Lung cancer is one of the most common malignant neoplastic diseases (1), with still poor prognosis despite the advances in the knowledge of the biology of this neoplasm. By far the most common cancer-related cause of death, lung cancer claims more lives each year than colon, breast, and prostate cancers combined (2). Small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC) are the two main types according to histopathological diagnosis, with the former having a noticeably dismal prognosis compared to the latter. NSCLC represents about 85% of all lung carcinomas, with two predominant histological subtypes: adenocarcinoma and squamous cell carcinoma (3,4).

Recently, the immune system in cancer development and therapeutic response has taken on a pivotal role, and immune checkpoint inhibitors (ICIs) have received increasing attention due to their growing use in lung, skin, genitourinary, and other cancers treatment (5,6).

Biomarkers, such as programmed cell death-ligand 1 (PD-L1), tumor mutational burden (TMB) and neoantigen load, which may reflect the state of the tumor immune microenvironment, seemed to be helpful in selecting patients who are more likely to benefit from ICI treatment. However, the detection of these biomarkers greatly depends on the adequacy of tumor tissue. Therefore, there is an unmet need for biomarkers that can be easily assessed without requiring invasive procedures.

In NSCLC, ICIs, by targeting inhibitory receptors on T cells, have achieved tumor responses in a subset of patients across a wide range of cancer types; however, a not so small subset of patients undergoing ICI treatment do not experience clinical benefit. In less than half of patients with PD-L1 expression ≥50%, pembrolizumab yields an objective decrement in neoplastic mass, while 45% of those with negative PD-L1 respond to therapy with nivolumab and ipilimumab, suggesting that PD-L1 is far from an ideal biomarker predicting response to immunotherapy (7,8). Among patients with newly diagnosed lung cancer, almost 15% have leukocytosis (9). Granulocyte colony stimulating factor (G-CSF), macrophage colony-stimulating factor (M-CSF), and interleukin 6 (IL-6) levels have been shown to be significantly higher in patients with lung malignancies, associated to a worse prognosis (10). Leukocytosis is likely an upshot defined by the extent of cytokine production and is thus referred to as tumor-related leukocytosis.

Furthermore, even thrombocytosis has shown a correlation to many malignancies, and has been suggested to be an independent prognostic factor in lung cancer (11-13).

Tumor-related leukocytosis predicts poor survival in lung cancer patients (14,15). The pro-oncogenic role of G-CSF has been evaluated in preclinical models; tumor-secreted G-CSF has been shown to stimulate proliferation of myeloid-derived suppressor cells (MDSC). MDSCs are a heterogeneous population of cells with neutrophil-like immunophenotypic features. Polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) have been found to hinder immune response activation in cancer and many other pathological conditions. Whereas in healthy individuals PMN-MDSCs are barely detectable, they accumulate in the blood, lymph nodes, and bone marrow of patients and cancer laboratory animals and inhibit T-cell-mediated antitumor activity (16). G-CSF mobilizes MDSCs from the bone marrow and makes them nest into distant organs, developing a pro-tumor microenvironment that fosters tumor cell diapedesis and metastasis (17,18). G-CSF is secreted by cancer cells and the tumor microenvironment, leading to the proliferation of neutrophils as well as MDSCs. MDSCs inhibit lymphocytes proliferation, increasing the number of neutrophils. Therefore, the neutrophil-to-lymphocyte ratio (NLR) represents only the epiphenomenon of a deeper immunoregulatory mechanism, which is indirectly related to disease progression and patients’ survival, while MDSCs directly contribute to disease progression and clinical outcome (19).

NLR represents a valuable and reliable candidate biomarker, because it is cost-effective and easily accessible (20,21). Biologically, NLR reflects systemic inflammation and could represent a sort of balance of the immune system in a malignant biological setting (22-24). Specifically, neutrophils can secrete immunosuppressive mediators and angiogenic factors, and can contribute to a pro-tumor microenvironment (25). In addition, low levels of circulating lymphocytes may correlate with lower levels of tumor-infiltrating lymphocytes (TILs) and a reduced anti-tumor T-cell response (26).

