First-line toripalimab in combination with platinum-doublet chemotherapy for advanced B3 thymoma and thymic carcinoma: an open-label prospective phase Ⅱ trial

Abstract

Background

Several studies have demonstrated the efficacy of programmed cell death protein 1/programmed death-ligand 1 (PD-L1) inhibitors in patients with thymic epithelial tumors (TETs). Previous trials suggest that the efficacy of immunotherapy in TETs is associated with specific features of the tumor immune microenvironment.

Patients and methods

Patients received a combination of paclitaxel, carboplatin, and toripalimab every 3 weeks for up to six cycles, followed by toripalimab maintenance every 4 weeks until disease progression or other discontinuation criteria. The primary endpoint was progression-free survival (PFS). Immunohistochemistry and next-generation sequencing were used to assess the association between various biomarkers and efficacy.

Results

Between November 2020 and January 2024, 24 patients were enrolled in the study. The median PFS was 19.3 months [95% confidence interval (CI) 12.8-25.8]. The objective response rate was 37.5% (95% CI 18.8%-59.4%) and the disease control rate was 100% (95% CI 85.8%-100%). Grade 3-4 treatment-related adverse events (TRAEs) occurred in 12 (50.0%) patients. Common grade 3-4 TRAEs included decreased neutrophil, white blood cell, and platelet counts. Immune-related adverse events (irAEs) were identified in 11 (45.8%) patients; grade ≥3 irAEs were reported in 1 (4.2%) patient. Patients with higher PD-L1 expression showed a trend toward longer PFS. Further analysis revealed a specific infiltration pattern of CD8+ T cells, M1/M2 macrophages, and CD56^bright natural killer cells in the tumor immune microenvironment and mutations in the nuclear factor κB pathway were associated with better clinical outcomes.

Conclusions

Toripalimab combined with carboplatin and paclitaxel is an emerging therapeutic option for patients with thymic carcinoma and B3 thymoma. The overall incidence and severity of adverse events are within the acceptable range. Tumor microenvironment features or mutational signatures are credible prognostic indicators.

Highlights

• •Toripalimab plus chemotherapy is a treatment option for TETs.
• •The median PFS was 19.3 months, ORR was 37.5%, and DCR was 100%.
• •Patients with higher PD-L1 expression showed a trend toward longer PFS.
• •Toripalimab plus chemotherapy is an emerging therapeutic option for patients with thymic carcinoma and B3 thymoma.
• •Mutations in the NF-κB pathway were associated with better clinical outcomes.

Introduction

Thymic epithelial tumors (TETs) are rare neoplasms that originate from the epithelial component of the thymus gland and are frequently found in the anterior mediastinum. According to the World Health Organization classification of tumors, thymic carcinomas (TCs) and thymomas are classified as subtypes of TETs.¹ Systemic chemotherapy and radiotherapy are crucial modalities for treating recurrent, advanced, or metastatic TETs. Carboplatin-paclitaxel, with or without ramucirumab, is recommended as first-line systemic therapy and is used more frequently for TCs.^2,3 Cisplatin and doxorubicin plus cyclophosphamide are recommended as a first-line option for patients with thymoma.⁴

The efficacy and safety of chemoimmunotherapy in TETs remained uncertain when this study was designed. Programmed cell death protein 1 (PD-1) is primarily expressed on the surface of activated T cells, binds to programmed death-ligand 1 (PD-L1) on tumor cells, and downregulates effector T-cell responses.^5,6 PD-1 and PD-L1 are key components of the brake mechanism involved in immune activation and function. Anti-PD-1/PD-L1 agents may be a treatment option for TETs, given the relatively high PD-L1 expression of 60%-70%.7, 8, 9 The first clinical trial¹⁰ evaluating pembrolizumab monotherapy in a cohort of 40 patients with advanced TCs reported a median progression-free survival (PFS) of 4.2 months, an objective response rate (ORR) of 22.5%, and a 5-year overall survival (OS) rate of 18%. It also demonstrated a prolonged duration of response of ∼3 years.¹⁰ Further retrospective studies with superior efficacy have indicated the therapeutic potential of combining immune checkpoint inhibitors with chemotherapy,^11,12 while prospective evidence is still lacking. Hence, we conducted a prospective clinical trial to investigate the efficacy, safety, and potential prognostic biomarkers of toripalimab in combination with carboplatin and paclitaxel as first-line therapy for patients with advanced B3 thymoma and TC.

