Assessment of Clinical Response Following Atezolizumab and Bevacizumab Treatment in Patients With Neuroendocrine Tumors: A Nonrandomized Clinical Trial

Key Points

Question  What is the response rate following combination treatment with the vascular endothelial growth factor inhibitor bevacizumab and the programmed cell death 1 ligand 1 inhibitor atezolizumab in patients with advanced neuroendocrine tumors (NETs)?

Findings  In this nonrandomized clinical trial of atezolizumab and bevacizumab that included 40 patients with NETs, responses were seen in 15% to 20% of patients.

Meaning  A subset of patients with NETs may be responsive to immunotherapy with bevacizumab and atezolizumab, potentially enriched for tumor programmed cell death 1 ligand 1 expression.

Importance  Therapies for patients with advanced well-differentiated neuroendocrine tumors (NETs) have expanded but remain inadequate, with patients dying of disease despite recent advances in NET therapy. While patients with other cancers have seen long-term disease control and tumor regression with the application of immunotherapies, initial prospective studies of single-agent programmed cell death 1 inhibitors in NET have been disappointing.

Objective  To evaluate the response rate following treatment with the combination of the vascular endothelial growth factor inhibitor bevacizumab with the programmed cell death 1 ligand 1 inhibitor atezolizumab in patients with advanced NETs.

Design, Setting, and Participants  This single-arm, open-label nonrandomized clinical study in patients with rare cancers included 40 patients with advanced, progressive grade 1 to 2 NETs (20 with pancreatic NETs [pNETs] and 20 with extrapancreatic NETs [epNETs]) treated at a tertiary care referral cancer center between March 31, 2017, and February 19, 2019. Data were analyzed from June to September 2021.

Interventions  Patients received intravenous bevacizumab and atezolizumab at standard doses every 3 weeks until progression, death, or withdrawal.

Main Outcomes and Measures  The primary end point was objective radiographic response using Response Evaluation Criteria in Solid Tumors, version 1.1, with progression-free survival (PFS) as a key secondary end point.

Results  Following treatment of the 40 study patients with bevacizumab and atezolizumab, objective response was observed in 4 patients with pNETs (20%; 95% CI, 5.7%-43.7%) and 3 patients with epNETs (15%; 95% CI, 3.2%-37.9%). The PFS was 14.9 (95% CI, 4.4-32.0) months and 14.2 (95% CI, 10.2-19.6) months in these cohorts, respectively.

Conclusions and Relevance  In this nonrandomized clinical trial, findings suggest that clinical responses in patients with NET may follow treatment with the combination of bevacizumab and atezolizumab, with a PFS consistent with effective therapies.

Trial Registration  ClinicalTrials.gov Identifier: NCT03074513

Well-differentiated neuroendocrine tumors (NETs) are rare, heterogeneous neoplasms with rising incidence and prevalence.¹ While metastatic NETs are often indolent, they are incurable and frequently life-limiting. Available therapies depend on the primary site, with patients with pancreatic NETs (pNETs) deriving benefit with temozolomide, sunitinib, lanreotide, and everolimus, while patients with extrapancreatic NETs (epNETs) also benefit from lanreotide, octreotide, and everolimus.² Peptide receptor radionuclide therapy with ¹⁷⁷Lu-Dotatate is supported by randomized clinical trials,³ single-arm studies, and retrospective data,⁴ leading to regulatory approval for gastroenteropancreatic (GEP) NETs. However, novel therapies with potential for more durable control are needed, spurring interest in immunotherapy for patients with NETs.

Checkpoint inhibition of the programmed cell death 1 (PD-1)/PD-1 ligand 1 (PD-L1) axis has been disappointing in well-differentiated NETs. Combined data from nearly 400 patients with grade 1 to 3 NETs have yielded response rates less than 10% and progression-free survival (PFS) less than 6 months.⁵ Therefore, strategies combining checkpoint inhibition with additional agents have been of particular interest in NETs.

