NRG/RTOG 1122: phase II double-blinded, placebo-controlled study of bevacizumab with/without trebananib in patients with recurrent glioblastoma or gliosarcoma

Abstract

Background:

Targeting vascular endothelial growth factor (VEGF) alone does not improve overall survival in recurrent glioblastoma (rGBM). The angiopoiein (Ang)-TIE2 system may play a role in tumor survival under VEGF inhibition. We conducted a phase II, double-blinded, placebo-controlled trial of bevacizumab + trebananib (a novel Fc fusion protein that sequesters Ang1/Ang2) over bevacizumab alone in rGBM.

Methods:

Patients 18 years or older with a KPS ≥ 70 and GBM or variants in first or second relapse were randomized to bevacizumab 10 mg/kg every 2 weeks + trebananib 15 mg/kg every week or bevacizumab + placebo. The primary endpoint was 6-month progression-free survival (6m PFS).

Results:

After an initial 6-patient lead-in cohort confirmed the safety of combining bevacizumab and trebananib, 115 eligible patients were randomized to the control (n=58) or experimental treatment (n=57). In the control arm, 6m PFS was 41.1%, median survival time (MST) was 11.5 months [95% CI: 8.4-14.2], median PFS was 4.8 months [95% CI: 3.8-7.1], and RR was 5.9%. In the experimental arm, 6m PFS was 22.6%, MST was 7.5 months [95% CI 6.8-10.1], median PFS was 4.2 months [95% CI 3.7-5.6], and RR was 4.2%. The rate of severe toxicities was not significantly different between arms.

Conclusion:

The combination of bevacizumab and trebananib was well tolerated but did not significantly improve 6m PFS rate, PFS, or OS for patients with rGBM over bevacizumab alone. The shorter PFS in the experimental arm with a HR=1.51 (p=0.04) suggests that the addition of trebananib to bevacizumab is detrimental.

Trial Registration:

ClinicalTrials.gov identifier NCT01609790

Precis:

In this phase II, double-blinded, placebo-controlled trial in patients with recurrent GBM through NRG oncology, the combination of bevacizumab plus trebananib (an angiopoietin inhibitor) did not significantly improve 6m PFS rate, PFS, or OS for patients with recurrent GBM as compared to bevacizumab plus placebo. The Ang1 blocking effects of trebananib may somehow counteract or negate the anti-tumor effects of Ang2-VEGF blockade.

Introduction

Bevacizumab, a monoclonal antibody against vascular endothelial growth factor (VEGF), is frequently used to treat patients with recurrent glioblastoma (GBM) and received FDA approval for this indication.^{1, 2} Although randomized, phase III trials in newly diagnosed GBM suggest an improvement in progression-free survival (PFS), the addition of bevacizumab to standard therapy is not associated with an improvement in overall survival (OS) compared to standard therapy alone.^{3, 4} Indeed, GBMs often develop resistance to bevacizumab treatment within months of starting therapy.^{1, 5} Targeting VEGF alone may not be sufficient for durable responses. Intrinsic and acquired resistance may be due to the redundancy of angiogenesis stimulators.^{6, 7}

Angiopoietins (Ang) may play a role in mediating resistance to anti-VEGF therapy in GBM. In a phase II clinical trial of cediranib [a pan-VEGF receptor (VEGFR) inhibitor] in recurrent GBM, progression on cediranib correlated with increases in the soluble form of the Ang-binding Tie2 receptor.^{8, 9} In a U87 glioma xenograft model, dual inhibition of VEGFR and Ang2 inhibited tumor growth, prolonged vessel normalization compared to VEGFR inhibition alone, and improved survival through reprogramming of tumor-associated macrophages toward an antitumor phenotype.^{10, 11} In addition, a subset of GBM patients on anti-VEGF therapy develops a non-enhancing tumor pattern of progression, believed to represent a more invasive phenotype.^{5, 12, 13} Upregulation of Ang2 correlates with the invasiveness of gliomas.^{14, 15} Blockade of Ang2 may decrease the invasiveness of gliomas and abrogate an escape mechanism for anti-VEGF therapy. Data from a colon cancer xenograft model suggest that combined Ang1/Ang2 inhibition confers better tumor suppression than Ang2 inhibition alone.¹⁶ Furthermore, in the same models, Ang1/Ang2 inhibition plus bevacizumab mediated better tumor suppression than Ang2 inhibition plus bevacizumab. For these reasons, dual inhibition of the VEGF and angiopoietin-TIE2 signaling pathways is an attractive therapeutic strategy in the recurrent GBM population.

