Background
Nasopharyngeal carcinoma (NPC) is a malignancy with remarkable difference in region distributions, which is of high incidence in the Southeast Asia, the Arctic region, and the North Africa, especially in the Guangdong Province in Southern China [
1]. In accordance with the coming era of intensity-modulated radiation therapy (IMRT) and increasing applications of potent chemotherapy, the overall survival and tumor local control rate of patients with nasopharyngeal carcinoma have been considerably improved [
2,
3]. Despite the best available treatment, local recurrence and distant metastasis remain to be the main reason for failure after NPC treatment, approximately 5–15% and 15–30%, respectively [
4]. Recently, increasing numbers of clinical trials concerning immunotherapy have shown promising effects on patients with nasopharyngeal carcinoma [
5,
6]. These studies indicate that the mechanisms of immune evasion exert an enormous function on the pathogenesis of NPC. For these reasons, it would be invaluable for optimizing the treatment in NPC patients that if we could elucidate the relationship between the molecules in immune system and NPC.
As is well known, tumor cells can display immune evasion by activating immune checkpoint molecules. Programmed cell death ligand-1 (PD-L1), an immunoinhibitory molecule, has the function of inducing T-cell-mediated immune tolerance, including anergy and apoptosis, by activating programmed cell death-1 (PD-1) located on the surface of T cells [
7]. Therefore, immune therapies targeting PD-1/PD-L1 axis have shown significant anti-tumor effect in some types of solid tumors, including melanoma, non-small-cell lung cancer, and head and neck carcinomas [
8].
PD-L1 has been proven to be overexpressed in many types of cancer cells and associated with different clinical outcomes, either better or worse, depending on the categories of tumor [
9‐
12]. Also, there are some controversial about the prognostic value of PD-L1 expression on particular types of cancer, such as breast cancer [
13,
14], and nasopharyngeal carcinoma [
15‐
17]. However, there is relatively fewer studies towards the prognostic value of PD-1 expression in patients with cancer. A meta-analysis study demonstrated that the positive expression of PD-1 in TILs correlated with poorer overall survival in patients with epithelial-originated cancer, while the study did not include any data about nasopharyngeal carcinoma [
18].
Since the prognostic value of PD-L1 or PD-1 expression in nasopharyngeal carcinoma remains unclear, the aim of the present study was to incorporate all available data using the method of meta-analysis to explore whether different expression status of PD-L1 or PD-1 in patients suffering for nasopharyngeal carcinoma have an effect on their survivals. Also, the correlation between clinicopathological features in patients with NPC and PD-1/PD-L1 expression has been evaluated.
Methods
Protocol and registration
This systematic review has been conducted following the “Preferred Reporting Items for Systematic Reviews and Meta-Analyses” (PRISMA) guidelines [
19] and the Cochrane Handbook [
20]. Also, this review was registered ahead on the online database of International prospective register of systematic reviews (PROSPERO) with the Registration Number of CRD42018109532.
Search strategy
We performed the literature research in the PubMed, Web of Science, EMBASE, and the Cochrane Library databases for all relevant original articles up to January 06, 2019. The comprehensive search strategies were based on the combinations of the following key words: “programmed cell death-ligand 1, PD-L1, CD274, B7-H1, programmed cell death 1, PD-1, CD279” AND “nasopharyngeal, nasopharynx” AND “carcinoma, tumor, cancer”, with language restricted in English. In addition to the search online, manual search was performed on the reference lists of retrieved articles as well to broaden the search. When it occurs that multiple articles include the same cohort of NPC patients, multiple aspect of evaluation would be carried out in order to decide the final included report.
Inclusion and exclusion criteria
The eligible criteria were the following: cohort studies were conducted in human with NPC and the diagnosis of NPC was confirmed by pathology; the expression status of PD-1 or PD-L1 was detected: the correlation between PD-1/PD-L1 and overall survival (OS), disease-free survival (DFS), progression-free survival (PFS), or distant-metastasis free survival (DMFS) were elucidated by hazard ratio (HR) and its 95% confidence intervals (95% CI) or Kaplan–Meier curve [
21]. Studies were excluded if they met either of the following exclusion criteria: review or basic research; conference abstract or letters; as well as case report or clinical trials.
Data extraction and quality assessment
Two investigators (Zi-Lu Huang and Shan Liu) independently extracted relevant data from the included studies and summarized it. Any disagreements appeared would be resolved by consulting an adjudicating senior author (YF Xia). The following data were extracted from the eligible studies: name of the first author, year of publication, country where the study was carried out, number of patients, clinicopathological features in different expression of PD-1/PD-L1, cut off value for PD-1/PD-L1 overexpression, detection area, detection methods, prognostic endpoints of interest, statistical analysis approach, and HR and its 95% CI for the endpoints of interest.
