397 Background: We previously found that adding D to induction FOLFOX and PET-directed CRT in resectable EA was safe and had promising pathologic responses and survival outcomes (Ann Surg 2023;278:e511). Here, we add T to the combination. Methods: Pts with locally advanced (cT3/4 and/or cN+) E/gastroesophageal junction A received 2 cycles of mFOLFOX6. PET responders (≥35% SUV reduction, PETr) received 5-FU/capecitabine + oxaliplatin with RT; PET non-responders (PETnr) received carboplatin/paclitaxel with RT. All received T 300 mg ×1 and D 1500 mg q4W ×2 starting 2 weeks before CRT. Esophagectomy was planned 6–8 weeks post-CRT. R0 resections were followed by adjuvant T 300 mg ×1 and D 1500 mg q4W ×6. Primary endpoint was pathologic complete response (pCR) rate. Results: 20 Pts were enrolled; 1 withdrew before treatment. Median follow-up was 23.9 (21.4–47.5) months. 1 Pt died of aspiration pneumonia while being treated for grade 4 colitis during CRT. Of 18 Pts who completed CRT, 3 achieved clinical complete response (cCR) and declined surgery and 1 Pt developed distant metastases. As such, 14 Pts had surgery. pCR rate was 26% (5/19, ITT) and 36% (5/14) in resected Pts. The combined pCR+cCR rate was 42% (8/19). Major pathologic response (MPR; Notable gd 1/2 adverse events (AEs) included fatigue (63%), nausea (63%), diarrhea (47%) and neutropenia (21%); gd ³3 events comprised neutropenia (16%) and acute kidney injury (5%). Immune-related AEs were rash/dermatitis (37%; gd ³3 11%), pruritus (21%; gd ³3 11%), colitis (21%; gd ³3 16%), arthralgia/arthritis (16%; all gd £2) and thyroid dysfunction (11%, all gd £2). PFS was 84.2% (95%CI 69.3-100.0) and 72.2% (95%CI 54.1-100.0) at 12- and 24-months. OS was 89.5% (95%CI 76.7-100.0) and 82.0% (95%CI 65.2-100.0) at 12- and 24-months, Of the 3 Pts with cCR who declined surgery, 1 relapsed locally at 14.7 months and was salvaged surgically (disease-free at 23.2 months), 1 developed brain/liver metastases at 16.2 months and 1 remains disease-free at 15.9 months. Conclusions: Adding D+T to PET-directed CRT in resectable esophageal/GEJ adenocarcinoma led to encouraging pathologic responses and survival outcomes. However, toxicities were qualitatively increased vs. D plus CRT. Although perioperative chemo is now standard, immunotherapy-based PET-directed CRT may remain relevant in Pts who wish to avoid esophagectomy. Clinical trial information: NCT02962063 .

AIMS:ASCL1, NEUROD1, POU2F3 and YAP1 are recently described markers of transcriptional subtypes in small cell lung carcinoma (SCLC), while DLL3, regulated by ASCL1, is a target of novel therapeutic agents in various neuroendocrine neoplasms. The expression of these markers in lung carcinoids is not well established. METHODS AND RESULTS:We examined these markers in 109 lung carcinoids and compared their expression with that in 191 enteropancreatic neuroendocrine tumours (EP-NETs) and with lung carcinoid markers (OTP, TTF1). ASCL1, NEUROD1, OTP and TTF1 were positive in 56%, 0%, 84% and 35% of lung carcinoids, respectively. Of the OTP-negative lung carcinoids (n = 18), 4 (22%) were ASCL1-positive, of which one was TTF1-positive. In contrast, 59% of EP-NETs were NEUROD1-positive, whereas only rare tumours focally expressed ASCL1 (1.1%) and OTP (0.5%) and none expressed TTF1. DLL3 was positive in 57 (52%) lung carcinoids versus 5 (2.6%) EP-NETs. All lung carcinoids and EP-NETs were completely negative for POU2F3 and YAP1. We also analysed clinicopathologic correlates of ASCL1, OTP, TTF1 and DLL3 expression in lung carcinoids, expanding on several previously suggested associations, including ASCL1 and TTF1 with peripheral location, OTP with low Ki67 (P = 0.002) and low stage (P = 0.002) and DLL3 with high Ki67 (P = 0.002). CONCLUSION:Unlike SCLC, lung carcinoids and EP-NETs completely lack the expression of POU2F3 and YAP1, which offers diagnostic applications. Our findings also nominate ASCL1 and NEUROD1 as site of origin markers for lung versus digestive NETs/carcinoids, respectively. Finally, the divergent expression of DLL3 in lung carcinoids and EP-NETs has therapeutic implications.