Trebananib (AMG 386) is a novel peptide-Fc fusion protein that sequesters Ang1 and Ang2, thereby preventing their interaction with Tie2 and inhibiting tumor endothelial cell proliferation and tumor growth. We performed a phase II, double-blinded, placebo-controlled trial in patients with recurrent GBM comparing bevacizumab plus trebananib versus bevacizumab plus placebo.

Patients and Methods

Patients

Eligibility criteria included age ≥18 years at registration with a histologic diagnosis of GBM or variants (gliosarcoma, glioblastoma with oligodendroglial features, giant cell glioblastoma) in first or second relapse with unequivocal radiographic evidence for tumor progression ≤14 days prior to registration. Patients with a secondary GBM were also eligible as long as GBM had been histologically proven. For patients with a recent tumor resection for progressive disease, registration on study could not occur any sooner than 28 days from surgery; Additional eligibility criteria included supratentorial disease only; Karnofsky performance status (KPS) ≥70; and adequate renal, hepatic, and bone marrow function.

No prior treatment with VEFG inhibitors including bevacizumab or with angiopoietin-TIE2 inhibitors was allowed. Exclusion criteria also included MRI evidence of recent intracranial hemorrhage, history of bleeding diathesis or coagulopathy, clinically significant cardiovascular disease including myocardial infarction and stroke within 180 days prior to registration, and uncontrolled hypertension. Therapeutic or prophylactic therapy with aspirin, a low-molecular weight heparin, or a Factor Xa inhibitor was allowed, but not warfarin.

All patients provided written informed consent. The study was approved by the institutional review board or the equivalent panel at each study center before patient enrollment. The study was registered at clinicaltrials.gov (NCT01609790).

Study Design, Treatment, and Endpoints

For the phase II portion, under the permuted block design,¹⁷ patients were stratified by age (<50 vs. ≥50), KPS (70-80 vs. 90-100), and recent resection (Yes vs. No/Biopsy), and then randomly assigned to receive trebananib or placebo with a 1:1 allocation ratio. All patients received bevacizumab at a dose of 10 mg/kg every two weeks. Trebananib (or placebo) was administered intravenously at a dose of 15 mg/kg every week until disease progression or severe treatment-related toxicity. At the time of tumor progression, patients were unblinded. Those who received bevacizumab monotherapy were allowed to cross over and to receive treatment with bevacizumab and trebananib, provided that they still met eligibility criteria.

The primary endpoint was 6-month progression-free survival rate (6m PFS) and was confirmed by central review (EQL, ERG). Secondary endpoints included PFS, OS, radiographic response (RR), and safety. Toxic effects were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0. Radiographic assessments were performed by the local investigator and were based on the Response Assessment in Neuro-Oncology (RANO) criteria.¹⁸ Additional details regarding study design can be found in the supplement.

Statistical Methods

The primary endpoint was 6m PFS. Based on the March 2009 FDA briefing¹⁹, 6m PFS for bevacizumab monotherapy in recurrent GBM patients is 36%. We hypothesized that the addition of trebananib to bevacizumab in bevacizumab-naïve patients would result in a 6m PFS rate of 55%. Assuming exponential distribution for PFS time, these rates corresponded to a median PFS time of 4.1 and 7 months for the control and experimental arm, respectively, with a hazard ratio (HR) of 0.59. With a one-sided significance level at 0.15, a total of 114 patients (57 per arm) would yield an 85% power to detect the projected effect size on 6m PFS, based on a two-sample test on proportions. To guard against up to a 10% ineligible rate, 127 patients were required to be accrued for the phase II portion of the study.

An interim futility analysis on 6m PFS was designed to be performed when 50% of the total required eligible patients had a minimum of 6-month follow-up. The analysis followed the principle of intent-to-treat, with all the eligible cases included in the treatment arm to which they were randomized regardless of what treatment the patients actually received. The decision rule was that if the 6m PFS rate for the experimental arm was worse than that for the control arm by any amount, the study would be stopped and reported early.

