PD-L1 and Survival in Solid Tumors: A Meta-Analysis

Background Numerous agents targeting PD-L1/PD-1 check-point are in clinical development. However, the correlation between PD-L1expression and prognosis of solid tumor is still in controversial. Here, we elicit a systematic review and meta-analysis to investigate the potential value of PD-L1 in the prognostic prediction in human solid tumors. Methods Electronic databases were searched for studies evaluating the expression of PD-L1 and overall survival (OS) of patients with solid tumors. Odds ratios (ORs) from individual studies were calculated and pooled by using a random-effect model, and heterogeneity and publication bias analyses were also performed. Results A total of 3107 patients with solid tumor from 28 published studies were included in the meta-analysis. The median percentage of solid tumors with PD-L1 overexpression was 52.5%. PD-L1 overexpression was associated with worse OS at both 3 years (OR = 2.43, 95% confidence interval (CI) = 1.60 to 3.70, P < 0.0001) and 5 years (OR = 2.23, 95% CI = 1.40 to 3.55, P = 0.0008) of solid tumors. Among the tumor types, PD-L1 was associated with worse 3 year-OS of esophageal cancer, gastric cancer, hepatocellular carcinoma, and urothelial cancer, and 5 year-OS of esophageal cancer, gastric cancer and colorectal cancer. Conclusions These results suggest that expression of PD-L1 is associated with worse survival in solid tumors. However, the correlations between PD-L1 and prognosis are variant among different tumor types. More studies are needed to investigate the clinical value of PD-L1 expression in prognostic prediction and treatment option.


Introduction
"programmed death-ligand 1" or "PD-L1" or "B7-H1" or "CD274" and the name of each specific solid tumor to recognize additional studies. We identified a total of 350, 346 and 249 entries, respectively. Eligibility criteria were the measurement of PD-L1 by immunohistochemistry (IHC), availability of survival data for at least 3 years, and publication in English. Studies evaluating gene expression of PD-L1 measured by polymerase chain reaction were excluded from the analyses. Citation lists of retrieved articles were manually screened to ensure sensitivity of the search strategy. Study selection was based on the association of PD-L1 and survival. Inter-reviewer agreement was assessed using Cohen's kappa coefficient. Disagreement was resolved by consensus.

Endpoints of interest
The primary endpoints were overall survival (OS) at 3 and 5 years. Tumors were classified by PD-L1 expression status using cut-offs as defined by individual studies.

Data collection process
Two authors (Pin Wu and Dang Wu) independently extracted information using predefined data abstraction forms. The following details were extracted by 2 reviewers (Pin Wu and Dang Wu): tumor type, number of patients, antibody used for the evaluation, technique used to quantify PD-L1, and cut-off to determine PD-L1 positivity. Survival data were extracted from tables or Kaplan-Meier curves for both PD-L1 negative or low (control group) and PD-L1 positive or high expression (experimental group).

Data synthesis
The relative frequency of survival at 3 and 5 years between the control and experimental groups was expressed as an odds ratio (OR) and its 95% confidence interval (CI). A number of sensitivity analyses were prespecified.

Statistical analysis
Data were extracted from the primary publications and analysed using RevMan 5.3 analysis software (Cochrane Collaboration, Copenhagen, Denmark). Estimates of ORs were weighted and pooled using the Mantel-Haenszel random effect model. Statistical heterogeneity was assessed using the Cochran's Q and I 2 statistics. Differences between subgroups were assessed using methods as previous described by Deeks et al [21]. Sensitivity analyses were carried out for different analytical methods and cut-offs for defining PD-L1 expression. All statistical tests were two-sided, and statistical significance was defined as P less than 0.05. No correction was made for multiple statistical testing.

Search results
The search results have been shown in Fig 1. The primary literature research retrieved 945 records. After screening the title of citations, 603 records were excluded because of the nonrelevance with the theme and duplicated literatures. Next, 314 citations were excluded after screening abstracts of the records. Then we carefully read the full text of the left citations and at last 28 studies were included.

