433 Background: Peritoneal metastasis (PM) from gastric cancer (GC) carries a dismal prognosis and remains refractory to systemic therapies. Oncolytic virotherapy offers a novel peritoneal-targeted approach. We evaluated the efficacy of CF33-hNIS, a chimeric orthopoxvirus, alone and in combination with anti–PD-L1 immune checkpoint blockade in an immunocompetent mouse model of gastric cancer peritoneal metastasis (GCPM). Methods: A syngeneic GCPM model was established using ACKPY3944 murine GC cells. Mice received intraperitoneal (IP) CF33-hNIS alone or in combination with intravenous (IV) or IP anti–PD-L1. Tumor regression, survival, and immune responses were assessed. Flow cytometry and immunohistochemistry characterized CD3⁺ T-cell subsets in the peritoneal cavity and tumor microenvironment. Complete tumor regression (CTR) mice were rechallenged with ACKPY3944 cells to evaluate memory T-cell responses. Results: IP CF33-hNIS monotherapy significantly prolonged survival and increased CD3⁺ and CD8⁺ T-cell infiltration in both peritoneal fluid and tumor tissue. The most effective regimen—a single high-dose CF33-hNIS (10⁸ pfu) combined with IP anti–PD-L1 (Combo 2)—achieved 75% CTR. Mice with CTR rejected tumor rechallenge and exhibited marked expansion of effector and central memory T cells in the peritoneal cavity and spleen. Conclusions: IP CF33-hNIS exerts potent anti-tumor activity in GCPM, enhanced by concurrent IP anti–PD-L1. The high-dose Combo 2 strategy induces durable tumor-specific immunity and provides strong rationale for clinical translation of simplified, peritoneal-targeted oncolytic viro-immunotherapy in gastric cancer.

574 Background: Hepatocellular carcinoma (HCC) is a leading cause of cancer-related mortality worldwide and first-line treatment options for advanced stages remain limited. Our study aims to evaluate the effectiveness and safety of different immunotherapy combination approaches in the treatment of HCC. Methods: We retrieved data from clinical trials published between 2018 and 2025. Individual patient data was extracted from already published Kaplan-Meier (KM) curves to permit comparisons across studies. A meta-analysis was performed to assess patient demographics, overall survival (OS), progression-free survival (PFS), and the incidence of adverse events (AEs). Results: Our analysis included a total of 3,312 patients, of whom 878 received an immune checkpoint inhibitor (ICPI) in combination with bevacizumab, 705 received ICPI combined with a tyrosine-kinase inhibitor (TKI), and 1,729 were treated with TKI monotherapy. The median OS was 19.09 months (95% CI: 17.09-21.27) for the ICPI+ bevacizumab group, 19.22 months (95% CI: 16.82-21.88) for the ICPI+ TKI group, and 15.29 months (95% CI: 13.96-16.27) for the TKI monotherapy group. Corresponding median PFS values were 5.76 months (95% CI: 5.55-6.74), 6.17 months (95% CI: 5.59-7.07), and 5.43 months (95% CI: 4.77-5.52), respectively. Regarding safety profile, the overall incidence of grade III/IV AEs was most common in the ICPI+ TKI group (72.35%), followed by TKI monotherapy (42.73%) and ICPI+ bevacizumab (38.65%) (p=0.1035). Hypertension (26.14% versus ~8%, p=0.0041), diarrhea (2.84%, p=0.0024), and platelet abnormalities (10.09%, p<0.0001) were significantly higher in the ICPI+ TKI group. Anemia was more frequent with ICPI+ bevacizumab (5.56%, p=0.0275). Other AEs showed no significant difference. Conclusions: Our analysis demonstrated improved OS in ICPI combinations with bevacizumab or TKIs in advanced HCC. However, the ICPI+ TKI combination regimen was associated with a higher incidence of certain AEs, which highlights the need to balance the efficacy and tolerability in the management of advanced HCC.

