Search

186 Background: Actionable mutations can predict therapeutic benefit in patients with advanced malignancies, though clinical relevance of fusion testing for advanced colorectal cancer (aCRC) remains undefined. Identification of fusions from circulating tumor DNA (ctDNA) has previously been restricted to defined oncogenic fusion partners. To improve the sensitivity for fusion detection, we evaluated a partner-agnostic fusion analysis from ctDNA of patients with aCRC. Methods: De-identified data from Guardant Health was reviewed for 18,558 patients with aCRC who underwent ctDNA NGS testing by Guardant360 (Redwood City, CA) between 2017-2022. Fusion results were analyzed with a partner-agnostic bioinformatic approach. A fusion was defined as “clonal” if the variant allele frequency (VAF) ratio exceed ≥50% of highest somatic VAF, and “subclonal” if < 50% maxVAF. Microsatellite instability (MSI) status [MSI-high (bMSI-H) or microsatellite stable (bMSS)] and anti-EGFR exposure signature were determined using prior methods. Associations between fusion occurrence and coexisting alterations were performed using Fisher’s exact test. Results: Fusions were detected in 221 (1.2%) of patients with aCRC. 258 activating fusions were detected in 187 patients; FGFR3 (N = 59, 23%), RET N = 55, 21%), BRAF (N = 43, 17%), and ALK (N = 41, 16%) were most frequent. There were 71 previously unreported fusions in 28 additional patients; RET (N = 16; 23%), MET (N = 15, 21%), and BRAF (N = 11; 15%) were most prevalent. Clonal fusions occurred in 7% (18/258) of all activating fusions; RET (5/18, 28%) and FGFR3 (3/18, 17%) were most common and associated with bMSI-H status relative to bMSS (27% vs 4%, OR 8.165, 95% CI 2.332-33.99; p = 0.0076). Clonal fusions occurred less commonly in samples with a prior EGFR signature (OR 0.22, 95% CI 0.05-0.997, p = 0.049). Most detected fusions were subclonal including ALK, FGFR1-3, MET, RET and ROS1. Conclusions: Highly specific partner-agnostic fusion detection is feasible to increase sensitivity of ctDNA assay performance. Oncogenic fusions occurred in ~1% of all patients with aCRC. Clonal fusions as oncogenic drivers were infrequent and associated with bMSI-H status. Subclonal fusions were more common and occur in a setting consistent with prior exposure to anti-EGFR therapies. Reporting fusion partners and clonality from ctDNA may guide oncologists on the appropriate context for consideration of fusion-directed treatments.

3585 Background: Immune checkpoint blockade therapy improves survival in patients (pts) with microsatellite instability-high (MSI-H) advanced colorectal cancer (CRC). Oncologists often discontinue immunotherapy after 2 years of disease control based on prior trial data. Recurrence outcomes following discontinuation of immunotherapy and clinicopathologic features associated with recurrence remain underreported given the recent advent of these agents for pts with MSI-H advanced CRC. Methods: Records from pts with MSI-H CRC from MD Anderson Cancer Center who received immunotherapy between 2015-2022 and stopped after clinical benefit were reviewed. Median survival was estimated according to the Kaplan-Meier method. Associations between the event of recurrence and coexisting mutations ( KRAS, NRAS, BRAFV600E, PIK3CA, APC, TP53, POLE/POLD), metastatic site (lung, liver, lymph nodes, or peritoneum), primary tumor sidedness (right vs. left colon), and prior immunotherapy (anti-PD-(L)1 alone or with anti-CTLA-4 antibodies) were measured by Fisher’s exact tests. Results: Thirty-six pts with MSI-H CRC without progression on immunotherapy were reviewed. Of these 29 and 7 received anti-PDL1 antibody alone or in combination with anti-CTLA-4 antibody, respectively. Median exposure to prior immunotherapy was 24 months (range, 5-43). After a median follow-up of 19 months (95% CI, 14-26) after stopping immunotherapy, 30 of 36 pts (83%) remained without disease progression. For the 6 patients with progression after stopping, median time to relapse was 13 months (range, 5-31). Median disease-free survival (DFS) was not reached. The estimated 1-year, 2-year, and 3-year DFS probabilities were 90% (95% CI, 79-100), 79.1% (95% CI, 64-98), and 68% (95% CI, 47-98), respectively. Median overall survival from the time that immunotherapy was stopped was 54 months (95% CI, 47-NA). Only 1 pt died due to unrelated illness. There were no observed associations between disease recurrence and co-existing mutations, metastatic organ involvement, primary tumor sidedness, or immunotherapy used. Conclusions: Most pts with MSI-H advanced CRC who achieve initial clinical benefit and do not progress on immunotherapy do not recur after treatment is stopped. Our data suggest that favorable outcomes do occur following cessation of immunotherapy in this setting even with concomitant prognostically unfavorable clinical features (RAS, BRAFV600E mutations; liver, peritoneal metastases).

3598 Background: Treatment with encorafenib (E) and cetuximab (C) offers response and survival benefit for patients (pts) with MSS, BRAFV600E metastatic colorectal cancer (CRC). BRAF + EGFR inhibition induced a transient MSI-H phenotype in preclinical models of MSS, BRAFV600E CRC and may prime these tumors for response to immunotherapy with anti-PD-1 antibodies like nivolumab (N). Methods: In this single-arm, single-institution, phase I/II clinical trial, pts with treatment-refractory MSS, BRAFV600E metastatic CRC were eligible. No prior BRAF, MEK, or ERK inhibitors, anti-EGFR antibody, or immunotherapy was permitted. Pts received E (300 mg PO daily), C (500 mg/m2 IV q14 days), and N (480 mg IV q28 days). The primary endpoints were best overall response (RECIST 1.1) and safety/tolerability (CTCAE v5). A Simon two-stage design (H0: p≤.22; Ha: p≥.45, where p = percentage of pts with radiographic response) was employed using a one-sided α =.05 and β =.20. Median progression-free survival (PFS) and overall survival (OS) were estimated via Kaplan-Meier. To measure ex vivo treatment responses with an E-slice assay (EMPIRI), 300 µm fresh tissue slices from core biopsies were generated and cultured in serum-free media with E, C, and N. Longitudinal changes in viability were measured at days 4, 8, and 12 and compared to baseline viability in each tissue. Ex vivo “response” was defined if < 1X baseline tumor cell viability. Results: With a data cutoff of 2/8/2022, all pts are enrolled: 26 evaluable for toxicity and 23 for response. Median age is 60 years (range, 32-85), and 16 (62%) are female. Grade 3-4 treatment-related adverse events (AE) have occurred in 5/26 (19%) patients: colitis, maculopapular rash, leukocytosis, and myositis/myocarditis (all N = 1); asymptomatic elevated amylase/lipase (N = 2). Overall response rate is 48% (95% CI, 27-69), and disease control rate is 96% (95% CI, 78-100). Median PFS is 7.4 months (95% CI, 5.6-NA). For the 11 pts with responses, median duration of response is 7.7 months (95% CI, 4.5-NA). Median OS is 15.1 months (95% CI, 7.7-NA). E-slices showed concordance between pts with radiographic responses and reduction in cell viability, and between non-responders and increase in cell viability. Final results will be presented. Conclusions: E + C + N appears to be effective and well-tolerated for pts with MSS, BRAFV600E metastatic CRC. Ex vivo analysis of pretreatment tissue predicted eventual clinical response in matched patients. A follow-up randomized phase II trial (SWOG 2107) to evaluate encorafenib/cetuximab with or without nivolumab in pts with MSS, BRAFV600E metastatic CRC will activate in 2022. Clinical trial information: NCT04017650.

