Khorana, Alok A.

e15509 Background: Large language models (LLMs) are increasingly explored for oncology decision support, yet comparative evidence evaluating medical-specific versus general-purpose LLMs using expert-reviewed clinical cases is limited. We compared a medical-specific LLM (Google MedGemma 27B) with a general-purpose LLM (ChatGPT-5) using gastrointestinal (GI) and genitourinary (GU) oncology vignettes. Methods: This cross-sectional study evaluated 32 de-identified oncology vignettes (26 GI, 6 GU) created by a U.S. board-certified medical oncologist and informed by real-world cases. Vignettes included text-only information on the clinical context, description of pathology, tumor staging, biomarkers, and relevant comorbidities. Both LLMs were prompted to provide National Comprehensive Cancer Network(NCCN) guideline-based treatment recommendations, monitoring strategies, maintenance therapy considerations, and references, along with a self-reported confidence score (0–100). Board-certified U.S. medical oncologists independently reviewed outputs using a structured instrument; two disease specific expert oncologists evaluated each case. Outcomes included correctness and comprehensiveness (1–10 scales), hallucinations (factually fabricated or unsupported statements), NCCN traceability, potential patient harm, and clinician trust. Associations between accuracy and comprehensiveness were assessed using correlation coefficients. Inter-rater reliability was assessed using a two-way random-effects intraclass correlation coefficient (ICC). Results: ChatGPT-5 demonstrated higher mean correctness (8.65 vs 4.68) and comprehensiveness (8.68 vs 5.22) scores compared with MedGemma 27B. Accuracy and comprehensiveness were strongly correlated overall (Pearson r = 0.93, p < 0.001), with significant model-specific correlations (ChatGPT-5 r = 0.86, p < 0.001; MedGemma r = 0.77, p < 0.001). Hallucinations occurred less frequently with ChatGPT-5 (12.3% vs 65%), and NCCN-traceable recommendations were more common (75.4% vs 36.7%). Responses judged as potentially harmful occurred in 18.5% of ChatGPT-5 outputs versus 70% of MedGemma outputs. Clinician trust favored ChatGPT-5 (84.6% vs 15%). Inter-rater reliability improved with rater averaging (ICC(A,k): correctness 0.72; comprehensiveness 0.64). Conclusions: In expert-reviewed GI and GU oncology cases, ChatGPT-5 demonstrated higher clinical accuracy, comprehensiveness, and guideline alignment than a medical-specific LLM; however, a substantial rate of hallucinations and potentially harmful recommendations persisted across models. These findings underscore that, despite performance differences, current LLMs require rigorous clinician oversight and should not be used as standalone decision-support tools in oncology.

Abstract Sex differences in cancer outcome, including that of glioblastoma (GBM), are shaped by biological, hormonal, and immunological factors and impact disease progression, treatment responses, survival, and tumor microenvironment (TME) interactions. Platelets regulate the immune responses and tumor progression in many cancers, but it is not clear how they contribute to these sex-based differences by affecting the dynamics of the TME. Here, we show that GBM patients exhibit heightened platelet reactivity driven by PAR4 signaling. In murine GBM models, both pharmacological inhibition of PAR4 using BMS986120 and genetic deletion of PAR4 significantly prolong survival in females but not males. This survival advantage is estrogen dependent: it is preserved in chromosomal male-hormonal female mice within the four-core genotype model and is rescued in ovariectomized mice treated with estrogen. The survival benefit is TME specific and is mediated by platelet-driven enhancement of CD8⁺ T cell infiltration into the tumor. Inhibition of platelet PAR4 signaling increases calcium signaling through an estrogen-dependent interaction between PAR4 and estrogen receptor β (ERβ)—a receptor interaction not previously described. PAR4-activated platelets within the TME suppress CD8⁺ T cell function, and depletion of CD8⁺ T cells abolishes both the tumor-induced platelet reactivity and the survival benefit conferred by PAR4 inhibition. These findings establish platelet-mediated PAR4 signaling as a critical driver of tumor progression and identify sex-specific immune responses as key to therapeutic efficacy.

