Nasopharyngeal carcinoma (NPC) is one of the most common head and neck cancers in Southeast Asia and North Africa. The age-standardized incidence ranges from 20 to 50 per 100,000 males in southern China to 0.5 per 100,000 in white populations [1]. Recently, the first phase 3 trial in recurrent or metastatic NPC (RM-NPC), the landmark GEM20110714 study, has established gemcitabine plus cisplatin (GP) regimen as the standard first-line treatment [2]. However, no consensus has been reached beyond the first-line setting, in which the prognosis is extremely poor.

Endemic NPC is etiologically associated with Epstein-Barr virus infection. This virus-associated cancer represents the archetypal “inflamed tumor,” which exhibits a dense lymphocytic infiltrate and increased programmed death-ligand 1 (PD-L1) expression [3, 4]. These features make immunotherapy a promising treatment option for NPC patients. Recently in 2017, the landmark KEYNOTE-028 trial firstly reported promising antitumor activities and safety profiles of pembrolizumab in previously treated RM-NPC [5]. Subsequently, five additional phase 1–2 trials evaluating anti-PD-1 antibodies in RM-NPC were reported [6–9]. The NCI-9742 [6] and CheckMate-385 [7] trials demonstrated a manageable safety profile and clinical activity of nivolumab in multiply pretreated and/or treatment-naive RM-NPC patients. Fang and colleagues [8] reported that camrelizumab monotherapy was a well-tolerated and potentially effective treatment option for previously treated RM-NPC. They further reported that the combination of camrelizumab plus chemotherapy of GP regimen has a manageable toxicity profile and promising preliminary antitumor activity in treatment-naive RM-NPC [8]. Another latest trial, the JS001 study, reported in the European Society for Medical Oncology (ESMO) 2018 conference demonstrated the clinical activity of JS001 in multiply pretreated RM-NPC [9]. However, to date there is no head-to-head comparison of different anti-PD-1 drugs, standard first-line GP chemotherapy, and their combination therapy in RM-NPC. Therefore, we initiated this study to comprehensively compare the safety and efficacy of the abovementioned trials, and explore the optimal therapeutic regimens of anti-PD-1 approach in RM-NPC. We hypothesized that the efficacy and safety profiles differed across different anti-PD-1-based regimens.

The abovementioned anti-PD-1 trials were included in the analysis with GP arm from GEM20110714 trial as chemotherapy control [2]. The major assessed outcomes were adverse event (AE) and objective response rate (ORR). AE and ORR data were pooled up per regimen and described in percentage. The comparative incidences of AE between different regimens were evaluated by the odds ratio (OR) and corresponding 95% confidence interval (CI) using Fisher’s exact test. OR > 1 stands for fewer AEs. When AE rate in any comparative arm equaled 100% or 0%, the Haldane-Anscombe correction was adopted to evaluate OR and its 95% CI [10]. Given that ORR of PD-1 blockade may differ according to treatment lines (first-line vs. >1st line), we also evaluated the anti-PD-1 drugs per treatment setting and considered them as independent comparative groups when data was available. Given the evidence that high PD-L1 expression tended to be associated with favorable responses to PD-1 blockade in NPC [6], we further evaluated the pooled ORR of anti-PD-1 therapies stratified by PD-L1 positivity. Statistical analyses were performed using R version 3.5.1 (http://www.r-project.org). A two-tailed P < 0.05 was considered statistically significant.

This is the first report that compares the safety and efficacy of different anti-PD-1 drugs with/without chemotherapy in RM-NPC, which provides preliminary evidence and integrative insights into the future design and implementation of immune clinical trials in NPC. The general safety of nivolumab and pembrolizumab ranked high, while the incidences of grade 3–5 AEs were relatively low in camrelizumab and nivolumab. Integrating with the specific toxicity spectra of each drug, we postulate that camrelizumab has the least toxicity profile; the high incidence of all-grade AEs may be attributed to reactive capillary hemangiomas, which is generally unthreatening and self-resolved [8]. In general, anti-PD-1 therapy was safer than standard chemotherapy; however, once it was combined with chemotherapy, the incidence of grade 3–5 AEs and AE-related discontinue rate doubled over chemotherapy alone, which suggests potentially synergized toxicity. This is in accordance with the safety profiles in the KEYNOTE-048 trial investigating anti-PD-1 + chemotherapy versus chemotherapy alone in non-nasopharynx head and neck cancer [11].

