Furmonertinib-based triplet therapy improves survival in EGFR-mutant NSCLC with leptomeningeal metastases: a large-scale multicenter retrospective study (FURMO-LM) Authors: H. Chen, H. Xu, Z. Yin, X. Li, G. Lin, T. Xin, H. Yang, Q. Zhao, W. Cheng, L. Zhang, H. Li, W. Xu, H. Zheng, C. Fan, F. Wang, L. Guo, Y. Wu, Y. Wu, S. Yang, X. Tang, C. Zhu, S. Fang, Q. Wang Published: 2026 ABSTRACT BackgroundLeptomeningeal metastases (LM) in epidermal growth factor receptor mutant (EGFRm) non-small-cell lung cancer (NSCLC) represent a major therapeutic challenge. Despite the improved central nervous system penetration of third-generation EGFR tyrosine kinase inhibitors (TKIs), key uncertainties persist regarding the efficacy of combination regimens with antiangiogenic therapy and intrathecal chemotherapy and optimal dosing because of the limited evidence.Patients and methodsIn this large-scale, multicenter, retrospective study, 618 patients with EGFRm NSCLC and LM who were treated with furmonertinib-based regimens between March 2021 and December 2024 were enrolled. The primary endpoint was overall survival (OS). Secondary endpoints included OS from LM diagnosis (LMOS), time to treatment failure (TTF), clinical response rate (CRR), LM-specific objective response rate (ORR-LM), and treatment-related adverse events (TRAEs).ResultsThe median OS was 14.19 months [95% confidence interval (CI) 13.04-16.79], median LMOS was 17.05 months (95% CI 14.81-19.22), and median TTF was 11.23 months (95% CI 10.23-12.13). Triple therapy (furmonertinib plus antiangiogenic therapy and chemotherapy) was associated with longer median OS than mono or dual regimens (17.91 versus 10.31-14.19 months, P < 0.001). Notably, among patients with prior third-generation EGFR-TKI exposure, triplet therapy achieved a median OS of 17.18 months. In patients with poor performance status (Eastern Cooperative Oncology Group score 3-4), triplet therapy showed markedly improved OS compared with other regimens (17.31 versus 9.02-12.81 months, P = 0.003). Regarding furmonertinib dosing, 80 mg/day and 160 mg/day were associated with longer OS than 240 mg/day (16.79/14.68 versus 11.39 months, P = 0.002). The CRR was 66.7%, and the ORR-LM was 40.6%. TRAEs occurred in 59.2% of patients, with grade 3-4 events reported in 7.9%.ConclusionsThis large real-world study suggests that furmonertinib-based therapy, particularly the triplet regimen with antiangiogenic agent and chemotherapy, is associated with meaningful clinical outcomes over mono or dual therapy in patients with EGFRm NSCLC and LM, including those with prior third-generation TKI resistance or poor performance status. Introduction Leptomeningeal metastases (LM) of non-small-cell lung cancer (NSCLC) represent one of the most devastating complications of the disease, significantly shortening overall survival (OS) and worsening patients’ quality of life.1,2 In patients with epidermal growth factor receptor mutant (EGFRm) NSCLC, LM poses an even greater threat, with a notably higher incidence compared with EGFR wild-type disease.2,3 Moreover, with the prolongation of survival, the incidence of LM has steadily increased.4 However, the treatment for LM of EGFRm NSCLC remains highly limited, and the prognosis is still poor. In the management of LM, numerous recent studies have investigated novel therapeutic strategies, including systemic treatment with antibody–drug conjugates,5 intrathecal administration of PD-1 inhibitors,6,7 and intrathecal chimeric antigen receptor T-cell therapy (Clinical Trials.gov identifier NCT03696030). Proton craniospinal irradiation has also shown preliminary efficacy in treating LM in some solid tumors.8 However, these treatment approaches are not yet fully established. For patients with EGFRm NSCLC and LM, EGFR-tyrosine kinase inhibitors (TKIs) remain the cornerstone of therapy. Previous studies have shown that EGFR-TKI treatment prolongs OS compared with no EGFR-TKI therapy in patients with EGFRm NSCLC and LM.2,9 Nevertheless, the efficacy of first-generation and second-generation EGFR-TKIs in patients with LM has been unsatisfactory because of poor blood–brain barrier (BBB) penetration, with median OS of only ∼4 months.10, 11, 12 To increase intracranial drug concentrations, several studies have explored dose escalation of EGFR-TKIs, showing modest improvement.13,14 The third-generation EGFR-TKI osimertinib exhibits stronger BBB penetration and improved outcome in LM.15,16 Furmonertinib, another third-generation EGFR-TKI approved in China for EGFRm NSCLC, is typically dosed at 80 mg daily. High-dose furmonertinib (240 mg daily) has also demonstrated notable clinical efficacy in patients with EGFRm NSCLC and LM in our previous study.17 However, several challenges remain in the application of third-generation EGFR-TKIs for LM. First, although substantial evidence supports the value of intrathecal chemotherapy and antiangiogenic therapy for LM, leading to their widespread combination with EGFR-TKIs in clinical practice,8,18, 19, 20, 21 direct comparative evidence between combination regimens and EGFR-TKI monotherapy remains critically scarce. Second, most existing studies are limited by small sample sizes and considerable heterogeneity in TKI dosing across studies.15, 16, 17,22, 23, 24 To address these gaps, we conducted this large-scale