The Impact of Programmed Death-Ligand 1 Expression on the Prognosis of Early Stage Resected Non-Small Cell Lung Cancer: A Meta-Analysis of Literatures

Introduction Previous studies have demonstrated that programmed cell death-ligand 1 (PD-L1) serves as biomarker for poor prognosis and survival in advanced-stage non-small cell lung cancer (NSCLC) patients. However, the merit of PD-L1 expression to predict the prognosis of early stage NSCLC patients who underwent complete resection remains controversial. In the present study, we performed a meta-analysis to investigate the relationship between PD-L1 expression and prognosis in patients with early stage resected NSCLC. Methods Electronic databases, including PubMed, EMBASE, and the Cochrane Library, were searched until July 23 2020 for studies evaluating the expression of PD-L1 and the prognosis of resected NSCLCs. Hazard ratios (HRs) with 95% confidence intervals (CIs) of overall survival (OS) and disease-free survival (DFS) were pooled and analyzed. Heterogeneity and publication bias analyses were also assessed. Results A total of 15 studies involving 3,790 patients were considered in the present meta-analysis. The pooled HR indicated that PD-L1 expression related to a much shorter DFS (HR = 1.56, 95% CI: 1.18–2.05, p < 0.01), as well a significantly worse OS (HR = 1.68, 95% CI: 1.29–2.18, p < 0.01). Furthermore, our analysis indicated that PD-L1 expression was significantly associated with gender (male vs. female: OR = 1.27, 95% CI:1.01–1.59, p = 0.038), histology (ADC vs. SCC: OR = 0.54, 95% CI:0.38–0.77, p = 0.001), TNM stage (I vs. II–III: OR = 0.45, 95% CI:0.34–0.60, p = 0.000), smoking status (Yes vs No: OR = 1.43, 95% CI:1.14–1.80, p = 0.002) and lymph node metastasis (N+ vs N−: OR = 1.97, 95% CI:1.26–3.08, p = 0.003). Conclusions The results of this meta-analysis suggest that PD-L1 expression predicts an unfavorable prognosis in early stage resected NSCLCs. The role of personalized anti-PD-L1/PD-1 immunotherapy in the adjuvant settings of resected NSCLC warrants further investigation.


INTRODUCTION
Lung cancer is the most commonly diagnosed cancer and a leading cause of cancer-related deaths (1). Surgery is the standard treatment for early stage non-small cell lung cancer (NSCLC); however, postoperative prognosis remains unsatisfactory, with a 5-year survival rate ranging between 71 and 83% (2).Therefore, it is essential to identify new biomarkers for efficient clinical decision making and improve patient outcomes. Currently, blockade of the programmed cell death 1 (PD-1)/PD-1 ligand 1 (PD-L1) signaling pathway is one of the most promising immunotherapeutic strategies in boosting the immune system in the fight against cancer (3,4). Programmed death 1 (PD-1), an important immune checkpoint molecule, is an immune-inhibitory receptor expressed on the surface of activated T cells in response to persistent inflammatory stimuli (5,6). PD-L1 expressed on the tumor cells binds to the PD-1 receptors on activated T cells, resulting in the inhibition of the cytotoxic T cells. Blockade of the PD-1/PD-L1 pathway with monoclonal antibodies is a promising therapeutic strategy, with prominent clinical benefits of this checkpointblockade observed in recent clinical trials (7,8).
Previously, a study has demonstrated that PD-L1 is a marker of poor prognosis and survival in advanced-stage NSCLC patients (9). A meta-analysis, performed on six studies with 1,157 patients, demonstrated that NSCLC patients with positive PD-L1 expression exhibited a much poorer OS (10). Another meta-analysis by Li et al., which included the largest number of patients (fifty studies with 11,383 patients), also indicated that PD-L1 IHC expression was related to poor overall survival (11). However, the searching deadline for Li's study was January 2018, and more recent studies focusing on PD-L1 and prognosis in NSCLC were missing. More importantly, all of previous studies, including Li's study, were performed on the NSCLC patients of both early and advanced stages, and they only took OS into account. Therefore, the impact of PD-L1 expression in the prognosis of patients with early stage NSCLCs following complete resection remains controversial. In the present study, we performed a meta-analysis of all available evidence not only to analyze OS but also assess the correlation between PD-L1 expression and DFS in early stage surgically resected NSCLC patients, which is more accurate and valuable to reflect the influence of PD-L1 expression on the survival of NSCLC.