Intrahepatic cholangiocarcinoma (ICC) features poor survival due to frequent recurrences and limited prognostic markers. Using mass spectrometry-based proteomics, we analyze two independent cohorts comprising 80 and 62 treatment-naive ICC tumors, along with 9 independent patient-derived xenografts (PDX). In the first cohort, we identify two subclusters with distinct times-to-recurrence (TTR): An extracellular matrix (ECM)-enriched cluster (mean TTR 859 days) and a proliferation cluster (mean TTR 229 days). A 4-protein classifier trained on our cohort accurately stratifies these clusters in the Dong et al. dataset (2022) and in our second cohort, revealing similar proteomic motifs and clinical outcomes. The translation regulator EIF4A1, enriched in ICCs of both clusters, emerges as a therapeutic target, as its inhibition with eFT226 significantly reduces tumor growth in an ICC PDX model. Proteomic analyses of various PDX models also emphasize the critical role of tumor-stroma interactions in ICC. Overall, this study establishes two prognostic proteomic clusters, validates their relevance across datasets, and highlights EIF4A1 inhibition as a potential therapeutic strategy.

394 Background: EGFR (~7%) and c-MET (~5%) amplifications are recurrent events in GEA. While EGFR inhibitors were ineffective in an unselected population, retrospective data suggests benefit in amplified tumors. Amivantamab (ami) is a bispecific anti- EGFR and c-MET antibody. We report results of a multi-center investigator-initiated study of ami in patients with previously-treated EGFR and/or MET -amplified GEA (NCT05117931). Methods: Patients (pts) with advanced GEA who have received ≥1 prior treatment lines and have EGFR and/or MET amp tumors by tissue or plasma cell-free DNA (cfDNA) next generation sequencing (NGS) received ami at 1,050 mg IV weekly (or 1,400 mg if body weight ≥80 kg) in 28-day cycles. The primary endpoint is objective response by RECIST 1.1, and H0 is rejected if ≥ 6/25 (24%) patients achieve an objective response. Survival is stratified by EGFR vs MET and detection by enrollment assay (tissue vs plasma NGS), immunohistochemistry (IHC; 2-3+ by EGFR.113 or MET SP44), and fluorescence in situ hybridization (FISH). Results: At the data cut-off of 9/1/2025, 25 pts received ami with median follow up of 12.9 (range: 1.8-39.8) months. Pts received a median of 2 prior treatment lines (range: 1-5). One patient died of COVID19 after their initial ami dose and was inevaluable for response. Nine of 24 evaluable pts (38%) achieved an objective response ( EGFR amp: 5/17; MET amp: 3/5; both: 1/2). Six responses were confirmed, and the median duration of response was 6.4 (range 1.1-9.7) months. Six of 13 (46%) of patients with tissue amp and 5 of 15 (33%) with ctDNA amp achieved an objective response. Disease control was achieved in 17 of 24 (71%) pts. Median progression-free (PFS) and overall survival (OS) were 3.3 (95% CI 2.1-7.7) and 8.8 (95% CI 5.7-NR) months, respectively (Table). Despite a modest mPFS, 8 (32%) pts had PFS ≥ 7 months. At progression, 6 of 9 responders had persistent target lesion response with new escape lesions, including 2 with CNS metastases. Despite treatment discontinuation, one pt remains disease-free off treatment for over 