Description of studies
We identified 28 studies using IHC techniques for the assessment of PD-L1 expression. Characteristics of included studies are shown in Table 1. Four studies evaluated urothelial cancer, two evaluated colorectal cancer, three evaluated esophageal cancer, three evaluated gastric cancer, three evaluated hepatocellular carcinoma, four evaluated lung cancer, three evaluated melanoma, two evaluated oropharyngeal squamous cell carcinoma, and one each evaluated cervical cancer, glioblastoma, malignant mesotheliomas, ovarian cancer and pancreatic cancer. A total of 3107 patients were included in these studies. Evaluation and expression of PD-L1 A description of the antibodies used in the included studies is shown in Table 2. Various antibodies were used for the evaluation of PD-L1 expression. Seven studies used clone 5H1, five studies used clone MIH1, four studies used clone 2H11, two studies used clone 27A2 or ab82059, and one each used clone 236A/E7, NBP1-03220, ab58810 or primary antibody respectively. Four studies did not report the clone of PD-L1 antibody. The cut-off value for  Esophageal cancer 73.30% NR IHC; H-score 4; Scores were given separately for the stained area and for the intensity of staining. Quantification was made as follows: 33% of the cancer cells or less, 1; more than 33% to 66% of the cancer cells, 2; and more than 66% of the cancer cells, 3. Intensity of staining was stated as absent or weak, 1; moderate, 2; and strong, 3. Each section had a final grade that derived from the multiplication of the area and intensity scores.

Discussion
PD-L1 overexpression has been observed in a substantial number of solid tumors. Moreover, numerous studies have demonstrated that PD-L1 plays a key role in cancer immune escape PD-L1 and Solid Tumors Patient Outcome [40,41]. A decade ago some studies reported that blockade of PD-L1 could improve antitumor immunity [42][43][44]. Recently, several studies show that therapies targeting PD-L1/PD-1 display clinical responses in patients with multiple cancer types expressing high levels of PD-L1 [17][18][19][45][46][47][48]. These leading studies highlight that PD-L1 may serve as a biomarker for prognostic prediction and PD-L1/PD-1 targeted treatment option in solid tumors. In this study, we metaanalyzed the published data about the expression of PD-L1 in solid tumors and their association with survival for studies that evaluated PD-L1 by IHC.
Results showed that overexpression of PD-L1 was associated with worse 3-year OS for all studies analyzed except for one study of melanoma and lung cancer, respectively [6,11]. Studies reporting 5-year OS data demonstrated that overexpression of PD-L1 is associated with worse outcome except for one study of cervical carcinoma [29], melanoma [6], oropharyngeal squamous cell carcinoma [35] and two study of lung cancer [8,11]. Among the tumor types evaluated, esophageal cancer was the tumor type most linked with worse 3-year and 5-year outcome for patients with high levels of PD-L1 [31,34]. Overall, PD-L1 overexpression was reported to be associated with worse 5-year outcome of patients with digestive tract-derived tumors such as esophageal cancer [23,31,34], gastric cancer [26,37] and colorectal cancer [7, 39]. In addition, one study on hepatocellular carcinoma showed that PD-L1 expression on macrophages in tumors was significantly higher than paired normal tissues and correlated with tumor stage [14]. These results suggest that PD-L1 has an important role in the immune escape and progress of tumors especially digestive tract-derived.
A recent study reported that epithelial-originated cancer patients with positive expression of PD-L1 on tumor tissues were associated with significantly poorer OS when compared to those with negative expression of PD-L1 [49]. Consistently, our study showed that overexpression of PD-L1 in the tumor tissues of most epithelial-originated cancer types was associated with worse 3-year and 5-year OS except for lung cancer, melanoma and oral squamous cell carcinoma. However, recent clinic trials showed that patients with melanoma and lung cancer could acquire clinical benefit from anti-PD-L1 treatment [47,48]. In our meta-analysis, we noted that the results of studies using different clone of PD-L1 antibodies were conflicting in melanoma [5,6] (S1A and S1B Fig) and lung cancer [10,11] (S2C Fig). The difference between technologies used in studies may partly account for the contradictory results. Further studies are needed to confirm the impact of antibodies on the results of studies.
Besides, the PD-1 expression state of tumor infiltrating lymphocytes is another key point of the PD-1/PD-L1-mediated tumor immune escape. A recent study suggests that pre-existing CD8 + T cells distinctly located at the invasive tumor margin are associated with expression of the PD-1/PD-L1 immune inhibitory axis and may predict response to therapy [17]. Moreover, another study showed that across multiple cancer types, responses were observed in patients with tumors expressing high levels of PD-L1, especially when PD-L1 was expressed on tumorinfiltrating immune cells [19]. The evidence implys that further study is needed to clarify the different prognostic and therapeutic prediction value of both PD-1 and PD-L1 expression on various cell types in solid tumor tissue.

Conclusions
Our analyses show that overexpression of PD-L1 in solid tumor tissues, as measured by IHC, is associated with worse prognosis in different tumor types, which suggests that the development of strategies against the PD-L1/PD-1 axis would be a promising therapeutic approach for solid tumors. Moreover, further studies are required to investigate the potential role of PD-L1 expression in solid tumors for prognostic prediction as well as PD-L1 targeted treatment decision.