PFS was measured from the date of randomization to the date of first progression or death, whichever came first, or otherwise, the last follow-up date on which the patient was reported alive without experiencing disease progression. OS was measured from the date of randomization to the date of death, or otherwise, the last follow-up date on which the patient was reported alive. OS and PFS rates by treatment arm were estimated using the Kaplan-Meier method. The HRs for the treatment effects on OS and PFS were estimated using the Cox proportional hazard model and tested using the log-rank test. Multivariate analyses with the Cox proportional hazard model were performed with the stratification factors and other patient pretreatment characteristics as covariates. Differences in the rate of the reported severe toxicities and RR between treatment arms were tested using the Chi-square test. For all the secondary endpoints, two-sided tests with a significance level of 0.05 were used in the analyses.

The efficacy (as measured by OS, PFS, RR and 6m PFS) of the salvage treatment for patients who crossed over from the control arm to receive the experimental drug after disease progression was investigated as exploratory endpoints. For this group of patients, OS and PFS were measured from the date of initial progression (before cross-over). OS and PFS rates were similarly estimated using the Kaplan-Meier method. The incidence rates of RR and the reported severe toxicities were also calculated.

Results

Patients

One hundred and thirty patients were enrolled across 41 centers in the United States through NRG oncology and randomized on the phase II portion of the study between March 2013 and September 2014 (Figure 1; CONSORT diagram), of whom 15 (11.5%) were subsequently found to be ineligible. In total, 58 and 57 eligible patients randomized to the control and experimental arms, respectively, were included in the analyses. Table 1 shows the distributions of pretreatment characteristics by treatment arm. The distributions of the stratification factors were balanced between the two treatment arms.

Figure 1.

CONSORT diagram. Enrollment onto the safety run-in portion of the study began on June 4, 2012. It temporarily closed on October 2, 2012, when the target accrual of 7 patients for the safety run-in component was met. Enrollment onto the randomized portion of the study began on March 13, 2013, and accrual was competed on September 2, 2014.

Table 1.

Patient and tumor characteristics at baseline for all eligible patients.

	Bevacizumab + Placebo (n=58)		Bevacizumab + Trebananib (n=57)
Patient or Tumor Characteristic	n	%	n	%
Age (years)
Median	58		57
Min - Max	22 - 79		30 - 80
< 50	17	29.3	14	24.6
≥50	41	70.7	43	75.4
Gender
Male	36	62.1	32	56.1
Female	22	37.9	25	43.9
Race
Asian	2	3.4	0	0.0
Black or African American	3	5.2	2	3.5
White	53	91.4	53	93.0
Unknown or not reported	0	0.0	2	3.5
Ethnicity
Hispanic or Latino	3	5.2	2	3.5
Not Hispanic or Latino	55	94.8	54	94.7
Unknown	0	0.0	1	1.8
Karnofsky Performance Status
70-80	31	53.4	34	59.6
90-100	27	46.6	23	40.4
Neurologic Function
No symptoms	12	20.7	11	19.3
Minor symptoms	32	55.2	27	47.4
Moderate symptoms (fully active)	10	17.2	16	28.1
Moderate symptoms (required assistance)	4	6.9	3	5.3
Surgery (Initial Brain Tumor)
Biopsy only	8	13.8	3	5.3
Subtotal resection	19	32.8	12	21.1
Gross total resection	30	51.7	42	73.7
Other	1	1.7	0	0.0
Recent resection
No/biopsy only	30	51.7	30	52.6
Yes	28	48.3	27	47.4
Histologic Tumor Type
Glioblastoma	53	91.4	53	93.0
Gliosarcoma	2	3.4	2	3.5
Glioblastoma with Oligodendroglial Features	2	3.4	1	1.8
Giant Cell Glioblastoma	0	0.0	1	1.8
Other	1	1.7	0	0.0

Outcomes

The interim futility analysis was performed in March 2015. Out of the 27 and 30 eligible patients from the experimental and control arms, respectively, 7 (25.9%) and 11 (36.7%) were progression-free at 6 months after randomization. Of note, the total accrual to the study had already been reached and all patients had had 6-month follow-up by the time of this analysis. The report of this analysis was reviewed by the NRG Oncology Data Monitoring Committee, and it was recommended to unblind protocol treatment and to report the final analyses based on the data from all the randomized and eligible patients.

One hundred and nine (94.8%) patients were evaluable for 6m PFS (Table 2). Twelve of the 53 (22.6%) in the experimental arm were progression-free at 6 months after randomization, as compared with 23 of the 56 (41.1%) in the control arm (one-sided p-value=0.98).

Table 2.

