Importance:Preoperative and perioperative nivolumab improve event-free survival in resectable non-small cell lung cancer. The role of adjuvant nivolumab after upfront surgery is unknown. Objective:To determine whether adjuvant nivolumab improves disease-free survival and overall survival in patients with resected non-small cell lung cancer with any tumor programmed death-ligand 1 (PD-L1) expression and in those with at least 50% PD-L1 expression. Design, Setting, and Participants:This open-label, randomized phase 3 study enrolled participants from May 2016 through September 2019, with median follow-up of 72.6 months at the data cutoff in December 2025. The study was conducted at 378 centers in the US National Clinical Trials Network. Patients were identified through a screening trial. Those with resected tumors at least 4 cm and/or who were lymph node positive (N1/N2) were eligible for inclusion after completion of planned standard adjuvant therapy if the tumor was adenocarcinoma without sensitizing sequence variants in EGFR and ALK or squamous cell carcinoma. Intervention:Patients were randomized in a 1:1 ratio to receive nivolumab 480 mg intravenously every 4 weeks for up to 1 year or standard care observation. Main Outcomes and Measures:Co-primary end points were disease-free survival in the intention-to-treat population and in those with tumoral PD-L1 expression at least 50%. Overall survival was examined if the corresponding test of disease-free survival was statistically significant. Results:A total of 466 patients (median age, 66 years; 241 [52%] male) were assigned to receive nivolumab and 469 (median age, 67 years; 245 [52%] male) to undergo standard care observation. The median duration of follow-up was 72.6 months. The trial was stopped for futility at 75% information. In the intention-to-treat population, median disease-free survival was 71.3 months with nivolumab and 68.8 months with observation (hazard ratio for progression or death, 0.97 [97% CI, 0.79-1.20]; [95% CI, 0.81-1.17]; 1-sided P = .39). In the subset of participants with PD-L1 of at least 50%, median disease-free survival was 89.8 months with nivolumab and 78.5 months with observation (hazard ratio for progression or death, 0.86 [98% CI, 0.55-1.34]; [95% CI, 0.59-1.25]; 1-sided P = .22). Conclusions and Relevance:Adjuvant nivolumab was not associated with improved disease-free survival in patients with resected non-small cell lung cancer without sensitizing EGFR and ALK alterations when given after planned adjuvant chemotherapy and/or radiotherapy. Trial Registration:Clinicaltrials.gov Identifier: NCT02595944.

ABSTRACT Purpose Clinical assessment of vertebral lesion quality (osteolytic, osteoblastic, mixed) remains subjective, with limited interobserver reliability. This study evaluated a novel application of 3D convolutional neural networks (3D-CNNs) for classifying lesion quality from CT volumes in metastatic cancer patients. Materials and Methods This retrospective study used CT data from 151 cancer patients planned for radiotherapy for metastatic spine disease (September 2020–July 2024). Leveraging vertebra- level expert annotations, we introduced an unconventional U-Net-based strategy converting coarse voxel-wise predictions into vertebra-level lesion classifications. The final dataset comprised 2,125 vertebrae across four classes (no lesion, osteolytic, osteoblastic, mixed), split into a 3-fold cross- validation set and an independent holdout test set. Model performance was benchmarked against a DenseNet121 baseline and a musculoskeletal radiologist, with Cohen’s kappa assessing inter- rater agreement. Results The 3D model achieved an ensemble accuracy of 84.7%, outperforming DenseNet121 (72.1%), with substantial gains in F1 score, precision, and balanced accuracy. It showed high concordance with the radiologist (Cohen’s kappa = 0.76) and comparable sensitivity and specificity across all lesion subtypes. We found both models and the radiologist to struggle with osteolytic lesions, reflecting the difficulty of distinguishing this class from age-related changes in vertebral bone density and architecture caused by benign bone lesions, age-related systemic skeletal disorders and cancer treatments. Conclusions 3D-CNNs trained with vertebra-level labels can accurately and reliably classify vertebral metastatic lesion quality from CT scans, offering a scalable path toward automated characterization of metastatic spine disease to support clinical decision-making and large-scale radiomics research.

