Ahn, Daniel H.

837 Background: A recent randomized trial suggested that CR may influence ICI activity in patients with lung cancer. However, the same has not been established in GI cancers, where reports have shown mixed results with variable definitions of 'early' vs 'late' IT. We evaluated multiple approaches to assess the impact of CR on first-line ICI in GI malignancies. Methods: This retrospective study included patients from three Mayo Clinic sites (N=455). Both previously reported IT cutoffs and a new cutoff derived from ROC analysis were evaluated. Patients were categorized by the median of all ITs, the median of the first four ITs, and whether ≥20% of ITs occurred before the identified cutoff. Univariate and multivariate Cox regression (adjusting for age, MSI status, type of cancer, metastasis status, number of cycles, and receipt of chemotherapy with ICI) assessed associations with OS. KM analyses were performed to illustrate survival differences. The primary endpoint was OS. Results: When patients were categorized based on whether ≥20% of infusions occurred before 14:00 (ROC-derived cutoff), a significant association with OS was observed. On univariate analysis, early infusion timing was associated with improved survival (HR 2.30, 95% CI 1.58–3.32, p <0.001). This remained significant after multivariable adjustment (HR 2.11, 95% CI 1.41–3.16, p < 0.001). Median OS was 14.2 months in the early group vs 3.8 months in the late group (log-rank p < 0.001). Using previously defined cutoffs, categorization by the median IT of all infusions or the median of the first four infusions did not demonstrate a significant association with OS in either univariate or multivariate Cox regression analyses. Conclusions: These findings highlight the potential role of infusion timing as a chronotherapy factor in GI cancers. Given the retrospective design and heterogeneous population, prospective studies are warranted to validate this signal and clarify its clinical relevance. Baseline demographics of patients by infusion timing (early vs late). Variables Early N=412 (%) LateN=43(%) Total N=455(%) p-value Age <70 249(60.4) 21(48.8) 270 (59.3) 0.1 >70 163(39.6) 22(51.2) 185 (40.7) Gender F 118(28.6) 16(37.2) 134 (29.5) 0.3 M 294(71.4) 27(62.8) 321 (70.5) Cancer Type Gastroesophageal 109 (26.5) 8 (18.6) 117 (25.7) 0.1 Hepatocellular 159 (38.6) 24 (55.8) 183 (40.2) Biliary 86 (20.9) 4 (9.3) 90 (19.8) MSI-H 58 (14.1) 7 (16.3) 65 (14.3) Metastasis No 150 (36.4) 22 (51.2) 172 (37.8) 0.1 Yes 262 (63.6) 21 (48.8) 283 (62.2) Immunotherapy Cycle Number <6 189 (45.9) 33 (76.7) 222 (48.8) <0.001 >6 223 (54.1) 10 (23.3) 233 (51.2)

776 Background: Surgery is the only potentially curative option for PC. However, only a minority of patients (pts) undergo resection with current perioperative (periop)-chemotherapy (CT). In the absence of phase III trials, selection between mFOLFIRINOX and gemcitabine/nab-paclitaxel (gem-nab) is based on limited evidence. We assessed whether NGS–based tumor profiling can guide tailoring of CT. Methods: pts with resected PC (resectable, borderline resectable, and downstaged locally advanced) that underwent Tempus xT, xF (DNA), and/or xR (RNA) testing were selected (89% tissue, 11% blood). Most samples (73%) were collected post-CT. Samples were stratified by KRAS status into KRAS -mutated ( KRAS mut) and KRAS wild-type ( KRAS wt). PD-L1 IHC was reported as tumor proportion score (TPS). Primary endpoint was median overall survival (mOS) defined from CT initiation to death with censoring of pts alive at the last follow-up or at a study cutoff of 5 years. Secondary endpoint was molecular profiling by KRAS status. Pearson chi-square and Wilcoxon rank-sum tests were used for descriptive comparisons; univariate Cox regression and log-rank tests (p<0.05) for survival. Results: We included 1,325 pts (median age 66 years [IQR 59–72]) with resected PC [stage II (15%), stage III (12%) and stage I (10%), unknown (62%)]. 