Introduction
Pancreatic cancer is one of the malignant tumors of the digestive system characterized by a dismal prognosis and limited potential for oncologic treatment, having the fourth place in cancer related mortality. One of the major histological types of pancreatic cancer is pancreatic ductal adenocarcinoma (PDAC or PDA), which is often diagnosed at an advanced stage due to the difficulty of early detection and resistance to chemotherapy and radiotherapy [1]. In addition, undifferentiated carcinomas with osteoclast-like giant cells of the pancreas (UCOGCP) are a rare variant of PDAC, which represents about 0.4% of pancreatic carcinomas in resection specimens [2]. The overall 5-year survival rate of PDAC is 26% in resectable patients and the rate drops to 2% if the tumor was unresectable [3]. This poses a great threat to human health and survival. The reasons underlying such a poor prognosis have been postulated to relate to the advanced stage at the time of diagnosis, and resistance to standard chemotherapies as well as its ability to evade the host immune surveillance [4, 5].
PD-L1 (B7-H1) expressing on the cell membrane of several types of cells, including tumor cells, is a member of the B7 family, and it binds the PD-1 receptor to induce T cell apoptosis within various kinds of cancer tissues [6-9]. Studies show that PD-1 and its ligand (PD-L1) checkpoint pathway play a critical role in tumour immune escape and cancer immunoediting [9, 10]. Recently, increasing numbers of studies have investigated the prognostic significance of PD-L1 expression in pancreatic cancer patients [11-15], but the results are controversial.
A previous meta-analysis performed in 2017 which included 10 articles implied that PD-L1 expression was correlated with a poor overall survival outcome in PC patients (hazard ratio = 1.76, 95% CI: 1.43-2.17,
p < 0.00001) and PD-L1 could serve as a negative predictor for the overall survival of pancreatic cancer patients [16]. Another meta-analysis carried out in 2018 which included 9 studies with 993 patients showed that elevated PD-L1 expression was related to relative poor overall survival (OS) (HR = 1.63, 95% CI: 1.34-1.98, p < 0.001) and CSS (HR = 1.86, 95% CI: 1.34-2.57, p < 0.001), indicating that PD-L1 is a potential prognostic biomarker and may be helpful for clinicians to select the appropriate immunotherapy for pancreatic cancer patients [17]. However, other
studies showed non-significant prognostic value of PD-L1 in pancreatic cancer patients [15, 18, 19]. Hence, there is still insufficient evidence to implicate PD-L1 with poor prognosis in pancreatic cancer patients. Furthermore, several pertinent larger cohort studies have been published in recent years and some studies included in previous meta-analysis have been retracted by the authors. Therefore, we performed an updated meta-analysis of published studies to evaluate the prognostic value of PD-L1 expression in pancreatic cancer patients.
Material and methods
Search strategy
A systematic literature search of PubMed, EMBASE, Web of Science, Cochrane library, Scopus and Ovid was conducted. The literature was restricted to publications in English language. The following search terms were used: “PD-L1”, “programmed cell death ligand 1”, “B7-H1”, “CD274”, “pancreatic neoplasm”, “pancreatic cancer”, “pancreatic adenocarcinoma”, “prognosis”. The literature search stopped on February 10th, 2023. In addition, a recursive search of the reference articles of included studies was conducted manually to identify possible relevant articles. Studies were included or excluded based on the consensus between two authors (Chenchen Liu and Yingjie Shi) and when necessary with the assistance of Qian Wu. All selections were performed in duplicate. All analyses were performed based on previous published studies, thus no ethical approval or patient consent was needed.
Selection criteria
Two investigators (Chenchen and Lijuan Fan) independently assessed all the eligible studies and then extracted the data. We included studies that met the following inclusion criteria: (1) studies focusing on pancreatic cancer; (2) the histologic target was pancreatic tissues; (3) the association between PD-L1 expression, prognosis, and clinicopathologic features was investigated; (4) the expression of PD-L1 was categorized into high (and/or positive) and low (and/or negative) groups; (5) statistically acceptable methods of data collection and analysis; (6) hazard ratios (HRs) for survival rates and their 95% confidence intervals as well as those with enough information for calculating these data by using Tierney’s method; (7) full manuscript publication in English language. Studies were excluded by the following exclusion criteria: (1) duplicates, reviews, conference abstracts, or letters; (2) studies about PC cell lines, animal studies, and other types of cancer; (3) studies not about PD-L1; and (4) incomplete data.
