Safety Signals of CRS and ICANS With Tarlatamab in Relapsed Small Cell Lung Cancer: Insights From a Small Case Study Authors: Juwhan Choi, Seunghun Lee, Sung Yong Lee Published: 2026 ABSTRACT ABSTRACTBackgroundTarlatamab, a delta‐like ligand 3‐targeted bispecific T‐cell engager, has shown activity in previously treated small cell lung cancer (SCLC); however, real‐world data on cytokine release syndrome (CRS) and immune effector cell‐associated neurotoxicity syndrome (ICANS) remain limited. We evaluated the efficacy and safety of tarlatamab and laboratory findings associated with CRS in patients previously treated for extensive‐stage SCLC.Materials and MethodsWe retrospectively analyzed 11 patients who received tarlatamab through an expanded access program between October 2024 and May 2026. Repeated laboratory measurements were analyzed using generalized estimating equations and mixed‐effects models.ResultsThe objective response and disease control rates were 54.5% and 72.7%, respectively. The mean and median tumor size changes from baseline were −9.1% and −25.0%, respectively. The median progression‐free survival and overall survival were 3.8 months (95% confidence interval [CI]: 1.9–not reached) and 10.1 months (95% CI: 4.5–not reached), respectively. CRS occurred in seven patients (63.6%), with 10 events, all occurring during the first cycle and limited to Grade 1 or 2. No CRS event required tocilizumab treatment, intensive care unit admission, dose reduction, or treatment discontinuation. One patient developed Grade 3 ICANS and recovered fully. CRS events were associated with higher neutrophil percentages, lower lymphocyte percentages, and higher neutrophil‐to‐lymphocyte ratios than no‐CRS events.ConclusionTarlatamab shows encouraging antitumor activity and manageable toxicity in heavily pretreated Korean patients with SCLC. The neutrophil‐to‐lymphocyte ratio may represent an exploratory, concurrent hematologic correlate of CRS.Keywords: cytokine release syndrome, delta‐like ligand 3, immune effector cell‐associated neurotoxicity syndrome, small cell lung cancer Introduction Non‐small cell lung cancer (NSCLC) has witnessed remarkable therapeutic advancements over the past decade, largely driven by the development and implementation of targeted therapies and immune checkpoint inhibitors (ICIs). In contrast, small cell lung cancer (SCLC) continues to present a major therapeutic challenge due to the absence of actionable targeted genomic alterations [1]. Although ICIs have been incorporated into first‐line treatments for extensive‐stage SCLC, the median overall survival (OS) remains approximately 1 year, underscoring the need for more effective therapies [2]. Although recent efforts to expand the therapeutic options for SCLC have led to the development of agents such as lurbinectedin, these agents have not demonstrated dramatic improvements in prognosis [3]. Furthermore, available cytotoxic chemotherapies for SCLC are fewer and generally less effective than those for NSCLC [4]. As a result, the prognosis of SCLC remains significantly worse than that of NSCLC, particularly in patients who progress after first‐line treatments, including ICIs. Tarlatamab is a novel bispecific T‐cell engager (BiTE) antibody construct that targets delta‐like ligand 3 (DLL3), a protein that is highly expressed in most SCLC tumors but minimally expressed in normal tissues [5, 6]. Tarlatamab simultaneously binds to DLL3 on tumor cells and CD3 on T cells, thereby promoting T‐cell activation and expansion [7]. This interaction facilitates the formation of immunological synapses, leading to T cell‐mediated cytotoxicity against DLL3‐expressing tumor cells. In a recent Phase 2 study (DeLLphi‐301), 10 mg tarlatamab administered every 2 weeks demonstrated antitumor activity with durable objective responses and encouraging survival outcomes in patients with previously treated SCLC [8]. However, the mechanism of action of tarlatamab is associated with a distinct toxicity profile, including cytokine release syndrome (CRS) and immune effector cell‐associated neurotoxicity syndrome (ICANS), which differ from the adverse events typically observed with traditional chemotherapy or ICIs [9]. These immune‐mediated toxicities present unique challenges for clinicians in real‐world practice, potentially affecting their broader clinical adoption. Therefore, in this study, we aimed to evaluate the real‐world efficacy and safety profile of tarlatamab, with particular focus on CRS and ICANS, in a cohort of 11 patients with previously treated extensive‐stage SCLC who received tarlatamab through an expanded access program (EAP). Methods We retrospectively analyzed data of 11 patients with extensive‐stage SCLC previously treated with tarlatamab through an EAP between October 2024 and May 2026. Data were collected until May 15, 2026. All patients were hospitalized during the administration of tarlatamab and were discharged only after stabilization of symptoms