Citations

e14017 Background: The role of stereotactic radiosurgery (SRS) is well established in the setting of one to four brain metastases (BM), offering effective control with reduced neurocognitive toxicity compared with whole-brain radiotherapy (WBRT). However, for patients with five or more BM, the optimal radiation strategy remains uncertain, and comparative outcome data are limited. We performed a systematic review and meta-analysis to compare overall survival (OS) between SRS and WBRT in patients with ≥ 5 BM. Methods: A systematic review was conducted to identify comparative studies of SRS versus WBRT in patients with ≥5 brain metastases. Retrospective and prospective studies reporting median OS with available cohort sizes were included. Comparative survival was summarized using pooled ratios of median survival within a random-effects meta-analytic framework. Results: Five comparative studies were included, comprising 226 patients treated with SRS and 261 treated with WBRT, including both retrospective and prospective cohorts. Sample-size–weighted summary (with reported medians) showed median OS was 8.1 months with SRS and 7.7 months with WBRT. Pooled analysis demonstrated no significant difference in OS between treatment strategies (ratio of median survival, 1.03; 95% CI, 0.73–1.46; p = 0.81). Conclusions: Among patients with ≥5 BM, SRS and WBRT were associated with comparable OS. These data suggest that OS alone may not guide treatment choice in this population, emphasizing the importance of careful patient selection. As focal therapies are increasingly incorporated into clinical practice for patients with higher intracranial disease burden, prospective, multi-institutional randomized clinical trials are needed to better define optimal management strategies. Summary of included studies reporting median overall survival (OS). Study Study Design SRS (n) WBRT (n) Median OS in SRS cohort (months) Median OS in WBRT cohort (months) Kim et al., 2009 Retrospective 29 39 5.6 7.2 Mizuno et al., 2019 Retrospective 24 20 7.3 7 .2 Li et al., 2020 Prospective, randomized 35 34 7.8 8.9 Bodensohn et al., 2023 Prospective, non-randomized 40 70 10.4 6.5 Aizer et al., 2025 Prospective, randomized 98 98 8.3 8.5

e14016 Background: Postoperative stereotactic radiosurgery (SRS) is utilized following resection of brain metastases (BM), however use is limited by target delineation uncertainty and an increased risk of leptomeningeal disease (LMD). Preoperative SRS allows for precise targeting and is being increasingly utilized in treatment paradigms of resectable BM. Reported outcomes across studies are promising and there are several phase 3 trials which are underway. Therefore, we pooled multiple studies and conducted a meta-analysis to estimate 1-year local control (LC), LMD incidence, and overall survival (OS) following single-fraction preoperative SRS for resectable BM. Methods: We systematically searched PubMed, Embase, and Scopus to identify studies of single-fraction preoperative SRS for resectable BM. For each cohort, data for 1-year LC, 1-year LMD, and 1-year OS, along with corresponding sample sizes, were extracted. A proportion-based meta-analysis was performed to derive pooled estimates and 95% confidence intervals (CIs), with between-study heterogeneity quantified using I². Results: Ten studies published between 2016 and 2025, including both prospective trials and retrospective cohorts, met the inclusion criteria. There were 617 patients treated with single-fraction preoperative SRS. The median age was 60.8 years (range 58-65 years), and the median prescribed dose was 15.8 Gy (range 14.5-17.0 Gy). The pooled 1-year LC rate was 81.9% (95% CI, 76.9-85.9; I² = 43.4%). The pooled 1-year incidence of LMD was 4.6% (95% CI, 2.8-7.4; I² = 0%), and the pooled 1-year OS rate was 60.2% (95% CI, 55.5-64.8; I² = 0%). Sensitivity analyses showed stable pooled estimates, with no single study driving the results. Conclusions: Single-fraction preoperative SRS is associated with low rates of LMD and acceptable LC and OS. In the context of historically reported postoperative SRS outcomes, these results suggest that preoperative SRS represents a promising strategy in selected patients. Randomized trials are needed to define the comparative role and optimal patient selection for preoperative SRS, with ongoing studies (NCT05341739, NCT03741673) expected to clarify its role in the management of resectable BM. Study characteristics and reported outcomes of single-fraction preoperative SRS for BM. Study Study design Number of patients Median age (years) Median dose (Gy) 1-year LTC (%) 1-year LMD (%) 1 year OS (%) Agarwal, 2024 Trial 47 59 NR NR 4.8 NR Higazy, 2025 Trial 35 61 17 82 2.86 NR Li, 2022 Retrospective 24 64 NR 87.6 NR NR Mallela, 2024 Retrospective 26 64.5 NR 77.2 NR NR Murphy, 2024 Trial 35 63 NR 76.6 NR 59 Patel, 2018 Retrospective 12 57.5 NR 49.1 10 74.1 Patel, 2016 Retrospective 66 58.2 14.5 84.1 3.2 NR Prabhu, 2025 Retrospective 124 NR 17 87.2 1.7 64.5 Prabhu, 2025 Retrospective 183 NR 15 86.4 7.3 58.1 Prabhu, 2020 Retrospective 65 NR NR 77.1 4.2 56.3