Importance Neoadjuvant therapy (NT) is an increasingly used treatment strategy for patients with localized pancreatic ductal adenocarcinoma (PDAC). Little research has been conducted on cancer care delivery during NT, and the standards for optimal delivery of NT have not been defined. Objective To develop consensus best practices for delivering NT to patients with localized PDAC. Design, Setting, and Participants This study used a modified Delphi approach consisting of 2 rounds of voting, and a series of virtual conferences (from October to December 2023) to reach expert consensus on candidate best practice statements generated from a systematic review of the literature and expert opinion. An interdisciplinary panel was formed including 47 North American experts from surgical, medical, and radiation oncology, radiology, pathology, gastroenterology, integrative oncology, anesthesia, pharmacy, nursing, cancer care delivery research, and nutrition as well as patient and caregiver stakeholders. Main Outcome and Measures Statements that reached 75% agreement or greater were included in final consensus statements. Results Of the 47 participating panel members, 27 (57.64%) were male, and the mean (SD) age was 47.6 (8.2) years. Physicians reported completing training a mean (SD) 14.6 (8.6) years prior and seeing a mean (SD) 110.6 (38.4) patients with PDAC annually; 35 (77.7%) were in academic practice. Final consensus was reached on 82 best practices for delivering NT. Of these, 38 statements focused on pre-NT practices, including diagnosis and staging (n = 15), evaluation and optimization (n = 20), and decision-making (n = 3); 29 statements defined best practices during NT, including initiation (n = 3), delivery of therapy (n = 8), restaging practices (n = 12), and management of complications during NT (n = 6); and 15 best practices were identified to guide treatment post-NT, focusing on surgery (n = 7), pathology (n = 4), and follow-up (n = 3). Conclusions Using a modified Delphi consensus technique, best practice guidelines were developed focusing on the optimal standards for delivering NT to patients with localized PDAC. Given the prognostic importance of completing multimodality therapy, efforts to standardize and optimize the delivery of NT represent an immediate opportunity to decrease care variation and improve outcomes for patients with PDAC. Future research should focus on validating and implementing best practice standards into clinical practice.

e19503 Background: The risk of venous thromboembolic events (VTE) despite mandatory thromboprophylaxis (TPx) in patients with multiple myeloma (MM) receiving contemporary lenalidomide (Len)-based regimens is not well-defined. Methods: Medline, Embase, and CENTRAL databases were queried to identify phase 1-3 clinical trials of Len-based regimens with mandatory TPx published until April 2018. Random effects meta-analysis was performed using CMAv3 software to obtain crude, and per 100 patient-cycle incidence rate (IR). Prespecified subgroup analyses were performed to obtain IR for VTE by (1) line of treatment (newly diagnosed [ND], relapsed/refractory [RR] and maintenance [Mx] phase) and (2) different Len-based regimens. Results: We screened 1069 citations and analyzed data from 43 trials with 8696 patients, including 17 in ND, 23 in RR and 3 in Mx phase. 12 out of 43 were Phase 3 randomized controlled trials. The overall incidence of VTE in 8696 patients was 4.6% (95% CI, 3.7-5.8%), with 1.0 event (95% CI, 0.7-1.4%) per 100 patient cycles. In patients with NDMM (n = 3295), 193 (5.9%) experienced VTE with a pooled IR of 0.9 (95% CI, 0.3-2.3) per 100 patient-cycle. Among RRMM patients (n = 3981), 251 (6.3%) experienced VTE with a pooled IR of 1.0 (95% CI, 0.4-2.3) per 100 patient-cycle. The incidence of VTE was highest with the following triplet regimens: Rd + Proteasome inhibitor (PIs) and Rd + Anthracycline (Ac) [+/- Vincristine], with the latter having a high risk of VTE regardless of the treatment phase (table). With Len Mx (n = 1420), 38 (2.7%) patients had VTE, with a pooled IR of 0.2 (95% CI, 0-2.1) per 100 patient-cycle. Conclusions: Patients with MM undergoing treatment with Len-based regimens remain at risk of VTE despite current TPx strategies. The estimates of VTE IR provided by this analysis can guide clinicians in assessing treatment-related risk of VTE in this setting. [Table: see text]

