Toward optimizing patient selection for EGFR antibody therapies in metastatic colorectal cancer: outcomes and resistance features in real-world data

Abstract

Background

Patients with metastatic colorectal cancer (mCRC) with RAS- or BRAF-mutant tumors do not benefit from epidermal growth factor receptor (EGFR) monoclonal antibody (mAb) therapy. Among patients with RAS/BRAF wild-type (WT) tumors, a substantial portion still do not benefit from EGFR mAb treatment. Using real-world clinicogenomic data, we investigated the impact of primary and acquired genomic resistance alterations upon treatment outcomes and determined the prevalence of alterations before and after EGFR mAb treatment.

Materials and methods

This study utilized a de-identified mCRC clinicogenomic database from ∼280 US cancer clinics between March 2014 and April 2023. We examined real-world progression-free survival (rwPFS) and overall survival (rwOS) between patients with and those without pre-specified genomic alterations (PSGAs) by Cox models and an adjusted risk score. Genomic alterations were also compared between samples collected before and after EGFR mAb therapy.

Results

Nearly, one-third of microsatellite stable (MSS) RAS/BRAF WT tumors harbor intrinsic resistance alterations before treatment. MSS mCRC patients with WT RAS/BRAF tumors having resistance alterations within the PSGA set [non-canonical RAS/RAF/MAPK and PI3K/PTEN/AKT pathway components; ERBB2 alterations; alternative receptor tyrosine kinases (RTKs) including FGFR1, FGFR2, EGFR, MET, RET, PDGFRA, and NRTK1 fusion] demonstrated decreased rwPFS and/or rwOS on first-line EGFR mAb treatment. The prevalence of RAS/RAF/MAPK and RTK alterations was higher in samples collected after EGFR mAb therapy. The risk of acquiring an RTK resistance alteration increased with the total duration of EGFR mAb treatment.

Conclusions

Detection of genomic resistance alterations in MSS RAS/BRAF WT patients confers less favorable EGFR mAb treatment outcomes. The duration of EGFR mAb treatment increased the risk of emergence of an acquired resistance alteration.

Highlights

• •One-third of RAS/BRAF WT mCRC has EGFR mAb primary resistance gene alterations.
• •ERBB2, RTK, and RAS/PI3K pathway gene alterations were associated with EGFR mAb resistance.
• •RAS/BRAF/EGFR mutations and KRAS/MET amplification are more prevalent post-EGFR mAb treatment.
• •Acquired RTK resistance alterations are associated with EGFR mAb time of exposure.
• •Comprehensive genomic profiling detects a broad set of gene alterations of EGFR mAb primary resistance in mCRC.

Introduction

Colorectal cancer (CRC) is the second most common cause of cancer mortality in the United States with an anticipated 52 550 deaths in 2023.¹ Among those newly diagnosed with CRC, ∼23% will have metastatic (m) disease and the 5-year survival rate for mCRC is roughly 15%.¹ For unresectable mCRC, the primary treatment is systemic therapy with fluoropyrimidine-based cytotoxic chemotherapy, biologic therapies, immunotherapy, or a combination of these agents. Genomic and molecular profiling facilitates tailoring targeted therapies to mCRC, which has improved overall survival (OS) in prospective clinical trials.^2,3 For the majority of mCRC cases which respond poorly to immunotherapy, knowing the presence of genomic variants in KRAS/NRAS and BRAF genes as well as primary tumor sidedness4, 5, 6 provides important clinical information for the selection of guideline-directed therapies.^2,7

Randomized and retrospective studies of mCRC with wild-type (WT) KRAS have evaluated first-line systemic therapy with FOLFOX or FOLFIRI combined with either an epidermal growth factor receptor (EGFR) monoclonal antibody (mAb) (cetuximab or panitumumab) or a vascular endothelial growth factor mAb (bevacizumab). Several of these trials have demonstrated greater benefit of an EGFR mAb in patients with left-sided mCRC tumors.^6,8, 9, 10, 11 The phase III PARADIGM trial established an improved OS for FOLFOX plus panitumumab for patients with left-sided RAS/BRAF WT mCRC and this remains a first-line systemic therapy recommendation for unresectable left-sided mCRC.¹² However, retrospective data show that <10% of patients with left-sided RAS/BRAF WT mCRC receive an EGFR inhibitor in the first-line setting in the United States, but there is a trend toward their increased use in later treatment lines,¹³ possibly owing to treatment-related side-effects and a perceived lack of clinical effectiveness. Frontline EGFR mAb usage outside the United States is more variable with rates as high as 35% reported,^14,15 and clear endorsement on European Society for Medical Oncology (ESMO) guidelines.¹⁶ Moreover, it has been observed that upward of 50% of mCRC patients with WT RAS/BRAF genetic sequences are in fact refractory to EGFR blockade,¹⁷ likely attributable to the prevalence of mutations that confer resistance, underlying chromosomal changes, as well as intrinsic differences in cellular signaling, gene expression, or epigenetic processes.