multicenter study to evaluate furmonertinib-based treatment strategies, dosing regimens, and prognostic factors, providing real-world evidence to help inform the management of EGFRm NSCLC with LM. Patients and methods Study design and patients The present study was a retrospective multicenter real-world investigation of patients with EGFRm NSCLC and LM. A total of 707 patients from eight centers were screened, of whom 618 were enrolled. This study was approved by the Institutional Review Board of Henan Cancer Hospital and was registered on ClinicalTrials.gov (identifier NCT06926803). Patients were eligible if they met the following criteria: (i) aged ≥18 years, male or female; (ii) histologically or cytologically confirmed NSCLC; (iii) presence of classic or uncommon EGFRm confirmed by genetic testing; (iv) LM diagnosis according to European Association of Neuro-Oncology–European Society for Medical Oncology criteria25: type I (definite), which requires positive cerebrospinal fluid (CSF) cytology, or type II (probable), which requires typical clinical and magnetic resonance imaging (MRI) findings with negative cytology. Ambiguous cases were centrally adjudicated by a blinded expert panel before analysis; and (v) received furmonertinib treatment for >1 month between 3 March 2021 and 31 December 2024, with or without combination therapy, including radiotherapy, intrathecal chemotherapy, and antiangiogenic therapy. Patients were excluded if they had concurrent malignant tumors. Assessment Patients were followed up by a dedicated team of investigators. Objective response of LM on MRI was assessed according to the Response Assessment in Neuro-Oncology (RANO)-LM radiological criteria, and classified as complete response, partial response, stable disease, or progressive disease.26 All imaging was independently reviewed by two experienced radiologists, with excellent interrater agreement for RANO-LM response categories [kappa = 0.86, 95% confidence interval (CI) 0.81-0.91]; discrepancies were resolved by consensus. Clinical response, defined as sustained (≥4 weeks) improvement in neurologic symptoms or performance status,20 was assessed by the treating physician and verified retrospectively by the study team; discrepancies were resolved by a second investigator. The primary endpoint was OS, defined as time from furmonertinib initiation to death from any cause or last follow-up. Secondary endpoints included LMOS (time from LM diagnosis to death or last follow-up), time to treatment failure (TTF, time from first dose to permanent discontinuation for any reason), clinical response rate (CRR), LM objective response rate (ORR-LM), and treatment-related adverse events (TRAEs), graded according to the Common Terminology Criteria for Adverse Events version 5.0. Statistical analysis Descriptive statistics were used to summarize baseline data. Missing baseline data were imputed using multiple imputation (Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.107743). Survival curves for OS, LMOS, and TTF were generated using the Kaplan–Meier method and compared with log-rank test. Cox proportional hazards models were applied to estimate hazard ratios (HRs) with 95% CIs. Variables with P < 0.10 in univariate analysis were included in multivariable models. Intrathecal chemotherapy was excluded from the multivariable model owing to multicollinearity with treatment regimen (variance inflation factor >5). Inverse probability weighting (IPW) analyses were conducted to evaluate OS and LMOS across treatment regimens using the imputed dataset. Response rates were expressed as proportions with exact binomial 95% CIs and compared using χ2 or Fisher’s exact tests. All tests were two-sided (α = 0.05). Analyses were carried out in R (version 4.5.1) using the survival and survminer packages, without adjustment for multiple comparisons. Sensitivity analyses compared complete-case and imputed datasets to confirm robustness. Results Patient characteristics The baseline characteristics of the 618 patients with EGFRm NSCLC and LM are summarized in Table 1. The median age of the patients was 57.7 years. All cases were adenocarcinoma. LM was diagnosed as primary in 52 patients (8.4%) and secondary in 562 patients (91.6%); the median interval from lung cancer diagnosis to LM was 25.1 months (range 0-176.8). EGFRm subtypes were as follows: 221 cases (35.8%) with exon 19 deletion (19del), 319 cases (51.6%) with exon 21 point mutation L858R, 24 cases (3.9%) with exon 20 insertion (20ins), and 54 cases (8.7%) with other uncommon mutations. Details of the uncommon mutations are provided in Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2026.107743. Type I LM was present in 478 patients (77.3%) and type II in 140 patients (22.7%). Of the patients, 187 (30.3%) had an Eastern Cooperative Oncology Group (ECOG) score of 0-2, and 325 (52.6%) had an ECOG score of 3-4. Of the patients, 370 (59.9%) had undergone at least two lines of systemic therapy, and 472 patients (76.4%) had prior exposure to a third-generation EGFR-TKI. Among these, 111 (18.0%) patients had received furmonertinib before LM diagnosis. Treatment regimens included furmonertinib alone [80 patients (12.9%)] and combination therapies: 54 patients (8.7%) received antiangiogenic therapy, 178 patients (28.8%) received chemotherapy, and 306 patients (49.5%) received both antiangiogenic therapy and chemotherapy. Baseline characteristics of the four treatment groups and detailed combination treatment regimens are summarized in Supplementary Tables S3 and S4, available at https://doi.org/10.1016/j.esmoop.2026.107743. Among the 484 patients who received chemotherapy, 454 (93.8%) underwent intrathecal chemotherapy. Intrathecal chemotherapy was administered via lumbar puncture or Ommaya reservoir. The most commonly used regimen for intrathecal chemotherapy was pemetrexed (75.6%). Regarding furmonertinib dosing, 88 patients (14.2%) received 80 mg/day, 407 patients (65.9%) received 160 mg/day, and 123 patients (19.9%) received 240 mg/day. Eighty-three patients (13.4%) received brain radiotherapy during furmonertinib treatment, of which 66 (10.7%) received whole-brain radiotherapy (Supplementary Table S5, available at https://doi.org/10.1016/j.esmoop.2026.107743). Given the presence of missing data in variables such as ECOG score and prior treatment lines, these were addressed through statistical imputation. The imputed values are detailed in Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.107743.Table 1Basic characteristics of the participantsVariablesN = 618Sex Female377 (61.0) Male241 (39.0)Age (years)57.7 ± 9.6Age group (years) <65472 (76.4) ≥65146 (23.6)Smoking No498 (80.6) Yes120 (19.4)Time of LMa (months)25.1 (0-176.8)EGFR mutation type 19del221 (35.8) 20ins24 (3.9) L858R319 (51.6) Uncommon mutations54 (8.7)Brain metastases No200 (32.4) Yes418 (67.6)Type of LM Type I478 (77.3) Type II140 (22.7)Intracranial pressure before treatment (mmH2O) ≤180111 (18.0) >180184 (29.8) NA323 (52.3)WHO/ECOG performance status 0-2187 (30.3) 3-4325 (52.6) NA106 (17.2)Prior EGFR-TKI therapy No58 (9.4) Yes560 (90.6)Prior third-generation EGFR-TKI therapy No146 (23.6) Yes472 (76.4)Treatment regimens Furmonertinib monotherapy80 (12.9) Furmonertinib + antiangiogenic therapy54 (8.7) Furmonertinib + chemotherapy178 (28.8) Furmonertinib + antiangiogenic therapy + chemotherapy306 (49.5)Dose of furmonertinib (mg) 8088 (14.2) 160407 (65.9) 240123 (19.9)Intrathecal chemotherapy No164 (26.5) Yes454 (73.5)Brain radiotherapy No531 (85.9) Yes83 (13.4) NA4 (0.6)Number of prior systematic therapy lines 061 (9.9) 1179 (29.0) 2132 (21.4) 3114 (18.4) >3124 (20.1) NA8 (1.3)Data are shown as n (%), mean ± standard deviation, or median (interquartile range).ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; LM, leptomeningeal metastases; NA, not applicable; TKI, tyrosine kinase inhibitor; WHO, World Health Organization.aTime from diagnosis of lung cancer to LM. Table 1Basic characteristics of the participantsVariablesN = 618Sex Female377 (61.0) Male241 (39.0)Age (years)57.7 ± 9.6Age group (years) <65472 (76.4) ≥65146 (23.6)Smoking No498 (80.6) Yes120 (19.4)Time of LMa (months)25.1 (0-176.8)EGFR mutation type 19del221 (35.8) 20ins24 (3.9) L858R319 (51.6) Uncommon mutations54 (8.7)Brain metastases No200 (32.4) Yes418 (67.6)Type of LM Type I478 (77.3) Type II140 (22.7)Intracranial pressure before treatment (mmH2O) ≤180111 (18.0) >180184 (29.8) NA323 (52.3)WHO/ECOG performance status 0-2187 (30.3) 3-4325 (52.6) NA106 (17.2)Prior EGFR-TKI therapy No58 (9.4) Yes560 (90.6)Prior third-generation EGFR-TKI therapy No146 (23.6) Yes472 (76.4)Treatment regimens Furmonertinib monotherapy80 (12.9) Furmonertinib + antiangiogenic therapy54 (8.7) Furmonertinib + chemotherapy178 (28.8) Furmonertinib + antiangiogenic therapy + chemotherapy306 (49.5)Dose of furmonertinib (mg) 8088 (14.2) 160407 (65.9) 240123 (19.9)Intrathecal chemotherapy No164 (26.5) Yes454 (73.5)Brain radiotherapy No531 (85.9) Yes83 (13.4) NA4 (0.6)Number of prior systematic therapy lines 061 (9.9) 1179 (29.0) 2132 (21.4) 3114 (18.4) >3124 (20.1) NA8 (1.3)Data are shown as n (%), mean ± standard deviation, or median (interquartile range).ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; LM, leptomeningeal metastases; NA, not applicable; TKI, tyrosine kinase inhibitor; WHO, World Health Organization.aTime from diagnosis of lung cancer to LM. Basic characteristics of the participants Data are shown as n (%), mean ± standard deviation, or median (interquartile range). ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; LM, leptomeningeal metastases; NA, not applicable; TKI, tyrosine kinase inhibitor; WHO, World Health Organization. Time from diagnosis of lung cancer to LM. Efficacy evaluation and survival analysis At the data cut-off date (28 February 2025), 394 patients (63.8%) had discontinued furmonertinib, and 336 patients (54.4%) had died. The median OS was 14.19 months (95% CI 13.04-16.79), the median LMOS was 17.05 months (95% CI 14.81-19.22), and the median TTF was 11.23 months (95% CI 10.23-12.13; Figure 1).Figure 1(A) Kaplan–Meier curves of OS, (B) LMOS, and (C) TTF in patients with EGFRm NSCLC with LM.EGFRm, epidermal growth factor receptor mutant; LM, leptomeningeal metastases; LMOS, overall survival from leptomeningeal metastases diagnosis; NSCLC: non-small-cell lung cancer; OS, overall survival; TTF, time to treatment failure. (A) Kaplan–Meier curves of OS, (B) LMOS, and (C) TTF in patients with EGFRm NSCLC with LM. EGFRm, epidermal growth factor receptor mutant; LM, leptomeningeal metastases; LMOS, overall survival from leptomeningeal metastases diagnosis; NSCLC: non-small-cell