The Randomized Phase II ARC-9 Study of Etrumadenant-Based Therapy versus Regorafenib in Patients with Previously Treated Metastatic Colorectal Cancer

Abstract

Purpose:

Targeting the adenosine pathway may enhance the efficacy of chemo/immunotherapy regimens in patients with heavily pretreated advanced metastatic colorectal cancer (mCRC), for whom treatment options are limited.

Patients and Methods:

The phase II ARC-9 study, Cohort B (NCT04660812), evaluated the efficacy and safety of etrumadenant (A_2a and A_2b receptor antagonist), zimberelimab (anti–PD-1 mAb), FOLFOX, and bevacizumab (EZFB) versus regorafenib in patients with third-line mCRC who previously progressed on oxaliplatin- and irinotecan-containing regimens.

Results:

From September 21, 2021, to September 12, 2022, 112 patients were randomized 2:1 to EZFB (n = 75) or regorafenib (n = 37). As of November 13, 2023, the median survival follow-up was 20.4 months. The primary endpoint of progression-free survival (PFS) was improved with EZFB (6.2 months) versus regorafenib [2.1 months; hazard ratio (HR), 0.27; 95% confidence interval (CI), 0.17–0.43; nominal P < 0.0001], as was the secondary endpoint of overall survival (OS; EZFB, 19.7 months; regorafenib, 9.5 months; HR, 0.37; 95% CI, 0.22–0.63; nominal P = 0.0003). The confirmed overall response rate was 17% (90% CI, 10.6%–26.1%) with EZFB and 3% (90% CI, 0.1%–12.2%) with regorafenib. Treatment-emergent adverse events (TEAE), grade ≥3 TEAEs, and TEAEs leading to discontinuation of all study treatments were reported in 99%, 82%, and 5% of the EZFB arm and in 87%, 49%, and 17% of the regorafenib arm, respectively.

Conclusions:

EZFB significantly improved survival outcomes compared with regorafenib in patients with mCRC as a third-line treatment, with a manageable safety profile. Further investigation is warranted, given the clinically meaningful improvements in PFS and OS.

Translational Relevance.

Adenosine blockade may improve the efficacy of anti–PD-1 and oxaliplatin-containing chemotherapy in patients with treatment-refractory advanced metastatic colorectal cancer (mCRC). Etrumadenant, a small-molecule, selective dual antagonist of adenosine receptors A_2a and A_2b, has demonstrated clinical benefit in phase I studies. The phase II ARC-9 study, Cohort B, evaluated etrumadenant, zimberelimab, FOLFOX, and bevacizumab (EZFB) versus regorafenib in patients with third-line mCRC. The median progression-free survival and median overall survival significantly improved with EZFB versus regorafenib, with a manageable safety profile. The clinically meaningful improvements in survival outcomes support further investigation of EZFB.

Introduction

Colorectal cancer is the third most common cancer and the second leading cause of cancer death worldwide (1). 5-Fluorouracil–based chemotherapy with oxaliplatin and/or irinotecan is a standard approach for first- and second-line treatments of metastatic colorectal cancer (mCRC), in combination with appropriate biological therapy based on RAS status and tumor location. Patients who have disease progression after receiving first- and second-line treatments with these therapies are considered to have chemotherapy-refractory disease (2).

Patients with refractory mCRC will often not progress while actively receiving oxaliplatin as part of their initial systemic therapy. Due to oxaliplatin-related toxicities, including cumulative sensory neuropathy, oxaliplatin treatment is typically limited to 3 to 4 months, and patients commonly progress on maintenance therapy rather than while actively receiving oxaliplatin as part of first- or second-line treatment (3, 4). The median progression-free survival (PFS) of patients who receive FOLFOX with appropriate biological therapy is approximately 11 months in the first-line setting and 6 months in the second-line setting (3–6). Oxaliplatin reintroduction in late-line mCRC is frequently used in real-world clinical practice (7); however, there are insufficient data to justify routinely reintroducing oxaliplatin in late-line mCRC. Moreover, larger randomized studies, such as the OPTIMOX1, OPTIMOX2, and CAIRO3 studies, involved first-line patients with no intervening treatments other than maintenance therapy (8, 9). Trifluridine/tipiracil with or without bevacizumab, regorafenib, and fruquintinib are recommended treatments for chemotherapy-refractory disease and are associated with a median overall survival (OS) of 7 to 11 months, highlighting the unmet need for novel therapies for patients with treatment-refractory mCRC (10–13).

Targeting the adenosine pathway is a promising therapeutic strategy in mCRC (14). The tumor microenvironment is rich in adenosine, which is produced via CD73 in response to hypoxia, inflammation, and ischemia (14, 15). Adenosine stimulates the G protein–coupled receptors A_2a (A_2aR) and A_2b (A_2bR) to drive downstream immunosuppressive signaling (16–18). In preclinical studies, A_2aR and A_2bR inhibition protected against metastasis and increased immune activation against tumors (19). Through A_2bR signaling, adenosine regulates key stromal cell types that promote tumor progression in colorectal cancer, including cancer-associated fibroblasts and macrophages (18, 20).

