Clinicopathological and prognostic significance of PD-L1 expression in colorectal cancer: a meta-analysis

Purpose To systematically evaluate the correlation between PD-L1 expression and clinicopathological features and prognosis of colorectal cancer (CRC). Methods Seven databases (PubMed, Cochrane Library, EMBASE, Web of Science, CBM, Wanfang, and CNKI) were searched through May 2020. Risk of bias and quality of evidence were assessed by using the Newcastle–Ottawa scale (NOS), and meta-analysis was carried out by using the Review Manager 5.3 software on the studies with the quality evaluation scores ≥ 6. Meta-regression analysis was used to determine the independent role of PD-L1 expression on CRC prognosis after adjusting clinicopathological features and treatment methods. Results A total of 8823 CRC patients in 32 eligible studies. PD-L1 expression was correlated with lymphatic metastasis (yes/no; OR = 1.24, 95% CI (1.11, 1.38)), diameter of tumor (≥ 5 cm/< 5 cm; OR = 1.34, 95% CI (1.06, 1.70)), differentiation (high–middle/low; OR = 0.68, 95% CI (0.53, 0.87)), and vascular invasion (yes/no; OR = 0.80, 95% CI (0.69, 0.92)). PD-L1 expression shortened the overall survival (hazard ratio (HR) = 1.93, 95% CI (1.66, 2.25)), disease-free survival (HR = 1.76, 95% CI (1.50, 2.07)), and progression-free survival (HR = 1.93, 95% CI (1.55, 2.41)). Meta-regression showed that PD-L1 expression played a significant role on poor CRC OS (HR = 1.95, 95% CI (1.92, 3.98)) and disease-free survival (HR = 2.14, 95% CI (0.73, 4.52)). Conclusion PD-L1 expression independently predicted a poor prognosis of CRC. Electronic supplementary material The online version of this article (10.1007/s00384-020-03734-4) contains supplementary material, which is available to authorized users.


Introduction
Colorectal cancer (CRC) is one of the most common malignant tumors of the digestive system all around the world [1]. Its incidence and mortality rate ranked third and second in the world, respectively [2]. In 2018, both new cases and deaths were close to 30% of the total number of CRC cases in the world [3,4]. China's cancer statistics indicated that the incidence and mortality of CRC ranked fifth among all malignant tumors in China, bringing about 380,000 new cases and 190,000 deaths annually [5]. Furthermore, most patients have already been in the severe stage when they were seeking the medical examination [6,7]. Thus, it has become a major public health problem in many countries [8,9]. Surgery, chemotherapy, and radiation therapy are the main treatments for cancer; unfortunately, the recurrence rate and metastasis rate (approximately 30% and 10%) in advanced CRC patients still remain high [10,11]. In addition, some treatments showed only mild effects in reducing tumor load, such as cytokine therapy, toll-like receptors, and autologous cell therapy [12]. In recent years, immune card control point drugs have provided a new therapy for CRC, especially the programmed death 1 (PD-1)/programmed death ligand-1(PD-L1) monoclonal antibody as an immunodetection point inhibitor and an antibody-type tumor immune drug [13,14]. PD-L1, also known as CD274 or B7-H1, is the ligand PD-1 and a sort of immune checkpoint inhibitors and belongs to the CD28 family and is expressed on the surface of activated T cells to regulate proliferation and activation [15]. The binding of PD-L1 on tumor cells to PD-1 on lymphocytes can lead to immune escape of tumor cells and ultimately promote the generation and development of tumors by inhibiting the release of cytokines, restricting lymphocyte function, and inducing lymphocyte apoptosis [16]. It was reported that PD-L1 correlated with the clinicopathological features and affected the prognosis of cancers (such as breast, gastric, and ovarian cancers) [17][18][19].
The correlation between PD-L1 expression and clinicopathological features of CRC was inconsistent, and the independent impacts of PD-L1 expression on CRC prognosis were unclear in the previous meta-analyses [20][21][22][23]. Additionally, some limitations reduced the reliability because of small sample sizes [21,23] or the high heterogeneity [21,23] or incorrect model selection [21,23]. Thus, we aimed to update a meta-analysis of cohort studies to confirm the correlation between PD-L1 expression and clinicopathological features, and perform a meta-regression analysis to determine the independent role of PD-L1 on CRC prognosis after adjusting confounders.