Methods

Patients and regimen

This was a single-center, open-label, single-arm, prospective phase II study. Eligible patients were aged >18 years and had an Eastern Cooperative Oncology Group Performance Status (ECOG PS) of 0-1, adequate organ function, and histologically confirmed Masaoka stage III/IV B3 thymoma or TC or progression/recurrence after local treatment, without prior systemic anticancer therapy, major surgery or chemoradiotherapy for at least 4 weeks before enrollment. Patients with autoimmune disorders, concurrent malignancy, human immunodeficiency virus infection, immunodeficiency, active hepatitis B, autoimmune hepatitis, mental disorders, or other poorly controlled background diseases were also excluded. Patients with myasthenia gravis were excluded based on clinical history. The regimen consisted of paclitaxel (175 mg/m², day 1), carboplatin (area under the curve = 5, day 1), and toripalimab, a PD-1 monoclonal antibody (240 mg, day 1), every 3 weeks for up to six cycles, and a toripalimab maintenance dose every 4 weeks until disease progression, intolerable toxicity, withdrawal of consent, death, or other discontinuation criteria were met. Radiologic imaging assessment was carried out every two cycles. Due to concerns of potential myocarditis, myocardial enzyme levels were closely monitored at baseline, before treatment, or at the appearance of symptoms related to cardiac injury.

An electrocardiogram and echocardiography were conducted at baseline and when myocarditis was suspected. Written informed consent was obtained from all patients. This trial was registered as ChiCTR2000039155.

Outcomes

Response was assessed using RECIST version 1.1.¹³ The primary endpoint was PFS. The secondary endpoints included ORR, disease control rate (DCR), OS, duration of response (DoR), and safety. PFS was defined as the time from the first dose to progressive disease (PD) or death from any cause, whichever occurred first. OS was defined as the time from the date of the first dose to death from any cause. DoR was defined as the time from the date of complete response (CR) or partial response (PR) to disease progression. DCR was calculated as the sum of CR, PR, and stable disease (SD) rates. ORR was calculated as the sum of CR and PR rates. Patients with CR, PR, and SD lasting at least 6 months were included in the durable clinical benefit (DCB) group. Patients with PD or PR and SD lasting <6 months were included in the no durable benefit (NDB) group. Patients with best response of CR and PR were defined as responders. Patients with best response of SD and PD were defined as nonresponders. The severity of adverse events (AEs) was assessed using the Common Terminology Criteria for Adverse Events version 5.0. AEs were assessed by two or more independent medical professionals.

Immunohistochemical staining

Immunohistochemical staining was conducted according to the manufacturer’s instructions. PD-L1 expression was evaluated using the Dako PD-L1 IHC 22C3 pharmDx kit (Agilent Technologies, Santa Cl ara, CA) in conjunction with the Dako Autostainer Link 48 system (Agilent Technologies). Multiplexed immunohistochemistry (mIHC) was performed on FFPE tissue sections using the Opal 6-Plex Detection Kit (Akoya, Waltham, MA), enabling the simultaneous detection of multiple biomarkers within a single section through tyramide signal amplification and sequential antibody staining. Cells with positive staining were defined as those with yellowish-brown staining in the cytoplasm or cell membrane. Tissue samples were collected and analyzed from 21 patients. Three patients were not tested for PD-L1 expression due to insufficient tissue sample. Two independent observers determined and discussed the staining results to reach a consensus on contradictory cases. The PD-L1 tumor proportion score (TPS) was determined as the percentage of PD-L1-positive tumor cells with at least partial membrane staining relative to the total number of viable tumor cells, excluding tumor-associated interstitial cells and necrotic, normal, or nonneoplastic cells from the evaluation. Multiplex immunohistochemistry staining of CD8, CD56, CD68, human leukocyte antigen-DR isotype (HLA-DR), and pan-cytokeratin (PANCK) was conducted via Opal 6-Plex Detection Kit (Akoya, Waltham, MA). Cell subsets in the tumor immune microenvironment were identified by mIHC. Tumor-associated macrophages (TAMs) were divided into CD68+/HLA-DR+ M1 cells and CD68+/HLA-DR- M2 cells. Tumor-infiltrating lymphocytes (TILs) were also identified. T cells were divided into CD8+ and other T cells. Natural killer (NK) cells were divided into two subgroups, CD56^dim and CD56^bright, according to differences in CD56 expression density. Density was calculated by dividing the number of positive cells in the tumor parenchyma or stromal area by the tissue area of the tumor parenchyma or stromal area. Positive rate was calculated as the number of positive cells in the tumor parenchyma or stromal area divided by the total number of cells in the tumor parenchyma or stromal area.