A_2aR and A_2bR antagonism enhances the efficacy of chemotherapy and could lead to prolonged survival (21). Dual inhibition of A_2aR and PD-L1 can enhance lymphocyte tumor infiltration, whereas immunosuppression can be mitigated by inhibition of PD-L1/PD-1 and vascular endothelial growth factor or vascular endothelial growth factor receptor [VEGF(R)] (22, 23). Early-stage trials indicated that adenosine blockade and checkpoint inhibition induced sustained antitumor activity and improved the efficacy of FOLFOX reintroduction (24–26). In colorectal cancer, targeting adenosine in combination with oxaliplatin may have a complementary effect because platinum agents induce immunogenic cell death and the subsequent release of ATP, which is converted into adenosine that suppresses immunogenicity via interaction with A_2aR (16–18, 27).

Etrumadenant is a small-molecule, selective dual antagonist of A_2aR and A_2bR that prevents extracellular adenosine-mediated immunosuppression (28, 29). In multiple phase I studies, etrumadenant combined with chemotherapy and/or zimberelimab, a PD-1 inhibitor, was well tolerated and demonstrated clinical benefit in heavily pretreated patients across multiple solid tumor disease indications (24–26, 29). Biomarker analyses showed a correlation between lymphocyte infiltration and the extent of clinical benefit from etrumadenant and modified FOLFOX-6 (mFOLFOX-6) in late-line patients with mCRC (30). We report the results of the ARC-9 study, Cohort B, an international, phase II, randomized clinical trial in patients with refractory mCRC evaluating the efficacy and safety of etrumadenant and zimberelimab with mFOLFOX-6 and bevacizumab (EZFB) compared with regorafenib, which was the standard-of-care regimen at the time of the study design.

Patients and Methods

Study design and patients

The ARC-9 study, Cohort B, was conducted at 29 sites in South Korea, France, Italy, Spain, the United Kingdom, and the United States. Eligible patients were ≥18 years of age (≥19 years for South Korea) with confirmed unresectable mCRC, measurable disease per RECIST v1.1, an Eastern Cooperative Oncology Group performance status of 0 or 1, and a life expectancy of 3 or more months. Patients were required to have experienced disease progression during or following no more than two separate lines of treatment for mCRC, consisting of oxaliplatin- and irinotecan-containing chemotherapy in combination with anti-VEGF(R) or anti–epidermal growth factor receptor (EGFR). Patients could not have experienced disease progression within 2 months of the last dose of oxaliplatin for metastatic disease; however, if patients had disease progression within 6 months of the last dose of oxaliplatin for adjuvant treatment, this would count as a line of therapy. Patients with a BRAF V600E mutation and those previously treated with immune checkpoint blockade therapies were excluded. Complete inclusion and exclusion criteria are shown in Supplementary Table S1. Microsatellite instability (MSI) status was not obtainable for many patients, so polymerase chain reaction (PCR) analysis was used to assess MSI in all patients with tissue samples available at baseline.

This study was conducted in full conformance with the International Council for Harmonization E6 guideline for Good Clinical Practice, the Declaration of Helsinki, Institutional Review Board regulations, and all other applicable local regulations (ClinicalTrials.gov Identifier, NCT04660812; Supplementary Table S2). All patients provided written informed consent before any study procedures.

Procedures

The ARC-9 study, Cohort B (NCT04660812), is an ongoing, phase II, open-label, multicenter platform trial to evaluate the efficacy and safety of etrumadenant-based combination therapy in patients with mCRC. Patients were randomized 2:1 to receive EZFB or regorafenib via an interactive response technology system, stratified by region (United States vs. rest of the world). EZFB was administered as once-daily oral etrumadenant 150 mg on a 28-day continuous dosing cycle; intravenous zimberelimab 480 mg every 4 weeks on day 1 of each 28-day dosing cycle; intravenous mFOLFOX-6 every 2 weeks on days 1, 2, 15, and 16 of a 28-day dosing cycle; and intravenous bevacizumab 5 mg/kg every 2 weeks (24, 31, 32). The mFOLFOX-6 dosing consisted of oxaliplatin 85 mg/m² by intravenous infusion every 2 weeks, leucovorin 400 mg/m² by intravenous infusion every 2 weeks, fluorouracil 400 mg/m² by intravenous bolus, and fluorouracil 2,400 mg/m² by intravenous infusion (continuous 46-hour infusion) every 2 weeks (24, 25). Regorafenib was administered at a starting dose of 80 mg once daily for 1 week, followed by a dose escalation of 40 mg each week during cycle 1 if no significant drug-related toxicity was observed, reaching 120 mg per day during week 2 and 160 mg per day during week 3 (as per institutional guidelines). Subsequent cycles followed the current label recommendations, with 160 mg per day orally administered on days 1 to 21 of each 28-day dosing cycle (33).

Tumors were assessed at baseline and every 8 weeks thereafter until disease progression, new anticancer therapy, withdrawal of consent, study termination, or death. Treatment continued until disease progression, the occurrence of unacceptable toxicities, or withdrawal of consent. Patients who progressed on regorafenib were permitted to cross over to EZFB after a 5-day washout period. Patients who crossed over were evaluated following the same schedule as patients randomized to EZFB, starting at the time of the first cross-over dose (day 1).

Endpoints and assessments

The primary efficacy endpoint was investigator-assessed PFS in all randomized patients (efficacy-evaluable population). PFS was defined as the time from randomization until the first documentation of progressive disease or death, whichever occurred first. Patients without documented disease progression at the time of analysis were censored on the date of their last adequate tumor assessment. If no tumor assessment was performed after randomization, PFS was censored on the randomization date with a duration of 1 day.

