Phase II study of 1st-line durvalumab and platinum–etoposide in advanced large-cell neuroendocrine lung carcinoma (aLCNEC) Authors: Elizabeth Dudnik, Hovav Nechushtan, Ofer Rotem, Ekaterina Hanovich, Mira Wollner, Tatiana Boikaner, Niel Fridman, Jonathan Ofer, Merav A Ben-David, Daniel Reinhorn Published: 2026 ABSTRACT AbstractBackgroundLCNEC is a rare high-grade neuroendocrine malignancy with poor prognosis and limited evidence to guide systemic therapy.MethodsProspective, single-arm phase II study of 1st-line durvalumab 1500 mg + cisplatin or carboplatin and etoposide (EP) q3 weeks for 4 cycles, followed by durvalumab 1500 mg q4 weeks, in aLCNEC. The primary endpoint was 12-month PFS (RECIST v1.1, radiologist-assessed); secondary endpoints were ORR, OS, and safety. Sample size followed a single-stage Fleming design (one-sided α = 0.10, power 80%) testing a 12-month PFS rate ≤ 5% (H0) vs ≥ 18% (H1). The study closed early due to slow accrual after twelve of twenty-two planned patients were enrolled.ResultsTwelve patients were treated (median age 68y; 66.7% male; 91.7% smokers; 91.7% stage IVB). After median follow-up of 12.8 months (95% CI 8.5–21.3), 92% had progressed and 75% had died. Median PFS was 4.4 months (95% CI 3.1–12.9); 12-month PFS was 25.0% (95% CI 6.0–50.5), exceeding the ≤ 5% null hypothesis (H0) and compatible with the ≥ 18% target (H1). Median OS was 12.8 months (95% CI 5.2–NR); 12-month OS was 58.3% (95% CI 27.0–80.1). ORR was 50.0%, and DCR was 91.7%. All three patients with untreated brain metastases achieved complete intracranial response. Toxicities were mainly grade 1–2; 41.7% had grade 3 adverse events; three discontinued durvalumab for immune-related toxicity; and no treatment-related deaths occurred.ConclusionsThis first prospective evaluation of durvalumab + EP in aLCNEC, though underpowered, demonstrates efficacy and safety consistent with design expectations.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s00262-026-04464-2.Keywords: Carcinoma, Neuroendocrine, LCNEC, Immunotherapy, Immune checkpoint inhibitors, PD-L1 inhibitors Background Large-cell neuroendocrine carcinoma of the lung (LCNEC) is a rare, high-grade neuroendocrine malignancy comprising fewer than 3% of all lung cancers [1, 2]. LCNEC is characterized by aggressive clinical behavior, high proliferative rates, and a high likelihood of early recurrence [1]. LCNEC shares several features with SCLC [3]. According to 2015 World Health Organization criteria, LCNEC is composed of large cells with neuroendocrine differentiation (as demonstrated by typical morphology and immunoreactivity for chromogranin, synaptophysin, or CD56), and, typically, a high mitotic rate [4]. Platinum–etoposide chemotherapy (EP), typically extrapolated from SCLC treatment paradigms, remains the current standard first-line therapy; objective response rate (ORR) is high (37–52%), however, responses are short-lived, with median progression-free survival (PFS) of 4.6–6.1 months and median overall survival (OS) of approximately 10–11 months [3, 5–8]. Given its rarity, no randomized trials have defined an optimal treatment approach. Growing evidence indicates that immune checkpoint inhibitors (ICI), and specifically, anti-programmed cell death 1 (PD-1) and anti-programmed-death ligand 1 (PD-L1) antibodies, may provide meaningful clinical benefit in LCNEC. Retrospective cohort studies have reported ORR ranging from 13 to 60%, disease control rates (DCR) exceeding 70%, and median PFS of 4.2–14.2 months in patients with advanced disease [9–14]. Several real-world analyses suggest an association between ICI exposure and improved OS in advanced LCNEC [14–18]. However, the retrospective evidence remains contradictory, with other analyses, including the Netherlands Cancer Registry-based analysis, demonstrating no clear superiority of ICI or chemo-ICI combinations over chemotherapy alone [19, 20]. Subgroup analyses from three prospective phase II trials of ICI in rare malignancies included fewer than ten high-grade neuroendocrine carcinoma patients each, with limited responses (0/1, 0/5, and 2/5), and did not report LCNEC-specific outcomes [21–23]. The most robust prospective data to date come from the LANCE non-randomized trial, which evaluated atezolizumab combined with EP in ten LCNEC patients. Reported outcomes included an ORR of 50%, a 12-month PFS rate of 46.7% (95% CI 23.3–93.6), and a 12-month OS rate of 57.1% (95% CI 32.6–100%) [24]. Although outcomes in the EP cohort were inferior, the trial was not designed for direct comparison, and the true contribution of immunotherapy to OS in LCNEC remains uncertain. Notably, each prospective study was limited by the very small number of LCNEC patients, underscoring the difficulty of drawing firm conclusions and the need for additional dedicated trials. Durvalumab, a PD-L1-blocking IgG1 monoclonal antibody, has demonstrated significant survival benefit when combined with EP in extensive-stage SCLC, improving 12-month PFS from 5 to 17% and extending both PFS and OS in the CASPIAN trial [25]. Based on the