Data extraction and quality assessment
The two investigators (Yingjie Shi and Chenchen Liu) extracted the data independently and any discrepancies in interpretation were resolved by consensus of all authors. Relevant articles were reviewed in full to ensure suitability according to the predefined inclusion and exclusion criteria. The following characteristics of research articles were collected: first author’s name, country, publication year, cancer type, clinical stage of tumor, sample size, methods of detecting PD-L1, survival analysis, HRs of elevated PD-L1 for OS as well as their 95% CIs.
We utilized two methods to obtain the HRs in our article. In method 1, we obtained the HRs directly from the article. In method 2, we extracted the HRs from Kaplan-Meier curves and then we reconstructed the HR estimate by extracting some survival rates at specified times from the survival curves using the Engauge Digitizer software, and next we calculate these data by Tierney’s method [20-22].
We assessed the quality of all studies included under the criteria of Newcastle-Ottawa, which contain four parts including selection (4 points), comparability (2 points), and outcome (3 points) with a score range of 0-9. The NOS assigns a maximum
of 4 questions for selection, 2 questions for comparability, and 3 questions for exposure/outcome, with a maximum 1 point for each question. Points were scored only when the data were explicitly stated. Therefore, a higher score indicates better quality, with 9 points being the highest quality. The final decision and interpretation were based on the consensus of two authors (Chenchen Liu and Xuan Sun) and when necessary with the assistance of Yingjie Shi. All selections were performed in duplicate. Points of all eligible studies scored presented in Table I, with a higher score indicating a better methodological quality.
Statistical analysis
Our meta-analysis was performed using the Stata 12.0 software. The heterogeneity between studies included was determined by the chi square-based Q test and an I2 statistics. A p-value less than 0.05 for the Q test and I2 value above 50% were considered to be significantly heterogeneous, by which condition the random effects model was adopted. If the I2 value was below 50%, the fixed effects model was applied. Publication bias was assessed using a funnel plot and Egger’s test. To test the robustness of the main results, we utilized one-way sensitivity analysis to evaluate the stability of the meta-analysis by sequentially excluding one study each time. A p-value less than 0.05 was considered statistically significant.
Results
The baseline characteristics of the included studies are summarized in Table I.
The initial search identified 178 citations, the titles and abstracts were then carefully reviewed. After excluding duplicates, 58 irrelevant studies and duplicates were then excluded, and 120 citations were considered of potential value. Then, we further browsed the title, abstract, and full text of the literature; 79 citations were excluded based on the context. The full text of these 41 articles was further assessed for eligibility and retrieved for detailed evaluation. After further evaluation, 25 of them were subsequently excluded from the meta-analysis for not meeting the predefined criteria or insufficient data and outcomes. One additional article was obtained by a manual search of different sources. Eventually, 17 articles published from April 2007 to May 2022 comprising 2669 pancreatic cancer patients were included in our meta-analysis [11, 12, 14, 15, 18, 19, 23-33]. (Figure 1 shows the study flow diagram). These studies included a total of 646 cases of PD-L1 (+) patients and 2023 PD-L1 (–) controls. Among all the studies included, eight studies were performed on pancreatic cancer (PC) patients, eight on pancreatic ductal adenocarcinoma (PDAC or PDA) patients and one on UCOGCP patients. Of the 17 studies, five studies were conducted in Japan [11, 15, 18, 27, 30], seven in China [12, 19, 23, 24, 31-33], one in France [25], one in the UK [26], two in the USA [14, 28], and one in the Czech Republic [29] (Table I). Fifteen studies used the immunohistochemical (IHC) method to detect the PD-L1 expression, Quantitative reverse transcription-polymerase chain reaction (RT-PCR) was used in one study to detect
the PD-L1 expression, and gene set enrichment analysis was used in one study (Table I).
