Integration of Antiangiogenic Therapy with Cisplatin and Gemcitabine Chemotherapy in Patients with Nasopharyngeal Carcinoma

Nasopharyngeal cancer usually presents as locally advanced stage III or IV disease with T3-4 or N2-3 nodal status (1). The standard of care of patients with locally advanced nasopharyngeal carcinoma is concurrent cisplatin chemotherapy with radiotherapy (2). However, locally advanced stage IV patients with T4 or N3 disease have the highest risk of recurrence and metastasis; and may benefit from additional strategies to improve their outcome (3, 4). Emerging evidence suggests that induction chemotherapy improves patient survival in this group of patients when integrated with concurrent chemoradiation (5, 6). A meta-analysis also showed that the addition of induction chemotherapy improved overall survival, and reduced locoregional and distant failure (7).

Dysregulation of the VEGF/VEGF receptor-2 axis is a feature of many cancers including NPC. Elevated VEGF expression promotes disorganized tumor neovascularization with hyperpermeable, tortuous capillaries that have immature pericyte coverage (8, 9). Leakiness of vessels increases interstitial fluid pressure that collapses vessels with the collective effect of compromising blood flow and perfusion, limiting the delivery of therapeutic drugs. Furthermore, tumor expression of VEGF promotes immune tolerance directly and through hypoxic mechanisms, suppressing dendritic cell maturation, dampening immune activation through immunosuppressive cells like regulatory T cells, myeloid-derived suppressor cells, and M2 macrophage polarization (10). Recent meta-analysis has shown that high expression of VEGF in NPC was associated with inferior survival, and higher plasma levels of VEGF confounds the overall survival of patients (11). The VEGF/VEGFR-2 axis can be inhibited by small-molecule inhibitors like sunitinib, which target VEGFR 1–3, PDGFRα/β, stem cell factor receptor, Flt3, or with the humanized mAb bevacizumab, which binds to VEGF-A and prevents ligand-receptor engagement. These inhibitors have the potential to modulate the tumor microenvironment positively through maturation of tumor microvasculature thus alleviating hypoxia and enhancing chemotherapy delivery and immune cell trafficking of anticancer activated T cells.

To test the hypothesis that addition of anti-VEGF therapy to chemotherapy is tolerable and active in NPC, we conducted a phase II open-label study using two doses of two different antiangiogenic therapies consisting of either sunitinib or bevacizumab in combination with cisplatin and gemcitabine chemotherapy as systemic therapy in patients with previously untreated locally advanced or metastatic NPC (Supplementary Fig. S1). As full doses of antiangiogenic treatment may excessively trim blood vessels leading to collapse of vasculature that compromises drug delivery, lower doses were initiated 7 days prior to chemotherapy to exploit the vascular normalization window of 3–8 days after VEGF inhibition as demonstrated in a glioblastoma model (12). Metabolic response was used as the primary endpoint of tumor efficacy for several reasons: in locally advanced NPC, nonfunctional imaging may lead to difficulty in interpreting tumor response due to overlap with normal structures, and prior studies have shown that PET CT responses have been associated with clinical outcome (13). Correlative studies of serial tumor biopsies were performed to evaluate the pharmacodynamic effects of VEGF inhibition in the tumor microenvironment prior to chemotherapy.

Eligible patients were at least 21 years old with American Joint Committee on Cancer 2010 stage III–IVC previously untreated histologically confirmed EBV-encoded RNA (EBER)-positive WHO type II or III NPC; Eastern Cooperative Oncology Group performance status of 0 or 1; adequate bone marrow, renal, and hepatic functions; and evaluable disease as per RECIST (version 1.1). Patients were excluded if they had uncontrolled hypertension, New York Heart Association grade ≥2 congestive heart failure, active coronary artery disease, arterial or venous thromboembolic events in the last 6 months, clinically significant bleeding in the last 30 days, or persistent proteinuria > 3.5 g/24 hours. Written informed consent was obtained before any study-related procedures, and the study protocol was approved by the ethics committee of the institution and was conducted in compliance with ethical guidelines in the Declaration of Helsinki. (Clinicaltrials.gov ID: NCT01309633).