Secondary endpoints included confirmed objective response rate (ORR), defined as the proportion of patients with a best overall response of complete response (CR) or partial response (PR) according to RECIST v1.1; disease control rate, defined as the proportion of patients with a best overall response of CR, PR, or stable disease; duration of response (DOR), defined as the time from first CR or PR to disease progression or death; and OS, defined as the time (months) from randomization until death from any cause. If death due to any cause did not occur during the study, OS was censored on the date of last contact after the start of treatment.

Adverse events (AE) were evaluated in all patients who received any amount of study treatment (safety-evaluable population). Patients who crossed over from regorafenib to EZFB contributed safety information until the time of treatment cross-over.

Statistical analysis

The ARC-9 study, Cohort B, planned to enroll 105 patients in a 2:1 ratio (70 in the EZFB arm and 35 in the regorafenib arm). The sample size was calculated using 80% power with a two-sided significance level of 0.05 to detect a hazard ratio (HR) of 0.5 in PFS, with a 2:1 randomization ratio using a log-rank test. To control the type I error, the O’Brien–Fleming boundaries were approximated using the Lan–DeMets alpha spending function (34). The median PFS, OS, and DOR were estimated in each treatment arm using Kaplan–Meier methodology. HRs with 95% confidence intervals (CI) were calculated for PFS and OS using a Cox model, stratified by geographic region (United States vs. rest of the world). A stratified log-rank test was used to compare PFS and OS between treatment arms. The ARC-9 study, Cohort B, was not designed with sufficient power to control for alpha in multiplicity testing. Thus, all P values are nominal. The 90% CIs for ORR were calculated using the Clopper–Pearson exact method. A rank-preserving structural failure time (RPSFT) model was constructed to determine the OS in the regorafenib arm if cross-over to the EZFB arm had not been allowed.

Results

Patients

Overall, 112 patients were enrolled in the ARC-9 study, Cohort B, between September 21, 2021, and September 12, 2022, across 29 sites in five countries (Supplementary Fig. S1). In total, 75 patients were randomized to the EZFB arm and 37 to the regorafenib arm. Most patients in both arms had liver metastases (EZFB, 71%; regorafenib, 78%) and had received prior oxaliplatin treatment in the metastatic setting (EZFB, 84%; regorafenib, 84%; Table 1). One (1%) patient in the EZFB arm had MSI-high disease per PCR. The median (range) time to progression or discontinuation of the last prior oxaliplatin-containing regimen was 9 (0.7–57.2) months in the EZFB arm and 9.8 (0.9–33.7) months in the regorafenib arm. Patients had received prior anti-VEGF(R) treatment (EZFB, 85%; regorafenib, 97%) and prior anti-EGFR treatment (EZFB, 36%; regorafenib, 30%) at comparable rates in both arms.

Table 1.

Baseline demographics and clinical characteristics in the efficacy-evaluable population.

Characteristic	EZFB (n = 75)	Regorafenib (n = 37)
Age, years, median (range)	58 (35–79)	60 (35–81)
Sex, n (%)	​	​
Female	22 (29)	15 (41)
Male	53 (71)	22 (60)
Country, n (%)	​	​
France	15 (20)	4 (11)
Italy	5 (7)	2 (5)
South Korea	27 (36)	14 (38)
Spain	2 (3)	3 (8)
United States	26 (35)	14 (38)
Race, n (%)	​	​
Asian	28 (37)	14 (38)
Black or African American	2 (3)	1 (3)
White	25 (33)	14 (38)
Multiple	0	1 (3)
Not reported	20 (27)	7 (19)
ECOG PS, n (%)	​	​
0	30 (40)	18 (49)
1	45 (60)	19 (51)
Primary diagnosis, n (%)	​	​
Colon	56 (75)	27 (73)
Rectal	19 (25)	10 (27)
Primary tumor location, n (%)	​	​
Left	53 (71)	23 (62)
Right	22 (29)	12 (32)
Unknown	0	2 (5)
Overall TNM stage at initial diagnosis, n (%)	​	​
Stage I	1 (1)	0
Stage II	5 (7)	1 (3)
Stage III	13 (17)	7 (19)
Stage IV	56 (75)	29 (78)
Metastatic site, n (%)	​	​
Liver	53 (71)	29 (78)
Peritoneal	18 (24)	9 (24)
Lung only	4 (5)	2 (5)
Number of metastatic sites	​	​
1–2	46 (61)	17 (46)
3+	29 (39)	20 (54)
MSI status,an (%)	​	​
Stable (MSS or MSI-L)	61 (81)	30 (81)
Unstable (MSI-H)	1 (1)	0
Missing	13 (17)	7 (19)
KRAS,bn (%)	​	​
Mutant	38 (51)	20 (54)
Wild type	27 (36)	11 (30)
Missing	10 (13)	6 (16)
Prior irinotecan,cn (%)	73 (97)	37 (100)
Prior oxaliplatin,d,en (%)	74 (99)	36 (97)
Metastatic	63 (84)	31 (84)
1L	53 (71)	27 (73)
2L	10 (13)	5 (14)
3L	5 (7)	2 (5)
4L	1 (1)	0
Adjuvant/neoadjuvant	10 (13)	5 (14)
Locally advanced	4 (5)	2 (5)
Duration of the last prior oxaliplatin-containing regimen, months, median (range)f	​	​
All settings	9 (0.7–57.2)	9.8 (0.9–33.7)
1L metastatic	8.9 (0.7–57.2)	9.9 (0.9–33.6)
2L+ metastatic	9.2 (2.5–20.1)	8.5 (5.8–33.7)
Prior anti-VEGF(R),gn (%)	64 (85)	36 (97)
Prior anti-EGFR, n (%)	27 (36)	11 (30)

Abbreviations: 1L, first-line treatment; 2L+, second-line treatment or greater; 3L, third-line treatment; 4L, fourth-line treatment; ECOG PS, Eastern Cooperative Oncology Group performance status; EGFR, endothelial growth factor receptor; EZFB, etrumadenant and zimberelimab combined with mFOLFOX-6 and bevacizumab; mFOLFOX-6, modified FOLFOX-6; MSI, microsatellite instability; MSI-H, MSI-high; MSI-L, MSI-low; MSS, microsatellite stable; PCR, polymerase chain reaction; TNM, tumor–node–metastasis; VEGF(R), vascular endothelial growth factor or vascular endothelial growth factor receptor.

