Citations

TPS10057 Background: Ewing sarcoma (EWS) is an aggressive tumor of the bone and soft tissue primarily affecting adolescents and young adults. For patients with recurrent disease, outcomes remain poor with 5-year survival rates of 10%–15%, highlighting the need for novel approaches. STEAP1 is highly expressed in EWS cell lines, xenograft models, and patient tumor samples 1 and is transcriptionally regulated by EWSR1::FLI1 , the main oncogenic driver in EWS. Xaluritamig, a humanized bispecific XmAb 2+1 T-cell engager targeting STEAP1 and CD3, has demonstrated potent, antigen-dependent T-cell–mediated lysis of STEAP1-expressing EWS cell lines 1 and induced robust in vivo antitumor activity in EWS xenograft models with concomitant CD8+ T-cell activation. In metastatic castration-resistant prostate cancer, another STEAP1-expressing solid malignancy, xaluritamig has shown meaningful antitumor activity and a manageable safety profile 2 supporting its clinical evaluation in R/R EWS. Methods: This ongoing, phase 1b, single-arm, open-label, two-part, multicenter study (NCT07297979) is evaluating xaluritamig in pediatric, adolescent, and adult patients with a histologic diagnosis of EWS and an EWSR1 :: ETS fusion gene locally confirmed by next-generation sequencing. Patients must have R/R EWS following ≥1 line of systemic therapy and Karnofsky or Lansky performance status ≥70. Part 1 (dose confirmation) consists of cohort 1 (age ≥12 years) and cohort 2 (age ≥2 to <12 years), and part 2 (dose expansion) includes patients aged ≥2 years with no upper age limit at enrollment. This study includes a 21-day screening period, a treatment period, a safety follow-up period, and a long-term follow-up period. Patients receive xaluritamig every 2 weeks once reaching the target dose until radiographic disease progression per RECIST v1.1, clinical disease progression, unacceptable toxicity, initiation of other anticancer therapy not permitted per protocol, withdrawal of consent, death, or end of study as determined by the sponsor, whichever is earlier. Primary endpoints are safety, tolerability, and determination of recommended dose(s) for expansion. Secondary endpoints are pharmacokinetics, preliminary antitumor activity, and immunogenicity. Biomarker evaluation as an exploratory endpoint includes assessment of serum cytokines, other pro-inflammatory markers, immune-cell subsets, circulating tumor DNA, STEAP1 protein expression, and tumor genomic profile and their association with safety and efficacy. CRS mitigation plan includes step-up dosing, steroid premedication, and administration of interleukin-6 inhibitors as needed. Site activation commenced in January 2026 and enrollment is ongoing. 1. Nolan-Stevaux O. Cancer Res. 2020;80(16_Suppl):DDT02–03. 2. Kelly WK, Danila DC, Lin CC, et al. Cancer Discov . 2024;14(1):76-89. Clinical trial information: NCT07297979 .

e22006 Background: Metastasectomy is recommended for select pediatric patients with osteosarcoma with resectable pulmonary parenchymal metastases. It is unclear whether pulmonary metastasectomy confers the same benefit to patients with hilar or mediastinal lymph node metastases (LNM). This study aims to characterize pediatric patients with osteosarcoma with thoracic LNM and their outcomes compared to those with only pulmonary parenchymal metastasis. Methods: This is a retrospective cohort study of patients aged 0-21 with osteosarcoma metastatic to the lung who underwent pulmonary metastasectomy between 2000-2024. Patients were categorized into groups based on the presence or absence of LNM on imaging reviewed by pediatric surgeons and radiologists. Descriptive statistics and Wilcoxon rank sum, chi-square or Fisher’s exact tests evaluated differences between patients with and without LNM. Kaplan-Meier survival curves with log-rank tests evaluated the relationship between the presence of LNM and outcomes. The primary outcome was overall survival (OS) with secondary outcomes of recurrence-free survival (RFS) and disease-free survival (DFS). Results: 39 patients met inclusion criteria for this study, of which 13 (33%) had LNM (Table). Of the 13 patients with LNM, 8 (62%) had a pulmonary metastasectomy after their LNM diagnosis. Patients with LNM had poorer OS after first metastasectomy (log-rank p = .027), DFS after first metastasectomy (p = .018), and RFS after diagnosis (p = .039) than patients with pulmonary parenchymal metastasis without LNM. Patients who underwent metastasectomy after their LNM diagnosis had improved OS after LNM diagnosis than patients who did not undergo metastasectomy (pairwise log-rank test with Bonferroni correction p = .015). Conclusions: Pediatric patients with osteosarcoma metastatic to the lung with LNM had poorer OS, DFS, and RFS compared to patients without LNM. These results confirm the importance of attention to the presence of nodal involvement in patients with pulmonary metastatic osteosarcoma. Despite their poor prognosis, these results suggest a persistent survival benefit to surgical intervention. Total N = 39 No LMN N = 26 LMNN = 13 p Age 13.7 (9.5, 16.1) 13.9 (9.0, 15.9) 13.7 (10.7, 16.5) .777 Female 15 (38%) 10 (38%) 5 (38%) > .999 Histology Chondroblastic Osteoblastic Telangiectatic 11 (39%)16 (57%)1 (3.6%) 9 (50%)8 (44%)1 (5.6%) 2 (20%)8 (80%)0 (0%) .162 Primary tumor site Chest wall Femur Humerus Pelvis Radius Tibia 2 (5.1%)24 (62%)5 (13%)2 (5.1%)1 (2.6%)5 (13%) 2 (7.7%)14 (54%)3 (12%)2 (7.7%)1 (3.8%)4 (15%) 0 (0%)10 (77%)2 (15%)0 (0%)0 (0%)1 (7.7%) .821 Lung metastases present at time of diagnosis 20 (51%) 13 (50%) 7 (54%) .821 Treatment protocol AOST0331 AOST06P1 CCG 7921 P9754 30 (81%)2 (5.4%)4 (11%)1 (2.7%) 19 (79%)1 (4.2%)3 (13%)1 (4.2%) 11 (85%)1 (7.7%)1 (7.7%)0 (0%) > .999 Lung nodules removed during first metastasectomy 3 (2, 6) 3 (2, 5) 6 (1, 20) .567

Abstract Aims To characterize global variation in cardiac rehabilitation (CR) staffing, indications, and prescribed dose by country income classification and compare findings with the International Council of Cardiovascular Prevention and Rehabilitation’s 2016 Global Audit. Methods and results Cross-sectional REDCap survey of Phase II/post-discharge CR programmes (May–September 2025). Outcomes were multidisciplinary team size/composition, accepted cardiac and non-cardiac indications, and centre-based CR dose. Comparisons were made across World Bank country income classifications and Audit decade (2016 vs. 2025). Of 7025 programmes identified globally, 1505 (median programme response rate = 62%/country) initiated a survey, from 90/113 (80%) countries with CR. CR was delivered by a multidisciplinary team of 7 (median; P25–P75 = 5–8; nurses, physiotherapists, cardiologists, and dietitians most common); teams were smaller in low-income countries (median = 4; P &lt; 0.001), and physician specialty varied. Non-low-income programmes consistently accepted (≥90%) patients with acute coronary syndrome, with programmes in low-income countries less often accepting related interventions (e.g. surgery: low-income 0% vs. high-income 97%; P &lt; 0.01), but non-cardiac indications were more frequently accepted (e.g. primary prevention, diabetes, and lung disease). Acceptance of heart failure increased compared with the previous Audit (70% vs. 87%; P &lt; 0.001). Centre-based programmes prescribed a median of 20 h (P25–P75 = 12–30), typically two sessions/week over 8 weeks. Globally, 65 (72%) countries prescribed a median ≥12 sessions/programme. Programmes in upper-middle-income countries prescribed more sessions (median = 24, P25–P75 = 12–36; P &lt; 0.001). Dose did not differ meaningfully between Audits (P = 0.07). Conclusion CR programmes globally are offered by small, but multidisciplinary teams. They accept guideline-indicated patients, with applicable variation by context. Insufficient dose in one-quarter of countries underscores investment need.

