A cohort study of circulating biomarkers to predict the efficacy and prognosis of immune combination therapy in non-small-cell lung cancer

Abstract

Background

Immune checkpoint inhibitors (ICIs) bring significant clinical benefits to non-small-cell lung cancer (NSCLC), and convenient peripheral blood markers are still lacking. Human circulating cytokines play an important role in tumor growth and metastasis, and exploring their value in NSCLC immunotherapy helps to achieve precise treatment of patients.

Methods

This study was a mixed design of prospective blood collection and retrospective data collection. Patients with NSCLC who received the first ICI combined with chemotherapy were included, and plasma samples were collected at baseline and after 2 cycles of treatment. MILLIPLEX MAP technology was used to detect the levels of a panel of cancer biomarkers and to explore the predictive potential of cytokines for survival and treatment response in such patients.

Results

Baseline blood samples were collected from 79 NSCLC patients in this study, and survival analysis showed that high expression of 4 cytokines, carbohydrate antigen 125 (CA125), cytokeratin 19 fragment (CYFRA 21-1), human epididymis protein 4 (HE4), and hepatocyte growth factor (HGF), was associated with shorter overall survival (OS) and progression-free survival (PFS), low levels of stem cell factor (SCF) tended to have better OS than patients with high levels of SCF, and multivariate Cox regression showed that high levels of HGF were independent risk factors for OS (HR = 1.92, 95% CI: 1.02-3.70, P = .042) and PFS (HR = 3.23, 95% CI: 1.75-5.88, P < .001). HGF was more predictive of 1-year survival and 6-month PFS than programmed death ligand 1 expression. In addition, we collected blood samples from 53 patients after 2 cycles of treatment, CYFRA 21-1, HGF, interleukin-8 (IL-8), and tumor necrosis factor-related apoptosis-inducing ligand were associated with patient survival, and patients with increased HGF after treatment had shorter survival. In patients whose tumors responded to treatment, CA125 and CYFRA 21-1 levels increased from baseline, whereas soluble apoptosis-related factor (sFas) levels decreased.

Conclusions

Soluble cytokines, especially HGF, have certain clinical value in immunotherapy combination therapy and prognosis of NSCLC patients and are worthy of validation in a larger prospective cohort and exploration of their potential mechanisms.

Implications for Practice.

High levels of HGF, CA125, CYFRA 21-1, and HE4 are associated with shorter OS and PFS, providing a new blood marker reference for efficacy prediction of immune combination therapy in NSCLC. These findings support the use of circulating cytokines as an adjunct to prognostic assessment, efficacy monitoring, and individualized therapy for immunotherapy in NSCLC.

Introduction

Lung cancer has the highest mortality rate among solid tumors worldwide, with a 5-year survival rate of less than 10% in advanced cases, and non-small-cell lung cancer (NSCLC) accounting for 80%-85% of all cases.^1–4 In recent years, immunotherapy using programmed cell death-1 (PD-1) or programmed death ligand 1 (PD-L1) has produced durable responses and prolonged survival in patients with advanced NSCLC, radically changing treatment strategies.^5–7 In particular, immunotherapy combined with platinum-based chemotherapy has become the standard first-line treatment for advanced NSCLC.^8–10 While immunotherapy combined with chemotherapy offers a wider beneficiary population compared to monotherapy, patients who achieve durable responses remain a minority.^11,12 Appropriate predictive biomarkers need to be sought to help select NSCLC patients who have a long-term survival benefit from combined immunotherapy and chemotherapy.

Biomarkers associated with the prognosis of immunotherapy for advanced NSCLC have traditionally focused on tumor association, including PD-L1 immunohistochemistry (IHC) and tumor mutation burden (TMB).¹³ PD-L1 expression and TMB seem to show promise in first-line immunotherapy monotherapy, but their predictive value for chemotherapy and immunotherapy combinations is limited.^13–19 Although intracellular markers help elucidate the process of tumorigenesis, tissue biopsies are often required, which are not readily available and hinder dynamic monitoring. Complex dynamic interactions between tumor and host may influence the efficacy of immunotherapy,²⁰ and precise and easily accessible dynamic clinical biomarkers are urgently needed to predict the efficacy of immunotherapy combined with chemotherapy for advanced NSCLC.

Cytokines are small proteins produced by a variety of immune and non-immune cells that act as molecular messengers of cell-to-cell communication and exert effective immunomodulatory effects and play critical roles in cancer immunotherapy.^21,22 Hepatocyte growth factor (HGF) is a multifunctional cytokine that is associated with cell proliferation, migration, and patient prognosis, and recent studies have shown that HGF and its receptor cellular-mesenchymal epithelial transition factor (c-Met) play an important role in the development of a variety of tumors, however the role in NSCLC immunotherapy is unknown.^23,24 The aim of this study was to explore predictive biomarkers for first-line chemoimmunotherapy in patients with advanced NSCLC by detecting soluble cytokines in plasma before and after treatment.

