Hepatocellular carcinoma (HCC) accounts for 80% of cancer-related deaths worldwide, and is the second-most prevalent cancer in China [1]. Early diagnosis of HCC is often not possible due to the lack of specific symptoms, which translates to extremely poor prognosis [2, 3]. Cholangiocarcinoma(ICC) or bile duct cancer is the second most common form of reproductive HCC [4]. Although the exact pathogenesis is unclear, primary sclerosing cholangitis, parasitic infections, fibrotic polycystic epilepsy, hepatic cholestones, exposure to chemical carcinogens and viral hepatitis are common risk factors of bile duct cancer [5]. Cholangiocarcinoma is difficult to diagnose and treat, and its incidence rate is steadily increasing worldwide. The five-years survival rate for pancreatic cancer(PAAD) in developed countries is a depressing 8%, which is currently the worst amongst solid tumors [6]. The incidence of pancreatic cancer is higher in men, and most patients are diagnosed between 65 and 80 years of age [7, 8, 9]. Early diagnosis of pancreatic cancer is difficult, and it is often detected when the tumor has already metastasized.
Cancer immunotherapy is a novel treatment that involves modulation of the immune responses against cancer cells. The most common immunotherapy strategy used currently is the blocking of the immunosuppressive PD-1/PD-L1 checkpoint [10]. PD-1 is expressed mainly on lymphocytes and elicit inhibitory signals upon binding with PD-L1/L2. Although PD-L1/L2 show 60% amino acid sequence homology [11, 12], PD-L2 level is limited to activated dendritic cells, macrophagocytes and BMMCs whereas PD-L1 is more ubiquitous. The phase 2 Keynote-158 clinical trial verified the effect of the anti-PD-1 antibody pembrolizumab against cervical neoplasms, which was subsequently approved by the FDA for clinical applications [13]. In another study, 26% of the patients with terminal cervical cancer responded to the PD-1 inhibitor nivolumab, resulting in complete tumor remission [14]. There is also evidence that PD-L2 blockade in HCC, cholangiocarcinoma and pancreatic cancer may have a therapeutic effect. Our study was aimed to determine the diagnostic and prognostic relevance of PD-L2 in the aforementioned malignancies via a meta-analysis.
Randomized controlled trials (RCTs) were selected in our study according to the following criteria: (1) published in English language, (2) subjects diagnosed with HCC, cholangiocarcinoma or pancreatic cancer, (3) evaluation of PD-L2 expression levels in HCC, cholangiocarcinoma and pancreatic cancer samples, (4) relationship between PD-L2 and survival rates (OS/DFS), (5) relationship between PD-L2 and a minimum of two clinicopathological parameters, and (6) measurement of risk ratio (HR), odds ratio (OR), 95% confidence interval (95% CI) and other corresponding data. Articles lacking sufficient data were excluded to assure the study quality. In case of repetitive studies, the latest and most relevant articles were selected.
The PubMed, EMBASE and Google databases were searched for relevant studies published in English till March 4, 2021 using the following keywords: (“programmed cell death-ligand 2” OR “PD-L2” OR “B7-DC”) and (“hepatocellular carcinoma” or “hepatocellular cancer” or “HCC” or “liver cancer” or “cholangiocarcinomas” or “cholangiocellular carcinoma” or “neoplasm, pancreatic” OR “pancreatic neoplasm” OR “pancreas cancers”).
Two investigators independently extracted the first author, publication year, country, PD-L2 detection method, follow-up time and clinicopathological characteristics, and any disagreement was resolved by a third reviewer. HR and 95% CI were either acquired directly from the included studies, or calculated by the Kaplan Meier method using Engauge Digitizer version 4.1.
Two researchers independently evaluated the quality of the included studies using the 9-point Newcastle Ottawa scale (NOS), and scored according to the following aspects: (I) study group selection (four, one for each), (2) comparability (one, up to two points), (3) identification of risks or outcomes involved (three, one for each). This work is fully compliant with the PRISMA 2020 statement and PRISMA guideline, PRISMA checking list has been submitted [15, 16] and this work has been reported in accordance with PRISMA standards [16]. This work has been submitted to the Research Registry (the unique number: reviewregistry1440) and can be find here https://www.researchregistry.com/browse-the-registry#registryofsystematicreviewsmeta-analyses/. The AMSTAR 2 criteria was used to conduct a self-assessment of the quality of the systematic evaluation, and a completed AMSTAR 2 list has been uploaded [17].
