Safety Evaluation of Tarlatamab in SCLC: A Systematic Review and Meta-Analysis Authors: Elsa Vitale, Lorenza Maistrello, Alessandro Rizzo, Domenico Galetta, Vito Longo Published: 2026 ABSTRACT AbstractIntroductionTarlatamab, a bispecific T-cell engager targeting delta-like ligand 3 and CD3, demonstrated clinical activity both in previously treated SCLC and as intensification of maintenance immunotherapy. Owing to its action mechanism, tarlatamab exerts a new toxicity profile, needing careful management. Based on these premises, we performed a systematic review and meta-analysis to evaluate treatment-related adverse events (TRAEs) among available clinical trials enrolling patients with SCLC receiving tarlatamab monotherapy.MethodsThe research protocol was recorded in PROSPERO register with ID no. 1163787. We included the following keywords: phase I, phase II, phase III, small cell lung cancer, and tarlatamab. We conducted research in Embase, PubMed, Scopus, and Web of Science databases.ResultsA total of four studies with 457 observations were included. Although Egger’s test was performed, the results must be interpreted with caution as the small number of included studies (n = 4) limits the statistical power to reliably detect publication bias.For TRAEs, grades 3 to 4, 385 events were registered assessing the overall prevalence of 42% (95% confidence interval [CI]: 23%–64%). For serious TRAEs, 178 events were registered, and the overall prevalence was estimated at 40% (95% CI: 18%–67%). For TRAEs leading to dose interruptions and dose reduction, 105 events were included, and the prevalence was 24% (95% CI: 16%–34%). For TRAEs leading to discontinuation, 16 events were recorded, and the pooled prevalence was 4% (95% CI: 1.4%–8.2%). For TRAEs for grade 5 events, 12 events were analyzed, and the pooled prevalence was 3.2% (95% CI: 0.4%–22%). For cytokine release syndrome, 255 events were recorded, and the overall was 55.8% (95% CI: 48%–63%). For immune effector cell–associated neurotoxicity syndrome, 40 events were recorded, and the pooled prevalence was 8.9% (95% CI: 3.3%–21.8%). For decreased appetite, 146 events were reported, and the overall prevalence was 31% (95% CI: 23%–41%). For pyrexia, 139 events were analyzed, and the pooled prevalence was 30% (95% CI: 24%–38%). As regards dysgeusia, 457 observations and 139 cases were included, and the pooled prevalence was estimated at 32.5% (95% CI: 20%–48%).ConclusionsFuture research should be focused on identifying risk factors for TRAEs and developing prophylactic strategies to mitigate this risk without blunting antitumor efficacy in patients with SCLC receiving tarlatamab. By improving our understanding and management of tarlatamab TRAEs, we can maximize its clinical efficacy while minimizing harm and ultimately improving outcomes for patients with SCLC.Keywords: Small cell lung cancer, Tarlatamab, Safety Introduction Tarlatamab, a bispecific T-cell engager targeting delta-like ligand 3 (DLL3) and CD3, demonstrated clinical activity both in previously treated SCLC and as intensification of maintenance immunotherapy.1, 2, 3 Despite the incidence of treatment-related adverse events (TRAEs) of grade more than or equal to 3 and treatment discontinuation is lower with tarlatamab compared with systemic chemotherapy, tarlatamab has a drug-class safety profile, including cytokine release syndrome (CRS) and immune effector cell–associated neurotoxicity syndrome (ICANS). CRS in tarlatamab trials occurred in approximately 50% of patients; however, it has been described mainly of grade 1 or grade 2, with most events observed during the first cycle. As regards in the phase 2 Dellphi-301,1 CRS occurred during the second cycle, only in patients who missed one or more dose during the first cycle. At the same time, in the phase 1b Dellphi-303 evaluating tarlatamab as intensification of maintenance immunotherapy,3 four cases of CRS occurred 3 months after the start of the therapy, but all these events were reported after an interruption of the treatment. Similarly, ICANS events were generally of grade 1 or grade 2. Although ICANS mostly occurred