95% had adenocarcinoma. Pts received neoadjuvant (50%), neoadjuvant plus adjuvant (26%), neoadjuvant with indeterminate adjuvant allocation (13%) or adjuvant CT (11%). First-line periop-CT was mFOLFIRINOX in 73% and gem-nab in 27% (median duration 4.3 months (mo) [range, 2.6–6.7]). KRAS mutations were present in 75% (G12X 91%: G12D 44%, G12V 34%, G12R 21%, G12C <1%). mOS was 26.7 mo (95%CI, 22.4-35.7) in the total cohort; KRAS mut pts showed a clinically meaningful worse mOS compared to KRAS wt [24.1 vs 35.7 mo; HR 1.61 (95%CI, 0.98 - 2.63); p=0.06]. In the KRAS G12D pts, mFOLFIRINOX showed a trend towards longer mOS compared to gem-nab [17.79 vs 14.53 mo; HR 0.75, (95% CI, 0.56 - 1); p=0.050]; no significant differences were observed in other KRAS subgroups. Compared to KRAS wt, KRAS mut pts were enriched for TP53 (76% vs 22%; p<0.001), SMAD4 (28% vs 5.8%; p<0.001), CDKN2A (33% vs 8.5%; p<0.001), MTAP deletions (7.7% vs 1.2%; p<0.001), and chromatin-regulator alterations ( ARID1A 7.8% vs 1.2%; p<0.001; KMT2C 2.5% vs 0.6%; p=0.034). KRAS wt pts were enriched for BRAF V600E (1.5% vs 0%; p<0.001). PD-L1 TPS ≥1% was more frequent in KRAS mut compared to KRAS wt (11.0% vs 8.1%; p=0.033). Conclusions: KRAS status does not predict benefit from mFOLFIRINOX vs gem-nab in resected PC. Interestingly, resected PC showed a higher prevalence of KRAS wt compared to the metastatic setting. KRAS status is associated with distinct profiles of potentially targetable co-alterations. These findings may suggest the integration of genomic profiling in clinical trials to develop biomarker-driven tailored strategies in the early stage.

619 Background: Well-differentiated G3 GEP-NETs are rare and heterogeneous diseases, limiting phase III trials and definition of clear standards of care. Building upon our preliminary cohort of 76 patients (pts), we present an updated and expanded retrospective analysis across three Mayo Clinic tertiary centers. Methods: We included pts with histologically confirmed well-differentiated G3 GEP-NETs diagnosed since 2017 (new WHO classification) from the three Mayo Clinic sites (Rochester, Phoenix, Jacksonville). Endpoints were ORR, DCR, PFS, and OS across treatment lines. PFS/OS were estimated by Kaplan-Meier. OS was defined from treatment start. Multivariate Cox regression was used to evaluate the association of clinic-pathologic features (age, sex, tumor functionality, Ki-67, primary tumor site, stage at diagnosis, presence of liver metastasis, number of metastatic sites at diagnosis) with OS (p<0.05). Results: 111 pts (median age 62 y, range 22–83). Primary sites were pancreas (58%), small bowel (19%), unknown (9%), and rectum (5%). Most tumors were G3 at diagnosis (88%); the remainder evolved from prior G1/G2. Ki-67 was 20–55% in 83% of cases. Most were nonfunctional (70%); functional subtypes included carcinoid (13%) and gastrinoma (8%). Median plasma 5-HIAA was 38.5 ng/ml (5–1394). Advanced disease was present in 85%. Common metastatic sites included liver (85%), lymph nodes (32%), bone (26%), peritoneum (10%), and lung (9%). Median number of metastatic sites was 1 (range 0–5). Median number of systemic treatment lines was 2 (range 0–8). The most frequent regimens and outcomes are shown in the Table. Insulin-secreting tumors (HR 2.16, 95% CI 1.33–25.73, p=0.019) and ≥2 metastatic sites (HR 2.13, 95% CI 1.05–4.31, p=0.035) were