Double-dose firmonertinib as first-line treatment in patients with locally advanced or metastatic non-small-cell lung cancer harboring EGFR L858R mutation: a prospective, multicenter, phase II study (FIRM)

Abstract

Progression-free survival (PFS) with first-line third-generation epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors remains suboptimal in EGFR L858R-mutated advanced non-small-cell lung cancer (NSCLC), highlighting a need for strategies to delay resistance. This single-arm, phase II study (FIRM; ChiCTR2200060897) evaluates first-line double-dose firmonertinib (160 mg/day) in adults with L858R-mutated locally advanced or metastatic NSCLC. With a median follow-up of 27.5 months in 33 patients, primary endpoint was a median PFS of 21.1 months. Secondary endpoints include an unreached median overall survival, an objective response rate of 75.8%, a disease control rate of 90.9%, and an 18-month PFS rate of 63.1%. Grade ≥3 treatment-emergent adverse events occur in 6.1% of patients. Baseline circulating tumor DNA (ctDNA) variant allele frequency predicts ctDNA clearance at cycle 3 day 1, which correlates with longer PFS. Double-dose firmonertinib shows promising efficacy and tolerability, supporting its preliminary potential as first-line treatment for EGFR L858R-mutated advanced NSCLC.

Progression-free survival with epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors remains suboptimal in EGFR L858R-mutated advanced non-small-cell lung cancer (NSCLC), highlighting the need for strategies to delay resistance. Here this group reports a single-arm, phase II study evaluating first-line double-dose firmonertinib in 33 patients with L858R-mutated locally advanced or metastatic NSCLC.

Introduction

Third-generation epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors (TKIs) monotherapy remains a mainstay first-line therapy for advanced EGFR-positive non-small-cell lung cancer (NSCLC), particularly in patients intolerant of or unwilling to receive combination therapy¹. However, EGFR-TKIs have yet to meet expectations, with median progression-free survival (PFS) of approximately 20 months in patients with EGFR mutations^2–6. Notably, heterogeneity in efficacy has been observed across mutation subtypes—patients with exon 21 L858R mutation have consistently shown worse prognosis than those with exon 19 deletions across multiple studies (13.4 vs. 20.8 months; 14.4 vs. 21.4 months)^3,6. To overcome this therapeutic bottleneck and delay resistance, various combination strategies have been explored. Although combining EGFR-TKI with chemotherapy or bispecific antibody (e.g., amivantamab) has yielded promising survival benefits in the L858R-mutated population compared with monotherapy, widespread application in routine practice is limited by the overlapping toxicities and cumbersome dosing schedules^7–10. Therefore, it is critical to develop optimized regimens that balance efficacy enhancement with manageable toxicity profiles and maintain treatment convenience for EGFR exon 21 L858R-mutated NSCLC.

In recent years, a proportion of patients with resistance to first-line EGFR-TKIs in real-world practice have chosen dose intensification over second-line chemotherapy due to toxicity concerns. Firmonertinib, an oral third-generation EGFR-TKI approved in China, has demonstrated a favorable safety profile at the standard 80 mg dose for first-line treatment of adults with locally advanced or metastatic NSCLC harboring EGFR exon 19 deletions or exon 21 L858R mutations, with no dose-limiting toxicities observed during escalation up to 240 mg^11,12. Building on a favourable safety foundation, double-dose firmonertinib (160 mg) has been utilized in patients resistant to first-line third-generation EGFR-TKIs, showing potential to delay resistance without increased toxicities¹³. In light of these observations, we reviewed earlier preclinical studies and found that L858R mutations exhibit lower sensitivity to third-generation EGFR-TKIs than exon 19 deletions, with IC₅₀ values two or threefold higher, suggesting that higher drug exposure may be needed to achieve comparable therapeutic effects in NSCLC harboring L858R mutations¹⁴. In xenograft models harboring EGFR Ex19del/T790M/L858R mutations, firmonertinib demonstrated marked dose-dependent antitumor activity, providing a mechanistic rationale for dose escalation. Beyond the phase Ib dose-escalation study, other clinical data have also suggested that higher doses of firmonertinib (160 and 240 mg) may improve disease control, including promising central nervous system efficacy in patients with EGFR T790M-mutated NSCLC^15,16. Given these findings, and the patient preference for non-chemotherapy regimens in the first-line setting, investigating firmonertinib dose intensification as first-line treatment for L858R-mutated NSCLC is of potential clinical value.