Methods

Study design and participants

This study used a mixed design of prospectively collected blood samples and retrospectively collected clinical data. A flowchart of the study design and participant study design is shown in Figure 1. We enrolled NSCLC patients who received at least one cycle chemotherapy combined with PD-1/PD-L1 inhibitors at Beijing Chest Hospital, Capital Medical University between December 2018 and January 2022. The inclusion criteria for this study were patients aged >18 years, histopathologically diagnosed NSCLC, stage III/IV defined according to the eighth edition of Tumor, Node, Metastasis (TNM) staging, Eastern Cooperative Oncology Group (ECOG) performance status score of 0-2, and adequate liver and kidney and cardiac function. Patients who were not eligible for PD-1/PD-L1 inhibitor administration, such as patients with active autoimmune diseases and active infections, were also excluded.

Figure 1.

Flow chart of the study design. This figure was created with BioRender.com.

This study followed Good Practice Guidelines, was approved by the Institutional Review Board and Ethics Committee of Beijing Chest Hospital, Capital Medical University, and signed a written informed consent form prior to collecting blood samples from patients.

Sample collection and cytokine determination

Patients’ peripheral blood was collected prospectively at baseline, that is, without any antineoplastic agents, and in addition, blood samples were collected whenever possible from patients after 2 cycles of immunotherapy. Blood samples were collected in EDTA tubes, centrifuged at 1000 g for 10 minutes within 30 minutes of blood collection, and plasma was collected and aliquoted and subsequently stored at −80 °C. Samples were completely thawed for analysis and mixed by vortex and centrifuge while avoiding multiple freezing/thawing.

Cytokines were assayed by MILLIPLEX MAP’s Human Circulating Cancer Biomarker Magnetic Bead Panel 1 (Cat.# HCCBP1MAG-58K) was tested, Cytokines tested included alpha-fetoprotein, carbohydrate antigen 125 (CA125), carbohydrate antigen 15-3 (CA15-3), carbohydrate antigen 19-9 (CA19-9), carcinoembryonic antigen (CEA), cytokeratin 19 fragment (CYFRA21-1), soluble apoptosis-related factor (sFas), soluble apoptosis-related factor ligand (sFasL), fibroblast growth factor 2 (FGF2), human chorionic gonadotropin (b-HCG), human epididymis protein 4 (HE4), HGF, interleukin-6 (IL-6), interleukin-8 (IL-8), serum leptin (Leptin), macrophage migration inhibitory factor, osteopontin (OPN), prolactin (Prolactin), free prostate specific antigen (free-PSA), stem cell factor (SCF), transforming growth factor α, tumor necrosis factor α (TNFα), tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), and vascular endothelial growth factor (VEGF).

Add 25 µL of standard, control, sample and 25 µL of Serum Matrix to the standard, control, and sample wells accordingly. 150 µl of antibody beads for the above cytokines were taken and placed in a mixing flask for vortex mixing, and 25 μL of mixed beads were added to each well. Seal the plate and shake well for 2 hours at room temperature in the dark. The plate was washed 3 times and 25 µL of detection antibody was added to each well. Seal the plate and shake well at room temperature for 1 hour in the dark. 25 µL streptavidin-phycoerythrin was added to each well. Seal the plate, shake well for 30 minutes at room temperature away from light, wash the plate 3 times, add 100 µL sheath fluid to each well, shake on a shaker for 5 minutes to resuspend the beads.

The MILLIPLEX MAP kit comprises standard substances that necessitate gradient dilution, alongside antibody-immobilized beads tailored to specific markers. The Luminex technology employs proprietary methods to internally color-code beads using 2 fluorescent dyes, with each bead coated with a distinct capture antibody. Upon capturing an analyte from the standard samples, a biotinylated detection antibody is introduced to complete the reaction on the surface of each bead. We utilized the Luminex liquid chip instrument to identify each individual bead, and the bioassay results were quantified based on fluorescent reporter signals. Subsequently, data analysis was conducted using xPONENT software, and the standard curve was fitted using the formula: Y = a + ((b-a)/(1 + ((x/c) ^ d)) ^ f) (Logistic 5P Weighted).

