Background
Among the malignant tumors, lung cancer, a global public health issue, is the most frequent cause of tumor-related deaths globally [
1]. Small cell lung cancer and non-small cell lung cancer (NSCLC) are the two types of lung cancer based on their histology, with NSCLC accounting for about 85% of all lung cancer cases [
1]. Although there have been recent breakthroughs in the diagnosis and treatment of cancer, the overall 5-year survival rate for all NSCLC patients remains at 24.6%. Patients with stage I and stage IV have 5-year survival rates of 65% and 5%, respectively [
2].
Immune checkpoint inhibitors (ICIs) have improved the survival outcomes of patients with cancer [
3]. The US Food and Drug Administration recently approved ICIs to manage NSCLC [
4]. Currently, ICIs serve as the first-line and second-line treatments for NSCLC [
5,
6]. ICIs have been developed to target immune checkpoint molecules, such as programmed cell death protein 1 (PD-1) and programmed cell death ligand 1 (PD-L1), which are critical for regulating anti-tumor immune responses [
7]. The interaction of PD-L1 with PD-1 suppresses T-cell activity and proliferation, resulting in immunological tolerance. Pembrolizumab, an anti-PD-1 antibody, is used to treat patients with NSCLC exhibiting high PD-L1 expression in tumor tissue [
8]. However, clinical studies have reported that only 15%–30% of patients achieve favorable outcomes with ICIs [
9]. Some patients experience immune-related adverse reactions [
10]. Therefore, it is essential to identify biomarkers to assess treatment efficacy and disease prognosis, as well as to screen patients who will benefit from ICI therapy. Tissue PD-L1 (tPD-L1) expression is the most common biomarker for assessing the efficacy and prognosis of ICIs in NSCLC [
11]. Previous studies have suggested that highly expressed tPD-L1 improves the clinical prognosis of patients with tumors undergoing ICI therapy [
12,
13]. However, recent studies have reported that several individuals with high tPD-L1 expression do not respond to immunotherapy. In contrast, some patients with low tPD-L1 expression had long-term efficacy [
14]. A recent meta-analysis revealed that tPD-L1 alone is not an appropriate molecular biomarker for identifying eligible patients for immunotherapy in routine clinical practice [
15]. Furthermore, the detection of PD-L1 involves tissue sampling from patients with cancer. Repeated tissue biopsy analysis is challenging for patients with advanced lung cancer. Cancer develops over time due to dynamic molecular changes and harbors escape mutations and epigenetic modifications [
16,
17]. This can result in therapy resistance and disease progression, which cannot be evaluated using a single biopsy.
The use of blood-based biopsies to analyze PD-L1 expression in patients with cancer can overcome the challenges associated with the use of solid tumor biopsies. Compared with tissue biopsies, blood-based biopsy techniques have several advantages, including ease of sampling and reproducible monitoring of the dynamic biological changes in vivo [
18]. Some studies have reported that tumor cells and immune cells can both release PD-L1 into the bloodstream [
19]. Recent studies have revealed that three blood-based PD-L1 indicators, namely soluble PD-L1 (sPD-L1), exosomal PD-L1 (exoPD-L1), and PD-L1 in circulating tumor cells (CTCs), are linked to prognosis in patients with cancer [
20‐
22]. Several studies have shown that the elevated pre-treatment sPD-L, PD-L1 in CTCs, and exoPD-L1 levels are negatively correlated with the prognosis of patients with NSCLC undergoing immunotherapy [
23‐
25]. The dynamic upregulation of exoPD-L1 and positive PD-L1 expression in CTCs (PD-L1
+ CTCs) after immunotherapy were significantly and positively correlated with progression-free survival (PFS) [
26,
27]. However, Tiako Meyo et al. reported that upregulated post-treatment levels of sPD-L were not significantly correlated with overall survival (OS) [
28]. The roles of these three blood-based PD-L1 indicators in determining the prognosis of patients with NSCLC before and after immunotherapy are unclear. Thus, this study performed a meta-analysis to investigate if the three blood-based PD-L1 indicators are correlated with the prognosis of patients with NSCLC undergoing ICI therapy.
Methods
Protocol and registration
Search strategy
Relevant literature in the English language published up to February 23, 2023, was retrieved from the PubMed, Embase, and Cochrane Library databases. The following keywords were used: "non-small cell lung cancer," "lung cancer," "lung tumor," "lung neoplasm," "immune checkpoint inhibitor," "PD-1," "PD-L1," "anti-PD-1," "PD-L1," "programmed death 1 receptor," "programmed death ligand 1," "cytotoxic T lymphocyte antigen 4," "CTLA-4," "blood," "circulating tumor cell," "PD-L1," "soluble PD-L1," "exosomal PD-L1," "circulating PD-L1," "PD-L1 in CTCs," "sPD-L1," and "exoPD-L1." The references in the reviews were manually searched to identify additional related articles.