2 years. Additional biomarker data will be presented. Drug-attributed adverse events were seen in 24 of 25 pts, most commonly infusion reaction (36%; grade 3: 4%) and rash (52%; grade 3: 4%). Conclusions: Ami monotherapy was effective in heavily pre-treated pts with GEA, including tumors with both EGFR and MET amps, and the primary endpoint was met. Despite persistent target lesion control, responders developed mixed responses attributable to resistance via EGFR/MET heterogeneity, co-occurring alterations, and CNS metastasis development. EGFR/MET targeting warrants further evaluation in combination with chemotherapy to address heterogeneity, a common barrier to single agent targeted therapy in GEA. Clinical trial information: NCT05117931 . PFS and OS by gene. mPFS (95% CI) months mOS (95% CI) months EGFR amp 3.5 (2.1-11) 11 (5.7-NR) MET amp 3.3 (3.1-NR) 6.5 (3.6-NR) Both amp 1.5 (0.9-NR) 2.2 (0.9-NR)

e16141 Background: Although pathologic complete response (pCR) after neoadjuvant therapy is associated with improved outcomes in gastric and gastroesophageal junction (GEJ) adenocarcinoma, whether its prognostic significance varies by tumor location, radiotherapy exposure, or systemic regimen remains unclear. In addition, recurrence patterns among complete responders have not been reassessed in contemporary cohorts. In this single-institution cohort, we investigate the prognostic significance of pCR across key disease and treatment contexts and evaluate recurrence patterns, including central nervous system (CNS) involvement. Methods: Patients receiving neoadjuvant therapy followed by R0 resection for gastric and GEJ adenocarcinoma between 1990 and 2024 were identified from a single-institution database. Final pathology classified patients as pCR (ypT0N0) or non-pCR (residual tumor with or without nodal disease, including ypT0N+). Analysis of overall survival (OS) was performed and stratified by tumor location and radiotherapy (RT) exposure and by systemic treatment (CRT, MAGIC-era chemotherapy, FLOT, IO±chemo). Results: Among 1783 patients receiving preoperative therapy, 1498 underwent R0 resection. The overall rate of pCR was 12.5% [n=187/1498], but differed across tumor location and RT exposure (gastric: 6.6% [n=44/670], GEJ+RT: 19.6% [n=123/629], GEJ-noRT: 10.1% [n=20/199]) and systemic regimen (CRT: 19.6%, MAGIC-era: 6%, FLOT-era: 9.1%, IO±chemo: 15.9%). pCR was associated with improved OS in gastric (HR: 0.36, p=0.001) and GEJ+RT (HR: 0.48, p<0.001), but not among GEJ-noRT (HR: 0.54, p=0.18). Notably, among gastric cancer patients achieving pCR, no disease-related deaths were reported. The survival benefit of pCR was strongest with CRT (HR: 0.48, p<0.001) and MAGIC-era regimens (HR: 0.45, p=0.01), while not statistically significant in FLOT (HR: 0.26, p=0.18). Complete responders experienced a significantly higher rate of CNS-involved failures relative to non-pCR (35% vs. 4.5% of recurrences, p<0.001), with enrichment primarily observed among GEJ+RT tumors. Other recurrence patterns (hematogenous, locoregional, peritoneal) and median time to relapse did not differ between groups. Conclusions: The prognostic value of pCR was not uniform across gastric and GEJ adenocarcinoma. Survival gains were greatest among gastric and