A wide array of known primary and acquired resistance mechanisms diminish the effectiveness of EGFR mAb therapy.18, 19, 20 Retrospective and clinical trial post hoc analysis first demonstrated that RAS and BRAF V600E mutations are markers of primary resistance to EGFR mAb therapy, resulting in decreased treatment efficacy and OS.^4,5,21,22 Identified primary and acquired resistance mechanisms to EGFR-directed therapy include alterations in EGFR,^19,23 the downstream effectors RAS/RAF/MAPK^20,24,25 and PI3K/PTEN/AKT,^26,27 alterations in alternative receptor tyrosine kinase (RTK) signaling pathways,^18,28 as well as MET amplification,²⁹ which may contribute to the observed lack of objective response and shorter progression-free survival (PFS). Less common genomic events altering RTKs may negatively or positively affect the therapeutic susceptibility of mCRC to EGFR inhibition. mCRC with EGFR amplification has been independently associated with a better OS; however, exposure to EGFR mAb therapy did not result in significant changes in OS.³⁰ mCRC with ERBB2 [human epidermal growth factor receptor 2 (HER2)] amplification treated with EGFR mAb therapy has resulted in poorer overall response rate and PFS outcomes.^31,32 Observational studies utilizing circulating tumor DNA (ctDNA) have also suggested a relationship between genomic amplification of EGFR and FLT3 with possible EGFR treatment resistance^18,20; however, FLT3 frequently co-occurs with KRAS mutations²⁸ which significantly confounds prior interpretation.

Paradigms to more precisely identify the mCRC patients who may benefit most from EGFR-targeted therapy are emerging. Retrospective and prospective studies have explored ‘negative hyperselection’ approaches to best identify and select patients lacking known genomic alterations that may cause resistance to EGFR mAb therapy and initial studies have yielded encouraging treatment outcomes using this approach.^33,34 However, many of these studies have been limited in the breadth of genomic alterations analyzed. Additionally, whether this approach can be modeled in real-world data outside of a tightly controlled clinical trial is of interest to the mCRC field and remains unknown.

We investigated possible primary and acquired resistance alterations to EGFR mAb-directed therapy in a large real-world mCRC patient cohort, where patients underwent comprehensive genomic profiling (CGP) and were captured within an extensive clinicogenomic database (CGDB). Building off prior mCRC negative hyperselection studies, we hypothesized that (i) alterations in pre-specified pathways and genes of known biological significance are associated with less favorable clinical outcomes amongst patients treated with an EGFR mAb but not bevacizumab; (ii) these alterations are more prevalent following EGFR mAb treatment initiation; and (iii) development of unfavorable alterations is associated with duration of exposure to EGFR mAb.

Materials and methods

Study population

All patients included in this study had a confirmed diagnosis of mCRC, underwent tissue genomic testing using Foundation Medicine tissue CGP assays (Foundation Medicine, Cambridge, MA), and were included in the US-wide Flatiron Health and Foundation Medicine mCRC CGDB between March 2014 and April 2023. Retrospective de-identified longitudinal clinical data were derived from electronic health records (EHRs) from ∼280 US cancer clinics (∼800 sites of care), comprising patient-level structured and unstructured data, curated via technology-enabled abstraction of clinical notes and radiology/pathology reports. These were linked to genomic data derived from Foundation Medicine CGP testing by de-identified, deterministic matching.³⁵ The data are de-identified and subject to obligations to prevent re-identification and protect patient confidentiality. Clinical data included demographics, clinical and laboratory features, time of therapy exposure, and survival. Tumor-sidedness was determined as previously described.³⁶

mCRC patients with microsatellite stable (MSS), RAS/RAF WT tumors [defined as no known or likely pathogenic mutations in KRAS/NRAS exons 2/3/4 or BRAF^V600E mutation (biomarkers endorsed by National Comprehensive Cancer Network (NCCN) guidelines for first-line EGFR mAb treatment)], who received first-line therapy with EGFR mAb or bevacizumab, and had tissue biopsy collected before therapy started were included in the first step of the study. The second step of the study included mCRC patients who received therapy with EGFR mAb and had tissue biopsy collected before or after therapy started in any line. Those patients with tissue biopsy collected before therapy had no known or likely pathogenic mutations in KRAS/NRAS exons 2/3/4 or BRAF^V600E mutation. Figure 1 shows the cohort selection and analysis overview for this study. Institutional review board approval of the study protocol was obtained before the study start and included a waiver of informed consent based on the observational, non-interventional nature of the study (WCG IRB, Protocol No. 420180044).

Figure 1.

Cohort selection and analysis overview. EHR, electronic health record; ERBB2 alterations, ERBB2 mutations or amplification; LOT, line of therapy; mCRC, metastatic colorectal cancer; MSS, microsatellite stable; NCCN, National Comprehensive Cancer Network. Alteration in RAS signaling pathway: KRAS/NRAS/BRAF/HRAS/RAF1 mutations or amplification, BRAF rearrangements, or MAP2K1 mutations; alterations in PI3K signaling pathway: mutation in PIK3CA/PTEN/AKT1, or PTEN deletion; alterations in RTKs: MET/FGFR1/FGFR2/FGFR3 mutations, amplification, or rearrangements, EGFR/PDGFRA mutations or rearrangements, or NTRK1/NTRK2/NTRK3/ALK/RET/ROS1 rearrangements. ^aNo KRAS/NRAS exon 2/3/4 K/L mutations or BRAF^V600E mutation (biomarkers endorsed by NCCN guidelines for EGFR mAb first-line treatment).

