A Phase II Study Investigating Cabozantinib in Patients with Refractory Metastatic Colorectal Cancer (AGICC 17CRC01)

Colorectal cancer is the fourth most commonly diagnosed and second most deadly cancer in the United States (1). While several treatment options for metastatic colorectal cancer (mCRC) have led to improved survival and response rates over the past several decades, prognosis remains poor with an estimated median overall survival (OS) of 30 months and a 5-year OS rate of 14% (1). Standard-of-care agents in microsatellite stable (MSS) mCRC include first-line chemotherapy using sequential or in combination fluoropyrimidines, oxaliplatin, and irinotecan, and, depending on RAS and BRAF mutation status, mAbs targeting EGFR (cetuximab and panitumumab) and antiangiogenic therapy targeting the VEGF axis (bevacizumab: VEGFR; ramucirumab: VEGF-trap; ziv-aflibercept; ref. 2). On the basis of the phase III BEACON trial, encorafenib, an oral small-molecule BRAF inhibitor, plus cetuximab has been FDA approved for treatment of subsequent line, BRAF-mutant mCRC (3). mAbs targeting immune checkpoint proteins (i.e., PD-1, nivolumab, pembrolizumab; and CTLA-4, ipilimumab) have also been FDA approved for use as first-line agent for microsatellite instable (MSI) tumor types (4, 5). Clinically relevant biomarkers for treatment of mCRC also include HER2/Neu overexpression and NTRK fusions (6). Recently, studies have also shown potential for targeted inhibitors eliciting clinical responses in KRAS G12C-mutant colorectal cancer (7, 8). Regorafenib received FDA approval for treatment of metastatic, refractory colorectal cancer in 2015 based on the phase III “Regorafenib monotherapy for previously treated metastatic colorectal cancer (CORRECT) trial”, although regorafenib offers modest survival benefit of approximately 1.4 months over best supportive care and treatment resistance is inevitable (9, 10). TAS-102, an orally administered combination of trifluridine plus tipiracil hydrochloride, has also been FDA approved for marginal survival benefit in patients with refractory, mCRC. As more biomarker-driven studies have shown improved outcomes for a variety of molecular targets, the landscape for treating mCRC has become more complex as patients progress through therapeutic regimens. Although therapeutic options have led to improved survival, many patients ultimately progress on all standard-of-care agents regardless of tumor biology and targeted therapy, highlighting the need for ongoing research and novel therapies for mCRC.

Therapeutic resistance, both inherent and acquired, has been well described as an important reason for poor outcomes in mCRC. Angiogenesis has been well described as a hallmark for colorectal cancer tumorigenesis. Specifically, VEGF ligand binding to VEGFR2 has been shown as a driver of tumor neovascularization (11). While antiangiogenics have demonstrated incremental improvement in survival benefit, resistance to angiogenic inhibition uniformly develops over months in patients with mCRC (12, 13). One of the mechanisms by which acquired resistance occurs is through upregulation of the proto-oncogene mesenchymal–epithelial transition (cMET) receptor tyrosine kinase (RTK; ref. 14). Upon binding to its ligand hepatocyte growth factor (HGF), the cMET signaling pathway controls tumor cell survival and metastasis in part via activation of the PIK3CA/AKT/mTOR pathway as well as contributing to antiangiogenic therapeutic resistance via upregulation of MET expression (15–17).

Previous reports have observed cMET overexpression in 30%–75% and MET amplification in 3%–4% of patients with mCRC analyzed by IHC and PCR, respectively (18). Targeting the cMET/HGF axis may restore efficacy of the antitumor effects of antiangiogenesis, which has been previously reported by our group investigating cabozantinib (XL184), a multi-tyrosine kinase inhibitor (TKI) targeting VEGF and MET, in patient-derived xenograft (PDX) mouse models (19). However, the clinical utilization of cMET overexpression/MET amplification as a predictive biomarker in colorectal cancer is unproven and limitations such as a lack of consensus regarding method for analysis of expression/amplification and definition of positive values remain challenges for clinical interpretation.