In this work, we conducted a prospective, phase II study (FIRM) to evaluate the efficacy and safety of first-line double-dose firmonertinib (160 mg) in patients with locally advanced or metastatic NSCLC harboring the EGFR L858R mutation. The FLAURA study showed that early circulating tumor DNA (ctDNA) clearance was associated with outcomes¹⁷; thus, we investigated the correlation between plasma ctDNA dynamics from baseline to progression and the therapeutic efficacy following first-line firmonertinib.

Results

Patient characteristics

Between September 2022 and November 2023, 34 patients were screened, 33 of whom were enrolled and started on firmonertinib treatment, and one patient was excluded from the study (Fig. 1). The overaccrual by 8 patients was due to rapid accrual and the need to enhance the reliability of descriptive estimates in the absence of a formal hypothesis-testing framework.

Fig. 1.

Trial Profile.

The median age was 65 (range, 47–80) years (Table 1) and all patients had a baseline Eastern Cooperative Oncology Group performance status (ECOG PS) of 1. Thirty patients (90.5%) had metastatic disease at study entry, most commonly with lung (57.6%) and bone (54.5%) metastases. Six (18.2%) patients had brain metastases at baseline.

Table 1.

Baseline characteristics

	Total (n = 33)
Age, median (range), years	65 (47–80)
Sex, n (%)
Male	16 (48.5)
Female	17 (51.5)
ECOG performance status, n (%)
1	33 (100)
Baseline disease stage, n (%)
IIIBa	3 (9.1)
IV	30 (90.9)
Smoking status, n (%)
Never	22 (66.7)
Former	8 (24.2)
Current	3 (9.1)
Histologic type, n (%)
Adenocarcinoma	33 (100)
Number of metastatic sites, n (%)
<3	15 (45.5)
≥3	18 (54.4)
Brain metastasis, n (%)	6 (18.2)
TP53 mutation, n (%)
Yes	21 (63.6)
No	12 (36.4)
PD-L1 expression, n (%)
<1%	9 (27.3)
1%–49%	7 (21.2)
Unknown	17 (51.5)

^aOne had mediastinal and hilar lymph node metastases, one had involvement of the mediastinal and left supraclavicular lymph nodes, and one had bilateral supraclavicular, mediastinal, and hilar lymph node metastases.

ECOG Eastern Cooperative Oncology Group, PD-L1 programmed death ligand 1.

All 33 patients received at least one dose of firmonertinib and attended one safety follow-up, thus being included in the efficacy and safety analyses. At the data cut-off (August 4, 2025), 25 (75.8%) patients had discontinued the study, and 8 (24.2%) remained on study treatment.

Efficacy

With a median follow-up of 27.5 (range, 5.5–35.3) months, 22 (66.7%) of 33 patients had progressed or died. The median PFS was 21.1 (95% confidence interval [CI], 17.6–31.6) months (Fig. 2a), with an 18-month PFS rate of 63.1% (95% CI, 48.4%–82.1%). Of the 20 patients with available data who had progressed or died, 10 had locoregional progression, and 10 developed distant metastases. Detailed subsequent treatment information for patients following disease progression (PD) is presented in Table S1.

Fig. 2. Survival outcomes and tumor response.

a Kaplan–Meier curves of progression-free survival (PFS); b Kaplan–Meier curves of overall survival (OS); c Treatment exposure and response duration. d Best percentage changes from baseline in target lesions. Source data are provided as a Source Data file. NR not reached, PFS progression-free survival, OS overall survival, CI confidence interval, PR partial response, SD stable disease, PD disease progresssion, NE not evaluable.

OS data remained immature, with only 7 (21.2%) of 33 patients having died (OS events). Although the median OS was not reached, our analysis estimated 24- and 30-month OS rates at 90.0% (95% CI, 79.9%–100%) and 76.5% (95% CI, 61.5%–95.3%), respectively (Fig. 2b).