biological and clinical similarities between SCLC and LCNEC, and the encouraging retrospective data on immune checkpoint inhibition in LCNEC, combining durvalumab with EP may provide synergistic benefit in this population. Therefore, this study was designed to prospectively evaluate the efficacy and safety of durvalumab plus EP in patients with previously untreated advanced LCNEC. Methods Patients Eligible patients had histologically confirmed advanced-stage LCNEC (stage IV or stage III not suitable for definitive treatment, according to American Joint Committee on Cancer staging system, 8th edition [AJCC, 8th edition]; mixed histology with predominant LCNEC component was allowed) without prior systemic therapy for advanced disease; per protocol amendment, a single cycle of EP administered within 4 weeks before enrollment was permitted. Additional key inclusion criteria were measurable disease according to Response Evaluation Criteria in Solid Tumors, version 1.1 (RECIST 1.1), Eastern Cooperative Oncology Group performance status score (ECOG PS) ≤ 1, life expectancy ≥ 3 months, age ≥ 18 years, and adequate hematologic, renal, liver, and thyroid function. Asymptomatic or clinically stable brain metastases after surgery or radiotherapy, not requiring corticosteroids, were allowed. Main exclusion criteria comprised carcinomatous meningitis or a history thereof, active or prior autoimmune or inflammatory disorders, prior allogeneic bone marrow or solid organ transplant, active primary immunodeficiency, active tuberculosis, active hepatitis B or C infection, and known HIV/AIDS. The full list of inclusion and exclusion criteria is provided in Supplemental Material 3. Study design and treatments This was an open-label, non-randomized, single-center pilot phase II study. Patients received intravenous (IV) durvalumab 1500 mg on day 1 of each 21-day cycle in combination with IV etoposide 100 mg/m2 on days 1–3 and investigator’s choice of IV carboplatin at a target area under the plasma concentration–time curve (AUC) of 5 (calculated using the Calvert formula) or IV cisplatin 80 mg/m2 on day 1, for a total of four cycles, followed by maintenance IV durvalumab 1500 mg on day 1 of each 28-day cycle; for patients enrolled after one prior cycle of EP, durvalumab was initiated from cycle 2 onward. Treatment was continued until confirmed systemic disease progression (PD) per RECIST v1.1, unacceptable toxicity, or completion of 24 months of therapy in patients without PD. In the absence of clinically significant deterioration, radiologic progression required confirmation 4 weeks (± 3 days) after the initial PD assessment, and treatment was maintained between initial and confirmatory scans. In cases of unequivocal isolated central nervous system (CNS) progression, brain radiotherapy (stereotactic radiotherapy [SRS] or whole-brain radiotherapy [WBRT]) was given, with systemic therapy interrupted and then resumed at standard dose after radiotherapy. All patients were followed until confirmed PD or death. Study end points and assessments The primary endpoint was the 12-month PFS rate. PFS (RECIST v1.1) was defined as the time from treatment initiation to PD or death from any cause in the absence of PD and was estimated using the Kaplan–Meier method. Secondary endpoints included ORR, DCR, 12-month OS rate, and safety. ORR (RECIST v1.1) was defined as the proportion of patients with the best overall response of complete or partial response. DCR (RECIST v1.1) was defined as the proportion of patients with the best overall response of complete response, partial response, or stable disease. PFS and ORR were assessed by brain/chest/abdominal/pelvic computer tomography (CT) scans every 8 weeks (± 7 days), and patients presenting with brain metastases were additionally monitored with brain magnetic resonance imaging (MRI) every 8 weeks. PFS and ORR were determined by a board-certified radiologist with RECIST v1.1 expertise. OS, defined as the time from treatment initiation to death from any cause, was analyzed using the Kaplan–Meier method. Survival status was assessed every 8 weeks until death or withdrawal of consent. Efficacy was analyzed in the intent-to-treat (ITT) population, comprising all enrolled patients. Safety was assessed based on all-cause adverse events (AE) and treatment-related adverse events (TRAEs), graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0 (CTCAE v5.0) and collected up to 28 days after the last study dose. The safety population included all patients who received at least one dose of study treatment. Exploratory analyses examined the association of tumor PD-L1 expression (tumor proportion score [TPS] by 22C3 immunohistochemistry [IHC], cutoff ≥ 1%), tumor mutational burden (TMB, mutations per megabase [mut/Mb] by next-generation sequencing [NGS], cutoff ≥ 10), and LCNEC molecular subtype (SCLC-like vs non-small cell lung cancer [NSCLC]-like by NGS in accordance with the presence of TP53/RB1 co-mutations or loss) with PFS, OS, and depth of response. Statistical analysis Sample size was determined using