Combining antivascular endothelial growth factor and anti-epidermal growth factor receptor antibodies: randomized phase II study of irinotecan and cetuximab with/without ramucirumab in second-line colorectal cancer (ECOG-ACRIN E7208)

Abstract

Background

Early studies showed promise of combined anti-epidermal growth factor receptor (EGFR) plus anti–vascular endothelial growth factor (VEGF) antibodies for advanced colorectal cancer (CRC), yet this was later rejected as toxic and ineffective in studies not selected for RAS status. We studied advanced KRAS wild-type CRC, as second-line treatment, using irinotecan-cetuximab with or without the anti-VEGF receptor antibody ramucirumab.

Methods

Patients with 1 prior regimen including fluoropyrimidine, oxaliplatin, and bevacizumab, with KRAS wild-type tumors were stratified by Eastern Cooperative Oncology Group Performance Score, time since last chemotherapy, and progression on oxaliplatin to irinotecan-cetuximab (IC) (180 mg/m² and 500 mg/m² every 2 weeks) vs modified ICR (irinotecan-cetuximab with ramucirumab 150 mg/m² and 400 mg/m² plus 6 mg/kg, respectively). A total of 102 patients were compared for progression-free survival (PFS) as primary endpoint (85% power for 70% improvement in median PFS from 4.5 to 7.65 months).

Results

Of the 102 enrolled, 44 treated with irinotecan-cetuximab and 45 with modified ramucirumab were evaluable. Median PFS was 6.0 months vs 9.2 months, respectively (hazard ratio = 0.75, P = .07; statistically significant by study design for P < .128). Response rate was 23% vs 36% (P = .27), and disease-control rate was 52% vs 73% (P = .05). Grade 3 or higher toxicity was equivalent. Overall survival was not significantly different at approximately 19 months.

Conclusion

Previous phase 3 trials without RAS genotyping rejected combining anti-epidermal growth factor receptor and anti-VEGF drugs. In this randomized multicenter phase 2 study for KRAS wild-type CRC (all previously bevacizumab treated), the addition of ramucirumab to irinotecan and cetuximab improved PFS and disease control rate, showing the combination is feasible and effective. Further, phase 3 trials with appropriate patient-selection are required. (NCT01079780)

Colorectal cancer (CRC) remains the fourth most common cancer in the United States today for men and women, with approximately 150 000 new cases per year, and is the third most common cause of cancer deaths in men and women (1). Considerable progress has been made in the treatment of metastatic colon cancer since the 1990s with the introduction of new drugs and new targeted agents. Median survival was improved from approximately 13 months in the fluoropyrimidine-alone era to approximately 30 months with appropriate use of combination chemotherapy, targeted agents in C80405 (2).

Despite broad approval of anti-epidermal growth factor receptor antibodies for CRC in 2004 (3-5), retrospective molecular analysis of controlled trials showed that mutations in the RAS gene abrogate the benefit from these drugs, with change to the labels in 2009 (6). Altogether, approximately 45% of CRCs have RAS mutation and do not benefit from the use of epidermal growth factor receptor antibodies. Unfortunately, many key studies using these antibodies were done prior to the knowledge that RAS status must be determined with appropriate patient selection.

Although the benefit of combining anti-epidermal growth factor receptor and anti–vascular endothelial growth factor (VEGF) antibodies in early studies (6) was reported, it was subsequently rejected with the PACCE (Paniumumab Advanced Colorectal Cancer Evaluation) trial (7), which was stopped early for toxicity and reduced benefit. This trial combined FOLFOX or FOLFIRI and bevacizumab with or without panitumumab in the first-line setting. In this study, the double-antibody arm had inferior outcomes to the standard arm, which was attributed to substantial toxicity of the combination, and considered to be nonfeasible. However, subsequent analyses, particularly in the PRIME (Panitumumab Randomized trial In combination with chemotherapy for Metastatic colorectal cancer to determine Efficacy) trial of FOLFOX plus panitumumab vs FOLFOX in a nonselected population, demonstrated that the RAS-mutated patients had worse outcomes (8). The CAIRO2 (Combination Chemotherapy with irinotecan, capecitabine and bevacizumab) phase 3 Dutch trial of first-line CapeOX and bevacizumab with or without cetuximab also showed a negative effect when adding cetuximab in KRAS-mutated patients. This study overall did not show a benefit for adding cetuximab to CapeOX and bevacizumab but included RAS-mutated patients, with detrimental effects (9).