Cabozantinib is an orally administered, RTK inhibitor of multiple kinases effecting cancer cell growth, angiogenesis, and metabolism including MET, VEGFR2/KDR, AXL, TIE2, RET, and KIT. Cabozantinib has demonstrated clinical effectiveness in a variety of tumor types with NCCN guideline recommendations for use in metastatic renal cell carcinoma, metastatic/unresectable hepatocellular carcinoma, metastatic medullary thyroid cancer and imatinib-resistant gastrointestinal stromal tumors (20–23). Similarities among mechanism of action between other RTK inhibitors and cabozantinib exist including regorafenib, which also targets VEGFR2; however, cabozantinib provides the unique inhibitory profile of dual inhibition of VEGFR2 and MET.

Preclinically, cabozantinib has shown impressive antitumor activity in PDX mouse models with tumor growth inhibition in 80% of treated colorectal explants (24). Moreover, cabozantinib also demonstrated superior tumor growth inhibition, reduction in tumor vascularity, and increase in autophagy compared to regorafenib in colorectal PDX mouse models (19). On the basis of these data, we designed a phase II Simon two-stage, nonrandomized trial aimed to investigate the efficacy and safety of cabozantinib in metastatic, refractory colorectal cancer (NCT03542877).

This prospective study was performed through eight academic centers (members of the “Academic GI Cancer Consortium” or AGICC) in the United States and enrolled patients with histologically proven refractory mCRC between May 2018 and May 2019. The data cut-off date was August 4, 2021. Patients must have progressed on or were intolerant to fluoropyrimidine, irinotecan, oxaliplatin, and bevacizumab, and prior EGF inhibitor therapy was required for patients with left-sided, RAS wild-type (WT) tumors. Prior regorafenib or TAS-102 treatment was allowed but not required. Patients must have had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1 and had measurable disease per RECIST, version 1.1 (v1.1) as determined by the investigator. Patients with evidence of cavitating lung lesions, tumors invading or encasing major blood vessels, or demonstrating clinically significant gastrointestinal bleeding within 6 months before treatment start were excluded from the trial. Patients requiring concomitant anticoagulation at therapeutic doses with oral anticoagulants or platelet inhibitors were also excluded from the trial. Cabozantinib was administered at a starting dose of 60 mg orally daily based on previously established safety and efficacy data in other tumor types (25, 26). Two dose-level reductions were allowed per protocol (40 and 20 mg). Response-evaluable patients were those patients who had radiographic and/or clinical assessments to evaluate for disease progression or response per RECIST v1.1.

Investigators obtained written informed consent from the patients, procedures related to this trial were conducted in accordance with recognized ethical guidelines, and this study was approved by an Institutional Review Board.

Whole blood for circulating tumor DNA (ctDNA) analysis was collected at baseline and on treatment ± 5 days of cycle 3 day 1 from consenting patients to test the correlation of ctDNA response to clinical response. Whole blood draws (∼8 mL) were collected and processed to plasma by centrifugation at 1,600 × g for 10 minutes followed by a second spin at 16,000 × g for 10 minutes. Plasma was collected and frozen at −80°C. Two milliliters of plasma was used to isolate cell-free DNA (cfDNA) with the Qiagen circulating nucleic acid isolation kit (#55114) according to manufacturer's protocols. Following quantitation of the cfDNA using a Qubit fluorimeter, 10 ng of cfDNA was used to generate sequencing libraries with the Ion Torrent, Ion AmpliSeq Library kit and Ion AmpliSeq Cancer Hotspot V2 primer panel according to the manufacturer's protocols. Libraries were sequenced on either the Ion Torrent Proton P1 chip or the Ion Torrent Personal Genome machine 318 chip with a minimum average coverage of >1,200×. Variants were called using Ion Reporter software (v5.12) and filtered for known pathogenic and/or likely pathogenic mutations.

