Chronic obstructive pulmonary disease (COPD) caused 3.72 million deaths in 2021. Exacerbations drive morbidity and hospital admissions. Conventional oxygen therapy (COT) offers limited support, while non-invasive ventilation (NIV) improves outcomes but is often poorly tolerated. High-flow nasal cannula (HFNC) delivers heated, humidified oxygen at high flow rates, improving gas exchange and comfort. This systematic review (PRISMA-guided) examines HFNC use in acute and chronic hypercapnic COPD compared with COT and NIV.

Systematic searches of MEDLINE, Embase and CENTRAL (to 2025) identified 14 primary studies (7 randomized trials, 7 observational cohorts). Physiological, clinical and patient-reported outcomes were extracted. Risk of bias was evaluated using RoB 2 and ROBINS-I.

In acute settings, HFNC consistently improved PaCO₂ and reduced respiratory effort versus COT, lowering treatment failure (10.0% vs 19.4%, p=0.026). However, HFNC failed non-inferiority versus NIV for treatment failure (25.7% vs 14.3%) and showed higher intubation (14.2% vs 5.4%, p=0.026). Domiciliary HFNC reduced annual exacerbations and improved quality of life (SGRQ mean difference -8.12).

HFNC provides physiological benefits and lowers short-term failure versus COT and reduces home exacerbations. However, it is less effective than NIV for severe hypercapnic failure. HFNC should be reserved for mild-to-moderate exacerbations or NIV-intolerant patients, with close monitoring and predefined escalation criteria.

This systematic review and meta-analysis aimed to evaluate the association between obstructive sleep apnea and periodontitis and to quantify the magnitude of this relationship based on the available evidence.

A systematic literature search was conducted in PubMed, Scopus, and Web of Science up to November 2025, following PRISMA 2020 guidelines. Observational studies investigating the association between OSA and periodontitis in human adults were included. Study quality was assessed using the ROBINS-I tool. A random-effects meta-analysis was performed to pool odds ratios (ORs) with 95% confidence intervals (CIs). Additionally, Spearman correlation coefficients were synthesized when available.

Twenty-three studies were included in the qualitative synthesis, and 18 studies were eligible for quantitative analysis. The pooled meta-analysis demonstrated that OSA was significantly associated with an increased risk of periodontitis (OR = 2.22; 95% CI: 1.78-2.77), with moderate heterogeneity (I² = 57.4%). Furthermore, a random-effects meta-analysis of correlation coefficients revealed a moderate positive association between OSA and periodontal disease severity (r = 0.39; 95% CI: 0.17-0.58). These findings were consistent across different populations and diagnostic approaches, particularly in studies using polysomnography for OSA assessment.

The present meta-analysis provides evidence of a significant association between obstructive sleep apnea and periodontitis, with patients affected by OSA showing more than twice the risk of developing periodontal disease. Although causality cannot be established due to the observational nature of the included studies, shared inflammatory and microbiological pathways support a biologically plausible link. An interdisciplinary approach involving dental and sleep medicine professionals may be beneficial for early diagnosis and comprehensive management of both conditions.

Recurrent aphthous stomatitis (RAS) is a chronic autoinflammatory condition marked by recurring, painful sores in the mouth. It is often confused with Behcet's disease (BD), a rare systemic vasculitis that also presents with oral ulcers. Despite overlapping symptoms, BD has broader systemic implications, making an accurate diagnosis critical. This study aims to identify unique serum proteins that could reliably distinguish idiopathic RAS from BD.

We reanalyzed our previous mass spectrometry dataset comprising blood samples from 12 BD patients, 12 individuals with idiopathic RAS, and 21 healthy controls. Differentially expressed proteins (DEPs) related to RAS were identified and examined through Kyoto Encyclopedia of Genes and Genomes and Gene Ontology pathway enrichment. A protein-protein interaction (PPI) network was created to explore functional connections among the DEPs. Validation of three RAS-related proteins was carried out using enzyme-linked immunosorbent assay (ELISA) in a separate cohort of 26 RAS patients, 26 BD patients, and 30 healthy individuals. Their diagnostic utility was then evaluated via receiver operating characteristic (ROC) curve analysis.