Treatment and follow-up

All enrolled patients received standard immune checkpoint inhibitor (ICI) therapy according to local clinical practice: pembrolizumab (200 mg fixed dose every 3 weeks); atezolizumab (1200 mg fixed dose every 3 weeks); sintilimab (200 mg fixed dose every 3 weeks); and camrelizumab (200 mg fixed dose every 3 weeks). Other chemotherapeutic agents were selected based on the patient’s status. Patients were treated until disease progression, intolerable toxicity, or immune maintenance therapy for 2 years. Tumor assessments were performed every 6-8 weeks after immunotherapy, with the last follow-up date being December 2023. Overall survival (OS) was defined as the interval from the start of immunotherapy to the patient’s death or last follow-up, and progression-free survival (PFS) was defined as the interval from the start of treatment to the RECIST criteria for documentation of progressive disease (PD) or death and last follow-up, along with the patient’s survival status and progression status. Response to treatment was categorized as complete response (CR), partial response (PR), stable disease (SD), or PD according to RECIST 1.1. Objective response rate (ORR) was defined as the proportion of patients who achieved CR and PR. Immune-related adverse events (ir-AEs) as defined by the National Cancer Institute Common Terminology Criteria for Adverse Events (CTC-AEs) version 5.0 were used to assess the incidence of adverse reactions. ROC curves for 1-year survival and 6-month PFS of patients were plotted and area under the ROC curve (AUC) was calculated to measure the predictive ability of different markers.

Statistical analysis

Median values divided baseline and post-treatment cytokine levels into high and low groups, in addition, changes in cytokine levels before and after treatment were recorded. Patients in the response group were defined as having a best response to treatment of CR or PR, whereas patients in the non-response group were defined as having SD or PD. Correlations between continuous variables were performed using Spearman correlation tests, Student's t-test or Mann-Whitney U-test for comparison of differences in cytokine levels between 2 groups, and χ² test or Fisher’s exact test for comparison of baseline characteristics, treatment response, and irAEs among patients. Median values of OS and PFS were calculated by Kaplan-Meier, survival differences between cytokine groups were compared using Log-Rank test, and Cox proportional hazards model was used to calculate hazard ratio (HR) and 95% CI. Cytokines were matched before and after treatment to observe the trend of cytokines in the remission group. All statistical analyses were performed using R (v.4.1.3; https://www.r-project.org/), and P < .05 was considered statistically significant.

Results

Patient characteristics

A total of 79 patients with advanced NSCLC receiving first-line chemoimmunotherapy were included in this study, and the baseline characteristics of the patients are shown in Table 1. The cohort patients had a median age of 63 (58-69) years, including 14 (17.7%) women, 62 (78.5%) smokers, and 5 (6.3%) patients with poor performance status. Lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) accounted for about half of the patient cohort, respectively, and PD-L1 expression in lung tumor tissues was assessed in all but 8 patients, with 45.6% (n = 36) of patients having PD-L1 TPS ≥ 1%. Fifty-three (67.1%) patients were in clinical stage IV and 18, 20, 21, 6, 3, and 7 patients presented with lung, pleural, bone, brain, liver, and adrenal metastasis, respectively.

Table 1.

Patient baseline characteristics.

Characteristic	Level	Overall (N = 79)
Age, years (median [IQR])		63.0 [58.0, 69.0]
Gender (%)	Female	14 (17.7)
	Male	65 (82.3)
Smoke (%)	Current/former	62 (78.5)
	Never	17 (21.5)
ECOG (%)	PS0	13 (16.5)
	PS1	61 (77.2)
	PS2	5 (6.3)
Pathology (%)	LUAD	41 (51.9)
	LUSC	38 (48.1)
PD-L1 TPS (%)	<1%	35 (44.3)
	1%-49%	11 (13.9)
	≥50%	25 (31.6)
	NA	8 (10.1)
T (%)	T1	10 (12.7)
	T2	18 (22.8)
	T3	11 (13.9)
	T4	40 (50.6)
N (%)	N0	8 (10.1)
	N1	8 (10.1)
	N2	30 (38.0)
	N3	33 (41.8)
Lung metastasis (%)	Yes	18 (22.8)
Pleura metastasis (%)	Yes	20 (25.3)
Bone metastasis (%)	Yes	21 (26.6)
Brain metastasis (%)	Yes	6 (7.6)
Liver metastasis (%)	Yes	3 (3.8)
Adrenal glands metastasis (%)	Yes	7 (8.9)
TNM (%)	III	26 (32.9)
	IV	53 (67.1)

Abbreviations: ECOG, Eastern Cooperative Oncology Group; PD-L1, programmed death ligand 1.

Relationship between baseline cytokines and other parameters

We took baseline plasma samples from 79 patients, and Figure S1 shows cytokine levels normalized for these patients. As shown in Table S1 and Figure S2, we explored its association with lung cancer pathological type, gender, and stage, with significantly lower expression of free-PSA (P = .002), higher levels of leptin (P = .004) and VEGF (P = .005) in women compared with men; meanwhile, VEGF (P = .005) levels were lower in smoking patients than in non-smoking patients; CEA (P = .025) and leptin (P = .002) baseline levels were higher in adenocarcinoma patients than in squamous cell carcinoma patients, while sFasL (P = .036), TNFα (P = .032), and TRAIL (P = .036) were lower; CA199 (P = .047) and CEA (P = .033) were statistically higher, while IL-6 (P = .032) were lower in stage IV patients compared with stage III patients. Figure S3 shows the correlation between 24 cytokines, and Figure S4 shows the correlation between these cytokines and multiple blood parameters, such as albumin (ALB), C-reactive protein (CRP), lactate dehydrogenase (LDH), white blood cells (WBC), lymphocytes (L), monocytes (M), neutrophils (N), and platelets (PLT).