The primary endpoint for this trial was the binary endpoint of being progression free at 12 weeks. On the basis of the phase III CORRECT trial investigating regorafenib versus best supportive care, the median progression-free survival (PFS) for patients with refractory colorectal cancer was 1.7 months in the placebo arm. Therefore, the PFS at 12 weeks would be approximately 13%, which served as the historical control for the null hypothesis. This study was designed to have at least 90% power to detect an improvement in the 12-week PFS rate of at least 20%; therefore, the alternative hypothesis is that the 12-week PFS rate is 33%. The optimal design providing at least 90% power to detect the alternative while controlling the type I error rate at 0.05 uses at most 44 patients, thus 44 patients were enrolled to this trial. Using an optimal Simon two-stage design, the first 16 patients were enrolled and 12-week PFS was determined. An additional 28 patients were enrolled after meeting the predetermined criteria for trial continuation, resulting in a total of 44 patients enrolled. With this design, the therapy would be considered efficacious if 10 or more patients were free of progression at 12 weeks, suggesting that the treatment has sufficient efficacy to warrant future study.

Disease assessment was performed by CT or MRI methods at baseline and every 6 weeks for the first 12 weeks, then every 9 weeks thereafter during treatment until disease progression. Response evaluation was performed on the basis of RECIST v1.1. PFS was defined as the time from administration of the initial dose of cabozantinib to evidence of radiographic progression as defined by RECIST or death from any cause without evidence of disease progression, whichever occurred first. Kaplan–Meier estimates of PFS and OS rates were calculated along with their corresponding 95% confidence intervals. Objective response rate (ORR) was defined using the RECIST v 1.1 as the proportion of patients with a confirmed partial response (PR) or complete response. OS was defined as the time from administration of the initial dose of cabozantinib until death from any cause. Surviving patients who were taken off trial prior to clinical or radiographic disease progression were censored from primary endpoint analysis.

Historic results of tissue-based molecular profiling were retrieved on all enrolled patients. Retrospective analysis of PFS and response rate (RR) in patients based on RAS, BRAF, and PIK3CA mutation status was performed using point estimates with 95% confidence intervals and comparisons were made using a stratified log-rank test. Carcinoembryonic antigen serum levels were performed at baseline and every 3 weeks during treatment.

Safety and tolerability analysis of cabozantinib was summarized by dose and severity as assessed by the Common Toxicity Criteria for Adverse Events version 4.0 and relationship to study drug.

ctDNA correlation to clinical response was performed in patients who had pretreatment and posttreatment blood samples obtained, and in whom the mutation prevalence changed by ≥ 20% following therapy. Patients were grouped into those whose disease progressed or those whose disease remained stable or responded to therapy. A one-tail Fisher exact test was performed to determine whether there was a statistically significant association between the change in mutational prevalence and clinical response.

The human sequence data generated in this study are not publicly available due to patient privacy requirements but are available upon reasonable request from the corresponding author. Other data generated in this study are available within the article and its Supplementary Data files.

A total of 44 patients with mCRC were included in this study (male, 57%; median age, 60 years; Table 1). Of the 44 patients enrolled, 40 patients were response evaluable based on prespecified definition for primary endpoint analysis. The study met the predetermined criteria for expansion and total planned patient enrollment based on the Simon two-stage design. The median number of previous lines of therapy was 4. The median time from metastatic diagnosis to informed consent on this trial was 30.7 months. Eight (18%) patients had previous regorafenib treatment. KRAS mutations were present in 54% of patients. No additional clinically relevant molecular aberrations were reported including NRAS mutations, BRAF mutations, HER2 overexpression, NTRK fusions, or MSI-high (MSI-H) tumor types. Other patient characteristics and treatment course are summarized in Table 1 and Figure 1.

Among the 40 response-evaluable patients, 18 (45%) patients were free of progression at 12 weeks. Three patients discontinued treatment prior to disease assessment. The median PFS was 3.0 months, and the median OS was 8.3 months (Fig. 2). A trend toward improved PFS in patients with RAS WT versus mutant (MT) mCRC was observed (4.9 vs. 2.7 months, respectively; P = 0.16; Fig. 3). A trend toward improved OS in patients with RAS WT versus MT tumors was also observed (10.4 vs. 7.0 months, respectively; P = 0.13; Fig. 3). No clinically or statistically significant differences in PFS or OS were observed in the PIK3CA MT versus WT subgroup analysis (Supplementary Fig. S1). One patient demonstrated a PR and 27 other patients had stable disease (SD) as their best response, giving an ORR of 2.5% and a disease control rate of 70% (Supplementary Fig. S2). One patient maintained a PR through trial closure and was continued on cabozantinib through compassionate use at last follow-up (Fig. 4). Follow-up data are available for 34 subjects, 7 of which were alive at last follow-up. Of these 7 patients, 5 had PFS at 12 weeks.

