Biomarkers of response to camrelizumab combined with apatinib: an analysis from a phase II trial in recurrent/metastatic nasopharyngeal carcinoma

Abstract

Background

This study aims to develop a peripheral blood-based model that can predict the response to the combination therapy of camrelizumab and apatinib as a second-line or later-line treatment regimen in patients with recurrent/metastatic nasopharyngeal carcinoma (R/M-NPC).

Methods

We collected peripheral blood routine data from 72 patients with R/M-NPC from two clinical trial studies (NCT04547088, NCT04548271). Utilising the least absolute shrinkage and selection operator Cox regression model, we built a peripheral blood signature and developed a prognostic nomogram through multivariable analysis. Spectral flow cytometry analysed peripheral blood mononuclear cell immunophenotyping.

Results

Six indicators (WBC, MCV, HCT, MCHC, P-LCR, MLR) were included to construct the peripheral blood signature. By combining this signature with Epstein–Barr virus DNA, distant lymph node metastasis and previous PD-1 inhibitor treatment, we constructed a peripheral blood-based nomogram that showed favourable performance. High-risk individuals had lower overall survival than low-risk individuals (P < 0.05). Immunophenotyping revealed that the high-risk individuals had increased monocytic myeloid-derived suppressor cells, Tregs and decreased CD8 effector memory cells (P < 0.05).

Conclusions

We established a model that could predict the prognosis of combined therapy. The model could predict outcomes and reflect the systemic immune and inflammatory status, which is beneficial for risk stratification and therapeutic modification.

Background

Nasopharyngeal carcinoma (NPC) is a special type of head and neck tumour, which is characterised by distinct geographical distribution and is particularly prevalent in East and Southeast Asia [1]. Distant metastasis occurs in ~10% of newly diagnosed NPC patients, and about 30% of patients with locally advanced disease experience distant metastasis after treatment. Recurrence and distant metastasis have become the main causes of treatment failure and mortality in NPC [2]. In recent years, immune therapies, particularly programmed death-1 (PD-1) inhibitors, have shown promising results in the treatment of various haematologic malignancies and solid tumours [3, 4]. Previous studies have found intensive lymphocyte infiltration within the tumour microenvironment of NPC, with 89% to 95% of NPC cells expressing programmed death-ligand 1(PD-L1) on their surface, creating favourable conditions for immune checkpoint inhibitors in the treatment of NPC [5, 6]. Based on the results of phase III clinical trial of CAPTAIN-1st and Jupiter-02 study, Cisplatin/gemcitabine plus PD-1 inhibitor were recommended as the first-line regimens for recurrent or metastatic NPC (R/M-NPC) patients [7, 8]. However, for R/M-NPC patients who failed first-line therapy, there was no available standard-of-care treatment modality and therapeutic options remain limited.

NPC tissues contain abundant vascular distribution, and there are complex interactions between tumour immune microenvironment reprogramming and tumour vessel remodelling [9, 10]. Preclinical and clinical evidence indicated that PD-1 inhibitors combined with anti-angiogenesis exhibited promising antitumor efficacy and acceptable toxicity [11–15]. These findings lay the foundation for exploring novel and effective treatment strategies to improve outcomes and prolong survival in patients with R/M-NPC. Ding et al. reported a favourable efficacy with a median PFS of 10.4 months and an overall response rate (ORR) of 65.5% in camrelizumab combined with apatinib in R/M-NPC patients who failed first-line therapy [15]. Similarly, our phase II trial evaluated the safety and activity of the combination of camrelizumab and apatinib as a second-line or later-line treatment regimen for R/M-NPC patients. The ORR for this therapy was 65% in the platinum-resistant group (cohort 1) and 34.3% in the PD-1 inhibitor-resistant group (cohort 2), with tolerable toxicities in both cohorts [16]. Therefore, given the positive results obtained to date, identifying the patients that would benefit from this combined treatment regimen is necessary.

