Citations

8045 Background: Surgical resection is the primary therapy for stage I-II NSCLC but not performed in a subset of patients due to comorbidities or self-election, who instead undergo radiation with or without systemic therapy (RT±ST). Little is known about the demographic, tumoral, therapeutic, and quality-of-life (QOL) characteristics of this non-surgical group. Understanding their differences from the surgical group is essential for developing precision treatment strategies and tailored survivorship guidelines. Methods: In a prospectively enrolled (1997-2016) and followed (through 2025) cohort of 20,951 primary lung cancer patients, 6,486 (31%) had stage I-II NSCLC, including 5,359 (83%) treated surgically (Surgical) and 1,113 (17%) otherwise (Non-surgical). Preexisting diseases were grouped into cancer, lung, and other comorbidities. QOL data were available for 40% of all patients. Descriptive analysis of demographics, tumor features, treatment patterns, and QOL metrics, as well as survival analysis (median years and overall survival [OS] rates) were conducted. QOL measures, on overall and 15 symptom and functional domains, were scored 0 (worst) to 10 (best), with 1-unit difference considered clinically and statistically meaningful. Results: At time of diagnosis, compared to the Surgical, the Non-surgical patients were older (median 73 v 68 years), more likely to have smoked cigarettes (91% v 83%), squamous cell and unspecified NSCLC (53% v 28%), stage IIB (19% v 12%), and preexisting lung comorbidities (71% v 58%). Non-surgical patients more frequently received RT±ST than the Surgical (70% v 21%) and had shorter median survival (2.5 v 7.8) years. Within the Non-surgical, 28% had no cancer treatment (No-Tx), 44% RT-only, 7% ST-only, and 19% both (RT+ST); their respective median diagnosis ages were 74, 76, 68, 70 years; median survival were 1.7, 2.7, 2.2, 3.2 years; 5-year OS (95%CI) at 25% (21-31), 25% (21-29), 26% (17-38), 30% (25-37); and 10-year OS at 9% (7-14), 7% (5-10), 10% (5-21), 9% (6-14), respectively. RT-only group had highest rates of comorbidities (37-84%) and the No-Tx the lowest (31-66%). Although the overall QOL scored the same for all groups, RT-only patients reported the worst scores on multiple symptoms burden, and the No-Tx reported the least pain and lung cancer symptoms but the worst on mental well-being and social activities. Conclusions: Non-surgically managed stage I-II NSCLC cases present clinically distinct patients with the shortest median survival in No-Tx group and highest 5-year survival in RT+ST. RT-only patients were oldest, had the most comorbidities, and reported the worst symptom burdens. These differences highlight the need for more in-depth analyses and studies (considering age, histology, comorbidity, QOL, etc.) to support tailored therapeutic strategies and survivorship care for patients who decline or are not suitable for surgery.

Abstract Background Accurate and efficient dose calculation is essential for online adaptive planning in proton therapy. Deep learning (DL) has shown promising dose prediction results for pencil beam scanning proton therapy (PBSPT) in recent years, but existing DL‐based dose prediction methods still suffer from limited generalizability and an inability to effectively handle outlier clinical cases. This may lead to inaccurate dose delivery to targets or excessive irradiation to organs at risk (OARs), thereby compromising the safety and efficacy of online adaptive proton therapy. Purpose To design a physics‐aware and generalizable AI‐based PBSPT dose prediction method that incorporates underlying physics to enhance generalizability, particularly in handling outlier clinical cases. Methods This study analyzed PBSPT plans of 103 prostate (93 for training and 10 for testing) and 78 lung cancer patients (68 for training and 10 for testing) from our institution, with each case comprising CT images and structure sets. Using the doses generated by our Monte Carlo‐based dose engine as the reference standard, we compared three methods: the region of interest (ROI)‐based method, the beam mask and sliding window method, and the proposed noisy probing dose method, which rapidly generates a low‐statistics dose via uniformly weighted spots on an expanded spot‐placement target volume without optimization. To evaluate the generalizability of these methods to rare treatment planning scenarios, 12 cases with uncommon beam angles or prescription doses were used to assess their performance, which was evaluated using dose‐volume histogram (DVH) indices, 3D Gamma passing rates (3%/2 mm/10%), and Dice coefficients for dose agreement, while prediction times were measured to gauge model efficiency. Results The proposed noisy probing dose method consistently outperformed the ROI‐based and beam mask baselines across all evaluated metrics, with more accurate dose agreement and superior generalizability. For DVH indices, the noisy probing dose method achieved the smallest deviation in clinical target volume (CTV) dose coverage: in prostate cancer, CTV D98 deviation was reduced by 45% (from 0.53 ± 0.22 Gy [RBE] for ROI‐based) and 29% (from 0.41 ± 0.28 Gy [RBE] for beam mask) to 0.29 ± 0.06 Gy [RBE]; in lung cancer, similar improvements were observed, with CTV D98 deviations reduced to 0.34 ± 0.12 Gy [RBE]. The 3D Gamma passing rates improved to 99.65% ± 1.15% for prostate targets and 97.04% ± 1.17% for lung targets. The dice coefficients of the 90% iso‐dose lines were also the highest with the noisy probing dose method (prostate: 0.983 ± 0.005; lung: 0.967 ± 0.01). For the 12 outlier cases, the noisy probing dose method maintained superior generalizability, yielding higher 3D Gamma passing rates (prostate targets: 96.79% ± 0.83%, OARs: 94.29% ± 1.01%; lung targets: 93.38% ± 1.34%, OARs: 93.95% ± 1.32%), demonstrating robust generalizability to rare clinical scenarios. The dose predictions for all testing cases were completed within 0.3 seconds. Conclusions A novel noisy probing dose method was proposed for PBSPT dose prediction in prostate and lung cancer patients. By embedding more proton‐specific physics, this method demonstrated an improvement in the generalizability of dose prediction.