Background/Objectives: Outcomes in non–small cell lung cancer (NSCLC) remain heterogeneous, even within stage and molecular subtypes. We evaluated whether a Computational Histology Artificial Intelligence (CHAI) biomarker, derived solely from the diagnostic hematoxylin-and-eosin (H&E) whole-slide image (WSI), provides prognostic information independent of established clinicopathologic factors. Methods: Using The Cancer Genome Atlas (TCGA) lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) projects, 914 patients with an evaluable diagnostic digitized whole-slide images (WSI) and outcomes data were stratified by stage and histology and randomly split into a development set (30%, N = 270) and a held-out validation set (70%, N = 644). A CHAI histologic signature was developed and locked on the development set and applied without modification to the validation set. In development, image features were extracted from H&E-stained WSIs and used to train a Cox proportional hazards model to output a CHAI prognostic biomarker score and then dichotomized into high (CHAI positive (+))- and low (CHAI negative (−))-risk groups. The primary endpoint was overall survival (OS), and the secondary endpoint was progression-free interval (PFI). Associations were assessed by Kaplan–Meier and Cox proportional-hazards models and adjusted for traditional clinicopathologic and genomic variables. Results: A total of 644 patients were included in the validation cohort. Median age was 68 (IQR: 60–74), 388 (60%) were male, 317 (49%) had LUAD, 327 (51%) had LUSC, and 57 (9%) were never smokers. CHAI-positive patients had significantly worse OS than CHAI-negative patients (3-year OS: 53% vs. 69%; log-rank p < 0.001). The biomarker remained significantly associated with OS (HR 1.70, 95% CI 1.29–2.24, p < 0.001) and PFI (HR 1.39, 95% CI 1.05–1.84, p = 0.023) in multivariable analysis. The effect was consistent across both histologies, and a model combining the CHAI score with clinicopathologic variables was well-calibrated in validation. The biomarker was not associated with stage, age, sex, or histologic subtype, and its adjusted prognostic effect was unchanged after relevant genomic alterations were added to the model. CHAI-positive tumors were, however, enriched for KRAS mutations (21% vs. 11% wild-type, p adjusted = 0.005). Conclusions: An H&E-only CHAI NSCLC biomarker was developed and provided independent prognostic stratification in a held-out NSCLC validation cohort. Such a biomarker has the potential to refine risk stratification in NSCLC.

IntroductionGiven the high prevalence of vertebral fractures following radiotherapy in patients with metastatic spine disease, torso muscle segmentation is necessary for biomechanical modeling of vertebral loading, permitting individualized evaluation of fracture risk.MethodsIn this study, we developed and validated a deep-learning model for full volumetric segmentation of the thoracic and abdominal spinal musculature in cancer patients with metastatic spine disease from sparsely annotated clinical CT image data. We obtained CT data for 148 metastatic spine disease patients undergoing radiotherapy treatment, and an external set of randomly selected 30 subjects from the National Lung Screening Trial. We extracted 1924 axial CT images at the midpoint of each vertebral level (T4 to L4) and manually labeled the key extensor and flexor muscles (up to 8 muscles per side) at each level. We trained a 2D nnU-Net deep-learning (DL) model to segment each muscle and, using these sparse annotations, trained the model to segment each muscle’s 3D volume per spine. Two experienced radiologists independently and blindly evaluated the anatomical fidelity of the segmentations using a Likert scale, for 1) manual- and 2) DL-segmentation, 3) random test samples from the muscle’s 3D volume and 4) an external NLST CT data.ResultsThe DL method achieved comparable performance to manual segmentations with a mean Dice score above 0.769. Mann-Whitney test analysis showed that the radiologist ratings of DL-generated muscle segmentations were noninferior to the manual segmentation for each muscle.DiscussionDemonstrating excellent performance for rapid, high-anatomical fidelity 3D segmentation of the main flexor, extensor, and stabilizing thoracolumbar muscles, the DL model from clinical CT scans, this development holds significant potential for reducing the manual effort required to generate individualized musculoskeletal models in cancer patients.