Over the past decade, PD-L1 expression, tumour mutational burden, RNA expression and infiltrating immune cells are widely used as biomarkers for tumour immunotherapy [17–19]. These markers are derived from tumour samples, which are difficult to obtain, especially for patients with recurrent or metastatic. In contrast, peripheral blood routine can be repeatedly obtained during treatment and provides us with valuable information on the systemic immune and inflammatory status. Moreover, compared with newly diagnosed patients, recurrent or metastatic patients are more likely to experience cachexia, metabolic disarray and immune-suppression. It is important to evaluate the systemic conditions of patients, which may be reflected in peripheral blood routine. Correspondingly, neutrophil-to-lymphocyte ratio (NLR), monocyte-to-lymphocyte ratio (MLR) and platelet-to-lymphocyte ratio (PLR) have been explored as potential predictors of immunotherapy response in a variety of solid tumours [20–22]. Whether the peripheral blood routine can predict response of combinational therapy of immunotherapy and anti-angiogenic drugs in R/M-NPC patients who failed first-line therapy remains unclear.

In this study, we collected the peripheral blood routine and clinical characteristics in the phase 2 trial and developed a peripheral blood signature-based nomogram that could predict the response to combination therapy of camrelizumab with apatinib as a second-line or later-line treatment regimen in R/M-NPC patients. The potential link between nomogram and peripheral blood mononuclear cells (PBMCs) was discussed. Notably, our study demonstrates that expansion of the myeloid-derived suppressor cells (MDSCs) and T regulatory cell (Tregs) can exert a dominant immune-suppressive effect, which points toward a window of therapeutic opportunity and thus determines successful immunotherapy. Understanding these correlative predictive markers might enhance our comprehension of the underlying mechanisms that contribute to tumour resistance against immunotherapy and anti-angiogenic treatments.

Methods

Patients

A total of 72 R/M-NPC patients underwent camrelizumab plus apatinib as a second-line or later-line treatment regimen at Sun Yat-sen University Cancer Center (SYSUCC) between 8 September 2020, and 7 September 2021 were included. All R/M-NPC patients underwent chemotherapy in prior treatment. The inclusion criteria and detailed therapy regimen could be obtained in the published study [16]. This study was approved by the institutional review board of our institution. All patients provided written informed consent prior to the collection of specimens.

Peripheral blood routine

A total of 24 peripheral blood indicators were collected from each patient at admission through the medical records platform of SYSUCC. Three peripheral blood immune inflammation indices were also calculated based on the peripheral blood indicators. The NLR was calculated by dividing the total number of neutrophils by the total number of lymphocytes. The MLR was calculated by dividing the total number of monocytes by the total number of lymphocytes. The PLR was calculated by dividing the total number of platelets by the total number of lymphocytes. Table S1 lists detailed information on the 27 peripheral blood variables.

Peripheral blood spectral flow cytometry analysis

Blood samples were collected on the day of the clinical assessments. Of the 72 enroled patients, 61 PBMC samples met quality requirements for immune cell typing analysis. For the isolation of PBMCs for immune phenotyping, the SepMate™ PBMCs Isolation Tubes with Lymphoprep™ were used to isolate PBMCs from whole blood. Purified PBMCs were cryopreserved in 10% dimethyl sulfoxide in foetal bovine serum, and stored at − 80 °C. Before the flow cytometry experiment, cryopreserved PBMCs were thawed and washed with RPMI medium containing 10% FBS, 2 mM L-glutamine, 100 units/mL penicillin and 100 µg/mL streptomycin. 1 × 10⁶ cells were stained with Fc block and the Live/Dead Stain Kit and with indicated fluorescently-labelled antibodies for 30 min at 4°C and washed with FACS buffer.

Immune cell subpopulations were defined as follows: CD4 T cells (CD3+CD4+), CD8 T cells (CD3+ CD8+), MDSCs (CD33+CD11b+HLA-DR−/low). Monocyte MDSCs (M-MDSCs, CD14+CD15-). CD8 T cells were divided into four major subgroups, which named Naive cells (N, Q2, CD3+CD8+CCR7+CD45RA+), central memory cells (CM, Q1, CD3+CD8+CCR7+CD45RA-), effector memory cells (EM, Q4, CD3+CD8+CCR7-CD45RA-) and terminally differentiated effector memory cells (TEMRA, Q3, CD3+CD8+CCR7-CD45RA+). CD4 T cells were divided into CD4 Tregs (CD3+CD4+CD25+CD127-), CD4 naive T cells (CD3+CD4+CCR7+CD45RA+), CD4 effect T cells (CD3+CD4+CCR7-CD45RA-). CD4 effect T cells were also divided into four major subgroups, namely early (Q2, CD27+CD28+), early-like (Q3, CD27-CD28+), medium (Q1, CD27+CD28-) and terminal (Q4, CD27-CD28-). All antibodies were purchased from Biolegend and BD (Table S2). Samples were acquired on Aurora spectral flow cytometer (Cytek), and data were analysed using FlowJo v10.

