The 27th World Conference on Lung Cancer (WCLC) will be held in Seoul, Korea, from September 12 to 15, 2026, hosted by the International Association for the Study of Lung Cancer (IASLC). As one of the major international meetings in thoracic oncology, the congress will gather leading researchers from around the world.
According to the report, eight studies led by Prof. Yilong Wu from Guangdong Provincial People’s Hospital have been accepted to the meeting, including 4 oral and 4 mini-oral presentations. Topics cover radiomics-based prediction, bispecific antibody neoadjuvant therapy, immune microenvironment and biomarkers, and surgical consolidation strategies.
MedJ has compiled the studies for readers’ reference.
Oral Presentations
Abstract OA08.04
Title: Non-Invasive Prediction of Spread Through Air Spaces in Stage I Lung Adenocarcinoma Using CT-based Virtual PET (CPT-STAS): A Multicenter Study
Presenter: Prof. Z. Huang
Overview:
This study aimed to develop and validate a multimodal radiomics model, CPT-STAS, which integrates conventional computed tomography (CT) with generative adversarial network (GAN)-synthesized positron emission tomography (PET), for non-invasive prediction of spread through air spaces (STAS) in early-stage lung adenocarcinoma (LUAD), and to explore its biological basis. The study retrospectively included 1,066 patients with early invasive LUAD from three centers. A Pix2pixHD GAN was used to synthesize PET images from CT, and a support vector machine (SVM) was used to integrate radiomic features from CT and synthesized metabolic images. Model performance was evaluated by receiver operating characteristic (ROC) curves, calibration curves, and decision curves, and was further validated by subgroup analysis, recurrence-free survival (RFS) outcomes, and radiogenomics.
The synthesized PET images showed high consistency with real PET parameters. The CPT-STAS model achieved an internal validation AUC of 0.845, comparable to a real multimodal model (DeLong test P=0.707). In the 0.5≤CTR<1 subgroup, the AUC reached 0.977, and the model consistently outperformed unimodal models on external testing. Higher model scores were significantly associated with worse RFS (P<0.01). Radiogenomic analysis indicated that the model captured imaging heterogeneity associated with invasive phenotypes and metabolic reprogramming.
The findings suggest that CPT-STAS may provide a robust and cost-effective tool for pre-operative STAS prediction in LUAD. By integrating synthesized metabolic data, it may support individualized surgical decisions and precise risk stratification.
Abstract OA09.03
Title: A Phase 2 Study of JS207, a PD-1/VEGF Bispecific Antibody, Plus Chemotherapy in Patients With Unresectable Stage III NSCLC
Presenter: Prof. Yilong Wu
Overview:
This study is a multicenter, open-label phase II trial evaluating the efficacy and safety of programmed cell death protein 1 (PD-1)/vascular endothelial growth factor (VEGF) bispecific antibody JS207 plus platinum-based doublet chemotherapy in patients with unresectable Stage III NSCLC. Standard treatment for this population is definitive concurrent chemoradiotherapy (cCRT) followed by programmed death-ligand 1 (PD-L1) inhibitor consolidation, but survival outcomes remain limited. After neoadjuvant therapy, subsequent treatment was determined by multidisciplinary team (MDT) review. The neoadjuvant phase included a safety lead-in of at least 21 days to assess toxicity and determine the dose of JS207 (initial 10 mg/kg, with 15 mg/kg as an alternative) plus chemotherapy. The primary endpoint was pathological complete response (pCR) rate.
From May 23 to December 3, 2025, 42 patients were enrolled and received JS207 10 mg/kg intravenously every 3 weeks (Q3W); 92.9% were male, 61.9% had squamous carcinoma and 33.3% had non-squamous carcinoma, and 35.7% were stage IIIA, 42.9% stage IIIB, and 21.4% stage IIIC. As of March 15, 2026, 52.4% (22/42) underwent surgery after JS207 plus chemotherapy neoadjuvant treatment, with a pCR rate of 54.5% (12/22) and a major pathological response (MPR) rate of 72.7% (16/22).