In this study, we were able to obtain the anti-VEGF receptor drug ramucirumab as the anti-VEGF partner drug. Ramucirumab is a receptor-directed antibody that binds to VEGF receptor 1 and 2 rather than the VEGF-alpha ligand, as with bevacizumab. Ramucirumab had single-agent activity in CRC and was approved for second-line therapy of CRC in combination with FOLFIRI for patients in the RAISE (Ramucirumab and 5FU-irinotecan Second line colon cancer Evaluation) study (10) without prior bevacizumab. In this study, all patients were required to have prior treatment with bevacizumab, in addition to chemotherapy.

Eastern Cooperative Oncology Group (ECOG)–American College of Radiology Imaging Network (ACRIN) E7208 (NCT01079780) was designed as the first molecularly guided CRC second-line trial. It was a randomized comparison of the addition of an anti-VEGF receptor antibody to the standard regimen of irinotecan and cetuximab for patients with KRAS wild-type CRC, all of whom had prior anti-VEGF therapy (11). All patients were previously treated with 1 regimen of fluoropyrimidine, oxaliplatin, and bevacizumab. The primary endpoint of this study was progression-free survival (PFS) with secondary endpoints of response rate and overall survival.

Methods

The study was designed and approved by the relevant ECOG-ACRIN and National Cancer Institute (NCI) Cancer Therapy Evaluation Program committees and the US Food and Drug Administration. The study was approved by the NCI Central Institutional Review Board and institutional review boards as required for all sites. Drug supply for ramucirumab was provided by the manufacturer Eli Lilly through the NCI Cooperative Research and Development Agreements program. Drug was supplied for ramucirumab arms from NCI Pharmaceutical Management Branch, and commercial sources were used for irinotecan and cetuximab.

Eligible patients were those with measurable, pathologically diagnosed KRAS wild-type CRC, who had previously received only 1 prior regimen including fluoropyrimidine, oxaliplatin, and bevacizumab for metastatic disease. Other eligibility criteria included ECOG Performance Status (PS = 0-1); recent documentation of disease progression (within 6 weeks of enrollment); no clinically significant bleeding episode within 3 months; no uncontrolled hypertension (on antihypertensives); no arterial thrombotic event within 6 months; no major surgery within 28 days; no major abdominal surgery, perforation, or resection within 6 months; no inflammatory bowel disease requiring pharmacologic intervention within 12 months; and no major cardiovascular event within 6 months.

The primary study endpoint was PFS for the experimental arm compared with the standard arm (irinotecan-cetuximab). Other objectives included the response rate for the 2 arms, their grade 3-5 toxicity rates, and overall survival.

The study was approved and began accrual in October 2010. Patients were stratified by ECOG PS (0 vs 1), whether they had a progression while on oxaliplatin (yes vs no), and time since last chemotherapy prior to progression (<6 vs ≥6 months). Eligible patients were randomly assigned 1:1 to a standard second-line regimen for patients with KRAS wild-type tumors: irinotecan 180 mg/m² and cetuximab 500 mg/m² intravenously every 2 weeks (irinotecan-cetuximab) or the same with ramucirumab 8 mg/kg intravenously every 2 weeks (ICR, irinotecan-cetuximab and ramucirumb). After the first 34 patients were enrolled and an interim real-time toxicity analysis was performed, the study was suspended in June 2012 for excess toxicity in the experimental ICR arm. The study was reopened in June 2014 after dose modification to a modified ICR regimen with irinotecan 150 mg/m², cetuximab 400 mg/m², and ramucirumab 6 mg/kg every 2 weeks (mICR). These doses were based on the actual tolerated doses by the 16 patients treated initially with full dose ICR.

