High PD‐L1 expression in the tumour cells did not correlate with poor prognosis of patients suffering for oral squamous cells carcinoma: A meta‐analysis of the literature

Abstract Objectives Oral cancer represents one of the most common malignancies in humans. Its prognosis is still poor, despite the most recent improvements in therapies. An increasing attention is placed on the role of programmed death ligand 1 (PD‐L1) in the tumour immunity and its potential function as a marker for tumour prognosis. Whether PD‐L1 expression is a prognostic factor for the poor outcomes in oral squamous cell carcinoma is still controversial. This study aimed to investigate, through a meta‐analysis, a potential correlation between PD‐L1 expression and the prognostic outcomes in patients with oral squamous cell carcinoma. Materials and methods The studies were identified by searching PubMed, SCOPUS, Web of Science and were assessed by two of the authors. After the selection process, 11 articles met eligibility criteria and were included in the meta‐analysis. Quality assessment of studies was performed according to the REMARK guidelines, and the risk of biases across studies was investigated through Q and I 2 tests. Meta‐analysis was performed to investigate the association between the PD‐L1 expression either overall survival (OS), disease‐free survival (DFS), disease‐specific survival (DSS), gender and lymph node metastasis. Results A total of 1060 patients were analysed in the 11 studies included in the meta‐analysis. Pooled analysis revealed that the expression of PD‐L1 did not correlate with poor OS (HR, 0.60; 95% CI: [0.33, 1.10]; P = 0.10), DFS (HR, 0.62; 95% CI: [0.21, 1.88]; P = 0.40), DSS (HR, 2.05; 95% CI: [0.53, 7.86]; P = 0.29 and lymph node metastasis (HR, 1.15; 95% CI: [0.74, 1.81]; P = 0.53). Furthermore, results of the meta‐analysis showed that high expression of PD‐L1 is two times more frequent in female patients (OR, 0.5; 95% CI: [0.36, 0.69]; P < 0.0001) compared to males. For all the three outcomes analysed, a high rate of heterogeneity was detected (I 2 > 50%). Discussion High PD‐L1 expression did not correlate with poor prognosis of patients suffering for oral squamous cell carcinoma. Studies published on the topic showed a significant variation in results, limiting the use of PD‐L1 expression by immunohistochemistry as prognostic biomarker in clinical practice.


| INTRODUC TI ON
Oral squamous cells carcinoma (OSCC) represents one of the most common malignancies in humans. 1 An annual incidence of about 200 000 new cases per year has been estimated worldwide. 2 Both incidence and mortality rate are about 2.8 times higher in males than in females. 3 The most known risk factors for the onset of OSCC are tobacco smoke, betel chew and alcohol consumption. 4 The prognosis of OSCC is still poor, showing very little improvements in the last decades, despite advances in therapies. 5 Recently, immunotherapy showed promising effects for the treatment of such patients. 6 The results of several studies suggest an important role of immune evasion mechanisms in the pathogenesis of OSCC. For these reasons, a deeper understanding of molecules involved in the function of immune system is crucial for the development of future strategies of treatment.
As it is known, cancer cells can negatively regulate the immune response through the activation of inhibitory immune checkpoints.
To date, different inhibitory immune checkpoints have been studied, including cytotoxic T-lymphocyte protein 4 (CTLA4), programmed cell death protein 1 (PD-1), lymphocyte activation gene-3 (LAG3), T-cell immunoglobulin-3 (TIM3) and T-cell immunoglobulin and ITIM domain (TIGIT). 7 In this article, we focused on the PD-1 immune checkpoint as the pharmacological inhibition of this immune checkpoints has recently demonstrated to improve the survival rate of patients with head and neck squamous cells carcinoma (HNSCC), 8 while the power of evidence is still weak regarding the clinical efficacy of the pharmacological inhibition of the other immune checkpoints above mentioned. In particular, we reviewed studies focused on the analysis of the programmed cell death ligand-1 (PD-L1) as a prognostic factor of patients suffering for OSCC. PD-L1 is a cell surface glycoprotein which induces both anergy and apoptosis of T cells through the activation of PD-1 receptors located on their surface. 9 The biological importance of the PD-1 receptors influences significantly the immune responses because of a diffused ligand distribution in the body. In fact, such axis showed to play a crucial role in autoimmunity, 10 tumour immunity, 11 infectious immunity 12 and allergy. 13 PD-L1 is commonly expressed in some healthy tissues since it is involved in the normal immunological homeostasis. 14 However, in many types of cancer, the expression of PD-L1 on tumour cells is remarkably higher. This overexpression seems to be present also in subsets of immune cells, including B and T cells, macrophages and dendritic cells. 11 Several studies demonstrated a strong correlation between PD-L1 expression on various tumour cells and a worse patients' prognosis. [15][16][17][18] Many studies have also been conducted to discover a possible role of the PD-1/PD-L1 axis in the biology of OSCC. 19,20 Its potential clinical and pathological implication has also been investigated providing, however, non-homogeneous conclusions.
The aim of the present study was to systematically review the literature and perform a meta-analysis on the available data in order to summarize the possible correlations between PD-L1 expression and the prognosis of patients suffering for OSCC.

