---
title: "Genomic Stratification for First-Line Immunotherapy Combinations in Advanced Biliary Tract Cancer:"
id: "frontiers-in-immunology-18-exploratory-genomic-stratification-of-benefit-from-first-line-immunotherapy"
canonical_url: "https://medichelpline.com/clinical-feed/frontiers-in-immunology-18-exploratory-genomic-stratification-of-benefit-from-first-line-immunotherapy"
content_type: "clinical_feed_article"
specialty: "Infectious Disease"
source_name: "Frontiers in Immunology"
source_url: "https://www.frontiersin.org/articles/10.3389/fimmu.2026.1878780"
published_at: "2026-07-27T00:00:00.000Z"
evidence_level: "Journal Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# Genomic Stratification for First-Line Immunotherapy Combinations in Advanced Biliary Tract Cancer:
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/frontiers-in-immunology-18-exploratory-genomic-stratification-of-benefit-from-first-line-immunotherapy
- **Specialty:** [Infectious Disease](https://medichelpline.com/clinical-feed/infectious-disease.md)
- **Primary Source:** Frontiers in Immunology
- **Source URL:** [Original Journal Publication](https://www.frontiersin.org/articles/10.3389/fimmu.2026.1878780)
- **Published At:** 2026-07-27T00:00:00.000Z
- **Evidence Rating:** Journal Feed
## Executive GIST (TL;DR)
- The source article title indicates an exploratory **genomic stratification** and **biomarker analysis** of benefit from first-line **immunotherapy-based combination** therapy in **advanced biliary tract cancer** (BTC) within a randomized phase 2 trial. - The publicly available page content provided here contains site navigation and metadata only; the core manuscript text, methods, results, and numeric outcomes were not present in the source supplied. - Specifics about patient population, enrollment numbers, randomization arms, the exact immunotherapy regimen(s), combination partners, dosing, and treatment duration were not reported in the supplied source content. - Details on genomic assays used (panel size, platform, tissue vs blood, sequencing depth), definitions of molecular subgroups, reported biomarkers, or specific genomic alterations associated with differential benefit were not reported. - Efficacy endpoints (overall survival, progression-free survival, objective response rate), statistical comparisons between biomarker-defined subgroups, and any reported interaction tests were not reported in the supplied material. - Safety data, immune-related adverse events, tolerability comparisons between arms, and any biomarker-safety correlations were not reported. - The supplied source did not include conclusions, authors’ interpretations, clinical recommendations, or information on the trial registration, sites, funding, or ethical approvals. - Because the manuscript content was absent from the provided page capture, any detailed synthesis, numeric result reporting, or specific clinical implications cannot be inferred and would require access to the full article. - For clinicians and researchers, the title suggests a potentially important effort to identify genomic predictors of benefit to first-line immunotherapy combinations in BTC, but the supplied source does not present the data needed to evaluate validity or applicability. - Users seeking the trial’s findings, methods, or biomarker definitions should consult the full published article or the journal’s full-text record; such details were not available in the supplied source extract.
## Clinical Analysis & Structured Key Points
Frontiers | Exploratory genomic stratification of benefit from first-line immunotherapy-based combination in advanced biliary tract cancer: a biomarker analysis of a randomized phase 2 trial ORIGINAL RESEARCH article Front. Immunol. , 27 July 2026 Sec. Cancer Immunity and Immunotherapy Volume 17 - 2026 | https://doi.org/10.3389/fimmu.2026.1878780 Published in Frontiers in Immunology Cancer Immunity and Immunotherapy 7 impact factor 11.3 citescore Part of a Research Topic Beyond Conventional Biomarkers: Unlocking Immunotherapy Response Through Novel Biomarkers or Combinatorial Approaches 45k views 20 articles Editor & Reviewers Edited by M D Mutlu Demiray Reviewed by R M Ramon Mohanlal Y S Yunjie Song Outline Figures and Tables Figure 1 View in article Figure 2 View in article Figure 3 View in article Figure 4 View in article Figure 5 View in article Table 1 Baseline clinical characteristics of patients enrolled in this study. View in article ORIGINAL RESEARCH article Front. Immunol. , 27 July 2026 Sec. Cancer Immunity and Immunotherapy Volume 17 - 2026 | https://doi.org/10.3389/fimmu.2026.1878780 