lung cancer; OS, overall survival; TTF, time to treatment failure. We first compared clinical outcomes across the four treatment regimens. The median OS was 11.06 months (95% CI 8.53-14.65) in patients receiving furmonertinib monotherapy (group A), 10.31 months (95% CI 7.68-18.37) in those receiving furmonertinib plus antiangiogenic therapy (group B), 14.19 months (95% CI 12.15-17.81) in those receiving furmonertinib plus chemotherapy (group C), and 17.91 months (95% CI 14.35-21.00) in those receiving furmonertinib combined with both antiangiogenic therapy and chemotherapy (group D). Notably, patients in group D exhibited significantly longer median OS compared with those in group A and the other two combination therapy groups (P < 0.001; Figure 2A).Figure 2OS analyses by treatment regimen and furmonertinib dosage. (A) OS across four treatment regimens. Group A: furmonertinib monotherapy; group B: furmonertinib + antiangiogenic therapy; group C: furmonertinib + chemotherapy; group D: furmonertinib + antiangiogenic therapy + chemotherapy (triple therapy). (B) OS by furmonertinib dosage (80, 160, and 240 mg/day).CI, confidence interval; HR, hazard ratio; OS, overall survival. OS analyses by treatment regimen and furmonertinib dosage. (A) OS across four treatment regimens. Group A: furmonertinib monotherapy; group B: furmonertinib + antiangiogenic therapy; group C: furmonertinib + chemotherapy; group D: furmonertinib + antiangiogenic therapy + chemotherapy (triple therapy). (B) OS by furmonertinib dosage (80, 160, and 240 mg/day). CI, confidence interval; HR, hazard ratio; OS, overall survival. To assess the potential influence of performance status on treatment outcomes, patients were stratified by ECOG score. Among those with ECOG 0-2, the median OS was 10.37 months (95% CI 7.64-not applicable [NA]) in group A, 22.94 months (95% CI 8.43-NA) in group B, 22.02 months (95% CI 15.37-28.10) in group C, and 15.34 months (95% CI 12.51-27.87) in group D (P = 0.091; Supplementary Figure S1A, available at https://doi.org/10.1016/j.esmoop.2026.107743). Among patients with ECOG 3-4, median OS was 10.80 months (95% CI 6.89-NA) in group A, 9.02 months (95% CI 6.23-13.66) in group B, 12.81 months (95% CI 7.71-16.79) in group C, and 17.31 months (95% CI 13.46-22.02) in group D (P = 0.003; Supplementary Figure S1B, available at https://doi.org/10.1016/j.esmoop.2026.107743). Analyses carried out after imputation of missing ECOG score data showed similar trends to those before imputation (Supplementary Figure S1C and D, available at https://doi.org/10.1016/j.esmoop.2026.107743). We further carried out IPW analysis, which confirmed that triplet therapy was associated with longer OS compared with other regimens (P < 0.001; Supplementary Figure S2, available at https://doi.org/10.1016/j.esmoop.2026.107743). We further evaluated the association between furmonertinib dosage and survival. The median OS was 16.79 months (95% CI 12.05-25.50) in the 80 mg/day group, 14.68 months (95% CI 13.43-17.91) in the 160 mg/day group, and 11.39 months (95% CI 8.56-14.19) in the 240 mg/day group (P = 0.002; Figure 2B). Subgroup analysis by ECOG score revealed no statistically significant survival differences across dosage groups among patients with ECOG 0-2. Among patients with ECOG 3-4, the 80 mg/day and 160 mg/day groups showed superior OS compared with the 240 mg/day group (Supplementary Figure S3A and B, available at https://doi.org/10.1016/j.esmoop.2026.107743). The results of the analysis after imputation of the ECOG scores showed a relatively high consistency compared with the results before imputation (Supplementary Figure S3C and D, available at https://doi.org/10.1016/j.esmoop.2026.107743). Stratification by intrathecal chemotherapy administration revealed a median OS of 11.56 months (95% CI 10.31-14.02) in untreated patients and 16.10 months (95% CI 13.93-19.16) in treated patients (HR 0.67, 95% CI 0.53-0.85, P = 0.001; Figure 3A). Assessment of prior systemic therapy lines indicated median OS of 24.35 months (95% CI 19.81-NA) in treatment-naïve patients, 16.92 months (95% CI 12.77-20.01) after one prior line, 14.35 months (95% CI 12.81-22.02) after two lines, 13.27 months (95% CI 10.90-17.18) after three lines, and 10.47 months (95% CI 8.46-13.04) after more than three lines (P = 0.001; Figure 3B). Prior exposure to third-generation EGFR-TKIs was associated with a median OS of 19.06 months (95% CI 14.32-26.26) in unexposed patients and 13.46 months (95% CI 12.77-15.37) in exposed patients (HR 1.42, 95% CI 1.09-1.85, P = 0.009; Figure 3C). Among patients with prior exposure to third-generation EGFR-TKIs, 52 (11.0%), 42 (8.9%), 141 (29.9%), and 237 (50.2%) received furmonertinib monotherapy, furmonertinib plus antiangiogenic therapy, furmonertinib plus chemotherapy, and triplet therapy, respectively, compared with 28 (19.2%), 12 (8.2%), 37 (25.3%), and 69 (47.3%) among those without prior exposure (Supplementary Table S6, available at https://doi.org/10.1016/j.esmoop.2026.107743). Subgroup analysis of OS across different treatment regimens in patients with prior exposure to third-generation EGFR-TKIs showed a trend consistent with that of the overall population (Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.107743). Receipt of brain radiotherapy during treatment corresponded to