and vital signs. To enable a detailed assessment of adverse events, we collected comprehensive data, including laboratory findings, vital signs, and medical records, from the initiation of tarlatamab treatment to death. The mean follow‐up duration was 9.5 ± 4.6 months. Adverse events were evaluated according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. We analyzed abnormalities in the complete blood count, electrolyte levels, liver function tests, and creatinine and bilirubin levels, from the initiation of tarlatamab treatment to the first response assessment. CRS and ICANS were defined and graded according to the 2019 American Society for Transplantation and Cellular Therapy (ASTCT) consensus grading system. Of the 11 patients included in the analysis, six had previously received ICI‐based combination therapy, whereas the remaining five had relapsed after surgery or concurrent chemoradiation therapy and had not been treated with ICIs. This study was approved by the Institutional Review Board of Korea University Guro Hospital (approval number: K2025‐1456‐002). Given the retrospective nature of this study, the requirement for informed consent was waived. Patient confidentiality was maintained throughout the study. Statistical Analysis Data were analyzed using the SPSS software (version 20; SPSS for Windows, SPSS Inc., Chicago, IL, USA). The baseline characteristics were summarized according to the occurrence of CRS. Continuous variables are presented as mean ± standard deviation and were compared between patients with and without CRS using Welch's t‐test. Categorical variables are presented as numbers and percentages and were compared using Fisher's exact test or the Fisher–Freeman–Halton exact test, as appropriate. Additionally, because all observed cases of CRS occurred during the first two drug infusion events, we conducted a comparative analysis between days 1 and 8 of the first treatment cycle. Laboratory findings were compared between CRS and no‐CRS events. Because repeated laboratory measurements were obtained from the same patients, generalized estimating equations (GEE) and linear mixed‐effects models were used to account for within‐patient correlations. Patient‐level random intercepts were included in the mixed‐effects model. To account for multiple testing across the laboratory variables, p values from the GEE and linear mixed‐effects analyses were adjusted separately using both the Benjamini–Hochberg false‐discovery rate procedure and the Bonferroni method. The Kaplan–Meier method and log‐rank test were used to analyze progression‐free survival (PFS) and OS. All statistical tests were two‐sided, and a p‐value < 0.05 was considered statistically significant. Results Baseline Characteristics Eleven patients were included in this analysis (Table 1). Seven patients experienced CRS events, whereas four did not. All patients were male, and all except one had a history of smoking. The mean age of the overall cohort was 67.5 ± 7.3 years. Brain metastasis was identified in one patient, and liver metastases were observed in five patients. None of the patients had bone metastases. Most patients received tarlatamab as a later‐line treatment: two as third‐line therapy, five as fourth‐line therapy, two as fifth‐line therapy, and two as seventh‐line therapy. Eight patients had an Eastern Cooperative Oncology Group (ECOG) performance status score of 0. No significant differences were observed in the metastatic sites, histological results, comorbidities, smoking history, ECOG performance status, prior programmed cell death protein 1 (PD‐1)/programmed death‐ligand 1 (PD‐L1) inhibitor therapy, or duration of platinum sensitivity according to the occurrence of CRS. Overall, the occurrence of CRS was not associated with any distinct baseline clinical characteristics in our cohort. TABLE 1Baseline characteristics.CharacteristicTotal (N = 11)CRS present (N = 7)CRS absent (N = 4) p Age, years; mean ± SD67.5 ± 7.365.7 ± 8.070.5 ± 5.40.271Sex, male; n (%)11 (100.0)7 (100.0)4 (100.0)NAPrevious smoking history, yes; n (%)10 (90.9)6 (85.7)4 (100.0)1.000Disease characteristicsBaseline tumor size, mean ± SD8.3 ± 2.97.6 ± 2.69.5 ± 3.40.412Bone metastasis, n (%)0 (0.0)0 (0.0)0 (0.0)NALiver metastasis, n (%)5 (45.5)4 (57.1)1 (25.0)0.545Brain metastasis, n (%)1 (9.1)1 (14.3)0 (0.0)1.000Previous treatment historyPrevious line of therapy, median (IQR)3 (3–4)3 (3–5)3 (2.8–3.2)0.485Prior PD‐1/PD‐L1 inhibitor therapy, n (%)6 (54.5)4 (57.1)2 (50.0)1.000Platinum sensitivity duration, n (%)< 90 days4 (36.4)2 (28.6)2 (50.0)0.576≥ 90 days7 (63.6)5 (71.4)2 (50.0)Whole brain radiotherapy, n (%)1 (9.1)1 (14.3)0 (0.0)1.000ComorbiditiesHypertension, n (%)4 (36.4)3 (42.9)1 (25.0)1.000Diabetes mellitus, n (%)4 (36.4)2 (28.6)2 (50.0)0.545COPD, n (%)5 (45.5)3 (42.9)2 (50.0)1.000IPF, n (%)4 (36.4)3 (42.9)1 (25.0)1.000ECOG performance status, n (%)08 (72.7)6 (85.7)2 (50.0)0.27211 (9.1)1 (14.3)0 (0.0)21 (9.1)0 (0.0)1 (25.0)31 (9.1)0 (0.0)1 (25.0)Abbreviations: COPD, chronic obstructive pulmonary disease; CRS, cytokine release syndrome; ECOG, Eastern Cooperative Oncology Group; IPF, idiopathic pulmonary fibrosis; IQR, interquartile range; NA, not applicable; PD‐1, programmed cell death protein 1; PD‐L1, programmed death‐ligand 1; SD, standard deviation. Baseline characteristics. Abbreviations: COPD, chronic obstructive pulmonary disease; CRS, cytokine release syndrome; ECOG, Eastern Cooperative Oncology Group; IPF, idiopathic pulmonary fibrosis; IQR, interquartile range; NA, not applicable; PD‐1, programmed cell death protein 1; PD‐L1, programmed death‐ligand 1; SD, standard deviation. CRS Of the 11 patients, seven (63.6%) experienced CRS, with 10 CRS events observed (Table 2). Among these events, five were Grade 1, and five were Grade 2, with no occurrence of Grade 3 or higher CRS. All CRS events occurred during the first cycle, either on day 1 or 8. Specifically, three patients experienced CRS on both days 1 and 8, two experienced it only on day 1, and two only on day 8. The median time to CRS onset was 14.0 h (interquartile range [IQR]: 9.3–26.8), with a mean onset time of 18.4 ± 12.2 h. The median duration of CRS was 39.0 h (IQR: 29.3–45.8), and the mean duration was 55.3 ± 60.4 h. Most CRS events were managed with acetaminophen, intravenous hydration, and glucocorticoids and none required tocilizumab treatment. No adverse events leading to the discontinuation or dose reduction of tarlatamab were observed. When comparing the CRS event characteristics between days 1 and 8 during the first cycle, the time to onset of fever was significantly shorter on day 1 (9.0 ± 2.1 vs. 27.8 ± 10.5 h, p = 0.004), and the duration of CRS symptoms was longer (75.0 ± 73.4 vs. 35.6 ± 43.2 h, p = 0.044) (Figure 1). However, no significant differences were observed between the two groups in terms of peak fever, CRS grade, hypotension, or the requirement for oxygen supplementation. Importantly, none of the 10 CRS events required inotropic support or intensive care unit (ICU) admission or resulted in mortality. TABLE 2Comparison of the clinical features of CRS between days 1 and 8 of the first cycle.Total CRS events (N = 10)Day 1 of the first cycle (N = 5)Day 8 of the first cycle (N = 5) p Hours to fever, mean ± SD18.4 ± 12.29.0 ± 2.127.8 ± 10.50.004CRS duration, mean ± SD (hours)55.3 ± 60.475.0 ± 73.435.6 ± 43.20.044Peak fever, mean ± SD (°C)38.8 ± 0.538.9 ± 0.438.8 ± 0.60.317CRS grade15 (50.0%)1 (20.0%)4 (80.0%)0.10325 (50.0%)4 (80.0%)1 (20.0%)≥ 30 (0.0%)0 (0.0%)0 (0.0%)CRS grade defining eventsHypotension4 (40.0%)3 (60.0%)1 (20.0%)0.262Inotropic used0 (0.0%)0 (0.0%)0 (0.0%)Oxygen supplementation1 (10.0%)1 (20.0%)0 (0.0%)0.500Intubation0 (0.0%)0 (0.0%)0 (0.0%)ICU admission0 (0.0%)0 (0.0%)0 (0.0%)CRS managementAntipyretics (e.g., acetaminophen)10 (100.0%)5 (100.0%)5 (100.0%)Steroid5 (50.0%)4 (80.0%)1 (20.0%)0.206Tocilizumab0 (0.0%)0 (0.0%)0 (0.0%)Abbreviations: CRS, cytokine release syndrome; ICU, intensive care unit; SD, standard deviation. Comparison of the clinical features of CRS between days 1 and 8 of the first cycle. Abbreviations: CRS, cytokine release syndrome; ICU, intensive care unit; SD, standard deviation. Body temperature trends during CRS events on days 1 and 8 of the first treatment cycle. CRS, cytokine release syndrome. Laboratory findings were compared at the infusion‐event level between events followed by CRS and those not followed by CRS using repeated‐measures analyses (Table 3). Blood samples were collected, on average, approximately 5 h before tarlatamab administration. Because all CRS events occurred following the cycle 1 day 1 or day 8 infusions, the analysis was restricted to the 22 infusion events at these two time points, comprising 10 events followed by CRS and 12 events not followed by CRS. Of the 11 patients, four contributed both CRS and no‐CRS events, three contributed only CRS events, and four contributed only no‐CRS events. Therefore, the estimated associations incorporate both within‐patient contrasts and between‐patient differences, while the GEE and mixed‐effects models account for the correlation between repeated observations from the same patient. Accordingly, these results should not be interpreted as reflecting purely within‐patient effects. TABLE 3Comparison of laboratory findings according to CRS occurrence among administration events up to day 8 of cycle 1.VariableCRS events (N = 10)No‐CRS events (N = 12)GEE p Mixed model p WBC count (cells/mm3)8230 ± 19877625 ± 32660.4750.415Platelet count (cells/mm3)198 400 ± 70 984231 250 ± 94 8950.0590.071Absolute neutrophil count (cells/mm3)6043 ± 17874647 ± 21290.1130.077Absolute lymphocyte count (cells/mm3)1297 ± 4312136 ± 13130.1110.050Neutrophils (%)72.6 ± 7.361.3 ± 11.9< 0.0010.004Lymphocytes (%)16.2 ± 5.526.9 ± 10.8< 0.001< 0.001Inflammatory cell ratiosNLR5.42 ± 3.632.91 ± 2.080.002< 0.001PLR153.4 ± 46.3134.0 ± 64.20.5820.468MLR0.63 ± 0.380.38 ± 0.210.0170.003ESR (mm/h)71.0 ± 