Twenty-five patients exhibited a partial response (PR), yielding an objective response rate (ORR) of 75.8% (95% CI, 57.7%–88.9%) (Table 2). Five patients had stable disease (SD), resulting in a disease control rate (DCR) of 90.9% (95% CI, 75.7%–98.1%). For responders, the median duration of response (DoR) was 21.6 (95% CI, 18.3–not reached [NR]) months (Fig. 2c). Tumor shrinkage was seen in 29 patients, with 14 experiencing more than a 50% reduction (Fig. 2d). The median maximum tumor shrinkage was −48.3% (interquartile range [IQR], −56.7% to −33.2%).

Table 2.

Tumor responses

Best of response, n (%)	Total (n = 33)
Complete response	0
Partial response	25 (75.8)
Stable disease	5 (15.2)
Disease progression	2 (6.1)
Not evaluable	1 (3.0)
Objective response rate, % (95% CI)	75.8 (57.7, 88.9)
Disease control rate, % (95%CI)	90.9 (75.7, 98.1)

Among the 6 patients with brain metastases, the median systemic PFS was 27.3 (95% CI, 12.9–NR) months. Systemic objective responses were observed in 5 (83.3%) patients, and disease control was found in all 6 (100%) patients (Fig. S1).

Safety

At the cut-off date, the median exposure duration to firmonertinib was 18.9 (IQR, 9.9–24.3) months. Treatment-emergent adverse events (TEAEs) were reported in 30 of 33 (90.9%) patients. The most commonly reported grade 1–2 TEAEs were lymphopenia (39.4%), creatine phosphokinase increased (39.4%), diarrhea (36.4%), and creatinine increased (33.3%) (Table 3). Grade ≥3 TEAEs occurred in 2 (6.1%) patients (anemia and hyponatremia). During the study, no patients developed interstitial lung disease, and one patient reported grade 1 electrocardiogram QT corrected (ECG QTc) interval prolonged.

Table 3.

Treatment-emergent adverse events (incidences ≥ 10%)

	Total (n = 33)
	Any-grade	Grade 1–2	Grade ≥ 3
Any TEAE	30 (90.9)	28 (84.8)	2 (6.1)
Lymphopenia	13 (39.4)	13 (39.4)	0
Creatine phosphokinase increased	13 (39.4)	13 (39.4)	0
Diarrhea	12 (36.4)	12 (36.4)	0
Creatinine increased	11 (33.3)	11 (33.3)	0
Rash	9 (27.3)	9 (27.3)	0
Pruritus	9 (27.3)	9 (27.3)	0
Blood lactate dehydrogenase increased	9 (27.3)	9 (27.3)	0
Hyperuricemia	8 (24.2)	8 (24.2)	0
Aspartate aminotransferase increased	7 (21.2)	7 (21.2)	0
Alanine aminotransferase increased	6 (18.2)	6 (18.2)	0
Thrombocytopenia	6 (18.2)	6 (18.2)	0
Serum amylase increased	5 (15.2)	5 (15.2)	0
Mucositis oral	5 (15.2)	5 (15.2)	0
Anorexia	5 (15.2)	5 (15.2)	0
Anemia	4 (12.1)	3 (9.1)	1 (3.0)

No serious adverse events occurred. Dose interruptions were reported in 2 patients (6.1%) due to grade 2 anorexia and grade 3 anemia. No TEAEs led to dose reduction, treatment discontinuation, or death.

Exploratory analysis

Twenty-eight patients with available plasma samples were enrolled for ctDNA analysis. In the 28 baseline samples, ctDNA was detected in 26 (92.9%) patients (Fig. 3a). The positive rates decreased to 32.0% (8/25) after 2 cycles of treatment (at cycle 3 day 1 [C3D1]) and increased to 78.6% at progression. Two patients initially negative for ctDNA remained ctDNA-negative at C3D1. Sixteen of the 17 patients who were ctDNA-negative at C3D1 ultimately achieved a confirmed objective response.

