228 Background: We previously conducted a phase I/Ib study with regorafenib and nivolumab in patients with refractory metastatic mismatch repair proficient (pMMR) colorectal cancer (CRC). This study aimed to investigate the biomarkers that predict the treatment response. Methods: Out of the 51 patients who received regorafenib and nivolumab, 22 archival pretreatment tumor samples were subjected to the Xerna TME Panel, a machine learning-based RNA-sequencing biomarker assay and were classified into one of four TME biomarker subtypes: Angiogenesis (A), Immune Active (IA), Immune Desert (ID), or Immune Suppressed (IS). Potential predictive biomarkers including the TME subtypes, KRAS (wild type vs mutant), PD-L1 (negative vs. positive, samples with > 1% tumor cells for PD-L1 were considered positive), CD8 expression (low vs. high), and Treg cells (low vs. high) in tumor microenvironment were evaluated for correlation with overall survival (OS), progression free survival (PFS) and disease control rate (DCR, defined as complete response + partial response + stable disease). Results: Among the 22 patients, 16 (72.7%) had liver metastasis and 15 (68.2%) had lung metastasis. KRAS mutation was found in 16 (68.2%) patients. 11/21 (52.4%) were positive for PD-L1. 12 (54.5%) had high CD8 expression, whereas 9/21 (42.9%) had high Treg cells in tumor microenvironment. Ten (45.5%) patients were classified as biomarker-positive (IA + IS subtypes) and 12 (54.5%) were biomarker-negative (A + ID) based on Xerna TME panel. Two (9.1%) patients achieved partial response, 12 (54.5%) had stable disease, and five (22.7%) developed progressive disease. The median PFS was 5.6 months and median OS was 13.1 months. No significant correlation was observed between RAS mutation (p = 0.664, p = 0.609), PD-L1 expression (p = 0.287, p = 0.173), CD8 (p = 0.152, p = 0.456) and PFS or OS. Low Treg was found to be associated with prolonged PFS (median: 9.8 vs. 1.9 months, p = 0.011) but not OS (p = 0.280). Similarly, only low Treg level was related with DCR (83.3% vs. 33.3%, p = 0.032). While not reaching statistical significance, Xerna TME biomarker-positive patients showed trends for higher median PFS (7.9 months vs. 4.1 months, p = 0.254), median OS (15.75 months vs. 11.9 months, p = 0.378), and higher DCR (70% vs. 58%, p = 0.675) compared to biomarker-negative patients. Additionally, the two patients with partial responses were Xerna TME biomarker-positive. Conclusions: Our study demonstrated that low Treg in tumor microenvironment is correlated with better prognosis in patients with refractory metastatic pMMR CRC who were treated with regorafenib plus nivolumab. Xerna TME panel analysis of these patients also showed trends for predictive clinical benefit. Prospective and larger cohort studies are needed to better define predictive biomarkers for this combination in the future.

