Role of eosinophilia in patients with recurrent/metastatic head and neck squamous cell carcinoma treated with nivolumab: Prediction of immune‐related adverse events and favorable outcome

Abstract Introduction Immune‐related adverse events (irAEs) are prognostic factors for patients on nivolumab. However, predictors of irAEs have not yet been identified. We aimed to investigate the predictors of irAEs occurrence and nivolumab discontinuation due to irAEs. Methods Sixty‐two patients with recurrent/metastatic head and neck squamous cell carcinoma received nivolumab therapy between June 2017 and December 2020. Treatment outcome was compared between the groups with or without irAEs. The irAE (+) group was further divided by nivolumab discontinuation. Progression‐free survival (PFS) and overall survival (OS) were compared between the groups. Predictors of irAE occurrence were analyzed. Results Twenty‐one patients (33.9%) developed irAEs, and six (28.6%) discontinued nivolumab due to severe irAEs. The irAE (+) group had significantly longer PFS and OS than the irAE (−) group (median PFS, 12.7 vs. 1.9 months; median OS, 33.1 vs. 12.8 months). The treatment outcomes in the discontinuation group were comparable to those in the non‐discontinuation group. The maximum absolute eosinophil count (AEC) during nivolumab therapy was significantly higher in the irAE (+) group than in the irAE (−) group (548.8 vs. 182) and higher in the discontinuation group than in the non‐discontinuation group (729.3 vs. 368.6). The receiver operating characteristic curve showed that the maximum AEC had a moderate‐to‐high accuracy for predicting irAE occurrence (area under the curve [AUC], 0.757) and nivolumab discontinuation (AUC, 0.893). Discussion Monitoring AEC during nivolumab therapy may be useful in predicting irAE occurrence, nivolumab discontinuation, and disease prognosis.


| INTRODUCTION
Immune checkpoint inhibitors (ICIs), including monoclonal antibodies against programmed cell death, have a significant effect on the treatment of recurrent/metastatic head and neck squamous cell carcinoma (R/M HNSCC). 1,2s a remarkable feature, ICIs induce characteristic adverse events known as "immune-related adverse events (irAEs)" that differ from those induced by traditional cytotoxic agents. 38][9] Severe irAEs may result in deterioration of QOL or treatment-related death.Predicting the occurrence of irAEs and providing early interventions is important.However, it is not clear which patient populations are susceptible to irAEs and ICI discontinuation.On analyzing clinical data of patients with R/M HNSCC who received ICI treatment with nivolumab, we found that patients with a remarkable increase in blood eosinophil counts during ICI treatment and severe irAEs, such as colitis, interstitial pneumonia, and liver dysfunction, experienced nivolumab discontinuation.1][12] Regarding the relationship between eosinophil count and the risk of irAEs, several reports demonstrated that high baseline absolute eosinophil count (AEC) is a potential predictor of irAEs. 13,14However, it has not been elucidated whether a chronological change in AEC during ICI treatment predicts the occurrence and severity of irAEs.Hence, this study aimed to evaluate the relationship between irAEs and the prognosis of HNSCC patients and to assess the predictive role of eosinophilia in irAE occurrence and nivolumab discontinuation due to irAEs.

| Patients and clinical data
This study was approved by the Institutional Review Board of Niigata University Hospital (No. 2019-0172).This study was conducted in accordance with the principles of the Declaration of Helsinki.We retrospectively analyzed the clinical data of patients with R/M HNSCC treated with nivolumab between June 2017 and December 2020.
We reviewed the patients' electronic clinical records and extracted data on age, sex, primary site, recurrent/ metastatic site, Eastern Cooperative Oncology Group performance status (PS) score, and treatment outcomes.Programmed death-ligand 1 (PD-L1) expression in either surgical or biopsy specimens was evaluated by immunohistochemical testing (Dako 22C3 pharmDx, Agilent Technologies/Dako) and grouped according to expression levels, comprising levels of <1% and ≥1% in at least 100 evaluated tumor cells.Data on AEC were obtained from blood samples routinely collected after every nivolumab administration.