Statistical considerations

The revised statistical design for IC vs modified ICR required 48 new eligible patients per arm to achieve 85% power to detect a difference of 3 months in median PFS (4.5 months control arm vs 7.65 months in the experimental arm; 70% increase in median PFS) using a 1-sided 0.15 level stratified log-rank test. Real-time monitoring of adverse events occurred until 16 patients were followed through 2 cycles. Full information occurred at 67 PFS events, with 1 interim analysis at 40 PFS events. Type I error control for the interim analysis was accomplished using an O’Brien–Fleming type boundary, with Lan–Demets use function methodology.

PFS was defined as the time from random assignment to disease progression or death without a documented progression or the date of last disease assessment for censored patients. Kaplan–Meier method was used with log-rank tests to assess differences between groups. χ² tests were used to test the distributions of categorical variables, and the nonparametric Wilcoxon rank sum test was used where appropriate.

Results

This study was activated on October 8, 2010; suspended on June 14, 2012, for an interim toxicity analysis; reactivated on June 6, 2014, with the modified ICR regimen; and closed to accrual on July 7, 2017. A total of 136 patients were enrolled (34 before the study suspension and amendment). Subsequently, 102 patients were enrolled, but 3 did not start treatment, and 13 were found to be ineligible during central review. Of the 99 patients who were randomly assigned to the IC arm vs the modified ICR arm and started treatment, all were included in the toxicity analysis, whereas 89 were eligible for the primary efficacy evaluation (Figure 1).

Figure 1.

CONSORT diagram. AE = adverse events.

Table 1 shows the demographic characteristics of the patients enrolled. Those enrolled were equally balanced for gender, age, race, ethnicity, ECOG PS, and extent of disease. In this study, we wished to capture all patients who could receive second-line irinotecan-based therapy, and therefore they could enter the trial as second-line therapy in several scenarios—recurrence or progression while on observation (off all drugs), progression on maintenance (fluoropyrimidine alone, bevacizumab alone, or fluoropyrimidine and bevacizumab), or progression on fluoropyrimidine with oxaliplatin. We stratified patients on the basis of these considerations, as we reasoned that those progressing while receiving first-line CapeOX or FOLFOX and bevacizumab would have worse outcomes, and they did. Approximately 28% were on surveillance at time of progression or relapse, 56% were receiving fluoropyrimidine, 66% bevacizumab, and 29% progressed while on an oxaliplatin regimen.

Table 1.