Fig. 3. Prognostic value of ctDNA dynamics in the FIRM study.

a ctDNA status at baseline, cycle 3 day 1 (C3D1), and progression; b Changes in mean variant allele frequency (mVAF) from baseline (n = 28) to C3D1 (n = 25) and then to disease progression (PD) (n = 14). Boxplot centers, upper and lower bounds, and whiskers represent median, upper and lower quartiles, and quartiles +/− 1.5 inter-quartile range, respectively; c Two representative cases where ctDNA clearance predicted treatment response earlier than radiographic tests. Both patients with ctDNA clearance and stable disease (SD) at C3D1 achieved partial response (PR) at 24 weeks and 40 weeks, respectively; d Kaplan–Meier curves illustrating progression-free survival (PFS) stratified by ctDNA status at C3D1; e Kaplan–Meier curves illustrating PFS stratified by ctDNA clearance at C3D1. f Prediction of ctDNA clearance in C3D1 according to baseline ctDNA mVAF based on receiver operating characteristic curve (ROC). P values in (b) were determined by Kruskal–Wallis test or Wilcoxon test. The exact P value for the comparison among the three groups was 4.7e-6, and the exact P value between baseline and C3D1 was 4.5e-6. P values in (d, e) were calculated by log-rank test. Optimal cut-off for (f) was calculated using the Youden index. Source data are provided as a Source Data file. ctDNA circulating tumor DNA, C3D1 cycle 3 day 1, SD stable disease, PR partial response, HR hazard ratio, NR not reached, AUC area under the curve.

Change of mean variant allele frequency (mVAF) during treatment is shown in Fig. 3b. At C3D1, the median of ctDNA mVAF significantly decreased from 1.1% at baseline to 0 (P < 0.001) and rebounded to 0.08% at progression. Further analysis stratified by response showed a significantly higher mVAF at C3D1 in non-responders compared to responders (0.75% vs 0%; P = 0.044; Fig. S2a). In patients with available samples both at baseline, C3D1, and progression, ctDNA mVAF was lowest at C3D1 in all 9 responders, remaining lower at progression compared to baseline (Fig. S2b). Two patients with baseline VAFs of 0.07% and 0.31% showed ctDNA clearance (VAF = 0) at C3D1 despite SD on imaging; both ultimately achieved PR at 24 and 40 weeks, respectively, highlighting the potential of ctDNA kinetics as an early biomarker of therapeutic response (Fig. 3c).

We further explored the association between PFS and the status of ctDNA. Patients with positive ctDNA at C3D1 (32%, 8/25) presented a significantly shorter PFS than those with negative ctDNA (10.7 vs. 24.3 months; HR, 3.9 [95% CI, 1.4‒10.6]; P = 0.005; Fig. 3d). Among the patients who were initially ctDNA-positive at baseline, 65% patients (15/23) who achieved ctDNA clearance at C3D1 had significantly improved mPFS compared to those who remained ctDNA-positive (23.4 vs. 10.7 months; HR, 3.7 [95% CI, 1.3‒10.4]; P = 0.008; Fig. 3e). Baseline EGFR VAF and ctDNA mVAF effectively predicted ctDNA clearance, with an area under curve (AUC) of 68.3% (P = 0.146; cut-off = 1.5%; Fig. S3a) and 72.5% (P = 0.042; cut-off = 1.1%; Fig. 3f), respectively. Additionally, baseline EGFR VAF showed moderate correlations with baseline total tumor diameters (rho = 0.426, P = 0.042) and a strong correlation with baseline ctDNA VAF (rho = 0.897, P < 0.001; Fig. S3b).

Furthermore, post-progression biopsies from 8 patients underwent somatic mutational landscape to evaluate the molecular disease at PD. MET amplifications and TET2 frameshift/nonsense variants were each observed in 2 patients. KRAS G12V and PIK3CA mutations were identified in a patient without EGFR L858R at PD (Fig. S4).

