Eosinophil Dynamics May Help Differentiate Immune‐Related From TKI‐Related Adverse Events During IO–TKI Therapy for Metastatic Renal Cell Carcinoma: A Multicenter Retrospective Study

1. Introduction

Advanced and metastatic renal cell carcinoma (mRCC) remains a life‐threatening disease with a poor long‐term prognosis, although the five‐year relative survival rate for RCC has gradually improved over recent decades [1]. In recent years, the treatment landscape has rapidly evolved with the introduction of combination therapy using immune checkpoint inhibitors (ICIs) and tyrosine kinase inhibitors (IO–TKI therapy), which is now widely used as a first‐line treatment and has significantly improved clinical outcomes and prognosis [2, 3, 4, 5].

However, more than 90% of patients receiving IO–TKI therapy experience treatment‐related adverse events (AEs) [2, 3, 4, 5]. These adverse events can be broadly categorized into immune‐related AEs (irAEs) and tyrosine kinase inhibitor (TKI)‐related AEs (TKI‐AEs). Importantly, these toxicities often present with overlapping clinical manifestations, such as diarrhea, skin disorders, and endocrine dysfunction, despite having distinct underlying mechanisms and management strategies [6, 7, 8, 9]. While irAEs typically require immunosuppressive therapy, including corticosteroids, TKI‐AEs are generally managed through dose modification or temporary interruption of TKI treatment [7, 8, 9]. Therefore, accurate differentiation between irAEs and TKI‐AEs is critical for appropriate clinical management. However, no reliable biomarker has been established to distinguish between them.

Eosinophils have emerged as potential biomarkers reflecting immune activation during ICI therapy and have been associated with both treatment efficacy and irAEs [10, 11, 12, 13, 14]. Our previous studies demonstrated that early increases in eosinophils are associated with the development of irAEs [12, 13, 14]. Therefore, we hypothesized that early changes in eosinophil counts may help differentiate irAEs from TKI‐AEs in patients receiving IO–TKI therapy. Accordingly, we conducted this study to investigate whether eosinophil dynamics can serve as a clinically useful biomarker in patients with mRCC.

2. Methods

3. Results

4. Discussion

In this study, eosinophil counts at the start of the second treatment cycle were significantly elevated in patients who developed irAEs compared with patients who experienced TKI‐AEs, and eosinophil counts ≥ 380/μL were associated with approximately 8.39‐fold higher odds of irAEs in multivariable analysis. Although the time to AE onset and follow‐up duration were significantly longer in the irAE group, eosinophil counts remained associated with irAE occurrence after adjustment for time to AE onset. To our knowledge, this is the first study in which eosinophil dynamics were evaluated as a biomarker to differentiate irAEs from TKI‐AEs in patients receiving IO–TKI therapy. While the overall discriminatory performance was modest (AUC = 0.57, sensitivity = 0.30, specificity = 0.94), the absence of significant differences in other hematological parameters—including the NLR and PLR—suggests that eosinophil dynamics carry a degree of specificity within the broader hematological profile that warrants further investigation. Given that eosinophil counts are routinely available as part of standard complete blood count assessments, they may represent a simple and accessible adjunct to clinical judgment.

These findings also have important clinical implications. Given the high specificity (0.94) but low sensitivity (0.30) and modest overall discriminatory performance (AUC = 0.57), elevated eosinophil counts may be more useful to support the diagnosis of irAEs than to exclude them. In other words, markedly increased eosinophil counts during the early stages of treatment may be predictive of irAEs and prompt closer evaluation or earlier consideration of immunosuppressive management. Conversely, the absence of such an elevation should not be interpreted as evidence against the occurrence of irAEs, and clinicians should continue to rely on comprehensive clinical assessment, including symptom patterns, organ involvement, the temporal course, and treatment response. This approach is consistent with current clinical practice guidelines, which emphasize that the diagnosis and management of irAEs should be based on an integrated clinical evaluation rather than a single biomarker [8].

The mechanisms underlying eosinophil elevation during ICI therapy are increasingly recognized. Programmed cell death protein 1 and cytotoxic T‐lymphocyte‐associated protein 4 blockades restore T‐cell‐mediated antitumor immunity but simultaneously disrupt self‐tolerance, leading to autoimmune‐like irAEs across multiple organ systems [6]. This immune‐activated milieu promotes upregulation of interleukin‐5 (IL‐5), the principal driver of eosinophil differentiation and peripheral mobilization [17, 18]. Notably, ICI‐induced CD4+ T cells upregulate IL‐5 production, driving bone marrow eosinophil expansion and systemic eosinophilia, with additional IL‐33 induction further promoting intratumoral eosinophil recruitment; although this finding was reported in breast cancer, this T cell–eosinophil axis may represent a broader mechanism of ICI‐induced immune activation [17]. These findings provide a biological rationale for the early eosinophil elevation observed in patients developing irAEs, consistent with our previous reports across multiple ICI‐based regimens [12, 13, 14].

Although several AEs—including diarrhea, endocrine dysfunction, and skin disorders—are shared by TKI and ICI therapies, TKI‐AEs arise through immune‐independent mechanisms driven by vascular endothelial growth factor receptor (VEGFR) inhibition in normal tissues, fundamentally distinct from the immune dysregulation underlying irAEs [19]. TKI‐induced diarrhea arises from mucosal ischemia secondary to impaired splanchnic perfusion, endocrine dysfunction from disruption of thyroid vascularity, and skin toxicity from compromised vascular repair in cutaneous tissues—all of which reflect VEGFR inhibition in normal organs rather than immune dysregulation [19]. In contrast, the corresponding irAEs are driven by autoimmune T‐cell infiltration and tissue‐specific inflammatory responses [6, 8, 17]. Given that TKI‐AEs are not thought to be mediated through T‐cell activation or IL‐5‐driven eosinophilopoiesis, the eosinophil‐activating cascade triggered by ICIs would not be expected to occur in this context, which may partly account for the divergent eosinophil trajectories observed between groups in the current study.