Patient demographics and disease characteristics of eligible patients

Characteristic	Arm A (irinotecan-cetuximab)	Arm C (modified ICR)
	No. (%)	No. (%)
	(n = 44)	(n = 45)
Gender
Male	25 (56.8)	32 (71.1)
Female	19 (43.2)	13 (28.9)
Age, y
Mean (SD)	58.9 (10.7)	59.4 (10.5)
Median (min, max)	58.8 (34, 84)	60.4 (37, 80)
Race
Asian	2 (4.6)	2 (4.4)
Black	6 (13.6)	3 (6.7)
Multirace	1 (2.3)	1 (2.2)
Unknown	0 (0)	2 (4.4)
White	35 (79.5)	37 (82.2)
Ethnicity
Hispanic	3 (6.8)	5 (11.1)
Non-Hispanic	38 (86.4)	38 (84.4)
Unknown, missing	3 (6.8)	2 (4.4)
ECOG Performance Score
0	24 (54.5)	24 (53.3)
1	20 (45.5)	21 (46.7)
Primary tumor site
Rectal	17 (38.6)	15 (33.3)
Colon	27 (61.4)	30 (66.7)
Abdominal metastases
No	42 (95.5)	44 (97.8)
Yes	1 (2.3)	1 (2.2)
Unknown	1 (2.3)	0 (0)
Bone metastases
No	42 (95.5)	42 (93.3)
Yes	2 (4.6)	3 (6.7)
Unknown	0 (0)	0 (0)
Liver metastases
No	12 (27.3)	12 (26.7)
Yes	32 (72.7)	33 (73.3)
Lung metastases
No	27 (61.4)	22 (48.9)
Yes	17 (38.6)	23 (51.1)
Ovary metastases
No	44 (100)	44 (97.8)
Yes	0 (0)	1 (2.2)
Pelvic metastases
No	42 (95.5)	42 (93.3)
Yes	2 (4.55)	3 (6.7)
Peritoneal metastases
No	38 (86.4)	40 (88.9)
Yes	6 (13.6)	5 (11.1)
Pleural metastases
No	43 (97.7)	45 (100)
Yes	1 (2.3)	0 (0)
Primary site metastases
No	41 (93.2)	40 (88.9)
Yes	3 (6.8)	5 (11.1)
Other metastasis
No	37 (84.1)	40 (89.1)
Yes	7 (15.9)	5 (11.1)
Lymph nodes metastases
No	25 (56.8)	22 (48.9)
Yes	19 (43.2)	23 (51.1)
Determination of KRAS status
Primary tumor	30 (68.2)	33 (75.0)
Metastatic tumor	14 (31.8)	11 (25.0)
Missing	0 (0)	1
Oxaliplatin at Progressiona
No	32 (72.7)	32 (71.1)
Yes	12 (27.3)	13 (28.9)
Fluoropyrimidine at Progressiona
No	20 (45.5)	19 (42.2)
Yes	24 (54.5)	26 (57.8)
Bevacizumab at Progressiona
No	12 (27.3)	18 (40.0)
Yes	32 (72.7)	27 (60.0)
No treatment at Progression
No	33 (75)	31 (68.9)
Yes	11 (25)	14 (31.1)
Receiving warfarin
No	42 (95.5)	44 (97.8)
Yes	2 (4.6)	1 (2.2)

a

Not mutually exclusive.

Treatment duration and toxicity

Patients tolerated treatment well on the modified ICR regimen (Table 2). In this arm, patients received a median of 12 cycles (range = 1-56 cycles, mean = 15 cycles) compared with 8 in the IC arm (range = 1-44 cycles, mean = 11 cycles) (Wilcoxon rank sum P = .09). For modified ICR, overall toxicity rate was 48% grade 3 and 8% grade 4 compared with 49% and 10% in the IC arm, respectively. The toxicities experienced in both arms were quite similar except modified ICR patients experienced 4% grade 3 acneiform rash compared with 10% grade 3 and 2% grade 4 in the IC arm (possibly due to lower dose of cetuximab). Grade 3 hypertension was 6% with ramucirumab, and there was no grade 3 hypertension observed in the ramucirumab arm.

Table 2.

Treatment-related toxicities (%) (grade 3-5 incidence ≥ 10%)

Toxicity type	Treatment arm
	Irinotecan-cetuximab (n = 51)			Modified ramucirumab (n = 48)
	Grade			Grade
	3	4	5	3	4	5
	%	%	%	%	%	%
Anemia	4	—	—	6	—	—
Fatigue	2	—	—	10	—	—
Rash acneiform	10	2	—	4	—	—
Diarrhea	16	—	—	13	2	—
Mucositis oral	—	—	—	6	—	—
Allergic reaction	8	2	—	2	—	—
Neutrophil count decreased	10	2	—	8	2	—
Platelet count decreased	—	—	—	4	—	—
White blood cell decreased	4	—	—	10	—	—
Hypomagnesemia	4	2	—	—	—	—
Proteinuria	—	—	—	4	—	—
Hypertension	—	—	—	6	—	—
Worst degree	49	10	—	48	8	—

Efficacy

During an interim analysis occurring at approximately 90% information time, a statistically significant improvement in PFS was found on the modified ICR arm with a hazard ratio (HR) of 0.65 (1-sided P = .07). An updated PFS analysis (Figure 2) including 76 progression events and 13 censored observations resulted in a median PFS time of 5.98 months (95% confidence interval [CI] = 3.88 to 9.07 months) for patients treated with IC and 9.20 months (95% CI = 5.52 to 11.33 months) for those treated with modified ICR. There was no difference observed in overall survival between the arms, with median survival of approximately 19 months on both arms, presumably because of subsequent treatment with an anti-VEGF agent for control arm patients.