| Assessment of treatment outcomes and immune-related adverse events (irAEs)
Progression-free survival (PFS) was defined as the time from the date of the first nivolumab administration to the date of disease progression or death.Overall survival (OS) was defined as the time from the start of nivolumab treatment to the date of death from any cause.IrAEs were evaluated according to the Common Terminology Criteria for Adverse Events, version 4.0.We divided the patients into two groups based on irAE occurrence: irAE (+) and irAE (−) groups.Furthermore, we divided the irAE (+) group into two groups based on nivolumab discontinuation owing to irAEs: discontinuation and non-discontinuation groups.We compared the treatment outcomes between these groups.

| Analysis of predictive factors for irAEs
To identify the potential predictive factors associated with irAE occurrence, we compared the following data between the irAE (+) and irAE (−) groups: baseline characteristics, maximum AEC during nivolumab therapy, and the time lag between the maximum AEC and irAE onset (irAE onset -maximum AEC).The negative and positive values of the time lag indicated that the maximum AEC followed and preceded the irAEs, respectively.We further evaluated the predictive capability of the maximum AEC for irAEs and nivolumab discontinuation using receiver operating characteristic (ROC) curves.

| Statistical analyses
The cutoff date for the analyses of PFS and OS was August 31, 2021.Survival time was estimated using the Kaplan-Meier method and compared using log-rank tests.Analysis of predictive factors was performed using Fisher's exact test or the Mann-Whitney U test.Statistical significance was set at p < 0.05.All statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria).Precisely, it is a modified version of the R commander designed to add statistical functions frequently used in biostatistics. 15 3 |RESULTS

| Details of patients who discontinued nivolumab owing to irAEs
Table 3 shows the details of six patients who discontinued nivolumab owing to irAEs.All patients were male.The irAEs causing nivolumab discontinuation were interstitial pneumonia (grade 2) in three patients and colitis (grade 3), elevated liver enzyme levels (grade 3), and oral mucositis (grade 3) in one patient each.Systemic steroid was administered to one patient with grade 3 colitis, whereas the other three patients with interstitial pneumonia received medications, such as antitussive drug, expectorant, and topical steroid, and the remaining two patients were carefully observed without interventions.All patients recovered from the irAEs.The AEC began to increase from the start of nivolumab treatment (data not shown), and the median maximum AEC was 730 (range, 541-1218).Three out of the six patients in the discontinuation group received sequential chemotherapy, whereas the other three did not.Nonetheless, two of the latter three patients survived for over 1 year.

| Predictive factors for irAEs and nivolumab discontinuation
There were no significant differences between the irAE (+) and (−) groups in age, sex, recurrent/metastatic site, PS score, PD-L1 expression level, prior radiotherapy, and number of prior chemotherapy lines (Table 4).However, the duration of nivolumab administration was significantly longer in the irAE (+) group than in the irAE (−) group (10.7 vs. 1.4 months, p < 0.001, Table 4).The maximum AEC of the irAE (+) group was significantly higher (548.8 vs. 182, p < 0.001) than that of the irAE (−) group, whereas the baseline AEC did not differ between the groups.In the irAE (+) group, the maximum AEC of the discontinuation group was significantly higher (729.3 vs. 368.6,p = 0.018) than that of the non-discontinuation group (Table 4).The time to the maximum AEC was significantly longer in the irAE (+) group than in the irAE (−) group (5.8 vs. 0.9, p < 0.001) (Table 2).Figure 2 shows the time lag between the maximum AEC and irAE onset (irAE onset -maximum AEC) in the non-discontinuation and discontinuation groups.The median time lag in the non-discontinuation group was −3.2 (range, −17.7 to +7.5) months, whereas that in the discontinuation group was 1.0 (range, −0.4 to +6.3) months (p = 0.056, Table 4, Figure 2).Although not statistically significant (p = 0.056), the maximum AEC followed the irAEs in the non-discontinuation group, whereas it preceded the irAEs in the discontinuation group.