3520 Background: In RAS/RAF WT colorectal cancer (CRC), rechallenge with anti-EGFR therapy (EGFRi) in patients (pts) with prior response leads to clinical benefit, with response rates up to 30% in prior trials. However, secondary MTs in the MAPK signaling pathway have been implicated in resistance to EGFRi. We designed a phase 2 trial to evaluate the efficacy of EGFRi rechallenge +/- a MEK inhibitor (trametinib) based on pre-treatment ctDNA MTs. Methods: This trial evaluated the efficacy and safety of EGFRi rechallenge +/- trametinib in pts with RAS/BRAF WT, MSS, treatment refractory mCRC who achieved clinical benefit with prior EGFRi based therapy for ≥16 weeks with subsequent progression. Pre study ctDNA was used to enroll in one of 3 arms: Arm A: Pts with an acquired EGFR ECD MT but absence of RAS/BRAF/MAP2K1 or with absence of any acquired resistance MT (Arm C) at time of study initiation received panitumumab 6 mg/kg IV Q2 wks. Arm B: Pts with an acquired RAS/BRAF/MAP2K1 MT received panitumumab 4.8 mg/kg plus trametinib 1.5 mg PO daily. Pts in Arms A and C were allowed to cross over on progression. The primary endpoint was ORR by RECIST v1.1. Results: 54 pts were enrolled, with 52 evaluable for efficacy. Median age is 59 yrs (range, 37-78), and 23 (46%) are female. Median number of prior therapies was 3. Three, 20, and 31 pts were enrolled in Arms A, B, C, respectively. Grade 3 TREAs occurred in 29 (54%) pts (all receiving the doublet regimen) and included acneiform rash in 17 (31%) and others occurring in < 5% of pts. There were no grade 4 TRAEs. In pts with no acquired MTs (Arm C), ORR was 20% (6/30) (95% CI, 0.07-0.37), DCR 67% (20/30) (95% CI, 0.45- 0.81), and median PFS and OS 4.1 mo and 11.2 mo, respectively. The median DOR was 5.5 mo. 22 patients crossed over to add trametinib at time of progression, without any responses. In contrast, in pts with acquired RAS/RAF/MAP2K1 MTs (Arm B), there were no responses, with DCR of 63% (12/19) (95% CI, 0.36-0.81), and median PFS and OS 2.1 mo and 5.9 mo, respectively. Only 3 pts were identified with EGFR ECD MTs (Arm A), and ORR is 0% (0/3) in this cohort, with DCR 67% (2/3) (95% CI, 0.09-0.99). Pts with PR had a longer median interval from prior EGFRi and longer time on prior EGFRi than those with SD+PD (5.5 vs 3.6 mo; p = 0.03, and 9.5 vs. 8.8 mo; p = 0.03, respectively). Conclusions: CtDNA guided rechallenge leads to responses in 20% of pts without acquired resistance MTs, with DCR of 67%. This exceeds current third line standard options. While panitumumab has the potential to block EGFR ECD mutations arising from cetuximab, these mutations in isolation were uncommon and there were no signals of efficacy. Although the acneiform rash induced by the combination of MEK and EGFR inhibition was manageable with close dermatologic management, the combination failed to improve outcomes for pts with acquired resistance. Alternative approaches to downstream MAPK blockade should be explored to improve outcomes. Clinical trial information: NCT03087071.

3554 Background: The conventional theory for the development of treatment resistance to anti-EGFR for metastatic colorectal cancer (mCRC) is the selective growth advantage of pre-existing therapy-resistant subclones with genomic mechanisms such as RAS mutations, leading to treatment resistance and disease progression. However, the impact of cytotoxic chemotherapy in combination with anti-EGFR on the mechanisms of resistance has not been assessed. Methods: We analyzed paired plasma samples from RAS/BRAF/EGFR wild-type mCRC patients enrolled in three large randomized phase 3 trials of anti-EGFR rechallenge in whom paired baseline and time of progression plasma samples had been collected for sequencing of ctDNA on a platform optimized for very low allele frequencies. 569 patients had paired baseline and progression ctDNA samples analyzed, including 147 in the first line study of FOLFOX +/- panitumumab, 91 patients in third line with panitumumab vs best supportive care, and 331 patients in the third line study of cetuximab vs. panitumumab. The mutational signature of the alterations acquired with therapy was evaluated. We also established colon cancer cell lines with resistance to cetuximab, FOLFOX, and SN38, and profiled transcriptional changes. Results: Using serial plasma samples, we demonstrate that patients whose tumors were treated with and responded to anti-EGFR alone were approximately 5-times more likely to develop acquired mutations at progression compared to those treated with an EGFR inhibitor in combination with cytotoxic chemotherapy (46% vs. 9%, respectively; p < 0.001). Consistent with this clinical finding, cell lines with non-genomic acquired resistance to cetuximab were cross-resistant to cytotoxic chemotherapy and vice-versa, with transcriptomic profiles consistent with epithelial to mesenchymal transition. In contrast, common acquired genomic alterations in the MAPK pathway that drive resistance to EGFR monoclonal antibodies do not impact sensitivity to cytotoxic chemotherapy. Further, contrary to the generally accepted hypothesis of clonal expansion of acquired resistance, in our work we demonstrate that baseline resistant subclonal mutations rarely expanded to become clonal at the time of progression (8%), and most remained subclonal (44%) or disappeared (49%). Conclusions: Collectively, this work outlines a model of resistance where non-genomic mechanisms of resistance common to both EGFR inhibitors and cytotoxic chemotherapy predominate in patients treated with EGFR and chemotherapy combinations. With EGFR inhibitor monotherapy, genomic acquired resistance mechanisms predominate, although only rarely through expansion of pre-existing subclones. These findings have important implications for strategies of EGFR-inhibitor rechallenge studies.

