Ayse Ulgen

Background: Ivermectin is a widely used antiparasitic agent administered in topical and systemic forms. Although ivermectin is generally deemed to be safe, data on rare central nervous system (CNS) adverse events (AEs) are limited. Objectives: To evaluate, using the FDA Adverse Event Reporting System (FAERS) database, the incidence, demographic characteristics and safety signals of CNS AEs associated with ivermectin use. Methods: AEs reported between April 2016 and April 2026 were retrieved from FAERS via the Public Dashboard and OpenVigil 2.1. Four Medical Dictionary for Regulatory Activities (MedDRA) System Organ Classes were analysed. Events with fewer than three reports were excluded. A total of 1749 cases (75 Preferred Terms) were included. Using proportional reporting ratios and the reporting odds ratios with 95% confidence intervals, according to the Evans criteria, disproportionality analysis was performed on reports where ivermectin was the primary suspected drug. Results: A total of 7571 ivermectin-related reports were entered into the FAERS database between 2016 and 2026. Of these, 1749 included CNS AEs. The majority of reports were in male individuals (n = 188/414; 45.4%) and in 18–64 years age group (n = 169/414; 40.8%). Disproportionality analysis identified 13 statistically significant AEs associated with ivermectin. Among these, the most frequently reported and statistically significant neurological AEs were ‘confusional state’, ‘neurotoxicity’ and ‘altered state of consciousness’. Severe AEs involving the CNS, such as coma and encephalopathy, were also reported. Conclusions: FAERS data suggest that ivermectin may be associated with rare but serious neurotoxicity. Specifically, this study adds quantitative real-world evidence that the strongest disproportionality signals cluster around neurotoxicity, encephalopathy and altered or depressed consciousness rather than around common, mild symptoms such as headache or dizziness. Although generally safe, neurological symptoms should be monitored, particularly in older and polymedicated patients.

Healthcare environments present uniquely demanding constraints for the deployment of artificial intelligence (AI). Clinical decisions of significant consequence are rarely the product of a single isolated computation; rather, safe and effective patient care relies on the continuous collaboration of distributed entities, including clinicians, hospital networks, laboratory systems, monitoring devices, and patients themselves. As generative models and large language models are increasingly deployed in clinical settings, the transition from monolithic AI models to Multi‐Agent Systems (MAS) offers unprecedented capabilities for distributed monitoring, coordinated decision support, and service integration. However, multi‐agent collaboration simultaneously introduces severe novel vulnerabilities, including topological contamination, privacy leakage, cascading systemic failures, and the socio‐psychological risks of undue algorithmic persuasion. This work introduces a comprehensive, privacy‐preserving, explainable, and human‐centred MAS framework explicitly designed for collaborative clinical decision support. Rejecting the paradigm of a single, omniscient AI model with unrestricted data access, the proposed architecture operationalises clinical AI as a heavily regulated team of specialised, interacting agents. The framework defines six distinct agent classes: Data Stewardship, Modality, Explanation, Consensus, Safety Monitor, and Clinician Interface agents. Furthermore, the implementation is governed by six foundational design principles grounded in classic Autonomous Agents and MAS theory, adapted for modern generative systems. By enforcing minimum‐necessary data exchange, verifiable team‐level explainability, explicit hierarchical escalation, secure communication topologies, and human‐centred safety protocols, this framework provides both a domain‐grounded blueprint and a transferable set of robust design principles for deploying trustworthy MAS in high‐stakes environments.

Exciting modern statistical methods can bring about huge leaps forward in science. Elizabeth Eisenhauer, Maria Nareklishvili, Ayse Ulgen, Andrej Srakar and Dean Bodenham, members of the RSS Emerging Applications Section committee, describe some recent applications of three of these methods: natural experiments, citizen science and changepoint detection

Osmolality, concentration of solute particles, was rarely used for prognosis for COVID-19. By analyzing blood samples of more than 1300 COVID-19 patients from Tokat, Turkey (including 100 surviving and 30 deceased inpatients), we found calculated osmolality to be an excellent prognostic biomarker for mortality and significantly associated with hospitalization, independent from gender and age. Although calculated osmolality is defined as a weighted sum of sodium, glucose, and urea, the three are not necessarily independent. Other blood test biomarkers, ferritin, creatine, and chloride are also correlated with osmolality after conditioning on age. By applying a combination of collider analysis and mediation analysis, we design a pipeline to construct a causal model among all these variables in their relationship to osmolality. We confirm that while glucose and sodium are independent contributors of osmolality, glucose and urea, urea and sodium are correlated. We also conclude that ferritin and creatine are associated with osmolality through urea, and chloride’s association to osmolality is through sodium.