^aMicrosatellite status was assessed using PCR analysis of available baseline tissue samples.

^b KRAS based on next-generation sequencing data using a nonvalidated exploratory assay.

^cClinical database was subsequently updated after the November 13, 2023, data cutoff date, showing that all patients had received prior irinotecan per the eligibility criteria.

^dClinical database was subsequently updated after the November 13, 2023, data cutoff date, showing that all patients had received prior oxaliplatin per the eligibility criteria.

^eA patient could be counted more than once if they had prior oxaliplatin therapy in multiple settings.

^fCalculated for patients who received prior oxaliplatin therapy as the time from the first prior oxaliplatin dose setting to progression or discontinuation of the last prior oxaliplatin-containing regimen.

^gIncludes bevacizumab, ramucirumab, and aflibercept, regardless of the treatment setting.

Efficacy

At the data cutoff date of November 13, 2023, among the 112 patients in the efficacy-evaluable population (EZFB, n = 75; regorafenib, n = 37), EZFB was associated with improved PFS compared with regorafenib [HR, 0.27 (95% CI, 0.17–0.43); nominal P < 0.0001; Fig. 1A]. The median PFS with EZFB was 6.2 months compared with 2.1 months with regorafenib, with a median (range) duration of follow-up of 20.4 (14.1–25.8) months. At 12 months, 16% (95% CI, 6.5%–25.2%) and 3% (95% CI, <1% to 8.4%) of patients in the EZFB and regorafenib arms remained alive and progression-free, respectively.

Figure 1.

Survival outcomes. Kaplan–Meier curves for (A) PFS and (B) OS. CI, confidence interval; EZFB, etrumadenant and zimberelimab with mFOLFOX-6 and bevacizumab; mFOLFOX-6, modified FOLFOX-6; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; rego, regorafenib. ^aHR and 95% CIs were calculated using a Cox model, stratified by geographic region. ^bStudy was not designed with sufficient power to control for alpha in multiplicity testing.

The median OS was 19.7 months in the EZFB arm versus 9.5 months in the regorafenib arm [HR, 0.37 (95% CI, 0.22–0.63); nominal P = 0.0003; Fig. 1B]. OS rates at 12 months were 64% (95% CI, 51.9%–75.2%) and 34% (95% CI, 17.1%–50.7%) in the EZFB and regorafenib arms, respectively. OS rates at 18 months were 53% (95% CI, 40.4%–66.2%) and 14% (95% CI, 0.7%–27.6%) in the EZFB and regorafenib arms, respectively. Overall, 21 (57%) patients crossed over from the regorafenib arm to the EZFB arm at the time of progression. In the RPSFT model that did not allow cross-over, the OS HR for the EZFB arm versus the non-cross-over regorafenib arm was 0.13 (95% CI, 0.06–0.28; Supplementary Fig. S2).

The median OS was consistently longer in the EZFB arm versus the regorafenib arm, including in patients with baseline liver metastasis, peritoneal metastasis, KRAS mutation status per next-generation sequencing, and response to prior therapy, as well as in patients with both short (<9 months) and long (≥9 months) durations to progression or end of the first oxaliplatin-containing chemotherapy regimen in the metastatic setting (Fig. 2). The risk of disease progression or death was consistently lower in the EZFB arm versus the regorafenib arm in all subgroups analyzed (Supplementary Fig. S3). In the subgroup of patients with baseline liver metastasis (EZFB, n = 53; regorafenib, n = 29), the median PFS was consistent with the primary analysis [EZFB, 5.7 months; regorafenib, 2 months; HR, 0.19 (95% CI, 0.10–0.35)], as was the median OS [EZFB, 19.7 months; regorafenib, 9.1 months; HR, 0.36 (95% CI, 0.20–0.66); Fig. 3].

Figure 2.

OS analysis by subgroups. 1L, first-line therapy; BOR, best overall response; CI, confidence interval; CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; EZFB, etrumadenant and zimberelimab with mFOLFOX-6 and bevacizumab; HR, hazard ratio; mFOLFOX-6, modified FOLFOX-6; MSI-L, MSI-low; MSS, microsatellite stable; NA, not applicable; NGS, next-generation sequencing; PCR, polymerase chain reaction; PD, progressive disease; PR, partial response; rego, regorafenib; SD, stable disease. ^aPer the inclusion criteria, disease progression occurred >2 months after the last prior oxaliplatin dose in the metastatic setting (n = 11), or inclusion was approved by the medical monitor (n = 1).

Figure 3.

Survival outcomes in patients with baseline liver metastasis. Kaplan–Meier curves for (A) PFS and (B) OS. CI, confidence interval; EZFB, etrumadenant and zimberelimab with mFOLFOX-6 and bevacizumab; mFOLFOX-6, modified FOLFOX-6; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; rego, regorafenib. ^aHR and 95% CIs were calculated using a Cox model, stratified by geographic region.