3588 Background: The success of tailored systemic therapies in treating distinct molecular subsets of patients (e.g., deficient mismatch repair, BRAF mutant, HER2 amplified) has spurred further exploration of novel targetable subsets within the heterogeneous landscape of metastatic colorectal cancer (mCRC). Human epidermal growth factor receptor 3 [HER3 (ErbB3)], a member of the HER (ErbB) receptor tyrosine kinase family, plays an important role in tumorigenesis and metastases and has emerged as a promising therapeutic target in a diverse array of cancers. For example, patritumab deruxtecan (U3-1402; HER3-DXd) is a HER3-directed antibody drug conjugate that has demonstrated clinically meaningful antitumor activity and acceptable safety profiles in metastatic breast cancer and EGFR-mutated non-small cell lung cancer. There is limited data, however, on the clinicopathological characterization of HER3 expression in mCRC. Methods: Tissue samples (surgical-metastatectomy) (N = 115) were obtained from a clinical cohort of patients (N = 99) with histologically proven mCRC and liver metastases who underwent liver resection with/without perioperative systemic chemotherapy. HER3 expression was analyzed on whole-mount preparations by immunohistochemistry (IHC). Staining was performed and visualized using the HER3 (D22C5) XP Rabbit-mAb (Cell Signaling Technology). Patients were categorized based on membranous intensity score as follows: Low with IHC 0 (absence of staining or staining in < 10% of tumor cells), 1+ (faint/barely perceptible staining in ≥10% of tumor cells) or 2+ (weak to moderate staining in ≥10% of tumor cells), or High with IHC 3+ (strong staining in ≥10% of tumor cells). Clinicomolecular and treatment data, including gender, tumor sidedness, mutational status (RAS or BRAF), and prior chemotherapy were collected by review of patient electronic medical records. Chi-squared (or Fisher’s exact) test were used to determine associations between groups. Overall survival (OS) was calculated using Kaplan-Meier method and compared using log-rank tests. Results: Among 99 analyzed patients, 98 were evaluable for HER3 expression. Of these 25.5%, 26.5%, 40.8% and 7.2% showed HER3 IHC scores of 3+, 2+, 1+ and 0, respectively. No significant association was seen with HER3 expression and clinicopathological variables, mutational status, or prior treatment. Among patients with 2 samples analyzed from the same liver surgery, there was a moderate level of heterogeneity with concordance of 78.5% (kappa 0.43). Patients with high HER3 expression had poorer OS (5-year OS: 52%; median: 90.2 months) compared to low HER3 expression (5-year OS: 85%; median: not reached). Conclusions: In this large cohort of mCRC, HER3 expression was observed in 92.8% of patients and across diverse clinical and molecular features, supporting HER3 as a promising targetable biomarker in a large subset of mCRC.

3524 Background: Markers of systemic inflammation including neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and monocyte-to-lymphocyte ratio (LMR) are prognostic in patients with metastatic colorectal cancer receiving systemic chemotherapy. The presence of liver metastases has also been hypothesized to modulate response to immunotherapy. In this study, we assess the prognostic role of these markers in patients with microsatellite high (MSI-H)/deficient mismatch repair (dMMR) tumors receiving immunotherapy for metastatic or unresectable colorectal cancer (CRC). Methods: This was a single-institution retrospective analysis of patients with dMMR/MSI-H CRC who received anti-PD-(L)1 and/or anti-CTLA-4 therapy for metastatic or unresectable disease at between 2015 and 2021 (n = 59). NLR, PLR, and LMR were calculated based on the complete blood count obtained within 1 week prior to treatment. Patient and tumor characteristics were obtained from the clinical record. Patient characteristics were compared using Fisher’s exact test and Mann-Whitney U where appropriate. Progression free survival (PFS) and overall survival (OS) were the primary endpoints and log-rank test was used for comparison of survival distribution among groups. Results: 59 patients with metastatic dMMR/MSI-H CRC were identified. Median age was 60, 53% (n = 31) had right-sided tumors, 35% (n = 35) of patients with testing available had RAS-mutated tumors, and 37% (n = 22) received prior chemotherapy. Most common sites of metastatic disease were peritoneum (n = 23, 39%) and liver (n = 17, 29%). Patients were divided into NLR-High (NLR ≥ 3, n = 20) and NLR-Low (NLR < 3, n = 39), and both groups had similar baseline characteristics. The rate of progressive disease as best response was not different in NLR-Low versus NLR-High (15% vs 30%, p = 0.3). At a median follow-up of 32 months, neither median PFS nor median OS were reached. 74% (n = 29) remained progression free at 1 year in the NLR-Low group versus 60% (n = 12) in NLR-High group which was not statistically significant (p = 0.37); 90% (n = 35) remained alive at 2 years in the NLR-low versus 80% (n = 16) in the NLR-High group (p = 0.4). Similarly, using a cut-off of 150 and 3 for PLR and LMR respectively, there was no significant difference between PFS at 1 year in the PLR-Low (n = 32) vs PLR-High (n = 27) (66% vs 74%, p = 0.58) and LMR-Low (n = 35) vs LMR-High (n = 24) (60% vs 83%, p = 0.084) groups. The presence of liver metastasis or the presence of a RAS mutation did not influence PFS at 1 year (p = 0.35 and p = 1.00, respectively). Conclusions: Markers of systemic inflammation may have a limited prognostic role for patients with dMMR/MSI-H CRC receiving immunotherapy.