A total of 99 proteins showed differential expression in RAS samples but not in BD cases when compared to healthy controls-85 were upregulated, and 14 were downregulated. Enrichment analyses indicated these proteins are primarily involved in metabolic and infection-related pathways, particularly influencing keratinocyte differentiation and oxidative stress responses. PPI network analysis highlighted key metabolic and keratinocyte-related proteins as central hubs, suggesting a role in RAS pathology. ELISA validation confirmed significantly elevated levels of ANXA2, ENO1, and S100A7 in RAS patients compared to both BD patients and healthy subjects.

Our findings identify a set of RAS-related serum proteins with potential diagnostic value. These serum proteins may enhance the clinical differentiation of RAS from BD, aiding in more accurate and timely patient care.

Frail older patients are at high risk of both thromboembolism and bleeding, making anticoagulant therapy challenging. Direct oral anticoagulants (DOACs) have largely replaced vitamin K antagonists because of comparable efficacy and lower bleeding risk. However, evidence supporting their use in frail older patients remains limited. Recent studies have demonstrated that a substantial proportion of frail older patients have DOAC plasma levels outside the expected on-therapy range, which may be associated with an increased risk of bleeding or thromboembolic events.

The DOAC-FRAIL randomized controlled trial (RCT) aims to determine whether adjustment of DOAC therapy based on peak plasma concentrations improves the safety and efficacy of anticoagulant treatment compared with standard care in frail older patients.

The DOAC-FRAIL study is an international, multicenter, randomized controlled trial including frail older patients aged ≥70 years with a Clinical Frailty Scale score >3 receiving DOAC therapy. Participants are randomized to either a Single Cycle Measurement and Adjustment Strategy based on peak DOAC plasma concentrations or standard care. Patients with a life expectancy of less than 3 months or previous DOAC dose adjustment based on plasma level measurements are excluded. The primary outcome is the composite of bleeding events, including clinically relevant non-major bleeding, and thromboembolic events during 2 years of follow-up. Secondary outcomes include feasibility, determinants of deviant peak DOAC plasma concentrations, quality of life, cost-effectiveness, and all-cause mortality.

The DOAC-FRAIL trial is currently ongoing. Participant recruitment has been initiated, and study results are expected after completion of follow-up.

The DOAC-FRAIL RCT will provide evidence on whether a plasma concentration-guided DOAC dosing strategy improves clinical outcomes compared with standard care in frail older patients. The findings may contribute to more personalized anticoagulant therapy in this vulnerable population.

EU CT-2025-521362-10-00.

Individuals with hypertension are at risk to reduced quality of life (QoL). Explainable machine learning (ML) can be used for domain-specific risk stratification and prioritization of modifiable determinants of low QoL.

To train ML classifier algorithms for QoL risk stratification in hypertension, where meaningful determinants were explored through Shapley additive explanations (SHAP).

Data from hypertensive individuals (n = 534) completed WHOQOL BREF, Quick Physical Activity Rating, Morisky Medication Adherence Scale 8, and standardized questionnaires for acceptance and knowledge were analyzed utilizing Decision Tree, Gradient Boosting, XGBoost, AdaBoost, Random Forest, and Naive Bayes ML classifiers. The trained ML algorithms were evaluated using stratified 10-fold cross-validation, where the stability was examined using rank-based metrics. SHAP were applied to the gradient boosting, as the most stable model.

For physical QoL, Random Forest (AUC 0.850; sensitivity 0.835; specificity 0.738) and Gradient Boosting (AUC 0.850; sensitivity 0.801; specificity 0.764) showed good reduced QoL identification. For psychological domain, best classifications were obtained from Gradient Boosting performed best (AUC 0.833; sensitivity 0.818; specificity 0.651) and XGBoost (AUC 0.831; sensitivity 0.824; specificity 0.660), with the former observed as the most stable SHAP analysis identified acceptance and medication adherence as the dominant shared drivers of risk across both QoL domains. Physical QoL risk was further influenced by physical activity-related factors, whereas Psychological QoL risk showed additional contributions from age and educational attainment.

Ensemble tree-based classifiers, particularly Gradient Boosting, had the most optimal performance in discriminating reduced and good QoL. Acceptance and medication adherence are the most influential shared drivers of risk, while physical activity, age, and educational attainment contributed to domain-specific heterogeneity.