Search strategy
Seven databases (PubMed, Cochrane Library, EMBASE, Web of Science, CBM, Wanfang, and CNKI) were searched through May 2020, and the search strategies were ("PD-L1" OR" B7-H1" OR "Programmed Cell Death Ligand 1" OR "CD274" OR "PD-1" OR "Programmed death 1") AND ("Colorectal Cancer" OR "Colorectal Neoplasm" OR "Colorectal Tumor" OR "Colorectal Carcinoma" OR "Colorectal Cancer" OR "Rectal Cancer" OR "Colon Cancer" OR "Rectal Neoplasm" OR "Colon Neoplasm"). Furthermore, we reviewed the reference list of original and review articles to search for more studies. Only studies that were published as full articles and in Chinese and English were considered.

Inclusion and exclusion criteria
Inclusion criteria for study enrollment were (1) cohort studies; (2) patients had confirmed colorectal cancer; (3) PD-L1 expression detected method: immunohistochemistry (IHC); (4) the literature provides the relationship between PD-L1 expression and clinicopathological features, such as sex, age, lymphatic metastasis, differentiation, TNM stage, and tumor location; (5) studies that provided detailed pathological parameters and survival

Data extraction
Two researchers (Shuxia Wang and Yun Wang) identified and classified the literature that met the inclusion criteria independently and excluded the study that obviously did not meet the inclusion criteria after reading the full text. For studies with insufficient information, we contacted the primary authors to acquire and verify data when possible. In cases of disagreement, the two researchers can make an attempt to reach a consensus. We extracted these objective data which were analyzed for aims of this study : (1) the basic information of the study including first author, year of publication, country, number of subjects, their demographic features, (2) type of study, (3) treatment method, (4) outcomes including the pathological parameters (sex, age, tumor location, TNM stage, lymphatic metastasis, differentiation, infiltration degree, tumor diameter, distant metastasis, and vascular invasion), and (5) prognostic values including overall survival (OS), diseasefree survival (DFS) and progression-free survival (PFS).

Quality assessment
Study quality was assessed by using the Newcastle-Ottawa score [24], which consists of three factors: patients selection, comparability of study groups, and assessment of outcomes. A score of 0 to 9 was assigned to each study, and studies achieving a score of 6 or higher were considered high quality.

Statistical analysis
If the numbers of included studies were less than 3, the metaanalysis could not be used. All statistical analyses were conducted by using Review Manager 5.3. Odds ratios (OR) and 95% CI were analyzed for the relationship between PD-L1 expression and basic clinicopathological features including sex (male/female), age (≥ 60/< 60 years old), tumor location (right + rectum/left + colon), TNM stage (III-IV/I-II), lymphatic metastasis (yes/no), differentiation (high-middle/low), tumor diameter (≥ 5 cm/< 5 cm), vascular invasion (yes/no), infiltration degree (3-4/1-2), and distant metastasis (yes/no). Hazard ratio (HR) and its 95% CI were presented for PD-L1 on CRC prognosis. Subgroup analysis was used to find the source of heterogeneity according to treatment methods (surgery or surgery combined with chemoradiotherapy (CRT)). Moreover, meta-regression analysis was used to analyze the independent role of PD-L1 on the prognosis of CRC after adjusting for above clinicopathological features and treatment methods. If the numbers of included studies were less than 10, the meta-regression could not be used. Depending on the results from the tests of heterogeneity, a fixed effect model or a

Description of studies and quality assessment
Thirty-two eligible studies  with Newcastle-Ottawa scale (NOS) score ≥ 6 were included in meta-analysis, including five in Chinese and twenty-seven in English, with a total of 8823 CRC patients. The follow-up duration was from 4 months to 7.3 years, and the sample size was from 65 to 1414. The selection process of literature is detailed in Fig. 1.
Basic information and quality evaluation of included studies are presented in Table 1 and Table 2.