median OS of 13.56 months (95% CI 12.77-16.10) in nonrecipients and 22.02 months (95% CI 14.09-29.62) in recipients (HR 0.63, 95% CI 0.44-0.91, P = 0.012; Figure 3D). The median OS was 13.04 months (95% CI 12.21-14.65) in patients with brain metastases compared with 19.12 months (95% CI 14.58-22.68) in those without (HR 1.43, 95% CI 1.14-1.81, P = 0.002; Figure 3E). The other insignificant results are presented in Figure 3F-H and Supplementary Figure S5A and B, available at https://doi.org/10.1016/j.esmoop.2026.107743.Figure 3Subgroup analysis of OS. Subgroups include: (A) intrathecal chemotherapy, (B) number of prior lines of therapy, (C) prior third-generation EGFR-TKI therapy, (D) brain radiotherapy, (E) brain metastases, (F) ECOG performance status, (G) EGFR mutation subtype, and (H) type of LM. HRs with 95% CIs are shown. P values are from log-rank tests.CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; HR, hazard ratios; LM, leptomeningeal metastases OS, overall survival; TKI, tyrosine kinase inhibitor. Subgroup analysis of OS. Subgroups include: (A) intrathecal chemotherapy, (B) number of prior lines of therapy, (C) prior third-generation EGFR-TKI therapy, (D) brain radiotherapy, (E) brain metastases, (F) ECOG performance status, (G) EGFR mutation subtype, and (H) type of LM. HRs with 95% CIs are shown. P values are from log-rank tests. CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; HR, hazard ratios; LM, leptomeningeal metastases OS, overall survival; TKI, tyrosine kinase inhibitor. Univariate and multivariate Cox regression analyses were carried out to identify baseline factors associated with OS (Table 2). Univariate analysis indicated that prior EGFR-TKI therapy, prior third-generation EGFR-TKI therapy, treatment regimen, furmonertinib dosage, intrathecal chemotherapy, brain metastases, and number of prior systemic therapy lines were significantly associated with OS. Multivariate analysis confirmed treatment regimen, furmonertinib dosage, brain metastases and number of prior systemic therapy lines were independent prognostic factors for OS.Table 2Univariate and multivariate Cox regression analysis of OSCharacteristicUnivariate analysisMultivariate analysisHR95% CIPHR95% CIPSex Male1--- Female0.92(0.74-1.14)0.447--Age (years)1.01(1.00-1.02)0.266--Smoking No1--- Yes1.14(0.86-1.50)0.357--EGFR mutation type 19del1--- 20ins1.51(0.92-2.48)0.104-- L858R1.18(0.93-1.49)0.175-- Uncommon mutations1.30(0.87-1.92)0.198--Prior EGFR-TKI therapy No1-1- Yes1.56(1.04-2.35)0.0320.97(0.51-1.85)0.931Prior third-generation EGFR-TKI therapy No1-1- Yes1.42(1.09-1.85)0.0091.43(0.98-2.08)0.065Treatment regimen Furmonertinib monotherapy1-1- Furmonertinib + antiangiogenic therapy0.96(0.63-1.44)0.8260.97(0.60-1.55)0.891 Furmonertinib + chemotherapy0.74(0.54-1.03)0.0750.87(0.57-1.34)0.526 Furmonertinib + antiangiogenic therapy + chemotherapy0.56(0.41-0.76)<0.0010.56(0.38-0.83)0.004Dose of furmonertinib (mg) 801-1- 1601.02(0.75-1.39)0.9121.06(0.72-1.58)0.760 2401.62(1.13-2.33)0.0091.60(1.02-2.50)0.040Intrathecal chemotherapy No1--- Yes0.67(0.53-0.85)<0.001--Brain radiotherapy No1-1- Yes0.63(0.44-0.91)0.0130.73(0.48-1.11)0.141Brain metastases No1-1- Yes1.43(1.14-1.81)0.0021.45(1.10-1.91)0.009Type of LM Type I1--- Type II0.88(0.67-1.16)0.363--Intracranial pressure before treatment (mmH2O) ≤1801--- >1800.86(0.64-1.16)0.320--WHO/ECOG performance status 0-21-1- 3-41.26(0.97-1.64)0.0831.30(0.99-1.72)0.059Number of prior systematic therapy lines 01-1- 11.67(1.05-2.66)0.0291.27(0.70-2.32)0.428 21.65(1.03-2.65)0.0381.50(0.82-2.74)0.186 32.23(1.39-3.58)<0.0011.91(1.03-3.54)0.040 >32.37(1.50-3.77)<0.0012.09(1.13-3.87)0.019The bolded values indicate P-values that are statistically significant (less than 0.05).CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; HR, hazard ratio; LM, leptomeningeal metastases; TKI, tyrosine kinase inhibitor; WHO, World Health Organization. Table 2Univariate and multivariate Cox regression analysis of OSCharacteristicUnivariate analysisMultivariate analysisHR95% CIPHR95% CIPSex Male1--- Female0.92(0.74-1.14)0.447--Age (years)1.01(1.00-1.02)0.266--Smoking No1--- Yes1.14(0.86-1.50)0.357--EGFR mutation type 19del1--- 20ins1.51(0.92-2.48)0.104-- L858R1.18(0.93-1.49)0.175-- Uncommon mutations1.30(0.87-1.92)0.198--Prior EGFR-TKI therapy No1-1- Yes1.56(1.04-2.35)0.0320.97(0.51-1.85)0.931Prior third-generation EGFR-TKI therapy No1-1- Yes1.42(1.09-1.85)0.0091.43(0.98-2.08)0.065Treatment regimen Furmonertinib monotherapy1-1- Furmonertinib + antiangiogenic therapy0.96(0.63-1.44)0.8260.97(0.60-1.55)0.891 Furmonertinib + chemotherapy0.74(0.54-1.03)0.0750.87(0.57-1.34)0.526 Furmonertinib + antiangiogenic therapy + chemotherapy0.56(0.41-0.76)<0.0010.56(0.38-0.83)0.004Dose of furmonertinib (mg) 801-1- 1601.02(0.75-1.39)0.9121.06(0.72-1.58)0.760 2401.62(1.13-2.33)0.0091.60(1.02-2.50)0.040Intrathecal chemotherapy No1--- Yes0.67(0.53-0.85)<0.001--Brain radiotherapy No1-1- Yes0.63(0.44-0.91)0.0130.73(0.48-1.11)0.141Brain metastases No1-1- Yes1.43(1.14-1.81)0.0021.45(1.10-1.91)0.009Type of LM Type I1--- Type II0.88(0.67-1.16)0.363--Intracranial pressure before treatment (mmH2O) ≤1801--- >1800.86(0.64-1.16)0.320--WHO/ECOG performance status 0-21-1- 3-41.26(0.97-1.64)0.0831.30(0.99-1.72)0.059Number of prior systematic therapy lines 01-1- 11.67(1.05-2.66)0.0291.27(0.70-2.32)0.428 