Citations

Background: Sinonasal undifferentiated carcinoma (SNUC) is an extremely rare, high-grade, and aggressive tumor of the sinonasal tract. Due to the rarity of this malignancy, current treatment guidelines are based on small and often/mainly single-center retrospective datasets. In the absence of a universally accepted standard of care for SNUC, treatment approaches vary across countries and institutions, reflecting real-world clinical practice. The primary aim of this study was to describe real-world treatment and outcomes for patients with confirmed SNUC. Methods: This was an international, multi-center, retrospective, observational cohort study that pooled patients into the largest SNUC dataset to date. Fifteen centers were enrolled to contribute data, including seven from Europe, four from the United States, three from the United Kingdom, and one from Canada. In the absence of a universally accepted standard of care for SNUC, treatment approaches varied across countries and institutions, reflecting real-world clinical practice. Patients included were those with histologically confirmed SNUC who were treated between 1997 and 2021. Results: This study yielded 485 patients treated for SNUC. The median age at diagnosis was 55.6 years (IQR: 44.5–67.6), and 63.7% were male. Most cases presented at advanced stages, with 70.8% as T4a or T4b. Overall survival (OS) outcomes were available for 412 patients, with a median follow-up of 26.0 months. The 5- and 10-year OS were 47.2% (95% CI: 40.8–53.3%) and 39.6% (95% CI: 32.5–46.6%), respectively. Advanced age, dichotomized T-stage (T4a/b vs. T1–3), M-stage, and orbital involvement were significant poor prognostic factors on univariable analysis (p’s < 0.01). On multivariable analysis, orbital involvement (HR: 2.73, 95% CI: 1.42–5.27, p = 0.003) and distance metastasis stage (HR: 3.00, 95% CI: 1.25–7.21, p = 0.014) were both independently associated with worse OS. Conclusions: This observational study presents the largest multi-center cohort analysis of SNUC to date, providing new insights into prognostic factors for a rare cancer treated at global centers of excellence. Orbital involvement and the presence of metastases are candidate independent risk factors associated with poorer OS.

Abstract Generative artificial intelligence (AI) is rapidly populating medical records with synthetic or partially AI-generated content, creating a feedback loop where future models are increasingly at risk of training on uncurated AI-generated data. However, the clinical consequences of this AI-generated data contamination remain unexplored. Here, we show that in the absence of mandatory human verification, this self-referential cycle drives a rapid erosion of pathological variability and diagnostic reliability of medical data at population scale. By analysing more than 800,000 synthetic data points across clinical text generation, vision–language reporting, and medical image synthesis, we find that models progressively converge toward generic phenotypes regardless of the model architectures. Specifically, rare but critical findings, including pneumothorax and effusions, vanish from the synthetic content generated by AI models, while demographic representations skew heavily toward middle-aged male phenotypes. Crucially, this degradation is masked by false diagnostic confidence. Models continue to issue reassuring reports while failing to detect life-threatening pathology, with false reassurance rates tripling to 40%. Blinded physician evaluation confirms that this decoupling of confidence and accuracy renders AI-generated documentation clinically useless after just two generations. We systematically evaluate three mitigation strategies that can be easily integrated into existing clinical workflows, finding that while synthetic volume scaling fails to prevent collapse, mixing real data with quality-aware filtering effectively preserves diversity. Ultimately, our results suggest that without policy-mandated human oversight, the deployment of generative AI threatens to degrade the very healthcare data ecosystems it relies upon.