9524 Background: Unresectable melanoma in-transit metastases (ITM) can pose a significant clinical challenge. Our recently published data comparing first-line treatments for ITM, isolated limb infusion or perfusion (ILI/ILP), immune checkpoint inhibitors (ICI) and intralesional talimogene laherparepvec (TVEC), found superior outcomes with TVEC. However, TVEC was used more often in patients with lower disease burden and earlier stage (IIIB vs. IIIC/D). Therefore, we sought to better investigate these differences with a propensity score-matched (PSM) analysis. Methods: A multi-institutional, international IRB- approved retrospective analysis was conducted for unresectable melanoma ITM from 2016-2022. PSM analysis comparing first-line ILI/ILP, ICI, and TVEC for was performed. Patients were matched for stage, number of ITM, anatomical site, CLND status, gender and age. Kaplan-Meier curves (OS, PFS, DMFS) and Cumulative incidence function (MSS) were used to visualize time-to-event outcomes in the PSM cohort, though both methods do not account for the matched structure. Results: 274 patients were identified, 120 female/154 male; 96 treated with ILI/ILP, 111 with ICI, and 67 with TVEC. Median follow-up was 40 months (range, 1-80). PSM was performed and identified 36 TVEC–ICI pairs and 27 TVEC–ILI pairs. These were combined resulting in a final cohort of 108 observations (45 TVEC, 36 ICI and 27 ILI). The groups were well balanced. No statistical difference was found for metastasis-free survival (DMFS), progression-free survival (PFS), melanoma-specific survival (MSS), and overall survival (OS) between treatment modalities. TVEC was associated with significant improvement in local PFS (median was 2.2 years (y) for ILI/ILP, 1.5y for ICI and not reached for TVEC, p=0.048). While not statistically significant, a clinical benefit was seen in PFS being longer for TVEC (2.1y for TVEC vs. 1.2y for ILI/ILP and 1.1y for ICI; p=0.2) and OS (not reached vs 2.9y for ILI/ILP and 5.2y for ICI; p=0.2). On mixed effect Cox regression proportional hazard model using the matched data set, there were no significant differences in any of the survival and recurrence outcomes. Conclusions: When patients are propensity-matched, there were no significant differences in first-line treatment modalities for unresectable melanoma ITM. Treatment with TVEC did result in clinically meaningful improved overall and local PFS and OS. This supports that selection of ITM therapy should incorporate an individualized approach, incorporating disease distribution, stage, burden, treatment side effects and patient priorities.

9535 Background: MUM carries a poor prognosis with few effective treatments, especially for HLA-A*02:01-negative pts. While a subset of MUM pts responds to immune checkpoint inhibitors, the comparative efficacy of IPI+PD1 versus PD1 monotherapy is unclear. This study compares objective response rate (ORR), progression-free survival (PFS), overall survival (OS), and safety between these regimens in MUM. Methods: MUM pts treated with PD1 or IPI+PD1 at 15 major melanoma centres (from Australia, Europe, United States and Israel) were included. Demographics, patient and disease characteristics, and clinical outcomes were examined. Univariate and multivariate (MVA) analyses were performed to identify clinical predictors of response and survival. Results: Of 412 MUM pts treated, 150 (36%) had PD1 and 262 (64%) received PD1+IPI. Compared to the PD1 group, PD1+IPI-treated pts were younger (64 vs. 70 years; p<0.001), had higher rate of elevated LDH (44% vs. 31%; p=0.037), and more frequently had prior treatment with tebentafusp (6.5% vs. 1.3%; p=0.031). Median follow-up from commencement of PD1+/-IPI was 4.9 years (95% CI 4.7 – 6.1). ORR was higher in IPI+PD1 group (18%) vs. PD1 (9%) (p=0.008), particularly in males (p=0.009), patients with liver metastases (p=0.018) and with lung metastases (p=0.011), with elevated LDH (p=0.047) and with no prior treatment with tebentafusp (p=0.007). PFS and OS at 1 and 2 years were numerically higher with IPI+PD1 (1- and 2-year PFS: 23% and 15%; 1- and 2-year OS: 63% and 39%) vs. PD1 (1- and 2-year PFS: 17% and 11%; 1- and 2-year OS: 54% and 