Efficacy, Safety, and Correlative Biomarkers of Toripalimab in Previously Treated Recurrent or Metastatic Nasopharyngeal Carcinoma: A Phase II Clinical Trial (POLARIS-02)

PURPOSE As yet, no checkpoint inhibitor has been approved to treat nasopharyngeal carcinoma (NPC). This study was aimed to evaluate the antitumor activity, safety, and biomarkers of toripalimab, a new programmed death-1 (PD-1) inhibitor for recurrent or metastatic NPC (RM-NPC) refractory to standard chemotherapy. PATIENTS AND METHODS In this single-arm, multicenter phase II study, patients with RM-NPC received 3 mg/kg toripalimab once every 2 weeks via intravenous infusion until confirmed disease progression or unacceptable toxicity. The primary end point was objective response rate (ORR). The secondary end points included safety, duration of response (DOR), progression-free survival (PFS), and overall survival (OS). RESULTS Among all 190 patients, the ORR was 20.5% with median DOR 12.8 months, median PFS 1.9 months, and median OS 17.4 months. Among 92 patients who failed at least two lines of systemic chemotherapy, the ORR was 23.9%. The ORRs were 27.1% and 19.4% in PD-L1+ and PD-L1− patients, respectively (P = .31). Patients with ≥ 50% decrease of plasma Epstein-Barr virus (EBV) DNA copy number on day 28 had significantly better ORR than those with < 50% decrease, 48.3% versus 5.7% (P = .0001). Tumor mutational burden had a median value of 0.95 muts/mega-base in the cohort and had no predictive value for response. Whole-exome sequencing results from 174 patients revealed that the patients with genomic amplification in 11q13 region or ETV6 genomic alterations had poor responses to toripalimab. CONCLUSION The POLARIS-02 study demonstrated a manageable safety profile and durable clinical response of toripalimab in patients with chemorefractory metastatic NPC. An early decrease in plasma EBV DNA copy number correlated with favorable response.


INTRODUCTION
Nasopharyngeal carcinoma (NPC) is distinctive from other head and neck cancers with unique geographical, etiological, and biological features. 1 It is distributed predominantly in South China, Southeast Asia, Middle East, and North Africa. [1][2][3] Nonkeratinizing NPC is the most common subtype in both North America and Southeast Asia, 4,5 which is more closely associated with Epstein-Barr virus (EBV) infection, more responsive to radiation therapy, and has better survival than keratinizing NPC. 5 Patients with recurrent or metastatic NPC (RM-NPC) have poor prognosis, with median overall survival (OS) ,20 months. 6 Platinum doublet, like cisplatin plus gemcitabine, is the standard first-line treatment for RM-NPC. 7 However, there is no standard treatment option in second-line settings and beyond. 6 EBV infection is instrumental in NPC development. 8 EBVinduced NPCs often present with intensive lymphocyte infiltration and overexpression of programmed death ligand-1 (PD-L1), 9 indicating the potential application of programmed death-1 (PD-1) blockade immunotherapy. Several small size phase II studies of anti-PD-1 immune checkpoint inhibitors (ICIs) in patients with RM-NPC showed objective response rate (ORR) from 20.5% to 34.1%. [10][11][12] However, predictive biomarkers for ICI therapy in NPC had not been established. [10][11][12] Toripalimab, a humanized IgG 4 monoclonal antibody against PD-1, 13 was first approved in December 2018 for the second-line treatment of advanced melanoma in China. 14 Here, we report the result of phase II pivotal study (POLARIS-02) evaluating the efficacy and safety of toripalimab in previously treated RM-NPC. It was to date the largest prospective study of anti-PD-1 monotherapy in chemorefractory NPC. Potential efficacy predictors were explored to identify the NPC population most likely to respond to ICI therapy.