In terms of safety, as of February 28, 2026, treatment-emergent adverse events (TEAEs) occurred in 92.9% of patients, with serious adverse events in 19.0%. JS207-related grade ≥3 TEAEs occurred in 45.2% of patients. Common TEAEs were anemia (66.7%), neutropenia (61.9%), and leukopenia (57.1%).
The findings suggest that JS207 plus chemotherapy showed encouraging surgical conversion and pCR rates with an acceptable safety profile in unresectable NSCLC, providing valuable clinical data for further confirmatory studies.
Abstract OA11.03
Title: Predicting NSCLC Neoadjuvant Immunotherapy Response Using H&E-Derived T-Cell States via Deep Learning
Presenter: Prof. Dongshen Li
Overview:
This study developed a deep learning workflow based on a multiple-instance learning (MIL) algorithm using routine hematoxylin and eosin (H&E) slides to predict neoadjuvant immunotherapy response in NSCLC. Neoadjuvant immunotherapy has significantly improved outcomes for patients with resectable NSCLC, but treatment response is highly heterogeneous and robust predictive biomarkers are lacking. T-cell functional phenotypes in the tumor microenvironment, such as exhaustion status, are key determinants of immune checkpoint inhibitor (ICI) efficacy. RNA sequencing (RNA-seq) can accurately characterize T-cell state scores (TCSS), but cost and quality-control complexity limit clinical application.
The study collected whole-slide images (WSIs) of H&E from treatment-naive NSCLC patients and used paired RNA-seq data to calculate eight T-cell functional state scores—cytotoxic, helper, regulatory, senescent, resting, proliferative, progenitor-exhausted, and terminally exhausted—as gold-standard labels. A regression model was built based on the CLAM architecture to directly predict pathology-derived T-cell state scores (pTCSS) from H&E image features, and a model for predicting pathological response to neoadjuvant immunotherapy was constructed.
Preliminary results showed that the model effectively learned and extracted spatial pathological features related to T-cell states from routine H&E slides. In the internal training set, model-predicted pTCSS showed good consistency with RNA-seq-derived TCSS, with key T-cell state correlations of approximately r=0.5 (proliferative r=0.570, cytotoxic r=0.570, senescent r=0.541, helper r=0.543).
The findings suggest that deep learning models based on the CLAM architecture can predict tumor microenvironment T-cell state scores from routine diagnostic H&E biopsy specimens, and may provide a promising tool for individualized response prediction in NSCLC neoadjuvant immunotherapy, with potential applications in pre-operative risk stratification and treatment decision-making.
Abstract OA13.01
Title: Intracranial Responders in Cerebrospinal Fluid and Peripheral-Intracranial Immune Linkage During ICI Therapy in NSCLC Brain Metastases
Presenter: Prof. Yangsi Li
Overview:
This study enrolled patients with NSCLC brain metastases (BM) planned to receive ICI-based therapy. Cerebrospinal fluid (CSF) and peripheral blood mononuclear cells (PBMCs) were collected longitudinally at baseline and during treatment for single-cell RNA sequencing (scRNA-seq, n=20) and T-cell receptor (TCR) sequencing (n=15). Paired BM tumor tissue was included when available and integrated with external datasets (total BM tumor tissue n=20), and three independent patient cohorts (n=64) were established for cross-cohort consistency evaluation.
Through high-dimensional single-cell analysis, the study mapped the cellular dynamics of CSF, PBMC, and BM tumor microenvironments in NSCLC and BM throughout ICI therapy. Within heterogeneous immune cell populations, the study identified a CSF CD4+PDCD1+CXCR6+ T-cell subset positively correlated with intracranial ICI response, suggesting its potential as a minimally invasive predictive biomarker. Transcriptomic and functional analyses showed that these cells exhibited significant similarity between CSF and BM tumor immune microenvironments, suggesting they may mediate local anti-tumor immunity.
TCR repertoire analysis showed significant clonotype overlap between paired CSF and PBMC samples, including expanded shared clones, suggesting that peripheral T cells may participate in intracranial immune responses during ICI therapy. The findings suggest that the CSF CD4+PDCD1+CXCR6+ T-cell subset may serve as a potential biomarker for predicting intracranial ICI response in patients with NSCLC BM. The significant TCR clonotype sharing between CSF and PBMC supports a mechanistic model in which peripherally activated T cells contribute to BM microenvironment immune responses.