Inclusion criteria
Studies on patients with pathologically diagnosed primary or metastatic NSCLC; studies on patients undergoing ICI mono-therapy or ICI com-therapy (chemotherapy or targeted therapy); studies evaluating the correlation between PD-L1-based blood markers (PD-L1 in CTCs, sPD-L1, and exoPD-L1) and OS/PFS of patients with NSCLC; studies reporting the hazard ratio (HR) and 95% confidence interval (CI). If the HR was not directly available from studies, it was extracted from the Kaplan–Meier curve. When the univariate and multivariate analysis results were provided simultaneously, only the multivariate analysis results were extracted.
Exclusion criteria
Case reports, reviews, meta-analyses, conference abstracts, and letters; studies in which the full text was not available for download; studies that included animal experiments; studies in non-English language.
Quality assessment
Based on selection, comparability, and outcomes, the quality of the included studies was evaluated using the Newcastle–Ottawa scale (NOS). The NOS quality score ranged from 0 to 9. Studies with NOS scores of ≥ 7 and < 7 were categorized as high-quality and low-quality studies, respectively [
29].
Information, such as the name of the first author, the publication year, the region, the sample size, the age, the gender, the cancer stage, the cut-off value, the time point, the study design, and the outcome, was retrieved from each included study. Two authors (Qian Cui and Dong Wang) independently extracted data from univariate and multivariate analyses with HR and 95% CI for OS or PFS. Any controversy was addressed via discussion.
Statistical analysis
The HR and 95% CIs were extracted to assess the association between these three blood-based PD-L1 indicators and survival outcomes of patients with NSCLC undergoing ICI therapy. Additionally, the association between clinicopathologic features and survival outcomes was evaluated by the HR and 95% CI. The HR values of > 1 and < 1 were considered adverse and favorable prognoses, respectively. Cochran's q-test and
I2 combined values were applied to analyze the heterogeneity of the results [
30].
P > 0.10 and
I2 < 50% indicated low-level heterogeneity [
29]. Subgroup analysis was utilized to investigate the heterogeneity of the studies. Furthermore, Begg's and Egger's tests were employed to assess publication bias. Differences were considered significant at
P < 0.05. All statistical analyses were performed using Stata 15.1 software.
The HR and 95% CIs were extracted to assess the association between these three blood-based PD-L1 indicators and survival outcomes of patients with NSCLC undergoing ICI therapy. Additionally, the association between clinicopathologic features and survival outcomes were evaluated by the HR and 95% CI.
Discussion
Cancers hide from the immune system through PD-L1 upregulation, which is associated with a poor prognosis [
39]. Additionally, the activation of the PD-1/PD-L1 pathway can lead to a series of alterations in the T-cell response, including an enhanced threshold of activated T-cells, reducing their proliferation and promoting their exhaustion and apoptosis [
40]. This weakens the anti-cancer immune response, allowing tumors to acquire immune escape capacity. PD-1/PD-L1 pathway inhibitors block the communication between PD-1 and PD-L1, improving the survival of patients with cancer by restoring the immune response against tumor cells [
41,
42]. However, the clinical efficacy of ICIs is unsatisfactory, with some patients exhibiting hyper-progression, leading to fatality. Recently, PD-L1 has been widely studied as a promising tumor marker. The tPD-L1 expression level is considered a predictive biomarker for the efficacy of ICIs [
43]. However, in patients with advanced NSCLC, performing reproducible detection is challenging as the tumor tissues are difficult to obtain. The emergence of precision medicine has enabled the analysis of molecular biomarkers based on liquid biopsies owing to their portability and reproducibility, which can overcome the spatial and temporal heterogeneity of malignancies. However, the predictive and prognostic values of blood-based PD-L1 indicators in patients with NSCLC undergoing ICI therapy are unclear, warranting a systematic meta-analysis. This is the first systematic and comprehensive analysis of the pre-treatment and post-treatment dynamic changes in the three blood-based PD-L1 biomarkers in patients with NSCLC undergoing ICI therapy. Although some studies have performed meta-analyses of sPD-L1 and PD-L1 in CTCs, they collectively analyzed various tumors [
44,
45]. These studies did not individually analyze patients with NSCLC, especially the dynamic changes before and after the application of ICIs. In terms of exoPD-L1, an analogous meta-analysis of exoPD-L1 has not been previously performed. The evaluation of dynamic PD-L1 levels before and after treatment is essential for patients with cancer undergoing ICI therapy. Therefore, it is particularly vital that the levels of blood-based PD-L1 are evaluated at different time points. The strength of this study is that subgroup analyses of the included studies were performed before and after immunotherapy. Meanwhile, the correlation between each marker and prognosis at different treatment times was elucidated.