36%), but these differences were not statistically significant (p>0.05). On MVA, adjusting for predefined variables (age, gender, ECOG PS, LDH, presence/absence of liver/lung metastases, prior treatment with tebentafusp), IPI+PD1 was associated with higher ORR (OR 2.71, 1.30 – 6.05; p=0.01) but not with PFS or OS compared to PD1. Presence of liver metastases (ORR [OR 0.28; 95% CI 0.13 - 0.64], PFS [HR 1.61; 95% CI 1.11 - 2.34], OS [HR 2.02; 95% CI 1.32 - 3.10]), ECOG PS≥2 (PFS [HR 1.82; 95% CI 1.07 - 3.10], OS [HR 2.24; 95% CI 1.25-4.01]), elevated LDH (PFS [HR 1.66; 95% CI 1.30 - 2.11], OS [HR 2.39; 95% CI 1.84-3.10]) and prior tebentafusp treatment (OS [HR 2.36; 95% CI 1.21-4.59]) were also independent predictors of response and/or survival. A higher percentage of pts experienced grade ≥3 immune-related adverse events (irAEs) in the IPI+PD1 group compared to the PD1 group (34% vs 13%, p<0.0001). Most pts ceased treatment due to progression (234, 57%), and more pts stopped due to toxicity in the IPI+PD1 vs. PD1 group (25% vs. 9%, p<0.0001). Conclusions: In pts with MUM, IPI+PD1 demonstrated a higher ORR but did not improve survival compared with PD1 alone. IPI+PD1 was more toxic, leading to early treatment discontinuation in one-quarter of pts. These findings may help guide treatment selection in MUM.

8579 Background: LOH may provide a means to target tumor versus normal cells, to augment the efficacy and safety of MSLN-targeted programs (Hecht et al. JCO . 2022). A2B694 is an autologous, logic-gated, Tmod CAR T therapy designed to improve tumor selectivity and decrease toxicity by integrating an MSLN CAR activator with an HLA-A*02 blocker (Hamburger et al. Mol Imm. 2020). Methods: The first-in-human, open-label, phase 1/2 EVEREST-2 (NCT06051695) study of A2B694 in patients with recurrent/metastatic MSLN-expressing cancers with tumor-associated HLA-A*02 LOH. The prescreening study BASECAMP-1 (NCT04981119) identifies eligible patients and cryopreserves leukapheresis product. Upon progression, A2B694 is manufactured and administered after lymphodepletion. Phase 1 primary objective: evaluate the safety and tolerability of A2B694 and identify a recommended phase 2 dose. Results: As of 05 January 2026, 13 patients were enrolled: 8 women/5 men, median age 59 years, 11 non-Hispanic White/2 Hispanic with unknown race. Tumor types included colorectal (n = 4), ovarian (n = 3), pancreatic (n = 3), non–small cell lung adenocarcinoma (NSCLC), gastro-esophageal, and mesothelioma (n = 1 each). A2B694 dose level (DL) groups were: DL1: 1×10 8 cells (n = 3), DL2: 2×10 8 cells (n = 4), DL3: 4×10 8 cells (n = 5), and DL4: 6×10 8 cells plus low-dose IL-2 (n = 1). Lymphodepletion prior to administration of A2B694 was well-tolerated, with expected, transient cytopenias. The only adverse event reported in more than 1 patient was grade 3 neutropenia. One patient had grade 3 ICANS and 1 patient had grade 1 CRS. There were no dose-limiting toxicities or new safety signals after up to 17 months follow-up. All 13 patients received A2B694, were efficacy-evaluable, and had A2B694 detected post-infusion in peripheral blood. While A2B694 was not detected in tumor biopsies collected in patients treated at DL1 (0/2), all patients treated at DL2-DL4 with available biopsies (3/3) had detectable A2B694 in the tumor microenvironment. A patient with KRAS G12V /STK11 co-mutated NSCLC who had progressed on carboplatin, pemetrexed, and pembrolizumab achieved a complete response (CR) at D90 post-infusion per RECIST 1.1 by central review and had a confirmed CR at D180. In addition, PET-CT scan and ctDNA on D190 demonstrated no evidence of disease. On D243, the patient had a CNS relapse, with an ongoing non-CNS CR per RANO-BM at D284. At M12, the patient's CT showed no new findings and persistence of A2B694 in the blood was confirmed by ddPCR. Conclusions: We report the first patient with NSCLC to have a CR after CAR T. Overall, A2B694 demonstrated manageable safety and tolerability in patients with advanced solid MSLN-expressing tumors with tumor-associated HLA-A*02 LOH. The maximum tolerated dose has not been reached; dose-escalation continues. Clinical trial information: NCT06051695 .