Discussion

The FIRM study prospectively evaluated the efficacy and safety of double-dose third-generation EGFR TKI as first-line therapy in patients with advanced NSCLC harboring the EGFR L858R mutation, indicating that double-dose firmonertinib is a feasible and well-tolerated option. Double-dose firmonertinib exhibited promising efficacy for patients with locally advanced or metastatic NSCLC harboring the EGFR L858R mutation, with a median PFS reaching 21.1 months and an 18-month PFS rate of 63.1%. OS was not reached with 24- and 30-month OS rates of 90.0% and 76.5%. Double-dose firmonertinib achieved durable and clinically meaningful disease control. ORR was 75.8%, median DoR was 21.6 months, and DCR was 90.9%, all of which fall within the historical range of standard first-line third-generation EGFR-TKI treatments^3,5,12,18,19. Notably, patients with brain metastases also achieved durable disease control with double-dose firmonertinib, with a median systemic PFS of 27.3 months. Given the added challenge of the L858R mutations that further worsen prognosis, this result was promising in the context of the median systemic PFS reported with routine-dose firmonertinib in EGFR-positive (exon 19 deletions or exon 21 L858R)¹²; however, this exploratory and descriptive finding should not be overinterpreted due to the small subgroup size (n = 6). Furthermore, a ctDNA status observed on C3D1 was associated with improved clinical outcomes for patients treated with the double-dose firmonertinib therapy, providing valuable information for predicting efficacy through continuous ctDNA monitoring.

Landmark trials such as MARIPOSA, FLAURA2, and NEJ009 have reported median PFS values of 20.9–25.5 months in EGFR-positive NSCLC^7,9,20. Nevertheless, subgroup analyses across these studies consistently showed higher hazard ratios for patients with L858R mutations compared with those harboring exon 19 deletions, underscoring a need for more intensive strategies in this population^7,9,20. Against this background, the median PFS of 21.1 months achieved with double-dose firmonertinib in advanced NSCLC with L858R mutation exceeded the 13.9 months reported for osimertinib monotherapy and approached the 24.7 months observed with osimertinib plus chemotherapy²¹. We also observed a numerically higher 24-month OS rate (90.0% vs. 75%–80%) compared to those reported for these combination regimens^22,23. While it remains unclear whether double-dose firmonertinib can provide long-term OS benefit, the systemic efficacy, manageable toxicity, and the availability of subsequent treatment options post-progression observed with double-dose firmonertinib are similar to the key factors that contributed to the FLAURA2 success, suggesting a potentially positive trend for long-term outcomes as follow-up continues¹⁰. However, these findings warrant further evaluation in phase III randomized trials comparing this regimen with current standards of care, including standard-dose firmonertinib and osimertinib plus chemotherapy, to clarify its role in L858R-mutated NSCLC. Additionally, compared with the combination regimens, double-dose firmonertinib showed a lower incidence of dose interruptions (6.1% vs. 43%–83%) and discontinuations (0 vs. 11%–35%)^7,9. This safety advantage could preserve performance status and enable the continuation of subsequent therapies, thereby contributing to survival benefits²⁴. Overall, despite the ongoing transition from monotherapy to combination regimens as the new standard of care, practical constraints such as tolerability concerns and patient preference render some patients unable or unwilling to receive combination therapy, for whom the double-dose firmonertinib regimen—with efficacy approaching that of combination therapy and acceptable safety—represents an additional clinical treatment option.

Emerging evidence supports the concept of dose intensification for the L858R-mutated subgroup. In a recent study, asandeutertinib (TY-9591)―a deuterium-stabilized derivative of osimertinib―administered at 160 mg (twice the standard 80 mg osimertinib dose) achieved a median PFS of 19.3 months in L858R-mutated patients with favorable safety, and has since entered further clinical investigation in this subgroup²⁵. Similarly, the INCREASE study demonstrated that high-dose icotinib (250 mg thrice daily) yielded a longer PFS (12.9 vs. 9.2 months) than the routine-dose (125 mg thrice daily) in L858R-mutated patients²⁶. Together with our findings, these results suggest that patients with L858R mutations may require higher EGFR-TKI doses than those with exon 19 deletions. Although confirmation in large-scale, prospective randomized studies is warranted, optimizing EGFR-TKI dosing by mutation types remains a valuable area for further investigation.