21.65(1.03-2.65)0.0381.50(0.82-2.74)0.186 32.23(1.39-3.58)<0.0011.91(1.03-3.54)0.040 >32.37(1.50-3.77)<0.0012.09(1.13-3.87)0.019The bolded values indicate P-values that are statistically significant (less than 0.05).CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; HR, hazard ratio; LM, leptomeningeal metastases; TKI, tyrosine kinase inhibitor; WHO, World Health Organization. Univariate and multivariate Cox regression analysis of OS The bolded values indicate P-values that are statistically significant (less than 0.05). CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; HR, hazard ratio; LM, leptomeningeal metastases; TKI, tyrosine kinase inhibitor; WHO, World Health Organization. Clinical response was evaluated in 484 patients, of whom 323 (66.7%) achieved a clinical response (Supplementary Figure S6A, available at https://doi.org/10.1016/j.esmoop.2026.107743). In the analysis stratified by baseline factors (Supplementary Table 7, available at https://doi.org/10.1016/j.esmoop.2026.107743), the highest CRR (75.3%) was observed in patients receiving the triple therapy, which was significantly higher than that in the other three groups. Of the 411 patients who underwent brain MRI imaging assessment (Supplementary Table S8, available at https://doi.org/10.1016/j.esmoop.2026.107743), 167 (40.6%) achieved an objective LM response, and the disease control rate (DCR) was 96.6% (Supplementary Figure S6B, available at https://doi.org/10.1016/j.esmoop.2026.107743). The ORR-LM was 39.0% in the 80 mg/day group, 37.6% in the 160 mg/day group, and 54.2% in the 240 mg/day group (P = 0.036). Subgroup analysis of LMOS revealed that the triple therapy maintained longer survival compared with other regimens (P = 0.005; Supplementary Figure S7A, available at https://doi.org/10.1016/j.esmoop.2026.107743), which was further confirmed by IPW analysis (P < 0.001; Supplementary Figure S8, available at https://doi.org/10.1016/j.esmoop.2026.107743). LMOS was longer in patients with type II LM compared with those with type I (HR 0.76, 95% CI 0.58-1.00, P = 0.046; Supplementary Figure S7B, available at https://doi.org/10.1016/j.esmoop.2026.107743). Furthermore, patients who received intrathecal chemotherapy exhibited longer LMOS than those who did not (HR 0.76, 95% CI 0.60-0.95, P = 0.016; Supplementary Figure S7C, available at https://doi.org/10.1016/j.esmoop.2026.107743). The other insignificant results are also presented in Supplementary Figure S7D-H, available at https://doi.org/10.1016/j.esmoop.2026.107743. Univariate Cox regression analysis of LMOS indicated that EGFRm type, prior EGFR-TKI therapy, treatment regimen, intrathecal chemotherapy, LM type, and number of prior systemic therapy lines were significantly associated with LMOS. Multivariate analysis confirmed EGFRm type, prior EGFR-TKI therapy, treatment regimen, and LM type were independent prognostic factors for LMOS (Supplementary Table S9, available at https://doi.org/10.1016/j.esmoop.2026.107743). Safety Safety data are summarized in Supplementary Table S10, available at https://doi.org/10.1016/j.esmoop.2026.107743. TRAEs occurred in 366 patients (59.2%), with grade 3-4 events reported in 49 (7.9%). The most common TRAEs (all grades, incidence >10%) were anemia (25.9%), leukopenia (19.6%), diarrhea (15.5%), decreased platelet count (13.4%), increased γ-glutamyltransferase (13.4%), neutropenia (12.1%), and nausea (10.8%). Combination regimens were associated with a higher incidence of hematological and gastrointestinal toxicity, and higher furmonertinib dosages were also associated with increased rates of leukopenia, neutropenia, elevated creatine kinase, diarrhea, and increased γ-glutamyltransferase (Supplementary Tables S11 and S12, available at https://doi.org/10.1016/j.esmoop.2026.107743). Discussion In this large-scale, multicenter, real-world study, we evaluated furmonertinib-based therapy in 618 patients with EGFRm NSCLC and LM, analyzing outcomes across different treatment regimens and furmonertinib doses. Median OS was 14.19 months, median LMOS was 17.05 months, and the CRR was 66.7%. Among patients with evaluable brain MRI, the ORR-LM was 40.6% and the DCR was 96.6%. TRAEs occurred in 59.2% of patients, with grade 3-4 events observed in 7.9% of patients. These findings provide real-world evidence of furmonertinib activity in LM and suggest that its combination with antiangiogenic therapy and intrathecal chemotherapy may be associated with improved outcomes. Third-generation EGFR-TKIs are standard first-line therapy for advanced EGFRm NSCLC.27,28 However, most patients with LM eventually develop resistance to conventional doses, rendering LM a major clinical challenge.29 Although dose escalation of third-generation EGFR-TKIs has shown preliminary efficacy in LM,15,23 monotherapy alone remains insufficient for many patients. For those with inadequate response to EGFR-TKI monotherapy, combining with antiangiogenic therapy or intrathecal chemotherapy may offer additional benefit.19,20,23,24,30 Previous studies have reported the activity of furmonertinib in EGFRm NSCLC with LM.17,22,31 However, these were limited by their single-center design, lack of control groups, small sample sizes, and heterogeneity in treatment regimens, precluding robust comparisons between monotherapy and combination