Statistical analysis and modelling

Categorical variables were compared between the groups using the χ2 or Fisher’s exact test. For continuous variables, following the Shapiro-Wilk test to assess normal distribution and Levene’s test to evaluate homogeneity of variance, parametric data were analysed using t-tests, while non-parametric data were analysed using the Wilcoxon rank sum test. The primary endpoint was overall survival (OS), which was defined as the interval between the date of treatment and death from any cause or the last follow-up. The secondary endpoint was progression-free survival (PFS), which was defined as the interval between the first date of treatment and disease progression or death due to any cause. Survival curves were estimated using the Kaplan–Meier method with the log-rank test. Associations between survival and clinical or laboratory features were evaluated using Cox proportional hazards regression analysis. To determine the strongest predictors for the final model, all variables from the univariate analysis were included in the multivariate analysis. All 27 peripheral blood indices were analysed using the least absolute shrinkage and selection operator (LASSO) with 10-fold cross-validation to identify the most effective predictive features for constructing a peripheral blood signature. Two nomogram models were constructed for predicting survival: Nomogram1, incorporating only clinical indicators, and Nomogram2, integrating both peripheral blood signature and clinical indicators. Both nomograms were based on variables that showed statistical significance in multivariate Cox analysis. The performance of the nomogram was evaluated using time-dependent area under the curve (AUC) and calibration curve. Analyses were performed using R version 4.2.0 and GraphPad Prism 9.0. All statistical tests were two-tailed, and statistical significance was set at P < 0.05.

Results

Patient characteristics

Baseline characteristics and previous treatment regimens are shown in Table 1. The median age was 45 years. 16 (22.2%) patients were female and 56 (77.8%) patients were male. Only 4 (5.56%) patients were with recurrence, 56 (77.8%) patients with distant metastasis, and 12 (16.7%) patients with both recurrence and metastasis. 51.4% of patients had liver metastasis, 44.4% of patients had lung metastasis, and 31.9% of patients had bone metastasis. In addition, 48.6% of patients had distant nodal metastasis. 44.4% of patients had been treated with PD-1 inhibitor. The median follow-up time was 17.4 months (range: 1.1–26.1 months). A total of 20 patients (27.8%) had died and 52 (72.2%) patients with disease progression during follow-up.

Table 1.

Patient characteristics (n = 72).

