Jin, Zhaohui

Abstract Prostate-specific membrane antigen (PSMA) is highly expressed in tumor-associated neovasculature in hepatocellular carcinoma (HCC), highlighting its promise as a theranostic target. This prospective study aimed to evaluate the utility of novel biomarkers derived from PSMA positron emission tomography computed tomography (PET/CT) to assess treatment response in advanced HCC patients (pts) receiving first line immune checkpoint inhibitors (ICI). Methods: This study (NCT05176223) enrolled unresectable, advanced HCC pts eligible for first-line ICI (atezolizumab/bevacizumab [A/B] or durvalumab/tremelimumab [D/T]) with measurable disease according to RECIST criteria. Primary endpoints: progression-free survival (PFS) per RECIST 1.1 and mRECIST; overall survival (OS); and best overall response. The survival distribution was estimated using the Kaplan-Meier method. Observed agreement is defined as the proportion of pts whose responses, as classified by PET/CT or CT, were concordant. Weighted kappa and Cohen’s kappa were used to quantify the strength of agreement between PET/CT and CT. Pts completed up to five 68Ga-PSMA PET/CT scans and blood collections. Results: Twenty-six pts were enrolled (25 evaluable) between Oct 2022 and Nov 2024. Median follow up was 22.1 months (m) (Q1, Q3: 14.5-23.7). Median age 68 yrs, 72% male, 88% White, 68% Child-Pugh [CP] A, 32% CP B, 32% Barcelona Clinic Liver Cancer [BCLC] stage B, 68% BCLC stage C, 36% with microvascular invasion; 56% received A/B and 40% received D/T treatment. PSMA was positive in 96% of pts at baseline with median SUVmax 14.3 (range 5.1-40.5) and median background liver SUVmax 3.9 (range 2.1-6.8). Median (95%CI) PFS was 5.6 m (4.6-12.5) by both RECIST and mRECIST. Median (95%CI) OS was 18.6 m (7.2-not estimable), with 56% deceased. The observed agreement between PET/CT adapted PERCIST criteria compared to RECIST and mRECIST in determining the best response was 46% and 50% and weighted kappa of 0.63 (95% CI: 0.37-0.90) and 0.66 (0.41-0.92), respectively. Observed agreement between PERCIST compared to RECIST and mRECIST for responders’ assessment was 67% and 71% with Cohen’s kappa of 0.36 (0.03-0.68) and 0.43 (0.11-0.76), respectively. Conclusion: In this cohort, PSMA PET/CT avidity was observed in all but one pt with advanced HCC, each demonstrating moderate to high SUVmax values. Median OS and PFS were consistent with outcomes reported in clinical trials. Response assessment using 68Ga-PSMA PET/CT according to PERCIST criteria showed good concordance with both RECIST and mRECIST standards. Analysis of immune biomarkers is ongoing. Citation Format: Nguyen H. Tran, Jacob Hirdler, Ajith Antony, Jacob Teske, Sudhakar K. Venkatesh, Amit Mahipal, Michael S. Torbenson, Scott M. Thompson, Thor R. Halfdanarson, Zhaohui Jin, Lionel Aurelien Kankeu Fonkoua, Leslie A. Washburn, Alexander Revzin, Haidong Dong, Robert R. Mcwilliams, Aminah Jatoi, Hani M. Babiker, Mitesh J. Borad, Tanios Bekaii-Saab, Gregory J. Gores, Lewis R. Roberts, Geoffrey B. Johnson, Ajit H. Goenka. Prospective evaluation of 68Ga-PSMA PET imaging in advanced hepatocellular carcinoma patients undergoing first line immunotherapy [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 2611.