approaches. Moreover, inconsistent dosing across studies, analogous to the variability observed with osimertinib in the BLOOM and BLOSSOM trials,15,16 has left the optimal furmonertinib dose for LM undefined. The present study was designed to address these knowledge gaps. This study enrolled patients with classic EGFRm (19del and L858R), as well as those with uncommon mutations and 20ins, reflecting the broad activity profile of furmonertinib.22,32 Consistent with their known relative resistance to TKIs, patients with uncommon EGFRm and 20ins showed numerically shorter OS, although this did not reach statistical significance. However, the activity observed in this subgroup supports furmonertinib as a potential therapeutic backbone for these patients with limited central nervous system (CNS)-penetrant options, likely necessitating even earlier integration of chemotherapy and antiangiogenic agents. Moreover, the cohort exhibited poor baseline performance status (52.6% with ECOG score 3-4) and extensive prior third-generation EGFR-TKI exposure (76.4%), reflecting a heavily pretreated, real-world LM population. Despite these adverse features, the median OS was comparable to or exceeded those in similar studies,16,17,23 supporting the clinical activity of furmonertinib-based therapies in this challenging setting. A comparison of key outcomes from the present study with those of pivotal trials of third-generation EGFR-TKIs in LM—including BLOOM, BLOSSOM, and a prior high-dose furmonertinib study—is provided in Supplementary Table S13, available at https://doi.org/10.1016/j.esmoop.2026.107743.15, 16, 17 Subgroup analysis revealed that patients receiving triple therapy had significantly longer OS and LMOS compared with those receiving monotherapy or dual regimens. The consistent benefit across both survival endpoints suggests the observed advantage is not merely a reflection of pre-LM prognostic factors but rather reflects genuine efficacy against leptomeningeal disease. Although prior small-scale studies have reported efficacy with EGFR-TKI combinations,23,24 our study provides the first large-scale, real-world evidence in a heavily pretreated population (76.4% with prior third-generation TKI exposure) suggesting that the addition of antiangiogenic therapy and intrathecal chemotherapy to furmonertinib may confer a survival benefit over continuing TKI monotherapy alone. Notably, a higher proportion of patients with prior third-generation EGFR-TKI exposure received combination therapy, reflecting a clinical tendency toward more intensive combination strategies in this population. The significant survival benefit observed with the triplet regimen in the overall cohort, which was largely comprised such patients, suggests that adding antiangiogenic and intrathecal therapy to a subsequent TKI (furmonertinib) may offer an advantage over TKI monotherapy. This supports a potential shift from sequential single-agent therapy to combination strategies upon LM. Furthermore, our findings provide insights into the role of TKI rechallenge in the LM setting. Despite prior exposure to a third-generation TKI, rechallenge with furmonertinib—particularly in combination with other modalities—resulted in meaningful clinical benefit. This challenges the notion of complete cross-resistance among third-generation TKIs and suggests that continuation of EGFR-TKI therapy in this class remains a viable option, especially when combined with CNS-active therapies. However, the benefit was clearly augmented by combination therapy, indicating that TKI monotherapy is likely insufficient for most patients with LM. Given the often compromised performance status of patients with LM, the tolerability of triple therapy is a relevant clinical consideration. In stratified analysis, the survival benefit of triplet therapy appeared more pronounced in the ECOG 3-4 subgroup. Although counterintuitive, this finding may reflect that poor performance status in these patients is frequently driven by reversible LM-related symptoms, which could be alleviated by intensive multimodal therapy. However, given the limited sample size in this subgroup after stratification and the potential for residual confounding, this observation should be considered exploratory and interpreted with caution. These findings do not support indiscriminate use of intensive therapy in all frail patients, and tolerability requires careful monitoring in clinical practice. Although the standard furmonertinib dose is 80 mg/day, higher doses (160 mg or 240 mg) are frequently used in LM based on evidence supporting dose escalation for intracranial disease control.13,14,17,23,24 In the present study, patients receiving 80 mg/day or 160 mg/day had longer OS compared with those receiving 240 mg/day in the overall population, with this difference primarily driven by the ECOG 3-4 subgroup. This discrepancy between higher radiological response and inferior OS in the 240 mg group highlights a dissociation between short-term tumor shrinkage and long-term survival in LM. Although high-dose furmonertinib may achieve regression of MRI-visible lesions, this does not necessarily translate into a survival benefit, potentially due to cumulative toxicity, poorer tolerability, and higher treatment discontinuation rates associated with high-dose therapy, particularly in