DNA sequencing and data analysis

Biopsy sample processing, tissue genomic DNA library preparation, and DNA sequencing were carried out as follows. Briefly, genomic DNA from tumor tissues and peripheral blood leukocytes (as a control) was extracted using the QIAamp DNA FFPE Tissue Kit and the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany), respectively.¹⁴ Sequencing libraries were constructed using the KAPA Hyper Library Prep Kit (KAPA Biosystems, Cape Town, South Africa). Hybridization capture-based targeted enrichment was conducted using the GENESEEQ PRIME™ 425-gene panel (Nanjing Geneseeq Technology Inc., Nanjing, Jiangsu, China). DNA libraries were captured using Dynabeads M-270 (Thermo Fisher Scientific, Waltham, MA) and xGen Lockdown Reagents (hybridization and wash kit; Integrated DNA Technologies, Coralville, IA). The captured libraries were amplified using Illumina p5 (5′-AAT GAT ACG GCG ACC GA-3′) and p7 (5′-CAA GCA GAA GAC GGC ATA CGA GAT-3′) primers and the KAPA HiFi HotStart ReadyMix (KAPA Biosystems), and purified using Agencourt AMPure XP Beads (Beckman Coulter, Brea, CA). Libraries were quantified by quantitative polymerase chain reaction (qPCR) using the KAPA Library Quantification Kit (KAPA Biosystems, Cape Town, South Africa). A Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA) was used to determine the library fragment size. The target-enriched library was then sequenced using the HiSeq 4000 next-generation sequencing (NGS) platform (Illumina, San Diego, CA).

Trimmomatic was used for quality control of FASTQ files by removing low-quality leading/trailing reads with fewer than 20 or N bases.¹⁵ PCR duplicates were removed using Picard (Broad Institute, Cambridge, MA) and local realignment for insertions/deletions (indels) was carried out. GATK3 (Broad Institute, Cambridge, MA) was used for base quality score recalibration. Sequencing data from B3 thymoma and thymic carcinoma tissue samples and the corresponding peripheral blood leukocyte samples were aligned, and sample identity was checked using single-nucleotide polymorphism (SNP) fingerprints with VCF2LR (GeneTalk, Bonn, Germany). MuTect (Broad Institute, Cambridge, MA) was used to detect somatic single nucleotide variants (SNVs) and indels were detected with Scalpel (scalpel-discovery with the somatic mode).¹⁶ The criteria for filtering SNVs and indels were as follows: (i) minimum ≥3 variant-supporting reads; (ii) variants were filtered if they were present at a population allele frequency of >1% in the 1000 Genomes Project or the Exome Aggregation Consortium (ExAC) database; (iii) filtered using an internally collected list of recurrent sequencing errors (≥3 variant reads and ≤20% variant allele frequency in at least 30 of ∼2000 normal samples) on the same sequencing platform.^17,18 The final list of mutations was annotated using vcf2maf (Memorial Sloan Kettering Cancer Center, New York, NY), which uses Ensembl’s Variant Effect Predictor. Copy number variants were detected using FACETS to obtain tumor purity, ploidy, and clonal heterogeneity-adjusted copy number data.^19,20 Fusion events were identified using the DELLY tool to determine the number of chimeric reads (paired-end sequencing reads mapped to different genes) and split reads (spanning a fusion breakpoint) in the targeted DNA sequencing data. Fusions were filtered for split reads. Pathway analyses were carried out using the oncogenic signaling pathways in The Cancer Genome Atlas and ImmPort databases.^21,22

Statistical analysis

At least 21 eligible patients were required for 80% power to demonstrate a benefit in PFS with a one-sided 2.5% type I error rate, assuming a median PFS of at least 8 months for chemotherapy and 13 months for PD-1 plus chemotherapy, considering an accrual time of 24 months and additional follow-up of 12 months. The shape parameter k was set to 1.5 for the Weibull distribution. All statistical analyses were conducted using R (version 4.4.3; R Foundation for Statistical Computing, Vienna, Austria). Two-sided P values < 0.05 were used to determine statistical significance. Exact binomial confidence intervals (CIs) were used to describe response rates. The Wilcoxon rank-sum test was used as a nonparametric statistical test to compare two groups. PFS and OS were estimated using the Kaplan–Meier method. We used the ‘survminer’ package in R to determine the optimal cut-off value for biomarkers.²³

