Citations

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%

PURPOSE Immunotherapy has demonstrated limited efficacy in the treatment of microsatellite-stable (MSS) metastatic colorectal cancer (mCRC) to date. We evaluated whether an off-the-shelf, multipeptide vaccine, PolyPEPI1018, in conjunction with standard chemotherapy can induce broad antitumor immune responses, potentially improving disease control in refractory mCRC. PATIENTS AND METHODS This single-arm, open-label study enrolled adults with MSS mCRC who had progressed after ≤2 previous lines of therapy for metastatic disease. PolyPEPI1018 was administered subcutaneously to four anatomical sites on days 1 and 15 and TAS-102 orally twice a day on days 1-5 and 8-12 of a 28-day cycle. Patients were treated for seven cycles or until disease progression. The primary end point was safety. Secondary end points were overall response rate, progression-free survival (PFS), overall survival (OS), and immune and genetic correlates. RESULTS Fifteen patients were enrolled; 80% had liver metastases and 67% had KRAS mutations. Adverse events (AEs) related to PolyPEPI1018 were grade 1 and 2 local skin reactions (n = 14). Grade 3 AEs (n = 15) at least possibly related to treatment were fatigue, decreased lymphocyte and neutrophil counts, and diarrhea. None had objective tumor response; the disease control rate was 53.3%. The median PFS (mPFS) was 4.0 (95% CI, 2.2 to 6.1) months and the median OS (mOS) was 8.7 (95% CI, 6.4 to 15.2) months. In post hoc subgroup analyses, patients without active liver metastasis had improved mPFS of 9.7 months and mOS of 14.8 months. In addition, patients who were predicted to have broader immune responses to vaccine based on human leukocyte antigen genotype had longer mOS (10.3 v 4.6 months, hazard ratio, 0.29 [95% CI, 0.07 to 1.17]). CONCLUSION The addition of PolyPEPI1018 to TAS-102 was safe and induced immunological responses that were associated with improved clinical outcomes. Further investigation of this chemo-immunotherapeutic is warranted.

426 Background: Testing for DPYD gene variations prior to administration of fluoropyrimidines ([FP] e.g. 5-FU or capecitabine) can determine if patients are at risk of significant toxicity and may benefit from dose reduction or elimination of chemotherapy. Barriers to testing include lack of recommendations in US guidelines for routine DPYD testing, absence of insurance coverage, and limitations in provider support. We designed an electronic health record (EHR)-based intervention to support shared decision-making (SDM) between providers and patients regarding DPYD testing. Methods: We built an EHR alert identifying all patients initiating FP without DPYD testing. The alert fired to remind providers upon signing FP. Clinical decision support (CDS) was included in the alert to help with DPYD testing selection along with an educational note template containing the incidence of DPYD deficiency, rationale for testing, test costs, and turnaround time. Interactions with the EHR were recorded to determine why providers declined ordering pharmacogenomic testing. Testing rates and results for patients initiating FP were collected from the EHR. Finally, providers were surveyed to determine the impact of the alert on conversations with patients as well as the utility of the embedded CDS and educational note template. Results: Within the first 8 months after alert activation, 34 patients completed DPYD testing of 444 starting FP (testing rate of 7.7%) vs the baseline testing rate in the 3 months prior to alert go-live of 6.8% (p = 0.86 via chi-squared analysis, two-tailed). Of the 34 patients, 5 were intermediate DPYD metabolizers. Reasons for alert dismissal per provider selection were: “Patient already tolerated prior fluoropyrimidines” (15.7%), “Patient declines” (12.9%), “Will order separately” (5.7%), “Not discussed as not appropriate” (5.3%), and “Not discussed, deferred” (60.4%). Thirty-nine providers (n = 77 eligible) responded to the survey; 22 prescribed FP in their practice and had encountered the alert. 21 (95.5%) of these reported the alert was “Very Useful” or “Somewhat Useful” and 18 (81.8%) used the DPYD testing options available in the alert. Sixteen (72.7%) discussed DPYD testing more frequently with patients as a result of the alert. Barriers to DPYD testing were “Concerned about the impact of lowering FP dosing” and “Lack of knowledge about DPYD testing”, while “Cost of DPYD testing” was scored lowest. Conclusions: Utilizing an EHR alert with embedded CDS had a positive impact on provider awareness and led to more discussions regarding DPYD pharmacogenomic testing with patients in line with current FDA recommendations. Limitations in provider knowledge regarding DPYD testing and concerns about lowering chemotherapy doses posed barriers to ordering testing. Further interventions including additional education are required.