| Receiver operating characteristic curve of the maximum absolute eosinophil count (AEC) for predicting irAEs and nivolumab discontinuation
Figure 3A shows the ROC curve of the maximum AEC as a predictor of irAE occurrence.The area under the ROC curve was 0.757 (95% CI, 0.625-0.89).With a cutoff maximum AEC of 312, the sensitivity and specificity for predicting irAEs were 76.2% and 75.6%, respectively.Figure 3B shows the ROC curve of the maximum AEC for predicting nivolumab discontinuation owing to irAEs.The area under the ROC curve was 0.893 (95% CI, 0.811-0.975).With a cutoff maximum AEC of 496, the sensitivity and specificity for predicting discontinuation were 100% and 82.1%, respectively.In this study involving patients with R/M HNSCC who received nivolumab therapy, the survival durations were significantly longer in the irAE (+) group than in the irAE (−) group (Figure 1A,B).These results are consistent with those of a recent systematic review that demonstrated that irAE occurrence was associated with favorable treatment outcomes in various cancer types. 16Among patients with irAEs, those who discontinued nivolumab owing to irAEs (discontinuation group) achieved comparable survival to those who did not discontinue nivolumab (non-discontinuation group) (Figure 1C,D).8][19] In addition, the durable effects of nivolumab, observed in the discontinuation group

F I G U R E 1 Survival analysis according to immune-related adverse event (irAE) occurrence and nivolumab discontinuation owing to (irAEs). (A) Progression-free survival (PFS) according to irAE occurrence:
The median PFS period of patients with irAEs was significantly longer than that of patients without irAEs (12.7 vs. 1.9 months, p < 0.001).(B) Overall survival (OS) according to irAE occurrence: The median OS period of patients with irAEs was significantly longer than that of patients without irAEs (33.1 vs. 12.8 months, p = 0.037).(C) PFS according to nivolumab discontinuation: The median PFS patients who discontinued nivolumab owing to irAEs was comparable to that of patients who did not.Both were significantly longer in those with irAEs than those without irAEs (11.6 vs. 12.7 vs. 1.9 months, p < 0.001).
(D) OS according to nivolumab discontinuation: The median OS period of patients who discontinued nivolumab was remarkably longer than that of the rest, although not at a significant level (not reached vs. 26.9 vs. 12.8 months, p = 0.062).
without sequential chemotherapy (Table 3), would also be responsible for the favorable survival outcomes.Similar durable response to nivolumab has been reported in non-small cell lung carcinoma (NSCLC), 7 metastatic renal cell carcinoma, 8 and HNSCC. 9In patients with NSCLC, the binding of nivolumab to peripheral T cells previously treated with nivolumab is prolonged for more than 20 weeks after the last infusion, 20 which may partly account for the durable effects of nivolumab.

| Predictive factors for irAEs and nivolumab discontinuation
In the present study, we further investigated the predictive factors associated with irAEs and nivolumab discontinuation.In their meta-analysis, Zhou et al. reported that high baseline AEC and a high neutrophil-lymphocyte ratio/platelet-lymphocyte ratio was significantly related to the risk of irAEs. 21Eosinophilia following the initiation of ICI treatment was also noted. 22,23The current results showed that maximum AEC during nivolumab therapy, but not baseline AEC, was significantly higher in the irAE (+) group than in the irAE (−) group (548.8 vs. 182, Table 4), and had an area under the ROC curve of 0.757 (Figure 3A), demonstrating moderate accuracy in predicting irAE occurrence.In the irAE (+) group, the maximum AEC was significantly higher in the discontinuation group than in the non-discontinuation group (729.3 vs. 368.6,Table 4).The maximum AEC had an area under the ROC curve of 0.893 (Figure 3B) for prediction of nivolumab discontinuation owing to irAEs, demonstrating high prediction accuracy.Maximum AEC was a significant predictor of irAEs and nivolumab discontinuation, which were associated with favorable survival outcomes.On analyzing data from the French pharmacovigilance database, Scanvion et al. found that 37 out of 1546 (2.4%) patients treated with ICIs had eosinophilia (≥1000). 23The median time to the AEC peak was 15 weeks, and 21 out of 37 (56.8%)patients manifested eosinophil-related adverse events. 23hile they observed patients with remarkable eosinophilia (≥1000) during ICI therapy, our results showed that patients who developed irAEs had mild eosinophilia (median, 548.8).Eosinophils infiltrate tumor tissues and regulate tumor progression either directly by interacting with tumor cells or indirectly by shaping the tumor microenvironment. 24Considering the latter, eosinophils release chemo-attractants that induce the migration of tumor-specific CD8+ T cells to the tumors, indirectly leading to tumor elimination. 25Certainly, migration of tumor-specific CD8+ cells is also a favorable predictor for response to ICI therapy. 26These data support that   not only baseline eosinophilia but also eosinophilia during ICI therapy may be an important factor for predicting irAEs and treatment outcome.