Patients and Study Design
This study is the NPC cohort from a phase Ib/II, multicohort trial (ClinicalTrials.gov identifier: NCT02915432) evaluating the safety and clinical activity of toripalimab in patients with RM-NPC, head and neck cancer, gastric cancer, and esophageal cancer. The study Protocol was approved by institutional ethics committees of all participating centers. This study was conducted in accordance with the Declaration of Helsinki and the international standards of good clinical practice.
Key eligibility criteria included histologically or cytologically documented RM-NPC refractory to prior standard chemotherapy, or disease progression within 6 months after adjuvant chemotherapy or chemoradiotherapy, age 18 years or older, measurable disease, Eastern Cooperative Oncology Group performance status of 0 or 1, and adequate organ function. Key exclusion criteria included anticancer monoclonal antibody therapy within 4 weeks before treatment initiation, any anticancer therapy within 2 weeks before treatment initiation, prior ICI treatment, systemic corticosteroid therapy within 7 days before treatment initiation, known additional malignancies, and active CNS metastases.
The definition of patient population with at least two prior lines (2L1) of systemic chemotherapy included: (1) received at least two lines of systemic chemotherapy; (2) first-line chemotherapy must include platinum-based regimen; (3) neoadjuvant, adjuvant, or concurrent chemoradiotherapy were considered as a line of systemic treatment if tumor recurrence or metastasis occurred within 6 months after the end of neoadjuvant, adjuvant, or concurrent chemotherapy; (4) stage IVb at enrollment as defined by Union for International Cancer Control and American Joint Committee on Cancer staging system for NPC, seventh edition; (5) clear evidence of disease progression to prior therapy at enrollment. Intolerance to chemotherapy should not be counted as a line of systemic treatment.

Treatment and Assessments
Patients received toripalimab 3 mg/kg once every 2 weeks via intravenous infusion until disease progression, intolerable toxicity, or voluntary withdrawal of informed consent. Patients experiencing initial disease progression could continue to receive toripalimab under the condition of potential benefit to the patients, with the consensus of investigator and sponsor.
Tumor response was assessed according to RECIST v1.1 and immune-related RECIST by independent radiologic review committee (IRC) and investigators. Adverse events were graded according to National Cancer Institute Common Terminology Criteria version 4.0.

Study End Points
The primary end point was ORR determined by IRC according to RECIST v1.1. The secondary end points included safety, duration of response (DOR), disease control rate (DCR), progression-free survival (PFS), and OS. Exploratory end points included PD-L1 expression, plasma EBV DNA copy number, tumor mutational burden (TMB), and genetic biomarkers by whole-exome sequencing (WES) as potential efficacy predictors.
PD-L1 expression was determined by immunohistochemistry (IHC) staining with SP142 antibody 15 in a central lab and evaluated by certified pathologists. PD-L1 expression on tumor cells (TCs, defined as tumor proportion score) and immune cells (ICs) were evaluated.
Plasma EBV DNA copy number was determined by the quantitative reverse transcription polymerase chain reaction (qRT-PCR) method with probes against EBV genes before treatment and every 4 weeks until disease progression.
WES was performed using SureSelect Human All Exon V6 kit (Agilent, Santa Clara, CA) on tumor and matched blood samples. Genomic alterations including microsatellite stability status, nucleotide variants, short and long insertions and deletions (INDELs), copy number variants, and gene rearrangement and fusions were assessed. TMB was determined by analyzing somatic mutations including coding base substitutions and INDELs according to megabase (Mb).

Sample Size Determination and Statistical Analysis
The sample size of the original NPC cohort was estimated according to Simon's two-stage design. In stage I, among 34 evaluable patients with NPC, if the number of responders was ,2, the study cohort would be terminated. Otherwise, an additional 14 patients would be enrolled. Considering a dropout rate of 10%, a total of 54 patients with NPC would be enrolled. The study Protocol was amended to increase the NPC cohort size to 100 after a promising result (26.5% ORR) among the first 34 patients at stage I. The sample size of the NPC cohort was further amended to 180 patients after a type-B meeting with the National Medical Product Administration (NMPA) of China for registration purpose and was based on the following assumptions. At a one-sided significance level of 0.025, 160 patients could provide at least 80% power to show the efficacy of toripalimab at targeted ORR of 24% versus 15% for alternative third-line therapy using Clopper-Pearson method. As an estimated 20 patients with one prior line of systemic treatment had been enrolled in the cohort, a total of 180 patients would be enrolled.
Safety analysis included all patients with at least one dose of study drug. ORR and its 95% exact CI were determined by Clopper and Pearson methodology. Fisher's exact test was used to compute two-tailed P values from contingency tables. PFS and OS were plotted using the Kaplan-Meier method, with median and corresponding two-sided 95% CI. Statistics analyses were performed using SAS version 9.4 or GraphPad Prism software.  Table 1. According to histology subtypes (WHO Classification of Tumors), 182 patients (95.8%) were nonkeratinizing and 8 patients (4.2%) were keratinizing NPC.