Abstract Background Unresectable hepatocellular carcinoma (uHCC) is a leading cause of cancer death. FDA-approved first-line systemic therapies include atezolizumab/bevacizumab (atezo/bev) and durvalumab/tremelimumab (durva/treme); however, there is a lack of comparative data. Methods We reviewed outcomes of patients with uHCC who initiated atezo/bev or durva/treme between 2017 and 2024, across six institutions. Overall survival (OS) and time to treatment discontinuation (TTD) were analyzed using the Kaplan–Meier and Cox models, adjusting for baseline characteristics. Results Four hundred fifty-two uHCC pts were included. Median age: 68 years; 77% male; 81% white. Most common etiologies were viral hepatitis (38.9%) and metabolic dysfunction-associated steatohepatitis (19.5%). Disease progression was the primary reason for treatment discontinuation, atezo/bev (56%) and durva/treme (42%). Outcomes were not statistically significant (median OS [month, m]: 14.0 vs 14.6 [P = .66]; median TTD [m]: 4.9 vs 3.9 [P = .42] for atezo/bev vs durva/treme). Outcomes were significantly different between Child-Pugh classes (CP: A, B7, B8/9, C) respectively, median OS(m): 19.0, 6.1, 5.1, 2.0 (P < .001); median TTD(m): 6.1, 2.3, 3.0, 1.3 (P < .001). Conclusions In this real-world study of uHCC, no significant difference in clinical outcomes was observed between atezo/bev and durva/treme in the first-line setting. CP scores were a key prognostic variable with both regimens.

This cross-sectional study analyzes patient-reported outcome findings from the Enhanced, Electronic Health Record–Facilitated Cancer Symptom Control (E2C2) trial to investigate level of symptom burden in adolescents and young adults with cancer.

PURPOSE Colorectal cancer (CRC) incidence and mortality have decreased since the 1970s, but the incidence in young adults (<50 years, named early-onset CRC [eoCRC]) has been increasing. PATIENTS AND METHODS Individual patient data on 13,365 patients with metastatic CRC enrolled between 2000 and 2012 in 17 first-line randomized trials in the Aide et Recherche en Cancérologie Digestive database were pooled. The distribution of demographics, clinicopathologic features, biomarkers, and outcome data were summarized and compared by age groups. Progression-free survival (PFS) and overall survival (OS) were assessed by Kaplan-Meier curves and Cox models stratified by treatment arms within studies, adjusting for potential confounders. Predictive value of age group on clinical outcomes was evaluated by testing interaction effect between treatment and age variables. RESULTS Overall, 2,045 patients with eoCRC (median age, 42.5) and 11,320 patients with average-onset CRC (aoCRC; median age, 63.8) were included in this analysis. Within the eoCRC population, treatment with bevacizumab in addition to chemotherapy improved PFS (9.9 v 6.8 months; hazard ratio [HR], 0.66 [95% CI, 0.54 to 0.80]; P < .0001), which was similar to the findings in aoCRC population (9.4 v 7.3 months; HR, 0.73 [95% CI, 0.67 to 0.80]; P < .001; interaction P = .5415). However, epithelial growth factor receptor inhibitor (EGFRi) did not improve PFS in RAS wild-type (WT) patients with eoCRC who had left-sided primary tumors (8.3 v 8.9 months; HR, 1.20 [95% CI, 0.81 to 1.77]; P = .36), whereas EGFRi significantly improved PFS in the aoCRC population (9.9 v 8.5 months; HR, 0.74 [95% CI, 0.64 to 0.86]; P < .0001; interaction P = .083). CONCLUSION Treatment-naïve patients with metastatic eoCRC appear to derive similar benefit from bevacizumab as patients with aoCRC. However, patients with eoCRC with left-sided RAS/RAF WT tumors did not appear to derive benefit from first-line EGFRi.

Objectives: Esophageal cancer (EC) often presents with dysphagia due to tumor obstruction. Esophageal stenting has the potential of palliating dysphagia, improving nutrition, preventing aspiration, and improving quality of life (QoL) but may be associated with risks. The present systematic review and guidelines are intended to assist treatment decision-making when considering stent placement in patients with EC based on the available evidence. 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-III trials and retrospective analyses published between January 1, 2010 and December 3, 2024 in the Ovid Medline database. These references were assessed by American Radium Society (ARS) Appropriate Use Criteria (AUC) methodology. RAND-UCLA consensus methodology was used to rate the appropriateness of the use of stents. Results: ARS AUC recommendations include (1) esophageal stenting is usually not appropriate in patients with early-stage EC in whom upfront surgery is planned; (2) esophageal stenting is usually not appropriate in patients with locally-advanced EC in whom neoadjuvant/perioperative therapy and esophagectomy or definitive chemoradiation is planned; (3) esophageal stenting may be appropriate in the setting of metastatic EC, especially in patients with short life expectancy with limited treatment options; (4) esophageal stenting is usually not appropriate for benign stricture following curative-intent therapy; (5) esophageal stenting is usually not appropriate for locally recurrent tumor in the setting of prior radiation; and (6) esophageal stenting is usually appropriate for management of tracheoesophageal fistula before curative-intent treatment. Conclusions: This ARS AUC summary provides guidelines for the use of esophageal stents in patients with EC provides based on available evidence.