Abstract Background The incidence of gastroesophageal cancer diagnosed at ages younger than 50 years is increasing. Assessment of the clinicopathologic and molecular characteristics of these cancers is limited currently. Methods This analysis used Tempus’ clinicogenomic database to compare patients diagnosed with gastroesophageal cancer at ages younger than 50 vs 50 years and older who completed tumor molecular profiling between December 2017 and July 2024. The spectrum of biomarkers, somatic alterations, and germline alterations were compared between those diagnosed with gastroesophageal cancer at ages younger than 50 years and those aged 50 years and older, and a multivariate analysis was conducted to identify factors associated with longer survival. Results Among 5863 gastroesophageal cancer patients, 785 had gastroesophageal cancer diagnosed at ages younger than 50 years. There were more females (35% vs 25%, P < .001) and fewer White patients (69% vs 80%, P < .001) with gastroesophageal cancer at ages younger than 50 years. Fewer tumors of patients with gastroesophageal cancer at ages younger than 50 years had high tumor mutational burden (3.2% vs 10.2%, P < .001) or high microsatellite instability (1.7% vs 5.1%, P < .001), but there was no difference in PD-L1 positivity (49% vs 51%, P = .3). Higher incidence of somatic CDH1 mutations was seen in those with gastroesophageal cancer at ages younger than 50 years (16.1% vs 6.6%, q < 0.001) with lower incidence of TP53 (65.6% vs 74.1%), CDKN2A (12.5% vs 19.8%), and KRAS (11.7% vs 18.3%, P < .001 for all) alterations. The most statistically significant germline difference was higher incidence of CDH1 (2.2% vs 0.3%, q = 0.001) and TP53 (0.6% vs 0%, q = 0.039) alterations among those with gastroesophageal cancer at ages younger than 50 years. Median survival for patients with metastatic disease was longer for gastroesophageal cancer at ages younger than 50 years (11 vs 8.5 months, P = .002), and gastroesophageal cancer at ages younger than 50 years, ERBB2 amplification, and high microsatellite instability were associated with superior survival. Conclusions Unique molecular and germline profiles were seen in patients with gastroesophageal cancer at ages younger than 50 years, which may suggest differences in causal factors yet undiscovered.

PURPOSE Gastroesophageal carcinoma (GEC) is increasingly managed using biomarker-guided therapies. Although both tissue DNA (tDNA) and circulating tumor DNA (ctDNA) testing are used clinically, the concordance of genomic alterations between primary tumors, metastatic sites, and liquid biopsy remains incompletely understood. METHODS We retrospectively analyzed 345 patients with GEC who underwent ctDNA testing at Mayo Clinic. tDNA next-generation sequencing and immunohistochemistry (IHC) for human epidermal growth factor receptor 2 (HER2)/erb-b2 receptor tyrosine kinase 2 ( ERBB2 ) and mismatch repair (MMR) were collected. Concordance was evaluated between ctDNA and tDNA, ctDNA and IHC, and across all three modalities, using 60-day and 30-day intervals. RESULTS Among 345 patients with ctDNA testing, 186 had tDNA testing, and 310 underwent IHC. ctDNA testing showed a shorter turnaround time than tDNA (median 6 v 13 days). The most frequent alteration in ctDNA, primary tDNA, and metastatic tDNA was TP53. Concordance between ctDNA and primary tDNA was ≥78.8% for key genes; between ctDNA and metastatic tDNA, it was ≥84.8%. IHC-ctDNA concordance was higher in metastatic tumors, with 91.1% for HER2/ ERBB2 and 97.7% for mismatch repair (MMR)/microsatellite instability (MSI) compared with 81.1% and 95.8%, respectively, in primary tumors. Triplet concordance for MSI/MMR was 97.6% in primary tumors and 100% in metastatic tumors, while ERBB2 /HER2 showed 84.4% and 93.9% concordance, respectively. ctDNA temporal heterogeneity analysis revealed 14 patients with fluctuating ERBB2 status and 3 with fluctuating MSI-H status. CONCLUSION ctDNA testing demonstrates high concordance with tDNA and IHC and offers faster turnaround. Given its minimally invasive nature and strong concordance with tissue-based testing, ctDNA is a reliable tool for genomic profiling and longitudinal monitoring in GEC.