3592 Background: While BRAFV600E accounts for the majority of BRAF mutations in mCRC, non-V600 BRAF variants (a BRAF) have emerged in recent years as a distinct molecular subtype. There are no consensus recommendations regarding management. This study provides a comprehensive profile of a BRAF, their clonalities and co-mutations in mCRC using a large genomic database as well as a prospective treatment cohort of patients with a BRAF and mCRC managed at a single center. Methods: A systematic analysis was performed of patients with mCRC who underwent ctDNA testing (Guardant360 platform, Guardant Health) from September 2014 to May 2021. A variant was defined as clonal if the mutant allele frequency (MAF) was greater than 50% of the highest somatic MAF in the sample; otherwise it was defined as subclonal. Co-mutation analysis was conducted with BRAF, KRAS, NRAS, NF1, ERBB2, PIK3CA and SMAD4. Treatment history and overall survival (OS) for patients with a BRAF mCRC from MD Anderson Cancer Center were included. Results: 1,733 out of 14,742 mCRC patients had at least one BRAF variant, including 6.5% of patients with BRAFV600E variants and 6.2% with a BRAF variants (1.1% with class II, 1.9% with class III, and 3.2% with unclassified variants). 431 unique BRAF variants were identified in a total of 1,905 BRAF variants. BRAF class II and III variants showed a higher rate of co-occurring KRAS mutations (25.6% and 21.5%) and co-occurring NRAS mutations (5.8% and 2.7%) compared with BRAFV600E variants (2.4% for KRAS and 0.1% for NRAS); however, co-occurring KRAS G12C was only noted in one patient. In our MDACC cohort, 38 patients were included in the analysis. The median age was 55, 81% were Caucasian, and 74 % had left sided primary tumors (45% rectal, 24% sigmoid) with 37% being exposed to at least 2 lines of therapy. The most common mutations in clinical practice were class III, D594G (39%), followed by class II G469A (10%), & class III G466E (7%). The median follow-up time was 23.8 months (mo). While there were no survival differences between a BRAF classes II and III, there was a significant difference in OS in patients with RAS co-mutation (28.3 mo vs not reached [NR], p = 0.05) or liver involvement (28.8 mo vs NR, p = 0.02). Patients < 50 years of age had extremely poor survival with OS of 16.3 mo (vs. NR) and HR 7.51 (95% CI 1.82-31.0, p = 0.005). Treatment with anti-EGFR or use of metastasectomy was not associated with improved survival. Conclusions: a BRAF mutations have historically been considered a favorable prognostic marker in mCRC. Co-mutation with RAS is frequent for both classes and portends poor survival in our real-world cohort. Furthermore, early onset a BRAF mCRC is associated with more aggressive disease. These factors highlight the need for dedicated clinical trials for this unique subset of mCRC and may represent an opportunity to improve management in early onset colorectal cancer.

22 Background: Inflammatory bowel disease (IBD) increases the risk of developing colorectal cancer (CRC), and colitis-associated CRC (CA-CRC) mortality is on the rise. It has been postulated that CA-CRC may be contributing to the increasing prevalence of early-onset CRC (EOCRC) but supportive studies are currently lacking. Molecular and clinical differences between CA-CRC and sporadic-CRC (S-CRC) have been reported, however outcomes for CA-CRC remains unclear. Signet ring cell carcinoma (SRC) is a rare subtype of CRC which is seen at higher frequencies, along with mucinous histology, in both CA-CRC and EOCRC. In this study, we validate the association of SRC and mucinous (SRC/M) histology with CA-CRC and EOCRC, and utilize it to estimate the amount of EOCRC attributable to undiagnosed or subclinical IBD. Methods: A retrospective study was conducted using three independent mCRC patient datasets from MDACC. The mATTACC discovery cohort consisted of 32 IBD- and 425 S-mCRC patients enrolled in a prospective biomarker trial. Validation of tumor histology was completed with a tumor registry (n=1696), excluding the MSI-High samples, and a real-world evidence (RWE) cohort from MDACC containing 269 CA-mCRC and 29,596 S-mCRC patients, was used as our validation cohort. Results: In the mATTACC cohort SRC/M histology was found in 37.5% of CA-mCRC and 11.7% of S-mCRC, showing a strong association between SRC/M and CA-mCRC (OR = 4.54, 95% CI: 2.19-9.43). The RWE cohort confirmed the correlation of SRC/M with CA-mCRC (28.6%) relative to S-mCRC (11.4%) patients (OR = 3.13, 95%CI: 2.39-4.09). An association was found between SRC/M and EOCRC (OR = 1.35; 95% CI: 1.24-1.47). By comparing the prevalence of SRC/M in EOCRC and late-onset CRC and correcting by the proportion of CA-CRC cases with SRC/M histology, we estimate that between 8.28% to 10.15% of EOCRC may attributable to undiagnosed/subclinical IBD. Using the RWE cohort, median overall survival was determined to be lower for CA-mCRC (31m) relative to S-mCRC (39m; p=0.007), yielding a HR of 1.26 (95% CI: 1.06-1.48). CA-mCRC patients with EOCRC (25m) were also found to have significantly worse outcomes than S-mCRC patients (40m) with EOCRC (p=0.0005; HR = 1.61, 95%CI: 1.23-2.11). Within CA-mCRC, patients with SRC or SRC/M histology (21m) had decreased OS compared to mucinous histology (51m), indicating the poor prognosis of SRC in CA-mCRC (p=0.028; HR=0.53, 95% CI: 0.3-0.94). Conclusions: Tumor biology consistent with CA-CRC, including SRC/M histology, may be present in 8.3% – 10.2% of patients with EOCRC without a clinical diagnosis of IBD, and harbors worse outcomes. Although other confounding biology may be underlying this association, recognition of undiagnosed IBD in CRC patients, especially those with metastatic disease, is important as it may impact prognosis and treatment strategies for this high-risk patient population.

12 Background: Encorafenib (E) and cetuximab (C) offers short-lived response and survival benefit for patients (pts) with MSS, BRAFV600E metastatic colorectal cancer (CRC). BRAF + EGFR inhibition induced a transient MSI-H phenotype in preclinical models of MSS, BRAFV600E CRC and may prime these tumors for response to immunotherapy with anti-PD-1 antibodies like nivolumab (N). Methods: In this single-arm, single-institution, phase I/II clinical trial, pts with treatment-refractory MSS, BRAFV600E metastatic CRC were eligible. No prior BRAF inhibitors, anti-EGFR antibody, or immunotherapy was permitted. Pts received E (300 mg PO daily), C (500 mg/m2 IV q14 days), and N (480 mg IV q28 days). The primary endpoints were best overall response (RECIST 1.1) and safety/tolerability (CTCAE v5). A Simon two-stage design (H0: p≤.22; Ha: p≥.45, where p= percentage of pts with radiographic response) was employed using a one-sided α=.05 and β=.20. In the first stage, ≥ 4/15 responses were needed in order for the trial to enroll 11 additional pts. Median progression-free survival (PFS) and overall survival (OS) were estimated via Kaplan-Meier. Results: All 26 pts have been enrolled - 23 patients treated, and 21 evaluable for response so far. Median age is 59 years (range, 32-85), and 14 (54%) are female. No dose-limiting toxicities occurred. Grade 3-4 treatment-related adverse events (AE) occurred in 4/22 (18%) patients. Grade 3 AEs included colitis, maculopapular rash, leukocytosis, and elevated amylase/lipase (all N=1). Grade 4 AEs in a single patient were myositis/myocarditis. Overall response rate is 45% (95% CI, 23-68), and disease control rate is 95% (95% CI, 75-100). Median PFS is 7.3 months (95% CI, 5.5-NA). Median OS is 11.4 months (95% CI, 7.6-NA). For the 9 pts thus far with responses, median duration of response is 8.1 months (95% CI, 7.3-NA). Updated results will be presented. Conclusions: E + C + N is effective and well-tolerated for pts with MSS, BRAFV600E metastatic CRC. The E+C+N regimen met its predefined efficacy endpoint and suggests a role for immunotherapy as a novel combination approach for this specific subpopulation of MSS metastatic CRC. A follow-up randomized phase II trial (SWOG 2107) to evaluate encorafenib/cetuximab with or without nivolumab in pts with MSS, BRAFV600E metastatic CRC will activate in early 2022. Clinical trial information: NCT04017650.