Study Outcomes

Outcome	Randomized Cohort			Cross-Over Cohort
	Bevacizumab + Placebo	Bevacizumab + Trebananib	P-value	Bevacizumab + Trebananib
6m PFS (n, %)	23/56 (41.1%)	12/53 (22.6%)	0.98	1/22 (4.5%)
OS (median, 95% CI)	11.5 months (8.4, 14.2)	7.5 months (6.8, 10.1)	0.09	4.9 months (3.8, 8.4)
PFS (median, 95% CI)	4.8 months (3.8, 7.1)	4.2 months (3.7, 5.6)	0.04	2.3 months (2.1, 3.2)
RR for evaluable patients			0.99*
CR	0	0		0
PR	3/51 (5.9%)	2/48 (4.2%)		0
SD	16/51 (31.4%)	9/48 (18.8%)		0
PD	32/51 (62.7%)	37/48 (77.1%)		9/9 (100%)

^*p-value by the Fisher’s exact test for PR vs SD/PD by treatment arms is 0.99

Abbreviations:

6m PFS: 6-month progression-free survival

CR: complete response

OS: overall survival

PD: progressive disease

PFS: progression free survival

PR: partial response

RR: radiographic response

SD: stable disease

The median follow-up time for the patients who were still alive at the time of the analysis was 15.0 months (2.5-26.4 months). The median survival time (MST) was 11.5 months (95% CI: 8.4-14.2 months) for the control arm and 7.5 months (95% CI: 6.8-10.1 months) for the experimental arm, with an HR=1.46 (95% CI: 0.95-2.27; p-value=0.09) (Figure 2A). The median PFS time was 4.8 months (95% CI: 3.8-7.1 months) for the control arm and 4.2 months (95% CI: 3.7-5.6 months) for the experimental arm, with an HR=1.51 (95% CI: 1.02-2.24; p-value=0.04) (Figure 2B). After adjusting for the stratification factors and other patient pretreatment characteristics, the adjusted HR of the treatment effect on OS was 1.46 (95% CI: 0.94-2.26) with a p-value=0.09 and that on PFS was 1.50 (95% CI: 1.01-2.23) with a p-value=0.05.

Figure 2.

Overall survival (A) and progression-free survival (B) by treatment arm.

Table 2 also contains the best RR by RANO criteria. Overall, 99 (86.1%) patients were evaluable. For the 51 patients in the control arm, 3 (5.9%) achieved a partial response (PR). For the 48 patients in the experimental arm, 2 (4.2%) achieved PR. No patient in either treatment arm achieved a complete response (CR). No significant difference was found between the two treatment arms.

Safety

Information for adverse events (AEs) without regard to attribution is presented by treatment arm for all the randomized and eligible patients who received protocol treatment (Supplementary Table 1). Although the proportion of patients with grade 3+ AEs was numerically higher for the control arm, 80.7% versus 64.9% for the experimental arm, this difference did not reach statistical significance (p-value=0.06). Similarly, the proportion of patients with grade 3+ AEs attributed to study treatment (36.8% of patients in the control arm vs. 35.1% in the experimental arm) was not significantly different (p-value=0.85). Overall, there were 2 patients (3.5%) in the control arm, and 5 (8.8%) in the experimental arm with reported grade 5 AEs, all of which were deemed unrelated to protocol treatment.

Cross-Over Study

A total of 25 patients crossed over from the control arm to receive open-label trebananib and bevacizumab after disease progression (Table 2). Twenty-two (88.0%) of them were evaluable for 6m PFS, and only 1 (4.5%) was progression-free at 6 months after initial disease progression on bevacizumab. Nine (36.0%) patients were evaluable for the best RR, all with progressive disease (PD) based on the RANO criteria. MST from the date of initial progression was 4.9 months (95% CI: 3.8-8.4 months), and median PFS was 2.3 months (95% CI: 2.1-3.2 months). No grade 5 AE was reported in the cross-over cohort, and 1 (4.0%) patient experienced grade 4 sepsis.

Pharmacokinetics

Four of six patients in cohort 1 and 17 of 25 in the cross-over arm participated in the optional PK and anti-trebananib antibody studies. The peak and trough concentrations of trebananib on Days 1 and 8 of Cycles 2 were comparable to those from previous studies.²⁰ Minimal variability of trebananib concentrations were observed between Days 1 and 8 of cycle 2, suggesting the steady state pharmacokinetics were reached.