Abstract Background Electronic health record (EHR)–based prognostic modeling is increasingly used in oncology, yet incorporating pharmacogenomic (PGx) knowledge derived from experimental systems into clinical prediction frameworks remains challenging. This gap is driven by fundamental mismatches between controlled drug–mutation assays and heterogeneous, incomplete real-world clinical data. Methods We propose a representation transfer framework that integrates PGx embeddings learned from large-scale in vitro pharmacogenomic screens into patient-level EHR models. A frozen pharmacogenomic encoder is used to generate interaction-aware embeddings from patient mutation profiles and administered therapies, which are aggregated into a fixed-length PGx Complementarity Representation. These representations are incorporated into multimodal survival prediction models alongside standard clinical features. Performance was evaluated using systematic modality ablation analyses, attribution analyses, and exploratory unsupervised representation analyses. Results Integrating PGx embeddings yielded consistent performance improvements across all evaluated modality combinations. Relative gains were largest in modality-sparse settings, where baseline EHR features encode limited biological context, and were attenuated—but remained significant—in biologically enriched configurations. Attribution analyses indicated that PGx embeddings contributed non-redundant predictive signal beyond standard clinical features. Exploratory unsupervised analyses further demonstrated that the learned representations exhibit interpretable association patterns aligned with known therapeutic exposures and pathway-level associations. Conclusion These findings suggest that externally learned pharmacogenomic representations can be transferred into real-world EHR models as a context-dependent, non-redundant augmentation. By framing PGx knowledge as an interaction-aware representation rather than a mechanistic model, this work provides an informatics framework for integrating experimental pharmacogenomic data into clinical prediction tasks in a reproducible and interpretable manner.

98 Background: Prior authorization (PA) purports to control costs and ensure adherence to evidence-based guidelines but negatively impacts cancer care. Radiation oncology (RO) represents a minority of overall cancer spending yet bears a disproportionately high PA burden. Limited research exists on the effects of PA on patient outcomes and administrative workload within RO clinics. Methods: From 9/2024 to 10/2024, an online survey was open to U.S. members of the American Society for Radiation Oncology assessing experiences with PA including disposition (approval/denial/appeal including peer-to-peer [P2P] review) and perceived impact on patient care and administrative workload. Outcomes included treatment delays, changes from recommended treatment, and patient event including hospitalization, ED visits, permanent disability, or death. Descriptive statistics evaluated responses. Results: Of 736 responding radiation oncologists (16% response rate), 83% reported a rise in PA over the past three years, with 58% noting the burden has increased greatly. In the past year, 50% reported that more than half of their cases required PA, with a median approval rate of 75% and median overturn rate of 82% for denials. For more than half of their cases, nearly half of physicians (43%) noted that P2P was a futile effort to overturn PA denials due to insurance policies. Most (75%) reported that PA resulted in their practice needing to reallocate staff; 91% reported worsening burnout due to PA. P2P reviews were particularly burdensome, averaging 1.65 hours per week. A third (30%) reported that PA requirements had caused serious complications for their patients, including 7% who reported PA had contributed to patient death. Almost all (91%) reported PA-related treatment delays for some of their patients, with 58% indicating average delays of ≥5 days. Most (81%) reported that at times they had to use radiation techniques they considered less optimal due to PA, including 58% who were forced to deviate from national guidelines. The most common reasons for not appealing a PA denial were anticipated delays (73%), perceived futility based on prior PA experiences (62%), and administrative burden (45%). Conclusions: The administrative and clinical burdens associated with PA have significant implications for patient care, physician well-being, and healthcare resource allocation in RO. Treatment delays, compromised therapies, and adverse patient outcomes are commonly reported and underscore the need for PA reform to improve patient outcomes, reduce physician burnout, and improve efficiency.