3594 Background: Recently, in metastatic colorectal cancer (mCRC), the detection of RAS mutations by circulating tumor (ct) DNA has recently emerged as a valid and non-invasive alternative approach, overall showing a high concordance with the standard tissue genotyping, giving information on response to EGFRi treatment and resistant mechanisms. However, RAS mutations may be missed due to low levels of any ctDNA in the blood (false-negative), and it has been difficult to distinguish this from patients without a RAS mutation in the tumor (true-negative). We propose a methodology that can be applied to multi-gene ctDNA testing panels to accurately distinguish true- and false-negative tests. Methods: 357 subjects with tissue and multi-panel ctDNA testing from MD Anderson (MDACC) were used as a training dataset and 295 subjects from Massachusetts General Hospital (MGH) dataset as the testing dataset. CtDNA panels contained between 65 and 70 genes, allowing evaluation of tumor ctDNA shedding from variant allele fraction (VAF) levels in the plasma from other genes (such as APC and TP53). Based on the relationship between KRAS and the VAFs of other gene, we established a Bayesian model providing a posterior probability of false negative in the ctDNA test, using thresholds of < 5% (low), 5-15% (medium), and > 15% (high). This model was validated on the MGH database. Results: Across both cohorts, 431 patients were ctDNA wild type for KRAS. Of those, 29 had tissue documenting a KRAS mutation for a false negative rate of 8%. The model provides the posterior probability that a KRAS mutation is indeed present in the tissue given the observed values of allele frequencies for other mutated genes in the plasma. In the validation cohort, a predicted low false negative had no false negatives (0/62, 95% CI 0%-5.8%), while a predicted medium false negative rate was associated with 3% false negative (1/32, 95% CI 0%-16%). In contrast, a high predicted false negative rate was associated with 5% false negative (5/100, 95% CI 1.6%-11%). The results demonstrate the ability of our tool to discriminate between subjects with true negative and false negatives, as a higher proportion of false negatives are observed at higher posterior probabilities. Conclusions: In conclusion, our approach provides increased confidence in KRAS ctDNA mutation testing in clinical practice, thereby facilitating the identification patients who will benefit from EGFR inhibition while reducing the risk of false negative tests. Extension of this methodology to NRAS and BRAF is possible, with clinical application enabled by a freely available online tool.

3514 Background: Colorectal cancers (CRC) lacking RAS MTs treated with EGFRi are thought to evolve by a repetitive process of genetic diversification and clonal evolution. Acquired MTs in KRAS, NRAS, BRAF, MAP2K1, and EGFR are known mechanisms of acquired resistance in the EGFRi refractory population. However, the prevalence of MTs in the first line (1L) setting is not well established as most experience with EGFRi has been beyond the 1L setting. Methods: We analyzed paired plasma samples from RAS/BRAF/EGFRWT mCRC patients (pts) enrolled in 3 large randomized phase 3 trials who had been treated with EGFRi and in whom paired baseline (BL) and time of progression (PRO) plasma samples had been collected for sequencing of ctDNA on a platform optimized for very low allele frequencies (Plasma Select-R™ and Resolution Bio™). Prevalence of MTs at BL and PRO from a 1L study (‘203; FOLFOX ± panitumumab) were compared with 2 studies in the third line setting (3L; ‘007; panitumumab + best supportive care [BSC] vs BSC; and 3L; ‘763; panitumumab vs. cetuximab), to assess the frequency of acquired resistance mutations via ctDNA analysis. Results: For pts with available paired plasma samples (n = 112 for ‘203; n = 89 for ‘007; n = 274 for ‘763), acquisition of at least one KRAS, NRAS, BRAF, MAP2K1, or EGFR MT was significantly less common in post-progression samples in the EGFR containing arms of the 1L ‘203 study compared to the 3L ‘763 and ‘007 studies (6.8% vs 50.4% vs 39.6%, respectively; p < 0.001). In the non EGFR containing arms of the ‘203 and ‘007 study, the rate of acquired MTs was 7.5% and 0%, respectively (p = 1). While this difference in the rate of acquired MTs between the EGFR and non EGFR containing arms was statistically significant for the 3L study (p < 0.001) it was not significant for the 1L study. Further, pts on both 3L studies treated with EGFRi who experienced CR, PR or SD acquired more MTs than those who had PD as best response (53.6% vs 33.3%, respectively; p < 0.001). This relationship was not significant in the 1L setting (7.7% vs 0%; p = 1). Subclonal MTs (rMAF < 25%) in KRAS, NRAS, EGFR, BRAF and MAP2K1 were present at BL in 129 pts (27%). Based on the hypothesis that EGFRi is selecting for rare existing mutated cells in the tumor, we would expect expansion of any preexisting subclones in the BL samples. However, in contrast to expectations, these subclones rarely expanded to become clonal at the time of progression (12.4%). Conclusions: In contrast to expectations, acquired KRAS, NRAS, BRAF, EGFR, or MAP2K1 MTs rarely develop after 1L therapy. While selective pressure appears to increase the frequency of acquired MTs in the 3L setting, preexisting subclonal MTs do not appear to be the dominant source of acquired MTs at progression, implying that there may also be a transient mutational process driving resistance rather than expansion of preexisting clones. These findings have significant implications for ongoing and planned EGFRi rechallenge studies.