Safety
All enrolled patients received a median of eight doses of toripalimab (range, 1-69 doses). Permanent discontinuation of toripalimab as a result of treatment-related adverse event (TRAE) occurred in four patients, and dose interruption as a result of TRAE occurred in seven patients.
Subgroup analysis of both intent-to-treat population and 2L1 population showed that female patients, patients with age $ 60 years, lactate dehydrogenase # 23 upper limit of normal, or without liver metastasis had numerically higher ORR (Appendix Tables A2 and A3, online only). None of the differences, however, were statistically significant.
Interestingly, compared with nonkeratinizing NPC (n 5 182), keratinizing NPC (n 5 8) had an ORR of 62.5%. However, given the small sample size, further study is warranted in a larger cohort of keratinizing NPC.
Dynamic monitoring of plasma EBV DNA copy number was performed during the study, and results were available from 148 patients. Patients with objective responses had a significant greater EBV titer decrease from baseline to day 28 than patients with stable disease or progressive disease (Fig 2). Furthermore, patients with $ 50% EBV titer decrease on day 28 (n 5 60) had significantly higher ORR than those with ,50% decrease (n 5 88) (48.3% v 5.7%, P 5 .0001). Notably, 14 patients who responded to toripalimab and later experienced disease progression had at least 100% increase of plasma EBV titer occurring at a median of 3 months before radiographic identification of disease progression.