Baseline cytokines predict response to cancer immunotherapy

By the time of last follow-up, a total of 43 deaths and 52 progression events had occurred, with median OS and median PFS of 24.4 (95% CI, 19.4-NR) months and 10.9 (95% CI, 7.9-16.2) months, respectively. Patients were assessed for response to treatment with an ORR of 43.0% (34/79, 0CR, 34PR) and a DCR of 93.6% (74/79, 0CR, 34PR, 40SD).

We observed a skewed distribution of baseline cytokine levels with extreme values (Figure S5). Extreme values may affect the regression coefficients if they are performed as continuous variables, reducing statistical power and making interpretation of the results difficult (Table S2). Median values were used to truncate each cytokine into high and low levels and survival analysis was performed. In univariate Cox regression analysis, high levels of CA125 (HR: 2.38, 95% CI; 1.25-4.55, P = .008), CYFRA 21-1 (HR: 1.96, 95% CI: 1.06-3.7, P = .03), HE4 (HR: 2.13, 95% CI: 1.16-4, P = .015), HGF (HR: 1.96, 95% CI: 1.08-3.57, P = .029), and SCF (HR: 1.85, 95% CI: 1.01-3.4, P = .047) had shorter OS (Figure 2A-B). However, this association was also observed in PFS for CA125 (HR: 2.5, 95% CI: 1.41-4.35, P = .002), CYFRA 21-1 (HR: 2.22, 95% CI: 1.27-3.85, P = .006), HE4 (HR: 1.85, 95% CI: 1.06-3.23, P = .03), and HGF (HR: 3.13, 95% CI: 1.72-5.88, P < .001) (Figure 2C-D). In addition, poor performance status, pleural metastasis, bone metastasis, and adrenal metastasis were found to be associated with poor patient prognosis in clinical features (Table 2). Further multivariate analysis showed that HGF was an independent predictive biomarker of response to immunotherapy in advanced NSCLC, with high levels of HGF predicting worse OS (HR: 1.92, 95% CI: 1.02-3.70, P = .042) and PFS (HR: 3.23, 95% CI: 1.75-5.88, P < .001). While pleural metastasis (HR: 2.12, 95% CI: 1.08-4.15, P = .028) and bone metastasis (HR: 3.14, 95% CI: 1.63-6.04, P = .001) were independent risk factors for patient survival, ECOG score 2 (HR: 8.51, 95% CI: 2.12-34.23, P = .003), bone metastasis (HR: 2.26, 95% CI: 1.2-4.28, P = .012), and adrenal metastasis (HR: 3.48, 95% CI: 1.15-10.56, P = .028) were independently associated with disease progression (Table 2).

Figure 2.

(A) Forest plot for univariate Cox regression of overall survival for 24 baseline cytokines grouped by median. (B) Survival curves for baseline cytokines with statistical differences in overall survival. (C) Forest plot for univariate Cox regression of progression-free survival (PFS) for 24 baseline cytokines. (D) Survival curves for baseline cytokines with statistical differences in PFS.

Table 2.

Univariate and multivariate Cox regression analyses for overall survival and progression-free survival.

