PD-L1 expression as a potential predictor of immune checkpoint inhibitor efficacy and survival in patients with recurrent or metastatic nasopharyngeal cancer: a systematic review and meta-analysis of prospective trials

Background The predictive value of programmed death-ligand 1 (PD-L1) expression in nasopharyngeal cancer (NPC) patients receiving immune checkpoint inhibitors (ICIs) remains controversial. This study aimed to evaluate the optimal threshold of PD-L1 expression in predicting the efficacy of ICIs in patients with recurrent or metastatic (R/M) NPC. Methods A meta-analysis was performed by retrieving relevant literature from PubMed, EMBASE, and Cochrane Library databases. Data on the pooled risk ratio (RR), mean overall survival (OS), progression-free survival (PFS), overall response rate (ORR) with 95% confidence interval, and 1%, 10%, and 25% PD-L1 expression cutoff points were obtained to examine the role of PD-L1 as a biomarker in R/M NPC patients receiving immunotherapy. Results In total, 1,312 patients from 14 studies were included. An improvement in PFS was observed in both patients with PD-L1 ≥ 1% (RR = 0.76, 95% CI 0.62–0.92, P = 0.005) and those with PD-L1 < 1% (RR = 0.68, 95% CI: 0.35–1.32, P = 0.26) who received first-line treatment with immunotherapy, with no significant difference between these subgroups. The pooled ORR was significantly higher in patients with PD-L1 ≥ 1% (ORR = 0.37) than in those with PD-L1 < 1% (ORR = 0.22) (P < 0.01) undergoing subsequent-line treatment. However, when we used the PD-L1 cutoff values of 10% and 25%, there was no significant difference between the positive (PD-L1 expression ≥ the cutoff value) and negative (PD-L1 expression < the cutoff value) subgroups. PD-L1 ≥ 1% also tended to be associated with better PFS and OS. Conclusions Our meta-analysis suggested that first-line immunotherapy could significantly improve PFS in R/M NPC patients, regardless of the PD-L1 expression levels. Positive PD-L1 expression (≥ 1%) might be a potential predictive biomarker for a better overall response to immunotherapy in R/M NPC patients in subsequent-line setting. Systematic review registration https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42024495841 PROSPERO, identifier CRD42024495841.

levels.Positive PD-L1 expression (≥ 1%) might be a potential predictive biomarker for a better overall response to immunotherapy in R/M NPC patients in subsequent-line setting.

Introduction
Nasopharyngeal carcinoma (NPC) is a common type of head and neck cancer with a skewed geographical, ethnic, and sex distribution.It is particularly prevalent in east and southeast Asia, where the highest age-standardized rates occur (1).According to GLOBOCAN 2020 data, approximately 133,354 new cases and 80,008 deaths from NPC were reported worldwide, of which 62,444 cases (46.8%) and 34,810 deaths (43.5%) were registered in China (2).
In the past decade, the global incidence and mortality rates of NPC have gradually declined (3), which could be attributable to lifestyle and environmental changes, the use of intensity-modulated radiation therapy, and the increasing application of adjuvant chemotherapy (4,5).However, approximately 15%-30% of patients who develop recurrent or metastatic (R/M) NPC have a median overall survival (OS) of less than 2 years (6).The main challenges in treating these patients are overcoming chemo-resistance and reducing the risk of adverse events (7).Currently, immunotherapies, especially immune checkpoint inhibitors (ICIs), represent a promising strategy to resolve these problems and effectively treat R/M NPC patients.
ICIs, particularly anti-programmed death-1 (PD-1)/ programmed death-ligand 1 (PD-L1) and anti-cytotoxic Tlymphocyte-associated antigen 4 (CTLA-4) antibodies, which activate CD8-positive T cells and induce cancer cell mortality, have revolutionized the treatment of advanced cancers.The tumor microenvironment of NPCs, characterized by massive inflammatory and immune cell infiltration, allows NPC patients to fully benefit from ICI therapy.ICIs have emerged as effective treatment options for patients with refractory R/M NPC.More recently, the Food and Drug Administration (FDA) approved toripalimab as a treatment for R/M NPC to be used in combination with first-line chemotherapies or subsequent-line monotherapies (8).The National Comprehensive Cancer Network (NCCN) Guidelines version 2.2024 refer to cisplatin/gemcitabine combined with ICIs as the first-line treatment in the management of R/M NPC (9).However, only about 50% of patients respond to treatment, indicating the major challenge of identifying patients who are suitable for immunotherapy (6).
The level of PD-L1 expression is one of the most commonly explored predictive biomarkers for the success of ICIs.Previous studies have shown that higher PD-L1 expression levels are associated with a higher response rate and better survival in patients with advanced stage melanoma treated with ICIs (10)(11)(12)(13).However, the predictive value of PD-L1 expression in NPC patients receiving ICIs remains controversial (14-16).Currently, there is no report of studies exploring the optimal cutoff value of PD-L1 expression to guide the clinical use of ICIs.
In this systematic review (SR), we comprehensively evaluated whether the expression level of PD-L1 influences the efficacy of anti-PD-1/PD-L1 monotherapy or combined therapy in NPC patients.Furthermore, subgroup analyzes were performed to assess and quantify the best cutoff value for PD-L1-positive tumors to guide future clinical practice.