8524 Background: Teliso-V is a c-Met–directed antibody-drug conjugate comprising telisotuzumab and the microtubule polymerization inhibitor monomethyl auristatin E (MMAE) payload. In the Ph2 LUMINOSITY study (NCT03539536), Teliso-V showed durable responses and manageable safety (Camidge et al, JCO 2024;42:3000-11) resulting in its accelerated approval in locally advanced/metastatic NSQ NSCLC with high c-Met protein OE (3+, ≥50%), as determined by an FDA-approved test. MET amp is a negative prognostic risk factor in advanced NSCLC, frequently associated with recurrent disease. High levels of MET amp and c-Met OE are hallmarks of MET -addicted tumors. Here, we analyzed the effects of MET amp on clinical responses to Teliso-V in LUMINOSITY. Methods: MET amp was evaluated by FISH and ctDNA in baseline samples from 108 NSQ EGFR WT NSCLC pts with c-Met OE (3+, ≥25%). MET amp by FISH was defined as having focal MET amp (MET/CEP7 ≥2.0) with MET gene copy number (GCN) ≥4. MET amp (focal) by ctDNA was defined as having plasma MET GCN ≥4 with no co-amplification in CDK6 and EGFR. c-Met OE by Immunohistochemistry (IHC; SP44) was defined as 3+ tissue staining intensity in ≥25% tumor cells (high c-Met OE: 3+, ≥50%; intermediate [int] c-Met OE: 3+, 25% to 49%). Exploratory analysis of tumor response was assessed in 76 efficacy evaluable pts with c-Met OE and MET amp assay results. Results: MET amp was detected in 34% (37/108) of pts with c-Met OE (3+, ≥25%) and was enriched by c-Met levels: 22% (11/50) in pts with Int c-Met OE (3+, 25% to 49%) and 45% (26/58) in pts with high c-Met OE (3+, ≥50%). The effects of MET amp on tumor response in LUMINOSITY are shown in Table 1. The majority of c-Met OE pts (79%) with PFS ≥10 mo (n=14) had GCN ≥10 and/or c-Met IHC 3+ ≥50%. No new safety signals were reported in pts with c-Met OE and MET amp. Conclusions: MET amp is more common in pts with high c-Met OE in this retrospective subgroup analysis. Tumor activity with Teliso-V was observed regardless of MET amp status. The impact of MET amp in pts with c-Met OE will be further evaluated in ongoing Phase 3 study (NCT04928846). Teliso-V efficacy in NSQ EGFR -WT NSCLC pts with c-Met OE with or without MET amp in LUMINOSITY. Total c-Met OE(3+, ≥25%) Total c-Met OE(3+, ≥25%) Intermediate c-Met OE(3+, ≥25%-49%) Intermediate c-Met OE(3+, ≥25%-49%) High c-Met OE(3+, ≥50%) High c-Met OE(3+, ≥50%) MET amp(N) No (N=53) Yes (N=23) No (N=30) Yes (N=7) No (N=23) Yes (N=16) ORR %(95% CI) 28 (18.0, 41.6) 39 (22.2, 59.2) 23 (11.8, 40.9) 57 (25, 84.2) 35 (18.8, 55.1) 31 (14.2, 55.6) PFS,median, mo(95% CI) 5.26 (3.71, 8.11) 8.02 (4.47, 14.65) 5.32 (2.69, 8.11) 7.52 (4.47, NA) 4.17 (3.25, 8.87) 8.02 (3.06, 25.92) OS,median, mo(95% CI) 14.5 (8.18, 17.38) 13.86 (6.54, 22.24) 14.03 (3.65, 17.02) 9.79 (6.54, NA) 16.26 (5.59, 36.44) 13.86 (3.06, 30.29)

4013 Background: Pancreatic ductal adenocarcinoma (PDAC) is driven predominantly by oncogenic KRAS signaling, yet prior attempts to target the MAPK pathway have been limited by toxicity and emergence of resistance. Atebimetinib is a next-generation, deep cyclic inhibitor (DCI) of MEK designed to achieve pulsatile pathway inhibition with improved tolerability. We report a Phase 2a cohort evaluating atebimetinib in combination with modified gemcitabine/nab-paclitaxel (mGnP) in the first-line treatment of advanced or metastatic PDAC (NCT05585320). Methods: In this multicenter, open-label, nonrandomized Phase 2a study, patients with previously untreated advanced or metastatic PDAC were eligible for participation and were treated with orally administered atebimetinib daily in combination with