Figure 2.

Progression-free survival. HR = hazard ratio; IC = irinotecan-cetuximab; mICR = modified irinotecan-cetuximab-ramucirumab.

Response was evaluated locally by RECIST 1.0 criteria with computed tomography scans every 4 cycles (8 weeks) and is shown in Table 3. Response rate (all partial response, no complete responses were documented) was higher with modified ICR (36% vs 23%) but did not reach statistical significance (P = .27). Disease control rate (partial response + stable disease) was also higher for the experimental arm (73% vs 52%, P = .05), and fewer patients had disease progression as best response in the experimental arm as well (36% vs 18%).

Table 3.

Best response by arm among eligible patients (arm A and arm C)

Best response	Irinotecan-cetuximab	Modified ICR
	No. (%)	No. (%)
	(n = 44)	(n = 45)
Partial response	10 (23)	16 (36)
Stable disease	13 (30)	17 (38)
Progression	16 (36)	8 (18)
Unevaluable	5 (11)	4 (8)

Stratification

We also examined the stratification factors to determine whether these were indeed important predictive factors. For patients who were still on treatment with oxaliplatin at the time of progression qualifying for study entry (n = 21), the median PFS was only 4.30 months compared with the remainder of patients with PFS of 9.07 months (HR = 0.57, P = .03; Figure 3A). An even larger effect was seen for those progressing within 6 months of last treatment compared with those who progressed 6 months or longer after last treatment (n = 14). For the long interval group, PFS was 12.58 months compared with 5.98 months for the others (HR = 0.49, P = .02; Figure 3B). This effect was also seen with a median overall survival of 37.2 vs 16.8 months in the 6-month or longer group vs less than 6-month group, respectively (HR = 0.44, P = .01).

Figure 3.

Progression-free survival by stratification factors. HR = hazard ratio.

Discussion

At the time we began this study, and still today, the addition of anti-epidermal growth factor receptor drugs to anti-VEGF agents is considered too toxic and possibly detrimental. The main study leading to this conclusion was the PACCE study (7), seeking to determine whether panitumumab could add benefit to combination chemotherapy (FOLFOX or FOLFIRI) with bevacizumab in the first-line metastatic setting. This international study of more than 1200 patients showed the experimental arm to be more toxic and less effective than FOLFOX and bevacizumab alone. However, it should be noted that this study was not selected for RAS status, and the detrimental effect of adding anti-epidermal growth factor receptor to oxaliplatin in the RAS-mutated patients was shown in subsequent analyses. This negative effect for those with RAS mutation was also true in the Dutch CAIRO2 trial (9) with a PFS of 8.1 months vs 10.5 months (P = .04, mutated vs wild type) and an overall survival of 17.2 months vs 21.8 months (P = .03). The RAS-mutated patients also did substantially worse than the Cetuximab-Bevacizumab control arm (12.5 and 24.9 m for PFS and overall survival, respectively). In the first-line PRIME study of FOLFOX and panitumumab vs FOLFOX alone, the study was amended mid-course to include KRAS wild-type–only patients, and the final analysis clearly showed a detrimental effect of using the anti-epidermal growth factor receptor antibody in RAS-mutated patients (8).

In contrast to above, the BOND-2 (Bowel Oncology with Cetuximab Antibody) and BOND-3 trials (though smaller studies) showed benefit without substantially increased toxicity. The BOND-2 trial investigated bevacizumab, in combination with cetuximab with or without irinotecan [adding the VEGF antibody to each arm of BOND trial (3,4)]. In this trial, the addition of the anti-VEGF agent was feasible and increased the response rate from 20% to 37% and time to progression from 4.9 months to 7.3 months, again in a RAS nonselected patient population. In the more recent trial, BOND-3 (12) utilized the same arms but in RAS wild-type selected patients. Despite only 36 patients, median PFS was 9.7 months for ICB (irinotecan-cetuximab-bevacizumab) vs 5.5 months for irinotecan-cetuximab (HR = 0.68) and overall survival 19.7 months vs 10.2 months (HR = 0.41). These survivals are similar to the findings in the current study. Finally, and most recently, the PARADIGM trial, selecting for left-sided RAS wild-type colon cancer, showed survival benefit for FOLFOX with panitumumab vs bevacizumab (HR = 0.82) (13), suggesting the need to further explore combined antibodies in a similar cohort.

