Citations

Abstract Objective Evaluate how RAG architecture, including corpus structure, retrieval strategy, and pipeline complexity, affects LLM-based medical problem solving and knowledge retrieval in sleep medicine. Materials and Methods We benchmarked four open-source LLMs (Llama-3-8B, Llama -3 -70B, Qwen 2.5-14B, and Qwen 2.5-235B) using a knowledge base of five sleep medicine textbooks. We compared performance across three dimensions: corpus structure (raw text vs table-of-contents aligned.), retrieval strategy (dense embedding vs hybrid sparse-dense), and pipeline complexity (baseline vs augmented). Evaluation metrics included board-style multiple choice question (MCQ) accuracy and clinical case vignette diagnostic ranking. Results RAG improved MCQ accuracy for all models. Llama-8B saw the largest gain of 10.6% (61.8% to 72.4%), while Qwen-235B reached 87.3%. In clinical cases, Llama-8B accuracy dropped by 7.1% when using raw text and dense retrieval due to context noise. This was corrected by using structured hybrid configurations. Hybrid retrieval consistently outperformed dense-only methods. Overall, structured corpora improved primary diagnosis accuracy by 6.1% on average, with Qwen-235B reaching a peak 10.2% increase. Discussion RAG effectiveness depends on the balance between model size and data structure. Large models handle uncurated text well, but smaller models are easily distracted by irrelevant data. Hybrid retrieval is necessary to maintain precision with specialized medical terms. A structured corpus paired with a baseline hybrid pipeline offers the best stability and speed for clinical use. Conclusion Rigorous data curation and hybrid retrieval are as essential as model scale for deploying safe, guideline-compliant AI in sleep medicine.

ABSTRACT Behavioral and social factors (BSFs) substantially influence the risk, onset, and progression of Alzheimer’s disease and related dementias (ADRD), yet BSF-related knowledge remains scattered across heterogeneous sources, posing critical challenges for data harmonization and evidence synthesis. To address this gap, we present the B ehavioral S ocial Data and Knowledge O ntology for AD RD (BSO-AD), a FAIR-compliant semantic resource for representing and harmonizing BSFs and ADRD-related knowledge. BSO-AD was developed following established ontology design principles with reuse of existing ontologies and controlled terminologies, including the Social Determinants of Health Ontology, Drug Repurposing-Oriented Alzheimer’s Disease Ontology, AD-Onto, ICD-9-CM, and ICD-10-CM. Relationships between BSFs and ADRD were derived through literature mining. BSO-AD contains 1,690 classes, 152 object properties, 49 data properties, and 40 annotation properties. Ontology evaluation through Hootation-based domain expert review and a scalable LLM-assisted ontology assessment framework demonstrated high domain coverage and strong semantic coherence. BSO-AD provides a semantic foundation for data harmonization and knowledge integration in BSF-related ADRD research, while the LLM-assisted evaluation framework supports scalable and automated ontology assessment.

Abstract Purpose: Standard care for high-grade gliomas (HGG) involves maximal surgical resection followed by radiation and temozolomide. Postoperative adjuvant therapy frequently causes lymphopenia, which is associated with poor prognosis. Interleukin 7 (IL7) is essential for lymphocyte development, homeostasis, and survival. NT-I7 (efineptakin alfa), a long-acting recombinant IL7, reverses lymphopenia and improves survival in murine glioma models. However, the safety, maximum tolerated dose (MTD), and impact of NT-I7 on immune cells in patients with HGG remain unknown. Patients and Methods: We conducted a phase I trial (NCT03687957) examining the MTD and effect of NT-I7 on lymphocytes in patients with newly diagnosed HGG. The primary endpoint was dose-limiting toxicity; secondary endpoints included absolute lymphocyte count (ALC) changes over time, overall response, progression-free survival, and overall survival. Exploratory endpoints included immune profiling at different time points using single-cell RNA sequencing (scRNA-seq) in a subset of patients. Results: NT-I7 was well tolerated with a MTD of 720 µg/kg. Moreover, NT-I7 significantly increased ALCs for more than 12 weeks in duration. Early elevations in CD4+, CD8+ T cells and NK cells further coincided with increased TNF and CXCL9 cytokine levels. Comprehensive immune profiling of peripheral blood T cells revealed selective clonotype expansion within CD8+, but not CD4+, T cells following NT-I7 administration. Finally, a subset of our patients with MGMT promoter–unmethylated glioblastoma, which are typically associated with a poorer prognosis, demonstrated promising clinical responses. Conclusions: NT-I7 has the potential to maintain and increase lymphocyte counts in patients with HGG and warrants further investigation, particularly in combination with immune-based therapies.