Background: Metastatic squamous cell carcinoma of the anus (SCCA) carries a poor prognosis, with systemic therapy remaining the standard of care. While metastasis-directed therapy improves outcomes in select gastrointestinal malignancies, the role of liver-directed intervention in metastatic SCCA remains undefined. We evaluated clinicopathologic features and oncologic outcomes of patients with liver-limited metastatic SCCA treated with curative-intent hepatic local therapy at a tertiary academic center. Methods: We conducted a retrospective cohort study of adults with histologically confirmed SCCA and liver-only metastatic disease who underwent curative-intent hepatic resection or ablation at Mayo Clinic between 1993 and 2023. Patients with extrahepatic disease or incomplete records were excluded. Demographic, tumor, treatment, and outcomes data were abstracted from electronic medical records. Overall survival (OS) and disease-free survival (DFS) were estimated using the Kaplan–Meier method. Prognostic factors were assessed using univariate Cox proportional hazards models. Results: Twenty-five patients met inclusion criteria. Median age was 56.7 years, and 92% were female. Most patients had metachronous metastases (76%), a single hepatic lesion (56%), and unilobar disease (76%). Pre-intervention systemic therapy was administered in 52% of patients, with radiographic complete or partial responses observed in all treated patients. Liver-directed therapy consisted of surgical resection in 80% and thermal ablation in 20%. Among surgical patients, 90% achieved microscopically negative margins. With a median follow-up of 22 months, disease recurrence occurred in 80% of patients, most commonly within the liver. Median DFS was 7.27 months. Median OS from the date of liver-directed therapy was 51.3 months. On univariate analysis, poorly differentiated tumor histology was associated with inferior OS (hazard ratio 4.67, p = 0.018). No other clinicopathologic variables were significantly associated with DFS or OS. Conclusions: In this single-institution experience, carefully selected patients with liver-limited metastatic SCCA undergoing curative-intent hepatic-directed therapy achieved prolonged overall survival, substantially exceeding historical outcomes with systemic therapy alone. Despite frequent recurrence, the observed median OS exceeding four years supports consideration of liver-directed therapy within a multidisciplinary framework for patients demonstrating favorable disease biology and response to systemic treatment. Prospective studies are needed to better define patient selection and optimal integration of local and systemic therapies in the modern treatment era.

Objectives: To summarize recent high-quality trials and provide an update to the original 2022 American College of Radiology Appropriate Use Criteria for Treatment of Locoregional Gastric Adenocarcinoma. This multi-specialty-led committee included gastrointestinal radiation and medical oncology, gastroenterology, radiology, and surgical oncology. Methods: Using the population, intervention, comparator, outcome, timing, and study design framework, the evidence was assessed using Cochrane and PRISMA 2020 methodology. Eligible studies included prospective Phase II/II trials and select retrospective studies published between 5/1/2020 and 8/1/2025 in the Embase, Medline, and Pubmed databases. Study type and quality were assessed. Well-established RAND-UCLA consensus methodology (modified Delphi) was used to rate the appropriateness of the treatment options. Results: ARS AUC recommendations include adjuvant FLOT chemotherapy for localized gastric adenocarcinoma after neoadjuvant FLOT and surgery, poCT or perioperative chemoimmunotherapy for localized gastric adenocarcinoma, adjuvant chemotherapy +/− CRT for localized gastric adenocarcinoma that undergoes upfront surgery, and systemic therapy alone for localized gastric adenocarcinoma in a medically inoperable patient. Conclusions: This updated ARS AUC summary provides guidelines for the multimodality treatment of resectable gastric adenocarcinoma.