Background. Acute cellular rejection (ACR) remains a major cause of morbidity after heart transplantation despite advances in immunosuppression. Whole genome transcriptomic profiling offers a systems-based, unbiased approach to elucidate the molecular mechanisms underlying ACR. However, noninvasive, longitudinal biomarker assessments capable of capturing the temporal dynamics of rejection biology remain scarce. Methods. RNA sequencing of peripheral blood from heart transplant recipients before, during, and after ACR was compared with nonrejection controls. Pathway analysis was conducted using differentially expressed genes (DEGs), and a machine learning approach was applied to assess gene-based prediction of ACR. Results. A total of 235 rejection-specific significant DEGs and 863 postrejection DEGs (false discovery rate < 0.05) were identified. During ACR, DEGs were enriched for T-cell activation/differentiation, apoptosis, and B-cell receptor signaling pathways. By combining the 2 sets of DEGs, a panel of 71 common genes was identified that reflected the significant, longitudinal transcriptomic dynamics of ACR. In an elastic net machine learning–based classifier, DYNLL1 and SERF2 were identified as ACR predictive genes, and achieved a cross-validated area under the receiver operating characteristic curve of 0.63. Conclusions. Peripheral blood transcriptomics identify dynamic temporal responses in ACR including T- and B-cell pathways with potential ACR predictive genes that warrant further investigation.

Abstract Background Integrating advanced artificial intelligence (AI) into clinical decision-support often requires the sharing of sensitive patient data with external services, raising privacy concerns. Homomorphic encryption (HE) allows computing directly on encrypted data, without revealing the underlying patient information. Objectives To develop a large language model (LLM)-assisted diagnosis framework while preserving patient privacy in the clinical text analysis, by leveraging HE and using rare disease (RD) diagnosis as a representative application. To demonstrate HE does not hinder the system performance. Materials and Methods Texts from patient histories and a RD knowledge base were embedded by LLMs into vectors, then encrypted using HE to obscure private information while retaining the semantic nuances. Diagnostic recommendations were generated by computing and ranking the similarities between the patient history and RD vectors in the encrypted space. The system was evaluated using 50 synthetic case reports (5 RDs, each with 10 reports). Results Applying HE did protect private information from reverse-embedding attacks. HE imposed little disruption to the diagnostic accuracy, with normalized discounted cumulative gains (nDCG) of 0.6108 ± 0.3412 (encrypted) versus 0.6083 ± 0.3415 (unencrypted). The accuracy and computational performance were tunable and consistent, as demonstrated across five different LLMs. Discussion Our privacy-preserving framework opens tremendous opportunities toward hosting and serving powerful AI solutions across institution boundaries, which would remove the need for local deidentification and incentivize users to access secure external decision-support services. Conclusions Integrating HE with LLM retrieval can promote the dissemination of nonredundant, high-capacity AI services by preserving both privacy and accuracy.

Introduction Placental inflammation is a major contributor to preterm birth (PTB), and there are currently few targeted strategies to prevent PTB and its associated adverse neonatal outcomes. Serum amyloid A (SAA), particularly the isoforms SAA1 and SAA2, are well-recognized inflammatory markers, but their functional roles in placental inflammation remain poorly defined. Methods Using a translational mouse model of sub-chronic maternal inflammation, we investigated the immune mechanisms and therapeutic potential of siRNA-mediated targeting of Saa2 (siSaa2). Placental expression patterns of SAA2 were examined in vivo , and macrophage responses to extracellular SAA2 were modeled in vitro using RAW264.7 cells to assess downstream P2X7R-dependent signaling and functional outcomes. Results SAA2 is primarily induced in placental trophoblast and endothelial compartments during inflammation, where it acts as an extracellular inflammatory mediator. In vitro , we model macrophage responses to extracellular SAA2 using RAW264.7 cells to examine downstream P2X7R-dependent signaling and functional outcomes. Maternal administration of siSaa2 improved PTB rates, placental morphology and fetal brain development. Additionally, SAA1 and SAA2 exhibited distinct expression patterns in the mouse and human placenta. Discussion These findings identify placental SAA2 as a key inflammatory mediator in inflammation-associated PTB and demonstrate that targeted silencing of Saa2 represents a potential therapeutic strategy to mitigate placental injury and adverse fetal outcomes.