Purpose While NRG-LU005 showed no overall survival (OS) advantage by adding concurrent and adjuvant immunotherapy to chemoRT (CRT), it reported longer OS in an exploratory comparison of twice-daily (BID) over once-daily (QD) radiation (RT). This planned analysis assessed longitudinal PROs. The pre-specified hypothesis was that clinically-meaningful-decline (CMD) in longer term PROs at 15 months would be lower with immunotherapy, as measured by the Functional-Assessment-of-Cancer-Therapy:Trial-Outcome-Index (FACT-TOI). Methods Patients (N=544) were randomized to standard CRT (platinum/etoposide + thoracic-RT [45Gy-BID or 66Gy-QD]) +/- atezolizumab starting cycle-2 of chemotherapy. Stratification factors (SFs) included RT-schedule, chemotherapy-type, sex, and performance-status(PS). PROs included validated instruments: FACT-TOI, EQ-5D-5L, and PROMIS-Fatigue (at baseline, end-of-CRT, and at 3, 6, 15, and 21 months). EQ-5D-5L scores were followed through 24 months. CMD and longitudinal trends were evaluated; multivariable logistic regression analysis (MVA) adjusted for treatment-arm, SFs, and baseline FACT-TOI. Results PRO compliance exceeded 85% at baseline and stabilized at 60-68% through 21 months. CMD in FACT-TOI was not significantly different by treatment arm at 15 months, though fewer patients on the immunotherapy arm had CMD at 21 months. EQ-5D-5L and PROMIS-fatigue were similar in both arms. BID-RT was associated with better FACT-TOI than QD-RT across all timepoints. On MVA, significant predictors of lower CMD for FACT-TOI (beyond baseline FACT-TOI) included BID-RT (at end-of-CRT, 15, 21 months), PS (at end-of-CRT), cisplatin (at 15 months), and immunotherapy (at 21 months). Conclusions The addition of atezolizumab to chemoradiation was not associated with a significant change in CMD in FACT-TOI at 15 months, with similar EQ-5D-5L and fatigue scores. While not randomized for RT-schedule, this analysis suggests that BID-RT (vs QD-RT) was associated with a consistently more favorable PRO trajectory.

BACKGROUND:There is renewed interest in resection of Stage III (N2+) NSCLC given impressive outcomes with neoadjuvant/perioperative chemoimmunotherapy (ChIO). We report surgical outcomes from an exclusively N2+ NSCLC clinical trial following chemotherapy + durvalumab. METHODS:This was a single arm phase II trial enrolled at 9 US hospitals. Eligible patients had resectable stage III NSCLC, pathologically proven N2+. Patients received 4 cycles of platinum doublet + durvalumab followed by lobectomy or greater, and adjuvant durvalumab for 1 year. Surgical approach, margins, extent of lymphadenectomy, complications, and treatment timeliness were analyzed. RESULTS:From 2021-2023, 37 patients were enrolled; 30 patients underwent resection (81%). Surgical outcomes are notable for R0 resection in 28/30 patients (93.3%), pneumonectomy rate of 6.7% (2/30), median stay 3.5 days, no mortality at 30 and 90 days. Minimally invasive surgery was possible in 19/30 (63.3%: 2 VATS, 17 robotic), with 2 conversions to thoracotomy (total thoracotomy rate: 11/30: 36.7%). Increased surgical difficulty was reported in 14/30 (46.7%). Median interval from neoadjuvant therapy to surgery was 46 days, and from surgery to adjuvant therapy was 35 days. All 23 patients recommended for adjuvant therapy received it. CONCLUSIONS:This trial provides insight into surgical outcomes for prospectively documented N2+ NSCLC treated with neoadjuvant ChIO. Surgery in this challenging scenario was accomplished with no mortality, high rates of minimally invasive surgery, R0 resection, and lobectomy, with prompt return to oncologic therapy. Following ChIO, resection of N2+ NSCLC can be achieved with excellent outcomes and warrants stronger consideration within multidisciplinary NSCLC care.

Purpose:This study investigated the effect of bone metastasis on the biomechanical environment of human vertebrae in patients with metastatic spine disease through the metric of load-to-strength ratio (LSR). Specifically, we compared the patients' LSRs to age and sex-similar noncancer controls from the Framingham Heart Study. Methods:Derived from clinical CT data of 135 metastatic spine disease patients planned for radiotherapy and 246 normative controls from the Framingham Heart Study, individualized spinal musculoskeletal models and vertebral strength estimates were used to compute level-specific LSR under natural standing and three weight-holding conditions (standing + weight, flexion + weight, and lateral bending + weight). Results:Adjusted for age, BMI, and spinal region, osteosclerotic and mixed lesion vertebrae had higher strength than osteolytic and control vertebrae. The musculoskeletal models suggested breast, prostate, and male lung cancer patients had higher compressive vertebral loading, and female lung cancer patients had lower compressive vertebral loading than controls. Male patients had higher standardized LSRs in natural standing, while female patients had lower LSRs for all activities than controls. Independent of sex, vertebrae with osteosclerotic and mixed bone metastasis had lower LSRs than controls, while, for osteolytic bone lesions, males had higher and females lower LSRs than controls. Vertebrae with no observed lesion on CT had higher LSRs than controls in males and lower LSRs in females. Discussion:Our findings highlighted that primary cancer and lesion type differentially affected task-specific vertebral loading and strength, thus modifying the vertebral LSRs. Sex-mediated differences in LSRs between FHS controls and vertebrae with no observed metastatic lesions suggest that considering the latter as "normal" should be taken with care. Our initial assessment supports further examination of whether vertebral LSR measurements are associated with vertebral risk and, if so, what threshold values indicate risk. Level of Evidence:3.