188 Background: Our previous phase I/Ib study revealed modest clinical efficacy of rego/nivo in refractory MSS-CRC, implying benefits in selected populations. This has prompted us to investigate predictive biomarkers. Methods: Pre-treatment tumor samples from the previous phase Ib rego/nivo study were obtained and assessed for mRNA sequencing (RNAseq). Response-associated transcripts were identified using DESeq2 method and annotated using gene ontology (GO) and Kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis. Results: A total of 19 pretreatment tumor samples were analyzed including 14 patients (pts) with disease control (DC) (2 partial response and 12 stable disease) and 5 pts with progression disease (PD). We observed significant upregulation of 89 genes including SFTPB ( P<0.001), SFTPA1 ( P<0.001), ERICH6B ( P<0.001) , XIST ( P<0.001), IGHV2-26 ( P<0.001), RETNLB ( P=0.005), VWA5B1 ( P=0.033), IGHV3-53 ( P<0.001), POTEH ( P=0.015), IGHV1-3 ( P=0.001) and downregulation of 70 genes including AFAP1-AS1 ( P=0.022), PRSS21( P<0.001), NTSR1 ( P<0.001), CALB1 ( P=0.033), FAM83A ( P=0.013), TBL1Y ( P=0.045), WNT7A ( P=0.022), PADI1 ( P<0.001), KRT6A ( P=0.030), KRT17 ( P=0.002) on RNA sequencing in the DC compared to PD. GO pathway enrichment analyses revealed 44 significantly upregulated pathways including positive regulation of B cell activation ( P<0.001), B cell receptor signaling pathway ( P<0.001), Phagocytosis engulfment ( P<0.001), Fc-gamma receptor signaling pathway involved in phagocytosis ( P<0.001), endocytosis ( P<0.001), immunoglobulin receptor binding ( P<0.001), serine-type endopeptidase activity ( P<0.001) in the DC compared with PD group. Nine downregulated pathways including cell-substrate junction assembly ( P<0.001), extracellular matrix (ECM) organization ( P=0.030), and ephrin receptor activity ( P=0.046) were observed in the DC compared with PD. KEGG pathway enrichment analysis revealed 8 significantly downregulated pathways including PI3K-Akt signaling ( P=0.009), ECM-receptor interaction ( P=0.006), Focal adhesion ( P=0.028), and Glycine, serine and threonine metabolism ( P=0.007) pathways in the DC compared with PD. Conclusions: RNAseq and following gene set enrichment analysis between pts with DC and PD revealed several dysregulated pathways involving immune regulation including macrophage and B cell activation, and also PI3-Akt signaling, ECM organization/receptor interaction, and adhesion, which are consistent with previous translational studies in other solid tumors. These dysregulated genes and pathways related to differentially expressed genes may need to be further evaluated as potential predictive biomarkers for rego/nivo or other tyrosine kinase inhibitor plus PD-1 blockade in MSS-CRC.

1567 Background: Objective and complete toxicity reporting in clinical trials is critical for patient-centered shared decision-making. Conference abstracts inform initial impressions of practice-changing treatments. Methods: We performed a systematic review of all abstracts of CRC and PaC phase 3 RCTs presented at ASCO annual meetings between 2012 – 2022; long-term follow-up, supportive care, and solely non-pharmacological studies were excluded. Objective minimization of adverse event (AE) reporting was defined as absent and/or incomplete reporting of cumulative grade 3-5 CTCAE (common terminology criteria for AE). We also assessed the use of subjective minimizing language (Chin-Yee et al ASH 2022), defined as use of “acceptable,” “tolerable”, “manageable”, “favorable” (primary minimization terms), or “feasible”, “safe”, “patients did well”, “limited” (secondary minimization terms), terms that falsely imply patients deemed the therapy as such. Presence/absence of PRO or QOL data was also assessed. Results: 63 RCTs met entry criteria (42 CRC, 21 PaC), detailed in Table. Most trials studied chemotherapy +/- other drugs (52; 83%). 17% of all abstracts did not provide any information on AE. Quantitative data on AEs were reported by 38 (60%) of abstracts. However, serious AE reporting was frequently absent (Table), with some trials reporting only specific toxicities (e.g. cytopenias) instead of cumulative CTCAE. Only 7 (11%) of abstracts noted the occurrence or absence of fatal AE. Any subjective-minimizing language was used in 15 (24%) abstracts. Notably, none of the abstracts using subjective-minimizing language provided information on fatal AE rates, nor reported on the patient perspective via QOL or PRO. Average grade ≥ 3 AE in the experimental arm were similar in abstracts with vs without minimizing language (44% vs 45%). Conclusions: Our systematic review of ph 3 RCTs in GI oncology presented at ASCO annual meetings reveals that subjective minimizing language is often used to describe serious toxicities, and without formally assessing the patient voice. Serious AE reporting is frequently absent or incomplete.[Table: see text]