Exploratory genomic stratification of benefit from first-line immunotherapy-based combination in advanced biliary tract cancer: a biomarker analysis of a randomized phase 2 trial Y H Yatong He 1,2 † Z R Zeyu Ruan 1,2 † X X Xiaoqing Xu 3 † Y H Yue Han 1,2 J Y Junrong Yan 4 J N Jiaojiao Ni 2 Q X Qi Xu 2 J Y Jieer Ying 1,2,5 * S Z Shurui Zhou 1,2,5 * 1. Postgraduate Training Base Alliance of Wenzhou Medical University (Zhejiang Cancer Hospital), Hangzhou, Zhejiang, China 2. Department of Hepato-Pancreato-Biliary & Gastric Medical Oncology, Zhejiang Cancer Hospital, Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences, Hangzhou, China 3. Department of Oncology, Hangzhou Xixi Hospital, Hangzhou, China 4. Geneseeq Research Institute, Nanjing Geneseeq Technology Inc., Nanjing, China 5. Key Laboratory of Prevention, Diagnosis and Therapy of Upper Gastrointestinal Cancer of Zhejiang Province, Hangzhou, China See more Article metrics View details Abstract Background: Benefit from first-line immunotherapy-based treatment in advanced biliary tract cancer (BTC) is heterogeneous, and selection biomarkers are lacking. We investigated whether genomic features could stratify benefit from intensified immunotherapy-based treatment versus chemotherapy. Methods: This exploratory biomarker analysis included 58 patients from a randomized phase 2 trial of sintilimab, anlotinib, gemcitabine, and cisplatin (SAGC) versus gemcitabine plus cisplatin (GC) with baseline tumor tissue available for 425-gene targeted next-generation sequencing. Treatment-by-biomarker interactions for progression-free survival (PFS) were evaluated using Cox models. A genomic classifier was developed, and its feature selection robustness and internal stability were assessed by leave-one-out cross-validation (LOOCV) and bootstrap resampling. Results: Among 58 biomarker-evaluable patients, SAGC improved overall PFS versus GC (median 7.7 vs 6.3 months; HR 0.48, 95% CI 0.27–0.86; P = 0.011), with no overall survival (OS) difference (HR 0.98; P = 0.946). Evaluating these interactions identified DNA damage response (DDR) alteration and high tumor mutation burden (TMB-H) as the most robust predictors. Applied to both endpoints, the classifier stratified patients into SAGC-predominant (SP; either feature, n=40) and GC-predominant (GP; neither feature, n=18) subgroups, revealing diametrically opposed clinical trajectories (interaction P = 0.001 for PFS; P = 0.004 for OS). In SP, SAGC markedly prolonged PFS (10.7 vs 4.5 months; HR 0.26, 95% CI 0.12–0.55; P = 0.0002) and showed a favorable OS trend (16.9 vs 10.3 months; HR 0.55; P = 0.113). Conversely, in GP, SAGC yielded shorter PFS (5.8 vs 8.0 months; HR 2.70; P = 0.078) and significantly shorter OS (8.2 vs 13.9 months; HR 3.73, 95% CI 1.18–11.77; P = 0.017) compared to GC. LOOCV supported the reproducibility of feature prioritization and the internal stability of the final classifier, while bootstrap resampling supported the robustness of subgroup-specific treatment effects. Conclusions: An exploratory genomic classifier based on DDR alteration and TMB-H may help identify patients with advanced BTC more likely to benefit from the intensified first-line SAGC regimen, pending external validation. Introduction For a long time, first-line treatment for advanced biliary tract cancer (BTC) remained largely limited to gemcitabine plus cisplatin (GC) ( 1 ). More recently, the treatment landscape has begun to change, with trials such as TOPAZ-1 and KEYNOTE-966 introducing immunotherapy into the first-line setting ( 2 , 3 ). This represented an important step forward, but the magnitude of benefit in unselected populations remained modest. In the overall study populations of these trials, the improvement in survival was generally less than two months, suggesting that many patients still do not derive substantial benefit from standard chemo-immunotherapy. To improve on these results, more intensive combination strategies have been explored. One approach that has attracted increasing interest is the addition of anti-angiogenic agents to chemo-immunotherapy ( 4 – 6 ). The rationale is that vascular targeting may modulate the tumor microenvironment and enhance the antitumor efficacy of immunotherapy through positive feedback loops that reinforce each other ( 7 ). We recently evaluated this strategy in our multicenter, randomized phase 2 SAGC trial, which tested the addition of sintilimab and anlotinib to standard gemcitabine plus cisplatin ( 8 ). This quadruplet regimen improved tumor response and prolonged median progression-free survival (PFS) to 8.5 months, compared with 6.3 months with chemotherapy alone. However, despite the clear PFS improvement, no overall survival (OS) benefit was observed. Median OS (mOS) was nearly identical between the two treatment arms (13.2 vs. 13.7 months) ( 8 ). This discrepancy is clinically important, because it suggests that the conventional intention-to-treat analysis may be obscuring biologically meaningful heterogeneity within the BTC population. It is plausible that some molecular subgroups derive substantial benefit from the SAGC regimen, whereas others may gain little while being exposed to additional treatment burden. At present, however, there is no practical way to distinguish these patients. A more practical biomarker-based stratification approach is needed. A similar problem has been addressed in lung cancer research through the MINERVA score, which used a broader genomic signature rather than a single-gene biomarker to identify differential treatment benefit ( 9 ). That strategy revealed survival benefits that were not apparent in the overall population. We reasoned that advanced BTC may also benefit from a more precise biomarker-based stratification strategy. Against this background, we performed large-panel next-generation sequencing (NGS) covering 425 genes in 58 patients from the SAGC trial with sufficient baseline tumor tissue available. The purpose of this analysis was not merely to describe the mutational landscape, but to explore whether genomic features could help explain differential benefit from SAGC versus GC. More specifically, we aimed to identify predictive biomarkers and to explore whether a simple biomarker-based classifier could stratify patients into subgroups with different relative benefit from SAGC versus GC. We also sought to determine whether such stratification could uncover an OS signal in biomarker-enriched subsets, which was not observed in the intention-to-treat population. If successful, this approach could provide a practical step toward more individualized first-line treatment for advanced BTC. Methods Patient enrollment and data collection The 58 patients included in this genomic sub-study were drawn from the original 80-patient cohort of our randomized SAGC trial (ClinicalTrials.gov: NCT04300959) ( 8 ). Baseline tumor tissue of sufficient quality for 425-gene targeted next-generation sequencing (NGS) was available for 58 patients, including 30 in the SAGC arm and 28 in the GC arm. Of note, 16 biomarker-evaluable patients from the SAGC arm were also longitudinally included in our previously reported prospective observational multi-omics study (NCT06048289) ( 10 ). Clinical data, including treatment response and survival outcomes, were obtained from the primary trial database. The trial was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Zhejiang Cancer Hospital (Approval No.: IRB-2019-203). All patients provided written informed consent. DNA extraction, library preparation, and targeted sequencing Genomic DNA was extracted from archived FFPE tumor tissue using the QIAamp DNA FFPE Tissue Kit (Qiagen). DNA quantity and purity were assessed using a Qubit 3.0 fluorometer and a Nanodrop 2000. Sequencing libraries were prepared with the KAPA Hyper Prep Kit and enriched using the 425-gene GeneseeqPrime ® pan-cancer panel with xGen Lockdown reagents. Captured libraries were amplified with KAPA HiFi HotStart ReadyMix (KAPA Biosystems), quantified with the KAPA Library Quantification Kit, and sequenced on Illumina HiSeq 4000 platforms to a mean target coverage depth of at least 1000×. Bioinformatics and variant calling Raw data processing and variant calling followed our previously described pipeline ( 10 ) with study-specific refinements. Briefly, Trimmomatic was used for quality control, and clean reads were aligned to the hg19 reference genome using BWA. After duplicate removal and recalibration with Picard and GATK3, somatic single-nucleotide variants (SNVs) and insertion/deletions (indels) were identified using Mutect2. Variants with population frequency >1% in the 1000 Genomes Project or ExAC were excluded. A minimum variant allele frequency threshold of 1% was applied, requiring at least 3 supporting reads for hotspot mutations (≥20 entries in COSMIC v92) and 6 for non-hotspot mutations. All candidate variants were manually reviewed using the Integrative Genomics Viewer (IGV). Gene fusions were identified using FACTERA, and copy number variations (CNVs) were analyzed using ADTEx. Copy number gain and loss were defined by log2 ratio cut-offs of 2.0 and 0.6, respectively. To characterize pathway-level