Materials and methods
The study was reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyzes (PRISMA) (17).The protocol for this SR and meta-analysis was registered in PROSPERO (no.: 495841).

Eligibility criteria
To be eligible for this SR, studies were required to satisfy the following Population, Intervention, Comparison, Outcomes, and Study design (PICOS) criteria.Patients with a pathological diagnosis of R/M NPC who received immunotherapy with/without other systematic treatments were included.The included studies were required to report at least one clinical outcome, namely OS, progression-free survival (PFS), or overall response rate (ORR), based on the PD-L1 expression levels of patients.Randomized control trials (RCTs) and non-RCTs were considered eligible.There was no restriction on the language or publication status of studies.Patients receiving radiotherapy were not eligible for this SR.Review articles, case reports, conference abstracts, protocols, editorials, and commentaries were also excluded.

Literature search
A comprehensive literature search was performed on PubMed, Embase, and the Cochrane Library to identify potential eligible studies published from January 2013 to December 6, 2023.We also manually searched for eligible studies by checking the reference lists of retrieved studies to minimize the risk of missing relevant information.The detailed search strategy is described in Supplementary File 1.

Literature selection
The titles and abstracts of potential studies were screened independently by two authors (C.H.L.W. and S.K.CH.), and then their full texts were assessed for eligibility.If there was any dispute, it was resolved through discussion between the two authors.A third author (C.L.C.) was consulted to settle unresolved disagreements.
A list of studies for inclusion was generated.For duplicate studies, the most recent and comprehensive version of each was selected for inclusion.SRs identified during the search were examined to ensure that no eligible studies were omitted.

Data extraction
Data were extracted by one author (R.Y.X.) and cross-reviewed by the other two authors (C.H.L.W. and S.K.CH.).Key information, including authors' details, year of publication, study population, sample size, patient characteristics, follow-up time, intervention, and results of all prespecified outcomes, were extracted from each eligible study using a pre-designed data-extraction table.

Methodological quality assessment
The methodological quality of all included studies was evaluated by two reviewers (C.H.L.W. and R.Y.X.) independently using the Cochrane's Risk of Bias in Randomized Trials (RoB 2) tool for RCTs (18) and the Cochrane's Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool for non-randomized studies (19).For the included single-arm non-randomized studies, risk of bias was assessed using a modified ROBINS-I approach (20).The risk of bias was categorized as low, moderate, serious, or critical.Publication bias was assessed using Egger's regression test and through a visual inspection of funnel plot asymmetry if there were more than 10 studies (21).

Data analysis
To examine the role of PD-L1 among R/M NPC patients receiving immunotherapy with/without other systematic treatment, we conducted a pairwise random-effects meta-analysis comparing immunotherapy plus chemotherapy patients with controls in the first-line therapy setting using RevMan version 5.4.We used pooled risk ratios (RRs) with 95% confidence intervals (CIs) to present PFS data.
Single-arm random-effects meta-analyzes were performed to synthesize the effects of immunotherapy with/without other systematic treatments on the clinical outcomes (i.e.OS, PFS, and ORR) in both first-line and subsequent-line settings using R version 4.2.3.The pooled estimated mean OS and PFS, as well as the pooled ORR with 95% CI, are presented.
For both pairwise and single-arm meta-analyzes, subgroup analysis was performed on each clinical outcome by stratifying patients into two groups (1): PD-L1 positive and (2) PD-L1 negative.Three cutoff points for PD-L1 expression level were used: 1%, 10%, and 25%.We also conducted a sensitivity analysis of the impact of treatments on the clinical outcomes by excluding patients who received combined immunotherapies and targeted therapies.We used I 2 values to quantify the level of heterogeneity, with I 2 < 25% indicating a low level of heterogeneity, 25%-50% indicating a moderate level of heterogeneity, and >50% indicating a high level of heterogeneity (22).