Atezolizumab targets the interaction of PD-L1 with PD-1 and B7-1,⁶ and bevacizumab targets vascular endothelial growth factor (VEGF).⁷ In the Rip-Tag murine model of pNET, VEGF blockade increased PD-L1 expression and tumor T-cell trafficking, improving survival when combined with PD-L1 blockade,⁸ consistent with prior clinical and preclinical data.^9,10 We therefore explored the safety and efficacy outcomes of combination therapy with bevacizumab and atezolizumab in patients with NETs within a basket study.

We report the pNET and epNET cohorts from an open-label, phase 2 basket trial with 8 parallel rare-tumor cohorts. This single-center study was approved by the MD Anderson Institutional Review Board. All patients had metastatic or locally advanced, grade 1 to 2 NETs (World Health Organization 2010 classification), with or without prior therapy, including anti-VEGF, radiographically progressing over the prior 12 months by investigator assessment. Somatostatin analogue therapy could be continued if dosed stably for 8 weeks before enrollment.

After providing written informed consent, patients received bevacizumab, 15 mg/kg, and atezolizumab, 1200 mg, intravenously on day 1 (21-day cycle) until disease progression, consent withdrawal, an adverse event preventing further administration, or conditions rendering further therapy unacceptable to the patient or investigator. Treatment beyond radiographic progression was permitted, provided an acceptable risk-benefit ratio. Toxic effects were assessed using Common Terminology Criteria for Adverse Events, version 4.0, and management guidelines were well delineated (see protocol in Supplement 1). Tumors were measured using computed tomography or magnetic resonance imaging every 4 cycles. On day 1 of cycles 1 and 2, tumor biopsies were collected.

The primary end point was intention-to-treat objective response rate (ORR), defined as the proportion of patients having 30% or greater reduction in the sum of the longest tumor diameters (Response Evaluation Criteria in Solid Tumors, version 1.1). Secondary end points included PFS and overall survival (OS). Radiologists blinded to study participation performed tumor metrics. Exploratory analyses included whole exome sequencing for tumor mutational burden, RNASeq for gene expression profiling and immune deconvolution, and PD-L1 immunohistochemistry staining. Complete methods are provided in eMethods in Supplement 2.

The ORRs and 95% CIs were estimated using the Clopper-Pearson method. Time-to-event outcomes were estimated using the Kaplan-Meier method. Patient characteristics and adverse events were provided with summary statistics. Statistical calculations were performed using SPSS, version 25 (IBM Corp). The Transparent Reporting of Evaluations With Nonrandomized Designs (TREND) reporting guideline checklist was used for reporting.

Between March 31, 2017, and February 19, 2019, we enrolled 20 patients with pNETs and 20 patients with epNETs. The data cutoff date was June 15, 2021. The study population was typical for these diseases (Table). The median (IQR) number of cycles administered was 17 (9.3-23.8) and 12.5 (8.25-31.5) for patients with pNETs and epNETs, respectively. Eighteen patients (6 with epNETs and 12 with pNETs) died, with a median follow-up of 43.8 (95% CI, 39.0-46.1) months.

At least 6 months after enrollment completion, the confirmed ORRs in patients with pNETs and epNETs were 20% (95% CI, 5.7%-43.7%) and 15% (95% CI, 3.2%-37.9%), respectively (Figure 1). Median PFS was 14.9 (95% CI, 4.4-32.0) months and 14.2 (95% CI, 10.2-19.6) months, respectively (Figure 2A). The median OS was 30.1 months (95% CI, 17.7 months-not reached) for patients with pNETs and not reached for patients with epNETs (Figure 2B). Among the 7 responders, median (range) duration of response was 27 (11-38) months.

The combination was well tolerated, with no unanticipated safety signals (eTable in Supplement 2). The most common adverse events were hypertension, proteinuria, and fatigue. Two (5%) patients discontinued protocol therapy for toxic effects—gastrointestinal hemorrhage and pain, respectively (eFigure 1 in Supplement 2 for flow).

