Peripheral Blood Eosinophil Count Optimizes Pembrolizumab-Based Immunotherapy in The First-Line Setting of Advanced or Recurrent Non-Small Cell Lung Cancer

For cancer immunotherapy, the tumor proportion score for the programmed death-1 ligand is not a robust biomarker. The peripheral blood eosinophil count (PEC) is a potential alternative. However, it is not yet established. To test the ecacy of PEC-guided selection of pembrolizumab monotherapy (MONO) or pembrolizumab plus chemotherapy (COMBO), we retrospectively reviewed data of patients with advanced or recurrent non-small cell lung cancer in the rst-line setting (April 2017 to April 2020). Among 137 patients enrolled, Kaplan–Meier analysis revealed no signicant difference between the MONO (n = 84) and COMBO (n = 53) therapies. The inuence of PEC before the second administration of each regimen (PEC 2 ) was evaluated. The low PEC 2 subgroup (<150 × 10 6 /L) receiving MONO had poorer survival rates than those receiving COMBO (median progression-free survival [mPFS]: 5.75 vs. 7.59 months and median overall survival [mOS]: 12.0 months vs. not reached [NR), respectively). In patients receiving MONO, the low PEC 2 showed worse prognosis compared with the high PEC 2 group (mPFS: 5.75 vs. 16.1 months and mOS: 12.0 months vs. NR, respectively). PEC 2 can be a potential predictive/prognostic biomarker for MONO, which encourages the switch from MONO to COMBO to avoid treatment failure.


Introduction
Lung cancer is the leading cause of cancer deaths worldwide. Despite the introduction of new anticancer therapeutics, such as molecular targeted agents for driver gene-mutated non-small cell lung cancer (NSCLC) and immune checkpoint inhibitors (ICIs) including pembrolizumab and nivolumab, the prognosis of patients with advanced-stage NSCLC is poor. Pembrolizumab, which is an inhibitor of programmed death 1 (PD-1) that interrupts the interaction with its ligand, PD-L1, has been evaluated for both monotherapy and combination therapy with platinum-doublet chemotherapy for the rst-line treatment of metastatic NSCLC [1][2][3][4][5][6]. Some patients exhibit a signi cant survival bene t from monotherapy or combination therapy using pembrolizumab; however, the PD-L1 tumor proportion score (TPS) is not always an accurate predictive/prognostic biomarker. Therefore, the methodology needed to identify the ICI-sensitive subpopulation has not been established. The appropriate biomarkers that guide the ICI treatment should re ect the anticancer immunity of each patient at the precise moment of testing with a simple sampling procedure at a low economical cost. Recent retrospective studies have indicated the potential use of peripheral white blood cell count as a predictive marker for ICIs, in particular, the peripheral blood eosinophil count (PEC) [7][8][9]. A PEC ≧ 150 × 10 6 /L before the administration of nivolumab was associated with improved progression-free survival (PFS) and overall survival (OS) in advanced or recurrent NSCLC [7]. However, baseline PEC showed no signi cant association with the treatment response or survival to anti-PD-1 antibodies (nivolumab and pembrolizumab) [8]. Other groups found that baseline PEC was not statistically associated with the treatment response; however, a signi cant elevation of PEC after three cycles of pembrolizumab or four cycles of nivolumab was observed in patients who had complete or partial responses based on the Response Criteria for Solid Tumours version 1.1 [9]. PEC varies in healthy people and is affected by many factors including age, sex, smoking, allergy, comorbidities, and medication [10]. Therefore, further studies are needed to determine the value of PEC as a practical biomarker for immunotherapy in NSCLC, especially for the establishment of PEC-guided optimization of the rst-line therapy using pembrolizumab in advanced and recurrent NSCLC. We veri ed the appropriate timing to evaluate PEC as a practical biomarker to avoid treatment failure and established three clinical questions that should be answered to determine the usefulness of PEC as a biomarker: 1) Is PEC useful for selecting pembrolizumab monotherapy (MONO) versus pembrolizumab plus chemotherapy (COMBO); 2) Can PEC predict the subgroup that receives a survival bene t from MONO; and 3) Can PEC predict the subgroup that receives a survival bene t from COMBO?
We performed a retrospective study to evaluate the clinical utility of PEC to identify pembrolizumabsensitive patients with advanced-stage NSCLC. We also determined whether PEC could guide the decision to choose MONO or COMBO in the rst-line setting. We focused on these speci c patients to establish a practical application of PEC as a biomarker for the pembrolizumab-based rst-line therapy.