BACKGROUND Aorto-aortic extra-anatomical bypass is a specialized surgical intervention primarily used to manage complex vascular complications, including stent graft infections after endovascular aortic repair. Stent graft infection is a rare but severe complication associated with high morbidity and mortality, which requires prompt treatment to prevent life-threatening sequelae such as pseudoaneurysm formation, aortic rupture, or fistula formation. CASE REPORT We describe 2 young male patients who developed infected thoracic aortic stent grafts after aortic arch debranching and endovascular repair. The first patient, a 37-year-old man with a history of traumatic aortic isthmus transection managed via tube interposition grafting 10 years earlier, presented with a proximal anastomotic pseudoaneurysm; he subsequently underwent arch debranching and endovascular stent grafting. He was later readmitted with systemic infection and mediastinal hematomas indicative of graft infection. The second patient, a 34-year-old man with Behçet disease and an aortic arch aneurysm, presented with an infected mediastinal hematoma involving the aortic stent graft after arch debranching and stent grafting. Both patients underwent stent graft explantation and extra-anatomical ascending-to-descending aorto-aortic bypass under peripheral cardiopulmonary bypass with deep hypothermic circulatory arrest. Both patients survived the complex aortic procedure and remained clinically stable at 38 and 17 months of follow-up (after the most recent procedure), respectively. One patient required additional wound management for a surgical site infection; the other underwent delayed repair of an ascending aortic pseudoaneurysm. CONCLUSIONS This case series demonstrates the feasibility of aorto-aortic extra-anatomical bypass as a potentially life-saving intervention for aortic endograft infection when perioperative management is optimized.

Large language models are increasingly applied in medical education and clinical decision support. However, comparative evaluations of their performance on pediatric content-spanning multiple subspecialties and question complexities-remain limited. This study sought to assess the performance of two advanced large language models, DeepSeek-R1 and GPT-4o, using a comprehensive set of Pediatrician Licensing Exam practice questions.

We administered 280 expert-validated pediatric questions covering 11 subspecialties and three levels of clinical complexity (A1: direct, A2: simple cases, A3: complex cases) to DeepSeek-R1 and GPT-4o. Each model was tested twice, with a 4-week interval. Performance metrics, including per-run accuracy, consistent accuracy (correct in both runs), aggregate accuracy (correct in either run), and inter-run consistency, were compared using chi-square tests and Bonferroni correction.

DeepSeek-R1 outperformed GPT-4o in both runs (1^st run: 90.7% vs. 81.8%, P=0.003; 2^nd run: 86.8% vs. 79.3%, P = 0.02). It also showed higher aggregate accuracy (92.5% vs. 85.7%, P=0.01) and consistent accuracy (85.0% vs. 75.4%, P = 0.01). Performance declined with increasing question complexity for both models, with DeepSeek-R1 demonstrating numerically higher accuracy across all question types. For consistent accuracy, DeepSeek-R1 achieved 89.2% vs. 79.2% on A1 questions (P=0.05), 83.8% vs. 71.6% on A2 questions (P=0.11), and 80.2% vs. 73.3% on A3 questions (P=0.36). At the subspecialty level, DeepSeek-R1 showed more uniform performance, achieving >90% consistent accuracy in four subspecialties: Neonatology (94.7%), Infectious Diseases (93.8%), Nephrology (91.9%), and Cardiovascular diseases (90%). In comparison, GPT-4o reached >90% consistent accuracy in three subspecialties-Genetics (100%), Infectious Diseases (100%), and Cardiovascular diseases (90%)-with greater variability across other domains.

DeepSeek-R1 demonstrated higher accuracy than GPT-4o on pediatric questions. The observed variability across different runs warrants attention, particularly in contexts that require highly consistent and reproducible performance.

Previous studies established handgrip strength (HGS) as a potential risk factor for cardiovascular diseases (CVD). However, existing studies focused exclusively on baseline HGS and neglected longitudinal changes in HGS during follow-up. Thus, our aim was to investigate the associations of transitions and dynamic burdens of HGS with incident CVD risk.