While the complete exploratory immune characterization of these cohorts is beyond the present scope, 5 patients with pNETs (25%; 95% CI, 9%-49%) and 0 patients with epNETs (0%; 95% CI, 0%-17%) had PD-L1 expression detectable by standard chromogenic immunohistochemistry (Ventana PD-L1, clone SP263) in greater than 1% of tumor cells at baseline. Four of 5 (80%; 95% CI, 28%-99%) demonstrated objective radiographic tumor response (eFigure 2 in Supplement 2). World Health Organization tumor grade was not significantly associated with PD-L1 expression or radiographic response. To further characterize the tumor immune milieu, we also performed CIBERSORT deconvolution from RNA, suggesting an immunosuppressive M2 macrophage score higher in nonresponsive patients (eFigure 2B in Supplement 2). Baseline tumor mutational burden was not significantly higher in responders (eFigure 2C in in Supplement 2).

Our data suggest that the combination of bevacizumab and atezolizumab is associated with tumor regression in select patients with NETs, without new or concerning safety signals. Furthermore, our data suggest that PD-L1 expression in greater than 1% of tumor cells by immunohistochemistry may be associated with efficacy. In contrast, unresponsive tumors demonstrated a gene expression signature suggesting an immunosuppressive M2 macrophage milieu, consistent with the established relationship between M2 macrophages and poor immunotherapy outcomes.¹¹

These data are intriguing because single-agent PD-1/PD-L1 blockade has previously shown minimal activity in well-differentiated GEP-NETs, with no association between responses and PD-L1 expression. In the KEYNOTE-158 study of pembrolizumab in 107 patients with lung or GEP-NETs,¹² objective responses were observed in 4 (3.7%; 95% CI, 1-9.3) patients, all with PD-L1–negative tumors. Cohort analysis is consistent with our observation of PD-L1 expression in 20% to 25% of patients with GEP-NETs,¹³ though no potential biomarkers of immunotherapy response have yet been identified in NETs to our knowledge.

In contrast, VEGF inhibitors have known efficacy in patients with NETs. In the pivotal randomized study of sunitinib compared with placebo in patients with advanced pNETs,¹⁴ an ORR of 9% (95% CI, 4%-18%) was observed, though an immune mechanism has not been shown to be associated with the observed responses. While we cannot assert synergy based on clinical activity alone and a larger randomized trial would be needed to determine the relative contributions of atezolizumab and bevacizumab, the associations of response with PD-L1 expression and M2 expression profile suggest that this combination may have achieved a novel benefit for patients with NETs, particularly in patients with PD-L1–positive tumors. Ongoing similar studies, such as that of pembrolizumab/lenvatinib,¹⁵ will be particularly interesting, especially if M2 macrophages are relevant to the NET immune milieu.

These early data have intrinsic limitations, principally the study size, which was necessary to maintain clinical and translational feasibility in this rare population. Nonetheless, the multidimensional translational insights are unique in this rare disease, to our knowledge.

In this nonrandomized clinical trial, findings suggest that the combination of atezolizumab and bevacizumab is the first regimen to demonstrate a clinical response profile potentially associated with the immune milieu of well-differentiated NETs, as well as a promising PFS overall. Furthermore, our data suggest that this regimen could be worthy of further investigation, with special attention to the 15% to 25% of patients with NETs whose tumors express PD-L1.

Accepted for Publication: December 22, 2021.

Published Online: April 7, 2022. doi:10.1001/jamaoncol.2022.0212

Corresponding Author: Daniel M. Halperin, MD, Department of Gastrointestinal Medical Oncology, Unit 426, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030 (dmhalperin@mdanderson.org).

Author Contributions: Dr Halperin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Woodman and Yao contributed equally.

Concept and design: Halperin, Estrella, Dervin, Wistuba, Woodman, Yao.

Acquisition, analysis, or interpretation of data: Halperin, Liu, Dasari, Fogelman, Bhosale, Mahvash, Rubin, Morani, Knafl, Overeem, Fu, Solis, Parra Cuentas, Verma, Chen, Gite, Subashchandrabose, Schulze, Darbonne, Yun, Wistuba, Futreal, Woodman, Yao.

Drafting of the manuscript: Halperin, Rubin, Overeem, Chen, Subashchandrabose.