Comprehensive genomic profiling

Hybrid capture-based next-generation sequencing (NGS) assays were carried out on patient tumor specimens in Clinical Laboratory Improvement Amendments (CLIA)-certified, College of American Pathologists (CAP)-accredited laboratory (Foundation Medicine, Inc.). Genomic alterations were identified via CGP of >300 cancer-related genes on Foundation Medicine’s NGS test (FoundationOne®CDx or FoundationOne®).37, 38, 39 We set pre-specified known or likely pathogenic genomic alterations (PSGAs) for this study based on previous evidence in the literature showing an association with either resistance or sensitivity to EGFR mAb therapy. The PSGAs were divided into six different subgroups: resistance alterations including (i) ERBB2 alterations (ERBB2 mutations, as defined by known or likely to be ERBB2 activating, and ERBB2 amplification)^28,40, 41, 42, 43, 44, 45, 46, 47, 48; (ii) PI3K signaling pathway alterations (mutations in PIK3CA, PTEN, AKT1, or PTEN deletion)^26,28,40; (iii) RAS signaling pathway alterations [KRAS, NRAS, HRAS, RAF1 mutations or amplifications; BRAF mutations (class I and II) and rearrangements; or MAP2K1 mutations]^18,49; (iv) RTK alterations (MET, FGFR1, FGFR2, FGFR3 mutations, amplifications, or rearrangements; EGFR, PDGFRA mutations or rearrangements; NTRK1, NTRK2, NTRK3 fusions; and ALK, RET, ROS1 rearrangements)^{18,28,40,42,43,50}; sensitivity alterations including (v) EGFR amplification⁴¹; and (vi) FLT3 amplification¹⁸ was evaluated as a separated group due to conflicting data.

Outcomes

Real-world (rw)PFS and (rw)OS were the primary endpoints for this study. rwPFS was calculated from the treatment start date until the time of disease progression or death, and patients not yet reaching progression or death were right-censored at the date of the last clinical note. rwOS was calculated from the start of treatment to death from any cause, and patients with no record of mortality were right-censored at the date of the last clinic visit or structured EHR activity. rwOS risk intervals were left truncated to the date of the CGP report to account for immortal time, as patients cannot enter the database until a CGP report is provided.^51,52 The mortality information in the Flatiron Health database is a composite derived from de-identified patient-level data within the EHR, the public Social Security Death Index, and a commercial death dataset mining data from obituaries and funeral homes. This mortality information has been externally validated in comparison to the National Death Index.⁵³ In addition, the Flatiron Health and Foundation Medicine CGDB has replicated associations with survival observed in biomarker subgroup analyses of randomized controlled trials.54, 55, 56

Statistical analysis

The analyses carried out in this study were pre-specified in a prospectively declared statistical analysis plan (SAP). The SAP also pre-specified inclusion and exclusion criteria, potential biases, primary outcome measures, handling of missing data, and all methods described herein, consistent with the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) guidelines.⁵⁷ The pre-specified analyses included the comparison of outcomes in patients with and without genomic alterations receiving first-line EGFR mAb therapy, the comparison of the prevalence of genomic alterations among tissue specimens collected before and after EGFR mAb treatment start, and the investigation of the association of the amount of time on EGFR mAb with selected genomic alterations. Additional exploratory analyses included the comparison of outcomes in patients with and without genomic alterations receiving first-line bevacizumab and the comparison of outcomes in patients RAS/RAF WT receiving first-line EGFR mAb versus bevacizumab.

Chi-square tests and Wilcoxon rank sum tests were used to assess differences between groups of categorical and continuous variables, respectively. Differences in rwPFS and rwOS were evaluated with the log-rank test and Cox proportional hazards models. To adjust for potential confounders across all analyses, the baseline clinical risk score was estimated from known prognostic features as the linear predictor from a Cox proportional hazards model for rwOS in mCRC, as previously described.⁵⁴ The prognostic features included the line of therapy, age at treatment start, gender, race, recurrent disease versus new diagnosis, Eastern Cooperative Oncology Group (ECOG) performance status (PS), practice type (academic or community), primary tumor location, albumin, alkaline phosphatase, serum creatinine, hemoglobin, lactate dehydrogenase, neutrophil-to-lymphocyte ratio, platelet, opioid prescription pre-therapy, and steroid prescription pre-therapy. Genomic alterations were compared between patients with biopsies collected before and after starting treatment with EGFR mAb by chi-square test, adjusted for multiple comparisons. The association between the duration of drug exposure and genomic alterations was evaluated by linear regression. R version 4.1.3 software (R Core Team, cran.r-project.org/doc/FAQ/R-FAQ.html) was used for all statistical analyses.