Literature search and selection
The literature search yielded 488 citations, among which 99 duplicate studies were removed.After screening the titles and abstracts, 116 eligible articles remained.As the full texts of 15 articles were not available, only 101 remaining papers proceeded to full-text assessment.Eighty-nine of these were excluded because (i) no recurrent or metastatic NPC adult patients were included (n = 22); (ii) treatment included radiotherapy or other therapies (n = 38); (iii) PD-L1 outcomes were not reported (n = 22); or (iv) they were retrospective studies (n = 7).With the identification of one additional reference through manual searches of the reference lists of included studies, a total of 14 studies in 13 articles were included in this systematic review.Details of the literature search and study selection are shown in the PRISMA flow diagram (Figure 1) (23).

Publication bias
Publication bias was assessed with an Egger's regression plot for 12 articles focusing on subsequent-line therapy.The plot revealed no presence of publication bias (P = 0.13), and no asymmetry was found in the funnel plot (Figure 7).

Discussion
To the best of our understanding, this is the first in-depth analysis of the predictive value of PD-1/PD-L1 status in clinical trials of immunotherapy and combined therapy for patients with advanced metastatic NPC.We comprehensively evaluated the correlation between different expression levels of PD-L1 and the ORR, PFS, and OS of R/M NPC patients, with data retrieved from 14 studies that included 1,434 patients in total.We aimed to determine the predictive value of PD-L1 and to identify an optimal PD-L1 cutoff value for the selection of patients likely to respond effectively to anti-PD-1/PD-L1 treatment.
The NCCN guidelines only recommend to use i) cisplatin/ gemcitabine alone, or ii) cisplatin/gemcitabine plus toripalimab, or iii) cisplatin/gemcitabine plus the other PD-1 inhibitors (pembrolizumab or nivolumab) as first-line therapies for R/M NPC (9).In our analysis of the two included studies in the firstline setting, the use of ICIs could significantly improve PFS, regardless of the PD-L1 expression levels.Various ICI monotherapies (toripalimab, pembrolizumab or nivolumab) are recommended as subsequent-line therapy based on the PD-L1 expression levels in NCCN guidelines (9).The availability of ample subsequent-line treatment data gave us the opportunity to draw more precise and accurate conclusions.The most compelling finding in our study was that patients with PD-L1 ≥ 1% who received ICI in the subsequent-line setting had significantly higher ORR than in those with PD-L1 < 1%.Our pooled results showed no significant difference between subgroups in analysis of PFS and ORR for PD-L1 cutoff value of ≥ 10%, and ≥ 25%.However, higher the PD-L1 expression, the higher the probability that the patient was able to achieve clinical benefit from ICIs in the subsequentline setting.
ICIs, which reactivate immune response in the tumor by preventing immunosuppressive factors from binding to their ligands, have fewer side effects compared to chemotherapy (37).Side effects of ICIs are usually mild.The most common side effects include fatigue, itchy rash, and diarrhea (38).In addition, as ICIs may also activate autoreactive T cells, they increase the risk of immune-related adverse events (irAEs).In our included studies, irAEs such as hypothyroidism, aspartate aminotransferase(AST) level increased, and rash, were also frequently reported.The detail information is shown in Table 1.
PD-L1, the most common immunosuppressive ligand, expressed on the tumor cell membrane combines with the PD-1 of tumor-infiltrating lymphocytes (TIL), contributing to tumor cell evasion from host immune system surveillance (39).In previous Forest plot of RR of PFS after first-line treatment.(A) 1% cut off; (B) 10% cutoff.PDL1, programmed cell death ligand-1; RR, risk ratio; PFS, progression-free survival.
studies, high expression of PD-L1 appeared to adversely affect the survival outcomes of NPC patients.A meta-analysis involving 13 studies showed that PD-L1 over-expression in NPC was associated with a poor OS (hazard ratio = 1.48, 95% CI: 1.00-2.18,P = 0.049) (40).Another study discovered a significant correlation between high PD-L1 expression and a short PFS/OS (41).In contrast to previous studies that examined the prognostic value of PD-L1 in patients with NPC, our study evaluated the predictive value of PD-L1 expression for ICI therapy.The results provide evidence that PD-L1-positive patients received more benefit than PD-L1-negative patients at a PD-L1 cutoff value of 1%, which sets a preliminary framework for the R/M NPC patient population suitable for ICI treatment.However, as the cutoff values varied across articles, coupled with the fact that 1% was the most widely used expression-level cutoff for PD-L1 detection, more comprehensive studies on PD-L1 expression levels and ICI treatment efficacy are warranted to accurately validate these results.