a single-stage Fleming design, with a one-sided type I error (α) of 0.10 and power (1–β) of 0.80. The design tested a null hypothesis of a 12-month PFS rate ≤ 5% (H0) against an alternative hypothesis of ≥ 18% (H1). The regimen was to be considered active if at least four of twenty-two treated patients were progression free at 12 months, requiring enrollment of twenty-two eligible patients who initiated study treatment. All statistical analyses were conducted with SAS software, version 9.4 (SAS Institute, Cary, NC, USA). Categorical variables were summarized as frequencies and percentages, continuous variables were reported as medians with corresponding 95% confidence intervals (CIs). PFS and OS were estimated using the Kaplan–Meier method; median PFS/OS and 12-month PFS/OS rates with 95% CIs were derived from the Kaplan–Meier curves. Follow-up time was estimated using the reverse Kaplan–Meier method. Univariate analyses of PFS, OS, and depth of response were performed using Cox proportional hazards regression. Results Baseline patient and tumor characteristics The study stopped early because of slow accrual. Between February 2022 and August 2024, fifteen patients were screened, and twelve patients were enrolled and received the treatment per protocol (Fig. 1). Baseline demographic, clinical, pathological, and molecular tumor characteristics of patients are presented in Table 1. CONSORT diagram. Abbreviations: AE—adverse events; CR—complete response; ECOG PS—Eastern Cooperative Oncology Group performance status score; PD—progressive disease/disease progression; PR—partial response; SCLC—small cell lung cancer; and SD—stable disease Table 1Baseline patient and tumor characteristicsCharacteristicsPatients (n = 12)Age at diagnosis (years), median (range)68 (47–76)Gender, n (%) Female4 (33.3) Male8 (66.7)Smoking, n (%) Current/past smoker11 (91.7) Never smoker1 (8.3)Smoking (pack/years), median (range)60 (30–150)Mixed LCNEC/SCLC histology with predominant LCNEC component, n (%)2 (16.7)Disease TNM stage (AJCC, 8th edition) T category, n (%) T1a0 (0.0) T1b1 (8.3) T1c1 (8.3) T2a0 (0.0) T2b1 (8.3) T32 (16.7) T45 (41.7) TX2 (16.7) N category, n (%) N01 (8.3) N10 (0.0) N26 (50.0) N35 (41.7) M category, n (%) M1a1 (8.3) M1b0 (0.0) M1c11 (91.7)Overall disease stage (AJCC, 8th edition), n (%) IVA1 (8.3) IVB11 (91.7)Disease status, n (%) De novo advanced stage10 (83.3) Recurrent disease2 (16.7)Prior neoadjuvant/adjuvant systemic treatment, n (%)1 (8.3) Platinum-based chemotherapy1 (8.3) ICI0 (0.0)Brain metastases, n (%) Yes3 (25.0) No9 (75.0)Liver metastases, n (%) Yes3 (25.0) No9 (75.0)ECOG PS, n (%) 01 (8.3) 111 (91.7)Ki-67 (%), median (range)72.5 (60.0–90.0)PD-L1 TPS (%), median (range)0.0 (0.0–30.0)PD-L1 TPS, n (%) ≥ 1%1 (8.3) < 1%6 (50.0) NA5 (41.7)TMB (mut/Mb), median (range)6.6 (2.8–41.0)TMB (mut/Mb), n (%) TMB ≥ 103 (25.0) TMB < 105 (41.7) NA4 (33.3)MSI status, n (%) MSI-high0 (0.0) MSI-low8 (66.7) NA4 (33.3)TP53 + RB1 mut/loss n (%) TP53 + RB1 mut/loss present1 (8.3) TP53 + RB1 mut/loss absent8 (66.7) NA3 (25.0)AGA de novo, n (%) AGA de novo present1 (8.3) AGA de novo absent8 (66.7) NA3 (25.0)AGA, Actionable genomic alterations; AJCC, 8th edition, American Joint Committee on Cancer staging system, 8th edition; ECOG PS, Eastern Cooperative Oncology Group performance status score; ICI, immune checkpoint inhibitors; Ki-67, Ki-67 proliferation index; M, metastasis category; MSI, microsatellite instability; mut, mutation; mut/Mb, mutations per megabase; N, lymph node category; NA, not available/not applicable; PD-L1, programmed-death ligand 1; SCLC, small cell lung cancer; T, tumor category; TMB, tumor mutational burden; and TPS, tumor proportion score Baseline patient and tumor characteristics AGA, Actionable genomic alterations; AJCC, 8th edition, American Joint Committee on Cancer staging system, 8th edition; ECOG PS, Eastern Cooperative Oncology Group performance status score; ICI, immune checkpoint inhibitors; Ki-67, Ki-67 proliferation index; M, metastasis category; MSI, microsatellite instability; mut, mutation; mut/Mb, mutations per megabase; N, lymph node category; NA, not available/not applicable; PD-L1, programmed-death ligand 1; SCLC, small cell lung cancer; T, tumor category; TMB, tumor mutational burden; and TPS, tumor proportion score Overall, the main patient and tumor characteristics of the study population were consistent with those commonly reported in LCNEC. The median age in the study cohort was 68 years (range 47–76). Two-thirds of patients were male (66.7%), and nearly all had an extensive history of tobacco use: 91.7% were current or former smokers (median—60 pack/years, range 30–150). Over half of patients (58.4%) had large primary tumors (T3 orT4), nearly all (91.7%) had extensive nodal involvement (N2 or N3) at baseline, and the majority (91.7%) had multiple metastatic sites. Most cases (83.3%) were de novo stage IV, of patients with recurrent disease only one received platinum-based chemotherapy in the adjuvant setting, and none were treated with ICI prior to study enrollment. Brain metastases