Results
Patient Characteristics: A total of 137 patients with advanced or recurrent NSCLC who received MONO or COMBO as their rst-line therapy were analyzed. The NSCLC staging was based on the eighth edition of the TNM classi cation for lung cancer from the International Association for the Study of Lung Cancer. None of the patients was known to have any treatable driver oncogenes, such as epidermal growth factor receptor gene mutations or anaplastic lymphoma kinase gene rearrangements. Among the included patients, 84 were treated with MONO and 53 were treated with COMBO ( Table 1). The median follow-up was 374 days (range; 6 to 1,220) for the MONO group and 394 days (range; 78 to 689) for the COMBO group. Patient characteristics, including smoking status, allergy history, histology, and brain metastases, were similar between the two groups. A history of allergy including asthma, pollinosis, food, drug, and contrast agents was observed in 44 patients (32.1%) in the whole population. Compared with the group treated with COMBO, patients in the MONO group were older (median: 73 vs. 69 years, p < 0.001) and had a poorer Eastern Cooperative Oncology Group (ECOG) performance status (p < 0.001). PD-L1 TPS, which was evaluated using the 22C3 pharmDx assay (Dako), was signi cantly higher in the MONO group compared with the COMBO group (p < 0.001). The patients with PD-L1 TPS ≧ 50% accounted for 84.7% in the MONO group and 17.0% in the COMBO group. However, because PD-L1 TPS was not always measured in the COMBO group, 12 patients (22.6%) had an unknown status. Compared to the COMBO group, prior anticancer therapies, such as surgery and radiation, were administered more frequently in the MONO group (p = 0.010). Of note, 23 patients (27.4%) receiving MONO had undergone surgery.
Discontinuation of the rst-line therapy because of disease progression, AEs, or patient death occurred in 55 of 84 (65.5%) patients in the MONO group and 34 of 53 (64.2%) in the COMBO group (p = 1.000). The second-line treatment after discontinuation of the rst-line therapy was administered to 27 of 84 patients (32.1%) in the MONO group and 20 of 53 patients (37.7%) in the COMBO group, with no statistical difference. The frequency of adverse events (AEs), including immune-related AEs and interstitial lung pneumonitis, was not signi cantly different between the two groups. There was also no association between the frequency of AEs and PEC before the rst or second administration of each therapy (data not shown). Absolute PECs were assessed on the nearest day before the administration of each anticancer therapy. In the whole population, the median PEC was 160 × 10 6 /L (range 0-2, 840 × 10 6 /L) before the rst administration of MONO or COMBO and 150 × 10 6 /L (range 0-1,590 × 10 6 /L) before the second administration. Therefore, we set 150 × 10 6 /L as the cut-off to distinguish high and low PEC for further analyses. PEC before the rst administration of therapy (PEC 1 ) was 155 × 10 6 /L (range 0-2,840 × 10 6 /L) in the MONO group and 169 × 10 6 /L (20-690 × 10 6 /L) in the COMBO group (p = 0.086). Interestingly, PEC before the second administration (PEC 2 ) was signi cantly higher in the MONO group (median: 185 ×   Supplementary Fig. 1). In the high PEC 1 subgroup, the median PFS was 13.6 months (95% CI: 8.87-NR, n = 46) in the MONO group vs. 10.6 months (5.29-NR, n = 27) in the COMBO group (p = 0.872) ( Supplementary Fig. 1a) Supplementary Fig. 1d).
Next, we evaluated PEC 2 with respect to survival analysis. The Kaplan-Meier results indicated that the high PEC 2 subgroup showed no statistical differences between the MONO (n = 53) and COMBO (n = 17) groups ( Fig. 1c and 1d) Fig. 1d). However, survival analysis using low PEC 2 revealed that patients with low PEC 2 receiving the MONO therapy (n = 31) had an unfavorable tendency of PFS (Fig. 1e) and a statistically shorter OS (Fig. 1f), compared with those receiving the COMBO therapy (n = 36). In the low PEC 2 subgroup, the median PFS was 5.75 (95% CI:    Table 4). The univariable analysis for OS revealed that the histology of squamous NSCLC (p = 0.031), discontinuation because of disease progression, AEs, death (p = 0.002), and low PEC 2 (p = 0.006) were signi cantly associated with shorter OS.  Table 5). The multivariable analysis for PFS revealed that low PEC 2 (p = 0.043), discontinuation because of disease progression, AEs, or death (p < 0.001) were signi cantly associated with unfavorable PFS. The multivariable analysis for OS revealed that discontinuation of treatment (p = 0.002) was signi cantly associated with unfavorable OS; however, PEC 2 did not reach statistical signi cance (p = 0.061).  Table 3).

Discussion
We conducted a retrospective study to determine whether PEC is a prognostic or predictive biomarker capable of predicting the outcome for advanced/recurrent NSCLC patients receiving MONO or COMBO in the rst-line setting. Our study focused on the rst-line treatment with pembrolizumab for metastatic NSCLC, although the confounding factors inevitable in a retrospective analysis should be avoided as much as possible. The importance of both MONO and COMBO was weighed as the rst-line treatment of advanced/recurrent NSCLC. To our knowledge, this is the rst study to show that low PEC before the second administration of therapy (PEC 2 ) predicts the poor survival rates in patients receiving the MONO therapy compared with those receiving the COMBO therapy. PEC 2 was also associated with signi cant differences in PFS, OS, and ORR between patients with high and low PEC 2 receiving the MONO therapy.