radiated GEJ tumors, attenuated in non-radiated GEJ tumors and in FLOT, and accompanied by persistent enrichment of CNS-involved relapse among complete responders. Survival benefit of pCR by tumor location, radiation exposure, and treatment regimen. Stratum Subgroup HR CI p-value Site and Radiation Gastric 0.36 0.16–0.69 0.001 GEJ+RT 0.48 0.35–0.65 <0.001 GEJ-noRT 0.54 0.22–1.33 0.183 Treatment Regimen CRT 0.48 0.35–0.65 <0.001 MAGIC-era 0.45 0.25–0.83 0.01 FLOT-era 0.26 0.04–1.89 0.183 IO±chemo 0.52 0.40–0.67 <0.001

Abstract Background: Outcomes for advanced EGC are poor, particularly in those with peritoneal disease. MSLN is overexpressed in up to 40% of EGC. Preclinical models of peritoneal carcinomatosis demonstrated superior efficacy and acceptable safety with IP versus systemic administration of M28z1XXPD1DNR, a MSLN-directed autologous CART, equipped with a modified CD3z (1XX), and a PD-1 dominant negative receptor that provides intrinsic T-cell checkpoint blockade (Cancer Discov 2021;11:2748). Methods: This phase I study evaluated the safety of IP M28z1XXPD1DNR in patients with MSLN-positive (IHC ≥25%) ECG with peritoneal carcinomatosis. Eligible patients may have extraperitoneal disease, must have received ≥1 prior line of therapy and have measurable or evaluable disease (RECIST 1.1). Between 4 and 18 patients will be treated to a maximum of 1x107 cells/kg (Table 1). Lymphodepletion with fludarabine 30mg/m2/day and cyclophosphamide 300mg/m2/day is given for 3 days prior to cell infusion. Adverse events (AEs), response rates, survival outcomes and immune correlates will be assessed. Results: Eleven patients have undergone apheresis; 6 received M28z1XXPD1DNR and were evaluable for dose-limiting toxicities (DLTs). Median age was 56.5 (range 38-66). Median lines of prior therapy was 3 (range 1-5). Dose escalation proceeded through DL2 (n=2), DL3 (n=2) and DL4 (n=2). No DLT was observed. Treatment-related AEs were predominantly low grade (G), including cytokine release syndrome (5/6; all G1), transient post-infusion abdominal pain (5/6), nausea (5/6), vomiting (4/6), anorexia (3/6; G≥3 in 1), constipation (3/6), diarrhea (3/6) and fatigue (3/6. All patients had progressive disease at 8 weeks. One patient showed improvement in peritoneal disease but developed a bladder serosal metastasis. Peripheral blood vector copy number showed CART expansion peaking in the first month and remaining detectable beyond 100 days following IP administration. Conclusions: IP M28zXXPD1DNR CART cells were well tolerated. CART persistence was noted beyond Day 100. Antitumor activity in MSLN-positive EGC was limited. Given the higher level of MSLN expression in peritoneal mesothelioma, we are amending the study to recruit such patients. Citation Format: Joan R. Choo, David Restle, Alfredo Amador Molina, Jasmeen Saini, Adrian Gonzalez Aguirre, Steven B. Maron, Samuel L. Cytryn, David Ilson, Parastoo Dahi, Laura H. Tang, Daniel Cameron, Stacy Fernstedt, Kay See Tan, Prasad Adusumilli, Geoffrey Ku. A phase I trial of intraperitoneal (IP) mesothelin (MSLN)-targeted CAR T-cell therapy in patients with MSLN-positive esophagogastric cancer (EGC) with peritoneal carcinomatosis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(8_Suppl):Abstract nr CT026.