were present in 25% of patients, and similarly, 25% had liver metastases at baseline. The cohort mainly comprised of patients with ECOG PS 1 (91.7%). Two patients (16.7%) with mixed LCNEC/SCLC histology with a predominant LCNEC component were included after a thorough pathological review. Tumors were highly proliferative, with a median Ki-67 proliferation index (Ki-67) of 72.5% (range 60–90%). PD-L1 expression was assessed in seven patients using 22C3 PharmDx antibody, PD-L1 expression was negative in most cases (median tumor TPS—0%, range 0–30). Nine patients underwent tumor genomic profiling by tissue NGS (Oncomine Comprehensive Assay Plus, n = 8; AVENIO Tumor Tissue Assay, n = 1). Of those, SCLC-like molecular subtype (defined by the presence of TP53/RB1 co-mutations or loss) was only diagnosed in one patient, whereas NSCLC-like molecular subtype (defined by the lack of TP53/RB1 concomitant alterations) predominated. Of patients with NSCLC-like subtype, one patient presented with a de novo actionable genomic alteration (AGA) EGFR exon 19 deletion (ex19 del), and one additional patient had co-occurring STK11 and KEAP1 mutations. The median TMB was 6.6 mut/Mb (range 2.8–41), and 25% of tumors were TMB ≥ 10 mut/Mb. No tumors had high microsatellite instability (MSI). Study treatment Eleven patients (92%) received durvalumab plus etoposide and carboplatin, and one patient (8%) received durvalumab plus etoposide and cisplatin (Fig. 1). The median treatment duration was 3.8 months (range 0.7–17.2; 95% CI 2.1–8.4) (Fig. 2). The median number of EP cycles was 4 (range 2–4), and the median number of durvalumab cycles was 5.5 (range 2–24). Swimmer plot illustrating treatment duration and clinical outcomes with durvalumab plus EP in patients with advanced LCNEC. Each horizontal bar represents an individual patient and depicts time on therapy. Blue bars correspond to ≥ 30% shrinkage, and pink bars correspond to < 30% shrinkage in SLD of target lesion(s) at best response (RECIST v1.1). Triangles indicate radiographic progression per RECIST v1.1, and X marks indicate death. *Death occurred at 34.5 months (beyond the plotted time axis). Abbreviations: EP—platinum–etoposide chemotherapy; LCNEC—large-cell neuroendocrine carcinoma of lung; RECIST v1.1—Response Evaluation Criteria in Solid Tumors, version 1.1; and SLD—sum of longest diameters Four patients (33.3%) had received a single cycle of EP chemotherapy within 4 weeks prior to study enrollment (as allowed per protocol). Three patients (25.0%) underwent palliative radiotherapy during protocol treatment. Post-protocol systemic therapy was given to eight patients (66.7%): All eight (66.7%) received second-line, three (25.0%) received third-line, and one (8.3%) received fourth-line systemic treatment (Fig. 1). Subsequent regimens included EP re-challenge (n = 3), topotecan (n = 5), lurbinectedin (n = 1), osimertinib (n = 1), carboplatin plus nivolumab and ipilimumab (n = 1), and pemetrexed (n = 1). PFS With a median follow-up of 12.8 months (range 3.1–34.5; 95% CI 8.5–21.3), eleven patients (92%) treated per protocol experienced PD according to RECIST v1.1 criteria (Figs. 1 and 2). The median PFS was 4.4 months (95% CI 3.1–12.9), and the 12-month PFS rate was 25.0% (95% CI 6.0–50.5) (Fig. 3). Progression-free survival (RECIST v1.1) (A) and overall survival (B) with durvalumab plus EP in patients with advanced LCNEC (vertical tick marks indicate censored observations). Abbreviations: CI—confidence interval; EP—platinum–etoposide chemotherapy; LCNEC—large-cell neuroendocrine carcinoma of lung; NR—not reached; OS—overall survival; PFS—progression-free survival; and RECIST v1.1—Response Evaluation Criteria in Solid Tumors, version 1.1 OS Nine patients (75%) had died (Figs. 1 and 2). The median OS was 12.8 months (95% CI 5.2- not reached [NR]), and the 12-month OS rate was 58.3% (95% CI 27.0–80.1) (Fig. 3). Objective response Among the twelve evaluable patients, six (50.0%) achieved a partial response (PR), five (41.7%) demonstrated stable disease (SD), and one (8.3%) showed PD as the best response, characterized by unequivocal progression with new lesion development (Fig. 1). No complete responses were observed. ORR was 50.0% (95% CI 21.1–78.9), and DCR was 91.7% (95% CI 61.5–99.8) (Table 2). The maximum reduction in the sum of target lesions ranged from 0% to –80.5% (Fig. 4). Table 2Summary of objective tumor response (RECIST v1.1)Durvalumab + platinum + etoposide (n = 12)Objective response rate, n (%, 95% CI)6 (50.0, 95% CI 21.1–78.9)Best objective response, n (%) Complete response0 (0.0) Partial response6 (50.0) Stable disease5 (41.7) Progressive disease1 (8.3)Disease control rate, n (%, 95% CI)11 (91.7, 95% CI 61.5–99.8) Summary of objective tumor response (RECIST v1.1) Waterfall plot depicting the percent change in SLD of target lesion(s) at best response (RECIST v1.1) following treatment with durvalumab plus EP in patients with advanced LCNEC. Blue bars represent ≥ 30% shrinkage (PR threshold or better), and pink bars represent < 30% shrinkage. PD-L1 status is