Revman 5.3 software (Revman, Cochrane Collaboration) and Stata 11.0 software (Stata, College Station) were used for all data analyses. The Cochrane q-test was used to evaluate the heterogeneity between the studies. Chi-square P < 0.1 or I2 statistic >50% were indicative of significant heterogeneity, in which event the random effects model was used. Otherwise, the fixed effect model was used. The prognostic effect of PD-L2 was evaluated in terms of HR and 95% CI. HR > 1 means that higher PD-L2 expression group has poorer survival. We adopted OR and 95% CI to assess the clinical significance of PD-L2 high expression. OR >1 indicates that PD-L2 with high expression has poor prognosis. Publication bias was evaluated using Begg’s and Egger test, and P < 0.05 was considered statistically significant. Sensitivity analysis was performed by leaving out one study at a time.
Database search yielded 264 relevant articles, of which 120 duplicate studies were excluded. After excluding articles published in languages other than English, meta-analyses, reviews and basic research studies, we finally obtained 8 articles for the meta-analysis. The screening procedure is outlined in Figure 1.
The flow chart of literature selection.
Table 1 shows the main characteristics of included studies, which contains 1725 subjects. The studies were published between the years 2007 and 2019 and were all conducted in Asia, including 6 in China, one in Japan and one in Korea. Kaplan-Meier method was used to calculated HR and 95% CI of three studies.
Table 1
Characteristics of included studies.
REFERENCES | COUNTRY | METHOD | OUTCOME (HIGH/LOW) | CASE NUMBER | FOLLOW-UP TIME |
---|---|---|---|---|---|
Lijie Ma 2018 [18] | China | Tissue Microarray | OS | 498 | 100 months |
Haotian Liao 2019 [19] | China | Tissue Microarray Immunohistochemistry |
OS,DFS | 304 | 96 months |
Baoju Wang 2011 [20] | China | Immunohistochemistry | Expression | 26 | — |
Hae II Ju 2017 [21] | Korea | Immunohistochemistry | OS,DFS | 85 | 125 months |
Yiyin Zhang 2019 [22] | China | Immunohistochemistry | OS | 455 | 24 months |
Takeo Nomi 2007 [23] | Japan | Immunohistochemistry | OS | 51 | 22 months |
Qiang Gao 2009 [24] | China | Tissue Microarray Immunohistochemistry |
OS,DFS | 240 | 60 months |
Jianzhen Lin 2018 [25] | China | Immunohistochemistry | OS | 66 | 60 months |
Seven studies had the overall survival (OS) data, and did not show any significant heterogeneity (Chi2 = 2.23; P = 0.90; I2 = 0%). The pooled results of the randomized model indicated that higher PD-L2 level was related to worse outcome and shorter OS (HR = 1.77, 95% CI: 1.48–2.10, P < 0.00001; Figure 2A). As shown in Figure 2B, there was significant heterogeneity between included studies including disease-free survival (DFS; Chi2 = 3.57; P = 0.17; I2 = 44%). The pooled results indicated that high PD-L2 predicted worse DFS (HR = 1.68, 95% CI: 1.08–2.63, P = 0.02; Figure 2B).
Forrest diagram showing the HR(Hazard Ratio) of OS (A) and DFS (B). The size of the square represents the relative contribution of each study. The horizontal solid line represents the 95% CI of each study. Diamonds represent the research after the summary.
Table 2 showed the correlation between PD-L2 overexpression and clinicopathological features.
Table 2
The correlation between PD-L2 overexpression and clinicopathological features.