DISCUSSION
NPC is distinct in its anatomic location and biology from other epithelial tumors in the head and neck region. No ICI has been approved to treat NPC. Phase III clinical trials which led to the approval of pembrolizumab or nivolumab for head and neck squamous cell carcinoma (HNSCC) specifically excluded NPC. 17,18 Standard treatment for patients with RM-NPC beyond the first-line setting has not  6 Several studies had shown promising activities of anti-PD-1 monotherapy for chemorefractory NPCs. Among 44 unselected advanced patients with NPC treated with nivolumab, the ORR was 20.5%. 10 In KEYNOTE-028 study, 19 patients with PD-L11 NPC treated with pembrolizumab had an ORR of 26.3%. 11 Among 91 unselected advanced patients with NPC treated with camrelizumab, the ORR was 34.1%. 12 In this study, with the largest NPC cohort treated with anti-PD-1 monotherapy to date, the ORR of toripalimab for all 190 patients was 20.5% (95% CI, 15.0 to 27.0), the response was durable with median DOR of 12.8 months and the median OS was 17.4 months. The definition of patient population with at least two prior lines of systemic chemotherapy (2L1) was clarified and redefined retrospectively by the Chinese regulatory agency, the NMPA, and 92 patients qualified as 2L1. Among 2L1 patients, toripalimab had an ORR of 23.9% (95% CI, 15.6 to 33.9), with a median DOR of 21.5 months and a favorable median OS of 15.1 months.
There are three histopathological types of NPC by WHO classification, keratinizing squamous cell carcinoma, nonkeratinizing carcinoma, and basaloid squamous cell carcinoma. Nonkeratinizing subtype is more common in  Southeast Asia and closely associated with EBV infection, whereas keratinizing subtype is associated with environmental factors, such as smoking and alcohol intake. 19 The frequency of keratinizing subtype is about 2% in Southeast Asia, whereas it accounts for about 25% of all NPC in North America. 4,5 Keratinizing NPC (n 5 8) had a 62.5% ORR in this study. Although it is interesting, considering the small sample size and possible selection bias, further study in a larger cohort of keratinizing NPC is warranted.
Virus DNA has been recognized as a prognostic biomarker in several virus-mediated malignancies. Latent infection with EBV is crucial to the development of NPC. 20 Measurement of plasma EBV DNA copy number has been recommended for NPC diagnostic and prognostic  monitoring. [21][22][23] The dynamic change of EBV titer is closely related to response to chemotherapy or radiotherapy. 24,25 Similarly, plasma HPV DNA was indicated as a promising biomarker for monitoring chemoradiotherapy response in patients with HPV-associated oropharyngeal squamous cell carcinoma. 26 However, the value of circulating viral DNA titer for immunotherapy has not been comprehensively investigated. In the current study, we found that patients with baseline EBV titer ,10,000 IU/mL had numerically higher ORR than those with EBV titer $ 10,000 IU/mL (54.5% v 39.5%, P 5 .088). Notably, responding patients (CR or PR) had more significant EBV titer reduction from baseline to day 28 than patients with stable disease or progressive disease (Fig 2). Furthermore, patients with $ 50% decrease of EBV titer on day 28 had significantly better ORR than those with ,50% decrease (ORR 48.3% v 5.7%, P 5 .0001). The results supported the positive association of plasma EBV DNA copy number reduction with improved disease control in response to immunotherapy.
PD-L1 expression status correlated with favorable response to immunotherapy in various cancers, but its value for NPC was unclear. Consistent with nivolumab in advanced NPC, 10 our study observed a numerically higher but not statistically significant ORR of PD-L1 TC1 patients than PD-L1 TC2 patients (27.1% v 19.4%). Although PD-L1 staining by SP142 on IC was used in other indications, 27 we found that PD-L1 IC expression had no apparent association with clinical efficacy in our study, suggesting different PD-L1 expression patterns and applications to predict clinical response in different cancer types. The high heterogeneity of PD-L1 expression in NPC 28-30 may also limit its value as a robust biomarker.
We also evaluated the predictive value of tumor TMB for RM-NPC in response to ICI therapy. Our study discovered that with a cutoff value of 2.9 muts/Mb (top 10% TMB value), TMB has no predictive value for clinical response. The median TMB value was 0.95 muts/Mb in the cohort, which was consistent with previous reports. 31,32 It was suggested that rather than the accumulation of DNA mutations, EBV infection might play a more important role in driving NPC tumorigenesis. 33 Consistent with our study, the results of KEYNOTE-012 study and KEYNOTE-055 study also found that TMB was less predictive for HNSCC patients with positive HPV or EBV, whereas high TMB correlated with a better response for HNSCC patients with negative HPV or EBV. 34,35 WES was performed to identify genomic biomarkers for immunotherapy in patients with RM-NPC. The frequently altered genes were consistent with previous reports. 36,37 Patients with genomic amplification in 11q13 region had no response and those with ETV6 genomic alterations had an ORR of 5.3%, indicating potential resistance mechanisms to immunotherapy. Genomic amplification in 11q13 region was previously reported to associate with poor response to toripalimab in melanoma 14 and esophageal squamous cell carcinoma. 38 Overexpression of CCND1, FGF, and ETV6 genes has been correlated with poor prognosis in patients with NPC. 39,40 The impact of genomic amplification in 11q13 region and ETV6 on immune function and immunotherapy needs to be further investigated.
In summary, POLARIS-02, the largest prospective clinical study of anti-PD-1 monotherapy in patients with chemorefractory RM-NPC, showed a manageable safety profile. The response rate and, more importantly, the DOR have demonstrated that toripalimab provides substantial benefit to patients with metastatic NPC who are receiving later-line therapy. This study also provided new perspectives on predictive biomarkers in NPC. Plasma EBV DNA copy number change might serve as a feasible predictor for clinical efficacy. A confirmatory phase III trial comparing toripalimab versus placebo in combination with cisplatin plus gemcitabine as first-line treatment for RM-NPC is ongoing (ClinicalTrials.gov identifier: NCT03430297). . T-test was used to determine statistical significance between two groups. Significant differences were observed between CR 1 PR and stable disease groups and CR 1 PR and progressive disease groups. Fold reduction 5 (baseline EBV tier)/(day 28 EBV titer). CR, complete response; EBV, Epstein-Barr virus; PR, partial response.