e13730 Background: Financial toxicity, the burden of treatment cost, affects up to half of people with cancer and can affect survival. Financial toxicity is a health equity issue, disproportionately affecting Black patients. This cross-sectional study used baseline data from the Discussions of Cost (DISCO) Trial, an RCT to improve patient-provider treatment cost discussions and, in turn, mitigate financial toxicity, to examine whether Black and White patients differed in patient-provider communication perceptions and financial toxicity correlates. Methods: Data were collected from 2021-2025 at an NCI-designated comprehensive cancer center. Participants were Black or White patients diagnosed with solid tumor cancers. At baseline, patients completed demographic measures; communication measures of self-efficacy in patient-provider interactions (e.g., “I am confident I know what questions to ask a provider”); self-efficacy in managing cancer treatment costs (e.g., “I am confident I can understand treatment costs”); and whether they have a provider- or patient-centered approach (e.g., “the provider should decide what gets talked about”); and financial measures including economic burden (e.g., “it’s difficult to live on my income”); treatment cost distress (e.g., “worries about paying for my cancer treatment keep me up at night”); and anticipated material hardship (e.g., “I worry about the financial problems I will have in the future as a result of my treatment”). Comparisons were conducted by patient race. Results: A total of 202 patients participated (116 Black patients, 86 White patients). White patients were more likely to have advanced degrees, a higher annual income, and less likely to have Medicaid compared to Black patients. Black patients reported higher self-efficacy for managing treatment costs ( p = .002) and a higher provider-focused orientation to the patient-provider relationship ( p = .001), than White patients. Black patients reported greater levels of financial burden ( p < .001), cost distress ( p = .005), and anticipated material hardship ( p = .003). There was no difference in self-efficacy for interacting with providers. Conclusions: These results suggest that a communication intervention for patients may help mitigate the experience of financial toxicity, especially for Black patients. Clinical trial information: NCT04766190 . Patient characteristics. Black Patients White Patients p -value Race 116 86 Education < High School High School/GED Some College 2 or 4-year College Degree Graduate/Professional College Trade/Technical/Vocational School 10352626712 0141631187 < .001 Annual Household Income 0 - $19,999 $20,000 - 39,999 $40,000 - 59,999 $60,000 - 79,999 $80,000 - 99,999 $100,000 - 149,999 $150,000 + 53201611431 51215991716 <.001 Insurance Type* Medicaid Medicare Medicare supplement Private None 604911330 113726393 <.001>.99<.001.20.15 *Patients could select more than one

e13729 Background: Financial toxicity, the burden of treatment cost, affects up to half of people with cancer and can affect survival. Financial toxicity is a health equity issue, disproportionately affecting Black patients. Patient education about cost and patient-provider cost discussions are recommended to mitigate financial toxicity but occur infrequently. The Discussions of Cost (DISCO) App, designed to mitigate financial toxicity through a tailorable education and communication intervention, was assessed for efficacy in an RCT with Black and White patients. Here we describe the DISCO App's feasibility and acceptability. Methods: Data were collected from 2021 to 2025 at an NCI-designated comprehensive cancer center. Participants were Black or White patients diagnosed with solid tumor cancers. While waiting to see their provider, patients randomized to the intervention arm used the DISCO App on a tablet. The DISCO App includes a video about treatment costs, ways to manage costs, and the importance