Characteristics	Overall survival				Progression-free survival
	Univariate analysis		Multivariate analysis		Univariate analysis		Multivariate analysis
	P	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)
Gender (male vs female)	.107	0.55 (0.27-1.14)			.614	0.84 (0.42-1.67)
Age	.878	1 (0.96-1.03)			.236	0.98 (0.95-1.01)
Smoke (never vs current/former)	.456	1.3 (0.65-2.58)			.657	0.86 (0.45-1.66)
ECOG (PS2 vs PS0/1)	.034	3.13 (1.09-8.95)	.085	2.7 (0.87-8.38)	<.001	9.28 (2.96-29.14)	.003	8.51 (2.12-34.23)
Pathology (LUSC vs LUAD)	.711	1.12 (0.61-2.04)			.193	0.69 (0.39-1.21)
PD-L1 TPS (≥1% vs < 1%)	.344	0.73 (0.38-1.41)			.629	0.87 (0.49-1.55)
PD-L1 TPS (≥50% vs < 50%)	.105	0.53 (0.25-1.14)			.167	0.64 (0.33-1.21)
T (T2 vs T1)	.502	1.48 (0.47-4.69)			.601	0.78 (0.31-1.96)
T (T3 vs T1)	.85	0.88 (0.24-3.29)			.429	0.66 (0.24-1.84)
T (T4 vs T1)	.485	1.46 (0.5-4.24)			.593	0.8 (0.36-1.78)
N (N1 vs N0)	.97	1.03 (0.21-5.13)			.932	0.94 (0.23-3.8)
N (N2 vs N0)	.405	1.69 (0.49-5.8)			.575	1.36 (0.46-4.01)
N (N3 vs N0)	.231	2.1 (0.62-7.05)			.292	1.77 (0.61-5.12)
Lung metastasis (yes vs no)	.391	1.37 (0.67-2.79)			.561	0.81 (0.41-1.63)
Pleura metastasis (yes vs no)	.047	1.92 (1.01-3.64)	.028	2.12 (1.08-4.15)	.085	1.69 (0.93-3.06)
Bone metastasis (yes vs no)	.009	2.25 (1.22-4.16)	.001	3.14 (1.63-6.04)	.009	2.16 (1.21-3.85)	.012	2.26 (1.2-4.28)
Brain metastasis (yes vs no)	.633	1.29 (0.46-3.61)			.615	1.27 (0.5-3.21)
Liver metastasis (yes vs no)	.76	0.8 (0.19-3.32)			.831	1.14 (0.35-3.66)
Adrenal glands metastasis (yes vs no)	.163	1.95 (0.76-4.98)			<.001	6.49 (2.62-16.08)	.028	3.48 (1.15-10.56)
TNM (IV vs III)	.137	1.68 (0.85-3.35)			.217	1.47 (0.8-2.72)
AFP (high vs low)	.46	0.79 (0.43-1.45)			.589	0.86 (0.5-1.49)
b-HCG (high vs low)	.473	1.25 (0.68-2.27)			.572	1.18 (0.68-2)
CA15-3 (high vs low)	.061	1.82 (0.97-3.33)			.113	1.56 (0.9-2.78)
CA19-9 (high vs low)	.111	1.64 (0.89-2.94)			.063	1.69 (0.97-2.94)
CA125 (high vs low)	.008	2.38 (1.25-4.55)	.234	1.54 (0.76-3.13)	.002	2.5 (1.41-4.35)	.102	1.69 (0.90-3.23)
CEA (high vs low)	.093	1.69 (0.92-3.13)			.589	1.16 (0.67-2.04)
CYFRA 21-1 (high vs low)	.03	1.96 (1.06-3.7)	.264	1.61 (0.69-3.70)	.006	2.22 (1.27-3.85)	.169	1.61 (0.81-3.23)
FGF2 (high vs low)	.63	1.16 (0.64-2.13)			.4	1.27 (0.73-2.17)
Free-PSA (high vs low)	.601	0.85 (0.47-1.56)			.282	1.35 (0.78-2.33)
HE4 (high vs low)	.015	2.13 (1.16-4)	.542	1.32 (0.55-3.13)	.03	1.85 (1.06-3.23)	.163	1.59 (0.83-3.03)
HGF (high vs low)	.029	1.96 (1.08-3.57)	.042	1.92 (1.02-3.70)	<.001	3.13 (1.72-5.88)	.001	3.23 (1.75-5.88)
IL-6 (high vs low)	.647	1.15 (0.63-2.08)			.999	1 (0.58-1.72)
IL-8 (high vs low)	.148	1.56 (0.85-2.86)			.16	1.47 (0.85-2.56)
Leptin (high vs low)	.321	1.35 (0.74-2.5)			.224	1.41 (0.81-2.44)
MIF (high vs low)	.644	1.15 (0.63-2.08)			.532	1.19 (0.69-2.08)
OPN (high vs low)	.858	0.94 (0.52-1.72)			.283	1.35 (0.78-2.33)
Prolactin (high vs low)	.925	0.97 (0.53-1.75)			.868	1.05 (0.61-1.82)
SCF (high vs low)	.047	1.85 (1.01-3.45)	.323	1.49 (0.68-3.33)	.179	1.45 (0.84-2.5)
sFas (high vs low)	.883	1.04 (0.57-1.92)			.688	1.12 (0.65-1.92)
sFasL (high vs low)	.075	1.75 (0.94-3.33)			.213	1.41 (0.82-2.44)
TGFα (high vs low)	.303	1.37 (0.75-2.5)			.053	1.72 (0.99-2.94)
TNFα (high vs low)	.845	1.06 (0.58-1.92)			.471	1.22 (0.7-2.13)
TRAIL (high vs low)	.568	1.19 (0.65-2.17)			.442	0.81 (0.46-1.41)
VEGF (high vs low)	.787	1.09 (0.6-2)			.572	1.18 (0.68-2)

Abbreviations: ECOG, Eastern Cooperative Oncology Group; MIF, migration inhibitory factor; OPN, osteopontin; PD-L1, programmed death ligand 1; SCF, stem cell factor; transforming growth factor α; TNFα, tumor necrosis factor α; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand; VEGF, vascular endothelial growth factor.