Results

Characteristics of mCRC patients

Firstly, 253 patients were identified with MSS mCRC and WT RAS/BRAF (as defined in the Materials and Methods section) who received first-line EGFR mAb treatment after tissue biopsy, with a median age of 60 years [interquartile range (IQR) 52.0-70.0 years]. At the start of treatment, 76 patients (30.0%) had detectable resistance genomic alterations within the PSGA set, while 177 patients (70.0%) had none (Figure 1). Most patients were male (62.5%), had an ECOG PS score of 0 (54.1%) or 1 (35.2%), had primary tumors of the left colon (45.5%) or rectum (30.0%), and received EGFR mAb in combination with chemotherapy (94.5%) (Table 1). The baseline clinical characteristics of patients in cohort 1, as subgrouped by genomic resistance alterations or genomic amplification of FLT3 or EGFR (Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2024.100036), exhibit similar patient profiles. Among the patients receiving first-line EGFR mAb, we observed numerous potential resistance alterations within genes in the RAS (7.1%) and PI3K signaling pathways (12.6%), ERBB2 (11.1%), and other RTKs (5.1%), as well as amplifications of FLT3 (18.2%) and EGFR (7.1%) which were considered separately (Supplementary Table S2, available at https://doi.org/10.1016/j.esmorw.2024.100036). Twenty-one patients had co-alterations in genes of the RAS and PI3K pathways, ERBB2, or other RTKs (Supplementary Table S3, available at https://doi.org/10.1016/j.esmorw.2024.100036), with four patients with alterations in both RAS and PI3K pathways, five patients with alterations in PI3K pathway and ERBB2, and three patients with alterations in both RAS pathway and ERBB2 (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmorw.2024.100036).

Table 1.

Baseline clinical characteristics of MSS mCRC patients RAS/BRAF WT receiving first-line EGFR mAb

	No resistance alterations (n = 177)	Resistance alterations (n = 76)	Total (N = 253)	P value
Age at Tx start				0.33
Median (Q1, Q3)	61.0 (52.0, 70.0)	59.5 (49.0, 67.0)	60.0 (52.0, 70.0)
Gender, n (%)				0.067
Female	60 (33.9)	35 (46.1)	95 (37.5)
Male	117 (66.1)	41 (53.9)	158 (62.5)
ECOG performance status, n (%)				0.947
0	88 (54.3)	38 (53.5)	126 (54.1)
1	57 (35.2)	25 (35.2)	82 (35.2)
2	16 (9.9)	7 (9.9)	23 (9.9)
3+	1 (0.6)	1 (1.4)	2 (0.9)
Missing, n	15	5	20
EGFR mAb regimen, n (%)				0.634
EGFR mAb + chemotherapy	168 (94.9)	71 (93.4)	239 (94.5)
EGFR mAb monotherapy	9 (5.1)	5 (6.6)	14 (5.5)
Primary tumor location, n (%)				0.043
Colon NOS	23 (13.0)	11 (14.5)	34 (13.4)
Left colon	84 (47.5)	31 (40.8)	115 (45.5)
Multiple	1 (0.6)	0 (0.0)	1 (0.4)
Rectum	55 (31.1)	21 (27.6)	76 (30.0)
Right colon	14 (7.9)	9 (11.8)	23 (9.1)
Transverse colon	0 (0.0)	4 (5.3)	4 (1.6)
Resistance alteration, n
ERBB2 alteration		28	28
PI3K signaling pathway		32	32
RAS signaling pathway		16	16
Other RTK alterations		13	13

ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; mAb, monoclonal antibody; NOS, not otherwise specified; RTK, receptor tyrosine kinase; Tx, therapy.

We next selected a total of 834 MSS mCRC patients with WT RAS/BRAF who received bevacizumab treatment following tissue biopsy, with a median patient age of 60 years (IQR 51.0-69.0 years). Similar to the previous cohort of patients receiving first-line EGFR mAb therapy, most patients were male (61.4%), had an ECOG PS score of 0 (51.6%) or 1 (39.2%), and had primary tumors of the left colon (42.1%) or rectum (24.0%). At the start of treatment, 521 patients (62.5%) demonstrated no genomic alterations and 313 (37.5%) had detectable alterations within the pre-defined gene set. Additional clinical baseline characteristics of patients receiving first-line bevacizumab are presented in Supplementary Table S4, available at https://doi.org/10.1016/j.esmorw.2024.100036.

For the second analysis of the study, a total of 1952 mCRC patients received EGFR mAb therapy in any line of therapy, with a median age of 60.0 years (IQR 52.0-69.0 years), majority male (59.5%), with an ECOG PS score of 0 (41.4%) or 1 (48.0%), and had primary tumors of the left colon (42.7%) or rectum (25.9%) (Supplementary Table S5, available at https://doi.org/10.1016/j.esmorw.2024.100036). The majority of patients received EGFR mAb in either the first (25.9%) or second (38.7%) line of therapy. Within this cohort, 1653 (84.7%) had tissue biopsies collected before treatment with EGFR mAb was started and 299 (15.3%) had tissue biopsies collected following treatment start (Figure 1).