27.363.2 ± 40.40.7500.804CRP (mg/L)19.38 ± 17.0728.16 ± 47.750.6550.631 Note: Because repeated laboratory measurements were obtained from the same patient, p‐values were calculated using GEE to account for within‐patient correlations and linear mixed‐effects models with patient‐level random intercepts. Descriptive statistics were calculated at the event level. Given the small sample sizes, these analyses should be interpreted as exploratory.Abbreviations: CRP, C‐reactive protein; CRS, cytokine release syndrome; ESR, erythrocyte sedimentation rate; MLR, monocyte‐to‐lymphocyte ratio; NLR, neutrophil‐to‐lymphocyte ratio; PLR, platelet‐to‐lymphocyte ratio; WBC, white blood cell. Comparison of laboratory findings according to CRS occurrence among administration events up to day 8 of cycle 1. Note: Because repeated laboratory measurements were obtained from the same patient, p‐values were calculated using GEE to account for within‐patient correlations and linear mixed‐effects models with patient‐level random intercepts. Descriptive statistics were calculated at the event level. Given the small sample sizes, these analyses should be interpreted as exploratory. Abbreviations: CRP, C‐reactive protein; CRS, cytokine release syndrome; ESR, erythrocyte sedimentation rate; MLR, monocyte‐to‐lymphocyte ratio; NLR, neutrophil‐to‐lymphocyte ratio; PLR, platelet‐to‐lymphocyte ratio; WBC, white blood cell. The neutrophil percentage was significantly higher during CRS events than during no‐CRS events (72.6% ± 7.3% vs. 61.3% ± 11.9%; GEE p < 0.001; mixed model p = 0.004). In contrast, the lymphocyte percentage was significantly lower during CRS events (16.2% ± 5.5% vs. 26.9% ± 10.8%; GEE p < 0.001; mixed model p < 0.001). Inflammatory cell ratios were also higher during CRS events, with significantly increased neutrophil‐to‐lymphocyte ratio (NLR; 5.42 ± 3.63 vs. 2.91 ± 2.08; GEE p = 0.002; mixed model p < 0.001) and monocyte ‐to‐lymphocyte ratio (0.63 ± 0.38 vs. 0.38 ± 0.21; GEE p = 0.017; mixed model p = 0.003). Absolute lymphocyte counts were lower during CRS events, although the difference was borderline or not consistently significant across models (1297 ± 431/cells/mm3 vs. 2136 ± 1313/cells/mm3; GEE p = 0.111; mixed model p = 0.050). To account for multiple testing, p values were adjusted using the Benjamini–Hochberg false‐discovery rate and Bonferroni methods. The association between NLR and CRS remained statistically significant in the standard GEE analysis (FDR‐adjusted q = 0.012; Bonferroni‐adjusted p = 0.036) and in the random‐intercept mixed‐effects model (both adjusted p < 0.001). However, the association did not remain significant after multiple‐comparison correction in a small‐sample bias‐reduced GEE analysis, and NLR was not statistically significant when CRS was modeled as a binary outcome. Therefore, because multiple‐comparison correction does not establish independence from other correlated hematologic variables, NLR should be interpreted as an exploratory, concurrent hematologic correlate of CRS rather than an independent or predictive indicator. Overall, given the small sample size and the potential influence of early post‐infusion timing, these findings should be regarded as exploratory and hypothesis‐generating and should not be interpreted as evidence of a confirmed or CRS‐specific inflammatory profile. Rather, the observed changes may represent concurrent, non‐specific hematologic responses associated with CRS or acute post‐infusion stress and require validation in larger prospective studies. ICANS One (9.1%) patient developed ICANS. The patient developed Grade 2 CRS on day 1 without any neurological symptoms. However, on day 8, in the absence of CRS, the patient presented with symptoms consistent with Grade 3 ICANS. The patient became disoriented and gradually developed a decreased level of consciousness. The patient was unable to write sentences, perform simple calculations, or follow basic instructions from the medical staff. Despite repeated questions from family members, the patient repeatedly stared blankly into space with an unfocused gaze and was unable to respond. The onset of ICANS occurred approximately 68 h after drug infusion. Owing to the limited prior experience of the research team with ICANS, corticosteroid therapy was not immediately administered. The patient experienced disorientation for approximately 2 weeks, after which full neurological recovery was achieved. From day 15 of cycle 1 onward, no further neurotoxicity or additional ICANS‐related symptoms were observed during the 117‐day follow‐up period until disease progression. Other Adverse Events Adverse events were evaluated based on the CTCAE version 5.0. The most frequently observed abnormality was elevated aspartate transaminase or alanine transaminase levels, which were detected in 27.3% of the patients. No neutropenia was observed. Most of the 7 documented adverse events were Grade 1, with one instance of Grade 2 anemia, and