QuestionCan nivolumab improve disease-free and overall survival in patients with resected non-small cell lung cancer (NSCLC) after standard care chemotherapy and/or radiotherapy?FindingsIn this phase 3 randomized clinical trial of resected NSCLC greater than or equal to 4 cm and/or lymph node positive without sensitizing EGFR or ALK sequence variants, a total of 466 patients were assigned to receive nivolumab and 469 to undergo observation. There were no statistically significant differences in the co-primary end points of median disease-free survival among patients who received nivolumab vs observation (hazard ratio for progression or death, 0.97) or in the subgroup of patients with tumoral programmed death-ligand 1 greater than or equal to 50% (hazard ratio for progression or death, 0.86).MeaningAdjuvant nivolumab did not reduce the risk of disease recurrence in patients with resected NSCLC who completed planned standard care adjuvant therapy. ImportancePreoperative and perioperative nivolumab improve event-free survival in resectable non-small cell lung cancer. The role of adjuvant nivolumab after upfront surgery is unknown.ObjectiveTo determine whether adjuvant nivolumab improves disease-free survival and overall survival in patients with resected non-small cell lung cancer with any tumor programmed death-ligand 1 (PD-L1) expression and in those with at least 50% PD-L1 expression.Design, Setting, and ParticipantsThis open-label, randomized phase 3 study enrolled participants from May 2016 through September 2019, with median follow-up of 72.6 months at the data cutoff in December 2025. The study was conducted at 378 centers in the US National Clinical Trials Network. Patients were identified through a screening trial. Those with resected tumors at least 4 cm and/or who were lymph node positive (N1/N2) were eligible for inclusion after completion of planned standard adjuvant therapy if the tumor was adenocarcinoma without sensitizing sequence variants in EGFR and ALK or squamous cell carcinoma.InterventionPatients were randomized in a 1:1 ratio to receive nivolumab 480 mg intravenously every 4 weeks for up to 1 year or standard care observation.Main Outcomes and MeasuresCo-primary end points were disease-free survival in the intention-to-treat population and in those with tumoral PD-L1 expression at least 50%. Overall survival was examined if the corresponding test of disease-free survival was statistically significant.ResultsA total of 466 patients (median age, 66 years; 241 [52%] male) were assigned to receive nivolumab and 469 (median age, 67 years; 245 [52%] male) to undergo standard care observation. The median duration of follow-up was 72.6 months. The trial was stopped for futility at 75% information. In the intention-to-treat population, median disease-free survival was 71.3 months with nivolumab and 68.8 months with observation (hazard ratio for progression or death, 0.97 [97% CI, 0.79-1.20]; [95% CI, 0.81-1.17]; 1-sided P = .39). In the subset of participants with PD-L1 of at least 50%, median disease-free survival was 89.8 months with nivolumab and 78.5 months with observation (hazard ratio for progression or death, 0.86 [98% CI, 0.55-1.34]; [95% CI, 0.59-1.25]; 1-sided P = .22).Conclusions and RelevanceAdjuvant nivolumab was not associated with improved disease-free survival in patients with resected non-small cell lung cancer without sensitizing EGFR and ALK alterations when given after planned adjuvant chemotherapy and/or radiotherapy.Trial RegistrationClinicaltrials.gov Identifier: NCT02595944 This randomized clinical trial examines whether adjuvant nivolumab improves disease-free survival and overall survival in patients with resected non-small cell lung cancer with any tumor programmed death-ligand 1 expression.

Pathologic vertebral fractures (PVF) are common and serious complications in patients with metastatic lesions affecting the spine. Accurate assessment of cancer patients' PVF risk is an unmet clinical need. Load-to-strength ratios (LSRs) evaluated in vivo by estimating vertebral loading from biomechanical modeling and strength from computed tomography imaging (CT) have been associated with osteoporotic vertebral fractures in older adults. Here, for the first time, we investigate LSRs of thoracic and lumbar vertebrae of 135 spine metastases patients compared to LSRs of 246 healthy adults, comparable by age and sex, from the Framingham Heart Study under four loading tasks. Findings include: (1) Osteolytic vertebrae have higher LSRs than osteosclerotic and mixed vertebrae; (2). In patients' vertebrae without CT observed metastases, LSRs were greater than healthy controls. (3) LSRs depend on the spinal region (Thoracic, Thoracolumbar, Lumbar). These findings suggest that LSRs may contribute to identifying patients at risk of incident PVF in metastatic spine disease patients. The lesion-mediated difference suggests that risk thresholds should be established based on spinal region, simulated task, and metastatic lesion type.