Characteristics	All	High risk	Low risk	P value
	n = 72	n = 36	n = 36
Age (yr)				0.492
≤45	37 (51.4%)	17 (47.2%)	20 (55.6%)
>45	35 (48.6%)	19 (52.8%)	16 (44.4%)
Gender				0.587
Female	16 (22.2%)	9 (25.0%)	7 (19.4%)
Male	56 (77.8%)	27 (75.0%)	29 (80.6%)
Smoking				0.229
No	47 (65.3%)	21 (58.3%)	26 (72.2%)
Yes	25 (34.7%)	15 (41.7%)	10 (27.8%)
WHO				0.493*
WHO II	2 (2.86%)	2 (5.71%)	0 (0.00%)
WHO III	68 (97.1%)	33 (94.3%)	35 (100%)
ECOG				0.065
0	34 (47.2%)	13 (36.1%)	21 (58.3%)
1–2	38 (52.8%)	23 (63.9%)	15 (41.7%)
R/M				1.000
Metastasis	56 (77.8%)	28 (77.8%)	28 (77.8%)
Recurrence	4 (5.56%)	2 (5.56%)	2 (5.56%)
Recurrence/Metastasis	12 (16.7%)	6 (16.7%)	6 (16.7%)
Liver metastasis				0.251
No	35 (48.6%)	15 (41.7%)	20 (55.6%)
Yes	37 (51.4%)	21 (58.3%)	16 (44.4%)
Lung metastasis				0.166
No	40 (55.6%)	17 (47.2%)	23 (63.9%)
Yes	32 (44.4%)	19 (52.8%)	13 (36.1%)
Bone metastasis				0.220
No	49 (68.1%)	22 (61.1%)	27 (75.0%)
Yes	23 (31.9%)	14 (38.9%)	9 (25.0%)
Distant LN metastasis				<0.001
No	37 (51.4%)	8 (22.2%)	29 (80.6%)
Yes	35 (48.6%)	28 (77.8%)	7 (19.4%)
Family NPC history				0.199*
No	66 (91.7%)	31 (86.1%)	35 (97.2%)
Yes	6 (8.33%)	5 (13.9%)	1 (2.78%)
EBV DNA				<0.001
<4000 copies/mL	37 (51.4%)	9 (25.0%)	28 (77.8%)
≥4000 copies/mL	35 (48.6%)	27 (75.0%)	8 (22.2%)
Previous PD-1 inhibitor treatment				0.020
No	40 (55.6%)	15 (41.7%)	25 (69.4%)
Yes	32 (44.4%)	21 (58.3%)	11 (30.6%)

Patients were stratified into two risk groups based on Nomogram2 (incorporates both peripheral blood signature and clinical indicators) using the median value (89.5) as the cut-off point: low-risk group (total score 0–89.5) and high-risk group (total score 90–289). *P value was calculated with the Fisher’s exact test.

WHO World Health Organisation, ECOG Eastern Cooperative Oncology Group, R/M recurrence/metastasis, LN lymph node, NPC nasopharyngeal carcinoma, EBV DNA Epstein-Barr Virus DNA, PD-1 programmed death-1.

Peripheral blood immune inflammation indices

It has been indicated that immune-inflammatory indices derived from peripheral blood routines, such as MLR, NLR and PLR, are closely related to the prognosis of immunotherapy [20, 21]. Spearman correlation analyses showed that MLR, NLR and PLR were significantly negatively correlated with OS and PFS (R < -0.2, P < 0.05, Fig. 1a, d, g). Using the median as a cut-off value, Kaplan–Meier analysis showed high MLR had significantly lower 1-year OS (70.2% vs. 88.6%, P = 0.032, Fig. 1b). In addition, high NLR was associated with lower 1-year PFS (24.8% vs. 51.2%, P = 0.051, Fig. 1f), though this difference was borderline significant. While other univariate survival analyses did not reach statistical significance (Fig. 1c, e, h, i), the consistent directional relationship across all indices suggests their biological relevance. These results collectively demonstrate that peripheral blood immune-inflammatory indices, particularly MLR, provide valuable prognostic information for R/M-NPC patients receiving immunotherapy, supporting their inclusion in our peripheral blood-based prediction model despite varying levels of statistical significance in individual analyses.

Fig. 1. Correlation between Peripheral blood immune inflammation indices and survival (n = 72).

a Spearman correlation of MLR with OS and PFS. b, c Kaplan–Meier analysis for OS and PFS of R/M-NPC patients divided into high- and low-MLR. d Spearman correlation of NLR with OS and PFS. e, f Kaplan–Meier analysis for OS and PFS of R/M-NPC patients divided into high- and low-NLR. g Spearman correlation of PLR with OS and PFS. h, i Kaplan–Meier analysis for OS and PFS of R/M-NPC patients divided into high- and low-PLR. Abbreviation: MLR monocyte-to-lymphocyte ratio, OS overall survival, PFS progression-free survival, R/M-NPC recurrent/metastatic nasopharyngeal carcinoma, NLR neutrophil-to-lymphocyte ratio; PLR platelet-to-lymphocyte ratio.