frail patients. This underscores the importance of balancing efficacy with tolerability when selecting a dose. Importantly, furmonertinib dose selection was based on clinician judgment rather than random assignment; thus, the inferior OS observed in the high-dose group may reflect underlying differences in patient prognosis rather than a direct effect of the dose itself. These findings should be interpreted as hypothesis-generating and require validation in prospective randomized trials. Additional subgroup and Cox regression analyses identified several prognostic factors. Intrathecal chemotherapy was associated with significantly improved OS and LMOS, consistent with previous studies20,21,33 and reinforcing its role in LM management. The high utilization rate of intrathecal chemotherapy in our cohort (93.8% of those receiving chemotherapy) reflects a distinct practice pattern in China, where intrathecal pemetrexed is widely used based on prospective data.20,23 In contrast, intrathecal chemotherapy is rarely used in the USA and Europe due to toxicity concerns, logistical challenges, and limited historical evidence. Our real-world data, suggesting a significant association between intrathecal chemotherapy and improved survival, may help inform global treatment guidelines. However, given that most chemotherapy in this study was administered intrathecally, we cannot disentangle the survival benefit of the chemotherapy agent itself from that of the intrathecal route. Therefore, our findings primarily support the value of incorporating CNS-directed chemotherapy into the treatment regimen. The number of prior therapy lines was negatively associated with OS; patients who had received three or more lines exhibited significantly shorter OS. Prior third-generation EGFR-TKI exposure and the presence of brain metastases at baseline were associated with inferior survival. Conversely, patients who received combined brain radiotherapy showed superior OS, likely reflecting effective control of concurrent brain parenchymal metastases and suggesting that aggressive local management may complement systemic therapy in this setting. Although no significant difference in OS was observed between patients with type I or type II LM, those with type I LM had significantly shorter LMOS. This aligns with a recent large-scale study establishing CSF positivity as an adverse prognostic marker in LM4 and highlights the value of LMOS as an endpoint that is more sensitive to factors directly related to leptomeningeal disease biology. The divergent results between OS and LMOS for this variable may be explained by the fact that OS captures survival from treatment initiation and may be influenced by pre-LM disease trajectory, whereas LMOS more directly reflects the clinical course after LM diagnosis. Notably, CRR and OS were not consistently correlated across baseline factors such as EGFR subtype, LM type, or ECOG status, suggesting that short-term response does not necessarily translate into long-term survival benefit in these subgroups. In contrast, treatment-related factors—including regimen type and intrathecal chemotherapy—showed concordance between CRR and OS, further supporting the value of combination therapy. Several limitations of this study should be acknowledged. First, its retrospective, nonrandomized design is inherently susceptible to selection bias and confounding. The absence of strict eligibility criteria regarding baseline characteristics such as ECOG performance status and prior lines of therapy resulted in considerable patient heterogeneity, which may influence the interpretation of subgroup analyses. Second, certain baseline variables were incomplete, potentially compromising the accuracy of corresponding subgroup analyses. However, multiple imputation was applied to address missing data, and sensitivity analyses demonstrated high consistency between imputed and complete-case datasets, supporting the robustness of our findings. Additionally, incomplete data on supportive interventions such as corticosteroid dosing and their dynamic effects on intracranial pressure precluded adjustment for these time-varying factors, so whether they impact long-term survival remains unclear. Third, the inclusion criterion requiring at least one month of furmonertinib exposure may have introduced survivorship bias by excluding patients with early death or rapid disease progression, potentially leading to overestimation of median survival. Consequently, our findings may not fully generalize to the broader LM population, particularly those with fulminant disease. Fourth, because combination therapies were not uniformly initiated concomitantly with furmonertinib, immortal time bias cannot be excluded. Finally, all subgroup analyses should be considered exploratory, as no adjustments were made for multiple comparisons; these findings should be interpreted as hypothesis-generating rather than confirmatory. Conclusions In this large-scale multicenter retrospective study of patients with EGFRm NSCLC with LM, furmonertinib-based therapy was associated with meaningful clinical outcomes. The addition of antiangiogenic and chemotherapy to furmonertinib appeared to offer a survival advantage over mono or dual therapy, even in patients with prior third-generation TKI resistance or poor performance status.