Purpose:Given the high prevalence of vertebral fractures post-radiotherapy in patients with metastatic spine disease, accurate and rapid muscle segmentation could support efforts to quantify muscular changes due to disease or treatment and enable biomechanical modeling for assessments of vertebral loading to improve personalized evaluation of vertebral fracture risk. This study presents a deep-learning approach for segmenting the complete volume of the trunk muscles from clinical CT images trained using sparsely annotated data. Materials and Methods:we extracted 2,009 axial CT images at the midpoint of each vertebral level (T4 to L4) from clinical CT of 148 cancer patients. The key extensor and flexor muscles (up to 8 muscles per side) were manually contoured and labeled per image in the thoracic and lumbar regions. We first trained a 2D nnU-Net deep-learning model on these labels to segment key extensor and flexor muscles. Using these sparse annotations per spine, we trained the model to segment each muscle's entire 3D volume. Results:The proposed method achieved comparable performance to manual segmentations, as assessed by expert radiologists, with a mean Dice score above 0.769. Significantly, the model drastically reduced segmentation time, from 4.3-6.5 hours for manual segmentation of 14 single axial CT images to approximately 1 minute for segmenting the complete thoracic-abdominal 3D volume. Conclusion:The approach demonstrates high potential for automating 3D muscle segmentation, significantly reducing the manual intervention required for generating musculoskeletal models, and could be instrumental in enhancing clinical decision-making and patient care in radiation oncology.

BACKGROUND:Adequate patient awareness and understanding of cancer clinical trials is essential for trial recruitment, informed decision making, and protocol adherence. Although large language models (LLMs) have shown promise for patient education, their role in enhancing patient awareness of clinical trials remains unexplored. This study explored the performance and risks of LLMs in generating trial-specific educational content for potential participants. METHODS:Generative Pretrained Transformer 4 (GPT4) was prompted to generate short clinical trial summaries and multiple-choice question-answer pairs from informed consent forms from ClinicalTrials.gov. Zero-shot learning was used for summaries, using a direct summarization, sequential extraction, and summarization approach. One-shot learning was used for question-answer pairs development. We evaluated performance through patient surveys of summary effectiveness and crowdsourced annotation of question-answer pair accuracy, using held-out cancer trial informed consent forms not used in prompt development. RESULTS:For summaries, both prompting approaches achieved comparable results for readability and core content. Patients found summaries to be understandable and to improve clinical trial comprehension and interest in learning more about trials. The generated multiple-choice questions achieved high accuracy and agreement with crowdsourced annotators. For both summaries and multiple-choice questions, GPT4 was most likely to include inaccurate information when prompted to provide information that was not adequately described in the informed consent forms. CONCLUSIONS:LLMs such as GPT4 show promise in generating patient-friendly educational content for clinical trials with minimal trial-specific engineering. The findings serve as a proof of concept for the role of LLMs in improving patient education and engagement in clinical trials, as well as the need for ongoing human oversight.