mGnP (modification: every other week dosing). The primary endpoint was objective response rate (ORR) per investigator assessment using RECIST 1.1 criteria in response-evaluable patients. The cohort was powered (~80%, one sided alpha 0.05) using an optimal Simon’s 2-stage design. Results: Fifty-five patients were enrolled and treated with 320 mg atebimetinib daily + mGnP. ECOG performance status was 0-1 (100%); median age was 68 years (range 42-85 years), 62% were ≥65 years, and 55% were male. No grade 5 treatment-related adverse events (AE) were observed. No AE related to atebimetinib were higher than grade 3. Grade 3 AE related to atebimetinib occurred in 29% of participants, with the most frequent being rash (5%), ALT increased (5%), and AST increased (5%). Among the 50 response-evaluable patients, best overall response (BOR) achieved were 21 PR, 21 SD, 7 PD, 1 NE. The ORR was 42%, and DCR was 84%. With a median follow-up of 10.4 months (data cutoff Jan 7, 2026), 6- and 9-month OS rates were 88% and 79%, respectively; median PFS was 8.3 months and median OS was not reached. Conclusions: Atebimetinib in combination with mGnP demonstrated favorable safety and promising efficacy compared to historic standard-of-care (SoC) GnP. 1 A phase 3 registrational trial is planned to start in mid-2026 for evaluation of atebimetinib + mGnP versus SoC GnP. 1. Von Hoff DD, Ervin T, Arena FP, et al. Increased survival in pancreatic cancer with nab-paclitaxel plus gemcitabine. N Engl J Med . 2013;369(18):1691-1703. doi:10.1056/NEJMoa1304369 (PMID: 24131140). Clinical trial information: NCT05585320 .

3626 Background: For patients (pts) with dMMR stage III colon cancer, the addition of atezolizumab (atezo) to adjuvant 5-fluorouracil, leucovorin plus oxaliplatin (mFOLFOX6) significantly improved disease-free survival in ATOMIC (NCT02912559). Here we present HRQOL data. Methods: Pts with surgically resected stage III dMMR colon cancer were randomized (1:1) to receive mFOLFOX6 + atezo (840 mg IV q2 weeks) for 12 cycles (6 months [mo]) followed by atezo monotherapy for 13 cycles (12 mo total) versus mFOLFOX6 for 12 cycles. HRQOL was an exploratory endpoint measured using the FACT-C (includes FACT-G), FACT/GOG-NTX, EQ5D-5L, and PRO-CTCAE. Optional participation was by paper surveys at trial registration, prior to treatment cycles 4 and 7, and at 6, 12, and 36 mo after registration. Mean changes from baseline (BL) were compared between treatment arms using general linear mixed models (negative mean changes indicate worsening). At 6 mo, a two-sided 95% confidence interval excluding a difference of 1.9 points (FACT/GOG-NTX) and 6 points (FACT-C Trial Outcome Index [TOI]) favoring the mFOLFOX6 arm would indicate non-inferiority of the addition of atezo. Rates of side effect bother (FACT-G item GP5) and pt-reported adverse events (AEs) by PRO-CTCAE (composite scores, baseline adjusted) were compared using Fisher’s exact tests. Results: Of 712 randomized pts, 581 (285 atezo+mFOLFOX6; 296 mFOLFOX6) consented to participate in the survey. Pts completed 2398/3021 (79%) of expected surveys across all time points (1218/1489 [82%] atezo+mFOLFOX6; 1180/1532 [77%] mFOLFOX6). The difference in mean changes from BL between arms indicated non-inferiority of the addition of atezo at 6 mo (table); differences remained small at 12 and 36 mo. No significant differences in mean changes from BL between arms were found at 6, 12, or 36 mo for any scale. At 6 mo, 16% vs 20% of pts in each arm reported at least “quite a bit” of side effect bother (p=0.34), favoring atezo. Across cycles, atezo was associated with higher rates of