Mini-Oral Presentations
Abstract MO06.04
Title: Neoadjuvant Chemo-Immunotherapy vs EGFR-TKI in EGFR-Mutant NSCLC: TP53 as the Predictive Biomarker for Treatment Selection
Presenter: Prof. F. Wang
Overview:
This retrospective study included 150 patients with stage IIA-IIIB epidermal growth factor receptor (EGFR)-mutant NSCLC who received neoadjuvant therapy between 2020 and 2025, and explored the predictive value of genomic alterations and treatment modality interaction for major pathological response (MPR), to inform the choice between EGFR-tyrosine kinase inhibitor (TKI) and chemo-immunotherapy (Chemo-IO). Although osimertinib has established adjuvant status (ADAURA trial), the optimal neoadjuvant regimen for resectable EGFR-mutant NSCLC remains unclear. The cohort included EGFR-TKI monotherapy (n=99) and Chemo-IO (n=51).
Interaction logistic regression analysis showed that, among 318 genes, only tumor protein p53 (TP53) had a significant interaction with treatment modality (interaction term P=0.002, odds ratio=0.054, 95% CI 0.008-0.345), which remained significant after Bonferroni correction (P=0.023). In the overall population, Chemo-IO had significantly higher pCR rate than EGFR-TKI (19.6% vs 3.0%, P=0.002) and higher MPR rate (41.2% vs 20.2%, P=0.011).
After TP53 stratification, treatment effects reversed: in TP53 wild-type patients (n=44), EGFR-TKI achieved higher MPR (38.7% vs 23.1%); in TP53-mutant patients (n=87), Chemo-IO achieved significantly higher MPR (48.6% vs 9.6%), with an absolute risk reduction of 39.0% in this subgroup favoring Chemo-IO (number needed to treat [NNT]=2.6). Bootstrap validation showed stable model performance (AUC=0.708, 95% CI 0.659-0.724). Multi-gene models incorporating MDM2 or RB1 did not outperform TP53 alone (P>0.3).
The findings suggest that TP53 status should be considered as a routine biomarker for neoadjuvant treatment selection in EGFR-mutant NSCLC.
Abstract MO09.07
Title: Surgery Versus Continued ALK-TKI Therapy Alone in Stage IV ALK-Positive NSCLC: A Propensity Score-Matched Retrospective Study
Presenter: Prof. Z. Li
Overview:
This retrospective study included 145 patients with stage IV anaplastic lymphoma kinase (ALK)-positive NSCLC treated at the Guangdong Lung Cancer Institute between 2016 and 2025, comparing survival outcomes of surgical consolidation after effective systemic therapy versus continued ALK-TKI therapy alone. Patients who had received ALK-TKI for at least 6 months and showed tumor shrinkage without progression on imaging at the 6-month assessment were included. From this time point, the surgery group was defined as patients who underwent pulmonary resection after induction ALK-TKI, and the non-surgery group as those who continued ALK-TKI monotherapy. Propensity score matching (PSM) was used to adjust for baseline variables including age, metastatic burden, and ALK-TKI generation. The primary endpoint was progression-free survival (PFS); secondary endpoints were overall survival (OS) and perioperative outcomes.
Of 145 patients, 42 underwent conversion surgery after induction ALK-TKI (lobectomy in 19 [45.2%], segmentectomy in 4 [9.5%], and wedge resection in 19 [45.2%]), and 103 continued ALK-TKI monotherapy. After PSM, 35 matched pairs were included, with balanced baseline characteristics between groups.
Median PFS in the surgery group was numerically longer than in the drug-alone group (53.4 vs 32.2 months), although not statistically significant (P=0.226); survival curves continued to separate during follow-up, suggesting that surgery may provide more durable disease control. The R0 resection rate in the surgery group was 95.2%, with a 90-day mortality of 0%.
The findings suggest that surgery-based consolidation therapy may provide additional PFS benefit for selected stage IV ALK-positive NSCLC patients, and this multidisciplinary strategy warrants further study.