BackgroundPatients with gastric cancer and isolated positive peritoneal cytology have stage IV disease. The clinical significance of cytologic conversion is poorly understood. Our objectives are to (1) describe the clinicopathologic characteristics of patients with gastric cancer and isolated positive peritoneal cytology and their association with survival, (2) evaluate patients who underwent second cytologic evaluation for assessment of cytologic conversion, and (3) describe recurrence patterns in patients with cytologic conversion who underwent resection.Patients and MethodsPatients with gastric adenocarcinoma and isolated positive peritoneal cytology were identified from a prospectively maintained institutional database from 1996 to 2020 for this cohort study. Patients were characterized by selection for second cytologic evaluation following chemotherapy, cytologic conversion, and selection for surgical resection. Factors associated with overall survival (OS) were evaluated by time-dependent multivariable Cox regression models.ResultsOverall, 174 patients were identified, 62 (35.6%) were selected for second cytologic evaluation, 43 (69.4%) were cytologic converters, and 32 (74.4, 18.4% overall) underwent resection. Selection for second cytologic evaluation was associated with improved OS (hazard ratio [HR] 0.56, 95% CI 0.37-0.86), and among those patients, cytologic conversion was associated with improved OS (HR 0.17, 95% CI 0.08-0.35). Ten patients (31.3%) had no evidence of recurrence following resection during the study period. The peritoneum was the most common site of recurrence (n = 17, 53.1%).ConclusionsAmong patients with gastric cancer and isolated positive peritoneal cytology, second cytologic evaluation can provide prognostic information and inform treatment decisions. Cytologic conversion was documented in 25% of patients, with rare instances of long-term disease control.

Abstract Background: The addition of perioperative ICB to chemotherapy improves survival in localized gastroesophageal adenocarcinoma (GEA), but optimal treatment at recurrence remains undefined. A key question is whether prior perioperative ICB influences clinical response or immune sensitivity to subsequent ICB in the metastatic setting. We evaluated clinical outcomes and immune correlatives in patients who recurred after perioperative chemo-immunotherapy and underwent ICB rechallenge. Methods: We retrospectively analyzed patients with localized MSS GEA treated at Memorial Sloan Kettering (2020-2025) with curative intent perioperative chemo-immunotherapy (anti-PD-1 or anti-PD-L1). Patients treated on clinical trials were excluded. Clinical endpoints included RECIST response to ICB rechallenge at recurrence and progression-free survival (PFS) on rechallenge. Correlative studies (ongoing) include immune profiling and TCR clone-tracking from paired baseline and recurrence tumor biopsies and serial PBMCs in a subset of patients. Results: Among 66 patients with localized gastric (n=26) and esophagus/GEJ (n=40) cancer, 33 received neoadjuvant ICB (anti-PD-1 + FLOT, n=17; anti-PD-1 + FOLFOX/CAPEOX followed by surgery and adjuvant chemo/ICB, n=28; or non-operative management, n=5), and 33 received chemotherapy alone followed by surgery and adjuvant ICB only. Eighteen of 66 patients (27%) experienced disease recurrence or progression; of these, 55% (10/18) were rechallenged with ICB plus chemotherapy at recurrence (4 with prior neoadjuvant ICB; 6 with adjuvant-only ICB). Rechallenged patients had a longer interval from completion of perioperative therapy to recurrence compared with those not rechallenged (6.30 vs 1.27 months; p=0.315). Among rechallenged neoadjuvant-exposed patients, all four experienced tumor regression or disease stabilization with chemo-ICB (1 CR, 2 PRs, 1 SD). Among rechallenged patients who received adjuvant-only ICB, disease progression occurred in 50% (3/6), and two had short-lived responses; all ultimately progressed, including two deaths. Overall, median PFS from time of ICB rechallenge was 6 months (95% CI 4.27-NR): NR (0 events) in the neoadjuvant-exposed group versus 5 months (95% CI 4.14-NR) in the adjuvant-only group (log-rank p=0.056). Conclusions: In this perioperative chemo-immunotherapy cohort, prior neoadjuvant ICB exposure did not preclude - and may be associated with improved - response to ICB rechallenge at recurrence. Neoadjuvant-exposed patients showed numerically higher response rates and longer PFS on rechallenge compared with those who received adjuvant-only ICB. Ongoing correlative analyses of tumor and PBMC specimens aim to define immune determinants and TCR-based signatures associated with benefit from ICB rechallenge in MSS GEA. Citation Format: Jeremy M. Tchack, Samuel L. Cytryn, Steven B. Maron, Patrick Evans, Jessica Posada, Geoffrey Y. Ku, Jessica Yang, Ryan B. Sugarman, Ping Gu, Laura H. Tang, Amitabh Srivastava, Vivian E. Strong, Daniela Molena, Yelena Yuriy Janjigian. Immune and clinical determinants of response to immune checkpoint blockade (ICB) rechallenge after perioperative chemoImmunotherapy in gastroesophageal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6563.