Abstract Background Progressive supranuclear palsy (PSP) is a tauopathy for which there is limited understanding of epidemiological risk factors. Objective The objective of this study was to identify premorbid nonmedical and medical risk factors for PSP diagnosis in a large prospective population‐based cohort. Methods We performed a matched nested case–control study using the UK Biobank (UKB) cohort consisting of >500,000 adults aged 37 to 73 years. PSP diagnoses were identified from 2007 to censor date (August 2025) and matched 1:10 by age and sex to control subjects without parkinsonism or dementia diagnoses. We derived risk factors using an iterative logistic regression approach. Results There were 240 incident PSP diagnoses in the UKB. In the fully adjusted model, premorbid depression (odds ratio [OR], 3.22; 95% confidence interval [CI], 2.30–4.50; P < 0.001), delirium (OR, 6.76; 95% CI, 4.08–11.19; P < 0.001), and functional gastrointestinal disorders (OR, 1.91; 95% CI, 1.39–2.63; P < 0.001) were associated with increased risk. Heavy alcohol consumption was associated with higher risk compared with moderate consumption (OR, 1.65; 95% CI: 1.21–2.26; P = 0.0017). Body mass index (BMI) and alcohol consumption had nonlinear risk profiles. Cancer diagnosis was inversely associated with PSP (OR, 0.56; 95% CI, 0.40–0.79; P < 0.001). Depression and delirium diagnoses remained significant 5 to 10 years before diagnosis. Conclusions In this large prospective cohort, PSP was associated with premorbid lifestyle practices and neuropsychiatric diagnoses years before diagnosis. This suggests a long premotor phase and highlights opportunities for earlier diagnosis and further mechanistic investigation. The inverse association with cancer mirrors the relationship seen in other neurodegenerative conditions and may point toward shared mechanisms. © 2026 International Parkinson and Movement Disorder Society.

Abstract Models that estimate the probability of an adverse clinical outcome require an operational cutoff to translate continuous estimated probabilities into discrete labels that can trigger clinical action. Although statistical methods identify ‘optimal’ cut-offs, threshold selection ultimately reflects value judgments regarding harm tolerance, resource allocation, and workflow feasibility. We describe a governance-informed approach to selecting a deployment threshold for an asthma exacerbation (AE) prediction model integrated into clinical workflows. Using prevalence-adjusted performance metrics and real-world provider capacity modeling, we evaluated multiple candidate thresholds and quantified downstream workload and missed-event trade-offs. We demonstrate that statistically optimal thresholds may produce operationally infeasible alert volumes or unacceptable miss rates. We propose a structured threshold governance framework integrating statistical performance, clinical utility, stakeholder input, and human oversight safeguards. This case illustrates how threshold decisions should be treated as organizational governance processes rather than purely technical optimizations.

Abstract Radiotherapy is an established treatment for low-grade primary cutaneous B-cell lymphoma. Recommendations on its use differ internationally, which prompted our group to conduct this analysis. Twenty-two institutions participated in this international study. Patient eligibility required a diagnosis of limited (T1/T2) primary cutaneous marginal zone or follicle center lymphoma treated with radiotherapy between 1995 and 2023. Data were collected retrospectively until February 2024 within the framework of the International Lymphoma Radiation Oncology Group. Overall, 535 patients were analyzed. Predominant locations were the head (40%) and trunk (36%). Radiotherapy had a median dose of 24 Gy in fractions of 2 Gy. Complete responses were observed in 91% at a median time of 3.6 months following radiotherapy. There was no statistically significant difference between treatments ≤4 Gy and &gt;4 Gy for complete or overall response rates (P = .077 and P = .056, respectively). However, there was an inferior duration of local control with ≤4 Gy (5-year local control 73% ± 12% vs 96% ± 2%; P&lt; .001). Radiation dose was the main prognostic factor in the univariate and multivariable Cox analysis; however, higher doses did not translate into an overall survival benefit. Toxicities rarely exceeded grade 2 but were more frequent in the &gt;4 Gy group. Radiotherapy remains an effective treatment option for indolent skin lymphoma with low toxicity. High response rates are observed with low doses of ≤4 Gy. In comparison to conventional doses, these treatments have a shorter duration of local control but a favorable toxicity profile.

