Hornstein, Nicholas J.

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.

3516 Background: ICIs produce durable benefit in MSI-H CRC, yet optimal treatment duration and monitoring strategies remain unclear. ctDNA enables minimally invasive assessment of molecular response during and after ICI. We evaluated real-world treatment patterns, outcomes, and longitudinal ctDNA dynamics in pts with MSI-H CRC treated with ICI. Methods: Pts with MSI-H mCRC treated with first-line ICI were identified from Natera’s Real-World Database with linked longitudinal ctDNA results (Signatera, Natera, Inc.) and commercial claims data (Forian’s Hybrid data ecosystem, CHRONOS). Overall survival (OS) was estimated using available mortality data (Veritas Data Research, Fact of Death Mortality Data Index). ctDNA status was assessed pre-ICI when available and longitudinally following ICI initiation. Outcomes included OS and time to next trmt (TTNT); ctDNA dynamics following ICI discontinuation were descriptively evaluated. Results: Among 559 pts treated with first line ICI (pembrolizumab [70.1%], ipilimumab+nivolumab [15.7%], and nivolumab [11.8%]), the median duration of ICI was 266 days. Baseline (pre-ICI) ctDNA results were available for 180 pts, including 38% who were treated with chemotherapy, and most pts (71.7%, 129/180) were ctDNA-positive (ctDNA(+)) prior to ICI initiation. Of these, 89.9% (116/129) had ctDNA-timepoints after the start of ICI, and 81.0% (94/116) achieved anytime ctDNA clearance (from ctDNA(+) to negative), while 19.0% (22/116) remained ctDNA(+). ctDNA(+) at anytime after the start of ICI (HR: 8.0, p < 0.0001) or on first result after the start of ICI (HR: 3.4, p < 0.0001) was associated with inferior OS compared ctDNA-negativity (ctDNA(-)) at these time points. Among pts whose first post-ICI test was ctDNA(+), 32% switched treatments, with a median TTNT of 302 days, compared to only 6.6% of pts (median TTNT 683 days) whose first post-ICI test was ctDNA(-). First post-ICI ctDNA status was associated with longer ICI duration (median 192 days for ctDNA(+) vs 303 days for ctDNA(-)), and ctDNA(+) was predictive of inferior OS (HR: 4.9; p < 0.001). ctDNA after the end of ICI was available for 392 patients; it was persistently negative in 77.2% (303/392), persistently positive in 12.8% (50/392), cleared after the end of ICI (positive to persistently negative) in 4.1%, and the remaining 5.9% had variable patterns (e.g. positive to negative to positive). Conclusions: In this large real-world cohort of MSI-H mCRC treated with ICI, longitudinal ctDNA dynamics strongly correlated with treatment durability, time to subsequent therapy, and OS. These findings support further evaluation of ctDNA to inform treatment duration and surveillance strategies in MSI-H mCRC.