In NSCLC, higher neutrophil counts are correlated with reduced CD8+ T-lymphocytes counts (27). A high NLR in the pre-treatment setting seems to be independently and significantly associated with a short progression free-survival (PFS) and overall survival (OS) of patients with previously untreated NSCLC undergoing pembrolizumab monotherapy, even if PD-L1 expression on tumor cells is ≥50% (28). On the contrary, low NLR and high PD-L1 seem to be significantly associated with favorable outcomes (29), underlying the importance of this potential biomarkers for the choice of therapy, probably regardless of age, sex, tobacco use, performance status, histology, serum albumin level, and oncogenic drivers (30).

The aim of this retrospective study was to assess the prognostic role of NLR and platelet-to-lymphocyte ratio (PLR), alongside other variables, such as age, sex, metastatic sites, and PD-L1 tumoral proportion score (TPS) level in patients with advanced NSCLC treated with pembrolizumab.

Patients suffering from metastatic NSCLC, with TPS PD-L1 ≥50%, Eastern Cooperative Oncology Group (ECOG) Performance Status (PS) ≤2, and who had undergone pembrolizumab as first-line monotherapy were eligible for this analysis. Patients were ineligible if they had oncogene-addicted malignancies (EGFR, ALK mutations/translocations), autoimmune disease requiring systemic steroid treatment, or active interstitial lung disease. Pembrolizumab was initially administered at 2 mg/kg intravenously over 60 min and later at 200 mg every 3 weeks, or at 400 mg every 6 weeks. Patients were not required to provide any written informed consent, and the study was conducted in accordance with the Declaration of Helsinki and Italian Privacy Protection Commissioner.

In each center, tumor PD-L1 expression was determined by immunoassay using the Ventana PD-L1 rabbit monoclonal antibody SP263 (Roche, Basel, Switzerland) and the BenchMark Ultra platform system (Roche). Representative tumor sections (with more than 100 neoplastic cells) of formalin-fixed and paraffin-embedded tissues were used: sections of 4 μM were prepared, heated, and mounted on polarized slides for immunohistochemistry staining. Every test was carried out with a session control (a positive staining in human tonsilla epithelial cells). Microscopical assessment of TPS was evaluated by an experienced pathologist and given as a percentage of positive cells on total tumoral cells.

Collection of peripheral venous blood was routinely analyzed by local laboratories for complete blood count (CBC). White blood cell count (WBC), absolute neutrophil count (NEU), absolute lymphocyte count (LYM), absolute platelets count (PLT) at baseline were analyzed: NLR and PLR were calculated from pre-treatment complete blood counts. Clinicopathological features and treatment history data were extracted from the medical records: tumor response was assessed using computed tomography and/or positron emitting tomography every 8-12 weeks. Progression free survival (PFS) was measured from pembrolizumab treatment start to clinical, metabolic, or radiographic progression or death from any cause. OS was measured from pembrolizumab treatment beginning to death from any cause.

Statistical analysis. Clinically relevant features, such as age (continuous), sex, histological type (adenocarcinoma, squamous, NSCLC NOS), PD-L1 TPS (continuous), metastatic sites, and baseline peripheral blood counts were used as variables. Kaplan–Meier analysis of PFS and OS was performed. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated using the univariate and multivariate Cox proportional hazard model. Both log-rank test and Cox model were analyzed using SPSS (IBM, Chicago, IL, USA). The cut-offs for NLR and PLR were identified by Survival Analysis for Continuous Explanatory Variable maximally selected rank statistics (Jamovi Software) (31,32).

The analysis involved a series of 252 patients consecutively treated in four South Italy oncological centers according to clinical practice, between September 2017 and May 2022. Baseline characteristics and peripheral blood counts of these patients are shown in Table I. Of all patients, 205 (81.3%) were male. The median age was 70 years (range=39-87 years). One hundred eighty-two (72.2%) patients had adenocarcinoma histology. Twenty-seven (10.7%) patients had liver metastases (Table I).

The median±SD NLR was 3.8±4.5 and PLR was 171±113. To evaluate OS, the cut-off values for NLR and PLR used were 4.8 (p-Value<0.001) and 193 (p-Value=0.003), respectively (Figure 1A and B); for PFS they were 5.29 (p-Value<0.001) and 182 (p-Value=0.013), respectively. For simplicity, it was decided to use the cut-off calculated on the OS also in the analysis of PFS.