Association between PD-L1 and patient survival in pancreatic cancer
There was no significant heterogeneity among the studies (I2 = 38.6%, p = 0.053), and we used the fixed-effects model to calculate the pooled HR (Fig. 2). Fourteen studies reported the overall survival (OS), one study reported both the OS and recurrence free survival (RFS), one study reported both the OS and progression-free survival (PFS), and one study report disease-specific survival (DSS) as well
as PFS in our meta-analysis. Finally, we choose OS as the main parameter in our meta-analysis since OS remains the gold-standard efficacy endpoint for the development of new treatments in oncology. A significant association was observed between PD-L1 and OS in pancreatic cancer patients (HR 1.902, 95% CI: 1.657-2.184) (Fig. 2). When omitting the UCOGCP, the HR was 1.949, 95% CI: 1.608-2.362. The results revealed that patients with high PD-L1 expression were more likely to have significant shorter OS. We further divided patients into different groups under the criteria of different regions and methods of detecting PD-L1 and the survival outcomes. Due to geographic differences in PC prevalence, subgroup analysis according to studies conducted in Western and Asian countries were performed. In subgroup analysis of Asian countries, twelve studies with a total of 394 cases and 1330 controls were included. Meta-analysis of these studies showed that the pooled HR 1.946, 95% CI: 1.645-2.302, I2 46.3%) (Fig. 3A).
In subgroup analysis of Western country studies, four studies with a total of 252 cases and 693 controls were included, the pooled HR 1.814, 95% CI: 1.423-2.313, I2 25.5% (Fig. 3B). In subgroup analysis based on items of methods of detecting PD-L1, the pooled HRs for IHC was 1.821, 95% CI: 1.568-2.114 (Fig. 3C), the HR was 1.863, 95% CI: 1.514-2.293 in conventional PC patients omitting the UCOGCP. Among all the studies included, 16 studies provided OS as the main outcome, while the study by Yue Zhang utilized the DSS and PFS as the main parameter, when we omit this study the pooled HR of the rest was 2.070, 95% CI: 1.783-2.402, I2 13.0%) (Fig. 3D). No significant heterogeneity was observed in all the subgroup analysis. Collectively, this meta-analysis showed that PD-L1 was an independent prognostic factor for pancreatic cancers.
Sensitivity analysis
Sensitivity analysis was performed to examine the effect of each single study on the overall meta-analysis results by omitting one study each time. The results showed that no study markedly influenced the significance of the summary HRs and the leave-one-out ORs ranged from 1.840 (95% CI: 1.598-2.117) to 2.069 (95% CI: 1.783-2.402), similar to the overall result, indicating that the pooled HR of OS was relatively reliable (Fig. 4).
Publication bias
Egger’s p-value tests were used to assess the potential publication bias in this meta-analysis. The funnel plots were unsymmetrical (Fig. 5). Significant publication bias was observed across the studies, with a p-value of 0.160 for Egger’s test (Fig. 6). Therefore, we assumed that the main results of our meta-analysis should be interpreted critically and carefully.