We analyzed data from the UK Biobank (UKB), the China Health and Retirement Longitudinal Study (CHARLS), the Survey of Health, Ageing and Retirement in Europe (SHARE), and the Korean Longitudinal Study of Ageing (KLOSA). We defined HGS transitions based on HGS at baseline and the first follow-up, and created three indicators to reflect HGS burdens. Cox models were applied to examine the association of HGS transitions and burdens with incident CVD risk. The predictive value of HGS indices was also evaluated.

A total of 73,555 participants were retained, and 4722 (6.4%) incident CVD cases were identified during follow-up. Transition analyses revealed that increased HGS during follow-up was associated with reduced CVD risk, whereas decreased HGS was associated with an elevated risk. Per standard deviation decrement in HGS slope, cumulative HGS and relative cumulative HGS led to a 19.8% (95%CI 1.5∼41.5%), 44.0% (95%CI 10.8∼87.2%) or 26.7% (95%CI 9.4∼46.8%) elevated risk of CVD, respectively. These associations were independent of and more pronounced than HGS, with stronger effects observed in East Asian cohorts (CHARLS and KLOSA) compared to European cohorts (UKB and SHARE). Incorporating dynamic HGS metrics enhanced the predictive accuracy, with HGS burdens providing greater gains than HGS. For the optimal models, all HGS indices resulted in an increase of AUC up to 7.6% in Europeans and 5.9% in East Asians.

HGS burdens outperformed in predicting cardiovascular health compared to single cross-sectional HGS itself, suggesting the clinical utility of longitudinal HGS monitoring in clinical and public health strategies for CVD prevention.

CTNNA3 encodes αT-catenin, an intercalated disc (ICD) protein essential for cardiomyocyte coupling. Human omics studies have shown reduced CTNNA3 expression, ICD ultrastructural disruption, and dilated cardiomyopathy (DCM)-associated hyperphosphorylation of αT-catenin. Direct RNA-level evidence linking biallelic CTNNA3 variants to human cardiomyopathy has been lacking. Clinical exome sequencing was performed in a 21-year-old man with DCM, severe left ventricular systolic dysfunction (LVEF 20%), and atrial fibrillation (AF). The identified homozygous variant (NM_013266.4 : c.1733 - 1G > C) was assessed with multiple splicing prediction tools and functionally validated via a minigene hybrid assay in HEK293 cells. Splicing predictors indicated loss of the canonical acceptor site. The minigene assay confirmed three aberrant transcripts: out-of-frame exon 13 skipping with a premature stop codon, a 24-nucleotide in-frame deletion, and partial intron retention, resulting in a possibly nonfunctional protein. Under guideline-directed medical therapy, LVEF normalized, but a persistent arrhythmic phenotype remained, with recurrent AF, frequent ventricular ectopic beats, and nonsustained ventricular tachycardia. We report a recessive form of DCM associated with a homozygous canonical splice-site variant in CTNNA3, encoding the ICD protein αT-catenin. Our results are consistent with human omics studies. Altogether, these data provide evidence that biallelic CTNNA3 splice-disrupting variants can cause human cardiomyopathy driven by ICD dysfunction. The dissociation between ventricular recovery and persistent arrhythmia highlights the complex phenotypic spectrum of CTNNA3-related disease.

Collegiate sport environments often bring mental health (MH) and performance (MP) into close interaction; yet, screening, triage, referral, and MP support often remain operationally fragmented. This fragmentation may weaken follow-through after screening, limit coordination across clinical and performance services, and create uncertainty around privacy, scope boundaries, and accountability. In this Perspective, drawing on implementation science, integrated care, and systems-oriented organizational design, we suggest that a promising direction in athlete MH-MP support is not simply expanding screening availability, but building governed infrastructure that better connects identification to action. We propose a sport science hub model for coordinating two linked but distinct lanes: (1) structured MH screening, triage, and clinician-led referral workflows, and (2) MP profiling and skills-based support delivered within non-clinical scope. The model is organized around explicit guardrails, including role separation, privacy-secured and auditable access, non-punitive use protections, and emergency escalation pathways. We discuss the rationale for using structured screening and referral processes while acknowledging the limitations of current screening tools, including false-negative and false-positive concerns, and the importance of interpreting performance disruption in context rather than medicalizing normal variation. We conclude with governance principles and a forward-looking research agenda focused on implementation, trust, acceptability, screening accuracy, and prospective evaluation of integrated MH-MP systems.