418 Background: Research on the genetic profile of cribriform pattern (CP) and intraductal carcinoma (IDC) subtypes of aggressive prostate cancer (PCa) is limited. We evaluated germline mutations in CP/IDC PCa patients receiving radiation therapy (RT) to assess the role of familial high-risk (FHR) and DNA damage repair (DDR) mutations associated with these pathologies and clinical outcomes. Methods: This multi-institutional study included CP/IDC PCa patients treated with RT who consented to germline whole exome sequencing (WES). WES data was annotated using Ensembl Variant Predictor, ClinVar, and OncoKB to identify pathogenic/oncogenic mutations. Mutational frequencies were calculated in R. Clinical data and disease progression (biochemical recurrence [BCR], local relapse [LR], distant metastases [DM]) were assessed. Gene and clinical outcome associations were evaluated using the Kaplan-Meier method and log-rank test (p < 0.05 for significance, Benjamini-Hochberg FDR of 0.1 used for multiple hypothesis testing). Results: Of 1,392 CP/IDC patients treated with RT between 2010-2024, 80 had germline sequencing, and 49 had WES data (12 CP, 24 IDC, 13 CP+IDC); 45 were localized at diagnosis (28 high-, 16 intermediate-, 1 low-risk). Median age at diagnosis was 64 years. 39 (80%) received RT post-radical prostatectomy (RP, salvage) and 10 (20%) for intact prostate (definitive). Median follow-up (FU, months) was 64.5 after diagnosis and 60.3 after RT. At last FU post-RT, 14 (29%) exhibited BCR, 12 (24%) had LR, and 9 (18%) had DM, with progression rates of 83% in CP and IDC, and 77% in CP+IDC. Five-year OS from diagnosis was 94.8% (95% CI, 80.4-98.7), and BCR-free survival (defined as no BCR post-salvage or post-definitive RT) was 69% (95% CI, 47.8-83). Of the 98 FHR and 280 DDR genes analyzed, 4 (8%) and 20 (40%) patients had confirmed mutations, respectively. Of the 24 patients with germline mutations, 20 (83%) had IDC pathology. DDR mutations associated with improved BCR-free survival, HR 0.14, 95%CI 0.03-0.70, p=0.007. Notably, 19 patients had a pathogenic MSH3 deletion, which was associated with improved BCR-free survival, HR 0.17, 95%CI 0.03-0.83, p=0.02. Additionally, 20 patients had pathogenic stop-gains/deletions, 17 (85%) of whom had IDC pathology. Five oncogene mutations were identified in DDR genes ATR, TP53 and ATM, and tumor suppressor genes SHDA and DPYS previously implicated in PCa, all in IDC samples. Conclusions: Pathogenic germline mutations, particularly DDR variants, are significantly associated with aggressive IDC pathology. Continued evaluation in non-CP/IDC patients undergoing RT is needed to examine associations between mutations, subtypes, and outcomes. Mutations in DDR genes, such as MSH3 deletion, were linked to improved BCR-free survival, highlighting the potential role of DDR alterations in therapeutic response of RT patients with aggressive PCa subtypes.

To identify a dosimetric predictor for lung doses in stereotactic body radiotherapy (SBRT) for peripheral lung cancer and demonstrate the utilities of the predictor in aiding plan evaluation and treatment modality selection. We performed a retrospective review of 108 plans that were previously treated on a TrueBeam in our institute. The cohort was further stratified into three subgroups based on the level of the chest wall (CW) involvement. The achieved lung metrics including the lung V20, V5 and mean lung dose (MLD) were evaluated against an anatomy-based parameter RPTV/Lungs-ratio of the planning target volume (PTV) to lung volume. Linear regression and prediction interval were used to identify outlier plans that had "suboptimal" lung doses. Re-optimization using alternative strategies to improve the lung doses were carried out for these outlier cases. We demonstrated the utility in treatment platform selection through a comparison with a magnetic resonance (MR) guided system, MRIdian. Strong correlation (R2 ∼ 0.9) existed between the lung doses and RPTV/Lungs for each subgroup. Increasing CW involvement progressively reduced lung doses as is evident by the regression coefficient, which dropped from 414.68 when the PTV-CW distance was > 5 mm, to 274.34 when > 10% of PTV was within CW. Three outlier plans with high lung doses were identified for replanning. Alternative optimization approaches were successful in reducing the lung V20, V5 and MLD by 0.7%, 1.3% and 0.4 Gy on average, respectively. The correlation showed different characteristics between a TrueBeam and a MRIdian, enabling treatment modality selection. The achieved lung metrics in peripheral lung SBRT were highly correlated to RPTV/Lungs, which can be used as a simple and practical tool to guide plan evaluation and treatment modality selection.