no Grade 3 or higher hematological toxicities (Table 4). TABLE 4Analysis of adverse events during the first 2 months according to CTCAE version 5.0.Total (N = 11)Grade 1Grade 2Anemia1 (9.1%)1 (9.1%)Thrombocytopenia1 (9.1%)1 (9.1%)Hyponatremia1 (9.1%)1 (9.1%)Increased AST or ALT levels3 (27.3%)3 (27.3%)Elevated serum creatinine levels1 (9.1%)1 (9.1%)Abbreviations: AST, aspartate transaminase; ALT, alanine transaminase; CTCAE, Common Terminology Criteria for Adverse Events. Analysis of adverse events during the first 2 months according to CTCAE version 5.0. Abbreviations: AST, aspartate transaminase; ALT, alanine transaminase; CTCAE, Common Terminology Criteria for Adverse Events. Efficacy Based on the best response evaluation, no patient achieved a complete response. A partial response (PR) was observed in six patients (54.5%), stable disease in two patients (18.2%), and progressive disease in three patients (27.3%) (Table 5). The objective response rate (ORR) was 54.5%, and the disease control rate (DCR) was 72.7%. The waterfall plot shows individual changes in tumor size from baseline according to the best response (Figure 2). Overall, a reduction in tumor size was observed in eight of the 11 patients, whereas the tumor size increased in three patients. The mean change in tumor size from baseline was −9.1%, and the median change was −25.0%, indicating an overall tendency toward tumor shrinkage in the cohort. TABLE 5Tumor response at the time of best response evaluation.Total (N = 11)Complete response0 (0.0%)Partial response6 (54.5%)Stable disease2 (18.2%)Progressive disease3 (27.3%)Objective response rate6 (54.5%)Disease control rate8 (72.7%) Tumor response at the time of best response evaluation. Waterfall plot based on best tumor response evaluation. In the overall cohort, the median PFS was 3.8 months (95% CI: 1.9–not reached), and the median OS was 10.1 months (95% CI: 4.5–not reached) (Figure 3A,B). Survival outcomes were explored according to selected clinical factors using Kaplan–Meier analysis and log‐rank tests (Table 6). When patients were grouped according to CRS occurrence, the median PFS was similar between patients without and with CRS (3.8 months [95% CI: 1.4–not reached] vs. 3.6 months [95% CI: 1.9–not reached]; log‐rank p = 0.884). The median OS was also not significantly different according to CRS occurrence (4.5 months [95% CI: 2.9–not reached] vs. 10.1 months [95% CI: 4.6–not reached]; log‐rank p = 0.912). Kaplan–Meier survival curves for progression‐free survival and overall survival. Kaplan–Meier curves showing (A) progression‐free survival and (B) overall survival in the overall cohort; (C) progression‐free survival and (D) overall survival according to prior PD‐1/PD‐L1 inhibitor therapy; and (E) progression‐free survival and (F) overall survival according to the ECOG performance status. ECOG, Eastern Cooperative Oncology Group; PD‐1, programmed cell death protein 1; PD‐L1, programmed death‐ligand 1. TABLE 6Kaplan–Meier survival analysis according to clinical factors.FactorGroup N Median PFS (months)Median OS (months)PFS log‐rank p OS log‐rank p CRS occurrenceCRS (−)43.84.50.8840.912CRS (+)73.610.10.8840.912Best responsePR66.3NR0.0140.059SD + PD51.98.00.0140.059Prior PD‐1/PD‐L1 inhibitor therapyNo5NRNR0.0210.023Yes63.38.00.0210.023Platinum sensitivity duration< 90 days41.94.60.7920.728≥ 90 days73.810.10.7920.728Previous line of therapy2nd/3rd line73.810.90.3520.390Other lines41.94.60.3520.390Liver metastasisNo63.610.10.6390.421Yes53.88.00.6390.421ECOG performance statusECOG 083.610.90.2790.002ECOG 1–333.84.50.2790.002Abbreviations: CRS, cytokine release syndrome; ECOG, Eastern Cooperative Oncology Group; ICI, immune checkpoint inhibitor; NR, not reached; OS, overall survival; PD, progressive disease; PD‐1, programmed cell death protein 1; PD‐L1, programmed death‐ligand 1; PFS, progression‐free survival; PR, partial response; SD, stable disease. Kaplan–Meier survival analysis according to clinical factors. Abbreviations: CRS, cytokine release syndrome; ECOG, Eastern Cooperative Oncology Group; ICI, immune checkpoint inhibitor; NR, not reached; OS, overall survival; PD, progressive disease; PD‐1, programmed cell death protein 1; PD‐L1, programmed death‐ligand 1; PFS, progression‐free survival; PR, partial response; SD, stable disease. The best response was associated with PFS. Patients who achieved PR had a longer median PFS than those with stable or progressive disease (6.3 months [95% CI: 3.3–not reached] vs. 1.9 months [95% CI: 1.4–3.8]; log‐rank p = 0.014). The median OS was not reached in the PR group (95% CI: 4.5–not reached), whereas it was 8.0 months (95% CI: 2.9–10.9) in the stable disease/progressive disease group, although the difference was not significant (log‐rank p = 0.059). Prior exposure to PD‐1 or PD‐L1 inhibitor therapy was associated with shorter survival outcomes (Figure 3C,D). The median PFS was not reached in patients without prior PD‐1/PD‐L1 inhibitor therapy (95% CI: 1.9–not reached), whereas it was 3.3 months (95% CI: 1.4–3.8) in those with prior exposure (log‐rank p = 0.021). Similarly, the median OS was not reached in patients without prior exposure (95% CI: 4.6–not reached), compared with 8.0 months (95% CI: 2.9–10.9) in those with prior exposure (log‐rank p = 0.023). The ECOG performance status was significantly associated with OS but not PFS (Figure 3E,F). The median PFS was 3.6 months (95% CI: 1.9–not reached) in patients with an ECOG of 0 and 3.8 months (95% CI: 1.4–3.8) in those with an ECOG of 1–3 (log‐rank p = 0.279). However, the median OS was significantly longer in patients with an ECOG of 0 than in those with an ECOG of 1–3 (10.9 months [95% CI: 4.6–not reached] vs. 4.5 months [95% CI: 2.9–8.0]; log‐rank p = 0.002). In contrast, the duration of platinum sensitivity, previous line of therapy, and liver metastasis were not significantly associated with PFS or OS. Given the small sample size, these survival analyses should be interpreted as exploratory. Discussion This study provides real‐world clinical data on the efficacy and safety of tarlatamab in Korean patients with SCLC treated in the second‐line setting or beyond, mirroring the design of the DeLLphi‐304 trial. Compared with the DeLLphi‐304 cohort, our patient population was slightly older, presented with more comorbidities, and received a greater number of prior lines of therapy. Despite these differences, our study demonstrated tumor responses and shrinkage outcomes comparable to those reported in DeLLphi‐304 [10]. Direct comparison with the DeLLphi‐304 trial provides further context for our efficacy findings. In our cohort, the median PFS and OS were 3.8 and 10.1 months, respectively, compared with 4.2 and 13.6 months in the tarlatamab arm of DeLLphi‐304. The ORR and DCR were 54.5% and 72.7% in our study, compared with 35.0% and 68.1%, respectively, in DeLLphi‐304. Although our cohort showed numerically higher response and disease control rates and a broadly comparable median PFS, the median OS was shorter. These findings should not be interpreted as evidence of superior response efficacy because of the small sample size, retrospective design, and substantial differences in treatment setting. Notably, the characteristic CRS and ICANS events were predominantly limited to the first treatment cycle and were mostly mild to moderate in severity (Grade 1 or 2), suggesting that they were clinically manageable. Furthermore, our findings suggest a potential association between NLR and CRS, supporting its role as an exploratory, concurrent hematologic correlate of CRS. Overall, this study confirmed that tarlatamab demonstrates promising efficacy and a tolerable safety profile in real‐world practice, supporting its role as a valuable therapeutic option for patients with SCLC. In the present study, infusion events followed by CRS were associated with a higher neutrophil percentage, a lower lymphocyte percentage, and a higher NLR than events not followed by CRS. These associations were observed in both the GEE and linear mixed‐effects analyses. However, these findings should not be interpreted as evidence that CRS is fundamentally driven by neutrophils or by an increased NLR. The established pathophysiology of CRS primarily involves T‐cell activation and monocyte/macrophage‐mediated release of inflammatory cytokines, including IL‐6, IFN‐γ, and TNF‐α [11]. Furthermore, neutrophilia and lymphopenia are non‐specific hematologic stress responses that may accompany fever, acute illness, or systemic inflammation through mechanisms such as neutrophil demargination and lymphocyte redistribution [12]. Accordingly, the observed NLR elevation should be regarded as an exploratory, event‐level hematologic correlate associated with subsequent CRS rather than as a CRS‐specific mechanistic marker or a validated predictive biomarker. Given the small sample size, the limited number of informative within‐patient comparisons, and the event‐level nature of the analysis, these findings should be considered hypothesis‐generating and require validation through larger prospective studies with serial blood sampling before any predictive or clinical utility can be established. Tarlatamab is a BiTE therapeutic agent designed to simultaneously bind to CD3 on T cells and DLL3 on tumor cells [7]. By engaging CD3, tarlatamab induces T‐cell activation and expansion, promoting cytotoxic activity through the release of effector molecules such as granzyme and ultimately leading to apoptosis of DLL3‐expressing SCLC cells [13]. DLL3, the therapeutic target of tarlatamab, is an inhibitory Notch ligand predominantly localized to the Golgi apparatus in normal cells and is generally not expressed on the cell surface [5, 14]. Conversely, DLL3 is aberrantly overexpressed on the surface of SCLC cells in approximately 85% of the cases, with reported expression levels ranging from 25% to 100% [6]. Importantly, DLL3 expression appears to be independent