Univariate log rank test revealed that high NLR was associated with a significantly shorter PFS and OS - median 3.7 and 7.6 months vs. 19.6 and 34.8 months (mean value, median not reached) for the low NLR counterparts, respectively; p-Value<0.0000001 (Figure 2 and Figure 3). At Cox proportional hazard univariate analysis, NLR ≥4.8 increased the risk of death by more than three-fold compared with the low NLR (HR=3.26, 95%Cl=2.3-4.6, p-Value<0.0000001) and the risk of disease progression by more than two-fold (HR=2.68, 95%CI=1.9-3.7, p-Value<0.0000001). Multivariate analysis was performed using as risk factors all baseline characteristics. In both multivariate analysis for PFS and OS, NLR and PLR were analyzed as continuous variables, also because, despite the dichotomization easing the decision-making process in clinical practice, it may create rather than avoid problems by reducing the statistical power of the analysis (33).

The multivariate analysis of PFS revealed that for each unit increase in NLR there was a 14% increased risk of disease progression (p-Value<0.00000001); furthermore, each 10% increase in PD-L1 reduced the risk of progression by 6.7% (p-Value=0.022) (Table II). Similarly, the multivariate analysis of OS found a 13.7% increased risk of death for each unit increase in NLR (p-Value<0.00000001), with each 10% increase in PD-L1 reducing death risk by 7% (p-Value=0.005) (Table III). Furthermore, a very high NLR – over 10 – seemed to forecast an even worse prognosis in patients undergoing immunotherapy, with sudden deaths in the weeks immediately following therapy (Figure 4). Although in univariate log-rank analysis PLR seemed to discriminate patients’ survival (Figure 5), in multivariate Cox regression, only NLR and PD-L1 retained statistical significance.

Today, NLR is widely used across many medical disciplines, as a reliable and low-cost marker of cancer-related inflammation (34). Petrova et al. evaluated NLR and PLR in patients with NSCLC treated with pembrolizumab as a second line treatment, after they had undergone platinum-containing chemotherapy as a first line treatment. They assessed optimal cut-offs by receiver operating characteristics (ROC), with matching NLR and PLR with clinical benefit, defined as radiological response or stabilization for more than six months (35). Similarly, Hasegawa et al. found an NLR of 4.56 as the optimal cut-off by ROC, indicating as outcome variable disease progression, regardless of when it occurred (28). As also shown by Ksienski et al., we identified optimal cut-off points of slightly lower values for baseline NLR and PLR by using maximally selected rank statistics (36). Nevertheless, we suggest that both NLR and PD-L1 must be considered as continuous variables rather than dichotomous variables, and both should be included in the multivariate analysis: thus, their role can be fully understood, with an exponential trend in the risk as the NLR increases and PD-L1 decreases.

NLR and PD-L1 probably represent two sides of the same coin (or maybe 2 faces of an icosahedron), namely two mechanisms by which the tumor succeeds to evade the host's immune response: therefore, we believe it is restricting and inaccurate to evaluate them with a simple "high and low" score. MDSCs are an entity hiding behind neutrophilic leukocytosis, acting in synergy with PD-L1-expressing tumor cells to inactivate T cell immunosurveillance. Finally, alongside other common variables, such as metastatic sites, age, and sex, NLR and PD-L1 predict risk of disease progression and death; furthermore, a very high NLR seems to forecast an even worse prognosis in patients undergoing immunotherapy, with sudden deaths in the days immediately following therapy, strengthening the evidence that the prognosis worsens as NLR increases.

It could be interesting to evaluate immunotherapy treatment alone in patients with low NLR and PD-L1 values, and also explore the efficacy of the chemo-immunotherapy combination in patients with a high NLR. The prediction of prognosis could help identify patients who are likely to benefit more from chemo-immunotherapy rather than immunotherapy only: indeed, chemotherapy could be effective in killing MDSCs and therefore, hinder their immunosuppressive role.

The Authors have no conflicts of interest to declare in relation to this study.

Francesco Jacopo Romano conceived the study, performed statistical analysis, drafted and revised the manuscript. Ferdinando Riccardi contributed to the study concept, manuscript drafting and critical revision. All Authors contributed to data collection and manuscript revision.