e20675 Background: Clonal hematopoiesis (CH) results from fitness-enhancing mutations in hematopoietic stem cells. Although many CH SVs are in heme-related genes, some are in genes with solid tumor relevance ( TP53, ATM and CHEK2 ). In this study we report the prevalence of CH SVs in driver genes, emerging targets, and other clinically relevant genes in NSCLC. Methods: For 1813 pan-cancer LBx (572 NSCLC), plasma cell-free DNA and white blood cell (WBC) DNA were sequenced in parallel at equal depth with FoundationOne Liquid CDx. A novel variant origin prediction (VOP) algorithm trained and validated using WBC was used to classify origins for SVs detected in 21,456 NSCLC LBx submitted 8/2020-7/2024. Three possible variant origins were predicted using sequencing features including fragmentomics: germline, tumor-somatic, and CH. Results: Classic NSCLC driver SVs were not detected in WBC except for a KRAS G12D SV in a NSCLC patient (contrasted with 80 KRAS G12X SVs classified as tumor-somatic), and a KRAS G12S and a MET ex14 skipping SV in WBC from non-NSCLC patients. In contrast, other mutations in driver genes ( EGFR T415K, A743T, E829K, V843I, and A864V; KRAS V14I, I24N, A59G, and A146T; BRAF class 2/3 mutations) were more likely to be CH. Table shows SVs detected in WBCs and predicted to be CH-derived in the wider NSCLC LBx cohort. SVs in genes representing emerging targets or having clinical implications in NSCLC were detected in WBC at the following frequencies: 54 (3.0%) NF1 , 16 (0.9%) TP53 Y220C, 8 (0.4%) RB1 , 4 (0.2%) STK11 , 3 (0.2%) KEAP1 , and 3 (0.2%) SMARCA4 . In the larger NSCLC LBx cohort, frequencies of LBx with predicted tumor-somatic SVs vs only CH SVs detected were: NF1 (4.3% vs 3.9%), TP53 Y220C (0.5% vs 1.2%), RB1 (5.6% vs 0.3%), STK11 (7.6% vs 0.4%), KEAP1 (7.0% vs 0.3%), and SMARCA4 (3.6% vs 0.2%). Conclusions: This study shows that detection of currently targetable driver SVs in NSCLC is not confounded by CH in LBx, except rare KRAS G12X and MET ex14 skipping mutations. However, some emerging targets such as TP53 Y220C, BRAF class 2 mutations, STK11, and SMARCA4 SVs are CH-derived at appreciable rates. Ruling out CH origin through tissue testing, equal depth WBC sequencing, or algorithmic prediction of CH origin may be advisable. SVs in targetable genes CH SVs in WBC, non-NSCLC (n = 1241) CH SVs in WBC, NSCLC (n = 572) Tumor-somatic SVs, not detected in WBC, NSCLC (n = 572) NSCLC LBx with VOP-predicted CH SVs (n=21,456)n (%) NSCLC LBx with VOP-predicted tumor-somatic SVs (n=21,456) n (%) EGFR L858R, G719X, S768I, L861Q, ex19 deletions, ex20 insertions 0 0 70 6 (0.0%) 1940 (9.0%) EGFR, other 3 2 28 87 (0.4%) 667 (3.1%) BRAF V600E 0 0 7 8 (0.0%) 206 (1.0%) BRAF, other, class 2/3 8 2 7 130 (0.6%) 391 (1.8%) MET ex14 skipping 1 0 5 18 (0.1%) 256 (1.2%) MET, other 1 0 1 23 (0.1%) 51 (0.2%) KRAS G12X 1 1 80 14 (0.1%) 2891 (13.5%) KRAS, other 1 3 14 43 (0.2%) 696 (3.2%) ERBB2 0 0 6 9 (0.0%) 305 (1.4%)

Diffuse midline glioma (DMG) is an incurable pediatric brain tumor, with radiotherapy offering only transient benefit before inevitable recurrence. While recent studies have revealed the cellular complexity of DMG, the relationship between intratumoral heterogeneity and disease progression — particularly under therapeutic pressure — remains poorly understood. In this study, we integrate longitudinal single-cell- and spatial transcriptomics with single-cell chromatin accessibility profiling and radiomics to investigate how distinct DMG cell states and microenvironmental components evolve in response to treatment, contributing to resistance and the emergence of therapy-persistent niches. We applied this multi-omic approach to a unique cohort of 10 matched diagnostic and autopsy DMG samples obtained from children enrolled in the PNOC023 clinical trial. Spatial transcriptomic profiling revealed that post-treatment autopsy tissues showed consistent enrichment of oligodendrocyte progenitor cell-like malignant populations, suggesting that this stem-like subpopulation is selectively retained or expanded following radiotherapy in DMG. Moreover, integration of single-cell RNA and ATAC sequencing with cerebrospinal fluid microRNA analyses further revealed molecular signatures of resistance and uncovered minimally invasive biomarkers of treatment response. To complement our molecular analyses, we applied radiomic profiling of longitudinal MRI scans to non-invasively monitor changes in tumor heterogeneity over the course of treatment in this patient cohort. In parallel, we developed a data-driven mathematical model using patient-derived DMG xenografts to simulate tumor cell responses to radiotherapy and generate quantitative predictions of preclinical treatment outcomes. We employed this model to explore alternative radiation fractionation schemes and identify optimal dosing strategies that most effectively deplete therapy-resistant tumor cell populations in DMG. Together, these integrated spatial and molecular datasets provide a high-resolution framework for understanding how radiotherapy remodels the DMG tumor ecosystem. By identifying persistent cellular populations and spatial niches that survive treatment, this work lays the groundwork for future precision therapies aimed at targeting the cellular reservoirs that drive DMG recurrence.