We set 150 × 10 6 /L as the cut-off for the classi cation of high and low PEC 2 subgroups, which was the median value of PEC 2 in the patient population. There was no relationship between PD-L1 TPS and PEC (both PEC 1 and PEC 2 ) and the frequency of AEs was not statistically different between the high and low PEC groups (data not shown). The cut-off value of PEC was generally consistent with that observed in previous studies: 125 [11], 150 [7], 150, and 300 [12] × 10 6 /L. This cut-off value is less than the 500 × 10 6 /L value, which represents the threshold of eosinophilia. Peripheral blood eosinophilia was also reported to improve the response and prognosis to the ICI monotherapy [13]. The true cut-off value of the PEC should be veri ed in a well-designed, large-scale, prospective study, and the difference in the immuno-oncological signi cance between PEC elevation and eosinophilia are unresolved. The results of our study warrant further investigation into PEC as a prognostic biomarker and the role of eosinophils in lung cancer immunotherapy.
We found that low PEC 2 , but not PEC 1 , was associated with worse survival outcomes of patients receiving the MONO therapy compared with that of the COMBO group. However, the timing of when to measure PEC remains controversial. PEC 1 was associated with better survival with nivolumab for the second-line or later treatment [7] and the ICI monotherapy or the combination with an anti-angiogenesis agent or chemotherapy for the rst-line or later [11]. In contrast, PEC 1 was not associated with treatment response and survival with nivolumab or pembrolizumab as the rst-line or later therapy [8]. Elevation of PEC after the administration of three cycles of pembrolizumab or four cycles of nivolumab was con rmed in patients who achieved a better treatment response [9]. Another group also reported that the maximum value of PEC during ICI with or without chemotherapy was observed 5 weeks after the initiation of each therapy in patients who obtained disease control and at 2 weeks for those who exhibited uncontrolled diseases [12]. However, we did not con rm elevated PEC after the administration of pembrolizumab with or without chemotherapy in this study. Instead, we demonstrated the importance of PEC 2 in the pembrolizumab-based therapy for predicting treatment response and prognosis. Considering these ndings, PEC 2 , which was evaluated 3 weeks after the initiation of the pembrolizumab-based therapy, is a convenient biomarker that may be routinely evaluated in clinical practice to predict the treatment response.
We found that PEC 2 was signi cantly associated with improved survival and treatment response. The reason why PEC 2 may predict a more favorable outcome in the MONO therapy remains unclear. Inhibition of the PD-1/PD-L1 interaction by the MONO therapy did not result in signi cant changes in PEC, but the high PEC 2 subgroup was associated with better prognosis and response compared with the low PEC 2 subgroup. These ndings indicate that PEC 2 may re ect tumor response to PD-1 inhibition. Surprisingly, our study revealed that COMBO did not always result in a survival bene t compared with MONO, and patients with low PEC 2 showed signi cantly poorer survival with the MONO therapy. PEC dropped signi cantly after the rst administration of COMBO and PEC 2 was not associated with prognosis. This phenomenon may be explained, in part, by antiemetic treatment with steroids during the COMBO therapy.
In patients with metastatic NSCLC treated with ICIs as the rst-line and later treatment, early use of steroids was associated with worse clinical outcomes because of the modulation of peripheral blood immune cells, resulting in PEC suppression and decreased activation of antitumor immunity [14]. The combination therapy using COMBO is expected to enhance the antitumor effect; however, the use of steroids for antiemesis may inhibit eosinophils from acting not only as a prognostic biomarker, but also as a mediator of antitumor activity. These ndings indicate that the choice of MONO or COMBO in the rst-line settings for advanced or recurrent NSCLC is controversial. However, low PEC 2 in MONO may be su cient to select the COMBO treatment.
There are several limitations to the present study. First, this is a retrospective study with a relatively short follow-up period at a single hospital. However, this study provides evidence to support a larger prospective study with an adequate sample size to determine an accurate cut-off value and interpretation of PEC. To clarify the usefulness of PEC as a biomarker, we designed our study to focus on the rst-line therapy for metastatic or recurrent NSCLC in which the regimen only contained pembrolizumab for immunotherapy. Second, we did not improve our understanding of the precise role of eosinophils in cancer immunity through the administration of pembrolizumab. Eosinophils are known to in ltrate tumors, exert an antitumor response in the tumor microenvironment [15] and antitumor immunity in a melanoma model [19]. These ndings support the hypothesis that peripheral blood eosinophils do not always function as antitumor effectors, but once activated, they migrate into the tumor microenvironment and exert antitumor effects along with CD8-positive T cells and inhibit the PD-1/PD-L1 interaction.
In conclusion, PEC 2 in the pembrolizumab monotherapy is a potential predictive and prognostic Statistical Analysis: Comparisons between patient characteristics were performed using χ 2 or Fisher's exact test for discrete variables and unpaired t-test, Wilcoxon signed-rank test, or analysis of variance for continuous variables. Survival analyses were performed using the Kaplan-Meier method and the log-rank test. All p values <0.05 were considered statistically signi cant. The hazard ratio and 95% CI were calculated using the univariable Cox proportional hazard model. Parameters with a p value less than 0.1 in the univariable analysis were selected for inclusion in the multivariable analyses. All statistical analyses were performed using the EZR software (Saitama Medical Center, Jichi Medical University, Saitama, Japan) with a graphical user interface for R (The R Foundation for Statistical Computing,