Ramucirumab is an active agent in CRC and was approved in the second-line setting on the basis of the RAISE study (10). In this study, 1072 patients, none of whom had prior bevacizumab, were randomly assigned 1:1 to FOLFIRI and ramucirumab vs FOLFIRI alone with PFS of 5.7 months vs 4.5 months (HR = 0.79) and overall survival of 13.3 months vs 11.7 months (HR = 0.85), leading to its US Food and Drug Administration approval in this setting. We note these results are somewhat inferior to our findings in E7208. The main toxicities of ramucirumab were somewhat higher rates of hypertension, neutropenia, and fatigue. When we designed our study, we considered the addition of ramucirumab to irinotecan-cetuximab regimen (ICR) at full doses could be well tolerated, but we included a real-time toxicity analysis at 34 patients for safety considerations. With a 1-dose level reduction in all drugs, a new arm with modified ICR became the new experimental arm. Modified ICR was tolerated well with equivalent toxicity yet more activity than IC in the subsequent patients enrolled.

Using the modified ICR regimen, patients tolerated treatment well, with no grade 5 toxicity and overall grade 3-4 toxicity rate of 56% compared with 59% with IC. The results of modified ICR showed improvement in PFS with a median of 9.2 months vs 6.0 months in the IC arm (HR = 0.75, P = .07 by Cox regression at interim analysis, which is statistically significant given alpha of 0.128 required by study design). Response rate was 36% for modified ICR compared with 23% for IC (P = .27), with disease control rate of 73% vs 52%, respectively (P = .05). Additionally, we did not limit patients to left-sided colon cancer, as is now clear, and did not capture these anatomical data.

We designed the trial to accommodate any situation where patients may be treated with second-line irinotecan chemotherapy following prior fluoropyrimidine, oxaliplatin, and bevacizumab. All patients were required to have received these previously. This approach contributed to completion of this study as compared with other such second-line efforts. The scenarios for the second-line group could include progression on the first-line regimen; progression on maintenance follow-up therapy with fluoropyrimidine, bevacizumab, or both; and progression when on treatment break and off all drugs. We stratified, therefore, based on: progression while on oxaliplatin, and time from last chemotherapy (≥6 months vs <6 months), which both had a profound effect on PFS (Figures 3 and 4). The hazard ratio for progression on oxaliplatin was 0.57, and for progression while off treatment for more than 6 months was 0.49.

Figure 4.

Forest plot for progression-free survival by key factors. CI = confidence interval; ECOG = Eastern Cooperative Oncology Group; HR = hazard ratio; IC = irinotecan-cetuximab; mICR = modified ICR; PFS = progression-free survival; PS = Performance Status; tx = treatment.

In conclusion, E7208 was the first National Clinical Trials Network trial successfully conducted in second-line therapy of CRC, and the first to treat a molecularly selected cohort. We demonstrated that second-line addition of an anti-angiogenic agent to anti-epidermal growth factor receptor antibody is feasible and appears to improve efficacy in appropriately selected patients. We also demonstrated the prognostic important of time since last treatment to progression and progression on first-line oxaliplatin for second-line trials. We showed similar benefit of the antibody combination as in the original BOND-2 combination experience and in the BOND-3 trial. Although we used the anti-VEGF receptor antibody ramucirumab rather than bevacizumab (an anti-VEGF ligand), multiple studies have shown similar levels of activity of all the anti-angiogenic antibodies and receptor molecules. The current attitude of toxicity and lack of benefit for the combination of an anti-epidermal growth factor receptor and anti-angiogenic likely stem from studies without appropriate patient selection factors, specifically RAS-mutation status and sidedness. Reconsideration of first-line anti-epidermal growth factor receptor plus anti-VEGF in appropriate first-line colon cancer patients (with left-sided RAS wild-type tumors) is warranted given this new understanding on optimal use of anti-epidermal growth factor receptor agents.