during the first 2 cycles, the median time to onset from the first dose of tarlatamab is longer than that of CRS. Even if the CRS and the ICANS tarlatamab-related events are generally of low grade, same cases of life-threatening conditions have been reported, suggesting the need for adequate training of medical staff. Several questions are still open at the moment, in particular the relationship between patient features and risk of CRS or ICANS development, the role of metastases sites and patient comorbidities, the impact of previous treatments, and the impact of radiotherapy. Moreover, tarlatamab had other TRAEs than CRS and ICANS, such as hematologic AEs and dysgeusia, the latter potentially affecting quality of life. On the basis of these premises, we performed a systematic review and meta-analysis to evaluate TRAEs in clinical trials investigating tarlatamab monotherapy. Methods Searching Strategy The research protocol was recorded in PROSPERO register with ID no. 1163787. We included the following keywords: phase I, phase II, phase III, small cell lung cancer, and tarlatamab. We conducted research in Embase, PubMed, Scopus, and Web of Science databases, as found in Supplementary File 1. A total of 18 records were identified (Fig. 1). Of these, six documents did not present any data, and after reading the remaining manuscripts, further eight studies were excluded, and only four manuscripts were definitively included for further analysis. The PRISMA flow chart. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses. Criteria of Eligibility Because these studies were relatively recent, we covered all randomized clinical trials, quasi-experimental, also both oral presentations and abstracts with data recording safety in adult patients (≥18 y) with cancer receiving tarlatamab treatment. Data Selection and Coding Initially, records were identified through a systematic database search and uploaded to a reference management software where duplicate studies were excluded. Then, two independent reviewers (E.V. and L.V.) assessed the title and abstract of the identified studies for inclusion, and unsuitable reports were removed. After that, the articles were uploaded, and the full texts were assessed more closely for eligibility. Disagreements about whether a study should be included or not were resolved by discussion and consensus. If the disagreement remains, arbitration from another reviewer was provided (A.R.). Data collection was extracted considering study characteristics (author, year of publication, design, sample size), participants, and proportion of cases of TRAEs between patients with SCLC treated with tarlatamab. Risk of Bias (Quality) Assessment The presence of Risk of Bias (ROB) (Fig. 2) was investigated using the ROB 2 tool,4 thanks to the following five domains through which it was possible to investigate the presence of biases in the results: 1.Biases arising from the randomization process; Biases arising from the randomization process; 2.Bias due to deviations from the intended interventions; Bias due to deviations from the intended interventions; 3.Bias due to lack of outcome data; Bias due to lack of outcome data; 4.Bias in outcome measurement; Bias in outcome measurement; 5.Bias in the selection of the reported outcome. Bias in the selection of the reported outcome. Risk of bias in RCT included evidence. RCT, randomized controlled trial. Data Synthesis Studies were assessed for quality as per protocol recommendations. Data extracted from the final selected studies were presented using both a narrative approach and tables. Statistical analysis was performed using the R environment (version 4.5.1), and the significance level of the analysis was set at p less than 0.05. The primary outcome of this meta-analysis was the proportion of patients with SCLC treated with tarlatamab who experienced an adverse event. For each type and/or grade of TRAEs, we performed a meta-analysis to estimate the incidence (proportion of patients with the event) of the specific adverse event. Specifically, we performed the following seven meta-analyses: 1.Prevalence of TRAEs for grades 3 