3572 Background: Serial ctDNA can measure dynamic changes in disease burden over time, however utility of serial profiling to detect changes in actionable alterations remains unclear. Methods: We evaluated 501 patients with ≥3 serial Guardant360 assays performed between 09/2016 and 11/2020 and compared MSI, fusion, amplification and single nucleotide variant (SNV) detection over time. This comprised 2147 assays with a median of 4 assays per patient (min 3, max 18) occurring an average of 163 days apart (+/- SD of 147 days). Maximum detected variant allele frequency in samples (maxVAF) was assessed for relation to changes in detected alterations as a surrogate for tumor volume. Results: Among 406 patients with assays assessable for MSI-status, 17 (4.2%) had MSI detected. New MSI detection on a subsequent assay always occurred with a rising maxVAF (3/3) that was also ≥0.7%, while loss of detectable MSI between assays always associated with falling maxVAFs (7/7) with 6/7 occurring when maxVAF fell below 0.4%. Fusions were noted in 9/501 (2%) patients. Among 3 patients who lost a detectable fusion, maxVAF decreased in 1 patient and changed ≤0.2% between assays in 2, while 2/3 patients with new fusions had rising maxVAFs and 1 patient had a falling maxVAF. Amplifications were detected in 242/501 patients (48%). While most genes had highly variable amplification detection between assays (9% serially detected), ERBB2 amplifications were more consistent and serially detected in 39% of detected cases (P < 0.0001). New detection of amplifications occurred more commonly in cases with rising maxVAF (OR 11.70, 95% CI 7.61-18.00, P < 0.0001) and loss of detectable amplifications occurred more between samples with falling maxVAF (OR 12.37, 95% CI 8.35-18.66, P < 0.0001). Change in maxVAF correlated with change in number of detected amplifications (r = 0.62, P < 0.0001), but only partially explained changes seen (R2= 0.39). Between serial assays, SNVs changed a median of 0 variants (IQR -1 to 1), however some patients had significant changes (max gain 21/max loss 18). Among 1646 serial time points, 454 (28%) had no change in SNVs, 674 (41%) gained SNVs, and 518 (31%) lost SNVs on subsequent assays. Gains were more common in samples with rising maxVAF (OR 7.76, 95% CI 6.18-9.73, P < 0.0001) while losses were more common when maxVAF fell (OR 6.90, 95% CI 5.47-8.66, P < 0.0001). The correlation between maxVAF change and SNV change was significant (r = 0.29, P < 0.0001), but minimally explained SNV changes (R2= 0.086) and was a much weaker association than noted for amplification changes. Conclusions: We noted significant differences in detection of actionable alterations across serial ctDNA assays. Increased ctDNA volume (higher maxVAF) due to tumor progression may explain some variation over time, but variability also occurs outside these changes, likely reflecting clonal evolution following therapy.

119 Background: Molecular profiling is critical for oncologists in personalizing treatment decisions for patients (pts) with metastatic colorectal cancer (mCRC). In contrast to archival tumor tissue specimens classically used profiling, sequencing of circulating tumor DNA (ctDNA) is more sensitive at quantifying low mutation allele frequencies and characterize “real time” tumor biology. We assessed the relationship between detection of BRAFV600E mutations in ctDNA and the clinical management of pts with mCRC. Methods: We retrospectively analyzed mCRC patients evaluated at MD Anderson Cancer Center with BRAFV600E mutations on ctDNA. ctDNA was isolated and sequenced for somatic mutations using a 70-gene next-generation sequencing assay (MD Anderson/GuardantHealth LB70 panel). Variant allele frequency (VAF) was characterized as the ratio of mutant reads: total reads for a given gene. BRAFV600E mutations were classified as “clonal” if the relative VAF (rVAF) exceeded 50% of the maximum VAF. “Major” and “minor” subclonal mutations were called for a rVAF of 10-50% and < 10%, respectively. Associations between BRAFV600E clonality and treatment decision were performed using a Fisher’s exact test. Survival outcomes were estimated using the Kaplan-Meier method. Results: 64 patients with mCRC had a BRAFV600E mutation detected in ctDNA. Concordance between tissue and ctDNA for BRAFV600E mutation was occurred in 44/55 (80%) patients with evaluable tumor specimen. There were 9 patients with BRAFV600E mutations identified in the absence of evaluable tumor tissue. Median VAF for BRAFV600E in the ctDNA was 3.6% (interquartile range, 0.50 – 17%). The majority of patients had a clonal BRAFV600E mutation (50/64, 78%). There were 3 (5%) and 11 (17%) patients with major subclonal and minor subclonal BRAFV600E mutations, respectively. Among patients with minor subclonal BRAFV600E mutations, 91% (10/11) had developed resistance to anti-EGFR therapies for management of RASwild-type mCRC. Discordance between tissue and ctDNA BRAFV600E status was associated with minor subclones (odds ratio (OR) 56, p < .0001). Clonal BRAFV600E mutations in the ctDNA were associated with a higher likelihood for treatment with BRAF targeted therapies (OR 5.8, p = .008). Median progression-free survival among 37 evaluable patients was 6.4 months. Conclusions: Reported VAF in the ctDNA served to stratify BRAFV600E according to relative clonality. Lower VAF was linked to acquired resistance to anti-EGFR therapies, whereas higher VAF was associated with receipt of matched targeted therapies for BRAFV600E mCRC. ctDNA technologies for identifying BRAFV600E mutations are feasible and informative for conducting relevant molecular profiling for patients with mCRC.

TPS143 Background: Patients with KRAS and NRAS ( RAS) wild-type mCRC benefit from the epidermal growth factor receptor (EGFR) monoclonal antibodies (Abs) panitumumab and cetuximab, but nearly all patients experience resistance. Blood-based profiling of cell free DNA (cfDNA) can identify genomic alterations that drive acquired EGFR Ab resistance. After discontinuation of anti-EGFR Abs, acquired genomic alterations decay over time to undetectable levels. Some studies have suggested clinical benefit from EGFR Ab rechallenge, but there is limited evidence that EGFR Ab rechallenge improves survival compared to standard of care (SOC) therapies. We hypothesize that cfDNA profiling will identify patients appropriate for panitumumab rechallenge, and that these molecularly selected patients will have improved survival compared to current SOC therapies. Methods: This is a randomized phase II, open label study designed to compare the overall survival (OS) of panitumumab rechallenge versus SOC (investigator choice TAS-102 or regorafenib). Secondary objectives include comparisons of progression free survival, objective response rate, clinical benefit rate, and quality of life as measured by the linear analogue self-assessment (LASA) questionnaire. Eligible patients have radiographically measurable KRAS, NRAS, and BRAF codon 600 wild-type mCRC based on tumor tissue testing, and must have experienced progression or intolerance to treatment with a fluoropyrimidine, oxaliplatin, irinotecan, an anti-VEGF Ab, and an anti-PD-1 Ab if the tumor has mismatch repair deficiency or is MSI-H. Progression after at least 4 months treatment with an anti-EGFR Ab is required. All patients must be enrolled in the COLOMATE cfDNA screening protocol (NCT03765736) and meet molecular eligibility based on Guardant360 cfDNA profiling (absence of amplification of ERBB2, KRAS, NRAS, and MET; absence of mutations of BRAF, EGFR, ERBB2, KRAS, NRAS, and MET [mutant allele frequency > 0.5%]). Greater than 90 days must have elapsed between the most recent treatment with an anti-EGFR Ab and cfDNA profiling. Dosing for all study drugs is according to clinical SOC. 120 patients will be randomized 1:1 to panitumumab rechallenge or SOC. With 83 OS events, this study will have 80% power to detect an improvement in median OS from 6.5 to 10 months (HR=0.65; 1-sided α= 0.15). This study began enrollment in 6/2020. Recruitment is ongoing at 16 sites in the Academic and Community Cancer Research United (ACCRU) network (ACCRU-GI-1623). Clinical trial information: NCT03992456.