of discussing costs with providers. Once patients enter their socio-demographic information (e.g., employment) and financial concerns, they receive a tailored list of cost questions to ask their provider. Patients responded to measures immediately after using the DISCO App. Feasibility was assessed by patient reports of how much time they needed to use the DISCO App and if they were able to use the DISCO App before the provider entered the exam room. Acceptability was measured by items assessing patient perceptions of the DISCO App (1 = strongly disagree; 5 = strongly agree). Descriptive statistics summarized results and mean comparisons were used to assess perceptions by patient self-identified race. Results: Of the 202 patients recruited, 124 were randomized to the intervention arm (67 Black patients; 57 White patients). All patients completed using the DISCO App prior to their provider entering the exam room and 80% of patients needed 15 minutes or less to use the DISCO App, with no difference by patient race. Black patients ( M = 4.4) reported finding the DISCO App more helpful than White patients ( M = 3.9; p = .02) and it brought up cost issues they had not considered (Black patients = 3.5; White patients = 3.0; p = .05). Patients reported the questions were easy to understand ( M = 4.4), helped them feel it was okay to talk with their provider about cost ( M = 4.0), that it prompted cost questions ( M = 3.3), that they learned what costs to consider (55%), and learned ways to manage cost (62%) with no differences by patient race. Conclusions: The DISCO App was feasible and acceptable to Black and White patients with cancer, with Black patients reporting more favorable perceptions than White patients. Next steps include assessing effectiveness in improving outcomes including prompting patient-provider treatment cost discussions and related financial toxicity outcomes (e.g., treatment adherence). Clinical trial information: NCT04766190 .

e13683 Background: Immune checkpoint inhibitors (ICIs) have transformed cancer treatment but are frequently complicated by immune-related adverse events (irAEs), which may result in treatment interruption, hospitalization and mortality. Reliable pre-treatment tools to identify patients at higher risk for irAEs, particularly checkpoint inhibitor pneumonitis (CIP) remain limited. Artificial intelligence (AI)-based prediction models have emerged but their clinical performance and readiness for implementation are unclear. Methods: We systematically searched PubMed, Embase, Scopus and Web of Science (inception-January 2025) for studies developing AI-based models to predict irAEs or CIP in patients receiving ICIs. Eligible studies included radiomics, deep learning, natural language processing and clinical machine-learning approaches. Two reviewers independently extracted data on cancer type, toxicity endpoint, modality, algorithm, validation strategy and model performance. Study design and reporting characteristics were assessed. Due to substantial methodological heterogeneity, results were synthesized descriptively. Results: Fifteen studies met inclusion criteria, including mixed solid tumor cohorts and lung-predominant populations. Predictive targets included any-grade irAEs and organ-specific toxicities, most commonly pneumonitis. Model approaches ranged from clinical and electronic health record–based models to CT- and PET/CT-based radiomics and multimodal imaging-clinical frameworks. Reported discriminatory performance varied widely, with AUC values ranging from 0.64 to > 0.90 across model types. The most consistent performance was observed in deep learning and multimodal radiomics-clinical models for pneumonitis, with AUCs frequently between 0.75 and 0.92, whereas several clinical-only models demonstrated more modest performance. Most studies were retrospective and single-center, used heterogeneous irAE definitions and lacked standardized imaging protocols. External validation and calibration assessment were uncommon. Conclusions: AI-based models for predicting irAEs in ICI-treated patients demonstrate promising performance, particularly for pneumonitis using radiomics and multimodal approaches. However, current evidence is limited by heterogeneity, scarce external validation and uncertain generalizability. Prospective, multi-institutional validation and standardized outcome definitions are needed before AI-driven irAE risk prediction can be integrated into routine oncology practice to guide surveillance and patient counseling.