Peripheral blood signature construction

27 peripheral blood variables were screened for the selection of potential predictors of prognosis, with non-zero coefficients in the LASSO and logistic regression models (Fig. 2a, b). Using stepwise multiple Cox regression analysis, six most useful prognostic indicators (white blood cell count [WBC], mean corpuscular volume [MCV], haematocrit [HCT], mean corpuscular haemoglobin concentration [MCHC], platelet large cell ratio [P-LCR], MLR) under the optimal adjustment parameter were eventually used to construct the peripheral blood signature. Each patient was provided with a peripheral blood Score that was calculated using the formula derived from the levels of these elements weighted by the corresponding regression coefficient. The formula used is as follows: Score = 0.3107484 × WBC − 0.1060785 × HCT + 0.069964 × MCV − 0.0586550 × MCHC + 0.095208 × P-LCR + 1.928380 × MLR. Among them, increases in WBC, MLR, MCV and P-LCR were related to poor prognosis, while HCT and MCHC were associated with better survival. With the optimal cut-off value evaluated by R package ‘survminer’, NPC patients were divided into low- and high- peripheral blood Score groups. Compared with the patients in the low- peripheral blood Score group, high-Score patients had significantly lower 1-year OS (39.5% vs. 92.0%, P < 0.001, Fig. 2c) and 1-year PFS (12.0% vs. 47.3%, P = 0.006, Fig. 2d). These results show that peripheral blood signature could effectively distinguish patients who benefit from combined treatment in terms of survival.

Fig. 2. Development of peripheral blood signature in R/M-NPC patients (n = 72).

a LASSO coefficient profiles of the most useful prognostic factors. b 10-fold cross-validation for tuning parameter selection in the LASSO model. c, d Kaplan–Meier analysis for OS and PFS of R/M-NPC patients divided into high- and low- peripheral blood Score. Abbreviation: LASSO least absolute shrinkage and selection operator, OS overall survival, PFS progression-free survival, R/M-NPC recurrent/metastatic nasopharyngeal carcinoma.

Clinical features assessment and nomogram development

As shown in Table S3, the following variables were considered in the Cox proportional hazards model: age, gender, Eastern Cooperative Oncology Group, smoking, family NPC history, Epstein–Barr virus (EBV) DNA, recurrence/metastasis status, distant lymph node (LN) metastasis, liver metastasis, lung metastasis, bone metastasis, previous PD-1 inhibitor treatment and peripheral blood Score. In the multivariate analysis, distant LN metastasis (hazard ratio [HR] 4.481, 95% confidence interval (CI) 1.249–16.072, P = 0.021), EBV DNA (HR 5.255, 95% CI 1.286–21.476, P = 0.021), previous PD-1 inhibitor treatment (HR 3.073, 95% CI 1.025–9.217, P = 0.045) and peripheral blood signature (HR 11.644, 95% CI 3.128–43.343, P < 0.001) were independent predictive factors of OS. Peripheral blood Score demonstrated the strongest effect among all independent risk factors.

Based on the result of multivariate analysis, Nomogram2, which integrated both peripheral blood signature and clinical indicators, was built to predict the survival benefit of R/M-NPC patients who received camrelizumab plus apatinib as a second-line or later-line treatment regimen (Fig. 3a). Using Nomogram2, a total score was calculated for each patient (Table S4). We chose the median value as the cut-off (89.5), and patients could be divided into two risk groups: low-risk group (total score 0–89.5) and high-risk group (total score 90–289) (Table 1). Kaplan–Meier analyses showed high-risk patients had significantly inferior 1-year OS (58.0% vs. 100%, P < 0.001) and 1-year PFS (21.7% vs. 54.4%, P = 0.003) than low-risk patients (Fig. 3b, c). We calculated the 1-year OS time-dependent AUC to compare the predictive accuracy of distant LN metastasis, EBV DNA, previous PD-1 inhibitor treatment, peripheral blood Score, Nomogram1 (constructed with clinical independent factors, Fig. S1), and Nomogram2 (constructed with clinical independent factors and peripheral blood signature). The AUC of 1-year OS Nomogram2 model was 0.957 (95% CI 0.921–0.994), which was significantly superior to each predictor (previous PD-1 inhibitor treatment history: 0.576, 95% CI 0.448–0.704; distant LN metastasis: 0.788, 95% CI 0.706–0.871; and EBV DNA: 0.746, 95% CI 0.643–0.849) and Nomogram1(AUC = 0.864, 95% CI 0.773–0.955;) (Fig. 3d, Table 2). Additionally, the calibration curve showed favourable agreement between the prediction by Nomogram2 and the actual observation (Fig. 3e, f).