The confirmed ORR was 17% (90% CI, 10.6%–26.1%) in the EZFB arm and 3% (90% CI, 0.1%–12.2%) in the regorafenib arm (Fig. 4). The median DOR was 11.5 months (95% CI, 3.9–12.6) in the EZFB arm and not evaluable in the regorafenib arm because there was only one responder. Patients in the EZFB arm achieved PR and stable disease with and without prior anti-VEGF(R) treatment. The confirmed disease control rate was 63% (90% CI, 52.6%–72%) in the EZFB arm and 8% (90% CI, 2.2%–19.6%) in the regorafenib arm.

Figure 4.

Confirmed best overall response (CBOR) and change from baseline in the sum of target lesion diameters in (A) patients treated with EZFB and (B) patients treated with rego. EZFB, etrumadenant and zimberelimab with mFOLFOX-6 and bevacizumab; mFOLFOX-6, modified FOLFOX-6; PD, progressive disease; PR, partial response; rego, regorafenib; SD, stable disease; SLD, sum of target lesion diameters; VEGF(R), vascular endothelial growth factor or vascular endothelial growth factor receptor.

Safety

Among the 109 (EZFB, n = 74; regorafenib, n = 35) patients in the safety-evaluable population, the median (range) treatment duration was 26 (<1 to 98) weeks in the EZFB arm and 7 (<1 to 95) weeks in the regorafenib arm (Table 2). The median (range) number of cycles of treatment received was 7 (1–24) in the EZFB arm and 2 (1–22) in the regorafenib arm. Overall, 73 (99%) patients in the EZFB arm and 31 (89%) patients in the regorafenib arm reported a treatment-emergent AE (TEAE). In the EZFB arm, 44 (60%) patients reported a TEAE that led to the discontinuation of at least one study treatment, which was primarily because of oxaliplatin in 39 (53%) patients. Four (5%) patients in the EZFB arm discontinued all study treatments because of TEAEs. TEAEs in the EZFB arm were more commonly attributed to mFOLFOX-6 rather than etrumadenant, zimberelimab, or bevacizumab.

Table 2.

Safety summary in the safety-evaluable population.

Parameter	EZFB (n = 74)	Regorafenib (n = 35)
Treatment duration, weeks, median (range)	26 (<1–98)	7 (<1–95)
Any TEAEs, n (%)	73 (99)	31 (89)
Serious TEAEs, n (%)	37 (50)	9 (26)
Any TEAEs related to any study drug, n (%)	72 (97)	27 (77)
Related to etrumadenant	50 (68)	—
Related to zimberelimab	53 (72)	—
Related to mFOLFOX-6	72 (97)	—
Related to bevacizumab	49 (66)	—
Related to regorafenib	—	27 (77)
Any TEAEs leading to discontinuation of ≥1 study drug, n (%)	44 (60)	6 (17)
Leading to oxaliplatin discontinuationa	39 (53)	—
Any grade 3 or higher TEAEs, n (%)	61 (82)	17 (49)
Most common grade 3 or higher (>5%)	​	​
Neutropeniab	35 (47)	0
Nausea	8 (11)	0
Blood bilirubin increased	7 (10)	2 (6)
Diarrhea	6 (8)	0
Hypersensitivityc	6 (8)	0
WBC count decreased	6 (8)	0
Anemia	5 (7)	1 (3)
Neuropathyd	5 (7)	0
Vomiting	5 (7)	0
Abdominal pain	4 (5)	1 (3)
Thrombocytopeniae	4 (5)	1 (3)
Fatigue	2 (3)	3 (9)
Blood alkaline phosphatase increased	1 (1)	2 (6)
Hypertension	1 (1)	2 (6)
Palmar–plantar erythrodysaesthesia syndrome	0	2 (6)
Sepsis	0	2 (6)
Any TEAEs leading to death, n (%)	0	0
Any immune-mediated TEAEs,fn (%)	12 (16)	2 (6)
Any grade 3 or higherg	4 (5)	0
Hemolytic anemiah	2 (3)	0
Skin reaction	2 (3)	0
Adrenal insufficiency, any grade	2 (3)	0
Hemophagocytic lymphohistiocytosis, any grade	0	1 (3)
Hypothyroidism, any grade	4 (5)	2 (6)
Hyperthyroidism, any grade	1 (1)	0
Pneumonitis, any grade	2 (3)	0

Abbreviations: AE, adverse event; EZFB, etrumadenant and zimberelimab combined with mFOLFOX-6 and bevacizumab; mFOLFOX-6, modified FOLFOX-6; TEAE, treatment-emergent AE; WBC, white blood cell.

^aMost common AEs (>5%) leading to oxaliplatin discontinuation were peripheral neuropathy, infusion-related reaction, and peripheral sensory neuropathy.

^bNeutropenia was a grouped term that included neutropenia and decreased neutrophil count.

^cHypersensitivity was a grouped term that included drug hypersensitivity, hypersensitivity reaction, infusion-related reaction, and anaphylactic reaction.

^dNeuropathy was a grouped term that included peripheral neuropathy, peripheral sensory neuropathy, neuralgia, and peripheral motor neuropathy.

^eThrombocytopenia was a grouped term that included thrombocytopenia, decreased platelet count, and decreased platelet production.

^fImmune-related reactions were identified by investigators according to sponsor-defined cluster/grouping terms, preferred terms, and maximum reported severity; events with multiple occurrences per patient were counted once at maximum severity.