Patients with genomic alterations in components of the RAS and PI3K pathways, ERBB2, or other RTKs before receiving first-line EGFR mAb have less favorable outcomes than those without

We examined the 253 MSS WT RAS/BRAF patients who underwent tissue biopsy before the start of first-line EGFR mAb therapy. Comparison of patients treated with first-line EGFR mAb, who had a tumor biopsy with one of the resistance PSGAs, to patients with a tumor without those alterations showed significantly poorer rwPFS [median 7.4 versus 12.2 months, adjusted hazard ratio (aHR) 1.52, 95% confidence interval (CI) 1.14-2.02] and rwOS (median 21.8 versus 35.0 months, aHR 1.64, 95% CI 1.14-2.36) (Figure 2A). Examining the 21 patients with co-alterations in the RAS and PI3K pathways, ERBB2, or other RTKs, the outcomes were even less favorable (median PFS 6.34 months and OS 16.3 months).

Figure 2.

Comparison of outcomes between MSS mCRC patients with wild-type RAS/BRAF tumors treated with first-line EGFR mAb or bevacizumab and grouped by the presence or absence of genomic resistance alterations. Kaplan–Meier plots showing rwPFS and rwOS for (A) patients receiving EGFR mAb with genomic resistance alterations versus no alterations; (B) patients receiving bevacizumab with genomic resistance alterations versus no alterations; (C) patients receiving bevacizumab or EGFR mAb therapy with genomic resistance alterations; (D) patients receiving bevacizumab or EGFR mAb therapy with no genomic resistance alterations. aHR, adjusted hazard ratio; alt, alterations; CI, confidence interval; EGFR, epidermal growth factor receptor; HR, hazard ratio; mAb, monoclonal antibody; OS, overall survival; PFS, progression-free survival; rw, real-world; Tx, treatment.

We next evaluated the relationship and outcomes between each subgroup of resistance PSGAs to the rwPFS and rwOS. Among patients receiving first-line EGFR mAb, patients with PSGAs in the RAS signaling pathway (n = 18) versus no alterations (n = 177) tended to have less favorable rwPFS (median 7.1 versus 12.2 months, aHR 1.57, 95% CI 0.95-2.61) and rwOS (median 20.1 versus 35 months, aHR 1.66, 95% CI 0.85-3.22) (Supplementary Figure S2A, available at https://doi.org/10.1016/j.esmorw.2024.100036). Patients with alterations in the PI3K signaling pathway (n = 32) versus no alterations (n = 177) tended to have less favorable rwPFS (median 7.4 versus 12.2 months, aHR 1.39, 95% CI 0.93-2.08) and rwOS (median 21.8 versus 35.0 months, aHR 1.64, 95% CI 1.01-2.65) (Supplementary Figure S2B, available at https://doi.org/10.1016/j.esmorw.2024.100036). Similarly, patients with PIK3CA mutations (n = 23) have less favorable rwPFS (median 7.1 versus 12.2 months, aHR 1.63, 95% CI 1.01-2.61) and rwOS (median 21.8 versus 35.0 months, aHR 2.16, 95% CI 1.23-3.77) (Supplementary Figure S3A, available at https://doi.org/10.1016/j.esmorw.2024.100036). Patients found to have alterations in the specified RTKs (n = 13) versus those without (n = 177) had less favorable rwPFS (median 6.4 versus 12.2 months, aHR 1.86, 95% CI 1.05-3.3) and rwOS (median 15.4 versus 35.0 months, aHR 2.51, 95% CI 1.25-5.06) (Supplementary Figure S2C, available at https://doi.org/10.1016/j.esmorw.2024.100036). Patients with ERBB2 alterations (n = 28) versus no alterations (n = 177) demonstrated a less favorable rwPFS (median 7.1 versus 12.2 months, aHR 1.79, 95% CI 1.18-2.72) and trended toward decreased rwOS (median 24.0 versus 35.0 months, aHR 1.47, 95% CI 0.86-2.53) (Supplementary Figure S2D, available at https://doi.org/10.1016/j.esmorw.2024.100036). Similarly, patients found to have genomic amplification of ERBB2 (n = 23) have both shorter rwPFS (median 8.2 versus 12.2 months, aHR 1.74, 95% CI 1.1-2.75) and rwOS (median 24.0 versus 35.0 months, aHR 1.77, 95% CI 0.98-3.19) (Supplementary Figure S3B, available at https://doi.org/10.1016/j.esmorw.2024.100036). An additional multivariable analysis was carried out for both rwPFS and rwOS and showed the individual contribution of each subgroup of genomic alterations (Supplementary Figure S4, available at https://doi.org/10.1016/j.esmorw.2024.100036). Finally, to investigate if the genomic alterations observed were associated with overall increased mutational heterogeneity, we evaluated the tumor mutational burden (TMB) in patients with and without PSGAs. Patients with no PSGAs (n = 177) had a median TMB of 2.6 mut/Mb (IQR 1.3-5.0 mut/Mb) while patients with any resistance alterations (n = 76) had a slightly higher median TMB of 3.5 mut/Mb (IQR 2.5-5.0 mut/Mb), with a P value of 0.033.

To better assess the potential causality of the resistance alterations we observed, we carried out an exploratory treatment-interaction analysis. Our treatment-interaction analysis showed that patients with resistance alterations experienced less favorable outcomes receiving first-line EGFR mAb with chemotherapy compared to patients receiving first-line chemotherapy only (Supplementary Table S6, available at https://doi.org/10.1016/j.esmorw.2024.100036), but without reaching statistical significance (rwPFS interaction P = 0.229, aHR 1.29, 95% CI 0.85-1.95; rwOS interaction P = 0.919, aHR 1.03, 95% CI 0.60-1.76).