Discussion
Pancreatic cancer (PC) being a lethal disease with a dismal 5-year survival less than 10%, constitutes a major public health problem worldwide [1]. Given the poor prognosis of PC, it is of great importance to identify novel biomarkers, which could help to make
a diagnosis at early stage and provide more precise and valuable information for better therapy. Currently, tumor immunotherapy is a relatively new strategy in cancer treatment, the aim of which is to block the immunosuppressive effect of tumoral cells. PD-L1,
also known as B7-H1, was first cloned in 1999 [6]. The programmed cell death-1 protein receptor (PD-1)
and its ligand (PD-L1) checkpoint pathway play an important role in tumour immune escape, thereby obstructing effective immune surveillance and thus promoting tumour growth [9, 10]. Several studies have reported a poor prognosis in pancreatic patients with increased expression of PD-L1. In a study conducted by Tessier-Cloutier et al. on patients with resected pancreatic tumors with high expression of PD-L1, the results showed that PD-L1 was associated with a poor disease specific survival (DFS) [34]. The study by Tsukamoto et al. revealed that PD-L1-high patients with PDAC had poorer prognoses than PD-L1-low patients [35]. Another study carried out by Yamaki et al. in a small group using immunostaining with fluorescent phosphor-integrated dot (PID) nanoparticles showed similar results [15]. However, in another study it was demonstrated that pancreatic cancer patients with intense CD8+ TILS and
PD-1+ TILs (tumor-infiltrating lymphocytes) infiltrate had a better prognosis [36]. A previous meta-
analysis by Zhuan-Sun et al. which included 10 studies showed that PD-L1 may act as a negative predictor for the overall survival of PC patients with a pooled HR 1.76 (95% CI: 1.43-2.17), and in addition high expression of PD-L1 was correlated with poor differentiation and neural invasion [16]. Another meta-analysis which included 9 studies performed by Hu et al. demonstrated that high PD-L1 expression was associated with poor OS in patients with pancreatic cancer, and pooled HR was 1.63 (95% CI: 1.34-1.98) [17]. The prognostic significance of PD-L1 in pancreatic cancer remained inconsistent according to previous studies and the prognostic value of PD-L1 expression in pancreatic cancer was not well established. Hence, we carried out this meta-analysis to further clarify the prognostic value of PD-L1 expression in pancreatic cancers.
Our meta-analysis using a detailed search strategy which included 17 relevant studies with a total of 2669 patients, provided convincing evidence that PD-L1 expression is predictive of poor tumor survival, suggesting that PD-L1 may be used as a negative, unfavourable prognostic marker for pancreatic cancers. The combined results indicated that increased PD-L1expression was associated with a shorter OS in pancreatic cancer patients. A shorter overall survival time was observed in the patients of high PD-L1
expression compared with those of low PD-L1 expression. Subgroup analysis including region, an PD-L1 detection method showed that these factors did not alter the predictive value of PD-L1 for poor prognosis in pancreatic cancer patients and there was no evidence of statistically significant heterogeneity across the studies. Additionally, there was significant publication bias in our meta-analysis despite the fact that stable results were obtained in sensitivity analysis. There might be some explanations for this. First, the data collection may be incomplete because data from non-English language papers were not included. Second, most of the included studies reported positive results due to the fact that negative results are generally less likely to be published. Third, we only included studies with sufficient data to calculate the pooled HRs, omitting those with insufficient information for combined HRs. Thus, our results might overestimate the predictive significance of PD-L1 in prognosis of pancreatic cancer to some extent.
Nevertheless, we should note that there are several limitations in our study due to the discrete data across studies. First, not all the HRs are provided by the primary articles and we calculated some of them by reconstructing survival curves, which may cause bias, thus making the HR less accurate. Second, most of the patients included in the meta-analysis were from Asia countries, and thus our results may just represent patients from Asia. Third, the sample size was relatively small and we did not analyze the correlations between PD-L1 expression and clinicopathological characteristics of patients as the data were not complete.
In conclusion, our meta-analysis indicated that the expression of PD-L1 is associated with a worse OS in pancreatic cancer patients with a pooled HR of 1.902, 95% CI: 1.657-2.184. In addition, PD-L1 may act as a new parameter for predicting poor prognosis and a promising target for anticancer therapy in pancreatic cancer. However, considering the above limitations, larger, multi-center and higher-quality studies are recommended to further determine the prognostic value of PD-L1 expression in pancreatic cancer patients.
The authors declare no conflict of interest.
References
1. Jonathan DM, Rishi Surana, Juan WV, et al. Pancreatic cancer. Lancet 2020; 395: 2008-2020.
2.
Muraki T, Reid MD, Basturk O, et al. Undifferentiated Carcinoma With Osteoclastic Giant Cells of the Pancreas: Clinicopathologic Analysis of 38 Cases Highlights a More Protracted Clinical Course Than Currently Appreciated. Am J Surg Pathol 2016; 40: 1203-1216.