Moreover, the PD-L1 expression on ICIs effect shows differences between first-line treatment and subsequent-line treatment, which may be caused by many factors.First, it is known that tumor progression is influenced by the tumor immune microenvironment, one of the important mechanisms is escape from immune surveillance with the selection of poorly immunogenic cells (42, 43).When the disease becomes refractory, the tumor microenvironment (TME) becomes more immunesuppressive.As a result, in the first-line setting when TME is still favorable, the immunotherapy-chemotherapy combination would improve survival regardless of PD-L1 expression.However,in subsequent line settings when TME becomes more immunosuppressive, only those with higher PD-L1 expression derived benefit from checkpoint inhibitors.
Second, all first-line trials evaluate the immunotherapychemotherapy combination while most later-line studies are using immunotherapy-alone (44).Chemotherapy could activate the T-cell priming and recruitment and works synergistically with immunotherapy, therefore patients who accept first-line treatment of immunotherapy-chemotherapy combination would respond to the treatment regardless of PD-L1 expression.
Though PD-L1 is the most widely studied biomarker for immunotherapy, additional biomarkers have been evaluated in several studies.For instance, a meta-analysis showed that patients with lower baseline plasma Epstein-Barr virus (EBV) DNA levels had a higher ORR and longer median PFS than those with higher EBV DNA levels, but tumor mutational burden (TMB) was not significantly correlated with clinical prognosis in NPC patients treated with ICIs (16).Furthermore, a statistical difference in PFS was observed between patients with tumors showing loss of HLA-A and/or HLA-B expression, and patients with tumors expressing both HLA-A and HLA-B in trial NCI-9742 (26).A single-arm phase II clinical trial indicated that, in R/M NPC patients, a strong suppression of TGFb1 levels was associated with worse ORR and PFS (30).
With the development of bioinformatics and biotechnologies, novel forms of biomarkers, such as mutations/chromosomal abnormalities, have been made available that provide new perspectives on precision medicine.A recent clinical trial revealed that 11q13.3focal amplification and high MRGPRF expression are predictive of poor outcomes following gemcitabine plus apatinib and toripalimab therapy, but in another study (POLARIS-02), the genomic alternations had no statistically significant associations with clinical efficacy (28,33).However, our study of the PD-L1 biomarker has particular clinical relevance.PD-L1 status is readily  Forest plot of pooled results of OS after subsequent-line treatment (1% cutoff).Our meta-analysis has several limitations.First, there was significant variability in the literature with regards to the prevalence and prognostic significance of PD-L1 expression in NPC patients, probably because of differences in the assays and scoring methods used across studies.However, in a cross-correlation study performed using different PD-L1 immunohistochemical assays, the JS311 antibody had similar PD-L1 staining patterns and scores to the antibodies 22C3, 28-8, and SP263 (45).The predictive utility of PD-L1 expression may also depend on its differential expression in immune cells versus tumor cells.Second, there was a lack of sufficient clinical trials of first-line treatments reporting OS and ORR in patients with different PD-L1 expression levels that could be included in our analysis.Despite the encouraging outcomes, the limited number of articles means we are skeptical of the conclusions, and more clinical trials focusing on ICI treatments are needed for further validation.Third, only three and two studies were included in the analysis of the PFS and OS, respectively, for the PD-L1 10% level.More clinical trials are needed to further enrich and validate our conclusions and better guide the use of clinical PD-L1 levels to maximize the benefits and reduce the side effects of ICIs.Lastly, most of the studies included were conducted in Asian populations, and the regional characteristics of NPC may limit the generalizability of our findings.

Conclusions
Our meta-analysis suggested that first-line immunotherapy could significantly improve PFS in R/M NPC patients, regardless of the PD-L1 expression levels.Nonetheless, positive PD-L1 expression (≥ 1%) might be a potential predictive biomarker for a better response to immunotherapy in R/M NPC patients in subsequent-line setting.The higher the PD-L1 expression, the higher the probability that the patient was able to achieve clinical benefit from subsequent-line treatment.

FIGURE 5 Forest
FIGURE 5Forest plot showing pooled results of PFS after subsequent-line treatment (1% cutoff).

TABLE 1
Characteristics of the studies included in the meta-analysis.

TABLE 2
Technical information of PD-L1 measurement in the included studies.