^gThere were no grade 5 immune-mediated AEs.

^hHemolytic anemia was a grouped term that included autoimmune hemolytic anemia and hemolytic anemia. TEAEs were defined as any AEs that started on or after the start of study treatment and within 90 days of the last dose date for the EZFB arm and 30 days of the last dose date for the regorafenib arm, excluding any AEs that occurred after the initiation of a new anticancer therapy. Medical history and AEs were coded using the Medical Dictionary for Regulatory Activities version 24.0.

Grade ≥3 TEAEs were reported in 61 (82%) patients in the EZFB arm and 17 (49%) patients in the regorafenib arm. The most common grade ≥3 TEAEs in the EZFB arm were consistent with the safety profile of mFOLFOX-6 and included neutropenia and nausea. No TEAEs leading to death were reported in either treatment arm. Immune-related AEs were reported in 12 (16%) patients in the EZFB arm and 2 (6%) patients in the regorafenib arm. In the EZFB arm, 3 (4%) grade 3 immune-related AEs and 1 (1%) grade 4 immune-related AE (hemolytic anemia) were reported; there were no grade 3 or higher immune-related AEs reported in the regorafenib arm.

Discussion

In the phase II, international, randomized ARC-9 study, Cohort B, statistically significant improvements in PFS (nominal P < 0.0001) and OS (nominal P = 0.0003) were observed with etrumadenant combined with chemo/immunotherapy compared with regorafenib in heavily pretreated patients with primarily microsatellite-stable mCRC. The improvements in PFS and OS were clinically meaningful, with a median PFS of 6.2 months with EZFB versus 2.1 months with regorafenib [HR, 0.27 (95% CI, 0.17–0.43); nominal P < 0.0001] and a median OS of 19.7 months with EZFB versus 9.5 months with regorafenib [HR, 0.37 (95% CI, 0.22–0.63); nominal P = 0.0003]. The improvements in PFS and OS with EZFB versus regorafenib were maintained in patients with baseline liver or peritoneal metastases, KRAS mutation status per next-generation sequencing, and response to prior oxaliplatin therapy, as well as in patients with both short (<9 months) and long (≥9 months) durations to progression or end of the first oxaliplatin-containing chemotherapy regimen in the metastatic setting. In the EZFB arm, grade ≥3 TEAEs were reported in 82%, but the overall frequency and severity of TEAEs did not differ from the expected toxicity profile of mFOLFOX-6 and PD-(L)1 inhibitors (35). Overall, the clinically meaningful improvements in PFS and OS from the ARC-9 study, Cohort B, support future registrational studies in advanced colorectal cancer.

Selective A_2aR and A_2bR inhibition may potentiate the cytotoxic activity of infiltrating immune cells in the tumor microenvironment and complement the anticancer effects of chemo/immunotherapy to improve OS. Oxaliplatin chemotherapy drives increased adenosine via induction of immunogenic cell death and ATP release (27). Inhibition of the adenosine signaling pathway with etrumadenant and PD-1 with zimberelimab may ameliorate the immunosuppressive tumor microenvironment and improve the effectiveness of mFOLFOX-6 and bevacizumab treatment (16–18, 27).

Patients enrolled in the ARC-9 study, Cohort B, were required to have progressed on or after ≤2 separate lines of prior treatment with an oxaliplatin- and irinotecan-containing regimen in combination with anti-VEGF(R) or anti-EGFR. Patients who experienced disease progression within 2 months of their last dose of prior oxaliplatin were excluded. The patient population in the ARC-9 study, Cohort B, is considered to have refractory disease, which is typically associated with the start of third-line treatments of trifluridine/tipiracil with bevacizumab, fruquintinib, or regorafenib (2, 10–13). During enrollment in the ARC-9 study, Cohort B, regorafenib and single-agent trifluridine/tipiracil were the approved therapies for refractory mCRC, and the current survival in the fluoropyrimidine-refractory setting remains 7 to 11 months with the available agents (10–13). Some retrospective reports have suggested that FOLFOX rechallenge may achieve better outcomes than regorafenib in the third-line setting; however, these studies involved highly selected patient populations and are constrained by small sample sizes and inherent biases (36, 37). Notably, regorafenib has been used in multiple prospective phase III trials in refractory colorectal cancer as a comparator arm, and none of these trials met their primary endpoint of improved OS versus regorafenib (38–40). Moreover, 57% of patients in the regorafenib arm crossed over to EZFB, further reducing the likelihood that chemotherapy was the primary driver of the observed OS benefit. In this trial, we randomized patients to regorafenib, which was dosed using the more tolerable and efficacious regorafenib dose optimization method (33). Overall, the patients enrolled in the ARC-9 study, Cohort B, were similar to those enrolled in other major phase III registrational refractory mCRC studies, including in the prevalence of liver metastases, primary tumor sidedness, and KRAS mutation status per next-generation sequencing (10–12, 41).

There are limited high-quality prospective data to support the reintroduction of oxaliplatin in refractory mCRC, and its use is controversial in clinical practice (42). It is important to recognize that the only large, randomized, prospective clinical trials evaluating the reintroduction of an oxaliplatin-containing regimen were the OPTIMOX1, OPTIMOX2, and CAIRO3 studies, which were conducted in first-line patients with mCRC (3, 8, 9). Patients in these studies received the following treatment: (i) initial fluoropyrimidine and oxaliplatin-based treatment (FOLFOX or CAPOX) ± biological therapy, followed by (ii) maintenance treatment (fluoropyrimidine ± biologic) or a chemotherapy-free interval and (iii) retreatment with FOLFOX or CAPOX at progression (3, 8, 9).