PD-L1 expression responses to cancer immunotherapy

Median OS and PFS were 39.6 and 10.9 months in patients with PD-L1 TPS ≥ 1% and 22.8 and 12.2 months in patients with TPS < 1%, respectively, and no statistically significant differences were found between the 2 groups. In terms of ORR, the group of patients with TPS ≥ 1% was 47.2% compared to 40% of patients with TPS < 1% (P = .54). Although patients with PD-L1 TPS ≥ 50% showed longer median OS compared with the TPS < 50% group (NR vs 21.5 months, P = .105), statistical significance was not reached. ORR (48% vs 41.3%, P = .587) and survival analysis for PFS (11.2 vs 10.9 months, P = .167) also did not reveal a stratified difference (Figure 3A-H). Analysis of the correlation between cytokine and PD-L1 expression is an important step to investigate the mechanism of response to immunotherapy, and there was a linear significant negative correlation between plasma CEA levels and PD-L1 TPS in NSCLC patients treated with ICI (r = −0.392, P < .001), while leptin expression levels were lower in the subgroup with TPS ≥ 50% (Figure 3I-J and Figure S6). HGF had significantly better predictive power than PD-L1 TPS for 1-year survival (AUC = 0.653 and 0.534) and 6-month PFS (AUC = 0.701 and 0.402), indicating that HGF has a higher accuracy in predicting response to ICI (Figure 3K-L).

Figure 3.

(A-D) Kaplan-Meier survival curves for baseline PD-L1 levels and overall survival and progression-free survival; (E-H) correlation between baseline PD-L1 levels and objective tumor response and disease control rates; (I) correlation between baseline cytokine levels and PD-L1 TPS; (J) scatter plots for correlation between baseline HGF levels and PD-L1 TPS; (K-L) ROC curves for prediction of 1-year survival and 6-month progression-free survival of patients with HGF and PD-L1 TPS.

Cytokine changes and survival analysis before and after treatment

Subsequently, we collected 53 plasma samples after 2 cycles of treatment, and further survival analysis showed that patients with lower CYFRA 21-1 levels after treatment had longer survival than patients with higher levels (P = .015); patients with higher HGF (P = .010) and IL-8 (P = .048) levels had faster disease progression, whereas patients with higher TRAIL (P = .005) had significantly longer PFS after treatment (Figure 4A-D). Matching plasma samples at baseline and after treatment, we compared cytokine profile changes before and after treatment and explored the relationship between tumor immunotherapy responses. We observed a significant decrease in CA125 (P = .032) and CYFRA 21-1 (P = .026) levels and an increase in sFas (P = .032) levels after treatment in patients achieving objective response, whereas these changes were not significant in patients without response (Figure 4E-G and Table S3).

Figure 4.

(A) Survival curve of CYFRA21-1 level and overall survival after treatment; (B-D) survival curves of HGF, IL-8, and TRAIL level and progression-free survival after treatment; (E-G) changes of CA125, CYFRA21-1, and sFas levels before and after treatment in patients with partial response.

In addition, we recorded changes in cytokine levels before and after treatment, and box plots (Figure S7) showed patterns of changes, with extremes present less frequently. Therefore, we performed univariate Cox regression analysis of PFS directly using the continuous variable form of cytokine changes. Notably, increased HGF levels after treatment (HR: 1.32, 95% CI: 1.02-1.71, P = .038) were associated with disease progression, whereas patients with increased sFasL (HR: 0.32, 95% CI: 0.12-0.83, P = .019) and TRAIL (HR: 0.54, 95% CI: 0.33-0.89, P = .016) levels had longer survival (Table S4).

Prognostic relevance of irAE and its relationship with cytokines

A total of 34 (43.0%) patients experienced 42 irAE events, including pneumonitis in 12 patients, hypothyroidism in 11 patients, rash in 7 patients, reactive capillary hyperplasia in 5 patients, myositis in 3 patients, hepatitis in 2 patients, and pancreatitis and enteritis in 1 patient each. The cohort was divided into 2 subgroups based on whether or not patients experienced irAEs. Kaplan-Meier analysis showed no significant survival differences in PFS (P = .33) and OS (P = .35) between groups with and without irAEs (Figure 5A-B). In the irAE group, 19 (55.9%) patients achieved PR, 15 (44.1%) SD, and all patients achieved disease control; whereas in the non-irAE group, 15 (33.3%) patients achieved PR, 25 (55.6%) SD, and 5 (11.1%) PD, with an DCR of 88.9%. There was a difference in treatment response (55.9% vs 33.3%, P < .05) (Figure 5C-D). We also analyzed the association between baseline cytokines and the development of irAEs, which was statistically significant for HGF and TNFα (Figure 5E).