Sustentacular cell presence and the absence of keratin expression represent diagnostic hallmarks for distinguishing paragangliomas from neuroendocrine tumors/carcinoids. However, based on the literature and our practice, pulmonary carcinoids can exhibit both features. Here, we examined the prevalence and biological correlates of this phenomenon. Lung carcinoids (N = 109) were analyzed with several common keratins, including AE1/AE3, sustentacular cell markers (S100 and SOX10), nuclear markers of lung carcinoids (OTP, TTF1, and ASCL1) and paragangliomas (GATA3 and PHOX2B), and 505-gene next-generation sequencing. AE1/AE3 was strikingly variable (mean H score, 127 of 300; SD, 80), with 18 cases (17%) exhibiting low labeling (H score, <50), including 3 AE1/AE3-negative cases. CAM5.2 was similarly variable (mean H score, 105; SD, 94) and was low/negative in 16 of 18 AE1/AE3-low cases. Conversely, pan-keratin OSCAR and CK18 were higher overall (mean H scores, 253 and 264, respectively) and showed unequivocal positivity in all AE1/AE3-low/negative cases. Sustentacular cells were present diffusely in 35% of carcinoids, with 15 of 18 AE1/AE3-low carcinoids containing sustentacular cells. Overall, 15 of 109 lung carcinoids (14%) were AE1/AE3 low and sustentacular cell positive. In contrast, such features were seen in only 4 of 188 enteropancreatic neuroendocrine tumors (2%). All AE1/AE3-low lung carcinoids were GATA3 and PHOX2B negative, while TTF1, OTP, and/or ASCL1 positive. By next-generation sequencing, all cases lacked mutations typical of paragangliomas, and many harbored alterations typical of lung carcinoids. Notably, lower AE1/AE3 was associated with the presence of sustentacular cells (S100: P = .003; SOX10: P = .005), spindle morphology, peripheral location, and ASCL1+/TTF1+/OTP+/HNF4A- immunophenotype-the constellation reflecting the emerging concept of "proneuronal" carcinoids. We conclude that lung carcinoids are not keratin negative but instead keratin variable, and some may appear as negative in an antibody-dependent manner. Together with the common presence of sustentacular cells, this may yield a profile overlapping with paragangliomas. We suggest using additional keratins and tumor-specific transcription factors (eg, OTP and GATA3) as the updated approach for this differential diagnosis. Potential biological implications of paraganglioma-like features in lung carcinoids are discussed.