The results from a combined post hoc analysis of the OPTIMOX1 and OPTIMOX2 studies in patients pretreated with irinotecan or oxaliplatin showed that the 189 patients with refractory mCRC, who had a PR to initial oxaliplatin therapy, experienced a median OS of 14 months after the first reintroduction of FOLFOX (43). Furthermore, treatment with FOLFOX and biological therapy in second-line, oxaliplatin-naive patients is associated with an approximately 11-month median OS (6). Although these examples provide valuable benchmarks for the efficacy expectations with the reintroduction of oxaliplatin, patients in the ARC-9 study, Cohort B, received more cumulative oxaliplatin before initiation and were at a more advanced stage than the patients in OPTIMOX1 and OPTIMOX2, having received study treatment in the third line. Furthermore, the efficacy of EZFB was maintained, regardless of sensitivity to prior oxaliplatin (Fig. 2). These considerations should inform the interpretation of the 19.7-month median OS with EZFB for patients with refractory mCRC, as it seems unlikely that these results can be attributed to mFOLFOX-6 and bevacizumab retreatment alone. Importantly, to our knowledge, the ARC-9 study, Cohort B, is the first prospective study of FOLFOX rechallenge in the third-line treatment of patients with mCRC.

In the ARC-9 study, Cohort B, EZFB demonstrated clinically meaningful improvements in PFS and OS. A 2020 systematic literature review of studies investigating third-line treatments for mCRC reported a maximum median OS of 10 months across five randomized trials (44). Furthermore, in 2023, the phase III SUNLIGHT trial reported a median OS of 10.8 months in third-line mCRC with trifluridine/tipiracil and bevacizumab (12). Thus, this EZFB combination deserves further evaluation in mCRC.

Although the results are promising, the ARC-9 study, Cohort B, has several limitations. The study was open-label, and the sample size was small, meaning the patient population may not be representative of all patients (Supplementary Table S3), although all subgroups seemed to benefit. Enrollment was not restricted to patients with microsatellite-stable disease, given the known sensitivity of MSI-high tumors to immune checkpoint inhibitors, and not all patients had MSI status data. However, the efficacy results were consistent across MSI subgroups (Fig. 2). Indeed, the demonstrated activity of EZFB in patients with microsatellite-stable disease and patients with liver metastases could be a major advantage of using this regimen. A limitation of this study is the challenge in determining the contribution of components of etrumadenant, zimberelimab, mFOLFOX-6, and bevacizumab in the EZFB arm; however, the efficacy outcomes exceed expectations from the existing data with the recycling of standard-of-care agents in the fluoropyrimidine-refractory setting, with a manageable toxicity profile. Finally, the cross-over from regorafenib to EZFB may have influenced outcomes and could be a limitation of the study, but this limitation would only strengthen the efficacy outcomes observed in the EZFB arm.

In the randomized ARC-9 study, Cohort B, selective A_2aR and A_2bR inhibition with etrumadenant-based chemo/immunotherapy met the primary endpoint of longer PFS and provided encouraging efficacy compared with regorafenib in patients with advanced, microsatellite-stable mCRC. Moreover, EZFB did not demonstrate any unexpected safety signals, with most AEs attributed to mFOLFOX-6. Further investigation of this regimen is warranted.

Supplementary Material

Supplementary Table S1

Inclusion and exclusion criteria

Supplementary Table S2

Ethics committee approval information

Supplementary Table S3

Representativeness of study participants

Supplementary Figure S1

Patient dispositions. EZFB, etrumadenant, and zimberelimab with mFOLFOX-6 and bevacizumab; mFOLFOX-6, modified FOLFOX-6.

Supplementary Figure S2

OS if crossover from regorafenib to EZFB had not been permitted, using the RPSFT model. A key assumption of the RPSFT model is that the experimental therapy (EZFB) has the same treatment effect, regardless of whether the experimental therapy is given from the initial randomization or from the time of crossover. EZFB, original group that received etrumadenant and zimberelimab with mFOLFOX-6 and bevacizumab; mFOLFOX-6, modified FOLFOX-6; OS, overall survival; rego(o), original regorafenib group; rego(r), regorafenib group with RPSFT model; RPSFT, rank-preserving structural failure model. aHazard ratio and 95% CIs were calculated using a Cox model, stratified by geographic region.

Supplementary Figure S3

PFS subgroups analysis. 1L, first-line therapy; BOR, best overall response; CR, complete response; ECOG PS, Eastern Cooperative Oncology Group performance status; EZFB, etrumadenant and zimberelimab with mFOLFOX-6 and bevacizumab; HR, hazard ratio; mFOLFOX-6, modified FOLFOX-6; MSI, microsatellite instability; MSI-L, MSI low; MSS, microsatellite stable; NGS, next-generation sequencing; PCR, polymerase chain reaction; PD, progressive disease; PFS, progression-free survival; PR, partial response; rego, regorafenib; SD, stable disease. aDisease progression occurred >2 months after the last prior oxaliplatin dose in the metastatic setting (n = 11), or inclusion was approved by the medical monitor (n = 1)

Data Availability

Requests for data from any qualified researcher who engages in rigorous, independent scientific research will be considered if the clinical trial data are not part of an ongoing or planned regulatory submission. For information on the process or to submit a request, visit https://trials.arcusbio.com/our-transparency-policy.