4047 Background: BRAFV600E mutations occur in fewer than 10% of all patients (pts) with metastatic colorectal cancer (mCRC) and arise from sessile serrated adenomas. Despite efficacy with targeted therapies against MAPK signaling and with immunotherapies in this population, survival outcomes for pts with BRAFV600E mCRC in general are poor. Characteristics distinguishing pts with BRAFV600E mCRC with favorable versus unfavorable outcomes have not been well annotated. Methods: Records of 188 pts with BRAFV600E mCRC evaluated at MD Anderson Cancer Center between 3/2010-1/2020 were reviewed. Pts with the shortest and longest metastatic survival (N = 25 for each group) were compared. Associations between prognostic group and clinical/pathologic features were measured by odds ratio and for median survival by log-rank testing. Results: Median metastatic survival differed between the 2 BRAFV600E mCRC populations (8.6 vs 84 months, p < .0001). Pts with poor survival more commonly had primary tumors arising from the hepatic flexure/proximal transverse colon (44% vs 16%, p = .04) and more frequent hepatic involvement (75% vs 28%, p = .001). Pts with favorable survival were more likely to develop metachronous metastases (52% vs 16%, p = .01), have fewer distant organ involvement (median 1 vs 2, p = .02), and undergo definitive locoregional therapy to metastatic disease (44% vs 0%, p = .01). Microsatellite instability (36% vs 4%, p = .008) and a history of tobacco use (44% vs 16%, p = .04) were associated with a favorable prognosis. Durable responses to MAPK-targeted therapies (5/25) and immunotherapy (3/25) were noted in the favorable group. Conclusions: Pts with BRAFV600E mCRC can achieve excellent long-term survival which belies conventional context and is driven by locoregional and systemic treatment options alike. Anatomic localization of the primary tumor and prior exposures may highlight environmental influences on tumor biology which account for the clinical heterogeneity of pts with BRAFV600E mCRC.

4084 Background: Consensus Molecular Subtype 4 (CMS4) colorectal cancer (CRC) features increased TGFβ signaling, which may account for de novo resistance to immunotherapy for patients (pts) with microsatellite stable mCRC. To date, no prior trial has incorporated CMS status as an integral biomarker. Bintrafusp alfa (M7824) is a dual PD-L1 antibody/TGFβ trap with acceptable safety. Methods: Primary tumors from pts with metastatic CRC underwent CMS testing in a CLIA setting. In this Simon two-stage phase II trial (Ho: p < .05; Ha: p≥.25) for CMS4 mCRC, pts received bintrafusp alfa 1200mg IV every 14 days. RT (8Gy/day x 3 days) to a single metastatic lesion with abscopal intent was administered between doses 2 and 3. The primary objective was to estimate response rate (RR) per iRECIST. Correlative studies including RNA sequencing were performed on pre- and on-treatment biopsies. Results: 53 of 137 tested pts (39%) between June 2018-December 2019 had CMS4 mCRC. 13 of 15 treated pts received the agent with RT. All pts were evaluable for toxicity, and 13 for response. Median number of doses was 3 (IQR, 2-4). There was one grade 3 immune-related adverse event (colitis) requiring study discontinuation. There were 2 pts with stable disease and 11 with progressive disease as best response (RR 0%, 95% CI 0-22%). Enrollment was stopped after first stage for futility. Median PFS and OS were 1.6 months and 5.0 months, respectively. In paired samples, treatment with bintrafusp alfa resulted in an increase in the expression of IFNγ signature in nonirradiated metastatic lesions ( p< .001, q< .001). Updated results will be presented. Conclusions: This is the first reported clinical trial to utilize CMS status as an integral biomarker for pts with metastatic CRC and capitalizes on treating CRC subpopulations with targeted agents based upon validated RNA-based signatures. Although the efficacy for bintrafusp alfa and RT is low, changes in IFNγ signature provides a potential signal for refining therapeutic strategies based upon TGFβ enrichment in pts with mCRC. Clinical trial information: NCT03436563 .

152 Background: Monotherapy with immune checkpoint blockade (ICB) is ineffective for patients (pts) with MSS mCRC. Novel approaches to modulate the tumor microenvironment (TME) are needed. Here, we investigate whether the combination of trametinib (T) with durvalumab (D) can alter the immune TME by successfully priming and activating T-cells. Methods: An open-label, single center phase II trial with primary endpoint of immune-related response rate for T+D in refractory MSS mCRC pts (NCT03428126). T is 2mg/day orally starting 1 week prior to D, which is given 1500mg intravenously every 4 weeks. Dose de-escalation strategy performed to identify maximum tolerated dose (MTD). Simon 2-stage design utilized with plans to enroll 29 pts into the first stage, requiring response in 2 or more pts to proceed to stage 2 (n = 15). Results: Demographics for 29 treated pts: 48% female, median age 48 years (range 28-75), and median prior therapies was 2 (range 1-5). No grade (G) 4 treatment-related adverse events (TRAE). The most common G3 TRAE included autoimmune hepatitis (14%) and acneiform rash (10%). G1/2 TRAE included acneiform rash (69%), fatigue (24%) and anemia (21%). No fatal TRAE and 4 pts discontinued treatment due to TRAE. 1 of 29 pts had confirmed partial response (PR) lasting 9.3 months (mo) for an ORR of 3.4%. This pt had an ATM E221fs*14 mutation. 7 pts had stable disease (SD) with median time to progression (TTP) of 5.4 mo (range 3.9-9.3 mo). 1 pt remains on active therapy with SD ( > 10 mo). 5 pts (1 PR, 4 SD) demonstrated decrease in total CEA ng/mL (best percentage reduction: 94%, 95%, 42%, 34% and 21.6% respectively). Median TTP for the entire cohort was 3.2 mo (range 1.1-9.3 mo). Consensus molecular subtypes (CMS) were performed on the primary CRC in 23 pts: 12 CMS2, 2 CMS3, and 9 CMS4. 4 SD pts were CMS2, 1 SD pt was CMS4 and the CMS status of the pt with a PR was unknown. Conclusions: The combination of T+D did not meet efficacy criteria to proceed to the second stage of the study. Analysis of 15 paired on-treatment biopsies is ongoing and will be presented. Utilization of CMS characterization in mCRC clinical trials is feasible and may provide an improved biological understanding of treatment activity. Clinical trial information: NCT03428126.