3520 Background: Recent analyses highlight nonhierarchical outcomes using the 8 th Edition AJCC staging system for CC. For instance, the 5-year survival rate for stages I and IIIa patients (pts) closely align. Additionally, tumor deposits (TDs) have been established as significant prognostic indicators. The AJCCCCEP commissioned this study to develop an updated pathological staging system for CC focused specifically on pts without distant metastasis (M0), while retaining the existing stage IV classification. Methods: Individual patient data (IPD) from pts diagnosed with cc (2010- 2017) in the NCDB were divided into training (70%) and internal validation (30%) datasets. External validation used IPD from clinical trials. The primary endpoint was overall survival (OS). Risk classification development for M0 pts incorporated ungrouped data on pathologic T categories, the number of involved regional lymph nodes (LN+), and TD counts. Recursive partitioning and regression tree analyses were applied to construct hierarchical staging levels. Pre-specified criteria required survival probabilities to be consecutive and show clear separations using Kaplan-Meier (KM) estimates with pairwise log-rank test P of < 0.005 for the training and < 0.05 for validation analyses. Results: Data from 281,997 pts (median age 67 years, 50% male, 81% white, 55% T3, 19% T4, 44% N+, 26% M+, and 11% with ≥1 TD) were analyzed, with a median follow-up of 7.3 years. The updated staging system (Table) met pre-specified criteria, with all observed pairwise P < 0.0001 in the development and internal validation sets. KM OS curves displayed a hierarchical separation across all sub-levels after the 1 st year of diagnosis. Consistent results were seen in pts treated with adjuvant chemotherapy in 4 trials (all pair-wise P < 0.0001). Conclusions: The proposed pathological staging system for M0 pts fulfills pre-specified criteria for hierarchical risk stratification, validated both internally and externally, and provides an evidence-based update. Pending review process, the AJCCCCEP will recommend that these changes be made to the Version 9 staging protocol for colon cancer to improve prognostication for CC pts. Stage T, # of LN+, # of TD M % of pts 1y OS (CI), % 3y OS (CI), % 5y OS (CI), % I T1, 0, 0 0 5 96 (95-97) 91 (90-92) 84 (83-86) IIa T2, 0, 0 0 10 95 (94-95) 88 (88-89) 80 (79-81) IIb T1, 0, 1+T1, 1+, 0T2, 0, 1+T2, 1-4, 0T3, 0, 0 0 27 93 (92-93) 84 (84-85) 75 (75-76) IIIa T1, 1+, 1+T2, 1-4, 1+T2, 5+, 0T3, 0, 1+T3, 1-4, 0 0 14 92 (91-92) 80 (80-81) 71 (70-72) IIIb T2, 5+, 1+T3, 1-4, 1+T3, 5+, 0T4a, 0-4, 0T4b, 0-2, 0 0 13 86 (86-87) 69 (68-70) 58 (57-59) IIIc T3, 5+, 1+T4a, 0-4, 1+T4a, 5+, anyT4b, 0-2, 1+T4b, 3+, any 0 5 78 (77-80) 53 (51-54) 40 (38-41) IVa Any 1a 19 59 (58-60) 28 (28-29) 17 (16-18) IVb Any 1b 7 43 (42-44) 14 (13-14) 6 (6-7) CI: 95% confidence internal; Peritoneum involvement data were not available before 2018 in NCDB. Thus, IVa/b were based on 7 th Edition.