Objectives: The practice parameter was revised collaboratively by the American College of Radiology (ACR), the American Brachytherapy Society (ABS), the American College of Nuclear Medicine (ACNM), the American Radium Society (ARS), the Society of Interventional Radiology (SIR), and the Society of Nuclear Medicine and Molecular Imaging (SNMMI). This document summarizes current evidence-based guidelines for the administration of Yttrium radioembolic therapy to the liver, including training requirements, evidence-based guidelines for administration, and safe practice for administration. Methods: This practice parameter was revised according to the process described under the heading The Process for Developing ACR Practice Parameters and Technical Standards on the ACR website (https://www.acr.org/ClinicalResources/Practice-Parameters-and-Technical-Standards) by the Committee on Practice Parameters—Interventional and Cardiovascular Radiology of the ACR Commission on Interventional and Cardiovascular, Committee on Practice Parameters and Technical Standards—Nuclear Medicine and Molecular Imaging of the ACR Commission on Nuclear Medicine and Molecular Imaging and the Committee on Practice Parameters—Radiation Oncology of the ACR Commission on Radiation Oncology in collaboration with the ABS, the ACNM, the ARS, the SIR, and the SNMMI. Results: This review seeks not to be a comprehensive discussion of radiotherapy to the liver, but rather, seeks to provide a parameter for safe and effective therapy. We discuss the qualifications of physicians involved in this therapy, basic indications, contraindications, procedural work-up, safe-handling, and regulatory requirement for the administration of selective internal radiation therapy to patients that are likely to benefit. The goal of this document is not to define which patients are best treated by these therapies, as this is best determined for individual patients after multidisciplinary review. A consistent and evidence-based approach to therapy, however, would benefit all patients who are offered this therapy. This document seeks to provide a framework for current best practices for the administration of the 2 currently available radioembolization devices. Conclusions: As Yttrium-90 radiotherapy to the liver occupies a growing role in the treatment of primary and metastatic liver cancer, this review seeks to assist clinicians of all involved specialties to optimize the efficacy and safety of these procedures.

Abstract Background The radiation oncology clinical practice involves many steps relying on the dynamic interplay of abundant text data. Large language models have displayed remarkable capabilities in processing complex text information. But their direct applications in specific fields like radiation oncology remain underexplored. Purpose This study aims to investigate whether fine‐tuning LLMs with domain knowledge can improve the performance on Task (1) treatment regimen generation, Task (2) treatment modality selection (photon, proton, electron, or brachytherapy), and Task (3) ICD‐10 code prediction in radiation oncology. Methods Data for 15 724 patient cases were extracted. Cases where patients had a single diagnostic record, and a clearly identifiable primary treatment plan were selected for preprocessing and manual annotation to have 7903 cases of the patient diagnosis, treatment plan, treatment modality, and ICD‐10 code. Each case was used to construct a pair consisting of patient diagnostics details and an answer (treatment regimen, treatment modality, or ICD‐10 code, respectively) for the supervised fine‐tuning of these three tasks. Open source LLaMA2‐7B and Mistral‐7B models were utilized for the fine‐tuning with the Low‐Rank Approximations method. Accuracy and ROUGE‐1 score were reported for the fine‐tuned models and original models. Clinical evaluation was performed on Task (1) by radiation oncologists, while precision, recall, and F‐1 score were evaluated for Task (2) and (3). One‐sided Wilcoxon signed‐rank tests were used to statistically analyze the results. Results Fine‐tuned LLMs outperformed original LLMs across all tasks with p value ≤ 0.001. Clinical evaluation demonstrated that over 60% of the fine‐tuned LLMs‐generated treatment regimens were clinically acceptable. Precision, recall, and F1‐score showed improved performance of fine‐tuned LLMs. Conclusion Fine‐tuned LLMs demonstrated statistically significant improvements over original LLMs upon three clinically important tasks in radiation oncology. This study explored the feasibility of applying fine‐tuned LLMs in radiation oncology, inspiring further development of utilizing LLMs to assist with radiation oncology tasks.