3.
Siegel RL, Miller KD, Jemal A. Cancer statistics. CA Cancer J Clin 2016; 66: 7-30.
4.
Kadera BE, Toste PA, Wu N, et al. Low expression of the E3 ubiquitin ligase CBL confers chemoresistance in human pancreatic cancer and is targeted by epidermal growth factor receptor inhibition. Clin Cancer Res 2015; 21: 157- 165.
5.
Delitto D, Perez C, Han S, et al. Downstream mediators of the intratumoral interferon response suppress antitumor immunity, induce gemcitabine resistance and associate with poor survival in human pancreatic cancer. Cancer Immunol Immunother 2015; 64: 1553-1563.
6.
Dong H, Zhu G, Tamada K, et al. A third member of the B7 family, co-stimulates T-cell proliferation and interleukin-10 secretion. Nat Med 1999; 5: 1365-1369.
7.
Dong H, Strome SE, Salomao DR, et al. Tumor-associated B7-H1 promotes T-cell apoptosis: a potential mechanism of immune evasion. Nat Med 2002; 8: 793-800.
8.
Freeman GJ, Long AJ, Iwai Y, et al. Engagement of the PD-1 immunoinhibitory receptor by a novel B7 family member leads to negative regulation of lymphocyte activation. J Exp Med 2000; 192: 1027-1034.
9.
Ishida Y, Agata Y, Shibahara K, et al. Induced expression of PD-1, a novel member of the immunoglobulin gene superfamily, upon programmed cell death. EMBO J 1992; 11: 3887-3895.
10.
Keir ME, Butte MJ, Freeman GJ, Sharpe AH. PD-1 and its ligands in tolerance and immunity. Annu Rev Immunol 2008; 26: 677-704.
11.
Nomi T, Sho M, Akahori T, et al. Clinical significance and therapeutic potential of the programmed death-1 ligand/programmed death-1 pathway in human pancreatic cancer. Clin Cancer Res 2007; 13: 2151-2157.
12.
Wang L, Ma Q, Chen X, et al. Clinical significance of B7-H1 and B7-1 expressions in pancreatic carcinoma. World J Surg 2010; 34: 1059-1065.
13.
Chen Y, Sun J, Zhao H, et al. The coexpression and clinical significance of costimulatory molecules B7-H1, B7-H3, and B7-H4 in human pancreatic cancer. Onco Targets Ther 2014; 7: 1465-1472.
14.
Hutcheson J, Balaji U, Porembka MR, et al. Immunologic and metabolic features of pancreatic ductal adenocarcinoma define prognostic subtypes of disease. Clin Cancer Res 2016; 22: 3606-3617.
15.
Yamaki S, Yanagimoto H, Tsuta K, et al. PD-L1 expression in pancreatic ductal adenocarcinoma is a poor prognostic factor in patients with high CD8+ tumor-infiltrating lymphocytes: highly sensitive detection using phosphor-integrated dot staining. Int J Clin Oncol 2017; 22: 726-733.
16.
Zhuan-Sun Y, Huang F, Feng M, et al. Prognostic value of PD-L1 overexpression for pancreatic cancer: evidence from a meta-analysis. Onco Targets Ther 2017; 10: 5005-5012.
17.
Hu Y, Chen W, Yan Z, et al. Prognostic value of PD-L1 expression in patients with pancreatic cancer: A PRISMA-compliant meta-analysis. Medicine (Baltimore) 2019; 98: e14006.
18.
Imai D, Yoshizumi T, Okano S, et al. The prognostic impact of programmed cell death ligand 1 and human leukocyte antigen class I in pancreatic cancer. Cancer Med 2017; 6: 1614-1626.
19.
Wang Y, Lin J, Cui J, et al. Prognostic value and clinicopathological features of PD-1/PD-L1 expression with mismatch repair status and desmoplastic stroma in Chinese patients with pancreatic cancer. Oncotarget 2017; 8: 9354-9365.
20.
Tierney JF, Stewart LA, Ghersi D, et al. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials 2007; 8: 16.
21.