to 4; Prevalence of TRAEs for grades 3 to 4; 2.Prevalence of TRAEs for serious TRAEs; Prevalence of TRAEs for serious TRAEs; 3.Prevalence of TRAEs leading to dose interruptions and/or dose reduction; Prevalence of TRAEs leading to dose interruptions and/or dose reduction; 4.Prevalence of TRAEs leading to discontinuation; Prevalence of TRAEs leading to discontinuation; 5.Prevalence of TRAEs for grade 5 events; Prevalence of TRAEs for grade 5 events; 6.Incidence of CRS; Incidence of CRS; 7.ICANS; ICANS; 8.Decreased appetite; Decreased appetite; 9.Pyrexia; and Pyrexia; and 10.Dysgeusia Dysgeusia In each meta-analysis, the effect size for each study was assessed using a logit transformation of the proportion of subjects for whom the TRAE occurred, using the generalized linear mixed model method.5 In every meta-analysis, heterogeneity among studies was assessed using Chi-based Q test. A significant Q value (p < 0.05) suggested the presence of significant heterogeneity between studies. Furthermore, the I2 inconsistency index (I2) was used to estimate the percentage of true variability in the observed effects. I2 values less than 25%, 25% to 50%, 50% to 75%, and more than 75% were interpreted as indicators of low, moderate, high, and very high heterogeneity, respectively.6, 7, 8, 9 Because we expected there to be a high degree of heterogeneity among studies, random-effects pooling models were improved.10 Forest plots were generated for each meta-analysis.11 Publication bias was assessed using a contour-enhanced funnel plot and Egger's t test.12 Results Selected Studies A total of four documents were included in this systematic review and meta-analysis. Table 1 reported all data referring to tarlatamab, administered intravenously, and its related manageable safety profile in patients with SCLC. Because the approved dose for second-line treatment with tarlatamab is 10 mg every 2 weeks, data on other doses have been excluded. Table 1Tarlatamab and Its Related Manageable Safety Profile in Patients With SCLCAuthorPublication YearsRCT PhasePatient No.TRAEs for Grades 3–4 (n)Serious TRAEs (n)TRAEs Leading to Dose Interruptions/Dose Reduction (n)TRAEs Leading to Discontinuation, (n)TRAEs for Grade 5 Events (n)CRS (n)ICANS (n)Decreased Appetite (n)Pyrexia (n)Dysgeusia (n)Ahn et al.2 2023Phase 2n = 100n = 57n = 58n = 31n = 7n = 3n = 49n = 7n = 29n = 35n = 35Chiang et al.13 2025Phase 1n = 88n = 43n = 47n = 23n = 2n = 5n = 54n = 17n = 21n = 26n = 39Dowlati et al.14 2024Phase 1n = 17n = 6n = 3n = 3n = 0n = 3n = 10n = 1n = 7n = 9n = 5Mountzios et al.1 2025Phase 3n = 252n = 67n = 70n = 48n = 7n = 1n = 142n = 15n = 89n = 69n = 61CRS, cytokine release syndrome; ICANS, immune effector cell–associated neurotoxicity syndrome; RCT, randomized controlled trial; TRAE, treatment-related adverse event. Tarlatamab and Its Related Manageable Safety Profile in Patients With SCLC CRS, cytokine release syndrome; ICANS, immune effector cell–associated neurotoxicity syndrome; RCT, randomized controlled trial; TRAE, treatment-related adverse event. Meta-Analysis Meta-Analysis for the Prevalence of TRAEs for Grades 3 to 4 A total of four1,2,13,14 studies were included in the analysis, including 457 observations and 173 events. Heterogeneity among studies was very high and statistically significant (p < 0.001; τ2 = 0.24; I2 = 90.8%, with a 95% confidence interval [CI]: 79.5%–95.9%); thus, a random-effects model was applied. The forest plot (Fig. 3) suggested that the overall prevalence of grade 3 to 4 TRAEs was 42% (95% CI: 23%–64%). Both the symmetry of the funnel plot (Supplementary File 2) and Egger’s test (β = 2.98; 95% CI = −7.2 to 13.18; p = 0.623) suggested the absence of publication bias. Forest plot for the prevalence of TRAEs for grades 3 to 4, the prevalence of serious TRAE, and the prevalence of TRAEs for grade 5. TRAE, treatment-related adverse event. Meta-Analysis for the Prevalence of TRAEs for Serious TRAEs A total of four studies were analyzed, including 457 observations and 178 events. High and significant heterogeneity was detected (p < 0.001; τ2 = 0.41; I2 = 92.1%, with a 95% CI: 82.9%–96.3%), prompting