Critical revision of the manuscript for important intellectual content: Halperin, Liu, Dasari, Fogelman, Bhosale, Mahvash, Estrella, Morani, Knafl, Fu, Solis, Parra Cuentas, Verma, Gite, Dervin, Schulze, Darbonne, Yun, Wistuba, Futreal, Woodman, Yao.

Statistical analysis: Halperin, Liu, Rubin, Knafl, Overeem, Fu, Woodman.

Obtained funding: Halperin, Darbonne, Wistuba, Futreal, Yao.

Administrative, technical, or material support: Halperin, Dasari, Bhosale, Mahvash, Morani, Knafl, Solis, Verma, Chen, Subashchandrabose, Dervin, Schulze, Darbonne, Woodman, Yao.

Supervision: Halperin, Parra Cuentas, Woodman, Yao.

Conflict of Interest Disclosures: Dr Halperin reported receiving grants from Genentech/Roche during the conduct of the study; and receiving grants from Tarveda and Thermo Fisher Scientific; grants and personal fees from AAA/Novartis and ITM; and personal fees from Ipsen, Lexicon, TerSera, Curium, AlphaMedix, Camurus, Crinetics, and Chimeric Therapeutics outside the submitted work. Dr Dasari reported receiving grants from HutchMed, Merck, and Eisai and personal fees from Novartis outside the submitted work. Dr Fogelman reported being a salaried employee of Merck (began after completion of this study) outside the submitted work. Dr Mahvash reported receiving grants from Sirtex Medical, Boston Scientific Corporation, and Siemens Healthineers and personal fees from ABK Biomedical, Sirtex Medical, and Boston Scientific Corporation outside the submitted work. Dr Dervin reported receiving personal fees from Genentech, a member of the Roche Company, as a previous employee of Genentech, outside the submitted work. Dr Schulze reported being an employee of Genentech Inc and owning stock in Roche during the conduct of the study. Mr Darbonne reported personal fees for employment from Genentech during the conduct of the study and outside the submitted work. Ms Yun reported being an employee of Roche/Genentech with stock ownership of Roche. Dr Wistuba reported receiving grants from Genentech during the conduct of the study; and receiving grants and personal fees from Genentech/Roche, Bayer, AstraZeneca, Pfizer, Merck, GlaxoSmithKline, Guardant Health, HTG Molecular, Novartis, Amgen, Sanofi, and Daiichi Sankyo; grants from Adaptive, Adaptimmune, EMD Serono, Takeda, Karus, Johnson & Johnson, Iovance, 4D, and Akoya; and personal fees from Bristol Myers Squibb, Flame, Oncocyte, and Platform Health outside the submitted work. Dr Yao reported receiving personal fees from Hutchison Medi Pharma, Ipsen Biopharmaceuticals, Amgen, Chiasma, and Critics outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by Genentech/Roche, National Institutes of Health Cancer Center Support Grant Award (CA016672 [Institutional Tissue Bank and Research Histology Core Laboratory ]), the Adaptive Patient-Oriented Longitudinal Learning and Optimization (APOLLO) Moon Shots Program, the Translational Molecular Pathology-Immunoprofiling lab in the Department of Translational Molecular Pathology at MD Anderson Cancer Center, and the Strategic Translational Research Platform. Dr Halperin was supported by a Career Development Award from the Conquer Cancer Foundation of the American Society of Clinical Oncology.

Role of the Funder/Sponsor: The funders provided study design input, with no role in data collection, analysis, or interpretation; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: The authors gratefully acknowledge additional technical and administrative support from: Grace Mathew, MS, Wenhua Lang, BS, Celia Garcia-Prieto, PhD, Liren Zhang, PhD, Julia Mendoza-Perez, PhD, Wei Lu, PhD, Jianling Zhou, BS, Mei Jiang, BS, Auriole Tamegnon, BS, Renganayaki Krishna Pandurengan, BS, Shanyu Zhang, BS, Beatriz Sanchez-Espiridion, PhD, and Sandesh Subramanya, PhD.

Data Sharing Statement: See Supplement 3.