Results

Patient characteristics

Between November 2020 and January 2024, 24 patients were enrolled in this study. At the cut-off date (24 April 2025), 21 (87.5%) patients discontinued study treatment. Among them, 16 (66.7%) discontinued due to disease progression, 3 (12.5%) due to toxicity, and 2 (8.3%) died from causes unrelated to study treatment. Three (12.5%) patients were still receiving maintenance therapy (Figure 1). Seventeen (70.8%) patients had confirmed PD by the data cut-off. A total of four (16.7%) deaths were reported. All 24 patients were included in the analysis set. Patient ages ranged from 35 to 69 years, with a median age of 55 years; seven (29.2%) patients were aged >60 years. Sixteen (66.7 %) patients were male. All patients were Asian. Of the 24 enrolled patients, 22 were diagnosed with thymic squamous carcinoma and 2 (8.3%) were diagnosed with B3 thymoma. The majority of patients (n = 23, 95.8%) were diagnosed with Masaoka stage Ⅳ or American Joint Committee on Cancer Version 9 Cancer Staging System stage Ⅳ. All patients had an ECOG PS of 0-1. Fifteen (62.5%) patients had a tumor size ≥50 mm at baseline. In addition, 15 (62.5%) patients had pleural metastasis, 4 (16.7%) had lung metastasis, 4 (16.7%) had liver metastasis, 10 (41.7%) had distant lymph node metastasis, 1 (4.2%) had diaphragm metastasis, and 1 (4.2%) had thoracic vertebral metastasis throughout the entire course of disease. No enrolled patients developed myasthenia gravis during the treatment period based on their clinical manifestations. Prior treatment in the early stage of disease included thymectomy (n = 10, 41.7%), neoadjuvant chemotherapy (n = 2, 8.3%), adjuvant chemotherapy (n = 3, 12.5%), and thoracic radiotherapy (n = 4, 16.7%). Patients were enrolled after disease relapse and received systemic therapy during the study. Fourteen (58.3%) patients did not receive any treatment before enrollment (Table 1).

Figure 1.

Trial profile. Twenty-four patients were enrolled in this study. At the data cut-off date, 21 patients discontinued study treatment. Among them, 16 discontinued due to disease progression, 3 due to toxicity, and 2 died from causes unrelated to the study treatment. Three patients were still receiving maintenance therapy. All patients were included in the analysis set. (Cut-off date: 24 April 2025).

Table 1.

Patient characteristics

Characteristic	Eligible patients (N = 24), n (%)
Age (years)
≥60	7 (29.2)
<60	17 (70.8)
Age (years), median (range)	53 (34-68)
Sex
Male	16 (66.7)
Female	8 (33.3)
Histology
B3 thymoma	2 (8.3)
Squamous carcinoma	22 (91.7)
Masaoka stage
III	1 (4.2)
IV A	9 (37.5)
IV B	14 (58.3)
TNM staging (AJCC ninth edition)24
III	1 (4.2)
IV A	7 (29.2)
IV B	16 (66.7)
Smoking history
Yes	8 (33.3)
No	16 (66.7)
ECOG PS
0	9 (37.5)
1	15 (62.5)
Tumor size (mm)
≥50	15 (62.5)
<50	9 (37.5)
Metastasis
Pleura	14 (58.3)
Lung	4 (16.7)
Liver	3 (12.5)
Bone	1 (4.2)
Lymph node	10 (41.7)
Previous treatment
Thymectomy	10 (41.7)
Neoadjuvant chemotherapy	2 (8.3)
Adjuvant chemotherapy	3 (12.5)
Thoracic radiotherapy	4 (16.7)
None	14 (58.3)

A total of four patients had received chemotherapy >6 months before enrollment, while none had received treatment <6 months before enrollment. One patient underwent four cycles of neoadjuvant taxane/platin chemotherapy and experienced recurrence 6 months after surgery. Another patient had received a total of six cycles of neoadjuvant and adjuvant taxane/platin chemotherapy combined with immunotherapy, followed by post-operative radiotherapy. Liver metastasis was observed 7 months after the completion of radiotherapy. Two patients received four cycles of adjuvant chemotherapy. One received taxane/platin chemotherapy plus post-operative radiotherapy and the other received anthracycline-based chemotherapy plus post-operative radiotherapy. The former patient progressed 4 years later and the latter progressed after 3 years.

AJCC, American Joint Committee on Cancer; ECOG PS, Eastern Cooperative Oncology Group Performance Status; TNM, tumor–node–metastasis.

Efficacy

The median follow-up was 22.1 months (range 12.1-52.7; Figure 2B and C). Nine (37.5%) patients achieved the best PR and 15 (62.5%) achieved the best SD response. The median tumor reduction was −18.8% (range 0 to −81.8%; Figure 2A). The median PFS was 19.3 months (95% CI 12.8-25.8; Figure 2D). The ORR was 37.5% (95% CI 18.8% to 59.4%) and DCR was 100% (95% CI 85.8% to 100%). The median DoR was 26.4 months (95% CI 0-57.8). The median OS was not reached (Table 2).