e13677 Background: The development of AI led to the emergence of tools such as AIS, which have been increasingly adopted in medicine. However, formal evaluation of barriers and clinical care gaps in HemOnc is limited, particularly regarding treatment plan complexities and toxicity documentation. Methods: Structured investigator-developed surveys about AIS availability, adoption, benefits, and challenges were developed on Qualtrics and approved by the IRB. HemOnc faculty physicians (FP), advanced practice providers (APP), and trainees (HOT) were surveyed between 11/14/25-1/16/26. Surveys were distributed in the USA via email, social media (on X), and ASCO myConnection. Descriptive statistics were used for analysis. Results: 132 participants responded to the survey: 55 FP (42%), 14 APP (11%), and 63 HOT (48%). 80% of them reported having AIS at their institutions, and 54% of those reported using these tools in their clinical practice. Among FP and APP, 72% practiced in academic settings, 14% in community, and 13% in hybrid centers. Among those not using AIS, identified barriers were issues with note formatting (56%), lack of training (25%), and concerns that the note content is incorrect or inadequate (25%). 60% are willing to use AIS if note formatting matched their preference, and 42% would try it if contents better reflected the patient encounter. Table 1 details the proportion of encounters using AIS, the portion of documentation it was used for, and the perceived challenges with these tools. Among AIS users, the average satisfaction score was 5.3 on a 7 point scale, corresponding to participants being somewhat satisfied. Participants agreed slightly that AIS had reduced their feeling of work-related burnout, with an average score of 3.6 on a 5 point scale. Overall, participants felt neutral about how well-suited AIS are for HemOnc, with an average score of 4.0 on a 7 point scale. Conclusions: Despite AIS being widely available, 46% of the surveyed HemOnc workforce do not use them in clinical practice. Those who do reported decreased time spent on documentation outside work hours and a slight reduction in the feeling of burnout. Formatting issues and record integration remain barriers, with substantial concerns about AIS not adequately describing HemOnc treatment plans. To bridge the last mile of integration into HemOnc, partnership between AIS developers and the clinical workforce is essential to meet clinical needs, while prioritizing patient safety and data privacy. FP + APP % HOT % Patient encounters where AIS is used 62 44 Used AIS for the subjective portion of note 68 96 Used AIS for the assessment/plan portions of note 40 25 AIS decreased the time on clinical documentation outside work hours 96 93 AIS note formatting needs improvement 68 57 Issues integrating AIS and outside records 46 71 AIS note contents about the plan are suboptimal 32 61

e13663 Background: Artificial intelligence (AI) in oncology, including computer vision and radiomics, has expanded rapidly with substantial federal support. However, the extent to which growth in funding and publications is accompanied by growth in prospective interventional evaluation is not well quantified. We conducted an ecosystem-level analysis to compare temporal trends in federal awards, peer-reviewed publications, and interventional clinical trial registrations in AI-oncology. Methods: We performed a cross-sectional landscape analysis spanning 2015–2025 using three public data sources: (1) federally funded extramural awards indexed in NIH RePORTER (including NIH/NCI and other agencies where available); (2) interventional studies registered on ClinicalTrials.gov; and (3) PubMed-indexed publications. Records were identified using a consistent Boolean query requiring AI-related terms (“artificial intelligence,” “machine learning,” “deep learning,” “neural network,” “large language model,” “computer