64 Background: The incidence of eoCRC (diagnosed at age <50) is increasing. We have previously shown higher rates of severe anxiety and depression in this group compared to adults diagnosed at age ≥50 (average-onset CRC, aoCRC). We assessed the impact of ePROM collection coupled with a CC-based supportive care intervention on symptom burden in patients with eoCRC. Methods: The Enhanced, EHR-facilitated Cancer Symptom Control (E2C2) trial was a cluster-randomized, pragmatic trial conducted at Mayo Clinic Rochester and within the Mayo Clinic Health System (in Minnesota and Wisconsin) between 2019 and 2023. Enrolled patients serially reported the severity of six SPPADE symptoms (Sleep deficit, Pain, Physical function impairment, Anxiety, Depression, and Energy deficit/fatigue) on 11-point numerical rating scales. Symptom scores were interpreted as none to mild (0-3), moderate (4-6), or severe (7-10). Those with moderate symptoms were offered symptom management education and self-management materials, while those with severe symptoms were offered the option to work with a symptom care manager. In this post hoc analysis we examined the effects of the intervention on change in symptom scores among those with eoCRC, and using the Kaplan-Meier method, the median time to improvement to a mild/moderate level in those with a severe symptom, comparing eoCRC with aoCRC. Results: Among 4,422 patients with CRC, 887 had eoCRC. Forty-six percent of patients were female ( p =0.61), and a greater proportion of eoCRC patients were of a non-white race (11.6% vs 4.7%), lived in an urban setting (65.6% vs 50.5%), had a Bachelor’s degree or higher (36.8% vs 24.4%), were employed (78.5% vs 36.7%), on a non-government insurance plan (83.4% vs 35.3%), single (19.7% vs 11.2%) or married (74.2% vs 67.9%), and had metastatic disease (54.8% vs 46.1%, p <0.0001 for all). Compared to a pre-intervention baseline, the mean change in symptom scores for eoCRC patients following the intervention demonstrated improvements in anxiety (-0.238), depression (-0.134), and fatigue (-0.146, p <0.05 for all). Time to symptom improvement from severe to mild/moderate was longer for anxiety and depression in eoCRC patients compared to aoCRC patients (Table 1). Conclusions: Collection of ePROMs coupled with a CC-based intervention led to improvements in anxiety, depression, and fatigue among eoCRC patients, although time to symptom improvement was longer among eoCRC patients compared to aoCRC patients. Time to improvement of severe symptom (months), median (95% CI). eoCRC aoCRC p -value Sleep 4.2 (3.1-5.8) 3.8 (2.7-5.4) 0.4 Pain 4.1 (3.0-7.5) 3.2 (2.3-5.0) 0.5 Physical function impairment 3.1 (1.9-6.3) 2.8 (2.2-4.2) 0.5 Anxiety 4.3 (3.4-9.1) 2.6 (2.0-3.5) 0.009 Depression 4.3 (2.6-7.5) 2.6 (2.1-3.5) 0.04 Fatigue 4.3 (3.7-9.5) 3.4 (2.7-4.2) 0.06

622 Background: Colorectal small and large cell neuroendocrine carcinoma (SCNEC and LCNEC) comprise <1% of colorectal cancers and have poor prognosis. Detailed survival information is limited due to rarity. We analyzed survival outcomes using Survival, Epidemiology, and End Result (SEER) database (2000-2022) and compared with NETs and adenocarcinoma (AC). Methods: All patients with SCNEC, LCNEC, NETs or AC with colorectal primaries were extracted. AJCC stage was calculated with 8 th ed. TNM classification. Kaplan-Meier estimator, log-rank test, and Cox regression (including AJCC stage, surgery, radiation, and chemotherapy variables) were used to analyze survival outcomes. Results: Comparing colon and rectal primaries, the only significant mOS difference was in stage IV SCNEC (2 vs 7 months, p<0.0001). Comparing LCNEC and SCNEC, the only significant mOS difference was in stage IV colon SCNEC and LCNEC (2 vs 5 months, p<0.0001). Stage III and IV colorectal SCNEC and LCNEC had significantly shorter mOS than stage III and IV AC and NETs. In multivariate Cox regressions, surgery and chemotherapy were associated with significantly improved survival for both SCNEC and LCNEC at both primaries (p<0.001), while radiation was never