Fig. 3. Nomogram2 based on peripheral blood signature and clinical characteristics (n = 72).

a Nomogram2 for predicting the 1-year OS for R/M-NPC patients. b, c Kaplan–Meier analysis for OS and PFS of R/M-NPC patients stratified into high-risk and low-risk groups based on Nomogram2. d Receiver operating characteristic curve for predicting 1-year OS. e, f Calibration curve of Nomogram1 and Nomogram2 for predicting the 1-year OS. Abbreviation: OS overall survival, R/M-NPC recurrent/metastatic nasopharyngeal carcinoma, PFS progression-free survival.

Table 2.

ROC analysis for predicting 1-year OS (n = 72).

Variables	1-year OS
	AUC	95% CI	P value
Peripheral blood Score	0.791	0.678–0.905	0.012
Distant LN metastasis	0.788	0.706–0.871	<0.001
EBV DNA	0.746	0.643–0.849	<0.001
Previous PD-1 inhibitor treatment	0.576	0.448–0.704	<0.001
Nomogram1	0.864	0.773–0.955	0.042
Nomogram2	0.957	0.921–0.994	Ref.

Nomogram1, which incorporates only clinical indicators, and Nomogram2, which integrates both peripheral blood signature and clinical indicators.

ROC receiver operating characteristic, OS overall survival, AUC area under the curve, CI confidence interval, LN lymph node, EBV DNA Epstein-Barr Virus DNA, PD-1 programmed death-1.

High-risk group was associated with immune suppression status

In order to elucidate the underlying mechanisms that determine survival between the high- and low-risk groups, we examined the phenotype of the peripheral, cryopreserved leucocytes before therapy. In particular, PBMCs underwent flow cytometry analysis for the presence of MDSCs and T cell subsets (Fig. S2). The reason for choosing to detect MDSCs is that we have previously found that MLR and NLR are significantly negatively correlated with patients’ survival. MDSCs can be categorised into two main subtypes: M-MDSCs and Polymorphonuclear/Neutrophilic-MDSCs (G-MDSCs). In this study, we found that MDSCs in peripheral blood of NPC patients were mainly M-MDSCs. The high-risk group was significantly associated with higher proportions of M-MDSC (P = 0.0168, Fig. 4a) and lower proportions of CD8 T cell (P = 0.030, Fig. 4b). Although no significant differences were observed, the high-risk group was related to higher proportions of CD4 T cell (P = 0.098, Fig. 4b). When stratified into immunocyte subsets based on different markers, high-risk group was also significantly associated with lower proportions of effect memory CD8 T cell (P = 0.0206, Fig. 4c, d) and higher proportions of Tregs, effect and early effect CD4 T cell (P = 0.029, 0.044, 0.035, respectively, Fig. 4e, f). These results suggest that patients in the high-risk group might yield an immunosuppressive environment, rendering conditions adverse for optimal immunotherapy.

Fig. 4. Comparison of immune cell populations in peripheral blood samples between high- and low-risk R/M-NPC patients stratified by Nomogram2 score (n = 61).

a Monocyte MDSCs in high- and low-risk groups. b CD4 or CD8 T cells (% of CD3+ cells) in high- and low-risk groups. c, d Heat map and statistical analysis showed the subtypes of CD8 T cells in high- and low-risk groups. e, f Heat map and statistical analysis showed the subtypes of CD4 T cells in high- and low-risk groups. High-risk group (n = 31), low-risk group (n = 30). Abbreviations: MDSCs myeloid-derived suppressor cells.