Figure 2.

Efficacy of toripalimab in combination with platinum-doublet chemotherapy for patients with advanced thymic carcinoma and B3 thymoma. (A) Waterfall plot showing the best change from baseline in the sum of the longest dimensions for each of the 24 assessable patients. PD-L1 expression levels are displayed. (B) Swimmer plot showing the durability, events, and best overall response during treatment. (C) Longitudinal graph showing the change in the sum of the longest dimensions for each of the 24 patients. (D) Kaplan–Meier survival curves for combination therapy demonstrating a median PFS of 19.3 months (95% CI 12.8-25.8).

CI, confidence interval; NA, not available; PD-L1, Programmed cell death-ligand 1; PFS, progression-free survival; PR, partial response; SD, stable disease; TPS, tumor proportion score.

Table 2.

Response to treatment

Response	Value
Best response, n (%)
PR	9 (37.5)
SD	15 (62.5)
PD	0
PFS (months), median (95% CI)	19.3 (12.8-25.8)
Overall response rate, % (95% CI)	37.5 (18.8-59.4)
DCR, % (95% CI)	100 (85.8-100)
DoR (months), median (95% CI)	26.4 (0-57.8)

CI, confidence interval; DCR, disease control rate; DoR, duration of response; PD, progressive disease; PFS, progression-free survival; PR, partial response; SD, stable disease.

Safety

The treatment-related adverse events (TRAEs) are summarized in Table 3. TRAEs of any grade were reported in all 24 patients, although most had only grade 1-2 AEs. Grade 3-4 TRAEs occurred in 12 (50%) patients. Common grade 3-4 TRAEs included decreased neutrophil, white blood cell, and platelet counts. One patient discontinued combination therapy due to myocarditis and another due to dyspnea. Both patients are still undergoing follow-up. Immune-related adverse events (irAEs) were identified in 11 (45.8%) patients; grade ≥3 irAEs were reported in 1 (4.2%) patient. Hyperthyroidism (n = 4, 17%) and hypothyroidism (n = 6, 25%) of grade 1-2 were the most common irAEs. Patients developed hyperthyroidism and hypothyroidism continued after medications were stopped. Among the two patients with type B3 thymoma, one had a grade 1 irAE (hypothyroidism). In three (12.5%) separate patients with TC, one (4.2%) had grade 1 myocarditis, one (4.2%) had grade 3 myocarditis, and one (4.2%) had grade 2 pneumonitis. Clinicians subsequently decided to discontinue treatment in these three (12.5%) patients. Patients diagnosed with grade 3 myocarditis and grade 2 pneumonitis discontinued treatment and received steroid treatment in time. The patients recovered well afterward. None of the patients died of toxicity or severe irAEs (Table 3). No significant association was found between the best response and the incidence of irAEs (P = 0.206).

Table 3.

Treatment-related AEs

AEs	All grades	Grade 1	Grade 2	Grade 3	Grade 4
Neutrophil count decrease	15 (63)	1 (4)	4 (17)	5 (21)	5 (21)
Nausea	14 (58)	12 (50)	2 (8)	0	0
WBC count decrease	14 (58)	0	5 (21)	6 (25)	3 (13)
Alanine aminotransferase increased	13 (54)	11 (46)	2 (8)	0	0
Aspartate aminotransferase increased	12 (50)	12 (50)	0	0	0
Fatigue	11 (46)	6 (25)	5 (21)	0	0
Platelet count decrease	9 (38)	4 (17)	1 (4)	3 (13)	1 (4)
Rash	7 (29)	6 (25)	1 (4)	0	0
Hypothyroidism	6 (25)	4 (17)	2 (8)	0	0
Anemia	5 (21)	4 (17)	1 (4)	0	0
Creatinine increased	4 (17)	4 (17)	0	0	0
Growth hormone abnormal	4 (17)	4 (17)	0	0	0
Hyperthyroidism	4 (17)	1 (4)	3 (13)	0	0
Pruritus	4 (17)	1 (4)	3 (13)	0	0
Localized edema	3 (13)	3 (13)	0	0	0
Adrenal insufficiency	2 (8)	2 (8)	0	0	0
Lipase increased	2 (8)	2 (8)	0	0	0
Blood prolactin abnormal	1 (4)	1 (4)	0	0	0
Chest pain	1 (4)	0	1 (4)	0	0
Dyspnea	1 (4)	0	1 (4)	0	0
Myocarditis	2 (8)	1 (4)	0	1 (4)	0
Pneumonitis	1 (4)	0	1 (4)	0	0
Serum amylase increased	1 (4)	1 (4)	0	0	0

Data are presented as n (%).