In the exploratory analysis, longitudinal monitoring of ctDNA showed a marked initial reduction in both ctDNA positivity rates and VAF during study treatment, followed by a rebound at progression, which may reflect the early efficacy and disease course. A correlation was observed between ctDNA status and clinical outcomes, as patients who achieved ctDNA-negative status or ctDNA clearance at C3D1 had longer PFS. Together with existing evidence^27–31, these findings imply that plasma ctDNA-negative status or ctDNA clearance at C3D1 might serve as a potential surrogate molecular endpoint for survival benefits. Moreover, while both EGFR VAF and ctDNA mVAF effectively identified patients likely to achieve ctDNA clearance, ctDNA mVAF demonstrated superior predictive performance owing to its ability to reflect comprehensive mutational burden. The molecular disease pattern of patients undergoing FIRM treatment at the time of PD is similar to that of patients treated with regular dose EGFR-TKI^32,33. Two patients may experience acquired resistance due to their TET2 loss-of-function variants, which could result in drug resistance by upregulation of TNF/NF-κB signaling³⁴. Preliminary insights from these findings provide a deeper understanding of disease monitoring and the therapeutic response to double-dose firmonertinib, although these findings require further validation.

There were also several limitations for this study. First, the FIRM study had a relatively small sample size, which weakened the robustness and statistical power of the study. Second, this exploratory single-arm design, without a control group or formal sample size calculation, may affect generalizability of the findings. Third, the survival data are still being followed up to fully assess the long-term benefits of double-dose firmonertinib. Fourth, the findings from the dynamic biomarker exploration require validation in a larger cohort with an additional control arm.

In this phase II study, double-dose firmonertinib showed promising efficacy and a favorable safety profile in locally advanced or metastatic NSCLC patients with EGFR L858R mutations, warranting further investigation in larger-scale phase III randomized trials.

Methods

Study design and patients

The FIRM study was a multicenter, single-arm, phase II study conducted across five hospitals in China. This study was prospectively registered in the Chinese Clinical Trial Registry (ChiCTR2200060897) and has been completed.

Eligible patients were aged 18 years or older with histologically or cytologically confirmed locally advanced or metastatic NSCLC not amenable to operation or other radical treatment. Additional criteria included measurable disease according to the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1), ECOG PS of 0–1, a life expectancy of at least 12 weeks, adequate bone marrow reserve and organ function, and no previous systemic treatment for advanced or metastatic NSCLC (patients who had received local treatment were eligible if the lesions within the treated area were non-target lesions). All patients were required to be confirmed EGFR exon 21 L858R-mutated. Patients with asymptomatic stable central nervous system metastases not requiring steroids for at least 14 days before the first dose of firmonertinib were also included. Exclusion criteria included any prior treatment with EGFR-TKIs; confirmed EGFR exon 20 insertion mutations, exon 19 deletions, exon 20 T790M mutations, or other uncommon mutations (G719X, L816Q, S768I). Full criteria are provided in the Protocol (Supplementary Note 2) in the Supplementary Information.

Procedures

Eligible patients received firmonertinib 160 mg (twice the standard dose of 80 mg) orally once daily (in 21-day cycles) on a continuous dosing schedule, until PD (defined according to RECIST 1.1), death, intolerable toxicities, or withdrawal from the study. Dose interruption was permitted if a patient had a grade 3 or higher adverse event according to the National Cancer Institute-Common Terminology Criteria for Adverse Events (NCI-CTCAE) version 4.03 or unacceptable toxicity. If the adverse event resolved or returned to grade 2 or less within 21 days, firmonertinib could be resumed at the same dose or a lower dose level. The dose could be reduced sequentially from 160 mg to 80 mg and then to 40 mg per day, with discontinuation required thereafter. Patients should be discontinued from the study in the event of interstitial lung disease or non-infectious pneumonia, ECG QTc interval prolonged accompanied by severe arrhythmic symptoms or signs, or symptomatic congestive heart failure. Detailed information is available in the Protocol in the Supplementary Information.