Authors’ Disclosures

M. Cecchini reports personal fees from Arcus Biosciences during the conduct of the study, as well as personal fees from Seattle Inc., Taiho Pharmaceutical, Regeneron, Elevate Oncology, Loxo@Lilly, BeOne Medicines, Arcus Biosciences, Modifi Bio, Incyte, AbbVie, Parabilis Medicines, Wren Therapeutics, Takeda, Exelixis, Astellas, and HiTOP and grants from NIH (1K08CA255465-01A1 and 1K08CA255465-01A1) outside the submitted work. S.-W. Han reports grants from Arcus Biosciences during the conduct of the study, as well as grants and personal fees from AstraZeneca and AbbVie; personal fees from Bayer, Natera, and Ono Pharmaceutical; and grants from BeyondBio, Cell Biotech, Astellas, Boryung, Genentech, GC Biopharma, Hanmi, Hengrui Pharmaceutical, IMBdx, Janssen, Jeil Pharmaceutical, Leap Therapeutics, Loxo, Eli Lilly, Mirati Therapeutics, MSD, Roche, Seagen, and Turnstone Biologics outside the submitted work. K.-W. Lee reports grants (research funding to K.-W. Lee's institution for conducting clinical trials) from Arcus Biosciences during the conduct of the study. K.-W. Lee also reports grants (research funding to K.-W. Lee's institution for conducting clinical trials) from ABL Bio, ALX Oncology, Amgen, Astellas, AstraZeneca/MedImmune, Beigene, Bolt Biotherapeutics, Daiichi Sankyo, GC Pharma, Genome & Company, Five Prime Therapeutics, Ildong Pharmaceutical, InventisBio, Leap Therapeutics, LSK BioPharma, Macrogenics, MedPacto, Merck KGaA, MSD, Ono Pharmaceutical, Pfizer, Pharmacyclics, Roche, Seattle Genetics, Taiho Pharmaceutical, Trishula Therapeutics, Y-BIOLOGICS, and Zymeworks; personal fees (honoraria) from Boryung, JW Pharmaceutica, and Ono Pharmaceutical; and personal fees (a consulting or advisory role) from Bayer, Bristol Myers Squibb, Daiichi Sankyo, Metafines, MSD, Ono Pharmaceutical, Sanofi/Aventis, and Vifor Pharma outside the submitted work. S. Kopetz reports other support from Arvinas, Debiopharm, Larkspur, Janssen, Kivu, Genentech, Merck, Boehringer Ingelheim, Bayer Health, Pfizer, Mirati Therapeutics, Flame Biosciences, Carina Biotech, Frontier Medicines, Replimune, Bristol Myers Squibb–Medarex, Amgen, Tempus, Harbinger Health, Zentalis, AVEO, Tachyon Therapeutics, Agenus, Revolution Medicines, Kestrel Therapeutics, Roche, Arcus Biosciences, AstraZeneca Pharmaceuticals, BeOne Medicines, Clasp Therapeutics, Cytovation, Dewpoint Therapeutics, Marengo Therapeutics, SageMedic, Servier, Sibylla, T-Cypher Bio, Xaira, AmMax Bio, Ikena, Guardant Health, Genentech/Roche, EMD Serono, Amgen, Eli Lilly, Daiichi Sankyo, Pfizer, Boehringer Ingelheim, BridgeBio, Zentalis, Biomed Valley, Johnson & Johnson, Bristol Myers Squibb–Medarex, Cardiff, Jazz Pharmaceuticals, and Frontier Medicines outside the submitted work. J. Mizrahi reports personal fees from AstraZeneca, Exelixis, Bristol Myers Squibb, Intera Oncology, and Amgen outside the submitted work. A. Italiano reports grants from Bayer, Roche, Bristol Myers Squibb, AstraZeneca, and Amgen outside the submitted work. D. Tougeron reports personal fees from Bristol Myers Squibb, MSD, AstraZeneca, Gilead, Incyte, Roche, Amgen, Servier, Pierre Fabre, Merck Serono, Takeda, and Daiichi Sankyo outside the submitted work. M. Boakye reports employment with Arcus Biosciences. V. Khemka reports other support from Arcus Biosciences during the conduct of the study, as well as other support from Arcus Biosciences and Merck & Co. Inc. outside the submitted work. Z.A. Wainberg reports personal fees from Alligator Bioscience, Amgen, Astellas, AstraZeneca, Bayer, Daiichi, Ipsen, Lilly, Merck, EMD Serono, Novartis, and Pfizer and grants and personal fees from Bristol Myers Squibb and Arcus outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

M. Cecchini: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S.-W. Han: Resources, data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. S. Lee: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, writing–original draft, project administration, writing–review and editing. K.-W. Lee: Resources, investigation, writing–original draft, writing–review and editing. S. Kopetz: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, project administration, writing–review and editing. J. Mizrahi: Conceptualization, resources, data curation, supervision, validation, investigation, visualization, writing–original draft, project administration, writing–review and editing. Y.S. Hong: Resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. F. Ghiringhelli: Validation, investigation, visualization, writing–original draft, writing–review and editing. A. Italiano: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, writing–original draft, writing–review and editing. D. Tougeron: Conceptualization, resources, data curation, supervision, validation, investigation, visualization, writing–original draft, project administration, writing–review and editing. B. Beagle: Data curation, formal analysis, writing–original draft, writing–review and editing. M. Boakye: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, writing–review and editing. T. Zhao: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. V. Khemka: Data curation, formal analysis, writing–original draft, writing–review and editing. Z.A. Wainberg: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.