Background/Objectives: Gallstone disease, a prevalent and costly digestive system disorder, is influenced by multifactorial risk factors, some of which may predispose to malignancy. This study aims to evaluate the association between gallstone disease and malignancy using advanced machine learning (ML) algorithms. Methods: A dataset comprising approximately 1000 patients was analyzed, employing six ML methods: random forests (RFs), support vector machines (SVMs), multi-layer perceptron (MLP), MLP with PyTorch 2.3.1 (MLP_PT), naive Bayes (NB), and Tabular Prior-data Fitted Network (TabPFN). Comparative performance was assessed using Pearson correlation, sensitivity, specificity, Kappa, receiver operating characteristic (ROC), area under curve (AUC), and accuracy metrics. Results: Our results revealed that age, body mass index (BMI), and history of HRT were the most significant predictors of malignancy. Among the ML models, TabPFN emerged as the most effective, achieving superior performance across multiple evaluation criteria. Conclusions: This study highlights the potential of leveraging cutting-edge ML methodologies to uncover complex relationships in clinical datasets, offering a novel perspective on gallstone-related malignancy. By identifying critical risk factors and demonstrating the efficacy of TabPFN, this research provides actionable insights for predictive modeling and personalized patient management in clinical practice.

Background: Deep vein thrombosis (DVT) is a major cause of morbidity and mortality, including pulmonary embolism and post-thrombotic syndrome. This study aimed to assess the effectiveness of inflammatory indices, derived from routine laboratory parameters, in predicting DVT. Method: In this prospectively designed study, patients diagnosed with DVT through Doppler ultrasound at a tertiary healthcare center between December 2024 and February 2025, along with a control group confirmed to be DVT-free by Doppler ultrasound, were analyzed. Blood markers such as D-dimer, CRP, IL-6, and inflammatory indices (SIRI, MHR, PLR) were compared between groups. Statistical tests included chi-square, t -tests, logistic regression, and ROC curve analysis. Diagnostic performance was measured using odds ratios (ORs) with 95% confidence intervals (CIs). Result: Inflammatory markers (SIRI, MHR, D-dimer, CRP, IL-6) were significantly elevated in DVT cases. SIRI demonstrated high diagnostic accuracy (AUC: 0.934) with a threshold of 0.97. Combined SIRI and D-dimer analyses yielded 93% positive and 100% negative predictive accuracy. Conclusion: This study demonstrated that inflammatory markers, particularly SIRI, were elevated in patients with deep vein thrombosis (DVT) and carried high predictive value in this patient group. The combined use of SIRI and D-dimer provided high diagnostic accuracy for DVT. SIRI, a low-cost index calculable through routine blood tests, was shown to be more effective than other inflammatory markers in predicting DVT. Additionally, the combination of SIRI and D-dimer yielded high positive and negative predictive values for DVT diagnosis.

Diabetes mellitus (DM) is common among older adults hospitalized with lower respiratory tract infection, yet information on the impact of DM on disease severity is limited. This study retrospectively analyzed 46 Turkish patients infected with respiratory syncytial virus (RSV), with information on their comorbidities, co-infection status, and symptoms. Patients are grouped into four severity levels from mild to severe, according to lung parenchymal infiltration status and oxygen level. Similar to previously published studies, we found that comorbidities of diabetes, heart failure, hypertension, co-infection of any type, bacterial co-infection, and age are associated with the disease severity. Cough is the most common symptom (89%) followed by fever (26%) and myalgia, dyspnea, and weakness (around 20%). Using a second-order analysis (two-variable regression), we identified two independent risks for disease severity, the first is represented by diabetes, and the second is represented by bacterial co-infection. We observed two patients whose more severe symptoms were not associated with an older age, but associated with a combination of diabetes and bacterial co-infection. To confirm the true causality from the statistical correlation, further studies are needed.

COVID-19 survival data presents a special situation where not only the time-to-event period is short, but also the two events or outcome types, death and release from hospital, are mutually exclusive, leading to two cause-specific hazard ratios (csHR _d and csHR _r ). The eventual mortality/release outcome is also analyzed by logistic regression to obtain odds-ratio (OR). We have the following three empirical observations: (1) The magnitude of OR is an upper limit of the csHR _d : |log(OR)| ≥ |log(csHR _d )|. This relationship between OR and HR might be understood from the definition of the two quantities; (2) csHR _d and csHR _r point in opposite directions: log(csHR _d ) ⋅ log(csHR _r ) < 0; This relation is a direct consequence of the nature of the two events; and (3) there is a tendency for a reciprocal relation between csHR _d and csHR _r : csHR _d ∼ 1/csHR _r . Though an approximate reciprocal trend between the two hazard ratios is in indication that the same factor causing faster death also lead to slow recovery by a similar mechanism, and vice versa, a quantitative relation between csHR _d and csHR _r in this context is not obvious. These results may help future analyses of data from COVID-19 or other similar diseases, in particular if the deceased patients are lacking, whereas surviving patients are abundant.