Figure 5.

(A-B) Survival curves for presence of immune-related adverse events and overall survival and progression-free survival; (C-D) correlation of presence of immune-related adverse events with tumor response and objective response rate. (E) Differences between groups were compared using the Mann-Whitney U test for the presence or absence of immune-related adverse events in relation to baseline cytokine levels.

Discussion

Anti-PD-1/PD-L1-based chemotherapy is the standard first-line treatment strategy for advanced NSCLC patients without EGFR mutations or ALK rearrangements, but currently commonly used biomarkers have poor reliability for predicting long-term survival. In this study, we examined peripheral blood cytokines before and during treatment and analyzed their prognostic predictive value. Overall, some cytokines have baseline differences in gender and pathological type, which are associated with blood routine parameters, and patients with different cytokine levels have different survival time and incidence of side effects after cancer immunotherapy. The tumor microenvironment is in a complex system and shows dynamic changes. The detection of peripheral blood biomarkers may help to understand the tumor-host relationship and reveal the underlying mechanisms of drug sensitivity or resistance. According to the current findings, the potential clinical utility of circulating cytokines was identified to distinguish subgroups of patients with advanced NSCLC who would benefit.

The value of the HGF marker appears promising for the following reasons: first, for NSCLC patients receiving immunochemotherapy, there was a clear difference in prognosis between patients with different HGF levels; second, multivariate Cox regression analysis of PFS and OS preserved the independent predictive role of high levels of HGF after controlling for other confounding factors; furthermore, there appeared an interesting association between baseline HGF levels and immunotherapy toxicities, that is, patients with irAEs tended to have higher levels of HGF. HGF is a ligand for c-MET, and its overexpression has demonstrated a role in promoting tumor growth in a variety of cancers, which may be associated with tumor microenvironment effects.^23,24 This signaling pathway is aberrantly activated in NSCLC cell lines and tumor tissues, and c-Met has been validated as a potential therapeutic target by small interfering RNA targeting and small molecule inhibitors.²⁵ In addition, MET amplification tends to promote PD-L1 expression in NSCLC, so the combination of MET inhibitors with ICI may be an attractive therapeutic strategy.^26,27 High plasma HGF levels specifically affected the efficacy of immunotherapy but not in the chemotherapy alone group. The HGF/MET pathway drives immunotherapeutic resistance by inhibiting CD8 + T-cell function, and MET inhibitors enhance immunotherapeutic efficacy.²⁸ Future studies further investigate the specific mechanism of HGF in immunotherapy of NSCLC and validate the reliability and clinical utility of HGF as a biomarker in a larger cohort of patients.

In addition to HGF, we found that lower levels of CA125, CYFRA 21-1, HE4, and SCF at baseline were associated with longer survival, and changes in CA125 and CYFRA 21-1 levels after treatment were associated with treatment response to ICI. Although not all of the above tumor markers are NSCLC specific, their levels are increased in NSCLC. Similar results have been observed in other studies, and multiple serum tumor markers, including CA125 and CYFRA21-1, and their dynamic changes are reliable prognostic factors in monitoring immunotherapy for advanced NSCLC and are localized as complementary tools for clinical decision-making.^29–33 HE4 is a potential new biomarker for NSCLC and has reference value in diagnosing SCLC and NSCLC and monitoring recurrence after complete resection of LUAD,^34–36 but there is a lack of relevant studies on the efficacy of immunotherapy in NSCLC. SCF activates the c-kit receptor, acts on hematopoietic progenitor cells, and is involved in disease progression in a variety of solid tumors, including NSCLC.^37–39 In this study, SCF was found to identify a subset of patients with short survival from immunotherapy for NSCLC, suggesting that SCF may serve as a potential stratification factor.

Anti-apoptosis is a universal feature of tumor cells, and induction of apoptosis is a new cancer therapeutic strategy. TNFα showed a moderate association with inflammatory markers CRP and neutrophils, and TNFα levels were higher in patients who developed irAEs, which is consistent with a role for TNF in driving the inflammatory response.⁴⁰ In addition, the TNF superfamily is involved in inducing cell death, and TRAIL and Fas are important death receptors. TRAIL receptor is a target to investigate tumor-selective apoptotic cell death, and TRAIL expression is decreased in NSCLC tumor cells compared with normal lung epithelial cells, and its overexpression induces apoptosis and exerts anti-tumor effects.^41–43 Apoptosis may exhibit synergistic antitumor immune responses, and this study found that TRAIL changes during treatment may serve as immunotherapy-based biomarkers, and TRAIL levels after immunotherapy and increases in TRAIL values are associated with longer PFS. In addition, Fas receptor binds to its corresponding ligand FasL and plays different roles in promoting tumor proliferation and apoptosis.⁴⁴ According to our analysis, elevated sFas levels after treatment have prognostic value in patients with remission, and pretreatment levels of sFasL, TNFα, and TRAIL were found to be higher in LUSC than in LUAD in our patient population, the mechanism of which is unknown and warrants further investigation.