indicated by stars (purple: ≥ 1%; white: < 1%), TMB status by triangles (purple: ≥ 10 mut/Mb; white: < 10 mut/Mb), TP53/RB1 co-mutation or loss (SCLC-like subtype) by purple circles, STK11 + KEAP1 co-mutation by white circles, and EGFR ex19 del mutation by white squares. Percentages are rounded to the nearest 0.5%. Abbreviations: EP—platinum–etoposide chemotherapy; ex19 del—exon 19 deletion; Ki-67—Ki-67 proliferation index; LCNEC—large-cell neuroendocrine carcinoma of lung; M—metastasis category; mut—mutation; mut/Mb—mutations per megabase; N—lymph node category; PD—progressive disease; PD-L1—programmed-death ligand 1; PR—partial response; RECIST v1.1—Response Evaluation Criteria in Solid Tumors, version 1.1.; SCLC—small cell lung cancer; SLD—sum of longest diameters; T—tumor category; TMB—tumor mutational burden; and TPS—tumor proportion score Three patients presented with asymptomatic brain metastases at baseline. None received brain radiotherapy, and all were monitored with brain MRI every 8 weeks. All three patients achieved a complete intracranial response at best response. Safety profile All-cause AE (any grade) were observed in all patients (100%), and serious adverse events (SAE) occurred in five patients (41.7%) (Table 3). Most toxicities were grade 1–2 (91.7% and 100%, respectively), while grade 3 events were reported in five patients (41.7%). The majority of AE considered treatment-related, with 75% attributed to durvalumab, 66.7% to EP, and 75% related to both treatments. Table 3Summary of adverse events (n = 12): A. AE by category B. TRAEAAE by categoryIncidence, n (%)Any grade all-cause AE12 (100.0)SAE5 (41.7)Maximum toxicity grade Grade 1 AE11 (91.7) Grade 2 AE12 (100.0) Grade 3 AE5 (41.7)AE leading to treatment discontinuation3 (25.0)AE leading to death0 (0.0)Treatment-related AE Related to durvalumab9 (75.0) Related to platinum + etoposide8 (66.7) Related to both durvalumab and platinum + etoposide9 (75.0)Not treatment-related9 (75.0)BTRAEAny grade, n (%)Grade 3/4, n (%)Weakness7 (58.3)1 (8.3)Hypothyroidism5 (41.7)0 (0.0)Colitis2 (16.7)0 (0.0)Arthralgia2 (16.7)0 (0.0)Neutropenia2 (16.7)1 (8.3)Mucositis2 (16.7)0 (0.0)Atrial fibrillation2 (16.7)0 (0.0)ALT/AST elevation2 (16.7)1 (8.3)Alkaline phosphatase elevation2 (16.7)0 (0.0)Phlebitis2 (16.7)0 (0.0)Cellulitis2 (16.7)0 (0.0)COVID-191 (8.3)1 (8.3)Pneumonia1 (8.3)1 (8.3)Fever1 (8.3)0 (0.0)Stevens-Johnson syndrome1 (8.3)1 (8.3)Itching1 (8.3)0 (0.0)Myositis1 (8.3)1 (8.3)Infusion related reaction1 (8.3)0 (0.0)Allergic reaction1 (8.3)0 (0.0)Anemia1 (8.3)0 (0.0)Thrombocytopenia1 (8.3)0 (0.0)Hyperthyroidism1 (8.3)0 (0.0)GGT elevation1 (8.3)0 (0.0)Creatinine elevation1 (8.3)0 (0.0)Hypomagnesemia1 (8.3)0 (0.0)AE, adverse events; ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyl transferase; SAE, serious adverse events; and TRAE, treatment-related adverse events Summary of adverse events (n = 12): A. AE by category B. TRAE AE, adverse events; ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyl transferase; SAE, serious adverse events; and TRAE, treatment-related adverse events The most frequent TRAE were weakness (58.3%) and hypothyroidism (41.7%) (Table 3). Other commonly observed TRAEs, each affecting 16.7% of patients, included colitis, arthralgia, neutropenia, mucositis, atrial fibrillation, transaminase elevation, alkaline phosphatase elevation, phlebitis, and cellulitis. Grade 3/4 TRAEs occurred in a limited number of cases, each representing 8.3% of patients, and included weakness, neutropenia, transaminase elevation, COVID-19 infection, pneumonia, Stevens–Johnson syndrome, and myositis. Three patients discontinued study treatment due to immune-related AE: one case of Stevens–Johnson syndrome, one case of grade 3 transaminase elevation, and one case of grade 3 myositis (Fig. 1). All patients were managed with corticosteroid therapy, achieving complete resolution of symptoms. No treatment-related deaths occurred. Predictive effect of different patient and tumor characteristics In the univariate analysis of PFS, OS, and depth of radiological tumor response, only the Ki-67 proliferation index demonstrated a significant association with radiological tumor shrinkage (p-0.02), with higher Ki-67 values correlating with a higher likelihood of tumor reduction. However, the Ki-67 proliferation index did not have a significant impact on PFS or OS. None of the other evaluated variables—including age, gender, ECOG PS, disease status (de novo advanced vs recurrent disease), the presence of brain or liver metastases, PD-L1 TPS, TMB, or molecular disease subtype (SCLC-like vs NSCLC-like)—showed a significant association with PFS, OS, or depth of radiological tumor shrinkage (Supplemental Material 4). The relationship between tumor stage, morphological and molecular tumor characteristics, and the depth of radiological response is illustrated in Fig. 4. Among the evaluated parameters, only the Ki-67 proliferation index showed a significant association with radiological tumor shrinkage. Notably, four of the six observed responses occurred in PD-L1-negative