Background Thromboprophylaxis is routinely used with lenalidomide‐based regimens in multiple myeloma because of a substantial risk of venous thromboembolism (VTE). However, little is known about the incidence of VTE with contemporary lenalidomide‐based regimens. The objective of the current study was to estimate the incidence of VTE despite thromboprophylaxis with currently used lenalidomide‐based regimens in patients with myeloma. Methods The Ovid MEDLINE, Embase, and Cochrane databases were queried from study inception to January 2019 for keywords to cover the following concepts: “lenalidomide,” “venous thromboembolism,” and “multiple myeloma.” Phase 1, 2, and 3 clinical trials evaluating lenalidomide‐based regimens with thromboprophylaxis were included. The pooled incidence rate of VTE was estimated using a random‐effects model. Results The search generated 1372 citations, with 51 clinical trials and 9069 patients included for analysis. The most common thromboprophylaxis agents were aspirin, low‐molecular‐weight heparin or warfarin, administered either per risk‐stratification or at investigators' discretion. The pooled incidence of VTE in trials of patients who had newly diagnosed and relapsed/refractory myeloma was 6.2% (95% CI, 5.4%‐7.1%) over median treatment durations ranging from 2 to 34 cycles, which translated into 1.2 VTE events per 100 patient‐cycles (95% CI, 0.9‐1.7 VTE events per 100 patient‐cycles). Among contemporary regimens, the risk of VTE was low with combined lenalidomide and low‐dose dexamethasone (0.2 [95% CI, 0.1‐0.6] events/100 patient‐cycles) and lenalidomide maintenance (0.0 [95% CI, 0.0‐0.7] events per 100 patient‐cycles). VTE risk was higher with combined lenalidomide and low‐dose dexamethasone plus proteasome inhibitors (1.3 [95% CI, 0.7‐2.3] events per 100 patient‐cycles). Conclusions Despite adequate thromboprophylaxis, lenalidomide‐based regimens have a substantial risk of VTE in controlled clinical trial settings. Further studies are needed on new thromboprophylaxis strategies with regimens that have a high VTE risk.

Abstract Background: Patients on systemic therapy for cancer are at varying risk for venous thromboembolism (VTE) and its consequences. Thromboprophylaxis is recommended in hospitalized medical and surgical cancer patients, but most VTE occurs in ambulatory cancer patients where risk can be estimated with a validated score. However, the benefit of extended outpatient thromboprophylaxis is uncertain with heparins and has not been tested with direct oral anticoagulants. Methods: We conducted a double-blind, randomized, placebo-controlled, parallel-group, multicenter study in adult ambulatory patients with various cancers initiating a new systemic regimen and at increased risk for VTE (defined as Khorana score ≥ 2). Enrolled subjects were screened for deep-vein thrombosis (DVT) and if none found randomized 1:1 to rivaroxaban 10 mg once daily or placebo up to day 180. Subjects were screened with lower extremity ultrasounds every 8 weeks on study. Primary efficacy endpoint was a composite of objectively confirmed symptomatic or asymptomatic lower-extremity proximal DVT, symptomatic upper- or lower-extremity distal DVT, symptomatic or incidental pulmonary embolism and VTE-related death. ISTH-defined major bleeding was the primary safety endpoint. All endpoints were adjudicated by blinded independent committees. Analyses for efficacy endpoints were conducted for intent-to-treat (ITT) population (all randomized patients) for the up-to-day 180 observation period (primary) and the on-treatment period (supportive). Safety analyses