We conducted further exploratory analyses on the outcomes of patients who received bevacizumab treatment with or without resistance PSGAs. We then compared bevacizumab-treated patients to those treated with EGFR mAb in either the presence or absence of resistance PSGAs. Consistent with our hypothesis, patients treated with bevacizumab remained unaffected by tumor intrinsic resistance alterations which lead to poorer outcomes if treated with EGFR mAb, as bevacizumab operates through independent antiangiogenic and immunomodulatory pathways.^58,59 Possible genes that may lead to either intrinsic resistance or enhanced treatment efficacy with bevacizumab were not included within the PSGA panel. As expected, we observed that patients treated with first-line bevacizumab had similar outcomes regardless of the presence of resistance alterations (Figure 2B). WT RAS/BRAF MSS patients with resistance alterations tended to have more favorable outcomes receiving bevacizumab versus EGFR mAb (median rwPFS 10.0 versus 7.4 months, aHR 1.37, 95% CI 1.05-1.78/median rwOS 25.2 versus 21.8 months, aHR 1.32, 95% CI 0.96-1.83) (Figure 2C), whereas WT RAS/BRAF MSS patients without resistance alterations tended to have less favorable outcomes receiving bevacizumab than EGFR mAb (median rwPFS 11.1 versus 12.2 months, aHR 0.93, 95% CI 0.77-1.12/median rwOS 28.2 versus 35.0 months, aHR 0.82, 95% CI 0.64-1.05) (Figure 2D). Analysis stratified by tumor-sidedness demonstrated a trend toward improved treatment outcomes amongst patients with resistance PSGAs receiving bevacizumab; while patients with left-sided WT RAS/BRAF MSS without resistance PSGAs trended toward benefit from EGFR mAb, those with right-sided disease with resistance PSGAs trended toward improved outcomes on bevacizumab (Supplementary Figure S5, available at https://doi.org/10.1016/j.esmorw.2024.100036). Interestingly, patients with no resistance alterations and tumors located on the right side had similar rwPFS when receiving EGFR mAb versus bevacizumab (8.4 versus 11.8 months, aHR 0.91, 95% CI 0.47-1.79), and tended to have more favorable rwOS (45.8 versus 28.1 months, aHR 0.66, 95% CI 0.26-1.66) (Supplementary Figure S5, available at https://doi.org/10.1016/j.esmorw.2024.100036).

In addition, we analyzed clinical outcomes in patients receiving first-line EGFR mAb whose primary tumors had genomic amplifications of EGFR or FLT3. Amongst patients with amplification of EGFR or FLT3 receiving first-line EGFR mAb, those with EGFR amplification versus no amplification had a significantly more favorable rwPFS (median 15.3 versus 9.9 months, aHR 0.56, 95% CI 0.32-0.97) and rwOS (median 28.8 months versus not reached, aHR 0.41, 95% CI 0.18-0.94) (Supplementary Figure S6A and B, available at https://doi.org/10.1016/j.esmorw.2024.100036). For patients with FLT3 amplification versus no amplification, there was a more favorable rwPFS (median 13.1 versus 9.7 months, aHR 0.71, 95% CI 0.5-1.0) and trend for rwOS (median 38.8 versus 28.4 months, aHR 0.72, 95% CI 0.45-1.16) with first-line EGFR mAb therapy (Supplementary Figure S6C and D, available at https://doi.org/10.1016/j.esmorw.2024.100036). We carried out two additional sensitivity analyses excluding either patients with EGFR amplification or those with EGFR amplification plus resistance PSGAs, and the observation remained similar (Supplementary Figure S7, available at https://doi.org/10.1016/j.esmorw.2024.100036).

Tissue specimens collected after starting EGFR mAb therapy have a higher prevalence of alterations in the RAS signaling pathway and RTKs

A total of 299 patients underwent a tissue biopsy following the start of treatment with an EGFR mAb. As compared to tissue biopsies obtained before treatment started and without KRAS/NRAS exons 2/3/4 or BRAF^V600E mutations (n = 1653), alterations in the RAS signaling pathway were more prevalent in tissue specimens collected after initiation of EGFR mAb therapy (38.5% after versus 8.7% before, P < 0.001), as were RTK alterations (16.7% after versus 8.2% before, P < 0.001) (Figure 3A). In terms of specific genomic alterations observed, KRAS/NRAS/BRAF/EGFR mutations and KRAS/MET amplifications were more prevalent in tissue specimens collected after therapy initiation (Figure 3B). The KRAS Q61H mutation (21.3% of all KRAS mutations), EGFR V441D mutation (33.3% of all EGFR mutations), and the BRAF V600E mutation (65.4% of all BRAF mutations) were most prevalent following initiation of EGFR mAb therapy (Supplementary Table S7, available at https://doi.org/10.1016/j.esmorw.2024.100036).

Figure 3.

