S2992
Invited Speaker
ESTRO 2026
Early evaluation indicated limited routine use. Key barriers included time burden, missing/non-routine input variables, unclear interpretation, and lack of explicit action thresholds. Models linked to a clear decision pathway and concrete actions were viewed more favorably and were more readily integrated into practice. Other models faced practical constraints (data availability) or diminished relevance due to evolving techniques. Take home messages: Successful CDS implementation requires stringent model selection for clinical utility, seamless EHR integration, and explicit decision trees with actionable cut-offs. Education and phased, clinician-centered rollout are essential to improve adoption. 5294 Nodal sparing strategies Charlotte Robert Radiotherapy, Gustave Roussy, Villejuif, France Radiotherapy-induced immune toxicity is increasingly recognized as a key determinant of clinical outcomes, particularly in the era of combined radiotherapy and immunotherapy. Radiation-induced lymphopenia (RIL) is common and clinically relevant, with severe lymphopenia reported in up to half of treated patients and consistently associated with poorer survival (HR > 1.5). Preserving immune function is therefore emerging as a major objective of modern radiotherapy. Recent advances now enable estimation of the radiation dose delivered to circulating blood and lymphocytes. Proposed approaches range from surrogate metrics to blood-flow–informed and compartment-based models designed to estimate cumulative dose to immune cells. These developments introduce the concept of dynamic immune-related organs at risk and challenge conventional dose– volume–based planning paradigms. They provide a strong rationale for lymphocyte-sparing strategies, including nodal sparing and reduction of the low-dose bath. Importantly, immune toxicity extends beyond lymphocyte dose metrics alone. Bone marrow dose is increasingly considered as a surrogate for irradiation of the myeloid compartment, while the thymus is gaining renewed attention as a determinant of immune health through its role in lymphocyte maturation and renewal. Integrating lymphocyte dose information with bone marrow and thymic considerations may therefore offer a more comprehensive assessment of treatment-induced immune impairment. Overall, immune-informed dosimetry supports a shift toward immuno-sparing and personalized
showcased to illustrate how the addition of SBRT to standard-of-care systemic therapy yields profound improvements in progression-free survival (PFS), radiographic PFS, and castration resistance-free survival. Crucially, the presentation will detail how these MDT benefits persist across both de novo and metachronous presentations, extending through both mHSPC and mCRPC states, without introducing prohibitive toxicities or safety risks. High-level evidence firmly establishes RT as a critical, life-prolonging pillar in the multidisciplinary management of mPC. By synthesizing data from recent trials, this presentation will provide a clear, clinically focused guide on patient stratification. Attendees will leave with actionable insights into translating these findings into daily practice, optimizing both local and systemic control, and understanding future directions—including the synergy of RT with novel androgen receptor pathway inhibitors.
5290 AI in radiotherapy workflows Wouter van Elmpt
Department of Radiation Oncology (Maastro), Maastricht University Medical Center, Maastricht, Netherlands Predictive and prognostic models (e.g., NTCP and radiomics) are widely developed in radiotherapy, yet real-world clinical uptake remains limited. Many tools fail to translate into patient-level decisions because they are hard to interpret and do not clearly indicate what action to take. We implemented and evaluate a pragmatic framework for implementing model-based clinical decision support (CDS) in radiotherapy, focusing on (1) selecting clinically meaningful models, (2) embedding them in the electronic health record (EHR), and (3) enabling actionable, individual-level interventions. Models were selected with clinicians across various disease sites (neuro-oncology, thorax, pelvis, head- and-neck), prioritizing models with modifiable treatment-related predictors. We built a minimum viable CDS prototype launched from the EHR (ChipSoft HiX), using automated transfer of available patient parameters and minimal manual input. Implementation followed a two-step evaluation: (1) group introduction using real cases, followed by a brief survey on expectations; (2) individual case use for evaluation purposes only (not for clinical decision- making), followed by a survey assessing usability, interpretability, workflow fit, and implementation barriers. A targeted educational program was also developed to improve understanding and trust.
Made with FlippingBook - Share PDF online