Abstract Hypertensive disorders of pregnancy (HDP) are a leading cause of maternal and perinatal morbidity, yet modifiable environmental risk factors remain poorly characterized. Prior studies typically have only examined a limited number of exposures and have rarely distinguished HDP subtypes (ie, gestational hypertension, preeclampsia, eclampsia, and chronic hypertension with or without superimposed preeclampsia) or accounted for residential mobility during pregnancy. To address these gaps, we conducted a spatial and contextual exposome study of HDP using linked electronic health records (EHR) and vital statistics data in Florida. We analyzed 686 412 singleton pregnancies conceived between 2013 and 2018, using computable phenotyping to distinguish HDP subtypes. A total of 245 spatial and contextual exposome measures spanning natural, built, and social environments were linked to residential histories from conception through gestational week 19. Using a two-phase double machine learning (DML) framework with exposure-specific, directed acyclic graph-guided confounder adjustment, we conducted discovery and replication analyses, followed by multi-treatment DML to estimate effect sizes. In Phase 1, 26 exposome measures replicated for gestational hypertension and 34 for overall HDP. In Phase 2, 12 measures remained associated with gestational hypertension and 11 with overall HDP, including air toxicants, meteorological factors, ultraviolet radiation, neighborhood crime indicators, environmental noise, and proximity to coastline. No exposures passed multiple-comparison thresholds for preeclampsia or eclampsia. These findings demonstrate that the spatial and contextual exposome contributes to HDP in a subtype-specific manner. Integrating EHR-linked phenotyping, residential mobility, and causal machine-learning methods offers a scalable framework for identifying environmental factors relevant to HDP prevention.

Abstract Background Neuroendocrine tumors (NETs) are rare, heterogeneous neoplasms associated with prolonged survival and substantial symptom burden. However, patient-reported outcomes across NET subtypes remain poorly characterized, particularly in real-world settings. This study describes baseline health-related quality of life (HRQoL) and care experiences among patients with gastroenteropancreatic and lung NETs, examining differences by tumor site and time since diagnosis. Methods The Neuroendocrine Tumors–Patient Reported Outcomes study is a prospective, multi-institutional US cohort of adults (aged 18 years and older) with incident small intestinal, pancreatic, gastroenteropancreatic, or lung NETs diagnosed from January 2018 through September 2024, identified via a validated electronic medical record–based computable phenotype. Baseline surveys assessed HRQoL, symptoms, care experiences, and clinical characteristics using validated instruments. Descriptive statistics and standardized mean differences compared responses by NET site and time since diagnosis. Results Among 2367 participants (mean age = 57.8 years; 57.3% female), 1974 had gastroenteropancreatic NETs (659 small intestinal NET, 555 pancreatic NET) and 393 had lung NETs. Fatigue (mean = 33.0), insomnia (mean = 32.5), and diarrhea (mean = 25.7) were the most burdensome symptoms. Lung NET patients reported worse dyspnea (standardized mean difference = 0.58, P &lt; .001) and lower physical, role, and global QoL scores than those with gastroenteropancreatic NETs, while pancreatic NET patients reported better functioning. Diarrhea worsened over time, especially in small intestinal NETs. Most rated care highly (75.3%) but cited concerns about treatment side effects (80.4%), costs (60.7%), and travel burden (58.8%). Conclusions This large US cohort reveals persistent symptom burden and HRQoL variation by tumor site and disease duration, underscoring the need for longitudinal HRQoL assessment in NET care.