Abstract MO11.07
Title: Integrated Analysis of Lung Microbiome Predicts Response to Anti-PD-1/PD-L1 Immunotherapy in Patients With Non-Small Cell Lung Cancer
Presenter: Prof. L. Wu
Overview:
This prospective study enrolled 64 treatment-naive patients with stage III-IV NSCLC without driver gene mutations from Guangdong, China. Bronchoalveolar lavage fluid (BALF) was collected at initial diagnosis, and all patients received first-line immunotherapy-based regimens. The study aimed to systematically characterize lung microbiota and metabolomic profiles and identify microbial biomarkers predictive of anti-PD-1/PD-L1 immunotherapy efficacy. BALF samples underwent shotgun metagenomic sequencing and liquid chromatography-tandem mass spectrometry (LC-MS/MS) metabolomic analysis. Alpha and beta diversity were assessed using Observed Species and Bray-Curtis dissimilarity, with principal coordinates analysis (PCoA). Predictive models were constructed using random forest (RF), ridge regression, and elastic net (Enet).
Results showed no difference in alpha diversity between responders (R) and non-responders (NR), but beta diversity was significantly higher in R than in NR (P<0.05). Six prediction models based on differentially enriched microbiota all achieved AUCs above 0.75, and all algorithms showed significant differences in first-line immunotherapy PFS between high-score and low-score groups. The intersection of the six methods identified 20 common strains associated with immunotherapy PFS.
Patients enriched with S. copri and A. cellobiosedens had significantly longer PFS (mPFS 12.5 vs 2.03 months, P<0.001; not reached vs 9 months, P=0.049); patients enriched with Psychrobacillus, C. saccharoperbutylacetonicum, E. hirae, and T. desertorum had significantly shorter PFS. The findings suggest that this study represents the largest systematic investigation of BALF microbial biomarkers in NSCLC, and that specific bacterial signatures may predict immunotherapy response and survival with relatively high accuracy.
Abstract MO15.03
Title: Pathological Response and MRD Dynamics After Neoadjuvant Serplulimab Plus Chemotherapy in Stage IIB-IIIB(N2) LS-SCLC
Presenter: Prof. Z. Chen
Overview:
This multicenter, single-arm phase II trial (NCT06911606) evaluated the efficacy and safety of neoadjuvant serplulimab plus platinum-based chemotherapy followed by surgery in patients with stage IIB-IIIB(N2) limited-stage small cell lung cancer (LS-SCLC). Concurrent chemoradiotherapy with immune consolidation remains the standard of care for LS-SCLC, but the role of definitive surgery is still debated. Patients received 4 cycles of serplulimab 300 mg plus etoposide and cisplatin or carboplatin as neoadjuvant therapy, followed by MDT evaluation; eligible patients underwent definitive resection 4-6 weeks after the last dose. The primary endpoint was pCR rate; key secondary endpoints included R0 resection rate, MPR, objective response rate (ORR), disease control rate (DCR), event-free survival (EFS), and safety. Circulating tumor DNA (ctDNA)-based minimal residual disease (MRD) dynamics was also explored.
As of January 2026, 45 patients had completed enrollment (median age 62 years; 24.4% stage IIB, 48.9% stage IIIA, 26.7% stage IIIB). The ORR for neoadjuvant immunochemotherapy was 88.9% (40/45) and the DCR was 100%; 40 patients (88.9%) underwent definitive resection, with a pCR rate of 40.0% (16/40), MPR rate of 62.5% (25/40), and R0 resection rate of 92.5% (37/40). At a median follow-up of 11.0 months, median EFS had not been reached, and the 12-month EFS rate was 86.1%.
Among 28 patients evaluable for MRD, 96.4% were baseline-positive; the MRD clearance rate was 47.8% (11/23) after neoadjuvant therapy and 77.8% (21/27) after surgery. Patients who were pre-operative MRD-negative had a significantly higher pCR rate than MRD-positive patients (66.7% vs 15.4%, P=0.015). Grade 3-4 adverse events occurred in 60.0% of patients, with the most common being neutropenia (51.1%); no new safety signals were identified.
The findings suggest that neoadjuvant serplulimab plus chemotherapy followed by surgery showed encouraging efficacy and an acceptable safety profile in stage IIB-IIIB(N2) LS-SCLC. Neoadjuvant MRD clearance was associated with pCR, and surgery further promoted MRD clearance.