the use of a random-effects model. As found in the forest plot (Fig. 3), the overall prevalence of serious TRAEs was estimated at 40% (95% CI: 18%–67%). No evidence of publication bias emerged from either the symmetry of the funnel plot (Supplementary File 2) or Egger’s test (β = 0.98; 95% CI: −9.68 to 11.66; p = 0.873). Meta-Analysis for the Prevalence of TRAEs for Grade 5 Events Four studies comprising 457 observations and 12 events were analyzed. A very high and significant heterogeneity was observed (p = 0.006; τ2 = 1.39; I2 = 75.8%, with a 95% CI: 33.4%–91.2%), justifying the use of a random-effects model. The pooled prevalence of grade 5 TRAEs was 3.2% (95% CI: 0.4%–22%]), as illustrated in the forest plot (Fig. 3). No evidence of publication bias was detected based on both the symmetrical funnel plot (Supplementary File 2) and Egger’s regression test (β = −3.86; 95% CI: −11.97 to 4.25; p = 0.449). Meta-Analysis for the Prevalence of TRAEs Leading to Dose Interruptions and/or Dose Reduction A total of four studies, encompassing 457 observations and 105 events, were included. Heterogeneity was not statistically significant (p = 0.089; τ2 = 0.04; I2 = 53.9%, with a 95% CI: 0%–84.8%). The pooled prevalence of TRAEs leading to dose interruptions and/or dose reductions was 24% (95% CI: 16%–34%), as found in the forest plot (Fig. 4). Both the funnel plot (Supplementary File 3) and Egger’s test (β = 0.47; 95% CI: −4.2 to 5.1]; p = 0.863) did not reveal signs of publication bias. Forest plot for the prevalence of TRAEs leading to dose interruptions and/or dose reduction and the prevalence of TRAEs leading to discontinuation. TRAE, treatment-related adverse event. Meta-Analysis for the Prevalence of TRAEs Leading to Discontinuation In total, four studies were evaluated, comprising 457 observations and 16 events. There was no significant heterogeneity between studies (p = 0.272; τ2 = 0.04; I2 = 23.2%, with a 95% CI: 0%–88.2%). The forest plot (Fig. 4) illustrates that the pooled prevalence of TRAEs leading to treatment discontinuation was 4% (95% CI: 1.4%–8.2%). The symmetry of the funnel plot (Supplementary File 3) together with Egger’s test (β = −8.1; 95% CI: −4.03 to 2.4; p = 0.669) did not indicate any publication bias. Meta-Analysis for the Incidence of Cytokine Release Syndrome A total of four studies were considered, encompassing 457 observations and 255 events. There was no heterogeneity among studies (p = 0.383; τ2 = 0; I2 = 1.8%, with a 95% CI: 0%–85%). As depicted in the forest plot (Fig. 5), the overall incidence of CRS was 55.8% (95% CI: 48%–63%). The funnel plot symmetry (Supplementary File 4) and Egger’s test (β = 0.22; 95% CI: −2.86 to 3.3; p = 0.902) did not reveal any potential publication bias. Forest plot for the incidence of CRS and the incidence of ICANS. CRS, cytokine release syndrome; ICANS, immune effector cell–associated neurotoxicity syndrome. Meta-Analysis for ICANS The analysis included four studies, totaling 457 observations and 40 events. Significant heterogeneity was observed (p = 0.003; τ2 = 0.26; I2 = 78.3%, with a 95% CI: 42%–92%). Forest plot 3 (Fig. 5) reveals that the pooled prevalence of ICANS was 8.9% (95% CI: 3.3%–21.8%). No publication bias was suggested by both the funnel plot symmetry (Supplementary File 4) and Egger’s regression test (β = −1.15; 95% CI: −7.72 to 5.42; p = 0.764). Meta-Analysis for Decreased Appetite Four studies were included in the analysis, yielding 457 observations and 146 events. The moderate heterogeneity detected was not statistically significant (p = 0.174; τ2 = 0.01; I2 = 39.6%, with a 95% CI: 0%–79.5%]), prompting the adoption of a random-effects model. As illustrated in the forest plot (Fig. 6), the overall prevalence of decreased appetite as an immune-related adverse event was 31% (95% CI: 23%–41%). Both the funnel plot symmetry (Supplementary File 5) and Egger’s test (β = −0.74; 95% CI: −4.44 to 2.96; p = 0.733) supported the absence of publication bias. Forest plot for the prevalence of decreased appetite, the prevalence of pyrexia, and the prevalence of dysgeusia. Meta-Analysis for Pyrexia A total of