The response rate with anti-PD-1 monotherapy for pretreated RM-NPC approximated 20–30%, whereas ORR increased to 40% in treatment-naive patients. Additionally, the incidences of AEs with anti-PD-1 regimen were generally lower than those with chemotherapy. Though the sample size in first-line anti-PD-1 group is limited, it provides preliminary evidences that comparing first-line anti-PD-1 monotherapy versus standard chemotherapy is a “trial-worthy” approach.

Interestingly, we observed similar CR rates between GP chemotherapy and anti-PD-1 + chemotherapy. Considering that anti-PD-1 monotherapy only achieved limited CR rate (~ 0–2%), it might suggest that the addition of anti-PD-1 agents may not be adequate enough to increase complete elimination of tumor cells (CR rate), though this combination therapy can still substantially improve tumor killing efficacy (PR rate). Likewise, a phase 2 trial by Chia and colleagues [12] reported a similar CR rate (8.6%) and a relatively higher PR rate (62.9%) in treatment-naive RM-NPC receiving chemotherapy followed by EBV-specific cytotoxic T lymphocytes (CTLs), compared to that of GP chemotherapy (Table 2). These results implicate that the combination of immunotherapy agents (either anti-PD-1 or EBV-CTLs) with chemotherapy may not achieve synergic effects (similar CR rates) in NPC patients; however, the substantially increased PR rate may still translate into patient survival benefits. Our findings need to be verified in ongoing/future prospective randomized trials evaluating the combination of immunotherapy and chemotherapy versus chemotherapy alone.Tab2

Summary of characteristics and clinical endpoints of the Singapore trial evaluating the combination of chemotherapy and adoptive T-cell therapy in NPC

One major challenge of immunotherapy remains that it only benefits small subsets of patients. Prior trial data [6, 9] showed a numerically higher ORR in patients with PD-L1–positive RM-NPC than in those with PD-L1–negative tumors; and high-PD-L1 expression was associated with better survival outcomes in both NPC and other head and neck cancer [13, 14]. We further performed pooled analysis to increase statistical power. Unfortunately, a significant margin was still not reached; the possible reason would be limited sample size. Moreover, PD-L1 expression alone may not be the only determinant of treatment benefits; it might be contingent on other factors in the tumor microenvironment that are yet to be identified. Future studies are warranted to identify reliable biomarkers for tailoring anti-PD-1 therapies. Improvement in treatment efficacy can also be achieved by breakthroughs in combination therapy. It is shown that camrelizumab+GP combination approach achieved a remarkable ORR in first-line treatment. Additionally, previous exposure to ipilimumab significantly improved antitumor activity of camrelizumab [8]. Going forward, these preliminary findings construct a road-map for the design of future trials to assess the efficacy of immuno-oncology cocktail and/or dual inhibition of immune checkpoints approaches in NPC. Results from relevant ongoing trials (e.g. NCT03581786, NCT03707509, NCT03097939) are eagerly awaited.

One major limitation of this study is that all trials were in phase 1/2, therefore long-term survival data are still lacking and the sample sizes were limited, especially for first-line nivolumab (n = 5) and the combination therapy group (n = 23). Our findings need to be verified in future large-scale, head-to-head, phase 3 trials. Secondly, the results of subgroup analyses regarding PD-L1 expression level should be interpreted with cautions, in view of the different immunohistochemical assays used [5, 6, 9].

Our study comprehensively compares the safety profile and efficacy of anti-PD-1 monotherapy, chemotherapy and their combination in RM-NPC, which provides important evidence for the design of future trials and clinical management with respect to anti-PD-1 therapy.