Systemic inflammation impacts the prognosis of patients with malignancies, and we hypothesize that changes in pro-inflammatory cytokines and inflammation-based markers may predict patients’ risk of disease. IL-6 and IL-8 showed a mild positive correlation with CRP and neutrophils in this study, which confirmed the pro-inflammatory chemotaxis of interleukins. Serum IL-6, IL-8 levels are prognostic mark­ers for predicting multiple malignancies, and reversal of tumor-associated inflammation may eliminate potential immunosuppressive effects.^45–47 We found that higher levels of IL-8 after treatment were associated with NSCLC progression, and changes in IL-8 levels may reflect the inflammatory state of disease progression or may be a marker of immune system activation by immunotherapy. Future studies are needed to further explore the specific mechanism of IL-8 in the tumor microenvironment, as well as its potential as a biomarker.

VEGF promotes vascular permeability and angiogenesis during tumor cell growth.⁴⁸ And esophageal squamous cell carcinoma patients with lower VEGF levels or decreased VEGF levels after radiotherapy have been found to have a better therapeutic effect on radiotherapy.⁴⁹ Currently, VEGF/EGFR inhibitors are approved in combination with ICIs for a variety of solid tumors. In this study, VEGF levels were higher in women and never-smokers, but no correlation was found between VEGF and patient outcome, which may be related to the selection of the median as the cutoff value.

It is important to note that even in the PD-L1 TPS ≥ 50% subgroup, no predictive value for OS and PFS was shown in this cohort, possibly because small samples obscured their effect, and real-world patients had biological complexity, and the predictive efficacy of cytokines such as HGF significantly exceeded PD-L1 expression. This is consistent with the theory proposed by Assya Akli et al. that microenvironmental factors such as HGF may become more reliable predictors in subpopulations dominated by PD-L1-independent resistance mechanisms.²⁸ In addition, it is well-known that patients with brain metastasis have a poor prognosis, but there is a lack of association between brain metastasis and survival in this study. This may be because ICI activates systemic immune responses to attack the primary tumor and brain metastasis, and the systemic effects of immunotherapy, intracranial activity, and synergy with combined radiotherapy may bridge the prognostic gap in NSCLC patients with and without brain metastasis.^50,51

We focus on the cytokine profile that is simply accessible in peripheral blood and address the burden of disease imposed on patients by biopsy or even repeat biopsy. This non-invasive approach is cost-effective and identifies NSCLC populations at high risk of early progression or poor response to ICI. However, the main limitation of our study is the possible absence of critical variables caused by the mixed design, such as missing partial PD-L1 expression data. In addition, retrospective treatment decisions were heterogeneous, and patients receiving PD-L1 or PD-1 inhibitors may have a potential bias, but they had essentially the same principle of action, and univariate Cox regression did not find a survival difference between the two. In addition, different molecular profiles may influence cytokine distribution and immunotherapy outcomes in histological types, including LUAD and LUSC.

Conclusions

Analysis of peripheral blood markers showed that certain cytokine concentrations could differentiate NSCLC patients who might benefit from immunotherapy, particularly baseline HGF levels. However, this requires prospective validation of its predictive effect in larger samples to provide ideas for the development of targeted therapies in the future.

Author contributions

Yanxia Liu (Conceptualization, Data curation, Writing—original draft), Xiaomi Li (Data curation, Formal analysis, Writing—original draft), Minghang Zhang (Data curation, Formal analysis), Yuan Gao (Investigation, Methodology), Ying Wang (Investigation, Project administration), Mingming Hu(Data curation, Software), Shaofa Xu (Conceptualization, Writing—review & editing), and Tongmei Zhang (Conceptualization, Writing—review & editing)

Supplementary material

Supplementary material is available at The Oncologist online

Funding

This work was supported by the Beijing Municipal Public Welfare Development and Reform Pilot Project for Medical Research Institutes [grant number JYY2023-14 and JYY2023-15] to T.M.Z.

Ethics statement

This study followed Good Practice Guidelines, was approved by the Institutional Review Board and Ethics Committee of Beijing Chest Hospital, Capital Medical University, and signed a written informed consent form prior to collecting blood samples from patients.

Conflicts of interest

All authors declare no conflicts of interest.

Data availability

The corresponding author can provide the datasets used and/or analyzed during the current study upon reasonable request.