four studies involving 457 observations and 139 events were analyzed. Heterogeneity among the studies was not statistically significant (p = 0.117; τ2 = 0; I2 = 49.1%, with a 95% CI: 0%–83.2%); consequently, a random-effects model was used. The forest plot (Fig. 6) summarizes the pooled prevalence of pyrexia, which was 30% (95% CI: 24%–38%). No indications of publication bias were found based on funnel plot symmetry (Supplementary File 10) and Egger’s test (β = 3.01; 95% CI: 1.05–4.97]; p = 0.095). Meta-Analysis for Dysgeusia Four studies, encompassing 457 observations and reporting 139 cases of dysgeusia, were included in the analysis. The heterogeneity across studies was very high and statistically significant (p = 0.005; τ2 = 0.9; I2 = 76.8%, with a 95% CI: 36.6%–91.5%]). The forest plot (Fig. 6) illustrated the pooled prevalence of dysgeusia, which was estimated at 32.5% (95% CI: 20%–48%). Assessment of publication bias revealed no evidence of asymmetry in the funnel plot (Supplementary File 5), and Egger’s test was nonsignificant (β = 1.84; 95% CI: −4.58 to 8.25; p = 0.631). Discussion Tarlatamab is a novel, long-acting, bispecific CD3 T-cell engager targeting DLL3.15,16 This agent attaches to DLL3 found on tumor cell surfaces and CD3 on cytotoxic T lymphocytes, leading to the activation of T cells, the release of inflammatory cytokines, and the death of DLL3-expressing tumor cells through CTL action.17 Tarlatamab presents a specific, drug-class safety profile, with events such as CRS and ICANS which needs careful monitoring and prompt intervention.18 Given the remarkable results of phase 3 DeLLphi-304 study, tarlatamab is now considered a new benchmark for second-line therapy in extensive-stage SCLC. This systematic review and meta-analysis is the first in literature aimed to evaluate tarlatamab-related TRAEs among all available SCLC clinical trials available. Especially, we investigated several TRAEs in patients with SCLC receiving tarlatamab. A total of four studies with 457 observations were included.1,2,13,14 Although the symmetry of funnel plots and Egger’s test appears to suggest the absence of publication bias, because of only four studies included in this meta-analysis, these tests lack the statistical power required to be reliable. Therefore, the publication bias could not be reliably assessed due to the small number of included studies. For TRAEs grades 3 to 4, the prevalence was 42%; for serious TRAEs, the prevalence was 40%; and for TRAEs leading to dose interruptions and/or dose reduction, the prevalence was 24%. However, for TRAEs leading to discontinuation and for TRAEs grade5, the pooled prevalence was only of 4% and 3.2%, respectively. Overall, tarlatamab does not exhibit a higher cumulative toxicity compared with conventional chemotherapy; rather, it is characterized by a distinct safety profile with qualitatively different adverse events. Among these, for CRS, the prevalence was 55.8%, whereas, for ICANS, the pooled prevalence was only 8.9%. Regarding dysgeusia, a characteristic side effect of tarlatamab, its pooled prevalence was 32.5%. The management of tarlatamab-characteristic TRAEs represents a new and challenging landscape for medical oncologists. As regard, the knowledge of the kinetics of these specific adverse events may be very helpful for their correct management. Tarlatamab-induced CRS typically occurs with a median onset of 13.5 to 1 hours after dosing, and most often after the first dose in cycle 1 of treatment. However, it is worth noting that tarlatamab-induced CRS may also occur within 1 to 14 days, and the median resolution time is considered to be 3 days. As regards cycles of treatment, less than 50% of cases occur with the first dose, approximately the 30% with the second, and 9% with later doses. On the basis of these premises, patients require close and careful monitoring for 24 hours after the start of tarlatamab infusion, often with 48-hour observation within 1 hour of a health care setting, especially for step-up doses. As regards tarlatamab-induced ICANS, this event has a median onset of 16 days, with a wide range from 1 to almost