were conducted for on-treatment period only for patients who received at least one dose of study drug. Results: Of 1,080 patients who provided consent, 49 (4.53%) had DVT on baseline screening and another 190 screen-failed due to other reasons. Of 841 randomized patients, 274 (32.6%) had pancreatic cancer; 698 (83%) were white and 428 (50.9%) were male. The primary efficacy endpoint occurred in 25 of 420 patients (5.95%) and 37 of 421 patients (8.79%) (HR, 0.66; 95% CI, 0.40 to 1.09; p=0.101) in the rivaroxaban and placebo groups respectively (number needed to treat, NNT=35) in the up-to-day 180 observation period (Figure 1A). Of all patients with VTE, 38.70% experienced events after discontinuing study drug. In a pre-specified analysis of all randomized patients during the on-treatment period, the primary endpoint occurred in 11 of 420 patients (2.62%) and 27 of 421 (6.41%) in rivaroxaban and placebo groups respectively (HR 0.40, 95% CI 0.20 to 0.80, p=0.007) (NNT=26) (Figure 1B). Major bleeding occurred in 8 of 405 (1.98%) in the rivaroxaban group and in 4 of 404 (0.99%) patients in the placebo group (hazard ratio, 1.96; 95% CI, 0.59 to 6.49; p=0.265) (number needed to harm, NNH=101). Clinically relevant non-major bleeding occurred in 2.72% and 1.98% of rivaroxaban and placebo groups, respectively (HR, 1.34; 95% CI, 0.54 to 3.32; p=0.53) (NNH=135). Adverse events were comparable between groups. For secondary efficacy endpoints, a pre-specified analysis of the composite of primary endpoint with addition of arterial and visceral thromboembolic events in the up-to-day 180 observation period showed significantly fewer events in rivaroxaban versus placebo group (6.90% versus 10.70% respectively; HR, 0.62; 95% CI: 0.39, 0.99; p=0.04). All-cause mortality occurred in 20.0% of patients in rivaroxaban group and 23.8% in placebo group (HR=0.83, 95% CI 0.62, 1.11; p=0.213). A pre-specified composite of the primary endpoint with all-cause mortality occurred in 23.1% of patients in rivaroxaban group and 29.5% in placebo group (HR 0.75; 95% CI 0.57 to 0.97; p=0.03) (Figure 1C). Conclusions: Rivaroxaban significantly reduced VTE and VTE-related death during the on-treatment period but not during the full study period; over one-third of events occurred post discontinuation of study drug. The incidence of major bleeding was low. The Khorana risk score cut-off of ≥2 identified cancer patients at high risk of thrombotic events both at baseline (4.53%) and during study (8.79% with additional 1.66% arterial events in placebo group). These results should inform future recommendations regarding thromboprophylaxis in at-risk ambulatory cancer patients. (Funded by Janssen; ClinicalTrials.gov number, NCT02555878) Figure 1 Figure 1. Disclosures Khorana: Parexel: Other: Personal fees and non-financial support for travel; Sanofi: Consultancy, Other: Personal fees and non-financial support for travel; Pfizer: Consultancy, Other: Personal fees and non-financial support for travel; Janssen: Consultancy, Other: Personal fees, Research Funding; TriSalus: Other: Personal fees; Halozyme: Other: Personal fees and non-financial support for travel; Seattle Genetics: Other: Personal fees and non-financial support for travel; AngioDynamics: Other: Personal fees and non-financial support for travel; LEO Pharma: Other: Personal fees and non-financial support for travel; Medscape/WebMD: Other: Personal fees and non-financial support for travel; Pharmacyclics: Other: Personal fees; PharmaCyte: Other: Personal fees; Bayer: Consultancy, Other: Personal fees and non-financial support for travel. Soff:Janssen: Research Funding; Amgen: Research Funding. Kakkar:Bayer AG: Consultancy, Research Funding; Boehringer Ingelheim: Consultancy; Daiichi Sankyo Europe: Consultancy; Janssen Pharmaceuticals: Consultancy; Pfizer: Consultancy; Sanofi S.A.: Consultancy; Verseon: Consultancy. Vadhan-Raj:AMAG Pharmaceuticals, Inc: Other: received funding from Amag to support clinical trial; Janssen: Consultancy, Research