AE, adverse event; WBC, white blood cell.

PD-L1 expression and tumor microenvironment in patients

PD-L1 expression was assessable in 21 patients and 7 (33%) had a TPS > 1%. Among them, two exhibited high expression (TPS > 50%; Table 1). The ORR correlated positively with increasing TPS levels (Supplementary Figure S1A, available at https://doi.org/10.1016/j.esmoop.2026.108366). No statistically significant differences in baseline characteristics were observed between the DCB and NDB groups. Patients in the DCB group displayed significantly higher PD-L1 expression than those in the NDB group (P = 0.030; Supplementary Figure S1B, available at https://doi.org/10.1016/j.esmoop.2026.108366). Moreover, patients with a TPS ≥1% showed a notable trend toward longer PFS than those with a TPS < 1% (30.7 months compared with 19.2 months, P = 0.185). A similar trend was also observed for the TPS ≥50% cut-off (32.3 months compared with 19.3 months, P = 0.196; Supplementary Figure S1C and D, available at https://doi.org/10.1016/j.esmoop.2026.108366).

The tumor microenvironment was examined in 21 patients using mIHC staining. Stromal M1 macrophage positivity correlated with both favorable treatment response and prognosis, as well as with longer PFS (P = 0.022; Supplementary Figure S2A, available at https://doi.org/10.1016/j.esmoop.2026.108366), demonstrating a significant increase in the responder group (patients with the best CR and PR) compared with the non-responder group (patients with the best SD and PD response) (P = 0.04; Supplementary Figure S2C, available at https://doi.org/10.1016/j.esmoop.2026.108366). Notably, in patients with a TPS > 1%, both the density (P = 0.09) and positivity (P = 0.04) of M1 macrophages in the tumor parenchyma were significantly higher than in those with a TPS < 1% (Supplementary Figure S2D, available at https://doi.org/10.1016/j.esmoop.2026.108366). No such significant difference was observed in the stromal compartment or in other tumor-infiltrating immune cells (Supplementary Figure S2E-G, available at https://doi.org/10.1016/j.esmoop.2026.108366). Immune infiltration in both stromal and parenchymal regions was associated with PFS (Supplementary Figure S2H, available at https://doi.org/10.1016/j.esmoop.2026.108366). Specifically, in the tumor parenchyma, improved PFS was associated with a higher density of CD8+ T cells (P = 0.031; Supplementary Figure S2B, available at https://doi.org/10.1016/j.esmoop.2026.108366), lower densities of M2 macrophages (P = 0.048; Supplementary Figure S2H, available at https://doi.org/10.1016/j.esmoop.2026.108366), and a lower density and positive rate of CD56^bright NK cells (P = 0.002 and P = 0.003, respectively; Supplementary Figure S2H, available at https://doi.org/10.1016/j.esmoop.2026.108366). In the stromal regions, a lower positive rate of CD56^bright NK cells was associated with longer PFS (P = 0.045; Supplementary Figure S2H, available at https://doi.org/10.1016/j.esmoop.2026.108366).

Genomic analysis

We obtained gene sequencing results from baseline tissue samples from 20 patients. The most common altered genes were CDKN2A (n = 8, 40.0%), CDKN2B (n = 5, 25.0%), NFKBIA (n = 8, 40.0%), CYLD (n = 5, 25.0%), and TP53 (n = 5, 25.0%). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis indicated that the nuclear factor NF-κB, TP53, RTK-RAS, and WNT pathways were the most frequently altered cancer-related pathways (Supplementary Figure S3A, available at https://doi.org/10.1016/j.esmoop.2026.108366). Gene interaction analysis revealed that mutations in the CDKN2B and CYLD genes significantly cooccurred (Supplementary Figure S3B, available at https://doi.org/10.1016/j.esmoop.2026.108366). At baseline, the responders group was characterized by a higher frequency of mutations in the WNT and TP53 pathways. Conversely, the non-responders or NDB group more frequently harbored inactive NFKBIA mutations and MCL1 amplifications (Supplementary Figure S3C, available at https://doi.org/10.1016/j.esmoop.2026.108366).

Notably, patients harboring NF-κB pathway-related mutations, including CYLD and NFKBIA mutations, exhibited a significantly lower density (P = 0.005) and positive rate (P = 0.020) of CD56^bright NK cells in the tumor parenchyma compared with wild-type (Supplementary Figure S3D and E, available at https://doi.org/10.1016/j.esmoop.2026.108366). These mutations were further associated with a significantly prolonged PFS (30.7 months compared with 14.2 months, hazard ratio 0.21, P = 0.005; Supplementary Figure S3F, available at https://doi.org/10.1016/j.esmoop.2026.108366) and OS (median not reached, P = 0.033; Supplementary Figure S3G, available at https://doi.org/10.1016/j.esmoop.2026.108366).