three years. However, as in the case of CRS, ICANS often appears early, often concurrent with CRS itself in cycle 1, and with a median time to resolution of 4 days.1,2,13,14 With the growing use and availability of tarlatamab monotherapy for previously treated patients with SCLC, effective management strategies for TRAEs—and especially CRS and ICANS—are essential. Our results emphasize that the administration of single-agent tarlatamab necessitates careful selection for outpatient care, which could be even more important in patients with brain metastasis, who may be more vulnerable to ICANS than other patients as recently reported in real-world data.19 Moreover, other TRAEs as decreased appetite and dysgeusia, affect quality of life and at the same time potentially worsen sarcopenia in patients with SCLC, thus requiring a careful therapeutic approach. In particular, dysgeusia is different from chemotherapy-induced dysgeusia, and no treatment approach has been made. This meta-analysis has some limitations and strengths to be considered. First, our analyses relied on aggregate data rather than individual participant data. Consequently, we could not adjust for patient-level confounding factors that may have influenced colitis risk and severity, nor could we evaluate effect modification across these variables. Moreover, the analyses included a limited number of studies, reducing statistical power and precision, and collectively, these limitations indicate that our pooled estimates should be viewed as hypothesis generating rather than definitive. In addition, the phase 2 Dellphi 301 study included also patients treated with tarlatamab 100 mg dose, while regarding the phase 1b Dellphi 303 study. Tarlatamab was administered concurrently with atezolizumab or durvalumab. Therefore, their safety data could be confusing in a global analysis. For this reason, we excluded these patients. Furthermore, real-world toxicity profiles may differ or present higher rates of adverse events than highly selected clinical trial cohorts. In particular, this worse toxicity may be due to poorer performance status, multiple smoking-related comorbidities, and a higher rate of active brain metastases and whole brain radiation compared with highly selected clinical trial cohorts.19 However, our analysis is the first systematic review and meta-analysis in literature aimed at exploring this topic. It has some strengths, including the inclusion of the most recently presented and published data on tarlatamab. Future research should be focused on identifying risk factors for TRAEs and developing prophylactic strategies to mitigate this risk without blunting antitumor efficacy in patients with SCLC receiving tarlatamab. By improving our understanding and management of tarlatamab TRAEs, we can maximize its clinical efficacy while minimizing harm and ultimately improving outcomes for patients with SCLC. CRediT Authorship Contribution Statement Elsa Vitale, Lorenza Maistrello: Conceptualization, Methodology, Software. Elsa Vitale, Alessandro Rizzo, Vito Longo: Data curation, Writing - original draft preparation. Domenico Galetta: Supervision. Alessandro Rizzo: Software, Validation. Elsa Vitale, Alessandro Rizzo, Vito Longo: Writing - reviewing & editing. Disclosure The authors affiliated to the IRCCS Istituto Tumori “Giovanni Paolo II”, Bari are responsible for the views expressed in the article which do not necessarily represent the Institute. Acknowledgments Five per thousand funds for the financial year 2024, resolution 107 of 18 February 2026. Footnotes Drs. Vitale and Maistrello contributed equally to this work. Cite this article as: Vitale E, Maistrello L, Rizzo A, et al. Safety evaluation of tarlatamab in SCLC: a systematic review and meta-analysis. JTO Clin Res Rep 2026;7:101010. Note: To access the supplementary material accompanying this article, visit the online version of the JTO Clinical and Research Reports at www.jtocrr.org and at https://doi.org/10.1016/j.jtocrr.2026.101010. Supplementary Data Supplementary File 1 revised References Associated Data Supplementary Materials Supplementary File 1 revised