significant. Conclusions: Most colorectal SCNEC and LCNEC patients present with stage IV disease. Stage IV colon SCNEC had significantly shorter mOS than both stage IV colon LCNEC and stage IV rectal SCNEC. Further research should explore early diagnosis and effective therapies for this rare, aggressive malignancy. Median overall survival (months) Colon SCNEC (n/mOS) Colon LCNEC (n/mOS) Colon NETs (n/mOS) Colon AC (n/mOS) SCNEC vs NETs in colon: P-value LCNEC vs NETs in colon: P-value SCNEC vs AC in colon: P-value LCNEC vs AC in colon: P-value SCNEC vs LCNEC in colon: P-value Rectal SCNEC (n/mOS) Rectal LCNEC (n/mOS) Rectal NETs (n/mOS) Rectal AC (n/mOS) SCNEC vs NETs in rectum: P-value LCNEC vs NETs in rectum: P-value SCNEC vs AC in rectum P-value LCNEC vs AC in rectum SCNEC vs LCNEC in rectum: P-value Colon SCNEC vs rectal SCNEC P-value Colon LCNEC vs rectal LCNEC: P-value Stage I 19/32 17/122 4496/NA 114,213/144 <0.0001 <0.0001 <0.0001 0.34 0.25 27/23 10/28 5185/NA 54,702/160 <0.0001 <0.0001 <0.0001 0.00047 0.88 0.99 0.24 Stage II 17/38 30/49 259/NA 121,757/106 0.0015 <0.0001 0.4 0.0068 0.65 17/16 3/NA 33/NA 38,417/106 <0.0001 0.57 <0.0001 0.87 0.3 0.065 0.55 Stage III 40/14 107/13 582/184 104,793/82 <0.0001 <0.0001 <0.0001 <0.0001 0.48 50/19 37/17 103/102 46,687/107 <0.0001 <0.0001 <0.0001 <0.0001 0.45 0.48 0.47 Stage IV 245/2 217/5 393/49 83,542/13 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 260/7 57/6 158/36 36,652/17 <0.0001 <0.0001 <0.0001 <0.0001 0.25 <0.0001 0.95

561 Background: For patients with aHCC, second-line treatment options are not well established. The primary aim of this study was to evaluate the efficacy of Futi, an FGFR 1-4 inhibitor, in combination with Pem as a second line therapy in patients with FGF19-expressing aHCC. Methods: Patients aged ≥ 18 years, with histological/cytological diagnosis or radiological diagnosis of HCC, Child-Pugh (CP) class A-B7, Barcelona Clinic Liver Cancer (BCLC) A, B or C, ECOG performance status of 0-2, and FGF19+, by in-situ hybridization or immunohistochemistry, received Futi 20 mg once daily plus i.v. Pem 200mg every 3 weeks. The primary endpoint was progression-free survival at 6 months (PFS6). Secondary endpoints include overall survival (OS). The study was designed to enroll 22 evaluable patients to detect PFS6 from 25% (historical control) to 47%, assuming a 1-sided alpha of 0.1 with 80% power. All patients provided written consent. Eligible patients with consent who started treatment are considered evaluable. Results: The study was closed to enrollment early at the request of funding agency due to poor accrual. Between 8/2021 and 3/2024, 14 patients enrolled with 1 patient withdrawing consent before beginning protocol treatment. Among 13 evaluable patients, median age was 72 (range: 60-87) years, 7 (53.8%) were male, 12 CP A (92.3%), 7 BCLC C (53.8%), with 7 (53.8%) having received prior liver directed therapy. Median PFS was 15 weeks (95% confidence interval [CI]: 9.0-Not Estimable) and PFS6 was 23.1% (95% CI: 8.6%, 62.3%). Objective response rate was 0% (95% CI: 0%, 24.71%) and stable disease occurred in 7 (53.9%) of patients. Median OS was 55.7 weeks (95% CI: 47.3-NE). Nine (69.2%) patients experienced at least 1 grade 3+ adverse event (AE), and 2 (15.4%) patients experienced at least a grade 4+ AE. One patient had a grade 5 AE unrelated to study treatment (acute kidney injury). Most commonly reported AEs (occurring in more than 31% of patients) were hyperphosphatemia (76.9%), fatigue (69.3%), diarrhea (61.6%), AST (53.9%) and ALT (46.2%), dry eye (38.5%), nail changes (38.5%), rash-maculo-papular (38.5%), and mucositis oral (38.5%). Conclusions: This prospective study evaluated the use of a targeted treatment in combination with immunotherapy following first line immune-based therapy. This combination had modest activity as a second line therapy for patients with aHCC but did not meet the primary endpoint. The safety profile was acceptable and consistent with prior reports. An optimal second-line regimen remains a critical need. Clinical trial information: NCT04828486 .