This was an open-label, phase II trial of cisplatin and gemcitabine with four cohorts: two doses of sunitinib (arm A and B) and bevacizumab (arm C and D). Because of logistic reasons, the trial started with the sunitinib (arm A and B) cohorts and bevacizumab (Arms C and D) cohorts were recruited after completion of Arms A and B. All patients received cisplatin (75 mg/m² on day 1) and gemcitabine (1 g/m² on days 1 and 8) every 3 weeks. Patients were randomly assigned 1:1 to oral sunitinib 12.5 mg (arm A) or 25 mg (arm B) daily for 1 week, starting 1 week before each cycle of cisplatin and gemcitabine. Arm C and arm D patients received bevacizumab 7.5 mg/kg and 2.5 mg/kg, respectively, given every 3 weeks, starting 1 week prior to each cycle of the same chemotherapy. Patients with stages III–IVB NPC received three cycles of the study treatment as induction chemotherapy prior to concurrent chemo-radiotherapy. Patients with metastatic NPC received up to six cycles of the treatment until disease progression or intolerable toxic effects.

Prespecified treatment dose modifications were allowed after the occurrence and resolution of predefined grade 3–4 hematologic or grade 2–4 nonhematologic toxic effects. Radiotherapy employed simultaneous boost intensity-modulated technique (IMRT; Supplementary Methods).

Baseline tumor assessments were performed within 4 weeks before the start of treatment. ¹⁸FDG-PET-CT scan was repeated at end of three cycles of induction chemotherapy for patients with stage III–IVB, after the third and last cycle of systemic treatment for patients with metastatic NPC; and the best metabolic responses for each patient were analyzed. ¹⁸FDG-PET-CT scan was performed using a standard protocol (Supplementary Methods). The sum of SUVmax of the primary nasopharyngeal lesion and lymph nodes and any distant metastatic sites for each scan was compared between the baseline and the postinduction chemotherapy scans. Complete metabolic response (mCR) was defined by the resolution of all metabolically active sites to background levels of the relevant surrounding tissue. Plasma sampling for Epstein–Barr virus (EBV) DNA to evaluate tumor burden and soluble VEGF were collected at baseline, after 1 week of sunitinib/bevacizumab, on cycle 2 day 1, on cycle 3 day 1, and 28 days postinduction therapy [end of induction therapy treatment (EOT)]. Plasma EBV DNA was based on quantification of BamHI-W DNA fragments via an amplicon-based next-generation sequencing method at Lucence Diagnostics (14). VEGF-A was analyzed using dual oligonucleotide-tagged antibody binding and quantitative PCR (Olink Proteomics, https://www.olink.com). All patients who received at least one dose of chemotherapy were evaluated for toxicity every cycle and adverse events were graded according to the NCI Common Terminology Criteria for Adverse Events, version 4.0.

The primary endpoint of the study was objective tumor response (metabolic response and RECIST 1.1) to induction chemotherapy. Secondary endpoints include tolerability to study treatment as graded by the percentage of patients experiencing toxicities, toxicities leading to dose delays or discontinuation, tumor response after definitive concurrent chemotherapy and radiotherapy, and relapse-free survival (RFS) in radiotherapy-treated locally advanced NPC. RFS is defined as the time from assignment to disease progression or death from any cause, whichever occurred first. Nasopharyngeal biopsies were taken at baseline and 7 days after initiating the first cycle of sunitinib or bevacizumab and IHC studies were performed by two research pathologists blinded to the clinical data. The effect of VEGF inhibition on tumor microvasculature was measured using IHC and comparing the endothelial cells and pericyte coverage and expressing changes in microvessel densities. Changes in immune infiltration into the tumor microenvironment were determined by scoring for inflammatory and immune infiltrates (Supplementary Methods).

A primary endpoint of the study for sample size calculation was mCR to induction chemotherapy. This was estimated to be around 5% for the worst study arm and 30% for the best study arm. Sample size calculation utilized the randomized phase II design based on statistical selection theory. With four study arms, 19 patients per study arm would yield the probability of 0.91 of selecting the schedule that has a true mCR rate of 30%. Time-to-event outcomes were estimated using the Kaplan–Meier method. Comparisons of metabolic responses between the four arms were done first using Fisher exact test of independence. For correlative tissue analysis, one-tailed Wilcoxon signed rank was used to establish the significance of any changes observed. All statistical analyses were performed using either IBM SPSS statistics version 25 or GraphPad Prism version 6.

Between May 2012 and January 2019, 77 patients were enrolled in the study; 15 to arm A, 18 to arm B, 20 to arm C, and 24 to arm D. One patient in arm C was ineligible and excluded from analyses because he developed new cardiac arrhythmia prior to study initiation (Fig. 1).

Patient and disease characteristics at baseline were similar between treatment groups (Table 1). Twenty-nine patients (37%) had preexisting arterial hypertension and 7 patients (9%) had coronary artery disease.