tumors; one occurred in a patient with concurrent STK11 and KEAP1 mutations, and another in a patient with a de novo EGFR ex19 del. In contrast, high TMB values were not associated with a greater degree of tumor reduction. The only patient with a SCLC-like molecular subtype exhibited a pronounced radiological response. Discussion The present prospective, single-arm phase II pilot study was designed to assess the activity of durvalumab in combination with EP in metastatic pulmonary LCNEC, with 12-month PFS as the primary endpoint in a single-stage Fleming design. The trial closed early due to slow accrual, with 12 of the planned 22 patients initiating protocol therapy, and therefore did not achieve its prespecified statistical power. As a result, the primary endpoint must be interpreted descriptively rather than as a formal test of the study hypothesis. Nonetheless, the observed 12-month PFS rate was 25.0% (95% CI 6.0–50.5). This point estimate is compatible with the assumptions used for trial planning; however, the wide confidence interval reflects the small sample size and substantial statistical uncertainty. Importantly, the lower bound of the confidence interval is close to the null hypothesis threshold of 5%, indicating that the true 12-month PFS rate could be considerably lower than the observed estimate and may not represent a clinically meaningful improvement. Therefore, these data should be interpreted as descriptive and hypothesis-generating rather than confirmatory. While the findings do not establish efficacy, they provide preliminary evidence of activity that may justify further prospective evaluation of durvalumab plus EP in LCNEC, particularly given the biological and clinical rationale for exploring treatment strategies adapted from ES-SCLC, including the CASPIAN regimen [25]. In our cohort, durvalumab plus EP achieved a median OS of 12.8 months and a 12-month OS rate of 58.3% (95% CI 27.0–80.1). These figures are numerically favorable against historical series of EP alone in LCNEC (median OS approximately 10–11 months) [3, 5–8], and they align closely with the results of the LANCE pilot study, in which atezolizumab plus EP yielded a 12-month OS rate of 57.1% (95% CI 32.6–100) and a 12-month PFS rate of 46.7% (95% CI 23.3–93.6) in 10 LCNEC patients [24]. While the 12-month PFS rate in our study appears lower than that reported in LANCE, the overlapping CIs, limited sample sizes, and differences in patient characteristics do not allow for definitive cross-trial comparisons. Notably, the near-identical 12-month OS rates between the two studies suggest a reproducible “tail” on the survival curve with chemo-immunotherapy in LCNEC, echoing the pattern seen with durvalumab plus EP in ES-SCLC in CASPIAN trial, where durvalumab-based chemo-immunotherapy improved median OS to 12.9 months, with 12-month-, 24-month-, and 36-month OS rates of 52.8%, 22.9%, and 17.6%, respectively [26]. Within this context, our trial is the first to prospectively evaluate a durvalumab-based chemo-immunotherapy regimen in LCNEC, extending the evidence base generated by LANCE to this specific PD-L1 inhibitor and supporting the broader concept of importing ES-SCLC immunochemotherapy into LCNEC. The antitumor activity observed in this study was substantial, with an ORR of 50.0% (95% CI 21.1–78.9), and the maximum reduction in target lesion burden reaching − 80.5%, consistent with a deep radiological response in a subset of patients. These findings are highly concordant with the LANCE pilot study, in which atezolizumab plus EP achieved an ORR of 50% [24]. The observed activity within our trial is therefore in keeping with the expectation that adding PD-(L)1 blockade to EP does not compromise the intrinsic chemosensitivity of LCNEC and may contribute to more durable disease control in a proportion of patients. Particularly notable is the intracranial activity observed in our cohort: All three patients with untreated, asymptomatic brain metastases at baseline achieved complete intracranial responses without upfront brain radiotherapy. This finding is clinically relevant because brain metastases occur in approximately 20% of patients with pulmonary LCNEC and are associated with poor outcomes, with retrospective series reporting median OS after brain-metastasis diagnosis of approximately 7 months and frequent use of local therapies such as surgery, whole-brain radiotherapy, or stereotactic radiosurgery [27, 28]. The observation is clinically plausible, as durvalumab plus EP has demonstrated activity in ES-SCLC patients, and a dedicated phase II study in ES-SCLC with untreated brain metastases reported an intracranial response rate of 66.7%, although intracranial control was often not durable and close MRI surveillance remained necessary [29, 30]. Together, these data suggest that durvalumab in combination with EP may provide meaningful CNS activity in selected LCNEC patients, but this requires confirmation in larger, prospective studies with systematic brain imaging and predefined intracranial endpoints. The safety profile of