Discussion

This trial provides preliminary evidence and supports further evaluation of the PD-1 plus chemotherapy combination in randomized controlled trials. Our primary endpoint of median PFS was met. The manageable irAEs suggest that immune checkpoint inhibitors are feasible for TC. The addition of platinum-doublet chemotherapy may contribute to tumor burden reduction and exposure to tumor neoantigens compared with prior immune monotherapy regimens, thereby synergistically augmenting the normalization of immunity toward tumors based on the classic cancer-immunity cycle.²⁵

Four prior trials evaluating PD-1 inhibitors, pembrolizumab or nivolumab monotherapy, reported a median PFS of 3.8-6.2 months, a median OS of 12.7-24.9 months, an ORR of 0%-22.5%, and a DCR of 63%-78.8%.26, 27, 28, 29 In our study, patients received toripalimab plus carboplatin and paclitaxel. The comparable MARBLE study³⁰ evaluating atezolizumab, a PD-L1 monoclonal antibody, plus carboplatin and paclitaxel reported a median PFS of 9.6 months and an ORR of 56%. However, patients with PD-L1 TPS ≥50% or those with PD-L1 expression ≥10% in tumor-infiltrating immune cells achieved a median PFS of over 17 months.³⁰ In a meta-analysis involving 11 379 patients,³¹ anti-PD-1 antibodies exhibited superior PFS and OS compared with PD-L1 antibodies, while no statistically significant difference was observed in incidence or severity of AEs. This conclusion is supported by two retrospective studies.^11,12 One study demonstrated a similar result: the median PFS of PD-1 antibody plus chemotherapy was 16 months and the median OS was 38 months in 22 patients¹¹; the other study¹² indicated that the combination of immune checkpoint inhibitors with chemotherapy yields a better median PFS of 12.7 months and an ORR of 44.4%. This may be explained by the fact that PD-1 antibodies prevent binding to both PD-L1 and programmed death-ligand 2 (PD-L2), whereas PD-L1 antibodies cannot block the interaction of PD-L2 with PD-1. A pathological study reporting the presence of PD-L2 in the epithelium of TETs may support this hypothesis.³² Further studies should be conducted to test related hypotheses.

Targeted therapy is also a competitive option for B3 thymoma and TC. The RELEVENT trial (NCT03921671) of first-line ramucirumab plus carboplatin and paclitaxel reported a median PFS of 18.1 months, a median OS of 43.8 months, and an ORR of 57.6%.³³ This finding complemented the standard regimen in that study³³ and showcased the therapeutic potential of antiangiogenic therapy for TCs. The role of platinum-doublet chemotherapy plus immunotherapy and antiangiogenic therapy awaits further evaluation, with small studies hinting at its clinical potential.^11,34 The PECATI trial (NCT04710628)³⁵ of pembrolizumab plus lenvatinib demonstrated an ORR of 23.3% and a median PFS of 14.9 months. Our study provides evidence supporting the use of PD-1 antibody plus platinum-doublet chemotherapy in patients with TC and B3 thymoma. Although the overall incidence of irAEs remained within acceptable limits, one case of grade 3 myocarditis and one case of grade 2 pneumonitis were reported, highlighting the importance of early detection and management (such as systematic troponin monitoring in early treatment cycles). In the JUPITER-02 trial (NCT03581786), a similar trial in nasopharyngeal carcinoma (toripalimab plus gemcitabine-cisplatin), the incidence of irAEs and irAEs grade >3 was 21.7% and 1.4%, respectively.³⁶ In a previous study of pembrolizumab monotherapy,²⁹ 15% of patients had irAEs grade >3. The incidence and severity of irAEs tend to be lower with toripalimab than with atezolizumab or pembrolizumab. Possible explanations include progressively optimized irAE management and differences in the intrinsic properties of antibodies. Toripalimab is a fully humanized immunoglobulin G4 (IgG4) monoclonal antibody with low immunogenicity. IgG4 binds weakly to Fcγ receptors and induces low antibody-dependent cell-mediated cytotoxicity and complement-dependent cytotoxicity effects, thereby reducing collateral damage to normal tissues. As this was a single-center study conducted at a specialized oncology hospital, potential biases such as selection bias may exist. Future multicenter studies with larger sample sizes are warranted to provide further clinical evidence.