| Protocol and Registration
This systematic review has been carried out following the guidelines of the "Preferred Reporting Items for Systematic Reviews and Meta-Analyses" (PRISMA) guidelines 21 and the Cochrane Handbook. 22 In addition, the protocol for the development of this review was prospectively registered on the online database PROSPERO (International prospective register of systematic reviews) with the registration number CRD42018090716.

| Eligibility criteria
The inclusion criteria were the following: (a) both prospective and retrospective clinical cohort studies, written in English language, regarding the immunohistochemical evaluation of PD-L1 expression in samples from OSCC patients; (b) at least 20 patients were included in each study; (c) studies which analysed the prognosis calculating the hazard ratio (HR) and its 95% confidence interval (95% CI) for at least one of the following: overall survival (OS), disease-free survival (DFS), disease-specific survival (DSS), gender and lymph node metastasis. Some studies reported the HR and 95% CI in the article.
Others only reported the Kaplan-Meier graph. In this case, the HR and 95% CI were extracted by Kaplan-Meier graph using the method reported by Tierney et al. 22 If the article did not report both HR and 95% CI, or the Kaplan-Meier graph, author was contacted by email.
By this last method, we got the HR and 95% CI for two studies. 23,24 Studies on non-human model, case series with less than 20 patients and case reports were not considered for the inclusion in this review.
No restrictions were applied about the year of publication. Science. The research process was carried out by two reviewers in an independent manner. MeSH terms and free text words were combined using Boolean operators (AND, OR). The following protocol was used: ((((PD-L1 OR Programmed Death Ligand 1 OR checkpoint inhibitor OR immune system))) AND ((OSCC OR "oral cancer" OR Tongue OR gingiva))) AND ((survival OR prognosis OR biomarker)).

| Study selection, data collection process and data items
The selection process was performed in two rounds. In the first round, authors screened the studies reading only title and abstract of publications, while in the second phase, a full-text evaluation was performed. In case of disagreement between reviewers, a final decision for the inclusion was taken in a joint session with a third author (VCAC). This author also calculated a value of kstatistic to show the level of reviewers' agreement. At the end of the selection process, papers fulfilling all inclusion criteria were included in the quantitative synthesis. Data extraction was performed using an ad hoc extraction sheet by two authors (VCAC and CA) in a joint session and controlled by a third author (GT).
For each study, the following data were extracted: name of the first author, year of publication, name of the country where the study was performed, classification used for staging, number of patients included, cut-off values, gender, staging, tumour size, rate of lymph node metastasis, HRs and 95% CI for the survival outcomes considered.

| Risk of bias assessment
The risk of bias of the included studies was evaluated using a classification derived from the Reporting Recommendations for Tumour Marker Prognostic Studies (REMARK), 25

as previously reported by
Almangush et al. 26 The scale consists of six parameters evaluating

| Summary measures and planned methods for analyses
For the pooled analysis of PD-L1 expression as prognostic factor in OSCC patients, the natural logarithm of HR and its standard error (SE) were calculated and entered into the software: Review Manager version 5.2.8 (Cochrane Collaboration, Copenhagen, Denmark; 2014). The inverse of variance test was used to calculate the overall effect. Results of the meta-analysis were summarized in forest plots, and a P-value lower than 0.05 was considered as threshold of statistical significance for all the tests performed in this meta-analysis.
Sensitivity analyses were performed for the outcomes OS and DFS omitting articles on the basis of risk of bias, cut-off and geography, hence repeating meta-analysis through a random effect model.