Discussion

The combination of trebananib and bevacizumab did not significantly improve the 6m PFS rate for patients with recurrent GBM in first or second relapse as compared to bevacizumab alone. Indeed, median PFS was longer in the control arm than the experimental arm. In comparison to other randomized studies comparing bevacizumab alone with in combination with another systemic agent (such as carboplatin,²¹ dasatinib,²² or onartuzumab²³) in recurrent GBM, our study similarly fails to demonstrate a survival advantage to combination therapy. We also examined the benefit of adding trebananib to bevacizumab in patients who progressed on bevacizumab monotherapy. Survival times were similar to prior studies examining the role of combining bevacizumab with a systemic agent after progression on bevacizumab.¹

The best RR was similar for the two treatment arms (around 5.0%), but lower than the historical benchmarks from prior bevacizumab studies (35% in Kreisl et al.¹ and 46.4-65.8% in Friedman et al.²). This is partly explained by differences in RR assessment methods, with the historical benchmark studies utilizing Macdonald criteria²⁴ and our study utilizing RANO criteria¹⁸. When Friedman et al. was reassessed by independent reviewers using RANO criteria, RR dropped to 33.1%.²⁵ The even lower RR seen in our study could reflect a heightened understanding that decreased contrast enhancement on bevacizumab may not reflect true anti-tumor responses.¹⁸ In addition, RR assessment was performed by the local investigator, some of whom were not neuro-oncologists or neuro-radiologists and may have been more apt to call progression earlier than investigators in prior benchmark studies.

The MST of 11.5 months in the control arm was numerically but not statistically significantly longer than the experimental arm (7.5 months). Although PFS was numerically similar (4.8 months in the control arm versus 4.2 months in the experimental arm), PFS was significantly longer in the control arm (p=0.04). Except for the protocol specified salvage treatment (placebo arm participants crossing-over to experimental arm) and non-protocol surgery (6 patients who progressed on the placebo arm underwent non-protocol surgery as compared with 0 on the experimental arm), there was no significant difference in reported non-protocol treatments given after disease progression between the two arms (Supplementary Table 2). When comparing patients in the control arm that crossed-over (N=25) to receive the experimental treatment versus those who did not cross-over (N=33), the patients who did not cross-over lived numerically longer (with a MST of 12.1 months versus 9.9 months). These data suggest that adding trebananib to bevacizumab could be detrimental compared to bevacizumab alone. In a separate study of trebananib with or without bevacizumab, there was no clear benefit from either trebananib alone or in combination with bevacizumab.²⁶ In the trebananib + bevacizumab arm of that study, outcomes were similar to our study with 6m PFS 24.3%, median OS 9.5 months, and median PFS 3.6 months.

It is unclear why patients receiving bevacizumab alone fared better than patients receiving bevacizumab + trebananib. One potential reason relates to the complex interactions between Ang1 and Ang2. Ang2 can act as a Tie2 antagonist or partial agonist depending on cellular context.²⁷ Therefore, depending on the balance of Ang1 and Ang2, targeting the Ang-Tie2 pathway has the potential to promote or inhibit tumor growth.²⁷ Several studies examining the effects of Ang1 or Ang2 blockade suggest that Ang2 may play a more important role in tumor angiogenesis.²⁷ In addition, tumor models of several cancers suggest that Ang1 overexpression reduces tumor growth.²⁸ It is possible that the Ang1 blocking effects of trebananib may somehow counteract or negate the anti-tumor effects of Ang2-VEGF blockade.

Another possible explanation for the lack of benefit from combined therapy is upregulation of alternative proangiogenic pathways. Other mechanisms that have been implicated in angiogenesis resistance include upregulation of alternative proangiogenic pathways such as basic fibroblast growth factor (bFGF) pathway, hepatocyte growth factor/cMET pathway, and SDF-1α; recruitment of bone marrow-derived proangioenic cells; and increased pericyte coverage of tumor vessels leading to vessel co-option.^{29, 30} It is possible that co-inhibition of the VEGF and angiopoietin/TIE2 pathways is still insufficient to prevent angiogenesis escape.

In summary, the combination of bevacizumab and trebananib failed to demonstrate improvements in 6m PFS, PFS, OS, or RR compared to bevacizumab alone. Careful consideration should be given to future trials of angiopoietin inhibitors in combination with bevacizumab given the possible detrimental effects of Ang1 in this study. Dual inhibitors of Ang2 and VEGFR are in development and may have therapeutic potential for glioblastoma.