| Chronological relationship between maximum AEC and irAEs
The chronological relationship between irAE onset and maximum AEC was different between the discontinuation and non-discontinuation groups: The median time lag between the maximum AEC and irAE onset in the discontinuation group was +1 months, whereas it was −3.2 months in the non-discontinuation group (Table 4, Figure 2).If irAEs occur at an early phase after the start of nivolumab (median, 1.4 months), the AEC would reach a peak later (median, 6.8 months), and irAEs would not result in nivolumab discontinuation (Table 4).In contrast, if the irAEs did not occur in the early phase after the start of nivolumab, a remarkable increase in AEC (median, 4.3 months) would precede the occurrence of severe irAEs, requiring nivolumab to be discontinued 1 month after the maximum AEC (Table 4).Consistent with the current results, it has been reported that eosinophilia precedes irAEs, such as hypopituitarism and adrenal deficiency, which develop more than 2 months after nivolumab administration. 27,28All these findings suggest that an increase in AEC is involved in irAE occurrence and nivolumab discontinuation and that careful monitoring of the AEC can help predict the onset of severe irAEs, which is associated with favorable survival outcomes.

| Limitations
This study has some limitations.First, this study was retrospective in nature and the sample size was small.Second, we could not exclude factors other than nivolumab administration that might be related to the eosinophil count (e.g., allergic rhinitis and bronchial asthma).Third, it is unclear whether eosinophilia and subsequent irAEs are characteristic to HNSCC.Future prospective studies are needed to determine how eosinophils affect PFS as well as irAE onset.

T A B L E 3
Details of patients who discontinued nivolumab owing to irAEs.Bold: irAEs that led to nivolumab discontinuation.Abbreviations: AEC, absolute eosinophil count; AWD, alive with disease; DOD, dead on disease; irAEs, immune-related adverse events.T A B L E 4 Predictive factors for irAEs and nivolumab discontinuation.

F I G U R E 2
Time lag between the maximum absolute eosinophil count (AEC) and immune-related adverse event (irAE) onset.The box-and-whisker plot shows the time lag between the maximum AEC and irAE onset.The upper column shows the time lag in the non-discontinuation group.The median time lag was −3.2 (range, −17.7 to +7.5) months.The lower column shows the time lag in the discontinuation group.The median time lag was +1.5 (range, −2.3 to +6.4) months (p = 0.056).Time lag, time lag between the maximum AEC and irAE onset (months); middle vertical lines in columns, median time lag; left vertical lines in columns, first quartile; right vertical lines in columns, third quartile; left bars, minimum value; right bars, maximum value; circles, each data.F I G U R E 3 Predictive capability of absolute eosinophil count (AEC) for immune-related adverse events (irAEs) and nivolumab discontinuation.(A) Receiver operating characteristic curve of AEC indicates irAE occurrence.The area under the curve was 0.757 (95% confidence interval [CI], 0.625-0.89),sensitivity was 0.762, and specificity was 0.756.(B)Receiver operating characteristic curve of AEC as an indicator of nivolumab discontinuation owing to irAEs.The area under the curve was 0.893 (95% CI, 0.811-0.975).The sensitivity and specificity were 1 and 0.821, respectively.
irAE occurrence was a favorable prognostic factor in patients with R/M HNSCC who receive nivolumab.Patients who discontinued nivolumab owing to severe irAEs had comparable treatment outcomes to those who continued nivolumab treatment.The AEC was significantly higher in patients with irAEs and in those who discontinued nivolumab owing to irAEs.Monitoring AEC during nivolumab therapy may be useful for predicting irAE occurrence, nivolumab discontinuation, and prognosis.