Tumor assessments were performed at baseline, at week 4, and every 6–8 weeks thereafter until PD, using either computerized tomography or magnetic resonance imaging scans of the chest, abdomen, pelvis, and any other suspected areas. Adverse events were monitored and graded according to the NCI-CTCAE version 5.0 throughout the study and for at least 30 days after administration of the last dose of firmonertinib. Available tissue samples were collected at baseline and blood samples were collected at baseline, cycle 3 day 1 of the first study treatment, and at the time of disease progression for exploratory biomarker analysis.

Sequencing of tumor tissue and peripheral blood

All 33 enrolled patients underwent laboratory confirmation of EGFR exon 21 L858R mutation through tumor tissue biopsy analysis using ARMS-PCR, cobas-PCR, or next-generation sequencing (NGS). High-purity genomic DNA (gDNA) was extracted from formalin-fixed paraffin-embedded (FFPE) tumor specimens with ≥20% tumor cellularity. All PCR-based analyses were conducted in accordance with manufacturer protocols, incorporating both positive and negative controls. For NGS-based analyses, library preparation for tumor tissue sequencing was performed using KAPA Library Preparation Kit (Roche, Basel, Switzerland). Deep sequencing was subsequently conducted on the Illumina NextSeq 550 or NovaSeq 6000 platform (SanDiego, CA, USA), achieving a mean coverage depth of 500 ×.

Peripheral blood samples were processed within 2 h post-collection, with centrifugation separating components into plasma and buffy coat fractions. For parallel analysis of ctDNA and germline variants, cell-free DNA (cfDNA) extracted from plasma and gDNA isolated from white blood cells were subjected to library construction using the xGen^TM cfDNA & FFPE DNA Library Preparation Kit (Integrated DNA Technologies, Coralville, IA, USA) according to manufacturer’s protocols. Ultra-deep sequencing was performed on the Illumina NovaSeq 6000 platform using a customized 158-gene panel, generating 2 × 151 bp paired-end reads with an average sequencing depth of 60,000 ×.

Bioinformatics analysis of tumor-derived genomic DNA and ctDNA

A comprehensive bioinformatics pipeline was developed for simultaneous analysis of tumor tissue and ctDNA variants, comprising 4 core modules including sequence read alignment, somatic variant calling, structural variation detection, and copy number alteration analysis. Raw sequencing data underwent quality control through adapter trimming and base quality filtering using fastp v2.20.0. Alignment to the GRCh37/hg19 reference genome was performed with BWA-MEM v0.7.17. Single nucleotide variants and insertion-deletion mutations (InDels) were identified through VarDict v1.5.7 and subsequently annotated for clinical significance using InterVar. Gene fusion events were detected via FACTERA v1.4.4, while CNVs, including focal amplifications and deletions, were quantified using CNVkit v1.1. Longitudinal ctDNA monitoring was achieved through mean variant allele frequency (mVAF, mean frequency of all detected mutations in ctDNA) trajectory analysis across serial timepoints. ctDNA positivity was defined as detection of ≥1 somatic mutation above platform-specific detection thresholds. ctDNA clearance was defined as the conversion from positive ctDNA to negative status.

Endpoints

The primary endpoint was the investigator-assessed PFS, defined as the time from the date of the first study dose until the date of documented PD or all-cause death before PD. The secondary efficacy endpoints included ORR, defined as the proportion of patients who showed CR or PR; DCR, defined as the proportion of patients with CR, PR, or SD; 18-month PFS rate; OS, defined as the time from the date of first study dose to death from any cause; and safety. The exploratory endpoint included changes in the peripheral blood ctDNA gene profile.

Statistics and reproducibility

The experiments were not randomized. No statistical method was used to predetermine sample size, and the sample size was set to 25 patients based on its exploratory nature.

The full analysis set included all patients who received at least one dose of the study therapy and was used for efficacy assessments. Adverse events were assessed in the safety analysis set, consisting of all patients with safety data who had received at least one dose of received treatment. No data were excluded from the analyses.