The methodological quality of the retrieved articles was evaluated according to the Newcastle–Ottawa quality assessment scale (NOS) for cohort study. The scale is composed of eight items of assessment that can be divided into the following three subgroups: selection, comparability, and outcome, with the subtotal score of 4, 2, and 3, respectively. Finally, a total score of 0–9 was distributed to each eligible study.
Endpoints of interest and statistical analysis
The primary endpoint of our study was overall survival (OS). If adequate data is available, the primary outcomes would be subdivided in order to make further analysis of subgroups. The secondary endpoints were disease-free survival (DFS), progression-free survival (PFS), distant-metastasis free survival (DMFS), and the clinicopathological factors.
The statistical analysis was performed using Review Manager 5.3 (Cochrane Collaboration, Oxford, UK) and STATA 14. We estimated the prognostic significance of PD-1/PD-L1 expression in NPC by directly using HR and its 95% CI when the original articles reported, otherwise, the Kaplan–Meier curve would be used to obtain HR and its 95% CI using the method provided by Tierney et al. [
21]. Statistical heterogeneity between studies was quantified by using the Q test and the I
2 statistic. The p-value of Q test lower than 0.1 or the statistic of I
2 higher than 50% was considered that it existed heterogeneity between studies. When it appeared heterogeneity, a random effects model would be chosen to pool the data, otherwise, a fixed effects model would be used. Subgroup analysis and sensitivity analysis were performed to track the origin of the heterogeneity. In this article, we adopted HR > 1 as the benchmark of PD-1/PD-L1 overexpression indicating a poorer outcome. We evaluated the publication bias through the Begg’s and Egger’s test in quantitatively. A two-tailed p value < 0.05 was considered to be statistically significant.
Discussion
This meta-analysis of 18 retrospective cohort studies including 15 articles covering PD-L1 and six studies concerning PD-1, a total of 1836 patients, comparing different expression level of PD-L1/PD-1 and the prognosis of NPC showed that no statistical significance was found between higher/positive expression of PD-L1/PD-1 and the prognosis of NPC. Among the subgroup analysis, when the detection area of PD-L1 was the immune cells, a higher expression of PD-L1 displayed better overall survival in NPC. Observing all the clinicopathological characteristics, only the expression of PD-L1 was found correlate with PD-1, and no other clinicopathological features was related with the expression of PD-L1/PD-1.
PD-1, known as an inhibitory receptor, which is mainly expressed on activated T lymphocytes, but also can be observed on B cells and NK cells [
37]. PD-L1 as well as PD-L2, one of the ligands for PD-1 receptor, belonging to B7 family is found expressed on PD-L1 contributes to the tolerance and impairment of the immune system [
38]. The tumor microenvironment makes contribution to the upregulation of PD-1 expression in TILs, which leads to the impairment of antitumor immune. Since the upregulated PD-1 in TILs will result in an exhausted phenotype and impaired function of T cells in TILs [
39]. At the same time, it is known that PD-L1 has been discovered expressing on many types of tumor lines, and its combination with PD-1 in TILs serves as one of the pivotal mechanism of modulating tumor cells with immunogenicity to escape from the surveillance of the host immune system [
40].
On account of various of tumor types and therapeutics, the prognostic value of PD-L1/PD-1 in tumor has not reach a consensus. In a meta-analysis about solid tumors included 61 articles showed that the overexpression of PD-L1 suggested a worse prognosis [
41]. However, when it comes to some specific types of tumors, the results were not the same. For instance, a study pooled 11 articles revealed that the expression of PD-L1 had no statistical correlation with the prognosis of patients with oral squamous cell carcinoma, whatever in OS, DFS or disease-specific survival (DSS) [
42]. In addition, another meta-analysis concerning osteosarcoma involving eight original studies indicated that the overexpression of PD-L1/PD-1 significantly related with a higher incidence of metastasis and total mortality risk [
43].
In our study, the prognostic value of PD-L1 expression in NPC was reported in 15 eligible studies. A previous study regarding the correlation of PD-L1 expression and head and neck carcinoma (HNC) included 17 original articles totally, while only 2 of them are about NPC [
44]. The conclusions drawn from the present study are something different from the previous one. The meta-analysis study about HNC demonstrated that no association was found between PD-L1 expression and the prognosis of HNC, however, the subgroup analysis indicated that in the group of Asia regions/countries a poorer OS in HNC may correlated with positive expression of PD-L1. Compared with the previous study, original studies included in our meta-analysis were all done in Asia regions/countries and the pooled data revealed no statistical difference between the expression level of PD-L1 and prognosis of NPC. It may be because nasopharyngeal carcinoma and non-nasopharyngeal head and neck cancer vary in their clinicopathological features and therapeutics.