Abstract Background The accuracy and safety of generating medication orders by large language models (LLMs) must be demonstrated. Without standardization, performance evaluation is limited to time and resource-intensive clinician grading. This evaluation aimed to develop a standardized medication format that supports automated performance evaluation (MedMatch). Methods First, a survey of 40 medication prompts was given to clinicians to assess agreement in medication order communication. Second, a clinician panel developed a standardized medication format (MedMatch) for oral and intravenous medications. Third, a clinician-annotated dataset of medication prompts and standardized answers in the MedMatch format was developed for LLM testing. Finally, LLMs were retested with the same dataset, adjusted to exclude route information, to evaluate the appropriate categorization of medication route. Results The formal medication orders consistently showed low omission rates and high overlap for all entities, compared to the verbal and brief written communication types. Lexical overlap results demonstrated pattern norms amongst clinicians with entities appearing most commonly in positions 1-5 in the order of drug name, dose, unit, route, and frequency. In the second survey, the formal written group performed the highest with 78.3% of prompts considered appropriate as a computer-generated response. LLM accuracy on MedMatch order standardization was highest in oral solid (64.2-72.5%), intravenous intermittent (72.5-84.3%), and intravenous push (62.7-74.5%) categories. LLMs performed the worst at categorizing medication orders accurately into intravenous push (18-61%) and intravenous intermittent (51-100%) routes. Conclusions Standardized format for computer-based outputs may support automated performance analysis and enhance the clarity of medication communication.

Abstract Purpose: Rearranged during transfection (RET) alterations are oncogenic drivers across solid tumors. Selective RET inhibitors (SRI) selpercatinib and pralsetinib have transformed outcomes for patients with RET-altered malignancies. Limited knowledge exists on genomic mechanisms of resistance to SRI. Experimental Design: We established “RETgistry,” a global consortium of patients with advanced RET-altered solid tumors who received SRI and underwent postprogression tissue or plasma biopsies assessed by next-generation sequencing. Frequencies of secondary RET resistance mutations and acquired non-RET gene alterations were determined. Progression-free survival (PFS) and time to treatment discontinuation (TTD) on first SRI were estimated with the Kaplan–Meier method. Results: RETgistry included 109 patients with RET-altered advanced solid tumors (lung, n = 94; thyroid, n = 15) who underwent 143 post-SRI progression biopsies (tissue, 91; plasma, 52). The median PFS and TTD were 13.9 months [95% confidence interval (CI), 10.1–16.6] and 17.3 months (95% CI, 14–20.2), respectively. Secondary RET mutations were detected in 20 (14%) biopsies [lung cancer, 15 (12.4%) and thyroid carcinoma, 5 (22.7%)]. Common acquired off-target alterations involved MET (18.2%; amplification, 15%), TP53 (8.2%), APC (7.6%), KRAS (7.1%), KEAP1 (5.9%), and CDKN2A/B (5.3%). MET alterations were enriched in post-SRI versus pre-SRI specimens (full cohort, 17.6% vs. 2.0%, P = 0.022; lung cancer, 19.1% vs. 2.1%, P = 0.022). Conclusions: The prevalence of secondary RET mutations after SRI was low, underscoring a greater role for off-target resistance. Recurrent acquired alterations involving tumor suppressor genes or upstream regulators of MAPK and PI3K pathways were identified, most commonly MET amplification. Continued efforts to characterize SRI resistance biology are critical to guide the development of novel therapeutic strategies.

Low self-esteem (LoST) is a latent yet critical psychosocial risk factor that predisposes individuals to depressive disorders. Although structured tools exist to assess self-esteem, their limited clinical adoption suggests that relevant indicators of LoST remain buried within unstructured clinical narratives. The scarcity of annotated clinical notes impedes the development of natural language processing (NLP) models for its detection. Manual chart reviews are labor-intensive and large language model (LLM)-driven (weak) labeling raises privacy concerns. Past studies demonstrate that NLP models trained on LLM-generated synthetic clinical notes achieve performance comparable to, and sometimes better than those trained on real notes. This highlights synthetic data’s utility for augmenting scarce clinical corpora while reducing privacy concerns. Prior efforts have leveraged social media data, such as Reddit, to identify linguistic markers of low self-esteem; however, the linguistic and contextual divergence between social media and clinical text limits the generalizability of these models. To address this gap, we present a novel framework that generates context-enhanced synthetic clinical notes from social media narratives and evaluates the utility of small language models for identifying expressions of low self-esteem. Our approach includes a mixed-method evaluation framework: (i) structure analysis, (ii) readability analysis, (iii) linguistic diversity, and (iv) contextual fidelity of LoST cues in source Reddit posts and synthetic notes. This work offers a scalable, privacy-preserving solution for synthetic data generation for early detection of psychosocial risks such as LoST and demonstrates a pathway for translating mental health signals in clinical notes into clinically actionable insights, thereby identifying patients at risk.