Discussion

Based on the previous clinical trial cohorts and blood samples, we constructed a peripheral blood signature with WBC, MCV, HCT, MCHC, P-LCR, MLR to estimate the individualised risk of R/M-NPC patients undergoing combined camrelizumab and apatinib as a second-line or later-line treatment regimen therapy. Furthermore, we developed a nomogram based on the peripheral blood signature, incorporating vital clinical characteristics such as EBV DNA, distant LN metastasis and previous PD-1 inhibitor treatment. This nomogram can both precisely predict OS and reflect the individualised immune status of the patients. It could assist clinicians in assessing the systemic immune status of R/M-NPC patients and in evaluating the potential benefits they may derive from the combined PD-1 inhibitor and anti-angiogenic regimen before initiating therapy.

Besides the markers from the tumour environment, the therapeutic effect and survival outcome are related to multiple factors, such as systemic inflammation, immune status and nutrition [3, 23, 24], which could be reflected in peripheral blood routine [20, 21, 25–28]. In this study, the increases in WBC, MLR, MCV and P-LCR were correlated with poor prognosis. WBC is a widely used peripheral blood indicators of cancer and associated with poor prognosis and unsatisfactory therapeutic efficacy. MLR is an important indicator representing the inflammatory response and immune status. Multiple studies have indicated that high MLR level is associated with unfavourable survival in cancers treated with PD-1 inhibitors. Lymphocytes play a crucial role in immune surveillance, inhibiting the proliferation and migration of tumour cells by inducing apoptosis, thus exerting an anti-tumour effect. Monocytes could promote tumour development by producing pro-inflammatory cytokines. Higher MLR indicates a weaker immune status and development of immune suppression. MCV reflects the size of red blood cells. It has been reported that high MCV is associated with negative outcomes of various cancers. P-LCR reflects the percentage of large platelets. The role of P-LCR in tumours is unclear and studies are limited. In addition, the increases in HCT and MCHC level were with better survival in this study. HCT refers to the proportion of the volume occupied by red blood cells in a certain volume (L) of whole blood, and MCHC reflects the concentration of haemoglobin in red blood cells. HCT and MCHC can reflect the degree of anaemia in patients, and their decrease is associated with poor prognosis [29, 30]. Taken together, we propose that in R/M-NPC patients, a relatively low WBC count, MLR, MCV and P-LCR level, along with high HCT and MCHC levels, reflect reduced inflammation, decreased immunosuppression and adequate nutritional status. These conditions likely enhance response to combination therapy, ultimately contributing to improved clinical outcomes.

Clinical characteristics and EBV DNA are always regarded as important factors correlated with therapeutic response and prognosis in NPC patients. We hypothesise that combining peripheral blood signature with these significantly correlated clinical indicators may further enhance the robustness of the predictive model. As expected, when combining the peripheral blood signature with independent clinical factors (EBV DNA, distant LN metastasis, and previous PD-1 inhibitor treatment), the performance of nomogram significantly improved. Plasma EBV DNA is the most effective biomarker for prognostication, predicting treatment response and disease surveillance [31–33]. Pre-treatment EBV DNA ≥ 4000 copies/mL has a significantly worse prognosis than EBV DNA < 4000 copies/mL [34, 35]. For R/M-NPC patients undergoing immunotherapy, EBV DNA was also a useful biomarker for monitoring disease progression [33]. Similarly, EBV DNA was an independent prognostic factor for R/M-NPC patient in our study. The use of PD-1 inhibitor before combination therapy was a disadvantageous factor for R/M-NPC patients. The reason might be the dominant environment of immunosuppression that exists in those patients, which leads to the development of PD-1 inhibitor-resistant. Distant LN metastasis was usually regarded as a part of distant metastasis for analysis [36]. When conducting a separate analysis, it was found that the occurrence of distant LN metastasis was an important prognostic predictor for R/M-NPC patients. The colonisation of tumour cells in the LNs may lead to immune suppression, thereby promoting the metastasis of the tumour to distant organs. Immunotherapy often necessitate the activation of immune responses within LNs, which can be effectively induced by classical dendritic cell subsets presenting tumour antigens derived from the primary tumour [37, 38]. The distant LN metastasis might be associated with a weakening of anti-tumour immune response [39, 40]. Collectively, EBV DNA represents the tumour burden, distant LN metastasis indicates the tumour’s metastatic potential and the possibility of an immunosuppressive condition, a history of previous PD-1 inhibitor treatment suggests resistance to PD-1 inhibitors, and peripheral blood signature reflects the systemic immune and inflammatory status. These factors contribute to the good performance of the nomogram.