All 44 patients on treatment were assessed for safety. A total of 769 AEs were reported and all patients reported at least one AE. The most frequent cabozantinib-related AEs reported in ≥ 10% of patients included fatigue, diarrhea, hypertension, nausea, decreased appetite, weight loss, proteinuria, constipation, palmar-plantar erythrodysaesthesia syndrome, abdominal pain, vomiting, thrombocytopenia, alanine aminotransferase increase, and hypothyroidism (Table 2). A total of 93 AEs were recorded as ≥ grade 3 which occurred in 34 subjects (77%). Serious AEs (SAE) were reported in 20 (45%) patients with a total of 37 documented SAEs and fourteen (38%) of SAEs were recorded as at least possibly related to cabozantinib. Of the five grade 5 AEs that were observed, four were documented as not related to treatment (three disease progression, one disseminated intravascular coagulation) and one was documented as possibly related to treatment attributed to bowel perforation requiring ICU level care in the setting of radiographic evidence of disease progression.

Dose reduction of cabozantinib occurred in 24 (55%) patients and 5 (11%) patients were reduced to the lowest dose level of 20 mg/day. Two patients were discontinued on treatment due to toxicity.

The presence of pathologic mutations in study participants was determined using cfDNA extracted from whole blood. The association between mutation prevalence and response to therapy was analyzed in the patients for which a pretreatment and posttreatment sample was obtained, and in whom there was a greater than 20% change in mutation prevalence (n = 14). Of the 7 patients with SD or clinical response to therapy, 5 showed a decrease in the prevalence of known pathologic mutations, while 2 had an increase in mutation prevalence. In patients whose disease clinically progressed during the study, 5 showed increase pathologic mutation prevalence, while 2 had a decrease in pathologic mutation prevalence. While these results show association between clinical disease and mutational load, the association was not statistically significant (Fisher exact test P = 0.143).

We demonstrated that cabozantinib has clinical activity in patients with refractory, heavily pretreated mCRC. Of the 40 evaluable patients, 45% of patients (n = 18) met the 12-week PFS primary endpoint in a treatment refractory patient population with a median of 4 previous lines of treatment. The survival endpoints of metastatic PFS (mPFS) and median OS (mOS) were 3.0 months and 8.3 months, respectively. One patient with PR maintained response through study closure and despite dose reduction to 40 mg/day of cabozantinib. ctDNA analysis also showed a trend toward concordance between mutational load and clinical efficacy. The safety profile of cabozantinib is consistent with previous studies in other solid tumors with the most frequent AEs were fatigue (64%), diarrhea (52%), and hypertension (41%; refs. 20–22).

The patient demographics, tolerability of treatment, and efficacy of cabozantinib are comparable with the phase III CORRECT and RECOURSE trials that led to FDA approval of regorafenib and trifluridine/tipiracil (TAS-102), respectively, for treatment of refractory mCRC. In the CORRECT trial, the regorafenib group had a mPFS of 1.9 months and a mOS of 6.4 months (9). Similarly, the RECOURSE trial investigating TAS-102 showed a mPFS of 2.0 months and mOS of 7.1 months in the investigational arm (11). ORRs reported in these trials were 1% and 1.6%, respectively. Oral VEGFR TKI agents apatinib and fruquintinib have shown mPFS of approximately 4 months in Asian patients with refractory, mCRC (27, 28).