associated with inferior OS. Conclusions: This represents one of the largest real-world datasets of well-differentiated G3 GEP-NETs and provides a deeper evaluation of treatment strategies and prognostic factors. CAPTEM, SSAs, and PRRT with Lu-177 emerged as the most commonly used therapies, with PRRT showing the most favorable outcomes. Insulin secretion and higher metastatic burden were confirmed as adverse prognostic factors. CAPTEM SSAs PRRT Lu-177 Carboplatin plus etoposide FOLFOX Everolimus Sunitinib N of pts (%) 64 (58) 44 (40) 37 (33) 24 (19) 18 (13) 11 (10) 8 (7) Line of therapy (%) First 36 (56.3) 27 (61.4) 5 (13.5) 13 (59.2) 5 (7.2) 1 (9.0) 1 (12.5) Second 15 (23.4) 13 (29.5) 13 (35.2) 3 (13.6) 5 (35.7) 3 (27.3) 2 (37.5) Third or higher 13 (20.3) 4 (9.1) 19 (51.3) 6 (27.2) 8 (57.1) 7 (63.7) 5 (62.5) DCR (%) 60.9 56.8 75.0 52.2 71.4 45.5 37.5 ORR (%) 34.4 13.6 44.4 30.4 35.7 18.2 12.5 mPFS (95% CI) 8.0 (5.3-10.6) 6.2 (2.8-9.5) 11.2 (9.1-13.3) 7.3 (2.8-11.8) 4.4 (2.4-6.4) 3.9 (2.2-5.6) 3.9 (1.54-6.26) mOS (95% CI) 40.1 (23.0-57.2) 50 (22.6-77.4) 76.1 (16.3-135.8) 33.0 (21.0-45.0) 21.1 (15.0-27.2) NE NE

599 Background: Biliary tract cancers (BTC) remain aggressive malignancies with limited treatment options. Alterations in PIK3CA are implicated in tumor progression across solid tumors, but their prognostic impact in BTC remains unclear. We investigated the association of PIK3CA alterations(alt), alone and in combination with TP53 and ARID1A , with survival outcomes in advanced BTC. Methods: We conducted a retrospective analysis of 1,079 patients with advanced BTC treated with first-line chemotherapy (CT) or chemoimmunotherapy (CIO) across three sites in the USA, Japan, and Brazil. Genomic profiling was performed using next-generation sequencing on tissue and/or liquid biopsy samples. Survival outcomes were estimated by Kaplan-Meier analysis, with comparisons by log-rank testing and hazard ratios (HR) derived from Cox regression models. Results: PIK3CA alterations were identified in 11% (123/1,079) of patients. The median age at diagnosis was 68 yrs (range, 20–79), with 50% (62/123) female. The most common primary site was intrahepatic in 43.1% (53/123) while majority of patients received CT only 92.5% (111/123). When analyzed with TP53 , 68 pts had PIK3CA/TP53 co-alterations(co-alt), 501 had TP53 alt, 55 had PIK3CA alt alone, and 455 had neither PIK3CA nor TP53 alt (WT/WT). Overall survival (OS) was significantly worse in patients with PIK3CA alt, particularly in the setting of TP53 co-alt. Median OS (mOS) was 22.6 months(m) (95% CI, 20.4–24.8) for PIK3CA/TP53 WT, 15.9 m (95% CI, 11.6–18.5) for PIK3CA alt /TP53 WT, 14.6 m (95% CI, 13.1–19.5) for PIK3CA alt/ TP53 alt, and 17.1 m (95% CI, 15.5–18.8) for TP53alt/PIK3CA WT (p<0.001). In analyses with ARID1A , 32 pts had PIK3CA/ARID1A co-alt, 91 had ARID1A alt only, 120 had PIK3CA alt only, and 836 were WT/WT. PIK3CA/ARID1A co-alt tumors had a mOS of 12.8 m (95% CI, 9.1–NE) compared with 19.3 m (95% CI, 18.1–20.9) in PIK3CA WT /ARID1A WT (p=0.009). Multivariate analysis identified PIK3CA alt (hazard ratio [HR] 1.37, 95%CI 1.09–1.72; p=0.008), PIK3CA alt/ TP53 alt (HR 1.52 95% CI, 1.11–2.07; p=0.008), ARID1A alt/ PIK3CA WT (HR 1.40, 95% CI 1.08-1.82, p=0.011), and metastatic status (HR 1.82, 95% CI 1.52-2.19, P<0.001) as independent prognostic factors for shorter OS. Conclusions: In advanced BTC, PIK3CA alt were associated with significantly worse overall survival, particularly when co-occurring with TP53 or ARID1A alt. These findings establish PIK3CA as a prognostic biomarker of poor outcome and highlight the need to explore PI3K-targeted therapeutic strategies in BTC.