Peripheral blood lymphocyte subpopulations and percentages reflect the current state of immune function and are closely related to therapeutic efficacy and prognosis [41–43]. T cell-mediated cellular immunity is critical for the antitumor immune response. CD8 T cells are antigen-specific killer cells targeting tumour cells in the antitumor immune response [41]. MDSCs and Tregs exhibit immunosuppressive activity by inhibiting T cells and NK cells and play a significant role in various malignant tumours [43–45]. Consistent with those studies, high-risk patients are associated with a decrease in CD8 T cells, an increase in MDSCs and Tregs, which indicates that high-risk patients are more likely to be in a state of immune suppression. Therefore, high-risk individuals are recommended to participate in clinical trials, such as antibody-drug conjugates, adoptive cell transfer therapies (including tumour infiltrating lymphocyte therapy, chimeric antigen receptor T-cell therapy and T-cell receptor engineered T-cell therapy), and tumour vaccines.

This study has several limitations. First, potential selection bias may have affected the reliability and reproducibility of the model because the patient sample was derived from a clinical trial cohort. Second, the sample size is small, and large-scale randomised prospective clinical data are needed. Third, the nomogram model requires external validation in diverse clinical settings to assess its generalisability and predictive accuracy across different patient populations.

Conclusions

In summary, this study has assessed the clinical characteristics and peripheral blood routine data to predict the response to camrelizumab plus apatinib combination therapy in R/M-NPC patients. Then, we constructed an efficacy prediction model based on clinical indicators and peripheral blood signature, and the prediction performance was good. Moreover, there was a close connection between predictive model and the peripheral blood immunocytes, indicating that the predictive model can effectively reflect the patient’s peripheral blood immune status. Further prospective studies with a larger sample size are needed to confirm these findings.

Statement of translational relevance

This study’s translational relevance lies in its development of a peripheral blood-based biomarker signature to predict treatment response in patients with recurrent/metastatic nasopharyngeal carcinoma (R/M-NPC) undergoing camrelizumab and apatinib combination therapy. The identification of key blood indicators and their integration with clinical characteristics into a predictive nomogram offers a non-invasive method for patient stratification, potentially enhancing treatment efficacy. Additionally, the immunophenotyping insights provide a mechanistic understanding of why certain high-risk patients may not benefit from the therapy, paving the way for personalised treatment strategies.

Author contributions

LY and HM contributed to conceptualisation, funding acquisition, supervision and manuscript review and editing. KL, SL and GJ were responsible for data curation, formal analysis, conceptualisation, methodology and writing the original draft. SX and SL conducted data curation and formal analysis. LT performed data curation and methodology. All authors read and approved the final version of the manuscript.

Funding

This study was funded by grants from the National Key Research and Development Program of China (2022YFC2505800, 2022YFC2705005), National Natural Science Foundation of China (No. 32200651, 82203776, 82203125, 82222050, 82272739, 82272882, 82203259, 82173287, 82073003, 82003267, 82002852, 82373258, 82372980, 82361168664, 8247101588, 82473038), Guangdong Basic and Applied Basic Research Foundation (2021B1515230002, 2023B1515120092, 2023A1515010398, 2024A1515013021), Science and Technology Program of Guangzhou (202201011561, 2023A04J2127, 2023A04J2246, 2024B03J1248), Sun Yat-sen University Clinical Research 5010 Program (No. 201315, 2015021, 2017010, 2019023), Innovative Research Team of High-level Local Universities in Shanghai (SSMU-ZLCX20180500), Postdoctoral Innovative Talent Support Program (BX20220361), Planned Science and Technology Project of Guangdong Province (2019B020230002), Science and Technology Projects in Guangzhou (202201011533), Key Youth Teacher Cultivating Program of Sun Yat-sen University (20ykzd24), and Fundamental Research Funds for the Central Universities.

Data availability

All data included in this study are available upon request by contact with the corresponding author.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of Sun Yat-sen University Cancer Center. All patients provided written informed consent prior to the collection of specimens. All methods were performed in accordance with the relevant guidelines and regulations.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-025-03044-y.