490 Background: Patients with hepatocellular carcinoma (HCC) and decompensated cirrhosis (Child-Pugh C) have a poor prognosis, due to underlying liver dysfunction. There is lack of data regarding systemic therapy in patients with HCC and CPC cirrhosis.. Therapeutic decisions in this setting are primarily guided by risk–benefit considerations, as systemic therapy may precipitate further hepatic decompensation. Consequently, most patients receive supportive care. Real-world data regarding systemic treatment outcomes in this population remain scarce. Methods: We utilized de-identified data from TriNetX, a global federated health research network, to identify patients with HCC and CPC cirrhosis using inclusion criteria requiring: bilirubin >2, albumin <3.5, INR >1.7, ascites, and encephalopathy two months prior to diagnosis. We analysed those who initiated systemic therapy with tyrosine kinase inhibitor (TKI) or an immune checkpoint inhibitor (ICI). Overall survival (OS) was assessed using Kaplan–Meier analysis, with group comparisons performed using log-rank tests. Propensity score matching was applied to adjust for age, sex, and comorbidities. Statistical significance was defined as two-sided P ≤ 0.05. Results: A total of 532 patients were included in the analysis. The median age was 63.9 years (range 55–73); 75.40% were male, 57.70% were Non-Hispanic White, 16.16% Black, 9.02% Asian, and 18.04% Hispanic. The Overall Survival (OS) rates for the entire cohort were 36.6% at 1 year, 25.3% at 2 years, 22.3% at 3 years, and 17.5% at 5 years. Among patients who initiated treatment within 3 months of diagnosis, 320 received frontline TKI therapy and 143 received ICI. Among the therapies used, sorafenib was most frequently used (42%), followed by lenvatinib, cabozantinib, regorafenib, atezolizumab plus bevacizumab, pembrolizumab, and durvalumab. After propensity score matching, 136 patients were included in each cohort. No significant difference in OS was observed between the TKI and ICI groups (Table 1). Conclusions: In this real-world cohort of HCC patients with CPC cirrhosis, OS was 36.6% at 1 year, 25.3% at 2 years, 22.3% at 3 years, and 17.5% at 5 years. Frontline TKI therapy showed numerically higher survival than ICI-based regimens, though the difference did not reach statistical significance. These findings highlight the need for prospective studies to better define treatment strategies in this high-risk population. Survival by treatment group. Time TKI OS% ICI OS% HR (95% CI) P Value 1 year 40.2 28 0.74 (0.54–1.03) 0.79 2 years 30.2 24.2 0.78 (0.57–1.06) 0.67 3 years 26.4 24.2 0.79 (0.58–1.08) 0.49

390 Background: Surgery alone often provides suboptimal outcomes for patients with locally advanced (LA) gastric and gastroesophageal junction (G/GEJ) adenocarcinoma, and additional perioperative therapy is required to improve survival outcomes. Recent evidence favors systemic treatment with perioperative chemotherapy (CT) over neoadjuvant chemoradiotherapy. Given the range of systemic therapies in this setting, we conducted a network meta-analysis (NMA) to compare their efficacy in resectable LA G/GEJ adenocarcinoma. Methods: A systematic review identified randomized controlled trials published up to August 20th, 2025, that enrolled patients with resectable G/GEJ adenocarcinoma. Eligible interventions included surgery combined with adjuvant, neoadjuvant, or perioperative CT ± immune checkpoint inhibitors (ICIs) or antiangiogenic agents. Trials evaluating radiotherapy were excluded. Hazard ratios (HR) for event-free survival (EFS) and overall survival (OS) were extracted or reconstructed. A Bayesian NMA with