 0.01). Lesser effect was seen on OS. Conclusions: IHC assessment of AREG and EREG identified pts who did or did not benefit from Pan, as has been previously demonstrated through mRNA quantification. IHC represents a more practicable technique as it can be provided at the point of care and is associated with shorter turn-around times. AREG and EREG IHC may be of use in routine practice to identify patients who would benefit from anti-EGFR therapy and those for whom alternative treatment strategies should be explored.