alterations, curated variants were mapped to the 10 canonical oncogenic signaling pathways and a curated core DNA damage response/repair (DDR) gene set. The DDR panel was compiled from KEGG, Reactome, and Gene Ontology annotations with literature support and was restricted to genes directly involved in canonical DNA repair and checkpoint signaling, including BER, NER, MMR, FA/ICL repair, HR, NHEJ, TLS, and ATM/ATR-mediated checkpoint pathways. Broader DDR-associated genes involved mainly in chromatin remodeling, DNA replication, ubiquitin signaling, or general cell-cycle regulation were excluded. The full gene list is provided in Supplementary Table 1 . Tumor mutational burden (TMB) was defined as somatic nonsynonymous mutations per megabase (muts/Mb), including only samples with ≥ 20% ABSOLUTE-estimated tumor purity. Patients were dichotomized into high TMB (TMB-H; ≥5) and low TMB (TMB-L; 40% of loci were classified as MSI-high (MSI-H). Development of a genomic stratification score to evaluate the relative benefit of SAGC To identify molecular features associated with differential treatment benefit, treatment-by-biomarker interaction effects were assessed using Cox proportional hazards models. For each candidate biomarker, the interaction effect was evaluated using the following model: where denotes the baseline hazard, ​the treatment assignment for patient (GC = 0, SAGC = 1), ​the biomarker status, the treatment-by-biomarker interaction term. The interaction coefficient ​was used to estimate the differential treatment effect associated with each biomarker. To assess whether the interaction effect remained robust after accounting for clinical characteristics, an adjusted Cox model was further fitted: Where represents the k -th clinical covariate for patient , and is the corresponding regression coefficient. The statistical significance of the interaction term in the adjusted model was used to determine whether the biomarker-treatment interaction remained significant after clinical adjustment. For each biomarker, a standardized interaction statistic ( -score) was calculated as: where denotes the treatment-by-biomarker interaction coefficient for biomarker , and its standard error. The sign of the z-score reflected the direction of treatment preference, with negative values indicating relative preference for SAGC and positive values indicating relative preference for GC. Because treatment assignment was randomized, the primary biomarker-selection framework was based on the unadjusted interaction model in order to preserve the randomized comparison structure. To assess robustness, clinically adjusted interaction analyses were additionally performed using multivariable Cox models including prespecified baseline clinical covariates. Biomarkers showing directionally consistent interaction signals across the unadjusted and adjusted analyses, together with supportive evidence from the adjusted model, were considered for classifier development. Because the selected biomarkers were binary and showed similar interaction strength in the same SAGC-favorable direction, the final genomic classifier was defined using a simple rule-based framework rather than a continuous weighted score. Patients with neither selected biomarker were assigned to the GC-predominant group (GP), whereas those with either or both biomarkers were assigned to the SAGC-predominant group (SP). Kaplan–Meier analyses were then performed within each subgroup to compare PFS and OS between the SAGC and GC arms. Internal validation of the genomic classifier Internal validation was comprehensively conducted using a two-step leave-one-out cross-validation (LOOCV) approach and bootstrap resampling. First, to address potential overfitting in the biomarker prioritization step, an n−1 LOOCV feature-selection sensitivity analysis was performed. In each of the 58 iterations, one patient was sequentially omitted, and the treatment-by-biomarker interaction analysis was independently re-run on the remaining 57 patients for all 19 candidate genomic features. Features were ranked by interaction P-value in each fold to evaluate whether DDR alteration and TMB-H remained consistently prioritized after removal of individual patients. Second, to evaluate the stability of the final locked classification rule, the omitted patient was reassigned according to the prespecified DDR/TMB-H rule. After completion of all LOOCV iterations, Kaplan–Meier analyses and Cox interaction models were performed using these LOOCV-derived subgroup assignments. Furthermore, bootstrap resampling was performed for 1, 000 