TPS277 Background: Identifying non-responders to expensive salvage therapies with modest benefits and substantial treatment related adverse events (TRAEs) (e.g. regorafenib/TAS102 in mCRC) is necessary to maximize benefits and limit toxicities. Serial ctDNA sequencing is reliable for tracking tumor dynamics and appears to predict resistance to therapy earlier than radiographic progression. Methods: TACT-D is a randomized study (N = 100) to validate the ability of changes in ctDNA (ΔctDNA) to predict resistance early and in limiting toxicities. We hypothesize that increase in ctDNA (measured by variant allele fraction) at 2 weeks (wk) into treatment can predict resistance earlier than standard radiographic means [at 8-12 wk] and detecting resistance early can enable prompt change in therapy resulting in reduction of TRAEs. Pts with mCRC eligible for either regorafenib/TAS102 are randomized 2:1 to either standard of care (SOC) or ctDNA arm. On SOC arm, treatment is given as per current paradigm i.e. for 8 wk and then restaging. On ctDNA arm, decision to continue therapy is based on ctDNA change between baseline and 2 weeks [ΔctDNA = ctDNA (C1D15 – C1D1)]. Increase in ctDNA triggers early radiographic staging (4 wk). Treatment is continued for disease stability/regression and discontinued for progression. Study has 2 co-primary endpoints: 1) Association of Δ ctDNA and radiographic progression [62 pts on SOC arm, have 94% power (2-sided α 0.05) to detect difference of 95% vs. 58% in progressive disease between pts with increase vs decrease in ctDNA] and 2) Compare proportion of pts experiencing TRAEs within 4 months between study arms [67 in SOC arm and 33 in ctDNA arm have 82% power (2-sided α 0.05) to detect a 30% decrease in toxicity]. Key secondary endpoints include: patient-reported outcomes (MD Anderson Symptom Inventory and PRO-CTCAE), OS, clinical events of special interest (hospitalizations/ER visits/medical interventions such as blood transfusions/IV hydration), clinical trial referral and cost effectiveness. Study is now actively accruing pts (NCT03844620). Funding: MD Anderson Cancer Center, Houston, TX & Guardant Health Inc., Redwood City, CA. Clinical trial information: NCT03844620.

180 Background: RAS mutations are found in ~50% of patients (pts) with metastatic colorectal cancer (mCRC) and associated with resistance to anti-EGFR. Circulating tumor DNA (ctDNA) enables detection of resistant RASMUT arising from RASWT. Recently there has been interest in defining the converse: RASMUT tumors that revert to RASWT, with early results suggesting rates of ~7%. Clinical trials in this population are in development, though the incidence has not been validated with robust methodologies. Methods: 1) We identified 74 mCRC pts with baseline RASMUT and longitudinal ctDNA or tissue data enrolled in ATTACC (NCT01196130), a prospective genomic matching protocol utilizing paired tissue/ctDNA samples at baseline. We evaluated serial samples for RAS loss. 2) Using an external cohort of pts with mCRC and serial ctDNA with a targeted NGS assay sequencing all KRAS/ NRAS exons (Guardant360, Guardant Health), we screened pts for baseline RASMUT with no evidence of prior anti-EGFR exposure and evaluated for RAS loss. Results: 74 pts met criteria of RASMUT CRC with serial samples in ATTACC. Of these, 51 retained RASMUT. 22 pts had very low or absent levels of other clonal alterations such as APC or TP53 and are therefore unable to reliably detect RAS loss. One patient had true RAS loss with NRAS G13R, APC and TP53 mutations at baseline and persistent high-level APC and TP53 mutations without a detectable NRAS mutation, for an overall rate of RAS loss of 2% (1/52). In the second cohort we identified 162 pts, 34 of which had insufficient ctDNA to assess RAS loss on the serial sample as defined by loss of clonal alterations like APC and TP53. Of the remaining 128 patients, 11 had RAS loss (8.5%, with 1 NRAS, 10 KRAS). We next compared the relative mutant allele frequency (rMAF) between RAS retainers and RAS loss. The median baseline rMAF for pts who lost RAS was 0.74, compared to 0.86 in pts retaining RAS (p = 0.045). Conclusions: RAS reversion in mCRC from RASMUT to RASWT is uncommon and occurs at a rate between 2-8% in our two cohorts. RAS reversion is associated with a lower rMAF at baseline, suggesting subclonality. Liquid biopsies must be interpreted carefully, such that a determination of RAS mutation status is most informative in the presence of truncal APC and/or TP53 mutations.

238 Background: RAS mutations are negative predictors of response to anti-EGFR therapies such as panitumumab in mCRC. Mutations at baseline (BL) and follow-up (FU) during randomized phase 3 studies of first line treatment (1L; study 20050203 [‘203]; panitumumab + fluorouracil, leucovorin and oxaliplatin [FOLFOX4] vs FOLFOX4) were compared with those in third line treatment (3L; study 20100007 [‘0007]; panitumumab + best supportive care [BSC] vs BSC) to assess tumor heterogeneity via ctDNA analysis. Methods: Biomarker analysis was conducted for pts with plasma samples at BL and FU. Samples were analyzed using the Plasma Select-R 63-gene panel (Personal Genome Diagnostics, Inc.), with a limit of detection of 0.1%. Mutations were defined at the amino acid level. The Cox hazard ratio (HR) by sum of RAS mutant allele frequency (MAF) was determined, as were event-free survival (EFS) and best response by RAS mutation status. Results: For all pts with available samples (‘203, n = 120; ‘0007, n = 90), fewer mutations and fewer mutations/gene were observed in the 1L vs 3L setting at BL ( KRAS 2 vs 3 maximum mutations/gene; EGFR 1 vs 4 maximum mutations/gene). In 3L the Cox HR increased continuously with RAS MAF; while this was not found in 1L. In the 1L setting, emergent RAS mutations were not predictive of EFS for FOLFOX4, but were predictive of shorter EFS for panitumumab + FOLFOX4. More panitumumab-treated pts in the 3L setting had detectable RAS mutations emerging from BL to FU (28.6%, 57.1%) vs pts treated in 1L (24.5%, 26.5%). For pts who achieved a partial response, more treated in 1L maintained a higher frequency of wild-type RAS from BL to FU (84.4%, 78.1%) vs pts treated in 3L (95.7%, 43.5%). Conclusions: The overall mutational landscape differs between the 1L vs 3L setting in anti-EGFR–treated mCRC pts. Panitumumab monotherapy in the 3L setting appears to induce greater RAS-specific selective pressure than panitumumab FOLFOX4 combination therapy in 1L, resulting in increased RAS mutations at FU in the 3L setting. The combination of panitumumab + FOLFOX in 1L is associated with delayed emergence of expansion of RAS mutations compared to later line single agent panitumumab.