Funding. Riess:Janssen Scientific Affairs: Other: Travel reimbursement; Bayer: Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Boehringer Ingelheim: Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Daiichi Sankyo: Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Bristol-Myers Squibb: Speakers Bureau; Pfizer: Membership on an entity's Board of Directors or advisory committees. Wun:Eli Lilly, Inc.: Research Funding; Emmaus, Inc.: Membership on an entity's Board of Directors or advisory committees; Glycomimetics, Inc.: Research Funding; Janssen: Membership on an entity's Board of Directors or advisory committees; Pfizer, Inc.: Membership on an entity's Board of Directors or advisory committees. Streiff:Janssen: Membership on an entity's Board of Directors or advisory committees, Research Funding; Bayer: Consultancy, Membership on an entity's Board of Directors or advisory committees; CSL Behring: Other: outcome adjudication committee; Daiichi Sankyo: Membership on an entity's Board of Directors or advisory committees; Pfizer: Membership on an entity's Board of Directors or advisory committees; Boehringer Ingelheim: Research Funding; Roche: Research Funding; Portola: Consultancy, Research Funding. Garcia:Shingoi: Consultancy; Portola: Research Funding; Pfizer: Consultancy; Janssen: Consultancy, Research Funding; Incyte: Research Funding; Daiichi Sankyo: Research Funding; Bristol Meyers Squibb: Consultancy; Boehringer Ingelheim: Consultancy; Retham Technologies LLC: Consultancy; Genzyme Corporation: Consultancy; Alexion: Consultancy. Liebman:Janssen (Johnson & Johnson): Other: steering committee of the CASSINI trial during the conduct of the study. Belani:Janssen: Consultancy. O'Reilly:3DMed, Agios, AlignMed, Amgen, Antengene, Aptus, ASLAN, Astellas, AstraZeneca, Bayer, BeiGene, Bioline, BMS, Boston Scientific, Bridgebio, CARsgen, Celgene, CASI, Cipla, CytomX, Daiichi, Debio, Delcath, Eisai, Exelixis, Genoscience, Gilead, Halozyme: Consultancy; Hengrui, Incyte, Inovio, Ipsen, Jazz, Janssen, Kyowa Kirin, LAM, Lilly, Loxo, Merck, Mina, NewLink Genetics, Novella, Onxeo, PCI Biotech, Pfizer, PharmaCyte, Pharmacyclics, Pieris, QED, RedHill, Sanofi, Servier, Silenseed, Sillajen, Sobi, Targovax: Consultancy; Janssen: Research Funding; Acta Biologica, Agios, Array, AstraZeneca, Bayer, BeiGene, BMS, CASI, Celgene, Exelixis, Genentech, Halozyme, Incyte, Lilly, MabVax, Novartis, OncoQuest, Polaris Puma, QED, and Roche: Research Funding; Tekmira, twoXAR, Vicus, Yakult, and Yiviva: Consultancy. Patel:Janssen Pharmaceuticals: Research Funding. Yimer:Seattle Genetics: Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Celgene: Membership on an entity's Board of Directors or advisory committees; Janssen: Speakers Bureau; AstraZeneca: Speakers Bureau; Epizyme: Equity Ownership; Clovis Oncology: Equity Ownership; Puma Biotechnology: Equity Ownership. Wildgoose:Janssen Scientific Affairs: Employment; Johnson & Johnson: Equity Ownership. Burton:Janssen Pharmaceuticals: Employment; Johnson & Johnson: Equity Ownership. Vijapurkar:Janssen Pharmaceuticals, Inc.: Employment; Johnson & Johnson: Equity Ownership. Kaul:Janssen Pharmaceuticals, Inc.: Employment; Johnson & Johnson: Equity Ownership. Eikelboom:AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb/Pfizer, Daiichi Sankyo, GlaxoSmithKline, Janssen, sanofi-aventis, and Eli Lilly: Honoraria, Research Funding; Heart and Stroke Foundation: Other: Personnel award. Bauer:Janssen: Consultancy. Kuderer:Celldex: Consultancy; Mylan: Consultancy, Other: Travel, Accommodations, Expenses; Myriad Genetics: Consultancy; Halozyme: Consultancy; Coherus Biosciences: Consultancy, Other: Travel, Accommodations, Expenses; Janssen Scientific Affairs, LLC: Consultancy, Other: Travel, Accommodations, Expenses; Pfizer: Consultancy; Bayer: Consultancy. Lyman:Halozyme; G1 Therapeutics; Coherus Biosciences: Consultancy; Generex Biotechnology: Membership on an entity's Board of Directors or advisory committees; Amgen: Other: Research support.