of the disease stage or prior treatment history, suggesting that it may serve as a consistently reliable therapeutic target across a broad spectrum of clinical scenarios in SCLC. Representative clinical trials of tarlatamab include the Phase II DeLLphi‐301 study, which established the recommended dose and evaluated its efficacy, and the Phase III DeLLphi‐304 study, which compared tarlatamab with standard chemotherapy in a second‐line treatment setting. The results of DeLLphi‐301 demonstrated that the recommended dosing regimen for tarlatamab was 10 mg administered every 2 weeks [8]. The DeLLphi‐304 trial further demonstrated a statistically significant improvement in OS in patients with extensive‐stage SCLC who received tarlatamab as second‐line therapy [10]. Compared with these randomized controlled trials (RCTs), our patients were generally older and received a greater number of prior lines of therapy. Notably, 36.4% of our patients had idiopathic pulmonary fibrosis, a population that is typically excluded from RCTs. These characteristics suggest that our cohort may better reflect real‐world clinical settings than those of previous RCTs. Compared with conventional ICIs, targeted therapies, and cytotoxic chemotherapies, Tarlatamab exhibits a distinct adverse event profile. The most common tarlatamab‐associated toxicities are CRS, which typically presents with fever, and ICANS, which often manifests as mental status changes [15]. CRS is a well‐documented complication of BiTE and Chimeric Antigen Receptor (CAR) T‐cell therapies [16]. It results from a cascade of immune activation, wherein therapeutic agents induce excessive release of inflammatory cytokines such as interferon‐gamma, tumor necrosis factor‐alpha (TNF‐α), and interleukin‐6 (IL‐6) [17]. These cytokines activate macrophages and T cells, perpetuating a cycle of immune amplification and cytokine overproduction, leading to CRS symptoms [18]. In our study, most CRS events occurred during the first cycle and were primarily limited to Grade 1–2 in severity, consistent with findings from the DeLLphi‐301 and DeLLphi‐304 studies with comparable median time to onset and duration of CRS [8, 10]. CRS occurred in 63.6% of patients, with all events confined to the cycle 1 day 1 or day 8 infusion and limited to Grade 1 or 2. This incidence was numerically higher than the CRS rates reported with the 10‐mg dose in DeLLphi‐301 and DeLLphi‐304, which were 51% and 56%, respectively; however, the severity and temporal distribution were similar. In DeLLphi‐301, Grade 1, 2, and 3 CRS occurred in 30%, 20%, and 1% of patients, respectively, whereas the corresponding rates in DeLLphi‐304 were 42%, 13%, and 1%. Most CRS events in both trials occurred after the first two tarlatamab doses. The median time to CRS onset in our study was 14.0 h, closely approximating the 13.1 h reported in DeLLphi‐301, whereas the median duration was shorter in our cohort than in DeLLphi‐301 (39.0 h vs. 4 days). Collectively, these findings indicate that the early onset and predominantly low‐grade CRS profile observed in our real‐world cohort was broadly consistent with that reported in the DeLLphi‐301 and DeLLphi‐304 trials. CRS management was guided by the ASTCT consensus grading scale published in 2019, which classifies the CRS severity and provides corresponding treatment recommendations [19]. Supportive measures, including antipyretics (e.g., acetaminophen), intravenous hydration, and glucocorticoids, are the mainstay of therapy. In severe cases, tocilizumab (a monoclonal antibody targeting the IL‐6 receptor) may be considered [20]. Most CRS events are effectively managed with supportive care, including glucocorticoid administration. While neutropenia was reported in 17% of the patients in the DeLLphi‐301 study, no cases of neutropenia were observed during the 2‐month follow‐up period in our cohort. These findings suggest that tarlatamab is well‐tolerated in real‐world clinical practice. ICANS is characterized by symptoms such as confusion, impaired attention, tremors, and motor weakness [21]. Compared with CRS, ICANS is characterized by a later onset and may persist for a longer duration [22]. In the case of tarlatamab therapy, ICANS has primarily been reported at higher doses, particularly approximately 100 mg, which led to a recommended clinical dose of 10 mg, to mitigate this risk. ICANS events predominantly occur during the first or second treatment cycle. Tarlatamab is generally believed to have limited penetration across the blood–brain barrier (BBB) under normal physiological conditions. ICANS is believed to result from an inflammatory cascade triggered by cytokines such as IL‐6 and TNF‐α that are secreted in response to tarlatamab [23]. These cytokines may induce endothelial cell damage or dysfunction, thereby compromising the BBB integrity [24]. This breakdown could allow further cytokine diffusion or even direct penetration of tarlatamab into the central nervous system, contributing