| Study features and risk of bias within studies
A total of 1060 patients were analysed in the 10 studies included in the meta-analysis. 19 reported only results for univariate analysis. Three studies fully respected the REMARKS guidelines, 19,28,29 while the remaining seven proved to be lacking in some of the parameters analysed. 23,24,[30][31][32][33][34]51 Absence of risk of bias was detected only for the immunohistochemistry, while some deficiencies were present for the others parameters. Results of the risk of bias for each of the included study are reported in Table 1.

| Synthesis of results and risk of bias across studies
Meta-analysis of seven studies revealed no significant correlation between high/low expression of PD-L1 and OS (HR, 0.60; 95% CI:

| D ISCUSS I ON
PD-L1, also known as B7-H1 or CD274, is a cell surface glycoprotein, which leads to T-cell inactivity or apoptosis by binding PD-1, a receptor expressed by the T lymphocytes. 19 The interaction between inhibitors. Conversely, some PD-L1-negative tumours can respond to these agents. 58 However, the predictive role of PD-L1 expression in tumour samples is still controversial. 59 In this study, we focused on the analysis of PD-L1 expression in OSCC tissue as a prognostic (and not predictive) biomarker. In fact, such marker has demonstrated to be an independent prognostic factor in different cancer types, including NSCLC, 60 (Figures 1 and 2). No differences were also detected for the rate of lymph node metastasis in patients with higher PD-L1 expression ( Figure 3). Our findings are in discordance with the results of a previous meta-analysis on head and neck cancers in which authors revealed a significant association between PD-L1 expression and poor prognosis in a subgroup analysis. 66 Such discrepancy is in part due to the inclusion of the meta-analysis of two recently published studies in which PD-L1 expression correlated with a better prognosis. 28,29 The lack of correlation between PD-L1 expression and OS appears to contrast with the prognostic value that is attributed to this marker, based on its immunosuppressive function. Several studies regarding other tumour types found the same results, suggesting a more complex function of PD-L1 in immunosurveillance signalling. 67 A possible explanation is that PD-L1 expression by cancer cells can be considered as a marker of an active host anti-tumour immune response. 68 Another way to address the issue is to consider the heterogeneity of tumour microenvironment in different tumour types. the same cut-off value. Results of this study revealed a significant association between PD-L1 expression and female gender. In fact, in women, higher expression of PD-L1 seems to be more common as already reported for NSCLC. 75,76 In these studies, the female subset of patient also corresponds to patients who are more likely to harbour EGFR mutations, suggesting a relationship between PD-L1 expression and altered EGFR signalling pathway. 77 A recent meta-analysis revealed that the magnitude of benefit of patients treated with immunotherapy is sex-dependent, and in particular, women have lower rates of positive response to the treatment. 78 However, it is not clear whether such different outcomes are due to the more frequent expression of PD-L1 in females or to other sexrelated mechanism. Such findings underline the importance of performing future studies aiming to compare sex-related expression as independent prognostic factor, in order to clarify whether PD-L1 could be considered a prognostic factor in men but not in women.
Furthermore, as previously mentioned, there is a complex relationship between PD-L1 expression and the presence and pattern of inflammatory infiltrate. This must be considered in the evaluation of prognostic significance of PD-L1, because peritumoral inflammatory process seems to be more intense in female patients with OSCC, mainly due to postmenopausal inflammatory state. 19 Analysis of risk of bias in the included studies revealed deficiencies in some parameters of the REMARKS guidelines. In particular, the authors recorded ambiguity in some of the included studies in the distinction between OS and disease-specific survival. As it is known, in the calculation of OS, death for any reason is taken into consideration while in disease-specific survival only deaths for cancer are considered. It is to underline that direct contact of authors helped to clarify such discrepancy for two of the included studies. 23,24 To note, such meta-analysis presents some limits, first of all, it relied on published results rather than on individual patients' data. In addition, it presented, for the survival outcomes considered in the meta-analysis, a very high rate of heterogeneity was detected, that strongly limits the quality of

| CON CLUS ION
High PD-L1 expression did not correlate with poor prognosis of patients suffering for OSCC. The studies published on the topic showed a significant variation in results. Hence, results from the current available literature limit the use of PD-L1 expression by immunohistochemistry as prognostic biomarker in clinical practice.
Higher levels of PD-L1 expression are more frequent in females than in males, and such factor should encourage future studies on the sex-related role of this biomarker.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interests.