451 Background: In first-line E/GEJ/G adenocarcinoma, anti-PD-1 + chemotherapy improves outcomes, yet many patients do not respond. Beyond PD-L1 CPS, tumor-intrinsic biomarkers of primary resistance remain limited. We applied a validated IFN-γ–linked T cell-inflamed (Tinfl) framework to define exclusionary programs and nominate targets. Methods: Bulk RNAseq was analyzed from TCGA ESCA/STAD adenocarcinoma (n=79, discovery) and a de novo Mayo cohort of pretreatment samples (n=74, validation) with harmonized survival and response data. Tinfl scores (160-gene signature) defined quartiles (Q1 cold, Q4 hot). Cold versus hot tumors were compared by edgeR + limma-voom to identify differentially expressed genes (DEGs) (Bioconductor v3.20). Pathway and upstream-regulator/causal-network analyses used Ingenuity Pathway Analysis seeded by top DEGs (|log2FC|>1.5, p<0.05). Survival was assessed by Cox proportional hazards models (two-sided α=0.05) and KM log-rank testing (α=0.05). A 14-gene Tinfl-weighted cold signature produced a per-sample composite score. Results: In the Mayo cohort, median age was 62 years; 72% were stage IV at diagnosis; 24% were HER2-positive and 3% MSI-H; primary sites were GEJ/esophagus/gastric 50%/34%/16%; PD-L1 CPS <1/1–5/6–49/≥50 occurred in 18%/46%/27%/4%; patients received a mean of 2.4 lines of therapy. Across discovery and validation cohorts, Tinfl-cold tumors showed significantly higher expression of developmental and lineage transcription factors (HOXC12, HOXC13, ZIC1, DLX3, EVX1), an oncofetal post-transcriptional hub (IGF2BP1), cancer-testis and piRNA components (MAGEA4, TDRD12, THEG), and neuronal or transport genes (GABRQ, KCNF1, LDLRAD1, SLC6A10P) (logFC >2, p<0.05). Pathway analysis highlighted chromatin and lineage regulators (KAT6A, SIRT1, LEF1), TGF-β signaling (TGFBR2), and repressors of innate DNA/IFN sensing (TREX1, IRF2BP2) (Z-score >2, p<0.05). The 14-gene Tinfl-weighted cold signature was higher in cold versus hot tumors (p<0.05) and showed a directional association with inferior overall survival in the Mayo validation cohort. Among patients who received first-line anti-PD-1–containing regimens, several signature members and the composite score were higher in non-responders (p<0.05). Conclusions: Using a discovery–validation design, we define tumor-intrinsic expression pathways that characterize Tinfl-cold E/GEJ/G adenocarcinoma. We also derive a concise 14-gene Tinfl-weighted cold signature that is enriched in non-responders and associated with inferior overall survival. Targetable nodes such as IGF2BP1 and MAGEA4, together with chromatin and Wnt/TGF-β regulators, support rational combinations with PD-1 blockade. Planned validation in a third cohort, single-cell confirmation of tumor-cell expression, and in vivo cytokine profiling in knockdown models will refine mechanisms and nominate clinical candidates.

103 Background: Clinicians spend significant time extracting and synthesizing information to risk-stratify and make adjuvant chemotherapy (AC) recommendations for CRC. By automating parts of this process, LLMs could help reduce documentation burden and variability in decision-making. Methods: We selected 144 stage III, pMMR CRC patients treated at Mayo Clinic (2017–2021). HIPAA-compliant Gemini-2.0-Flash-001 and GPT-4o extracted clinical variables (age, neuropathy grade, pathological stage) from EHRs and generated AC recommendations. Two approaches were compared: (1) a rule-based algorithm with nested if/else statements translating variables into NCCN-based recommendations, and (2) a dynamic LLM agent generating recommendations using a structured prompt informed by NCCN guidelines. Concordance with treating physicians was assessed, with disagreements reviewed by a blinded GI oncologist. Results: The dataset included 144 patients with median age 67 (IQR 66–68), 48.6% female. The LLM pipeline extracted neuropathy grade with 100% accuracy and pathological T/N stages with 99.3% accuracy. Concordance with physician recommendations was 93.1% for the rule-based algorithm and 62.5% for the LLM agent. In discordant cases, blinded adjudication by a GI oncologist favored the algorithm and LLM agent over physicians in most cases (Table 1). Conclusions: This exploratory work shows LLM agents can reliably extract variables and generate evidence-based treatment recommendations for stage III CRC. The high rate of expert preference for LLM recommendations in discordant cases is intriguing but limited by small sample size and possible adjudicator bias. Discordant physician recommendations reflect the complexity of decision-making, including patient preferences, performance status, comorbidities, and subtle findings not fully captured in EHRs. Rather than replacing judgment, we envision LLMs as decision support tools—analogous to an advanced trainee reviewing charts and presenting evidence-based recommendations to the attending physician with ultimate authority. Larger multi-institutional studies with multiple adjudicators are needed to validate these findings and clarify AI’s role in oncology. Patient characteristics, LLM extraction accuracy, and concordance with physician recommendations. Metric Result Total patients 144 Median age (IQR) 67 (66-68) Female (%) 48.6% LLM extraction accuracy – Neuropathy grade 100% LLM extraction accuracy – Pathological T and N stages 99.3% Primary Concordance – Rule-based algorithm vs. physician 93.1% Adjudicator preference of rule-based algorithm over treating physician for discordant cases 100% Concordance – LLM agent vs. physician 62.5% Adjudicator preference of LLM agent over treating physician for discordant cases 82.0%