to neurotoxicity. ICANS management primarily involves supportive care, including the administration of corticosteroids when necessary [25]. In our study, ICANS occurred in one of the 10 patients. Due to the limited clinical experience with ICANS, corticosteroid treatment was not promptly administered, resulting in prolonged symptom duration. This underscores the importance of clinician awareness and familiarity with the early identification and management of potentially serious adverse events. This study had certain limitations. First, the sample size was small (11 patients), and the study had a retrospective, single‐center design. Additionally, only one case of ICANS was observed, compared with multiple CRS events, which limited our ability to fully characterize the clinical features of ICANS. However, considering that tarlatamab has not yet been fully approved in Korea and remains a high‐cost therapy, our findings provide valuable real‐world clinical data before broader clinical adoption. Moreover, all patients underwent intensive inpatient monitoring during the early treatment phase, which enabled the close observation and timely management of CRS events. Second, as no cases of grade ≥ 3 CRS occurred in our cohort, we were unable to evaluate the need for or the efficacy of interventions such as tocilizumab treatment or inotropic support, limiting our evaluation of management strategies for severe or life‐threatening adverse events. Similarly, the limited experience in treating ICANS may have resulted in suboptimal management, potentially affecting our ability to evaluate the treatment efficacy and outcomes related to ICANS. Ultimately, the absence of DLL3 expression analysis in tumor tissues and the lack of peripheral cytokine profiling before and after tarlatamab administration hindered a deeper understanding of the pathophysiology of CRS and ICANS. Another limitation is that the biological basis of the elevated pre‐infusion NLR observed in association with subsequent CRS remains uncertain. A higher NLR may partly reflect greater tumor burden, cancer‐related systemic inflammation, or lymphopenia associated with altered host immune status rather than a CRS‐specific predisposition [26, 27]. Moreover, because some CRS events occurred on Cycle 1 Day 8, the pre‐infusion NLR measured at that time point may have been influenced by subclinical immune perturbations induced by the preceding Day 1 tarlatamab exposure. Therefore, the elevated pre‐infusion NLR observed on Day 8 cannot be considered a purely treatment‐naïve antecedent marker and may, at least in part, represent a consequence of prior treatment‐related immune activation. Future studies incorporating such biomarker analyses may offer valuable insights into the predictive and mechanistic aspects of immune‐related toxicity. Overall, this real‐world study supports the clinical activity and manageable safety profile of tarlatamab in heavily pretreated patients with SCLC. Our findings also suggest that NLR may serve as an exploratory hematologic correlate of CRS, although its predictive value remains unproven and requires prospective validation. Given the small sample size and retrospective design, all exploratory associations should be interpreted cautiously. Larger prospective studies are warranted to validate potential biomarkers of treatment efficacy and immune‐related toxicities in patients receiving tarlatamab. Author Contributions Juwhan Choi: conceptualization, investigation, funding acquisition, writing – review and editing, writing – original draft, methodology, validation, visualization, software, formal analysis, project administration, data curation, supervision, resources. Seunghun Lee: conceptualization, methodology, data curation, validation, visualization. Sung Yong Lee: conceptualization, investigation, funding acquisition, writing – original draft, writing – review and editing, visualization, validation, methodology, software, formal analysis, project administration, resources, supervision, data curation. Funding The authors have nothing to report. Disclosure The authors have nothing to report. Ethics Statement This study was reviewed and approved by the Ethics Committee for Clinical Research at the Korea University Guro Hospital (IRB No. K2025‐1456‐002). This study was conducted in accordance with the principles of the Declaration of Helsinki. Conflicts of Interest The authors declare no conflicts of interest. Acknowledgments This work was supported by the Korea University Guro Hospital. Tarlatamab was provided by Amgen Inc. through a named‐patient early access program. This study was conducted independently without funding or involvement of Amgen. Amgen reviewed the manuscript, but the authors retained full control over its content and submission. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy and ethical restrictions. References Associated Data Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy and ethical restrictions.