Importance:Patients with gastric cancer face substantial risk of recurrence after surgical resection. Molecular profiling of primary tumors may improve risk stratification to guide postoperative management. Objective:To identify genomic features of primary gastric tumors associated with disease recurrence and patterns of metastatic spread. Design, Setting, and Participants:This single-center cohort study took place at an academic quaternary referral center and included patients who underwent curative-intent resection of gastric adenocarcinoma from 2010 to 2024. Patients were classified by recurrence status and pattern of metastatic spread. Patients with gastric cancer (stages I to III) who underwent a margin-negative resection and had genomic sequencing of their primary tumor were included. Primary analysis excluded patients who had no evidence of disease with less than 2 years of follow-up. These data were analyzed from June 2025 through March 2026. Exposures:Primary tumor specimens were sequenced using a targeted panel of cancer-associated genes (MSK-IMPACT). Main Outcomes and Measures:Correlation of disease-free survival and patterns of metastatic spread (hematogenous, peritoneal, or lymphatic) with primary tumor genomic profile. Results:Among 438 patients who underwent complete oncologic resection, 377 had sufficient clinical follow-up or developed recurrence (median [IQR] age, 64 [55-71] years; 113 female [30%] and 264 male [70%]). Recurrence was identified in 179 patients, whereas 198 patients had no evidence of disease. In a multivariable analysis, alterations in KRAS (hazard ratio [HR], 1.54; 95% CI, 1.04-2.28; P = .03) and PIK3CA (HR, 2.15; 95% CI, 1.25-3.69; P = .006) were independently associated with worse disease-free survival. Tumors of patients with hematogenous recurrence had greater chromosomal instability (fraction genome altered, 0.11 vs 0.03; P = .001), whole-genome duplication (47% vs 15%; P = .02), and more frequent alterations of genes modulating cell cycle regulation (39% vs 6%; P < .001) compared with peritoneal recurrence. Bone metastasis arose from more genomically stable primary tumors (fraction genome altered, 0.005 vs 0.114; P < .001) and tumors with Lauren diffuse-type histology (36% vs 3%; P < .001) compared with other sites of hematogenous spread. Conclusions and Relevance:This study identifies genomic alterations associated with disease-free survival and patterns of recurrence after resection of gastric cancer. These clinicogenomic risk factors may personalize postoperative surveillance strategies and inform perioperative treatment decisions.

e16149 Background: Perioperative durvalumab + FLOT improves overall survival and is a new treatment option for patients with resectable GEC. FDG-PET SUV decline of ≥35% is commonly used as a benchmark of treatment response. We correlated FDG-PET and endoscopic response with pathologic outcomes to evaluate their utility in selecting patients for non-operative management (NOM). Methods: Patients with localized MSS GEC treated with curative intent ICB+chemotherapy at MSK between 2020-2025 were retrospectively analyzed. Association of primary tumor baseline FDG avidity and %SUV change with surgical pathological outcomes [pathologic complete response (pCR), defined as absence of residual carcinoma; ypT3-4; and ypN+] was determined. High avidity was defined as SUV > 8, metabolic response as 35% reduction in FDG, and clinical complete response (cCR) as a negative biopsy on post-neoadjuvant therapy endoscopy (EGD). Results: Among 48 patients (20 gastric, 28 GEJ) who received neoadjuvant FLOT (n = 30) or FOLFOX (n = 18) + ICB, 43 underwent surgery. pCR occurred in 7/43 [16.3%; pCR with FOLFOX+ICB 1/13 (7.7%); 6/30 (20%) with FLOT+ ICB]. 35/48 underwent FDG PET at baseline and after therapy and had FDG-avid disease at baseline (SUV > 5). Median baseline SUV was 8.7 (range 5.1-37.5) with median %SUV change -55% (range +12% to -100%) after therapy. Baseline SUV (≥8 vs < 8), %SUV reduction (≥35% vs < 35 and ≥75% vs < 35%), and residual post-therapy avidity were not associated with pCR, ypT3–4 disease, or ypN+ disease. Findings were consistent by site (gastric vs GEJ/esophagus) and among patients receiving triplet chemotherapy. Compared to patients without cCR or pCR (n = 26 PET evaluable), patients with cCR or pCR (n = 9 PET evaluable) had significantly reduced median post-treatment avidity (SUV 2.8 vs 4.1; p = 0.03), however the groups did not significantly differ in median percent reduction in avidity or percent with post-treatment SUV < 3. Among 6 patients with complete resolution of PET avidity, 2/6 had pCR, 1/6 had cCR and proceeded to NOM, and 3/6 had residual disease including 2 with ypT4aN+ disease. 12/48 patients had an EGD after neoadjuvant therapy; 5 had cCR of whom 3 had residual FDG avidity on post-treatment PET. 1/5 of the cCR had pCR; the other 4 proceeded to NOM. In both cases of complete metabolic resolution and cCR, patients proceeded to NOM with no evidence of recurrence after minimum follow up of 3 years. Conclusions: Percent reduction in primary tumor PET avidity following neoadjuvant chemoimmunotherapy in resectable GEC was not associated with surgical pathology outcomes in our cohort. The combination of PET and endoscopy may better select patients for NOM paradigms, though higher resolution tools to assess for residual disease are needed. Multimodal predictive models for surgical pathology and long-term survival outcomes are being developed.