CHARACTERISTICS | STUDIES | CASE NUMBER | POOLED OR (95% CI) | P | HETEROGENEITY | MODEL | PUBLICATION BIAS BEGG’S P | REFERENCES | |
---|---|---|---|---|---|---|---|---|---|
I2 | P | ||||||||
Sex | 3 | 625 | 1.04 [0.70, 1.55] | 0.85 | 0% | 0.11 | Fixed | 0.158 | [18, 21, 22] |
Tumor size (<5 cm VS ≥ 5 cm) | 2 | 325 | 0.15 [0.04, 0.25] | 0.005 | 0% | 0.02 | Fixed | 1.000 | [18, 21] |
AFP (low VS high) | 2 | 325 | 0.84 [0.52, 1.35] | 0.46 | 0% | 0.15 | Fixed | 1.000 | [18, 21] |
Liver cirrhosis | 2 | 325 | 1.20 [0.65, 2.19] | 0.56 | 0% | 0.13 | Fixed | 1.000 | [18, 21] |
BCLC stage (A+B VS C+D) | 2 | 321 | 0.56 [0.34, 0.94] | 0.03 | 0% | 0.05 | Fixed | 1.000 | [18, 21] |
High levels of PD-L2 correlated with larger tumors (OR = 1.96, 95% CI: 1.22–3.15, P = 0.005; Figure 3A) and worse BCLC stage (OR = 0. 56, 95% CI: 0.34–0.94, P = 0.03; Figure 3B). No crucial relationship was observed between PD-L2 and gender, AFP or cirrhosis (Fig. S2).
Forrest plot of HR for (A) tumor size and (B) BCLC stage. “+”: positive, “–”: negative. The size of the square represents the relative weight of each study. The horizontal solid line represents the 95% CI of each study. Diamonds represent the studies after summarizing.
As shown in Figures 4 and 5, there was no obvious significant publication bias in OS and DFS analysis (Egger’s test: P = 0.548 for OS, P = 1.000 DFS). In addition, the sensitivity analysis showed that none of the trials significantly altered the OS or DFS, indicating that our estimation is reliable.
Begg’s funnel diagram showing publication bias for (A) OS and (C) DFS. (B, D) Sensitivity analysis for (B) OS and (D) DFS.
Begg’s funnel plot showing publication bias for (A) tumor size and (C) BCLC stage. (B, D) Sensitivity analysis for BCLC stage.
Compared to PD-L1, there is a very limited body of research on the clinical significance and immunotherapeutic potential of PD-L2. We assessed the prognostic effect of PD-L2 in HCC, cholangiocarcinoma and pancreatic cancer by evaluating the correlation of PD-L2 level and clinicopathological characteristics. In this study, 8 studies including 1725 subjects were obtained for further analysis.
The PD-1 deficient mice develop autoimmune diseases, which is indicative of the immunosuppressive role of this receptor. Here we identified PD-L2, and compared the Similarities of PD-L1 and PD-L2 [10]. The combination of PD-1 and PD-L1 can significantly inhibit the proliferation and cytokine production mediated by T cell receptor (TCR) mediated by CD4 + T cells. In recent years, clinical and basic researchers have done remarkable work in tumor immunotherapy and this provides a whole new approach to treatment especially PD-1/PD-L1 block.
Our study indicated that high level of PD-L2 was correlated significantly with larger tumors and worse BCLC stage. In the study Lijie Ma 2018., the tumor tissue samples of 498 HCC patients were subjected to microchip analysis and H&E staining, and reported OS and time to recurrence (TTR). Hae II Ju 2017 included 58 confirmed HCC patients, of which 23.5% expressed high levels of PD-L2 [21]. Multivariate analysis indicated that overall tumor size was related to the survival and tumor recurrence. It is necessary to further evaluate whether PD-L2 expression has a predictive value of outcome, and has the potential to develop as therapeutic target in HCC. In the studies of Yiyin Zhang 2019 and Takeo Nomi 2007, 305 and 51 pancreatic cancer patients were enrolled, the PD-L2 level in the tumor samples was analyzed by immunohistochemistry and showed that the difference in the expression patterns of PD-L1 and PD-L2 was related to the tumor microenvironment [22, 23]. A total of 66 patients with cholangiocarcinoma were followed up for 3–65 months (median follow-up 25 months) by Jianzhen Lin 2018 et al. Multivariate analysis showed that PD-1 and PD-L2 levels had no prognostic significance for OS, which contradicted our expected results. This can be attributed to the small number of samples, and the fact that the cohorts were limited to the local region [25]. Therefore, more relevant studies need to be conducted, so as to certify the effect of PD-L2 in cholangiocarcinoma.