Genomic alteration prevalence and risk assessed by tumor biopsies before or after EGFR mAb therapy. Scatter plots showing the prevalence of alterations by (A) genomic alteration subgroups and (B) specific resistance alterations within RAS signaling pathway and RTKs. (C) Forest plot demonstrating the association and risk of RTK alteration prevalence with the duration of EGFR mAb therapy. Alt, alterations; CI, confidence interval; EGFR, epidermal growth factor receptor; mAb, monoclonal antibody; OR, odds ratio; RTK, receptor tyrosine kinase; Tx, treatment.

We hypothesized that a longer duration of EGFR mAb therapy would also increase the risk of developing resistance alterations. As compared to patients unexposed to EGFR mAb, those patients on treatment with EGFR mAb have a higher risk of RTK alteration detection as the treatment duration increased with respective odds ratios (ORs) of 3.24 (95% CI 1.93-5.31) for 0-6 months of exposure (n = 146), OR of 2.04 (95% CI 0.91-4.11) for 6-12 months of exposure (n = 75), and OR of 3.80 (95% CI 1.97-6.95) for >12 months of exposure (n = 74) (Figure 3C). We restricted this analysis to RTKs because alterations in the RAS signaling pathway are enriched in unexposed patients.

An exploratory analysis was conducted on an additional group to compare the prevalence of genomic alterations in liquid biopsy specimens (FoundationOne Liquid CDx) within the CGDB collected before EGFR mAb start and up to 30 days after progression on EGFR mAb. This group consisted of individuals who had a ctDNA tumor fraction greater than zero and were either treated with EGFR mAb before (n = 76) or after (n = 25) treatment initiation. Results from the liquid biopsy demonstrate an increased prevalence of KRAS mutations (28.0% after versus 0% before, P < 0.0001), NRAS mutations (20.0% after versus 0% before, P = 0.0007), EGFR mutations (28.0% after versus 2.6% before, P = 0.0009), and MAP2K1 mutations (20.0% after versus 2.6% before, P = 0.017) following treatment initiation (Supplementary Figure S8A and B, available at https://doi.org/10.1016/j.esmorw.2024.100036). Eighteen out of the 25 patients with liquid specimens collected after treatment started had at least one resistance alteration and all liquid specimens with resistance alterations were collected after disease progression (Supplementary Figure S8C, available at https://doi.org/10.1016/j.esmorw.2024.100036).

Real-world first-line therapy patterns for mCRC patients with left-sided RAS/BRAF WT tumors

To evaluate the real-world treatment patterns of left-sided RAS/BRAF WT tumors, we show the first-line treatment received for 1731 real-world mCRC patients with left-sided RAS/BRAF WT tumors with data available in CGDB. In general, patients tended to receive chemotherapy alone (469, 27.1%) or with bevacizumab (803, 46.4%) more often than chemotherapy with EGFR mAb (284, 16.4%). Those patients with a CGP report received before the first-line therapy start tended to receive chemotherapy with EGFR mAb (132 out of 533, 24.8%) more frequently than those with a CGP report completed after first-line therapy was started (152 out of 1198, 12.7%) (Supplementary Figure S9, available at https://doi.org/10.1016/j.esmorw.2024.100036).

Discussion

The prevalence and morbidity of mCRC remains high and targeted therapy options are available according to tumor molecular characteristics. To identify mCRC patients who would benefit from treatment with EGFR mAb, guidelines recommend molecular characterization of tumors for RAS and BRAF^V600E mutations.^2,7 In addition to the increasing prevalence of tumor CGP testing, ‘negative hyperselection’ strategies may best identify patients likely to benefit from first-line EGFR mAb therapy. In our analysis of a large real-world mCRC CGDB, we demonstrate that patients with MSS tumors harboring resistance PSGAs have significantly worse clinical outcomes when treated with EGFR mAbs. Among the WT RAS/BRAF patients receiving EGFR mAb in the first line (n = 253), we observed genomic resistance alterations in RAS (7.1%), PI3K signaling pathway (12.6%), alternative RTKs (5.1%), and ERBB2 (11.1%). Thus, we observe that over a third of patients with intrinsic resistance alterations receive EGFR mAb therapy, which is associated with less favorable treatment outcomes when using guideline-directed criteria for choosing EGFR mAb therapy.^2,7 Moreover, the true clinical benefit of EGFR mAbs in the clinic, as well as in past and current trials, may be obscured by a high rate of intrinsic resistance. Our study demonstrates that EGFR-resistance alterations can be acquired during EGFR mAb treatment, that the risk of acquiring resistance alterations increases with the duration of EGFR mAb treatment, and that these acquired mutations can be detected using a liquid biopsy method.

In our real-world data, MSS mCRC patients with a WT RAS/BRAF tumor containing any one of the resistance PSGAs (in RAS or PI3K pathways, alternative RTKs, or ERBB2) who were treated with a first-line EGFR mAb demonstrated significantly worse rwPFS and/or rwOS as compared to patients harboring no PSGAs in their tumor. We observed a significant prevalence of genomic alterations in mCRC tumors thought to lead to intrinsic resistance to EGFR mAb therapy in patients who were treated with anti-EGFR blockade (25.9% of those treated in first line and 38.7% of those treated in second line). In addition, the prevalence of observed alterations in untreated, naive mCRCs suggests that the EGFR and downstream signaling pathways may be involved in functions other than tumorigenesis, including dependency on the alterations for the promotion of cell survival, proliferation, and growth. The observed lack of deep and durable treatment responses on EGFR mAb therapy suggests that substantial cross-talk between other signaling pathways may occur to circumvent the dependency upon the EGFR.