Abstract Background: In MBC, CTC enumeration is highly prognostic for both progression free survival (PFS) and overall survival (OS). Multiparametric technologies enable for expanded phenotypic profiling of CTCs, including ER and HER2 assessment. Limited information is available on the concordance of CTC ER and HER2 levels compared to synchronously obtained tissue. We assessed the prognostic performance of CTC enumeration and phenotypic profiling of ER and HER2 on CTCs. Methods: The prospective, multi-institutional Aurora study enrolled pts with MBC of any subtype either at diagnosis of metastatic disease or at time of progression on any line of systemic therapy. Pts underwent a metastatic site biopsy, consented to archival primary tissue collection and donated blood samples serially every 3-4 months until progression. CTCs, from blood collected at study entry, were identified as nucleated, EpCAM+/ cytokeratin+/ CD45- cells using an automated imaging system (RareCyte, Seattle). ER and HER2 expression levels were measured by mean fluorescence intensity (MFI), with thresholds above 80 (ER) and 160 (HER2) considered positive. Comparisons were made to total RNAseq from paired metastatic tissue biopsies. PFS and OS were calculated from the time of study entry to last follow up or death. Results: Baseline CTC enumeration was performed in 154 pts. The most common clinical subtype was HR+/HER2- (59.7%), followed by HER2-positive (16.8%) and TNBC (15.5%). The median lines of prior therapy were 1 (range 0-10). At least 1 CTC was identified in 68/154 pts (44%), 30 (19%) had ≥ 5 CTCs, and 15 (9.7%) had CTC clusters. The median CTC value was 0 (0-877). The presence of CTCs (≥1 or ≥5) and CTC clusters was adversely associated with PFS and OS (Table 1 Table 1. CTC survival analyses Number Median PFS PFS HR (95% CI) Median OS OS HR (95% CI) Presence of CTC 0 CTC 86 7.37 (0.03 - 43.1) reference 20.72 (0.03 - 50.6) reference ≥ 1 CTC 68 4.7 (0.43 - 19.5) 1.65 (1.16-2.36) 12.04. (0.43-44.1) 2.13 (1.44-3.16) Presence of ≥ 5 CTC &lt;5 CTC 124 7.7 (0.03 - 43.1) reference 19.31 (0.03 - 50.6) reference ≥ 5 CTC 30 2.88 (0.43 - 18.7) 1.77 (1.15 - 2.71) 5.74 (0.43 - 29.3) 3.59 (2.26-5.70) Presence of clusters 0 clusters 139 7.37 (0.03 - 43.1) reference 18.06 (0.03 - 50.6) reference ≥ 1 cluster 15 1.91 (0.43 - 18.7) 2.06 (1.16 - 3.68) 3.42 (0.43 - 18.7) 5.44 (2.94-10.09) ). Across all subtypes, the median CTC ER-MFI across was 36.2 (range 4.2-10,016.6) and the median HER2-MFI was 139.6 (range 4.2-2480.5). We observed high intra-patient heterogeneity of ER-MFI (CV: 0.65 [0.34-1.25]) and HER2-MFI (CV: 0.79 [0.42-1.54]). Clinical metastatic subtype from metastasis at study entry was available in 54/68 pts with CTCs. Pts with clinical ER+ MBC had a higher median ER-MFI compared to clinical TNBC (111.9 vs 68.6, p = 0.02). However, 11 pts (26.8%) with clinical ER+ MBC had all CTCs with an ER-MFI &lt;80. Lower ER expression was observed in CTCs collected after prior exposure to a SERD (median ER-MFI: post-SERD [n=8] 27 vs post-aromatase inhibitor [n=4] 252, p&lt;2.22e-16). No correlations were observed between the median CTC ER-MFI or HER2-MFI and paired tissue ESR1 or ERBB2 RNA expression (n=66). Conclusions: CTCs are strongly predictive of PFS and OS. CTC ER- and HER2-MFI were highly heterogeneous without clear correlation with bulk tissue RNA expression. The clinical impact of discordant clinical receptor status and CTC receptor status warrants further exploration. Citation Format: K. V. Giridhar, W. C. Nenad, H. Ye, G. L. Wheeler, T. Hinoue, A. R. Michmerhuizen, B. M. Felsheim, U. R. Chandran, B. J. Kelly, E. R. Mardis, P. W. Laird, J. M. Balko, A. C. Garrido-Castro, S. Coppens, J. Bowen, R. Nanda, N. U. Lin, C. K. Anders, C. Isaacs, J. J. Tao, J. D. Anampa, N. Jahan, L. Huppert, A. DeMichele, B. H. Park, S. Vinayak, A. L. Delson, M. M. Magbanua, M. Balic, H. A. Parsons, A. V. Lee, L. A. Carey, L. Norton, A. C. Wolff, N. E. Davidson, I. Krop, T. A. King, C. M. Perou, K. A. Hoadley, Aurora US Metastases Network. Characterizing circulating tumor cell (CTC) enumeration and phenotypic profiling in blood with synchronously obtained tissue in metastatic breast cancer (MBC) in TBCRC /AURORA US [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS2-07-17.

Abstract Introduction: Recruitment for rare disease studies is challenging due to small eligible populations. Traditional clinical research management systems often lack tools to track recruitment contacts prior to enrollment. The NET-PRO study, focused on neuroendocrine tumors (NETs), implemented a participation monitoring system to enhance recruitment efficiency and representativeness. Methods: NET-PRO is a multicenter cohort study of 2538 adults diagnosed with gastroenteropancreatic (GEP) or lung NETs between January 2018 and September 2024. Recruitment occurred from January 2022 to February 2025 across 14 U.S. medical centers. Sites used flexible recruitment methods (email, mail, phone, in-clinic) and tracked contacts using REDCap-based tools. Participant characteristics were analyzed by enrollment mode (online or mail) and recruitment difficulty (number of contacts required prior to enrollment) using standardized mean differences, chi-square tests, and ANOVA. Results: Of 9279 contacted patients, 2675 consented (28.8%) and 2538 enrolled (27.4%). Most enrolled online (83.2%), while 16.8% enrolled by mail. Mail respondents were older, had lower education and income, and more comorbidities. Among those enrolled, recruitment difficulty was associated with older age, lower education and income, but not comorbidity. Over half of the most difficult-to-recruit participants enrolled online. Contact methods varied by attempt, with email dominating early contacts and phone/mail used more in later attempts. Conclusions: A participation monitoring tool supported flexible, multimodal recruitment and improved sample representativeness in a rare cancer study. Tracking recruitment contacts enabled adaptive strategies and may reduce bias in observational research by enabling better outreach to harder-to-reach populations.