To summarize, this meta-analysis clarified the clinical significance of PD-L2 level, and showed that PD-L2 is a potential biomarker of the response to immunotherapy. There are only 8 literatures about PD-L2 expression and hepatocellular carcinoma, cholangiocarcinoma and pancreatic cancer screened by us, further research is needed. Other patients included in the literature were removed from the tumor, and the expression of PD-L2 reflects the autoimmune function. The data indicated that PD-L2 seems to play an important role as a biomarker in poor outcome of HCC, and may provide a possibility for assessing the risk of HCC patients. According to our meta-analysis, patients with PD-L2 overexpression may be suitable for corresponding immunotherapy. Although great efforts have been made, deficiencies in the meta-analysis still need to be pointed out. Although they are not found in this work, they may partly cause publication bias. In the literature search process of this meta-analysis, we excluded some review meta-analysis, basic repeat. In addition to language, it was also deleted for other reasons (such as too small sample size or insufficient data). The literature we searched mainly focused on Asia, so we may ignore the differences of race. Finally, We hope to expand the number of patients to explore the relationship between PD-L2 and HCC, cholangiocarcinoma and pancreatic cancer. Due to the lack of relevant data, We have not been able to analyze all the clinicopathological features in this work. Therefore, in order to further understand the prognosis and clinical value of PD-L2 in HCC, cholangiocarcinoma and pancreatic cancer, a larger sample studies are still needed to conduct and ensure the universality of race and regions.
Overexpression of PD-L2 can predict poor OS and DFS in HCC, cholangiocarcinoma and pancreatic cancer patients, and is associated with larger tumors and poor BCLC stage. Therefore, PD-L2 is a promising prognostic biomarker and therapeutic target for the aforementioned malignancies, and is clinical potential needs to be validated through a large-scale, multi-center study.
The additional files for this article can be found as follows:
Supplemental Material 1AMSTAR 2. DOI: https://doi.org/10.29337/ijsonco.142.s1
Supplemental Material 2PRISMA 2020 checklist. DOI: https://doi.org/10.29337/ijsonco.142.s2
This work was supported by the Program of Excellent Doctoral (postdoctoral) of Zhongnan Hospital of Wuhan University (Grant No. ZNYB2020004), the Fundamental Research Funds for the Central Universities (2042021kf0147) and Zhongnan Hospital of Wuhan University Science, Techonology and Innovation Seed Fund. Project (CXPY2020015).
The authors have no competing interests to declare.
WJM designed the meta-analysis, and MG and YHG screened the studies. MG, YHG and JHL confirmed the criteria for inclusion of the studies, extracted usable data, and evaluated the methodological quality of these studies. MG, YHG, JHL and XC analyzed the data. YFY and WJM revised the manuscript. All authors have read and approved the final manuscript.
Meng Gao and Yonghua Guo these authors have contributed equally to this work.
Not commissioned, externally peer-reviewed.
Parkin DM, Ferlay J, Curado MP, Bray F, Edwards B, Shin HR, Forman D. Fifty years of cancer incidence: CI5 I–IX. International journal of cancer. 2010; 127: 2918–2927. DOI: https://doi.org/10.1002/ijc.25517
Kulik L, El-Serag HB. Epidemiology and Management of Hepatocellular Carcinoma. Gastroenterology. 2019; 156: 477–491.e471. DOI: https://doi.org/10.1053/j.gastro.2018.08.065