The CGDB revealed that a minority (16.4%) of mCRC patients with left-sided WT RAS/BRAF primary tumors received EGFR mAb therapy plus chemotherapy in the first line, while 46.4% received bevacizumab plus chemotherapy, consistent with previous reports.¹³ To our knowledge, less emphasis has been placed on analyzing treatment outcomes for patients with WT RAS/BRAF tumors with and without PSGAs treated with bevacizumab in the first line. Our analysis showed nearly identical rwPFS and rwOS treatment outcomes between these groups. Treatment outcomes with first-line bevacizumab versus an EGFR mAb in patients harboring genomic resistance alterations showed significant improvement of both rwPFS and rwOS in the bevacizumab-treated group. These results are consistent with the antiangiogenic and immunomodulatory mechanisms of bevacizumab operating independently of EGFR. In WT RAS/BRAF patients (of any sidedness) without PSGAs, there was a trend toward improved rwOS and rwPFS with EGFR mAb therapy, which was similar to that observed in the subgroup analysis of left-sided tumors. Our retrospective analysis supports the utilization of patient-specific tumor CGP data and a ‘negative hyperselection’ approach to identify the most appropriate first-line treatment to benefit mCRC patients. However, it remains unclear if right-sided mCRC patients would similarly benefit from this treatment selection paradigm as most patients included in the study were left-sided.

Several tumors harbored genomic amplifications of the EGFR or FLT3 genes. While EGFR-amplified mCRC may be independently associated with an improved OS,³⁰ treatment of EGFR-amplified tumors with an EGFR mAb resulted in a clinically significant increase in PFS and OS compared to patients without EGFR amplification. Prior observations had suggested that FLT3 amplification was associated with EGFR mAb resistance but may be confounded by high co-occurrence with RAS mutations.^20,28 Our analysis of patients with FLT3-amplified mCRC showed improved rwPFS when treated with an EGFR mAb. Given the understudied nature of FLT3 amplification in mCRC, future trials and retrospective analyses may specify this subgroup for further analysis.

A comparison of the prevalence of genomic alterations in unmatched mCRC tumor biopsies before and after EGFR mAb treatment revealed a significantly higher prevalence of RAS and RTK alterations after treatment. Interestingly, we found a strong association between the prevalence of RTK alterations and the duration of EGFR mAb treatment. While intuitive, to our knowledge, this is the first study to formally demonstrate that the cumulative duration of treatment increases the risk of developing an RTK alteration associated with EGFR mAb treatment resistance. It also highlights the durable selective pressure on tumor cells to acquire genetic changes that abolish EGFR-signaling inhibition. This underscores the possible benefit of surveillance liquid biopsies to monitor the emergence of acquired resistance to EGFR mAb to guide both treatment discontinuation and later EGFR mAb rechallenge following resistant clone decay.^60,61 Thus, serial liquid biopsy may be particularly well-suited to monitoring the emergence of polyclonal resistance.

There are significant strengths to our study, including the robust collection and analysis of real-world CGDB from ∼280 different community cancer clinics, which reflects treatment outcomes outside of rigorously selected and monitored clinical trials. Significant efforts were undertaken to reduce possible false discovery rates by using pre-specified prospective plans for data handling and statistical analysis. We explored and met several principles of the Bradford Hill criteria⁶² before conducting exploratory treatment-interaction analysis to better assess the causality of PSGAs. Our real-world data are consistent with small prospective serial sampling trials in similar populations.⁶¹ Study limitations include its retrospective design, incorporation of rare genomic alterations limited by total patient population size, as well as the inability to determine the effect of individual alterations on clinical outcomes. Additionally, some possible prognostic clinical variables, such as location and degree of metastatic burden, were not able to be ascertained. However, we generated a prognostic risk score to address differences among our cohorts including all known prognostic factors available in our database. Specimens collected before and after EGFR mAb treatment start were unmatched; thus, we cannot definitively affirm that alterations detected after treatment start were indeed acquired and a validation study with longitudinal serial biopsies would be required.

CGP approaches are increasingly utilized in mCRC for characterizing RAS/RAF, mismatch repair/microsatellite instability, and HER2 (ERBB2) status. CGP can yield large amounts of clinically actionable results to guide treatment selection, and our data may argue that CGP is underutilized to optimize patient selection. Here, in addition to the value of reporting RAS/BRAF WT status, we retrospectively show that broad identification of genetic alterations can help guide therapy selection to provide significant therapeutic benefit to MSS mCRC patients. Previously, limited RAS/BRAF biomarker profiling was incorporated into recommendation guidelines^2,7 following retrospective subgroup analysis of clinical trials.^4,5,21 The retrospective data presented here, in addition to retrospective analysis of the PARADIGM trial,⁴³ support the use of tissue or liquid CGP to identify a broad set of genomic alterations potentially mediating primary resistance to EGFR mAb therapy to guide treatment selection.