A review of the literature identified 14 trials with 928 NSCLC cases. In this meta-analysis, seven studies on sPD-L1 involving 554 patients with NSCLC, six studies on PD-L1 in CTCs involving 247 patients with NSCLC, and two studies on exoPD-L1 involving 171 patients with NSCLC were included to investigate the prognostic effect of these blood markers in patients with NSCLC undergoing ICI therapy.
The analysis revealed that high levels of sPD-L1 were significantly correlated with worse OS (HR = 2.11, 95% CI 1.59–2.80,
P < 0.001) and PFS (HR = 2.14, 95% CI 1.59–2.88,
P < 0.001). The subgroup analysis results revealed that high pre-treatment levels of sPD-L1 were significantly associated with poorer OS (HR = 2.32, 95% CI 1.68–3.18,
P < 0.001) and PFS (HR = 2.52, 95% CI 1.72–3.68,
P < 0.001). However, after immunotherapy, the upregulation of sPD-L1 was not significantly correlated with OS (HR = 1.46, 95% CI = 0.65–3.26,
P > 0.05) or PFS (HR = 1.62, 95% CI = 0.92–2.86,
P > 0.05). This result suggests that high pre-treatment levels of sPD-L1 can be considered a biomarker to predict the prognosis of NSCLC patients receiving ICIs. These findings are consistent with those of Han et al., who reported that lower pre-treatment sPD-L1 levels predicted improved PFS and OS [
46]. Additionally, another study demonstrated that high baseline levels of sPD-L1 were associated with poor OS in patients who initially received immunotherapy [
47]. sPD-L1, a predictor of tumor proliferation, migration, and survival, is reported to be a potential predictive and prognostic biomarker for immunotherapy [
48]. However, the dynamic changes in post-treatment sPD-L1 were not associated with prognosis. This meta-analysis included three studies related to post-treatment. Costantini et al. demonstrated that high sPD-L1 levels in NSCLC patients receiving nivolumab were associated with adverse effects and a poorer prognosis [
31]. However, Tiako Meyo et al. demonstrated that the upregulated sPD-L1 levels were not significantly associated with PFS and OS in patients treated with nivolumab. Interesting, the authors proposed a composite biomarker, namely the combination of sPD-1 and sPD-L1, to predict the efficacy and prognosis of nivolumab in patients with NSCLC [
28]. This provides new insights for future studies. Yang et al. suggested that the dynamic changes of sPD-L1 were not correlated with treatment efficacy and OS but were significantly linked to PD-L1 mRNA and exoPD-L1 levels [
26].
Recently, several studies have proposed that high expression of PD-L1 in CTCs is associated with a poorer prognosis in patients with various cancers treated with other therapies, such as chemotherapy or surgery [
22,
49,
50]. In contrast, PD-L1 upregulation in CTCs predicts a favorable prognosis in patients with cancer undergoing ICIs [
51,
52]. However, the pooled results of this analysis indicated no significant correlation between PD-L1
+ CTCs and OS (HR = 0.75, 95% CI 0.49–1.13,
P = 0.170) or PFS (HR = 0.79, 95% CI 0.59–1.06,
P = 0.12). Furthermore, pre-treatment and post-treatment subgroup analyses did not reveal significant differences. These results suggest that PD-L1
+ CTCs may not be correlated with the prognosis of patients with NSCLC undergoing ICI therapy. Of the six included studies, three were pre-immunotherapy studies. Guibert et al. demonstrated that the PD-L1
+ CTCs before immunotherapy were not significantly correlated with prognosis, but this correlation was significant in patients with treatment failure and disease progression [
25]. Dhar et al. reported no significant correlation between PD-L1
+ CTCs before immunotherapy and PFS [
35]. This can be attributed to the decreased number of patients included in the study (
n = 17). However, Zhang et al. demonstrated that pre-treatment PD-L1
+ CTCs were positively correlated with PFS and OS in patients with NSCLC treated with immunotherapy plus chemotherapy [
38]. The results of these two studies were contrasting. We consider that the heterogeneity may be attributable to two factors. First, the treatment protocols were different in the two studies. Guibert et al. utilized nivolumab, whereas Zhang et al. used sintilimab plus docetaxel. Second, the cut-off values of the two studies were markedly different (Table
1). Three studies examined dynamic changes after immunotherapy. Dall'Olio et al. and Ikeda et al. revealed that, compared to patients with negative PD-L1 expression in CTCs, PD-L1
+ CTCs have significantly longer PFS after immunotherapy [
27,
37]. However, Zhang et al. demonstrated that PD-L1 expression in CTCs was not associated with PFS [
36].
In this meta-analysis, only two studies related to exoPD-L1 were included. The pooled results showed that elevated pre-treatment exoPD-L1 levels were significantly linked to poorer PFS (HR = 4.44, 95% CI 2.87–6.89,
P < 0.001; Fig.