Meta-regression analysis
Meta-regression analysis confirmed that PD-L1 expression was to be correlated with OS (HR = 1.95, 95% CI (1.92, 3.98)) and DFS (HR = 2.14, 95% CI (0.73, 4.52)). And the prognosis of patients with surgery treatment alone was worse than that of surgery combined with CRT. Patients with distant metastasis had a poor prognosis (Table 3).

Sensitivity analysis
Sensitivity analysis on OS, DFS, and PFS indicated that after excluding any single study individually, there was no separate study that significantly affected HR and 95% CI, suggesting that the results of this meta-analysis were stable (Fig. S3).

Publication bias
Results of Begg's test suggested that there may be no publication bias among studies for OS, DFS, and PFS (all P > 0.05) (Fig. 6).

Discussion
Studies reported that the PD-1/PD-L1 pathway has become a promising therapeutic target for various human malignancies [17,18,[57][58][59][60]. Nonetheless, the correlation between PD-L1 expression and clinicopathological features [26,30] and the prognosis of CRC patients are still controversial [36,51]. Therefore, this study comprehensively searched the literature to solve the above-existing controversies in order to draw more reliable conclusions.
Data of our meta-analysis from 32 studies (8823 CRC patients), the largest to date, indicated that PD-L1 expression was significantly positively correlated with lymphatic metastasis and tumor diameter, but negatively correlated with differentiation and vascular invasion. However previous meta-analysis found that PD-L1 expression was correlated with tumor stage [21] and gender [22] and tumor location [23], which results were  [21][22][23] and the incorrect analytical model (all selected the fixed effects model that is available for I 2 < 50%) [21][22][23]. In this study, the random effects model was selected for TNM stage and tumor location because of mild heterogeneity (I 2 = 57% for TNM stage and I 2 = 65% for tumor location).
In univariate analysis, PD-L1 was correlated with poor prognosis of CRC in this study, which was similar to the results of previous meta-analysis [20][21][22][23]. However, high heterogeneity existed in our study and those meta-analyses [20][21][22][23]. Furthermore, in subgroup analysis based on treatment, we found that the degree of statistical heterogeneity reduced both in subgroup for OS (Fig. 5). It meant that the treatment method was the source of heterogeneity for OS. In order to control other confounders, meta-analysis should be necessary to analyze the independent role of PD-L1 on CRC prognosis. We found that PD-L1 expression independently predicted a poor prognostic outcome with meta-regression analysis. Previous meta-analysis made a contradictory conclusion by univariate analysis [20][21][22][23]. Meta-regression analysis can get a more reliable and accurate outcome after adjusting confounders including clinicopathological features and treatment methods that influence the CRC prognosis.
In our sensitivity analysis, none of the inclusions and exclusions of specific studies one by one materially changed the results of the primary meta-analysis; it suggested that the results of this meta-analysis were stable.
From the perspective of publication bias, Begg's test on OS, DFS, and PFS found that there was no significant publication bias that existed among included studies, and the results of this study were relatively reliable.
Despite some positive findings from this meta-analysis, two limitations still existed to our study. Firstly, although Chinese and English studies were included in this meta-analysis, language bias still existed. Secondly, although the literature screening was carried out with a strict search strategy, a small number of literatures including gray literature and conference literature may still be missing.

Conclusions
In summary, PD-L1 expression was significant correlated with lymphatic metastasis, tumor diameter, differentiation, and vascular invasion, and could act as an independently poor prognostic factor for CRC.
Authors' contributions All authors have read and approved the final manuscript. The corresponding author of this manuscript is Jihong Hu, and contribution of the authors was mentioned below with their responsibility in the research. All authors of this research paper have directly participated in the planning, execution, or analysis and interpretation of data and have read and approved the final version submitted. The contribution of each author is as follows: conceptualization, Jihong Hu; data curation, Shuxia Wang, Xibo Liu, and Mingyang Li; formal analysis, Jing Cao and Bo Yuan; methodology, Changtian Li and Yun Wang; writingoriginal draft, Shuxia Wang; designed the research and revised, Jihong Hu.
Funding The research project was funded and supervised by the National Natural Science Foundation of China (NO.81960614).
Data availability The data supporting this meta-analysis are from previously published studies, which have been cited. The processed data are available from the first author (wangshuxialucky@163.com) upon request.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.