e14551 Background: Extracellular vesicles (EVs) are microscopic particles released by all cell types with a distinct fingerprint reflecting the cell of origin. In patients with prostate cancer, PSMA+EVs have been associated with clinical outcomes. In this study, we investigated tumor-derived EVs (tdEVs) using the tumor-specific CD326 marker as a novel biomarker for pancreas ductal adenocarcinoma (PDAC). Methods: The study included patients with localized PDAC treated on a prospective phase 2 clinical trial evaluating a 15-fraction chemoradiation (CRT) regimen. Blood samples were obtained after induction chemotherapy (C) (prior to CRT) and in follow-up. 20 healthy controls were included. High-resolution flow cytometry enumerated levels of fluorescent antibody labelled tumor-specific CD326 (EpCAM)-positive EVs. All samples were de-identified and randomly coded. The technologist responsible for EV analysis was blinded to patient information and clinical data. Each sample was run in triplicate. Flow cytometry data was analyzed with custom-built software to automatically determine and apply gates and generate reports with scatter plots and EV concentrations for each combination of markers and data summaries. Results: tdEV data were analyzed from 70 patients with pre-CRT samples. Patients had mean age 66 years (9.0 standard deviation), with 60% ECOG 0, 43% borderline resectable and 48% locally advanced. Patients had a median 8 cycles of C prior to CRT, and 70% proceeded to surgery after CRT. 2-year overall survival (OS) was 54%. Patient samples showed elevated total CD326 EVs compared to the healthy controls (p = 0.0005). Furthermore, patient CD326 EVs revealed two distinct subpopulations: small (120-1,000nm) and large (1,000-3,000nm) size CD326 EVs. No large CD326 subpopulation was found in healthy donor samples. Higher pre-CRT levels of total CD326 and large CD326 EVs were associated with decreased progression-free survival (PFS) (hazard ratio (HR) = 2.38, p = 0.013 and HR = 2.017, p = 0.021) on multivariable models with surgical resection. Resection was associated with increased PFS (p < 0.0001), and the interaction terms did not meet significance (p = 0.083 and p = 0.21). No differences in pre-CRT total or large CD326 EVs were identified between patients with or without surgical resection (p = 0.72 and p = 0.83). CA19-9 levels were not predictive of clinical outcomes in this trial and showed no association with EV levels. Conclusions: We present an association between clinical outcomes and tdEVs in PDAC, highlighting their potential as a novel, independent biomarker. If validated, tdEVs may offer an opportunity to personalize care and improve risk stratification.

Abstract Background Most oncology therapeutics have limited distribution into the brain, and developing strategies to overcome this limitation would be clinically impactful. While therapeutic radiation is often cited as a strategy to accomplish this, there are no published studies demonstrating the effect of radiation on drug distribution into the brain or brain tumors. Methods Mice were treated with brain radiation (6 Gy × 5, 4 Gy × 10; 40 Gy × 1) and dosed with drugs (levetiracetam, cefazolin, nedisertib, brigimadlin, apitolisib, or GNE-317) at times ranging from just prior to months after radiation. Plasma and tissue drug concentrations were measured by LC-MS/MS. Results Radiation did not significantly enhance drug delivery into brain tissue for levetiracetam, cefazolin, GNE-317, apitolisib, or nedisertib at any time post-radiation. Even a single, supra-therapeutic dose of radiation (40 Gy) did not significantly affect brain distribution of GNE-317 or apitolisib (P ≥ .07) from 16 to 160 hours post-radiation. For brigimadlin, radiation (6 Gy × 5) was associated with a modest but significant increase in drug accumulation only at 72 hours post-radiation (brain-to-plasma ratio 0.014 ± 0.006 vs. 0.025 ± 0.010, respectively; P = .04), but not at any other timepoint (24 hours, 15, 28, 94, 133, 183 days; P &gt; .05). Similarly, radiation (6 Gy × 5) of orthotopic tumors did not increase levels of brigimadlin in GBM10 or GBM108 or nedisertib in GBM108 (P &gt; .05). Conclusions Radiation had no meaningful impact on drug delivery into brain or brain tumors for the drugs tested.