A trend toward improved survival of patients with RAS WT versus RAS MT tumors was also apparent in this study. A 2.2-month difference in mPFS and a 3.4-month difference in mOS was observed in patients with RAS WT versus RAS MT tumors (P = 0.16 and P = 0.13, respectively). Of note, correlation between RAS mutational status and treatment response has not been reported with apatinib or fruquitinib. Importantly, this study did not intend to formally evaluate RAS mutational status as a primary endpoint, and the differences between groups could be confounded by prognosis, tumor sidedness, treatment history, and other factors. Despite preclinical data suggesting a possible improved response to cabozantinib in PIK3CA MT colorectal cancer tumors, there was no observed survival difference in subgroup analysis using PIK3CA mutational status. No additional clinically relevant mutations including HER2 overexpression, BRAF V600E mutation, NTRK fusions, and MSI-H tumor types were reported.

In this trial, we showed an association between reduction of ctDNA and patients who met the 12-week PFS mark. Although a limitation of our ctDNA analysis include small sample size, we believe our results are hypothesis generating as they are consistent with previous reports showing that ctDNA may offer early response analysis of treatment (29). In the metastatic setting, ctDNA may offer predictive and prognostic information although data from prospective trials are limited (30–32). Blood-based analyses to better characterize and quantify biomarkers in patients with heavily pretreated mCRC continues to evolve with the hope of offsetting the challenges of tumor heterogeneity, comorbidity, and costs inherent to tissue biopsies (33). Our understanding of how to best utilize ctDNA has been challenging due to mixed clinical results. For example, in a phase Ib study of 24 patients, evaluation of MET amplification in cfDNA as a predictive biomarker in patients with MET-amplified mCRC undergoing cabozantinib plus panitumuab showed response in only 1 of 5 patients (34, 35). Despite data showing the importance of aberrant MET expression in acquired therapeutic resistance in mCRC, the predictive value of MET expression by IHC has not been proven (36). While there are several blood-based assays commercially available to molecularly characterize tumors and monitor response via ctDNA, data on how to utilize these technologies as a predictive tool are limited (37). Large, prospective trials are needed to help establish the role for ctDNA testing as a predictive tool in the management of patients with mCRC.

Cabozantinib has also shown clinical activity in other trials investigating its use in mCRC. The EGFR antibodies panitumumab and cetuximab have been approved on the basis of incremental survival benefit in treatment of RAS WT mCRC (38–41). Analogous to antiangiogenic therapy in mCRC, acquired resistance to anti-EGFR therapies through aberrant MET overexpression has been reported previously (42–44). In a recently published phase Ib clinical trial of 25 patients with RAS WT, mCRC, the combination of cabozantinib plus panitumumab, a fully humanized mAb targeting EGFR, demonstrated an ORR of 16%, mPFS of 15.8 weeks, and mOS of 51.8 weeks (34). There were no reported grade 5 AEs, 5 patients discontinued treatment due to toxicity, and 72% of patients required a dose reduction of cabozantinib. A high degree of grade ≥ 3 AEs were observed, although these data were similar to safety outcomes noted in previously reported trials in similar patient populations. Tolerability did not appear to correlate with clinical response.

Immunomodulatory effects of cabozantinib via downregulation of myeloid-derived suppressor cells and regulatory T cells have also been reported previously (45, 46). The Tyro3, AXL, and MerTK (TAM) receptor family regulate tumor-associated macrophages that elicit an immunosuppressive tumor microenvironment, and blockade of TAM receptors via cabozantinib may invigorate antitumor immune activation (47, 48). Because of favorable immunomodulatory properties, cabozantinib in combination with immune checkpoint inhibitors has shown promise in a variety of solid tumor types and clinical settings including metastatic renal cell, hepatocellular carcinoma, and mCRC (49–52). In the phase Ib CAMILLA basket trial, the combination of cabozantinib plus durvalumab, an anti-programmed death-ligand1 (PD-L1) mAb, led to two PRs out of 13 treated patients with mCRC and an ORR of 21% (53). A phase I dose-expansion trial investigating this combination in mCRC is ongoing (NCT03539822). On the basis of preclinical data showing synergy between cabozantinib and checkpoint inhibitors in MSS colorectal cancer PDXs, a phase II trial investigating cabozantinib plus nivolumab, an anti-programmed death-1 (PD-1) mAb, in treatment of refractory mCRC is also ongoing (NCT03170960; ref. 54).