METHODS
This study was reported on the basis of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines (Supplementary Data 1).
Relevant studies were retrieved by searching PubMed, Embase, and the Central Registry of Controlled Trials of the Cochrane Library, using the following terms: "Carcinoma, Non-Small Cell Lung" AND "PD-L1 protein, human" AND "Prognosis". (Supplementary Data 2) The final search period was July 23 2020. Two authors (SX and SZ) performed the search independently. We restricted our research to studies published in English.
Eligible studies were in agreement with the following criteria: (1) the histology type of cancer was NSCLC; (2) valid TNM stage and cancer differentiation, as well as sufficient survival data, such as hazard ratio (HR) with 95% confidence intervals (CI), OS, and DFS were available (3) were published in English; (4) evaluated the association between PD-L1 expression and prognosis or pathological features; (5) involved early stage resectable NSCLC patients; (6) had similar research experimental design and methods; (7) PD-L1 expression was divided into high (positive) and low (negative) categories; and (8) relevant information could be extracted from the full-text study. Exclusion criteria included: (1) duplicate reports, ongoing studies, letters, conference papers, and reviews; (2) studies regarding lung cancer cell lines, animal models, and other types of cancer; (3) studies with insufficient survival data for which HR and CI could not be determined; (4) papers not published in English; (5) methods and experimental design differed from those of the selected studies.
The Newcastle-Ottawa quality assessment scale (NOS) and National Institute for Clinical Excellence (NICE) quality assessment scale was performed to assess methodological quality and risk of bias for cohort studies and case series studies, respectively. The primary outcomes of our study were diseasefree survival (DFS) and overall survival (OS). The characteristic details of the publications, including the first author's name, publication year, tumor type, PD-L1 level, stage, the evaluation method of the PD-L1expression, were extracted by two independent investigators. Any disagreement was discussed between investigators to reach a consensus. Multivariate analysis results were extracted as some included studies performed univariate analysis. We used the data directly from the included studies, providing precise HR (95% CI). In the case of studies only providing Kaplan-Meier survival curves, the Engauge Digitizer version 2.11 software was used to extract relevant numerical values from survival curves and calculate the HR (95% CI) (12,13).
The STATA 15.0 (Stata Corporation, College Station, USA) software and R package 4.0.2 were used for data synthesis and analysis.
The random-effects model was employed in case of potential heterogeneity and to avoid underestimation of standard errors of pooled estimates in this meta-analysis. HRs for DFS and OS with 95% CIs according to the expression status of PD-L1 were pooled. The pooled odds ratios (ORs) were used to investigate the correlation between PD-L1 expression and clinicopathological features.
The heterogeneity test was performed using the Cochrane's Q test (Chi-squared test; Chi 2 ) and I 2 metric (14). A chi-squared P value <0.1 or an I 2 statistic >50% was defined as statistically significant heterogeneity (15). Moreover, the potential publication bias was assessed through Begg's funnel plots (16). P <0.05 was considered as statistically significant based on the two-sided test. Subgroup analysis was conducted according to gender, histology, TNM stage, smoking status, and lymph node metastasis. An HR >1 reflected longer OS or DFS for PD-L1 negative patients.
Sensitivity analyses (17) were conducted to investigate the influence of a specific study on the pooled risk estimate by removing one study in each turn ( Figures 2D, 3D, Supplementary Data 3).