Among 61 patients with locally advanced (stage III–IVB) NPC (Table 1), 55 patients (83.3%) received all three cycles of induction chemotherapy: 11 (100%) in arm A, 15 (100%) in arm B, 15 (93.8%) in arm C, and 14 (73.7%) in arm D. Discontinuation of chemotherapy was mostly due to toxicities (n = 6; two in arm B, one in arm C, and three in arm D) and patients' withdrawal of consent midway of the study (n = 5; one in arm C and four in arm D). One patient with metastatic NPC each in arm B and D did not receive the intended six cycles of chemotherapy due to disease progression. Dose modification due to adverse events occurred in 3 (20%) patients in arm A, 4 (22%) in arm B, 3 (16%) in arm C, and 3 (13%) in arm D, and treatment dose was delayed in 7 (47%) patients in arm A, 9 (50%) in arm B, 11 (58%) in arm C, and 14 (58%) in arm D (Fig. 1).

Myelosuppression was the commonest toxicity with a higher incidence of leukopaenia (84.8% vs. 55.8%, P = 0.007) and thrombocytopaenia (90.9% vs. 18.6%, P < 0.001) in the arms containing sunitinib than those receiving bevacizumab (Table 2). Notably, no patient had clinically significant tumoral bleeding and all serial biopsies following antiangiogenic therapy were uneventful. There was no treatment-related death.

A comparison of mCR rates showed significant differences between the four arms (P = 0.001). Furthermore, pairwise comparisons with Bonferroni correction (P < 0.0083 considered significant) showed superior response rates in arm C (42% mCR) compared with arm A (0%, P = 0.001) and arm B (5.6%, P = 0.008) but similar to arm D (12.5%, P = 0.017) (Tables 3A and 3B). The waterfall plots of metabolic responses and tumor responses in the four study arms, respectively, are shown in Fig. 2A and B. An example of mCR from induction therapy of a patient in arm C is shown in Supplementary Fig. S2.

Sixty-one patients (11 patients in arm A, 15 in arm B, 16 in arm C, and 19 in arm D) had locally advanced disease and received chemoradiation with curative intent, and were followed up for RFS. The median follow-up for patients in arms A and B was 53.2 months and for arms C and D was 22.4 months.

Overall, 20 of 61 patients (32.8%) with stage III–IVB NPC had disease progression. The 3-year RFS rates were 64% [confidence interval (CI), 36–92], 60% (CI, 35–86), and 88% (CI, 72–104) for arms A, B, and C, respectively (Supplementary Fig. S3). The 2-year locoregional failure occurred in 3 (27.3%) patients in arm A, 2 (13.3%) in arm B, 2 (12.5%) in arm C, and 2 (10.5%) in arm D, respectively. The 2-year distant failure occurred in 2 (18.2%) patients in arm A, 4 (26.7%) in arm B, 2 (12.5%) in arm C, and 3 (15.7%) in arm D, respectively.

To validate the clinical utility of metabolic response, RFS of patients treated with chemoradiation was analyzed according to the depth of metabolic response (80% or more vs. less than 80%). Data showed a higher RFS for the better responders who underwent chemoIMRT: 5-year PFS was 80.2% versus 42.2%; log-rank test P = 0.016 (Supplementary Fig. S4). Comparison of the depth of metabolic response showed that 47% patients in arm A, 28% in arm B, 62.5% in arm C, and 33% in arm D experienced 80% or more reduction in the sum of SUVmax in lesions.

A total of 59 plasma samples were analyzed for baseline EBV DNA levels: the mean and SD were 14,480 ± 28,383 IU/mL in arm A, 6,518 ± 16,157 IU/mL in arm B, 13,615 ± 27,722 IU/mL in arm C, and 5,883 ± 10,382 IU/mL in arm D. A total of 57 patients had plasma EBV DNA viral load remeasured at day 1 of cycle 3 the mean and SD were 560 ± 1,569 IU/mL in arm A, 229 ± 315 IU/mL in arm B, 65 ± 135 IU/mL in arm C, and 1,115 ± 3,416 IU/mL in arm D. It reached undetectable levels (defined as less than 50 IU/mL of EBV DNA) in 7/12 (58.3%), 3/14 (21.4%), 13/18 (72.2%), and 8/13 (61.5%) of patients in arms A, B, C, and D, respectively (Supplementary Fig. S5A).