Guyot P, Ades AE, Ouwens MJ, et al. Enhanced secondary analysis of survival data: reconstructing the data from published Kaplan-Meier survival curves. BMC Med Res Methodol 2012; 12: 9-12.
22.
Williamson PR, Smith CT, Hutton JL, et al. Aggregate data meta-analysis with time-to event outcomes. Stat Med 2002; 21: 3337-3351.
23.
Liang X, Sun J, Wu H, et al. PD-L1 in pancreatic ductal adenocarcinoma: a retrospective analysis of 373 Chinese patients using an in vitro diagnostic assay. Diagn Pathol 2018; 13: 5.
24.
Wang L, Ma Q, Li D, et al. Indoleamine 2,3-dioxygenase and B7-H1 expressions as prognostic and follow-up markers in human pancreatic carcinoma. Pathol Res Pract 2018; 214: 1309-1314.
25.
Birnbaum DJ, Finetti P, Lopresti A, et al. Prognostic value of PDL1 expression in pancreatic cancer. Oncotarget 2016; 7: 71198-71210.
26.
Diana A, Wang LM, D’Costa Z, et al. Prognostic value, localization and correlation of PD-1/PD-L1, CD8 and FOXP3 with the desmoplastic stroma in pancreatic ductal adenocarcinoma. Oncotarget 2016; 7: 40992-41004.
27.
Tsukamoto M, Imai K, Ishimoto T, et al. PD-L1 expression enhancement by infiltrating macrophage-derived tumor necrosis factor-α leads to poor pancreatic cancer prognosis. Cancer Sci 2019; 110: 310-320.
28.
Danilova L, Ho WJ, Zhu Q, et al. Programmed Cell Death Ligand-1 (PD-L1) and CD8 Expression Profiling Identify an Immunologic Subtype of Pancreatic Ductal Adenocarcinomas with Favorable Survival. Cancer Immunol Res 2019; 7: 886-895.
29.
Hrudka J, Lawrie K, Waldauf P, et al. Negative prognostic impact of PD-L1 expression in tumor cells of undifferentiated (anaplastic) carcinoma with osteoclast-like giant cells of the pancreas: study of 13 cases comparing ductal pancreatic carcinoma and review of the literature. Virchows Arch 2020; 477: 687-696.
30.
Iwatate Y, Hoshino I, Yokota H, et al. Radiogenomics for predicting p53 status, PD-L1 expression, and prognosis with machine learning in pancreatic cancer. Br J Cancer 2020; 123: 1253-1261
31.
Li J, Yin L, Chen Y, et al. PD-L1 correlated gene expression profiles and tumor infiltrating lymphocytes in pancreatic cancer. Int J Med Sci 2021; 18: 3150-3157.
32.
Zhu Q, Qiao G, Huang L, et al. Restored CD8+PD-1+ T Cells Facilitate the Response to Anti-PD-1 for Patients With Pancreatic Ductal Adenocarcinoma. Front Oncol 2022; 12: 837560.
33.
Zhang Y, Chen X, Mo S, et al. PD-L1 and PD-L2 expression in pancreatic ductal adenocarcinoma and their correlation with immune infiltrates and DNA damage response molecules. J Pathol Clin Res 2022; 8: 257-267.
34.
Tessier-Cloutier B, Kalloger SE, Al-Kandari M, et al. Programmed cell death ligand 1 cut-point is associated with reduced disease specific survival in resected pancreatic ductal adenocarcinoma. BMC Cancer 2017; 17: 618.
35.
Tsukamoto M, Imai K, Ishimoto T, et al. PD-L1 expression enhancement by infiltrating macrophage-derived tumor necrosis factor-α leads to poor pancreatic cancer prognosis. Cancer Sci 2019; 110: 310-320.
36.
Fukunaga A, Miyamoto M, Cho Y, et al. CD8+ tumor-infiltrating lymphocytes together with CD4+ tumor-infiltrating lymphocytes and dendritic cells improve the prognosis of patients with pancreatic adenocarcinoma. Pancreas 2004; 28: e26-e31.