random-effects models (200,000 iterations, 20,000 burn-ins, 4 chains) was performed. HRs with 95% credible intervals (CrIs) were generated. Treatment rankings were assessed using the surface under the cumulative ranking curve (SUCRA). Results: A total of 50 trials (57 reports, 17,841 patients) were included. The most frequent comparators were surgery alone and adjuvant doublet CT. Data from 35 trials (15,328 patients) contributed to EFS analyses (16 regimens). Perioperative triplet CT + ICI showed a trend toward improved efficacy compared with perioperative triplet CT alone (HR 0.71; 95% CrI 0.50–1.00) and was superior to adjuvant triplet CT (HR 0.50; 95% CrI 0.30–0.79), but not to perioperative doublet CT + ICI (HR 0.76; 95% CrI 0.42–1.30). In ranking analyses for EFS, perioperative triplet CT + ICI had the highest SUCRA (0.94), followed by neoadjuvant triplet CT + adjuvant monotherapy CT (0.87) and perioperative doublet CT + ICI (0.76). For OS, 48 trials (16,601 patients) were analyzed (16 regimens). Perioperative triplet CT + ICI improved OS versus adjuvant triplet CT (HR 0.58; 95% CrI 0.35–0.94) and neoadjuvant triplet CT (HR 0.45; 95% CrI 0.23–0.88), but not versus perioperative triplet CT (HR 0.78; 95% CrI 0.53–1.14) or perioperative doublet CT + ICI (HR 0.82; 95% CrI 0.44–1.49). In ranking analyses for OS, perioperative triplet CT + ICI achieved the highest SUCRA (0.94), followed by perioperative doublet CT + ICI (0.81) and perioperative triplet CT (0.78). Conclusions: Perioperative triplet CT + ICI achieved the most favorable SUCRA for EFS and OS, supporting this therapy as a potentially preferred systemic strategy for LA G/GEJ adenocarcinoma. However, overlapping CrIs of this regimen with perioperative triplet and perioperative doublet CT + ICI highlight residual uncertainty.

526 Background: Hepatocellular carcinoma (HCC) is a leading cause of cancer-related mortality worldwide. Soy contains isoflavones and other bioactive compounds with anti-inflammatory and anti-proliferative effects that may influence liver carcinogenesis. However, epidemiologic studies have reported inconsistent findings. This review aims to investigate the association between soy food consumption and the risk of HCC. Methods: A systematic search was conducted in PubMed, Scopus, and Cochrane for relevant studies published up to 6 September 2025. We included case-control and cohort studies that reported risk estimates for HCC based on varying levels of soy consumption, and assessed heterogeneity using Cochrane’s Q and I² statistics. Random-effects meta-analysis with inverse-variance method was used to pool effect sizes comparing the highest versus lowest categories of soy food consumption. Results: Three studies involving 22,153 participants, all from Asian populations, were included. Based on two case-control studies, in unadjusted analysis, soy food consumption was associated with 10% reduced odds of HCC (odds ratio [OR] 0.90 [95%CI, 0.83-0.98]; I 2 =0.0%, p=0.69). However, after adjusting for key confounders including hepatitis B or C infection and other lifestyle factors such as alcohol use and smoking, this association was no longer significant (adjusted OR 0.46 [0.13-1.60]; I 2 =97%, p<0.001), consistent with findings from a prospective cohort study. Dose-response analysis was not feasible due to inconsistent measurement units. Conclusions: These findings suggest that soy consumption alone is unlikely to play a major role in HCC prevention. Large-scale prospective studies with standardized dietary assessments are needed to clarify whether soy has benefits in high-risk populations, particularly those with chronic liver disease.