pt-reported itchy skin (62% vs 48%), cough (53% vs 43%), and shortness of breath (51% vs 41%) [all p<0.05]. Conclusions: Addition of atezo to mFOLFOX6 slightly increases pt-reported AEs, but its effect on HRQOL is minimal and not clinically meaningful. Results support tolerability of atezo+mFOLFOX6 as standard of care adjuvant treatment for pts with dMMR stage III colon cancer. Support: UG1CA189823, U10CA180821, U10CA180882; Genentech, a member of the Roche group; https://acknowledgments.alliancefound.org. Clinical trial information: NCT02912559 . Scale Mo Mean change from BL – atezo+mFOLFOX6 Mean change from BL – mFOLFOX6 Diff (95% CI) FACT/GOG-NTX 6 -10.0 -10.6 0.6 (-0.8, 2.1) 12 -8.0 -9.4 1.4 (-0.2, 2.9) 36 -7.2 -6.9 -0.4 (-2.1, 1.4) FACT-C TOI 6 -4.1 -2.5 -1.6 (-3.7, 0.6) 12 2.1 1.8 0.3 (-1.9, 2.6) 36 0.7 1.8 -1.1 (-3.7, 1.5)

1617 Background: Large language models (LLMs) are increasingly evaluated for oncology decision support, yet guideline concordance and hallucinations remain variable across cancer types and tasks. Retrieval-augmented generation (RAG) may improve safety by constraining the evidence base used during generation. Prior oncology LLM studies report discordance with guideline-based care, but few describe a reproducible health-system “vetting mechanism” that couples guideline-only retrieval with clinician-scored safety rubrics across multiple oncology domains. Methods: We assembled 316 de-identified oncology cases in a tumor-board template: medical oncology (breast, GI, GU, leukemia; n=216), gynecologic oncology (n=50), and neuro-oncology/CNS metastatic cancer (n=50). Each case was run on three systems: baseline GPT-4 (O1), NCCN-guideline anchored RAG (O2; guideline-constrained retrieval), and a literature-anchored system (OpenEvidence; O3). Oncologists independently scored outputs using a modified Generative Performance Score (mGPS; −1 to +1) combining guideline concordance with hallucination penalties; readability/rationality was scored on a 1–5 Likert scale. Inter-rater reliability was assessed (κ =.63). This workflow was designed as a pre-deployment model credentialing step for health-system governance (source policy → case bank → clinician-scored acceptance criteria). Results: Across all 316 cases (case-averaged), O2 achieved the highest mean total mGPS (0.75) compared with O1 (0.43) and O3 (0.59). In the medical oncology cohort (n=216), O2 outperformed O1 and O3 (mean mGPS 0.72 vs 0.35 and 0.54), with higher guideline concordance (0.85 vs 0.68 and 0.75) and lower hallucination penalties (−0.09 vs −0.25 and −0.19); readability was also highest for O2 (4.17 vs 3.22 and 3.86). In gynecologic oncology (n=50), O2 again performed best (mGPS 0.827) relative to O3 (0.697) and O1 (0.647). In the CNS cohort (n=50), mean composite mGPS favored O2 (0.82) over O3 (0.68) and O1 (0.55), while readability remained stable across models (3.55–3.73). Severe hallucination/unsafe recommendation rate: O1 [26%], O2 [5%], O3 [17%] (pre-specified definition; e.g., major fabricated citation or guideline-discordant harmful recommendation). Conclusions: In a 316-case, multi-domain clinician-scored evaluation, limiting retrieval to oncology guideline sources improved safety-aligned performance relative to unconstrained generation or literature-only anchoring. A guideline-constrained RAG configuration paired with clinician-scored rubrics and tail-risk reporting can serve as a practical health-system vetting mechanism for governance, re-validation, and change control of LLM decision support, including in resource-limited settings where reliable guideline-consistent recommendations are essential.