Fattovich G, Stroffolini T, Zagni I, Donato F. Hepatocellular carcinoma in cirrhosis: incidence and risk factors. Gastroenterology. 2004; 127: S35–50. DOI: https://doi.org/10.1053/j.gastro.2004.09.014
Bergquist A, Von Seth E. Epidemiology of cholangiocarcinoma, Best practice & research. Clinical gastroenterology. 2015; 29: 221–232. DOI: https://doi.org/10.1016/j.bpg.2015.02.003
Zhang H, Yang T, Wu M, Shen F. Intrahepatic cholangiocarcinoma: Epidemiology, risk factors, diagnosis and surgical management. Cancer letters. 2016; 379: 198–205. DOI: https://doi.org/10.1016/j.canlet.2015.09.008
Shelton CA, Umapathy C, Stello K, Yadav D, Whitcomb DC. Hereditary Pancreatitis in the United States: Survival and Rates of Pancreatic Cancer. The American journal of gastroenterology. 2018; 113: 1376. DOI: https://doi.org/10.1038/s41395-018-0194-5
The global, regional, and national burden of pancreatic cancer and its attributable risk factors in 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. The lancet. Gastroenterology & hepatology. 2019; 4: 934–947. DOI: https://doi.org/10.1016/S2468-1253(19)30347-4
Lowenfels AB, Maisonneuve P. Epidemiology and risk factors for pancreatic cancer, Best practice & research. Clinical gastroenterology. 2006; 20: 197–209. DOI: https://doi.org/10.1016/j.bpg.2005.10.001
Ilic M, Ilic I. Epidemiology of pancreatic cancer. World journal of gastroenterology. 2016; 22: 9694–9705. DOI: https://doi.org/10.3748/wjg.v22.i44.9694
Latchman Y, Wood CR, Chernova T, Chaudhary D, Borde M, Chernova I, Iwai Y, Long AJ, Brown JA, Nunes R, Greenfield EA, Bourque K, Boussiotis VA, Carter LL, Carreno BM, Malenkovich N, Nishimura H, Okazaki T, Honjo T, Sharpe AH, Freeman GJ. PD-L2 is a second ligand for PD-1 and inhibits T cell activation. Nature immunology. 2001; 2: 261–268. DOI: https://doi.org/10.1038/85330
Yearley JH, Gibson C, Yu N, Moon C, Murphy E, Juco J, Lunceford J, Cheng J, Chow LQM, Seiwert TY, Handa M, Tomassini JE, McClanahan T. PD-L2 Expression in Human Tumors: Relevance to Anti-PD-1 Therapy in Cancer. Clinical cancer research: an official journal of the American Association for Cancer Research. 2017; 23: 3158–3167. DOI: https://doi.org/10.1158/1078-0432.CCR-16-1761
Garcia-Diaz A, Shin DS, Moreno BH, Saco J, Escuin-Ordinas H, Rodriguez GA, Zaretsky JM, Sun L, Hugo W, Wang X, Parisi G, Saus CP, Torrejon DY, Graeber TG, Comin-Anduix B, Hu-Lieskovan S, Damoiseaux R, Lo RS, Ribas A. Interferon Receptor Signaling Pathways Regulating PD-L1 and PD-L2 Expression. Cell reports. 2017; 19: 1189–1201. DOI: https://doi.org/10.1016/j.celrep.2017.04.031
Chung HC, Ros W, Delord JP, Perets R, Italiano A, Shapira-Frommer R, Manzuk L, Piha-Paul SA, Xu L, Zeigenfuss S, Pruitt SK, Leary A. Efficacy and Safety of Pembrolizumab in Previously Treated Advanced Cervical Cancer: Results From the Phase II KEYNOTE-158 Study. Journal of clinical oncology: official journal of the American Society of Clinical Oncology. 2019; 37: 1470–1478. DOI: https://doi.org/10.1200/JCO.18.01265
Naumann RW, Hollebecque A, Meyer T, Devlin MJ, Oaknin A, Kerger J, López-Picazo JM, Machiels JP, Delord JP, Evans TRJ, Boni V, Calvo E, Topalian SL, Chen T, Soumaoro I, Li B, Gu J, Zwirtes R, Moore KN. Safety and Efficacy of Nivolumab Monotherapy in Recurrent or Metastatic Cervical, Vaginal, or Vulvar Carcinoma: Results From the Phase I/II CheckMate 358 Trial. Journal of clinical oncology: official journal of the American Society of Clinical Oncology. 2019; 37: 2825–2834. DOI: https://doi.org/10.1200/JCO.19.00739
Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Shamseer L, Tetzlaff JM, Akl EA, Brennan SE, Chou R, Glanville J, Grimshaw JM, Hróbjartsson A, Lalu MM, Li T, Loder EW, Mayo-Wilson E, McDonald S, McGuinness LA, Stewart LA, Thomas J, Tricco AC, Welch VA, Whiting P, Moher D. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ (Clinical research ed.). 2021; 372: n71. DOI: https://doi.org/10.1136/bmj.n71
Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Shamseer L, Tetzlaff JM, Akl EA, Brennan SE, Chou R, Glanville J, Grimshaw JM, Hróbjartsson A, Lalu MM, Li T, Loder EW, Mayo-Wilson E, McDonald S, McGuinness LA, Stewart LA, Thomas J, Tricco AC, Welch VA, Whiting P, Moher D. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. International journal of surgery (London, England). 2021; 88: 105906. DOI: https://doi.org/10.1016/j.ijsu.2021.105906
Shea BJ, Reeves BC, Wells G, Thuku M, Hamel C, Moran J, Moher D, Tugwell P, Welch V, Kristjansson E, Henry DA. AMSTAR 2: a critical appraisal tool for systematic reviews that include randomised or non-randomised studies of healthcare interventions, or both. BMJ (Clinical research ed.), 2017; 358: j4008. DOI: https://doi.org/10.1136/bmj.j4008
Ma LJ, Feng FL, Dong LQ, Zhang Z, Duan M, Liu LZ, Shi JY, Yang LX, Wang ZC, Zhang S, Ding ZB, Ke AW, Cao Y, Zhang XM, Zhou J, Fan J, Wang XY, Gao Q. Clinical significance of PD-1/PD-Ls gene amplification and overexpression in patients with hepatocellular carcinoma. Theranostics. 2018; 8: 5690–5702. DOI: https://doi.org/10.7150/thno.28742
Liao H, Chen W, Dai Y, Richardson JJ, Guo J, Yuan K, Zeng Y, Xie K. Expression of Programmed Cell Death-Ligands in Hepatocellular Carcinoma: Correlation With Immune Microenvironment and Survival Outcomes. Frontiers in oncology. 2019; 9: 883. DOI: https://doi.org/10.3389/fonc.2019.00883
Wang BJ, Bao JJ, Wang JZ, Wang Y, Jiang M, Xing MY, Zhang WG, Qi JY, Roggendorf M, Lu MJ, Yang DL. Immunostaining of PD-1/PD-Ls in liver tissues of patients with hepatitis and hepatocellular carcinoma. World journal of gastroenterology. 2011; 17: 3322–3329. DOI: https://doi.org/10.3748/wjg.v17.i28.3322
Jung HI, Jeong D, Ji S, Ahn TS, Bae SH, Chin S, Chung JC, Kim HC, Lee MS, Baek MJ. Overexpression of PD-L1 and PD-L2 Is Associated with Poor Prognosis in Patients with Hepatocellular Carcinoma. Cancer research and treatment. 2017; 49: 246–254. DOI: https://doi.org/10.4143/crt.2016.066
Zhang Y, Xu J, Hua J, Liu J, Liang C, Meng Q, Wei M, Zhang B, Yu X, Shi S. A PD-L2-based immune marker signature helps to predict survival in resected pancreatic ductal adenocarcinoma. Journal for immunotherapy of cancer. 2019; 7: 233. DOI: https://doi.org/10.1186/s40425-019-0703-0
Nomi T, Sho M, Akahori T, Hamada K, Kubo A, Kanehiro H, Nakamura S, Enomoto K, Yagita H, Azuma M, Nakajima Y. Clinical significance and therapeutic potential of the programmed death-1 ligand/programmed death-1 pathway in human pancreatic cancer. Clinical cancer research: an official journal of the American Association for Cancer Research. 2007; 13: 2151–2157. DOI: https://doi.org/10.1158/1078-0432.CCR-06-2746
Gao Q, Wang XY, Qiu SJ, Yamato I, Sho M, Nakajima Y, Zhou J, Li BZ, Shi YH, Xiao YS, Xu Y, Fan J. Overexpression of PD-L1 significantly associates with tumor aggressiveness and postoperative recurrence in human hepatocellular carcinoma. Clinical cancer research: an official journal of the American Association for Cancer Research. 2009; 15: 971–979. DOI: https://doi.org/10.1158/1078-0432.CCR-08-1608
Lin J, Long J, Wan X, Chen J, Bai Y, Wang A, Yang X, Wu Y, Robson SC, Sang X, Zhao H. Classification of gallbladder cancer by assessment of CD8(+) TIL and PD-L1 expression. BMC cancer. 2018; 18: 766. DOI: https://doi.org/10.1186/s12885-018-4651-8