4175 Background: Genomic loss of MTAP and CDKN2A is frequently observed across cancers, with ongoing development of MTA-selective PRMT5 inhibitors aiming to achieve synthetic lethality in the setting of MTAP loss. However, the biomarker landscape of patients with gastrointestinal (GI) cancers and MTAP deficiency remains poorly defined, and whether subsets of GI cancers harbor non-genomic mechanisms of MTAP deficiency remains unknown. Methods: All GI cancer samples sequenced with MSK-IMPACT v7, a DNA-based next generation sequencing (NGS) panel with coverage for MTAP / CDKN2A were queried. MTAP and CDKN2A deletions (del) were defined by low read counts in coverage-based copy number analyses. Genomic and pathologic analyses were completed in tumors with high MTAP del prevalence: esophagogastric cancers (EGC), pancreatic adenocarcinoma (PDAC), pancreatic neuroendocrine tumors (PanNETs), and biliary tract cancers (BTCs). MTAP immunohistochemistry (IHC) was completed on PDAC, PanNET, small bowel NET, and BTC samples to investigate prevalence of MTAP protein loss. Results: Among 8346 GI cancer samples, MTAP del was identified in 456 (5.5%), co-occurring with CDKN2A del in 98.8% of cases, with 79.8% confirmed homozygous del by FACETS, and was associated with higher tumor purity compared to MTAP -intact samples (p < 0.001). MTAP del prevalence was 14.4% in esophageal squamous cell carcinoma (SCC), 5.9% in esophagogastric adenocarcinoma (EGA), 12.4% in PDAC, 7.4% in PanNET (grade 1 [G1]: 3.4%, G2: 4.3%, G3: 19.4%), 0% in small bowel NET, 10.2% in BTCs (including 11.5% in intrahepatic cholangiocarcinoma [iCCA]), and 0.7% in colorectal cancer. MTAP del was associated with wild-type (WT) KRAS amplifications (amp) in EGC, KRAS G12V and WT KRAS amp in PDAC, and KRAS mutations and FGFR2 fusions in iCCA. Among EGA, PDAC, and BTCs, MTAP del tumors were associated with higher fractions of genome altered (p < 0.001 for each) and rates of whole genome duplication (p < 0.001 for each). Among 463 samples profiled by MTAP IHC, MTAP deficiency was found in 29/93 (31%) PDAC, 3/33 (11%) BTC, 15/60 (25%) PanNETs, and 196/277 (71%) small bowel NETs. Analysis of MTAP deficiency and activity of standard systemic therapies, as well as metabolomic and epigenomic profiling of MTAP-deficient GI cancers are ongoing and will be presented. Conclusions: MTAP genomic del are recurrently identified across GI cancers, most notably in esophageal SCC, PDAC, G3 pancreatic NETs, and iCCA. MTAP del are associated with distinct genomic drivers across GI cancers and markers of chromosomal instability more broadly. We found evidence of discrepancy between MTAP loss by NGS and IHC across multiple GI cancers, most notably small bowel NETs, which demonstrate frequent MTAP IHC deficiency in complete absence of genomic loss, suggesting non-genomic mechanisms of MTAP silencing. These observations may help define the role for therapies targeting MTAP deficiency in GI cancers.