A total of 9/11 in arm A, 14/14 in arm B, 18/18 in arm C, and 15/15 in arm D showed above normal circulating VEGF levels in their baseline samples. In patients receiving sunitinib (arms A and B), the VEGF levels remained high at the beginning of cycle 1 (C1D1) and in 4/8 (50%) of patients of arm A and 6/9 (66%) of arm B at EOT. In contrast, in patients receiving 7.5 mg/kg of bevacizumab (arm C), circulating VEGF levels in 17/18 (94%) of the cohort fell to within normal limits within 7 days of bevacizumab treatment (C1D1) and 15/15 (100%) remained within normal limits at EOT (Supplementary Fig. S5B).

Enumeration of mature pericyte covered microvasculature showed increase in mature vessels post anti-VEGF treatment in arms A and C (Fig. 3A; Supplementary Fig. S6A) without change in microvessel density with all four arms (Supplementary Fig. S6B).

Morphologic examination of the hematoxylin and eosin sections suggested an increase in immune and inflammatory cell infiltrate into the primary lesion in response to 7.5 mg/kg of bevacizumab treatment (Fig. 3B; Supplementary Fig. S7) prior to start of chemotherapy, indicating better chemotaxis and trafficking of immune cells into the tumor microenvironment. The predominant cellular infiltrate were lymphocytes comprising CD4 helper and CD8 cytotoxic T cells. Macrophages, neutrophils, and sometimes eosinophils were also present albeit in much lower numbers compared with the lymphocytic cells. To confirm this observation, the number of leukocytes in representative arm C samples (n = 6) were quantified using Image J (NIH) Software with the Colour Deconvolution Plugin. The median leukocyte density and range for the baseline and C1D1 samples were 6,501 cells/mm² (range: 2,622–8,999 cells/mm²) and 9,695 cells/mm² (range: 6,976–15,796 cells/mm²), respectively (Fig. 3B).

VEGF is highly expressed in NPC through oncogenic EBV latent membrane protein 1 (15), and expression correlates with metastases and prognosis (16), inspiring much interest in developing anti-VEGF therapy for NPC. Previous phase II studies in recurrent/metastatic NPC with anti-VEGF multikinase inhibitors yielded modest activity, with added risk of catastrophic hemorrhage from previously irradiated head and neck region (17, 18). As first-line therapy, bevacizumab combined with standard cisplatin-based chemoradiation was shown to be tolerable and achieved favorable disease control in RTOG 0615 study, raising the potential of incorporating antiangiogenic therapy with chemotherapy (19). Our study supports the concept of normalization of microvasculature with antiangiogenic therapy, which may improve chemotherapy delivery and immune infiltration that is associated with favorable prognosis (20). The antiangiogenic dosage used in all arms of the study did not result in excess of tumor hemorrhage, allowed serial biopsies to be done safely, and did not compromise definitive chemoradiation.

Among the four study arms with two different doses of sunitinib and bevacizumab, the arm with bevacizumab at 7.5 mg/kg was superior in tolerability, antitumor, plasma EBV, and VEGF responses to induction therapy. The 3-year RFS of 88% for predominantly stage IVA/B patients with locally advanced disease in this arm compares favorably with concurrent chemoradiation, where the 3-year RFS was ranged from 66% to 73% (21–25).

Induction chemotherapy is known to have a profound effect on NPC and the response rates in all four arms of our study were above 80%, which is consistent with effectiveness of the cisplatin and gemcitabine combination. Previously, investigators have shown that PET CT scan after induction chemotherapy may be a good early indicator of clinical outcome as metabolic responses observed during induction chemotherapy precedes anatomic responses on MRI (13, 26). Our study validated this in patients with locally advanced disease, where RFS was significantly prolonged for patients with 80% or better reduction in sum of SUV_max for the primary and neck lymph nodes. mCR rates were highest in the 7.5 mg/kg bevacizumab arm and translated to favorable RFS in this study arm. It should be noted that there are no current guidelines for defining mCR in NPC, and our criteria was consistent with both EORTC and PERCIST guidelines (27).

The favorable clinical outcome of arm C is supported by observed effects in both plasma biomarkers and tumor microenvironment. Arm C had the greatest suppression of both plasma EBV DNA titre and VEGF levels by the second cycle of treatment. Of note, patients with detectable plasma EBV postinduction chemotherapy and higher VEGF levels have worse prognosis in NPC (28). The IHC studies supported the hypothesis that anti-VEGF therapy promotes microvasculature maturation and enhanced immune infiltration within a week, with bevacizumab 7.5 mg/kg demonstrating both effects. Anti-VEGF treatment has been proposed to improve immune cell trafficking and alleviate hypoxia to activate immune function in tumors, and survival in NPC is associated with abundance of tumor lymphocytic infiltration (20). Though the optimal dose and duration of normalization of vasculature, following anti-VEGF treatment is uncertain, the sunitinib-containing regimen is unlikely to be tolerable at higher doses due to off target effects that cause myelosuppression (29).