190 Background: Microsatellite stable colorectal cancer (MSS-CRC) in advanced-stage disease remains a major therapeutic challenge due to its molecular heterogeneity. While classification frameworks such as CMS (Consensus Molecular Subtypes) and IMF (Intrinsic CMS, MSI, Fibrosis) have improved understanding of localized CRC, they offer limited insight into the biology of metastatic tumors. To address this gap, we implemented a multi-omic study of a cohort of treatment-naïve MSS-CRC patients with synchronous liver metastases (CRLM) and a large MSS-CRC Real World Dataset (RWD). Methods: Paired primary CRC and CRLM specimens from 20 patient underwent comprehensive profiling, including Whole Exome, Transcriptome: bulk and single-nucleus RNA sequencing (snRNA-seq) analysis. The intrinsic CMS framework (iCMS) was applied to characterize tumor-intrinsic states across primary CRC and CRLM. iCMS scores were inferred using previously published templates. Key findings were validated in a real-world dataset (TEMPUS MSS-CRC) profiled using multi-omics approaches. Results: Analysis of 388,018 single nuclei revealed 6 epithelial clusters associated with specific iCMS subtypes, supported by concordant bulk RNA-seq calls. Subtype composition at snRNA-seq level revealed that approximately one-third of tumors exhibited a hybrid phenotype with admixture of iCMS2 and iCMS3 epithelial cells. Tumors that were iCMS-unclassified in bulkRNAseq consistently displayed hybrid composition in snRNA-seq. Similarity metrics (cosine distance) from bulk data correlated with admixture proportions in snRNAseq. Genomic analysis revealed mutations in APC and KRAS as the predominant genetic events in non-hybrid iCMS2 and iCMS3 tumors, aligned with their differential WNT and MAPK pathway activity, respectively. In contrast, iCMS-hybrid tumors were significantly enriched for clonal co-mutations in WNT and MAPK drivers, with overrepresentation of specific KRAS variants (G12D, G13D). These iCMS-hybrid tumors exhibited intermediate WNT-MAPK co-activation. In the metastatic setting, iCMS2 and iCMS3 tumors maintained their subtype identity, whereas iCMS-hybrid primary tumors showed a shift toward iCMS2 dominance in liver, suggesting that the liver microenvironment promotes expansion of the iCMS2 state. Conclusions: Our study identifies a novel iCMS-hybrid phenotype with transcriptomic plasticity and co-activation of WNT and MAPK pathways, offering refined molecular resolution and potential to guide precision therapy in advanced MSS-CRC.

TPS250 Background: Fluoropyrimidine (FP) is a common chemotherapy class used across multiple tumor streams, including first line colorectal cancer management. Approximately 30% of patients develop serious toxicities with standard doses, potentially requiring hospitalization, intensive care admission and occasionally death. FP chemotherapy is primarily metabolized by dihydropyrimidine dehydrogenase (DPD, encoded by DPYD ), with DPD deficiency accounting for a significant number of serious toxicities. DPYD genotyping is a valid method of indirect measurement of DPD activity and has been implemented into routine care in some regions, enabling genotype-guided dosing. Despite mounting evidence, uptake of DPYD testing in Australia is currently sporadic and typically self-funded by the patient. This study will be the first to provide nation-wide prospective data on the feasibility and effectiveness of DPYD genotyping to direct FP dosing within the Australian health care system. Methods: This is a prospective, single arm, non-randomized study, aiming to recruit 5000 adult patients across Australia with solid organ malignancies intended to receive FP-based chemotherapy for the first time (trial registered 12/13/2023, Australian and New Zealand Cancer Trials Registry ACTRN12623001301651). DPYD genotyping will be performed for the key variants; c.1905+1G>A, ( DPYD *2A, rs3918290), c.2846A>T (rs67376798), c.1679T>G (DPYD*13, rs55886062.1) and c.1236G>A (rs56038477), with dose adjustments performed in accordance with published Australian guidelines. The primary outcome is incidence of serious (CTCAE v5.0 Grade 3+) FP-associated toxicities in the first 60 days in heterozygote DPYD variant carriers who receive a genotype-guided dose-adjustment, compared to: i) current standard of care (as assessed by a retrospective review of 500 consecutive cases); ii) contemporaneous cases without these variants treated with standard doses; and iii) variant carriers in this study that do not receive genotype-guided dosing. Primary endpoint data is anticipated to be completed by Q4 2027. Secondary outcomes include DPYD variant frequency, health economic and cost effectiveness analysis, and impact of dose adjustment on cancer outcomes. To date, 27 sites have been activated, with over 1200 participants enrolled. Completion of this study will inform a standardised approach for DPYD genotyping to minimise FP complications in Australia. Clinical trial information: ACTRN12623001301651 .