e15682 Background: Small cell carcinoma (SCC) of gastrointestinal tract is a highly aggressive rare malignancy with an estimated prevalence of 0.1% to 1% of all GI neoplasms. Treatment and survival outcomes data are very limited. Here we present an analysis of outcomes of patients with intestinal SCC using data from the Survival, Epidemiology, and End Result (SEER) database. Methods: SEER (2000-2020) was used to identify patients with SCC involving of the small intestine, colon and rectum. Variables included age, race, site, stage, chemotherapy (chemo), radiation therapy (RT) and surgery. Cox regression was used to model survival outcomes based on site, stage, and treatment received in the overall group and stage subgroups. Results: 1,169 patients with SCC of the small intestine, colon, and rectum were identified. Median age was 65 (24-90). 631 (54%) were male; majority of patients were white (989; 85%), followed by black (128; 11%) and other races (50; 4.3%). Colon (569; 49%) and rectum (514; 44%) were the predominant sites while small intestine only comprised 7.4% of the patients. Distant, regional and local diseases were reported in 55%, 17% and 7.5% of the patients. Median overall survival (mOS) of the entire group was 8 months (m). Primary tumor sites of small intestine and rectum were significantly associated with lower mortality rate in comparison to colon (11 m and 9 m vs 5 m, p < 0.001 for both). Patients with localized disease had the lowest mortality rate when compared with those with regional or distant metastatic diseases (38 m vs. 16 m and 5 m, p < 0.001 for both). Receiving surgery (HR: 0.458, p < 0.001), chemo (HR: 0.473, p < 0.001), and RT (HR: 0.733, p = 0.003) were all significantly associated with lower mortality rate (Table). These differences remained significant in each stage-based subgroup except for chemotherapy in localized disease subgroup, which encompassed only 87 patients. Conclusions: Intestinal SCC has poor prognosis, with SCC of the colon being most prevalent and with the highest mortality. Majority of patients presented with distant metastatic disease which led to high mortality. Multimodality treatment may improve survival for these patients. Number of cases (%) HR (CI) p-value mOS (months) vs reference Site (ref: colon) Rectum 514 (44%) 0.750 (0.648-0.867) <0.001 9 (ref: 5) Small intestine 86 (7.4%) 0.574 (0.44-0.749) <0.001 11 (ref: 5) Stage (ref: localized) Regional 201 (17%) 2.087 (1.507-2.891) <0.001 16 (ref: 38) Distant 639 (55%) 4.222 (3.110-5.733) <0.001 5 (ref: 38) Unknown/unstaged 242 (21%) 1.639 (1.160-2.315) 0.005 8 (ref: 38) Chemo (ref: none/unknown) Yes 590 (50%) 0.473 (0.404-0.553) <0.001 10 (ref:3) RT (ref: none/unknown) Yes 175 (15%) 0.733 (0.598-0.897) 0.003 12 (ref: 7) Surgery (ref: no surgery) Yes 363 (31%) 0.458 (0.385-0.546) <0.001 12 (ref: 6) Unknown 138 (12%) 0.717 (0.549-0.936) 0.01 7.5 (ref: 6)

Abstract Background Optimal treatment for resectable gastric and gastroesophageal junction (GEJ) adenocarcinoma remains unclear due to limited head-to-head comparisons among chemotherapy and chemoradiation regimens. This network meta-analysis aimed to determine the relative efficacy of available multimodality treatment regimens in these patients. Methods MEDLINE, EMBASE, Scopus, Web of Science, and CENTRAL were searched till September 20, 2024. Phase 3 randomized trials evaluating perioperative/neoadjuvant systemic therapy ± radiation in resectable gastric/GEJ adenocarcinoma were included. Primary outcomes were disease-free survival (DFS) and overall survival (OS). Frequentist network meta-analysis was conducted to compare hazard ratios (HRs) and 95% confidence intervals (CIs). Results Fifteen trials (8072 patients) were analyzed. pFLOT (perioperative fluorouracil, leucovorin, oxaliplatin, and docetaxel) ranked highest for DFS in the overall population (P-score 91%), achieving significant improvements compared to surgery alone (HR 0.48, 95% CI, 0.38-0.60) and nCROSS (neoadjuvant paclitaxel, carboplatin, and radiotherapy) (0.67, 0.56-0.81). For OS, nPLF + nCRT(EP) (neoadjuvant cisplatin, leucovorin, and fluorouracil for induction, followed by etoposide, cisplatin, and radiotherapy) (P-score 90%) and pFLOT (P-score 87%) were the top regimens. pFLOT significantly outperformed surgery alone (0.57, 0.45-0.72) and nCROSS (0.73, 0.60-0.89). Based on the limited data available, adding neoadjuvant chemoradiation to pFLOT provided no additional OS (1.14, 0.76-1.72) or DFS (1.22, 0.83-1.82) benefit. Similarly, adding pembrolizumab to perioperative chemotherapy (cisplatin and fluorouracil/capecitabine) was not superior to pFLOT (DFS: 1.10, 0.71-1.69; OS: 1.09, 0.69-1.72). Conclusions pFLOT demonstrated superior efficacy in resectable gastric/GEJ adenocarcinoma, outperforming surgery alone, nCROSS, and alternative perioperative regimens. The additive role of immunotherapy requires further investigation to optimize patient selection and outcomes.