Addition of bevacizumab to cisplatin/gemcitabine in metastatic disease potentially adds therapeutic benefit, but the issue of hemorrhagic events needs to be evaluated in the context of prior chemoradiation. The combination of antiangiogenic therapy with PD-1/PDL-1 inhibitors has shown promising tumor responses in cancers such as non–small lung cancers and renal cell cancers (30). The positive effects of antiangiogenic therapy on tumor perfusion and immune cell trafficking into the nasopharyngeal cancer seen in our study encourage new studies combining immunotherapy and antiangiogenic therapy in metastatic nasopharyngeal cancers. For locally advanced NPC, this study together with the data from RTOG 0615 support the need for further studies to establish the role of bevacizumab in the induction chemotherapy and concurrent chemoradiotherapy phases of treatment.

W.-Q. Chong reports nonfinancial support from MSD (conference support) and Amgen (conference support) outside the submitted work. R. Sundar reports other from BMS (honararia for talks/advisory board), Merck (honararia for talks/advisory board), Eisai (honararia for talks/advisory board), Bayer (honararia for talks/advisory board), Taiho (honararia for talks/advisory board), MSD (honararia for talks/advisory board), Eli Lilly (honararia for talks/advisory board), Roche (honararia for talks/advisory board), and Astra Zeneca (honararia for talks/advisory board) and grants from MSD and Paxman Coolers outside the submitted work. A. Jeyasekharan reports grants and personal fees from Janssen and personal fees from AstraZeneca and MSD outside the submitted work. S.-C. Lee reports grants and other from Pfizer (research grant for investigator-initiated studies; honorarium for advisory boards and as invited speaker at conferences), Eisai (research grant for investigator-initiated studies; honorarium for advisory boards and invited speaker at conferences), and Taiho (research grant for investigator-initiated studies) and other from Amgen (travel support to conference), Novartis (honorarium for advisory boards and invited speaker at conferences), Astra Zeneca (honorarium for advisory boards and invited speaker at conferences), Roche (honorarium for advisory boards and invited speaker at conferences), ACT Genomics (honorarium for advisory boards and invited speaker at conferences), MSD (honorarium for advisory board), and Eli Lilly (honorarium for advisory board) outside the submitted work. B.-C. Tai reports other from Boehringer Ingelheim Singapore Pte Ltd (conduct workshop), Boehringer Ingelheim (Malaysia) Sdn Bhd (conduct workshop), and Wilely-Blackwell (royalty) outside the submitted work. B.-C.Goh reports grants from National Medical Research Council, Singapore during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

W.Q. Chong: Data curation, formal analysis, investigation, writing-original draft, writing-review and editing. C.M. Lim: Investigation. A.K. Sinha: Investigation, methodology. C.S. Tan: Data curation, investigation, methodology, project administration. G.H.J. Chan: Investigation. Y. Huang: Investigation. N. Barr Kumarakulasinghe: Investigation. R. Sundar: Investigation. A.D. Jeyasekharan: Investigation. W.-S. Loh: Investigation. J.K. Tay: Investigation. K. Yadav: Investigation. L. Wang: Investigation, methodology, project administration. A.L. Wong: Investigation, methodology, project administration. L.R. Kong: Investigation. R.A. Soo: Investigation, writing-review and editing. J.A. Lau: Investigation. Y.Y. Soon: Formal analysis, investigation. R.M. Goh: Data curation. F.C.H. Ho: Investigation. S.M. Chong: Formal analysis, investigation, methodology. S.-C. Lee: Conceptualization, investigation. K.S. Loh: Investigation. B.C. Tai: Formal analysis, methodology. Y.C. Lim: Formal analysis, investigation, methodology, writing-original draft, project administration, writing-review and editing. B.C. Goh: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing-original draft, project administration, writing-review and editing.

This study is supported by Clinical Scientist Award awarded by National Medical Research Council of Singapore to B.-C. Goh (NMRC/CSA-SI/0006/2016 and NMRC/CG/012/2013).

We like to thank all the patients and their families who participated in this trial. We also wish to express our gratitude and respect to our collaborator and colleague, the late S.-M. Chong for his contributions including development of methodology and analysis of data in the project. S.-M. Chong passed away in August 2019.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.