Abstract Background Large language models (LLMs) are increasingly used in medication-related tasks despite limited evidence supporting their accuracy and safety. LLMs are vulnerable to adversarial attacks and may confabulate in response to fabricated information in the prompt. Given the substantial risk of harm from medication errors, the purpose of this study was to establish a benchmark task to evaluate LLM performance in distinguishing fabricated from real medications. Methods Two datasets (brand and generic), each consisting of 250 medication lists were developed, each including 4-6 medications and one fabricated medication (specifically a Pokémon character), with complete dosing information provided. Six LLMs were evaluated (GPT-5-Chat, GPT-4o-mini, DrugGPT, Gemma-3-27B-IT, Llama-3.3-70B-Instruct, and Qwen3-32B). For each medication list, models were queried for dosing information and disease indication across three experimental conditions: standard decoding, standard decoding with mitigation prompting, and deterministic decoding (temperature = 0). Each experiment was performed in triplicate. Outputs were deemed confabulations if the LLM failed to recognize and reject the fabricated medication. The primary outcome was the rate of confabulations. Exact paired-permutation tests were used to compare confabulation rates across LLMs and prompting approaches. Results Confabulation rates under standard conditions ranged from 2.7% to 99.6%. DrugGPT demonstrated the lowest baseline confabulation rates (2.7-6.4%), whereas other models demonstrated substantially higher rates (47.2-99.6%). Mitigation prompting significantly reduced confabulation rates compared to standard and deterministic decoding across all tasks and datasets (p < 0.001): GPT-5-Chat achieved confabulation rates of 0% with mitigation prompting from baseline rates of 66.4-88.5%. Deterministic decoding did not substantially impact confabulation rates. Conclusions LLMs performing medication-related tasks are susceptible to confabulation in response to invalid or fabricated inputs, raising important safety concerns for clinical use. While mitigation prompting reduced confabulation rates, this study underscores the importance of robust safeguards when deploying LLMs in medication-related decision-making.

Purpose: Cutaneous squamous cell carcinoma has a propensity for perineural invasion. Treating perineural spread has involved surgery and adjuvant chemoradiation, and more recently, immune checkpoint inhibitors. This study aims to evaluate motor and sensory functional outcomes in patients with cranial neuropathies from perineural spread of squamous cell carcinoma, particularly those undergoing immunotherapy. Methods: This was a multicenter case series and literature review. Results: Eighteen patients with cranial neuropathies from perineural spread had a mean age of 72.5 ± 9.7 years. Cranial nerves V and VII were most often involved. Treatments included radiotherapy (8/18; 44.4%), chemotherapy (7/18; 38.9%), and/or immunotherapy (11/18; 61.1%), with many receiving a combination of modalities (8/18; 44.4%). Of those receiving immunotherapy, 72.7% (8/11) demonstrated at least partial clinical and/or radiologic tumor response. Among those with functional outcomes documented, half had improvement in sensory/motor nerve function following immunotherapy (4/8; 50%). A literature review identified 55 articles describing 449 patients total with cranial neuropathies from squamous cell carcinoma perineural spread. Combinations of surgery, chemotherapy, and/or radiotherapy were employed, with immunotherapy used in 16 patients. From 5 patients on immunotherapy with documented functional outcomes, all demonstrated at least some improvement. Conclusion: Immunotherapy plays an evolving role in managing advanced squamous cell carcinoma. Functional improvements in cranial neuropathy were demonstrated in 50% of patients receiving treatment with immunotherapy at a mean of 32.6 months of follow-up. This may hold important implications for the timing of surgical intervention, particularly in the case of cranial nerves V and VII palsies.