vision,” “natural language processing,” “radiomics”) AND oncology-related terms (“cancer,” “oncology,” “tumor,” “neoplasm,” “carcinoma,” “melanoma,” “leukemia,” “lymphoma”). Annual absolute counts were extracted (without normalization) to preserve volume differences across the ecosystem. We summarized growth over time and calculated publication-to-trial and grant-to-output ratios. Results: Between 2015 and 2025, AI-oncology federal awards totaled 11,503. PubMed-indexed AI-oncology publications increased 38-fold from 558 (2015) to 21,414 (2025), totaling 81,782. Interventional AI-oncology trial registrations increased from 2 (2015) to 84 (2025), totaling 372 across the period. Overall, the publication-to-interventional-trial ratio was 220:1; interventional trial registrations represented 0.45% of publication volume. For every federal award, approximately 7 publications were produced, but only 0.03 interventional trials were registered. Conclusions: Across 2015–2025, AI-oncology funding and publications increased dramatically, while interventional trial registrations remained comparatively sparse. This persistent publication-to-trial gap highlights a translational bottleneck and supports targeted initiatives to accelerate prospective validation, such as dedicated funding mechanisms, multi-center evaluation networks, and trial-ready implementation infrastructure, to realize clinical benefit from AI in cancer care.

e13661 Background: The process of data abstraction and curation represents a significant proportion of clinical trial costs, requiring up to $30,000 per patient depending on the data complexity. Research variables such as stage, performance status (PS), and cancer disease status are typically present in the electronic health record (EHR) in free text but require abstraction for research use. Structuring these real-world data (RWD) in an exportable format utilizing minimum common oncology data elements (mCODE) could reduce the need for abstraction and the potential for erroneous transcription that has been shown to be 5.51-7.72% in a meta-analysis. We assessed the feasibility, availability, and accuracy of RWD for use across Alliance for Clinical Trials in Oncology sites by comparing RWD to matched electronic case report form data (CRF data) at 2 sites where EHR-integrated tools collect structured RWD for all solid tumor patients. Methods: Stage, PS, and cancer disease status (CDS) were extracted from the EHR as RWD and from eCRFs for 100 patients enrolled in National Clinical Trials Network trials across multiple disease sites from 2022 through 2025. Data from baseline and day 1 of cycles 1-4 were compared (500 encounter timepoints). Results: Stage, PS, and CDS were available from RWD more often than CRFs (85 vs 70 / 100 patients, 436 vs 416 / 500 encounters, and 345 vs 83 / 500 encounters respectively). When stage was present in both (n = 58), there were 5 mismatches (8.6%). PS had 22 mismatches (4%) and CDS had 45 mismatches (11.4%). After correcting for differences in CRF vs RWD definitions, CDS mismatches dropped to 1.2% (n = 4). Discordance between RWD and CRF data was related to differences in the timepoint of data collection (e.g., systemic staging scans completed after the initial consult), CRF data obtained from other EHR notes (e.g., oncology research nurse notes containing PS), inherent discordance between RWD and CRF data choices (e.g., responding vs complete or partial response), and failure to update EHR data capture tool fields. RWD were available 21 days (median) before CRF data. Conclusions: Our results suggest that RWD captured directly within clinical notes match CRF data, are more often present, and are available sooner. RWD could serve as a data source for trials, decreasing the need for manual curation and cost. Although some discordance exists, overall rates are low and close to the published error rate for abstraction.