durvalumab plus EP in this small cohort was manageable and broadly consistent with known toxicities of chemo-immunotherapy in lung cancer. Most toxicities were grade 1–2, and grade 3–4 TRAEs were observed in a minority. Although the proportion of discontinuations appears numerically high, this must be interpreted in the context of the very small denominator. Overall, the observed pattern and spectrum of toxicities resemble those reported in CASPIAN and LANCE [24, 25], with no new safety signals emerging in LCNEC, and confirm that durvalumab plus EP is feasible in a patient population often characterized by heavy smoking exposure and comorbid cardiopulmonary disease. Exploratory analyses of prognostic and predictive biomarkers did not identify a robust factor that clearly enriched for benefit from durvalumab plus EP. In univariate analyses, only the Ki-67 proliferation index was significantly associated with the depth of radiological tumor shrinkage, with higher Ki-67 values correlating with greater tumor reduction, but Ki-67 did not significantly influence PFS or OS. Neither PD-L1 expression (predominantly negative in this cohort), TMB (with only 25% of tumors ≥ 10 mut/Mb), nor LCNEC molecular subtype (SCLC-like vs NSCLC-like) showed a clear association with PFS, OS, or response. Responses were documented in PD-L1-negative tumors, in a patient harboring co-occurring STK11 and KEAP1 mutations, and in a patient with a de novo EGFR ex19 del, as well as in the single SCLC-like case, underscoring both the biological heterogeneity of LCNEC and the limitations of extrapolating biomarker paradigms from NSCLC or SCLC to this entity. These findings echo prior retrospective reports and emphasize the need for larger, genomically annotated cohorts. Although the maximum OS observed in the present cohort was 34.5 months, this single long-survival event should not be interpreted as evidence of a distinct long-term survivor subgroup or durable survival plateau. Exploratory review of the molecular profile of this outlier did not identify a specific molecular feature that could explain the prolonged outcome. LCNEC is increasingly recognized as a biologically heterogeneous disease, encompassing SCLC-like and NSCLC-like molecular subsets, as well as a rare subgroup of LCNEC harboring de novo actionable genomic alterations. Therapeutic value of matched targeted therapy in de novo actionable genomic alteration-driven LCNEC remains undefined, although the available retrospective and case-series evidence suggests a strong activity signal [31–34]. For instance, in the series of Heijboer et al., partial responses were reported in 19 of 38 small-molecule-inhibitor treatment lines, with a median OS of 14.6 months [31]. In our prospective cohort, one patient with a de novo EGFR ex19 del LCNEC demonstrated a prolonged response to durvalumab plus EP, suggesting that although genotype-matched targeted therapy is an important consideration in this LCNEC subset, the potential value of chemo-immunotherapy in this subgroup should not be overlooked. This study has several limitations. Early closure and the resulting sample size of 12 patients markedly limit the precision of efficacy and safety estimates, preclude the planned formal Fleming design analysis, and make all comparisons with historical series and LANCE [24] purely exploratory. The wide CIs around PFS and OS further hinder interpretation of modest differences across studies. In addition, the small cohort prevents robust correlations between outcomes and clinicopathologic or molecular features, and the single-arm, single-center design may introduce selection and institutional bias. Despite these limitations, this study has important strengths. It is one of the few prospective, protocol-driven evaluations of chemo-immunotherapy in LCNEC, with standardized eligibility criteria and treatment, and rigorous imaging follow-up using RECIST v1.1, reviewed by an experienced board-certified radiologist. The long duration and maturity of follow-up constitute an additional strength of this study. The efficacy signals for PFS, OS, and ORR are consistent internally and align closely with both the LANCE pilot study in LCNEC and the CASPIAN trial in ES-SCLC [24, 25], reinforcing the biological plausibility and clinical relevance of PD-L1 blockade in combination with EP in this histology. The detailed safety characterization, including immune-mediated events and their management, further supports the feasibility of this regimen in routine practice. Importantly, the observation of durable benefit in a subset of patients and complete intracranial responses without radiotherapy highlights the potential for meaningful long-term disease control with chemo-immunotherapy in LCNEC. In conclusion, despite being underpowered due to premature closure, this pilot phase II trial represents the first prospective evaluation of a durvalumab-based chemo-immunotherapy regimen in advanced LCNEC. The results suggest that durvalumab plus EP provides clinically relevant antitumor activity and manageable toxicity in advanced LCNEC, with PFS, OS, and ORR outcomes that are concordant with expectations from the study’s statistical design and consistent with external evidence. These findings support the integration of immune checkpoint inhibition into first-line treatment for advanced LCNEC and provide a rationale for ongoing and planned multicenter studies, including NECTAR/LOGIK2401 [35], ALPINE (NCT05470595), DUPLE (NCT06418087), FIRST-NEC/GFPC 01–2022 (NCT06393816), that are needed to confirm these results. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (PDF 1261 kb) Supplementary file2 (PDF 84 kb) Supplementary file3 (PDF 107 kb) Supplementary file4 (PDF 157 kb) Acknowledgements None. Abbreviations Adverse events Actionable genomic alterations American Joint Committee on Cancer staging system, 8th edition Alanine aminotransferase Aspartate aminotransferase Area under the plasma concentration–time curve Confidence interval Central nervous system Complete response Computer tomography National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0 Disease control rate Eastern Cooperative Oncology Group performance status score Platinum–etoposide chemotherapy Extensive-stage Exon 19 deletion Gamma-glutamyl transferase Hazard ratio Immune checkpoint inhibitors Immunohistochemistry; Institutional Review Board (Ethics) Committee Intent-to-treat Intravenous Ki-67 proliferation index Large-cell neuroendocrine carcinoma of lung Metastasis category Magnetic resonance imaging Microsatellite instability Mutation Mutations per megabase Lymph node category Not available/not applicable Next-generation sequencing Not reached Non-small cell lung cancer Objective response rate Overall survival Progressive disease/disease progression Programmed-death 1 Programmed-death ligand 1 Progression-free survival Partial response Response Evaluation Criteria in Solid Tumors, version 1.1 Serious adverse events Small cell lung cancer Stable disease Sum of longest diameters Stereotactic radiotherapy Tumor category Tumor mutational burden Tumor proportion score Treatment-related adverse events Whole-brain radiotherapy Author contributions Conceptualization and methodology—ED; data acquisition, analysis and interpretation—all authors; investigation, data curation, validation—ED; original draft, visualization—ED, DR; review & editing—all authors; supervision, project administration, funding acquisition—ED; final approval and accountability—all authors. Funding This investigator-initiated trial was funded by AstraZeneca as an externally funded research (funding provided to the participating center). Data availability The datasets generated and/or analyzed during the current study are not publicly available due to privacy and ethical restrictions, but are available from the corresponding author on reasonable request and subject to applicable approvals and data-sharing agreements. Supporting documents, including the full study protocol and Helsinki Committee approval, are provided in the Supplementary Materials (Supplemental Material 1, 2). Declarations Conflict of interest Conflict of interest statement ED reports research grants from AstraZeneca (Inst.), Boehringer Ingelheim (Inst.), and Takeda (Inst.), honoraria/consulting/advisory fees from Roche, Merck Sharpe & Dohme, Bristol-Myers Squibb, AstraZeneca, Takeda, Pfizer, Novartis, Janssen, Sanofi, Merck Serono, Medison, Rhenium, and Progenetics, also serves a member of the Assuta IRB Committee. HN is a co-inventor on a patent for LY6K-targeting antibodies and a member of the Hadassah IRB Committee. OR reports honoraria/consulting/advisory fees/travel support from AstraZeneca, Boehringer Ingelheim, Merck Sharpe & Dohme, Bristol-Myers Squibb, Novartis, Pfizer, Roche, Takeda, Teva, Rhenium. DR reports honoraria/consulting/advisory fees from AstraZeneca, Bristol-Myers Squibb, Merck Sharpe & Dohme. All other authors declare no conflicts of interest. Ethical approval The study was conducted in accordance with the ethical standards. Approval from the Assuta Institutional Helsinki Committee was obtained prior to study initiation (protocol approval number ASMC-0072–21). Consent for publication Written informed consent for trial participation was obtained from all participants; no identifiable individual-level data are reported in this manuscript. Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work, the authors used ChatGPT (version 5.1; OpenAI) to provide language editing and writing assistance. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary file1 (PDF 1261 kb) Supplementary file2 (PDF 84 kb) Supplementary file3 (PDF 107 kb) Supplementary file4 (PDF 157 kb) Data Availability Statement The datasets generated and/or analyzed during the current study are not publicly available due to privacy and ethical restrictions, but are available from the corresponding author on reasonable request and subject to applicable approvals and data-sharing agreements. Supporting documents, including the full study protocol and Helsinki Committee approval, are provided in the Supplementary Materials (Supplemental Material 1, 2).