4b). However, the dynamic increases in post-treatment exoPD-L1 were associated with superior PFS (HR = 0.36, 95% CI 0.24–0.54,
P < 0.001) and OS (HR = 0.20, 95% CI 0.07–0.53,
P < 0.001). In conclusion, the role of exoPD-L1 in determining the prognosis was reversed with the increase in treatment duration. Similarly, one study on solid tumors treated with anti-PD-1 therapy reported that the high pre-treatment exoPD-L1 levels predicted a short PFS [
53]. Exosomes belong to one of the subgroups of extracellular vesicles (EVs). EVs, which are lipid-bound vesicles secreted by cells into the extracellular space, comprise the following three primary subtypes: microvesicles (MVs), exosomes, and apoptotic bodies. These subtypes can be differentiated based on their release pathways, dimensions, contents, and functions [
54‐
56]. Tumor cells can utilize EVs to evade immune responses and promote cancer progression and metastasis [
57]. Additionally, tumor cells can secrete a large amount of PD-L1 through EVs. ExoPD-L1 binds to PD-1 through its extracellular domain, and the removal of exoPD-L1 can inhibit tumor growth [
58]. Chen et al. found that exoPD-L1 can interact with activated T-cells [
59]. Circulating exoPD-L1 was closely associated with cancer progression and immune suppression in patients with cancer. This is consistent with the findings of some reviews [
60]. However, elevated exoPD-L1 levels after several treatment cycles predicted a good prognosis in patients with NSCLC. Similarly, some studies found that the higher the level of exoPD-L1 increased after applying PD-1 antibodies, the longer the prognosis of cancer patients [
24,
26,
59]. Several studies have also confirmed an increase in tPD-L1 expression in the early stages of treatment in patients who respond to ICIs [
61]. During the early stages of immunotherapy, PD-L1 expression in tissues and exoPD-L1 in circulation may be upregulated. This is because immunotherapy blocks the interaction between PD-1 and PD-L1, leading to the release of most PD-L1 through exosomes. Therefore, increased PD-L1 expression after immunotherapy may exert a positive immune regulatory effect. However, de Miguel-Perez et al. found that the dynamic increase of EV PD-L1 in the blood of patients with NSCLC was strongly associated with treatment failure and decreased survival [
62]. This finding is in contrast to that of this study. The reason for the heterogeneity can be attributed to two factors. First, Yang et al. and Wang et al. detected exoPD-L1 using the Simoa platform and enzyme-linked immunosorbent assay, respectively. However, de Miguel-Perez et al. used the immunoblotting method for the detection of EV PD-L1. Second, two studies included in this meta-analysis analyzed PD-L1 in the exosomes, whereas de Miguel-Perez et al. analyzed PD-L1 in EV. Recent studies have confirmed that PD-L1 can be detected in both exosomes and MVs [
59,
63]. Exosomes are only one subtype of EVs, and the effect of MVs on the prognosis has not been previously reported. In addition, Del Re et al. evaluated the relationship between PD-L1 mRNA in plasma-derived exosomes and the efficacy of ICIs in patients with NSCLC. The results showed that the upregulation of exosomal PD-L1 mRNA was significantly and negatively associated with treatment efficacy [
63]. Notably, recent studies have suggested that the PD-L1 mRNA level is distinct from the protein level, owing to post-translational regulatory mechanisms [
58,
64].
Limitations
This meta-analysis has several limitations. First, the number of included studies and the sample size were small. Small-sample studies may overestimate the results of efficacy and prognostic evaluations. Second, the pooled results revealed that post-treatment sPD-L1 levels and PD-L1 expression in CTCs were not correlated with the prognosis of patients with NSCLC undergoing ICI therapy. This can be attributed to the high heterogeneity among the studies. Most studies have no definite consensus on the cut-off values of biomarkers and utilize various assay platforms, which may have contributed to the high heterogeneity of these studies and biased outcomes. Therefore, a criterion threshold must be established through a uniform and standardized protocol. In the future, the correlation of PD-L1 expression in CTCs and post-treatment sPD-L1 levels with the prognosis of NSCLC patients receiving ICI therapy needs to be validated using large-sample and multi-center clinical trials. Third, only three indicators were included in the analysis. EVs and some immune cells in the blood also contain PD-L1, such as dendritic cells and myeloid-derived suppressor cells [
65], which were not included in this meta-analysis. In the following work, we plan to systematically analyze all blood-based PD-L1 markers and apply more analytical tools such as bioinformatic analysis and Mendelian randomized trials to screen valuable signature genes. Fourth, most of the included studies analyzed only individual markers but not multi-target or multi-gene combinations. Furthermore, these three markers were not validated in clinical trials. In the future, clinical studies combining multiple markers will be performed to validate the characteristic markers.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.