Lung cancer patients often experience increased metastasis formation after radiotherapy. However, it is incompletely understood whether radiation affects the migratory behavior of tumor cells and how altered radiotherapy schedules might mitigate this risk. To address these questions, we performed live-cell microscopy experiments to profile changes in cell migration during radiation across 12 cancer cell lines and developed a predictive computational modeling platform describing tumor volume and dissemination during radiotherapy. Using this platform, we identified optimal fractionation schedules and then performed extensive in silico clinical trials, establishing that our optimized schedules substantially reduce metastatic seeding relative to the standard of care schedule. Training transformer models on the in silico clinical trial data enabled us to recover mechanistic parameters with high accuracy, demonstrating that the features determining optimal radiotherapy can be inferred from longitudinal tumor data. Our integrative predictive approach enables the rational design of optimum clinical trials across indications. ### Competing Interest Statement D.K. is a consultant for AstraZeneca and Genentech/Roche. D.K. declares that none of these relationships are directly or indirectly related to the content of this manuscript. F.M. is a co-founder of and has equity in Harbinger Health, has equity in Zephyr AI, and serves as a consultant for both companies. She is also on the board of directors of Recursion Pharmaceuticals. F.M. declares that none of these relationships are directly or indirectly related to the content of this manuscript.

TPS8113 Background: There are currently three approved approaches for patients with resectable NSCLC including neoadjuvant chemoimmunotherapy, adjuvant chemoimmunotherapy and perioperative treatment with neoadjuvant chemoimmunotherapy followed by adjuvant immunotherapy. All regimens were approved after showing event-free survival (EFS) or disease-free survival (DFS) benefit with the addition of immunotherapy to chemotherapy compared to chemotherapy alone. Each approach has its benefits and risks. Starting immunotherapy prior to surgery may improve treatment compliance and efficacy of immunotherapy. Nevertheless, neoadjuvant chemoimmunotherapy may result in missing an opportunity for curative surgery and increase the complexity of tumor resection. PROSPECT-LUNG (NCT06632327) is a randomized study evaluating whether starting chemoimmunotherapy before or after surgery leads to better outcomes. Methods: This is a randomized phase 3 trial in which patients will be randomized 1:1 to surgery followed by chemoimmunotherapy (adjuvant arm) or neoadjuvant chemoimmunotherapy followed by surgery and adjuvant therapy (perioperative arm). Patients with histologic or cytologic confirmation of surgically resectable stage IIA-IIIB NSCLC (per AJCC 9th edition) or stage IIA to IIIB per AJCC 8 th edition up to single ipsilateral mediastinal station (N2a), ECOG PS ≤ 2 (or Karnofsky ≥ 60%), no prior treatment for NSCLC and no previous malignancy within 3 years are eligible. The dual primary endpoints are real-world event free survival (rwEFS) defined as date from randomization to date of the first of the following events: failure to undergo resection for any reason, progression prior to surgery that precludes resection, recurrence or progression at any time after surgery or death from any cause, and overall survival (OS) defined as time from randomization to death from any cause. The target accrual is 1,100 patients assuming one-sided type I error of 0.03 for OS endpoint and 0.02 for rwEFS endpoint. This sample size will enable the detection of a 3-year rwEFS improvement from 55% in the adjuvant therapy arm to 64% in the perioperative arm with an 84% power. This sample size would detect an HR of 0.73 (improvement in median OS from 8.1 to 11 years in favor of the perioperative arm, 5-year OS from 65% in the adjuvant arm to 73% in the perioperative arm, assuming exponential survival) with 83.6% power. The study has a pragmatic design with minimal data collection, reporting of adverse events that lead to discontinuation of therapy, hospitalization or death only. It allows providers to choose therapy per standard of care (FDA approved or on NCCN), includes patients with ECOG performance status 2, permits use of local laboratory testing and imaging studies and determination of recurrence/progression will be done by local treating physicians with no use of RECIST. Clinical trial information: NCT06632327 .

A validated AI system accurately segments cardiac substructures, reproduces dose-outcome relationships, enables large-scale surveillance, and point-of-care alerts for high-risk patients. Automated cardiac dose monitoring could facilitate adoption of coronary-sparing therapy and follow-up.