131 Background: For patients with refractory metastatic colorectal cancer, current therapeutic options are with limited efficacy both rechallenge treatment and the immunotherapy. Both preclinical and clinical studies have shown that EGFR antibodies and immune checkpoint inhibitors have synergistic effects. This study investigates the clinical efficacy and safety of ICE regimen (Irinotecan Plus Cetuximab and Envafolimab. Envafolimab is a new PD-L1 agent and was used by subcutaneous injection). Methods: This is a retrospective observational and single-center study, conducted in Beijing Hospital involving patients with RAS/BRAF wild-type /MSS mCRC who had received at least two previous lines of treatment. Molecular status was reconfirmed via circulating tumor DNA (ctDNA) analysis. The primary endpoint was objective response rate (ORR) according to RECIST v1.1. The secondary endpoints included disease control rate (DCR), progression-free survival (PFS), overall survival (OS), and safety. Results: Between July 2022 and August 2025, 20 patients received the ICE regimen: irinotecan (150 mg/m², Q2W), cetuximab (500 mg/m², Q2W), and envafolimab (200 mg, Q2W). The median age was 58.4 years (range: 40–74). Tumor locations included the left colon (n=11) and rectum (n=8)， only one right colon. Seventy percent of patients (14/20) had liver metastases and 40%（8/20）with both liver and lung metastases. The median number of prior therapies was 4.4 (range: 3–8). Among 19 evaluable patients, the ORR was 47.3% (9/19) and the DCR was 100% (19/19). At a median follow-up of 16.3 months, 85% (17/20) remained on treatment. Median PFS was 13.2 months (range: 2.1–31.3). Median OS was not reached due to limited events (3 deaths). Grade ≥3 treatment-related adverse events were mostly dermatologic, with rash occurring in 35% (7/20) of patients. Notably, one patient was treated with ICE as the third line regimen and achieved a clinical complete response (cCR) till now. Another patient experienced disease progression after a 1-year interruption of irinotecan, was re-challenged with ICE based on ctDNA confirmation of RAS/BRAF WT/MSS status, and he got SD. It is interesting that he had original FBXW7 mutation and resisted to prior three lines of systemic therapies before with short PFS. We also observed some patients had rechecked ctDNA after progressed and founded some new mutations. Conclusions: The ICE regimen demonstrates promising antitumor activity and manageable toxicity in heavily pretreated RAS/BRAF WT, MSS mCRC patients. A prospective, multicenter, open-label phase II trial (NCT06321081) investigating ICE as a rechallenge strategy is currently underway. Due to the favorable outcomes observed, we have initiated development of the “ICE2” protocol and are applying for its use as a first-line treatment in elderly patients with RAS/BRAF WT/MSS mCRC. Clinical trial information: NCT06321081 .

417 Background: Interleukin-1 Receptor Associated Kinase -4 (IRAK4) drives pro-survival NF-kB signaling in various gastrointestinal cancers. In preclinical animal models, the oral IRAK4 inhibitor, emavusertib (CA-4948), augments the efficacy of cytotoxic chemotherapy by suppressing cell intrinsic survival mechanisms and prolongs survival when combined with cytotoxic chemotherapy as well as reduces desmoplasia in preclinical models. Methods: This is a single-institution, Phase I, dose escalation/expansion clinical trial of emavusertib in combination with FOLFOX and PD-1 inhibitor (nivolumab or pembrolizumab) and anti-Her2 therapy (trastuzumab if Her2 positive) as first-line therapy for metastatic or unresectable esophageal, gastroesophageal junction, or gastric cancers (adenocarcinoma or squamous cell carcinoma) (NCT05187182). The primary objectives of this study are to determine the safety, dose-limiting toxicities, and Recommended phase 2 dose (RP2D) of emavusertib in combination with FOLFOX/ PD-1 inhibitor +/- trastuzumab. Emavusertib is given orally at escalating doses of DL0 (150mg BID) and DL1 (200mg BID) with FOLFOX/ PD-1 inhibitor +/- trastuzumab on a 14-day cycle. Dose escalation is according to the BOIN design to determine the MTD of emavusertib with the combination. Toxicities are graded according to CTCAE v5.0. Response is evaluated according to RECIST v1.1 criteria. Results: We have treated 7 patients in dose escalation (DL0 N=3, DL1 N=4) with no DLTs at either dose level. One patient in DL1 was replaced due to non-compliance with study therapy. DL1 (emavusertib 200mg BID) was determined to be the expansion dose. Four patients have been treated at DL1 in dose expansion cohort A (Her2 negative), and 2 patients have been treated at DL1 in dose expansion cohort B (Her2 positive). There have been no SAEs related to emavusertib at either dose level. Of the 3 patients treated at DL0 (escalation), there was 1 CR, 1 PR, and 1 SD ( the CR patient progressed after 25 months on study) . Of all patients treated at DL1 who have reached their first scan on study (escalation and expansion), there are 2 PRs and 4 SDs. For all patients, the median duration on study is 4.9 months (range 0.5-25 months). Nine patients continue to receive study therapy. Conclusions: At this early stage of the study, emavusertib in combination with FOLFOX/ PD-1 inhibitor +/- trastuzumab as first-line therapy for metastatic or unresectable gastroesophageal cancers has a manageable toxicity profile and shows encouraging preliminary results. Enrollment in dose expansion arms at emavusertib 200mg BID (Her2 positive and negative) is ongoing. Clinical trial information: NCT05187182 .

TPS793 Background: mPDAC has a poor prognosis with a median overall survival < 1 year with chemotherapy. Gremlin-1, secreted by cancer-associated fibroblasts and tumor cells, represses bone morphogenetic protein (BMP) signalling, thereby promoting proliferation and invasiveness. GREM1 mRNA is detected in the stroma of approximately 70% of PDAC cases. Gremlin-1 protein expression correlates with poor prognosis in PDAC. In pre-clinical models, Ab7326 (murine version of ginisortamab), blocks gremlin-1 BMP binding, restores BMP-2, 4, & 7 signalling and downstream SMAD pathway activation, and increases survival in the gremlin-1 expressing KPC murine model of PDAC in combination with either gemcitabine or MEK inhibition. Ginisortamab, a fully humanised IgG monoclonal antibody, was well tolerated when administered as monotherapy by intravenous (IV) infusion on days 1 & 15 of a 28-day cycle in pts with advanced refractory cancers in a first-in-human ONCO01 study (https://clinicaltrials.gov/study/NCT04393298). Methods: Phase II multi-centre study (UK, Norway, Spain, Germany) with safety run-in and dose expansion (proof of concept) in 2 modules in pts with mPDAC, ECOG ≤ 1, adequate organ function, RECIST1.1 measurable disease. Module 1 (no prior chemotherapy for mPDAC): non-randomised exploratory efficacy study (n = 60) with primary outcomes of progression-free survival (PFS) and disease control rate (DCR) at 16 weeks. Pts receive ginisortamab 2000mg IV infusion on days 1 & 15 with AG (standard doses) IV on days 1, 8, 15 (28-day cycle). Radiological assessment (RECIST1.1), QoL, CA19-9 every 8 weeks. PK collected days 1, 2, 4, 8, 15, 16, 18, 22 (cycle 1), days 1, 8, 15 (cycle 2) and pre-infusion days 1, 15 other cycles (safety run-in pts). Mandatory pre- and on-treatment (cycle 2 day 8 +/- 5 days) tumor biopsy to include SMAD4 and gremlin-1 expression (all pts). Module 1 commenced recruitment Q2 2024. Module 2: mPDAC (n = 60) with ongoing stable disease or response after ≥ 16 weeks of standard of care 1 st -line chemotherapy, randomised 2:1 to ginisortamab (IV days 1, 15) + MEK inhibitor vs observation with primary outcome of PFS. Module 2 is planned to commence in Q3 / Q4 2025. This study is funded and sponsored by Cancer Research UK in collaboration with UCB. The EU CT Number is 2024-514129-43, and the IRAS Identifier is 1007925. Clinical trial information: NCT04393298 .