RESULTS
We identified 1,719 potentially relevant records through our search. A total of 1,558 studies were excluded after reviewing the title and abstract as their research contents failed to meet our inclusion criteria. Subsequently, the full texts of 92 articles were carefully screened, and a total of 15 studies (18-32) were found eligible for the final analysis. Figure 1 summarizes the flow chart depicting the process followed for study selection. No article was excluded by methodological quality and risk of bias and sensitivity analysis for significant heterogeneity (Figures 2D, 3D).
In all articles, resectability was a necessary intervention. All 15 studies were retrospective and published between 2016 and 2020. assays to evaluate PD-L1 expression in tumor cells and/or tumor-infiltrating lymphocytes (TILs). Table 1 summarized the characteristics of all involved studies, including perioperative therapy and PD-L1 detection information.
As shown in Figure 2A, the overall pooled HR indicated that the high PD-L1 expression was related to a significantly shorter DFS (HR = 1.56, 95% CI: 1.18-2.05, p < 0.01). Further analysis demonstrated heterogeneity among the eight studies (I² = 64%, p<< 0.01), and we hence performed Egger's and Begg's tests. According to Egger's test (p = 0.671, Figure 2B) and Begg's test (p = 0.620, Figure 2C), no publication bias influenced the HRs for DFS. Additionally, this confirmed that a negative correlation existed between PD-L1 expression and DFS in the case of resectable NSCLC patients.
The correlation of PD-L1 expression and clinical characteristics was further analyzed, including gender, histology, tumor stage, smoking status, and lymph node metastasis. We observed that PD-L1 expression was significantly associated with gender (male vs.  Table 2).
In addition, we also investigated the different cutoff values of PD-L1 expression on the survival. We found that PD-L1 expression at 5% cutoff value was not correlated to DFS (HR = . We propose that the discrepancy for these findings may be attributed to limited data and non-uniform PD-L1 detection platforms.

DISCUSSION
Immune checkpoint inhibitors targeting PD-1/PD-L1 have improved survival in patients with advanced NSCLC in secondand first-line settings (33)(34)(35)(36)(37). The PACIFIC trial reported a DFS benefit in patients with locally advanced, unresectable stage III NSCLC who received durvalumab consolidation therapy (38)  biomarker indicating poor prognosis and survival in advancedstage NSCLC patients. However, as reported by previous studies, the expression of PD-L1 on the prognosis of early stage resected NSCLC remains controversial. A comprehensive analysis is required to integrate all available data and provide further insight on this issue. By summarizing the data available from included studies, our results confirmed that the PD-L1 expression indicates an unfavorable prognosis in early stage resected NSCLC as well. Our conclusion that PD-L1 positive or high expression indicated a significantly inferior OS in early stage resected NSCLC, is consistent with the previous analysis for all stages or advanced stage NSCLC patients (9)(10)(11). Additionally, we demonstrate that DFS, which was not explored by the previous studies, is also negatively correlated with PD-L1 expression in early stage resected NSCLC. Lastly, based on subgroup analysis, we observed that the PD-L1 expression was associated with gender, histology, tumor stage, smoking status, and lymph node metastasis.
Our analysis provided evidence to support that the PD-L1 expression may have prognostic value in predicting survival of patients with resected NSCLC. However, there were several limitations to our study. Firstly, this was based on a retrospective analysis. A prospective analysis is required to further clarify these issues. Secondly, 12 out of 15 included studies were performed in Asian population. Although it was not our intention to ignore the data from non-Asian population, the lack of data from non-Asian population is still a limitation for our meta-analysis. Therefore, future investigation should focus more on the PD-L1 expression in early stage non-Asian NSCLC patients. Thirdly, the different adjuvant treatment strategies post-surgery and follow-up also influence the survival of patients with NSCLC undergoing resection, which could have influenced the analysis. Moreover, the cutoff value of the defined PD-L1 expression was rather  different among the included studies. We have to categorize high (positive) and low (negative) PD-L1 expression and study the correlation between PD-L1 expression and post-operative survival. Finally, the platform and antibody of PD-L1 detection was not uniform either in the included studies. We have mentioned this issue in Table 1.

CONCLUSIONS
In conclusion, our meta-analysis suggests that PD-L1 expression indicates an unfavorable prognosis in early stage resected NSCLC patients. In the adjuvant settings resected NSCLC, the role of individualized anti-PD-L1/PD-1 immunotherapy merits further investigated.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

AUTHOR CONTRIBUTIONS
TS, SZ, and HG retrieved and analyzed all of the data in the study. XFL, YW, XL, and SKZ revised the manuscript for important intellectual contents. LP and DH reviewed and edited the manuscript. SX, ZL, and LP designed, checked, and supervised all study processes. All authors contributed to the article and approved the submitted version.

FUNDING
The present study was funded by the National Natural Science