Purpose An ASCO provisional clinical opinion (PCO) offers timely clinical direction to ASCO’s membership and other health care providers. This PCO addresses identification and management of patients and family members with possible predisposition to pancreatic adenocarcinoma. Methods ASCO convened an Expert Panel and conducted a systematic review of the literature published from January 1998 to June 2018. Results of the databases searched were supplemented with hand searching of the bibliographies of systematic reviews and selected seminal articles and contributions from Expert Panel members’ curated files. Provisional Clinical Opinion All patients diagnosed with pancreatic adenocarcinoma should undergo assessment of risk for hereditary syndromes known to be associated with an increased risk for pancreatic adenocarcinoma. Assessment of risk should include a comprehensive review of family history of cancer. Individuals with a family history of pancreatic cancer affecting two first-degree relatives meet criteria for familial pancreatic cancer (FPC). Individuals (cancer affected or unaffected) with a family history of pancreatic cancer meeting criteria for FPC, those with three or more diagnoses of pancreatic cancer in same side of the family, and individuals meeting criteria for other genetic syndromes associated with increased risk for pancreatic cancer have an increased risk for pancreatic cancer and are candidates for genetic testing. Germline genetic testing for cancer susceptibility may be discussed with individuals diagnosed with pancreatic cancer, even if family history is unremarkable. Benefits and limitations of pancreatic cancer screening should be discussed with individuals whose family history meets criteria for FPC and/or genetic susceptibility to pancreatic cancer. Additional information is available at www.asco.org/gastrointestinal-cancer-guidelines .

AbstractVenous thromboembolism (VTE) is a frequent complication of cancer associated with morbidity, mortality, increased hospitalizations and higher health care costs. Cancer patients at increased risk for VTE can be identified using a validated risk assessment score, and the incidence of VTE can be reduced in high-risk settings using anticoagulation. Rivaroxaban is a potent, oral, direct, factor Xa inhibitor approved for the prevention and treatment of thromboembolic events, including VTE. CASSINI is a double-blind, randomized, parallel-group, multicentre study comparing rivaroxaban with placebo in adult ambulatory patients with various cancers who are initiating systemic cancer therapy and are at high risk of VTE (Khorana score ≥ 2). Patients with primary brain tumours or those at risk for bleeding are excluded. Approximately 700 patients will be randomized 1:1 to rivaroxaban 10 mg daily or placebo for up to 6 months if there is no evidence of VTE from compression ultrasonography (CU) during screening or from routine care imaging within 30 days prior to randomization. Mandatory CU will also be performed at weeks 8 and 16 (±7 days), and at study end (±3 days). The primary efficacy hypothesis is that anticoagulation with rivaroxaban reduces the composite of objectively confirmed symptomatic or asymptomatic, lower-extremity, proximal deep-vein thrombosis (DVT); symptomatic, upper-extremity DVT; symptomatic or incidental pulmonary embolism; and VTE-related death compared with placebo. The primary safety objective is to assess major bleeding events (Clinical trial information: NCT02555878).