Clinical data using other multi-TKIs plus checkpoint inhibition has shown mixed results. The Japanese phase Ib REGONIVO trial investigating regorafenib plus nivolumab demonstrated an impressive ORR of 29% in patients with refractory, mismatch repair proficient mCRC (55). However, a follow-up North American phase II trial using regorafenib plus nivolumab combination showed an ORR of 7.1% with a mPFS of 2.0 months in patients with refractory mCRC. Subgroup analysis demonstrated higher response rates and survival in patients without liver metastases compared with patients with liver metastases (ORR of 21.7% vs. 0%, respectively). Whether cabozantinib plus immunotherapy will demonstrate improved outcomes in mCRC is eagerly awaited, and further insight into subgroup analysis through biomarker-driven assessment is needed.

There were several limitations to this trial including the open-label, nonrandomized study design. Without a control group, the interpretation of the observed PFS and survival endpoints become less clear as the outcomes may be influenced by a highly selected patient population with a more favorable prognosis. Also, the benefit of a surrogate endpoint of PFS rate at 12 weeks may not translate to an improved OS benefit, which is the gold standard for assessing efficacy of cancer therapeutics (56). In addition, we used investigator assessment of RECIST v1.1 for radiographic assessments, making investigator bias difficult to control. Finally, information regarding molecular profiling from tumor tissue was collected via historical records performed prior to study enrollment using various platforms based on investigator discretion.

We believe the results of our trial support further investigation of cabozantinib as a single agent and/or in combination in the treatment of mCRC. In addition, this study may provide a useful baseline in evaluating efficacy in combinational trials. On the basis of the number of dose reductions required and durable response observed at 40 mg daily dosing, future designs may consider investigating cabozantinib 40 mg orally daily dosage. Studies using cabozantinib plus checkpoint inhibitors in mCRC are currently underway and results are highly anticipated (NCT04963283; NCT03539822; NCT03170960). On the basis of subgroup analysis, patients who harbor a RAS WT tumor may derive greater benefit from cabozantinib treatment; however, this is strictly a hypothesis generating observation. Studies using ctDNA as a predictive biomarker may also improve upon assessment of treatment response. Future research should aim to identify and validate predictive biomarkers, refine ctDNA as a predictive tool for disease response, and will hopefully refine the patient population that may derive the most benefit from cabozantinib.

E. Dotan reports grants from Iilly, Incyte, Relay, AstraZeneca, Zymeworks, Ipsen, Medimmune, NGM Biopharmaceuticals, Pfizer; personal fees from Incyte, Helsinn, Taiho, G1 therapeutics, Pfizer, Basilea, QED, and Boston Biomedical outside the submitted work. H.S. Hochster reports personal fees from Merck and TRIGR; personal fees and non-financial support from Processa outside the submitted work. S. Subramaniam reports personal fees from Exelixis outside the submitted work. P.J. Blatchford reports other from Exelixis during the conduct of the study. W.A. Messersmith reports other from Exelixis during the conduct of the study. No disclosures were reported by the other authors.

This trial was sponsored by Exelixis.

A.J. Scott: Conceptualization, formal analysis, supervision, investigation, methodology, writing-original draft. A.B. Mallick: Investigation, writing-review and editing. E. Dotan: Investigation, writing-review and editing. S.J. Cohen: Conceptualization, investigation, writing-review and editing. P.J. Gold: Conceptualization, investigation, writing-review and editing. H.S. Hochster: Conceptualization, investigation, writing-review and editing. S. Subramaniam: Investigation, writing-review and editing. A. Barzi: Conceptualization, investigation, writing-review and editing. G.S. Watts: Data curation, formal analysis, methodology, writing-review and editing. P.J. Blatchford: Conceptualization, data curation, software, formal analysis, validation, writing-review and editing. W.A. Messersmith: Conceptualization, resources, data curation, formal analysis, supervision, investigation, methodology, writing-review and editing.

We thank the patients and their families for their invaluable participation in this trial as well as the dedicated investigators and research staff for their hard work in successfully completing this trial.