Distinguishing irAEs from TKI‐AEs is clinically critical, as these toxicities require fundamentally different management strategies. irAEs necessitate prompt immunosuppression with systemic corticosteroids, escalating to infliximab or mycophenolate mofetil in refractory cases, whereas TKI‐AEs are managed through dose reduction or temporary interruption [8, 19]. Misclassification carries substantial consequences in both directions. Unrecognized irAEs risk delayed immunosuppression, leading to irreversible organ damage and preventable morbidity, whereas misattributed TKI‐AEs risk unnecessary corticosteroid exposure, with attendant infectious complications and the attenuation of antitumor immunity. In addition, emerging real‐world evidence, including our previous studies, suggests that treatment interruption in response to AEs in IO–TKI therapy is associated with poorer clinical outcomes, and inadequate management of TKI dose intensity may increase discontinuation rates and adversely affect prognosis [20, 21]. Therefore, inappropriate interruption of TKI therapy owing to misclassification of AEs may have detrimental effects on patient outcomes. In this context, early monitoring of eosinophil dynamics may represent a simple and accessible first step toward distinguishing these two toxicity profiles. In patients with clinically unclassifiable AEs, eosinophil dynamics showed heterogeneous patterns (Table S4), suggesting that eosinophil monitoring may provide supportive information when clinical differentiation is challenging.

This study has several limitations. First, the retrospective design and physician‐based AE classification without centralized adjudication may have introduced misclassification bias. Because AE classification was performed independently at each participating institution, inter‐institutional agreement in classification was not formally assessed. Although eosinophil counts were not used as a criterion for AE classification, the treating physicians were not blinded to laboratory data, and therefore the potential influence of eosinophil values on clinical assessment cannot be completely excluded. The classification incorporated clinical course and response to immunosuppressive therapy, which may be associated with unmeasured biological factors related to eosinophil levels. Therefore, a potential circularity between outcome classification and the biomarker under investigation cannot be completely excluded. Second, the small sample size (n = 123) limits statistical power and generalizability. Third, and most critically, the AUC of 0.57 and sensitivity of 0.30 indicate limited discriminatory performance; although the high specificity (0.94) suggests that substantially elevated eosinophil counts may support an irAE diagnosis when present, normal or mildly elevated counts cannot be used to exclude irAEs, and eosinophil monitoring should be regarded as a supplementary observation rather than a validated biomarker. In addition, the optimal eosinophil cutoff was derived by ROC analysis and subsequently evaluated in the same cohort. Therefore, the observed odds ratio may have been affected by optimistic bias resulting from in‐sample cutoff optimization. External validation using an independent cohort will be required before this cutoff can be considered for clinical application. Fourth, the inclusion of four distinct IO–TKI regimens may have introduced biological heterogeneity. Additionally, the interval to the second treatment cycle varied across regimens, which may have affected the timing of eosinophil assessment. Fifth, AE classification was performed at each participating institution without centralized review, which might have introduced inter‐institutional variability in diagnostic assessment. Sixth, the exploratory analysis of eosinophil dynamics in patients with unclassifiable AEs was limited by a small sample size (n = 18); two patients lacked eosinophil data at the 2‐course sample. Statistical analysis was not feasible, and thus these findings should be interpreted as hypothesis‐generating. Prospective multicenter studies with larger, regimen‐stratified cohorts and serial hematological profiling are warranted to validate these findings.

5. Conclusion

Eosinophil counts at the start of the second treatment cycle were significantly elevated in patients who developed irAEs compared with patients with TKI‐AEs. Early eosinophil monitoring may provide clinically useful supportive information for distinguishing irAEs from TKI‐AEs during IO–TKI therapy for mRCC. Prospective validation in larger cohorts is needed to confirm these findings and establish their clinical utility.

Author Contributions

Takahiro Yasui: investigation, writing – review and editing, supervision. Ryuji Matsumoto: investigation, writing – review and editing. Kengo Kawase: investigation, writing – review and editing. Yoshihiko Tasaki: conceptualization, formal analysis, investigation, data curation, writing – original draft, writing – review and editing. Shimpei Yamashita: investigation, writing – review and editing. Misato Tomita: conceptualization, data curation, writing – original draft, writing – review and editing. Takahiro Osawa: investigation, writing – review and editing. Shuzo Hamamoto: conceptualization, formal analysis, investigation, writing – review and editing. Yoko Furukawa‐Hibi: investigation, writing – review and editing, supervision. Yosuke Sugiyama: investigation, writing – review and editing. Yoshihisa Mimura: investigation, writing – review and editing. Hiroaki Ikoma: investigation, writing – review and editing.

Funding

This study was supported by JSPS KAKENHI (no. 25K18657 to Y. Tasaki).

Ethics Statement

This study was approved by the institutional review boards of all participating centers (approval number: 60‐25‐0072) and conducted in accordance with the ethical standards set out in the 1964 Declaration of Helsinki and its later amendments. The study was reported in accordance with the STROBE guidelines.

Consent

Given the retrospective design, the requirement for written informed consent was waived by the institutional review boards of all participating centers, and an opt‐out approach was applied.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.