e13660 Background: We previously showed success with collection of structured data elements within the electronic health record using the minimum common oncology data elements (henceforth, mCODE tool) supported by soft-stop close visit validation (CVV) and staff support interventions in a limited cohort. Since the additional interventions are resource-intensive and difficult to apply across an entire department, we therefore implemented a department-wide roll-out supported by the least resource-intensive interventions while monitoring uptake. Methods: The mCODE tool collects five data elements: stage, disease status, performance status, intent of treatment, and intent to change therapy. The Department of Oncology providers received education on the mCODE tool and were asked to implement use at a faculty meeting; soft-stop CVV was enabled at that time. The percentage of data captured for each data element was collected from the notes of providers who did not participate in prior pilots for the 4 months post-implementation. mCODE tool usage was defined as entry of at least one data element except stage (as stage has long been a structured data element in Epic, can be entered independently of the mCODE tool, and many providers were entering stage prior to mCODE tool introduction). Results: 86 MDs and 20 APPs for a total of 106 providers were included, resulting in 14,973 visits and 8,745 unique patients. 73.6% of providers (n =78) utilized the mCODE tool. Of those who utilized the tool, median overall data capture across the four data elements was 24.0%. 22.6% of providers (n =24) achieved > 80% overall data capture and 32.1% (n =34) had over 50% overall data capture despite minimal intervention. 75.6% of MDs (n = 65) utilized the tool with a median data capture of 16.3% compared to 65% of APPs (n = 13) with a median data capture of 90.3%. Rates of data capture for individual data elements were similar. Conclusions: Use of soft-stop CVV appears to drive utilization of the mCODE tool but is insufficient to promote > 80% data capture for most providers without more intervention. Supportive interventions such as individualized assistance with incorporation into existing templates and directed communication from leadership remain necessary to promote use of structured data elements and will be implemented in sequence to understand how we achieve the target data capture ( > 80%) at scale.

e13656 Background: Effective multidisciplinary cancer care is bottlenecked by the cognitive burden of synthesizing fragmented electronic health records across different specialties. Current AI tools often function as isolated chatbots rather than integrated workflow assistants. We developed and validated "OncoSphere AI," an Agentic AI multidisciplinary clinical workflow support tool designed to autonomously orchestrate the multidisciplinary review process by deconstructing complex cases into domain-specific insights. Methods: OncoSphere AI is a novel, multi-agent large language model–driven platform that simulates a multidisciplinary tumor board for cancer care. The architecture uses a parallel reasoning framework in which eight virtual specialty agents—Medical Oncology, Radiation Oncology, Surgical Oncology, Pathology, Radiology, Pharmacy, Clinical Research, and Nurse Navigation—independently analyze structured and unstructured patient data against specialty specific guidelines and contemporary peer-reviewed literature. System performance was assessed via built-in telemetry (task completion, latency, inter-agent agreement) and blinded review by eight expert oncologists using a structured survey to rate data extraction fidelity, guideline concordance, safety, and clinical utility compared with standard manual review. Results: The study cohort (n=20, median age 59.5 years; 55% female) had a median of 2.5 (range 0-7) prior lines of therapy and multiorgan metastatic disease (median: 2, range 1–4), including lung, breast, gastrointestinal (CRC, pancreatic, hepatobiliary), and genitourinary (prostate, renal) malignancies. OncoSphere AI demonstrated high workflow utility (mean rating: 4/5) and strong extraction fidelity, with more than 90% of biomarker extractions rated as a “perfect match” by expert reviewers. Operational efficiency gains were substantial with AI processing averaged 4.7 minutes per case versus 29.3 minutes for manual review, corresponding to an 84% reduction in time (p<0.001). System telemetry revealed adaptive compute behavior, with the platform autonomously allocating 1.73-fold more reasoning tokens to complex, refractory cases than to standard scenarios. Safety review identified only one critical contraindication issue in a heavily pretreated, multi-line setting. Conclusions: OncoSphere AI effectively automates the clinical “perception layer,” consistently transforming heterogeneous longitudinal data into decision-grade insights with high fidelity and marked time savings. By operating as a coordinated ensemble of domain-specialized AI agents rather than a single generic model, it offers a scalable approach to reduce physician cognitive burden and to standardize guideline-concordant care planning in diverse, heavily pretreated oncology populations.