Data availability

It is the National Institutes of Health policy that results and accomplishments of the activities that it funds should be made available to the public. Furthermore, per NCI National Clinical Trials Network Guidelines, any investigator may submit a request for data from ECOG-ACRIN trials. Data are available from ECOG-ACRIN and can be shared upon review and approval of a data use request.

Author contributions

Howard S. Hochster, MD (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Supervision; Validation; Writing—original draft; Writing—review & editing), Paul Catalano, PhD (Data curation; Formal analysis; Supervision; Writing—original draft; Writing—review & editing), Michelle Weitz, PhD (Data curation; Formal analysis; Writing—original draft; Writing—review & editing), Edith P. Mitchell, MD (Investigation; Writing—review & editing), Deirdre Cohen, MD (Investigation; Methodology; Writing—review & editing), Peter J. O’Dwyer, MD (Conceptualization; Investigation; Methodology; Project administration; Resources; Writing—original draft; Writing—review & editing), Brian A. Faller, MD (Investigation; Supervision; Writing—review & editing), Jeremy S. Kortmansky, MD (Investigation; Supervision; Writing—review & editing), Mark H. O’Hara, MD (Investigation; Supervision; Writing—review & editing), Sheetal M. Kricher, MD (Investigation; Supervision; Writing—review & editing), Jill Lacy, MD (Investigation; Supervision; Writing—review & editing), Heinz-Josef Lenz, MD (Investigation; Project administration; Supervision; Writing—review & editing), Udit Verma, MD (Investigation; Supervision; Writing—review & editing), and Al B. Benson III, MD (Conceptualization; Formal analysis; Investigation; Methodology; Project administration; Supervision; Writing—original draft; Writing—review & editing).

Funding

This study was coordinated by the ECOG-ACRIN Cancer Research Group (Peter J. O’Dwyer, MD and Mitchell D. Schnall, MD, PhD, Group Co-Chairs) and supported by the National Cancer Institute of the National Institutes of Health under award numbers: U10CA180820, U10CA180794, U10CA180888, UG1CA233320, UG1CA233302, UG1CA233337, UG1CA189830, UG1CA180830, UG1CA233341 P30CA014089, and P30CA072720. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Previously presented at ASCO Annual meeting, June 2018.

Conflicts of interest

HH, who is a JNCI deputy editor and co-author on this paper, was not involved in the editorial review or decision to publish the manuscript.

HH: consultant: Genentech, Bayer.

POD: Research Support: Pfizer, Genentech, BMS, AZ, GSK, Five Prime, FortySeven, Merck, Syndax, BBI, Novartis, Celgene, Incyte, Lilly/Imclone, array, h3biomedicine, Taiho, Minneamrata, pharmacyclics/abbvie, Mirati, Expert testimony: Dai-ichi Sankyo.

MO: Institutional: BMS, Celldex, Natera, Genmab, Hibercell; consulting: Strike Pharma, Alligator.

JL: Consulting: Ipsen, BMS, Genentech, Novartis, Deciphera, Equinox, Techspert, KeyQuest, First Word Group.

HJL: Consulting: Merck KG, Bayer, Merck, Isofol, Oncocyte, Invitae, Affini-T, 3 T Bioscience, Repimmune, G1 Therapeutics, Jazz Therapeutics, Adagene; Fulgent.

Travel support: BMS.

Data committee: Veloxis.

Stock: Biobreak, Fulgent.

Acknowledgements

We dedicate the publication to the memory of Edith P. Mitchell, MD, FACP, who unexpectedly and suddenly passed away during the submission of this manuscript. She was an inspiration to us all. We also wish to thank the staff of the ECOG-ACRIN operations and statistical centers for their unwavering assistance. Finally, we wish to thank all the NCTN sites nationally and all the patients who contributed to this trial.