399 Background: The standard of care for resectable GEA adenocarcinomas is perioperative chemotherapy with the FLOT (fluorouracil, leucovorin, oxaliplatin, docetaxel) regimen. Recent studies investigating the addition of checkpoint inhibitors (CPIs) to perioperative chemotherapy have shown mixed results. An NMA can leverage both direct and indirect comparisons to evaluate the relative efficacies of available regimens in this setting. Methods: MEDLINE, EMBASE, Scopus, Web of Science and CENTRAL were searched until August 27th, 2025. Eligible studies included phase 3 randomized trials evaluating perioperative or neoadjuvant chemotherapy +/- CPIs ± radiation in resectable GEA. The primary outcomes were disease-free survival (DFS) and overall survival (OS). Hazard ratios and 95% confidence intervals (CIs) were estimated using a frequentist NMA framework. Results: Fifteen trials including 8072 patients were analyzed. For DFS, perioperative durvalumab + FLOT (dFLOT) ranked highest (P-score = 0.99), followed by FLOT (P-score = 0.84) (HR for dFLOT vs. FLOT was 0.71, 95% CI 0.58-0.86). dFLOT also ranked higher than perioperative pembrolizumab + cisplatin and fluorouracil/capecitabine (KN-585), though this comparison was not statistically significant (HR 0.65, 95% CI 0.40-1.04). Compared with CROSS, dFLOT showed superior DFS (HR of 0.48, 95% CI 0.36-0.63). For OS, dFLOT ranked the highest (P-score=0.95) and was superior to CROSS (HR 0.57, 95% CI 0.42-0.77). While dFLOT appeared favorable versus FLOT (HR 0.78, 95% CI 0.62-0.98), this finding was primarily driven by MATTERHORN, which did not meet its prespecified OS significance boundary (P = 0.03 vs threshold P < 0.0001), and the proportional hazards assumption was violated. Similarly, no significant OS difference was observed between dFLOT and the KN-585 arms (HR 0.75, 95% CI 0.45-1.23). Conclusions: In this NMA, perioperative FLOT remained the benchmark regimen for resectable GEA. dFLOT has shown promising early results, ranking the highest for DFS and showing favorable, but not statistically definitive OS trends in comparison to FLOT. Importantly, the network enabled indirect comparisons not tested in head-to-head trials, demonstrating that dFLOT ranked above pembrolizumab+CF/CX (KN-585) without reaching statistical significance, and also outperformed CROSS. However, these findings should be interpreted with caution, as they are primarily driven MATTERHORN data in which OS did not cross the prespecified threshold and proportional hazards assumptions were violated. Perioperative immunotherapy combinations appear promising, but final MATTERHORN results and further confirmatory studies are needed before dFLOT can be considered as a new standard of care.

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%