iterations. For each iteration, a bootstrap sample equal in size to the original cohort was generated by sampling with replacement. Cox proportional hazards models were then fitted within each bootstrap sample to estimate treatment hazard ratios in the SP and GP subgroups and the treatment-by-subgroup interaction. Bootstrap estimates were summarized using empirical distributions. Statistical analyses Fisher’s exact test was used to compare categorical variables, and the Wilcoxon rank-sum test was used for continuous variables. PFS and OS were estimated using the Kaplan–Meier method, and differences between groups were assessed with the log-rank test. Hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were estimated using univariate Cox proportional hazards models. Overlap and association between DDR alteration and TMB-H, the two components of the genomic classifier, were evaluated using contingency counts, overlap proportions, the phi coefficient, and Fisher’s exact test. Collinearity among molecular variables was assessed using variance inflation factors (VIFs). All tests were two-sided, and P 0.05). Table 1 Patient chracteristics a SAGC group (n = 30) GC group (n = 28) P value Age (years), median (Range) 62 (34, 73) 60 (32, 78) 0.498 ≥ 65 12 (40.0%) 11 (39.3%) 1.000 < 65 18 (60.0%) 17 (60.7%) Sex 1.000 Female 10 (33.3%) 10 (35.7%) Male 20 (66.7%) 18 (64.3%) ECOGb performance status score 1.000 0 1 (3.3%) 0 (0%) 1 29 (96.7%) 28 (100.0%) Primary tumor type 0.276 Extrahepatic cholangiocarcinoma 4 (13.3%) 3 (10.7%) Gallbladder carcinoma 9 (30.0%) 4 (14.3%) Intrahepatic cholangiocarcinoma 17 (56.7%) 21 (75.0%) Metastases Liver 20 (66.7%) 21 (75.0%) 0.570 Lung 5 (16.7%) 6 (21.4%) 0.744 Bone 2 (6.7%) 5 (17.9%) 0.246 Peritoneum 7 (23.3%) 5 (17.9%) 0.749 Distant lymph node 20 (66.7%) 18 (64.3%) 1.000 Baseline clinical characteristics of patients enrolled in this study. a Data are n(%) or medium (Range). b ECOG, Eastern Cooperative Oncology Group. In this biomarker-evaluable cohort, the SAGC arm showed significantly longer progression-free survival than the GC arm. Median PFS (mPFS) was 7.7 months in the SAGC arm versus 6.3 months in the GC arm (HR, 0.48; 95% CI, 0.27–0.86; P = 0.011) ( Figure 1A ). Among the 56 patients with available OS follow-up data, mOS was generally comparable between the two treatment arms (median 11.3 vs. 11.4 months; HR, 0.98; 95% CI, 0.53–1.81; P = 0.946), with no apparent difference observed in this subset ( Figure 1B ). These findings were broadly consistent with the efficacy trend observed in the overall trial population. Figure 1 Clinical outcomes and genomic landscape of the biomarker-evaluable cohort. Kaplan–Meier curves for (A) progression-free survival (n=58) and (B) overall survival (n=56; 2 patients lacked evaluable OS follow-up data) in the SAGC and GC arms. (C) Oncoprint of the genomic landscape for 58 patients (30 SAGC, 28 GC). Only features occurring in five patients overall and two patients per arm are shown. Genomic landscape of the biomarker-evaluable cohort Comprehensive genomic profiling using the 425-gene panel revealed a mutational landscape consistent with the known biology of advanced BTC. As shown in the oncoprint ( Figure 1C ), only genes, genomic features, and pathway alterations occurring in at least five patients in the overall cohort and i
## Related Clinical Research

- [Acquired immune-mediated TTP after ivonescimab chemoimmunotherapy with cholestatic liver injury an](https://medichelpline.com/clinical-feed/frontiers-in-immunology-18-acquired-immune-mediated-thrombotic-thrombocytopenic-purpura-with-severe.md)
- [PD-L1 expression and survival in unresectable or recurrent gastric cancer treated with first-line](https://medichelpline.com/clinical-feed/british-journal-of-cancer-0-pd-l1-expression-and-survival-in-unresectable-recurrent-gastric-cancer-treated.md)
- [SLFN11 as a biomarker of cisplatin sensitivity in pediatric solid tumors](https://medichelpline.com/clinical-feed/british-journal-of-cancer-0-slfn11-enhances-cisplatin-sensitivity-in-pediatric-cancer-via-activation-of.md)
- [PCK2 inhibition restores cisplatin sensitivity in NSCLC by inducing ferroptosis](https://medichelpline.com/clinical-feed/pubmed-42599961.md) (DOI: 10.1096/fj.202602464R)
- [GLUT10 (SLC2A10) Expression in Breast Cancer and Its Role in Cisplatin Resistance](https://medichelpline.com/clinical-feed/pubmed-42595549.md) (DOI: 10.3760/cma.j.cn112152-20251018-00521)

## Navigation
- [← Back to Infectious Disease Feed](https://medichelpline.com/clinical-feed/infectious-disease.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.