Both NFKBIA and CYLD genes act as inhibitors of NF-κB pathway activation.37, 38, 39, 40, 41, 42, 43, 44 This study found that patients harboring NFKBIA and CYLD mutations in the NF-κB signaling pathway had a substantial prolonged PFS. Mutations that cause the loss of NFKBIA and CYLD gene function may be associated with diminished suppression of NF-κB, thereby promoting pathway activation. The relationship between NF-κB pathway activation and its impact on the tumor immune microenvironment or prognosis is still being explored. The underlying mechanism may involve NF-κB activation, enhancing T-cell activity and M1 polarization while reducing immune evasion. It potentiates antitumor response and ultimately improves response to immunotherapy.45, 46, 47, 48, 49 PD-L1 expression is significantly higher in TCs and B2/B3 thymomas than in AB/B1 thymomas,⁵⁰ while some researchers contend that PD-L1 expression in TETs signifies the thymic origin of the cells rather than an ongoing immune response, since cells in the normal thymus also express PD-L1. In the post hoc analysis of pembrolizumab trial,²⁹ PFS and OS of pembrolizumab monotherapy were longer in patients with TC and high PD-L1 expression than in those with low or no PD-L1 expression. In addition, five patients with a CYLD mutation had high PD-L1 expression.²⁹ A similar trend is observed in this trial. Supporting research suggests that lower CYLD levels may upregulate interferon-γ (IFN-γ)-mediated activation of the STAT1/IRF1 axis. This induces high PD-L1 expression, thereby promoting an immunosuppressive microenvironment.³⁸ Immune infiltration of TCs by TILs mainly involves fully differentiated CD4+ and CD8+ T lymphocytes.⁵¹ In our study, infiltration of CD8+ T cells in the tumor parenchyma was associated with better PFS; a similar trend was also observed in the tumor stroma. Parenchymal CD56^bright NK cells were associated with better PFS. We also found an association between NF-κB mutation and reduced infiltration of CD56^bright NK cells. CD56^bright NK cells can produce immunomodulatory cytokines, such as IFN-γ and tumor necrosis factor-α, whereas CD56^dim cells are commonly recognized as cytolytic lymphocytes.⁵² A previous study showed that monoclonal antibodies inhibit the transition from CD56^bright to CD56^dim, thereby limiting cytotoxicity.⁵³ A meta-analysis revealed that NK-cell infiltration corresponds to a decreased risk of death.⁵⁴ This correlation may be attributed to the suppression of downstream pathways related to CD56^bright NK-cell recruitment, thereby remodeling the tumor immune microenvironment and positively impacting clinical outcome. Lower densities of parenchymal M2 macrophages and higher stromal M1 macrophage positivity are associated with favorable prognosis. A similar study of various types of neoplasms also found that TAM infiltration density in tumors and the M1/M2 ratio are closely associated with tumor prognosis and response to immunotherapy.⁵⁵ In the current study, higher M1 infiltration positivity in tumor parenchyma was observed in patients with PD-L1 TPS over 1% (Supplementary Figure S2E, available at https://doi.org/10.1016/j.esmoop.2026.108366). M1 macrophages activated by IFN-γ are associated with the secretion of stimulatory cytokines and a better antitumor response. In contrast, M2 macrophages activated by interleukin 4 are related to a suppressed adaptive tumor-specific immune response.55, 56, 57, 58 NF-κB activation also promotes the polarization of M1 macrophages.⁵⁹ The underlying mechanism may be associated with a potent T-cell–mediated immune response in the PD-L1 TPS > 1% group, generating high IFN-γ levels, which subsequently upregulates PD-L1 expression and promotes the recruitment of cytotoxic immune cells and polarization of M1 macrophages.

The scarcity of robust clinical studies can be partly attributed to the rarity of TETs, which may lead to underpowered sample sizes. In future studies, sample size should be increased to further investigate the efficacy and safety of certain approaches. The precise identification of patient subgroups with low immune-related risk through biomarkers and genetic signatures may enable clinicians to reduce the incidence of irAEs and improve the overall welfare of patients with TETs. In addition, multiomics approaches, including genomics and spatial transcriptomics, may be employed to elucidate the impact of immunotherapy on the pathogenesis and progression of TETs, as well as to identify other factors influencing the safety and efficacy of immunotherapy.

Conclusion

This trial suggests that toripalimab in combination with carboplatin and paclitaxel may serve as an emerging therapeutic option for advanced TC and B3 thymoma in patients with favorable baseline conditions. TRAEs primarily include myelosuppression. The overall incidence and severity of AEs are within acceptable ranges. Specific characteristics of the tumor immune microenvironment and genome might aid clinicians in identifying favorable patient subgroups and enhancing clinical benefits.