The results of another recently published meta-analysis elucidating the prognostic value of some immune checkpoints in head and neck cancer were completely different from our meta-analysis and the previous study of head and neck that we just mentioned above [
45]. The analysis demonstrated that PD-L1 expression higher indicated better OS in HNC, especially in NPC (p = 0.01). While, by comparison with our study, several obvious distinctions deserve to be concerned. First, we found that it only included seven studies in the meta-analysis of PD-L1 expression in NPC about OS, by contrast, the present study included 12 eligible studies to pool the data regarding the impact of PD-L1 expression in OS, containing the previous seven studies. Second, some details of the data (HR and 95% CI) adopted were nuance. Since a few of survival data were derived from Kaplan–Meier curve, data deviation was inevitable. Also, when the eligible original articles provided two or more alternative survival data, choice among researchers could not reach unanimous unless prior discussion. For example, the study of Ono et al. [
27] offered the results of both univariate and multivariate, however, Jia et al. [
45] utilized the univariate one and our meta-analysis chose multivariate one. Similarly, Chang et al. [
23] supplied survival data in accordance with two PD-L1 cut-off value, while Jia et al. [
45] used the 1% cut-off value and we adopted the most common one, 5% cut-off value, to pool data.
Among all the original reports eligible in our study, some of the studies demonstrated that relatively higher expression of PD-L1 or PD-1 in NPC predicts worse prognosis [
22,
24,
26,
28,
29,
31,
34‐
36]. The most common and convincing explanation for the result was the interactions of PD-1 and PD-L1 would lead to immune suppression and promote tumor progression [
38,
40]. None of the multivariate analysis results in eligible studies approved for the main conclusion drew in Jia et al. [
45]. Remaining original studies declared no statistically significant difference in the expression of PD-1 or PD-L1 and the prognosis of NPC, which were consistent with our conclusion.
Interferon-gamma (IFN-γ) has been shown to be detected in tumor cells with PD-L1 positive and TILs, and act as a primary inducer for PD-L1 expression, while it could not be detected in PD-L1 negative tumor cells [
46]. Therefore, PD- L1 expression may not directly result in immune evasion and it may represent an outcome of on-going immune system triggers immune inhibition by producing IFN-γ or other cytokines [
25,
46]. In the tumor microenvironment of NPC, the condition of infiltrating lymphocytes is more complexity, on account of the high infectious rate of Epstein–Barr virus (EBV) in NPC [
47]. Sustainably exposure in viral contributes to more active inflammatory reactions resulting in the expression of PD-L1 [
48,
49]. It can be inferred that the expression of PD-L1 in NPC may just reflect the inflammatory response in the tumor microenvironment, which may need further research to identify. The expression of PD-1 in TILs have been shown to be dynamically changed in conformity to the status of immune system and tumor microenvironment, which may illustrate the undefinition of the prognostic value of PD-1 expression in NPC [
30,
50]. Recently, more and more clinical trials are focusing on immunotherapy in HNC, especially anti-PD-L1/PD-1 antibody in NPC. Hsu et al. [
5] reported a phase Ib clinical trial about anti-PD-1 monoclonal antibody, pembrolizumab, in NPC with PD-L1 positive, and revealed its antitumor activity and manageable adverse events. Another phase I trials concerning anti-PD-1 antibody, camrelizumab, in NPC also derives positive results [
51]. Colevas et al. [
52] conducted phase Ia trials regarding anti-PD-L1, atezolizumab, in HNC, and get encouraging antitumor activity regardless of PD-L1 expression. Since increasing researchers are concerning about the immunotherapy of anti-PD-L1/PD-1 antibody in NPC, it is worthwhile to figure out the role of PD-L1/PD-1 plays in the tumor microenvironment and prognosis of NPC.
Our meta-analysis performed sensitivity analysis on the outcome of OS regarding PD-L1 by cut-off value, detection area, calculation of HRs, and estimation method of HRs. Though we found that higher expression of PD-L1 in immune cells correlated with better OS, it only included three studies evoking the conclusion may not be robust. When we exclude the study of Li et al. [
26], heterogeneity decreased most obviously compared with deleting other single studies. We speculated that it may because of its unique cut-off value for PD-L1 and the estimation of HR was based on Kaplan–Meier curve. Furthermore, we discovered that its quality score was the lowest. However, excluding the study did not influence the main conclusion we drew. We failed to perform subgroup analysis among studies about PD-1, due to the deficiency of related original studies, which is a limitation of our study. Another limitation in our analysis is that the heterogeneity is obvious among studies, and it may originate from varied cut-off value for PD-L1 positive based on the subgroup analysis and summarization of main characteristics of eligible studies. The cut-off value for PD-L1/PD-1 positive in tumor microenvironment may need further discussion to determine suitable world-wide criterions in accordance with different types of tumors.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.