The demographic and clinical characteristics, safety outcomes, and tumor responses were summarized using means, standard deviations, medians, ranges, or IQRs for continuous variables, and frequencies and percentages for categorical variables. The median PFS, median OS, and DoR, along with their corresponding two-sided 95% CIs, were estimated using the Kaplan–Meier method. For OS, patients who were alive or lost to follow-up were censored at the time of last contact. For both PFS and DoR, patients who discontinued the study or initiated new anticancer therapy without radiographic evidence of PD were censored at the respective time of discontinuation or initiation of new anticancer therapy. The ORR and DCR were calculated based on the confirmed best overall response of tumors during the study, and the corresponding two-sided 95% CIs were determined using the Clopper-Pearson method. A post-hoc subgroup analysis of PFS was done based on the subgroups according to baseline characteristics. All statistical analyses were performed using R Studio version 2024.04.2.

Ethics statement

The study protocol was reviewed and approved by the Ethics Committee of Jiangsu Cancer Hospital (approval no.: 2022-033). Written informed consent was obtained from all patients before enrolment. Investigators obtained informed consent from each participant or each participant’s guardian. This study was conducted in accordance with the provisions of the Declaration of Helsinki, Good Clinical Practice guidelines, and applicable regulatory requirements. An insurance policy was in place to compensate patients for any injury related to the study drugs or other trial procedures. No compensation was provided to participants. Both male and female patients were eligible for enrollment, and sex information was collected through self-reporting. All collected data are reported in a sex-disaggregated manner.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conception and design: M.S.; Provision of study materials or patients: B.S., C.W., L.Z. (Liqin Zhang), Y.Z. (Yingying Zhu), X.Z., X.W. (Xiaoxuan Wang), Z.G., L.W., X.W. (Xiaohua Wang), L.Z. (Liqun Zhu), Y.Z. (Yun Zhou), D.L., M.S.; Collection and assembly of data: B.S., C.W., L.Z. (Liqin Zhang), Y.Z. (Yingying Zhu), X.Z., X.W. (Xiaoxuan Wang), Z.G., L.W., X.W. (Xiaohua Wang), L.Z. (Liqun Zhu), Y.Z. (Yun Zhou), D.L., M.S.; Data analysis and interpretation: M.S.; Manuscript writing: B.S., C.W., L.Z. (Liqin Zhang), Y.Z. (Yingying Zhu), X.Z., X.W. (Xiaoxuan Wang), Z.G., L.W., X.W. (Xiaohua Wang), L.Z. (Liqun Zhu), Y.Z. (Yun Zhou), D.L., M.S.; Final approval of manuscript: B.S., C.W., L.Z. (Liqin Zhang), Y.Z. (Yingying Zhu), X.Z., X.W. (Xiaoxuan Wang), Z.G., L.W., X.W. (Xiaohua Wang), L.Z. (Liqun Zhu), Y.Z. (Yun Zhou), D.L., M.S.; Accountable for all aspects of the work: B.S., C.W., L.Z. (Liqin Zhang), Y.Z. (Yingying Zhu), X.Z., X.W. (Xiaoxuan Wang), Z.G., L.W., X.W. (Xiaohua Wang), L.Z. (Liqun Zhu), Y.Z. (Yun Zhou), D.L., M.S.

Peer review

Peer review information

Nature Communications thanks Li-Ching Huang and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The protocol of the study is available in the Supplementary Information. Source Data are provided with this paper. All data generated in this study are included in the Supplementary Information and Source Data files. The raw sequence data generated in this study have been deposited in the Genome Sequence Archive in the National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (https://ngdc.cncb.ac.cn/gsa-human) under accession code HRA014922. The raw sequence data are available under restricted access for Chinese regulations on human genetic resources, access can be obtained by sending reasonable requests to the corresponding author, which will need the approval of the institutional ethical committees. Clinical data were not publicly available due to involving patient privacy, but can be accessed from the corresponding author (shimeiqi1963@163.com), upon request for 3 years; individual de-identified patient data will be shared for clinical study analyses. The remaining data are available in the manuscript, supplemental information, or source data file. Request for access to raw sequencing data for non-commercial research purpose will be considered and provided within three months, after obtaining approval by the sponsor and relevant ethics committees and the signing of a data use agreement. Source data is provided with this file. Source data are provided with this paper.

Competing interests

X.Z. and X.W. (Xiaoxuan Wang) were employed by Jiangsu Simcere Diagnostics Co., Ltd. All other authors declare no conflict of interest.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68554-6.