Djordje S Popovic,1 Theocharis Koufakis,2 Dimitrios Patoulias,2 Anca Pantea Stoian,3 Nikolaos Papanas4 1Clinic for Endocrinology, Diabetes and Metabolic Disorders, Clinical Centre of Vojvodina, Medical Faculty, University of Novi Sad, Novi Sad, Serbia; 2Second Propaedeutic Department of Internal Medicine, Hippokration General Hospital, Aristotle University of Thessaloniki, Thessaloniki, Greece; 3Department of Diabetes, Nutrition and Metabolic Diseases, “Carol Davila” University of Medicine and Pharmacy, Bucharest, Romania; 4Diabetes Centre, Second Department of Internal Medicine, Democritus University of Thrace, University Hospital of Alexandroupolis, Alexandroupolis, GreeceCorrespondence: Djordje S Popovic, Clinic for Endocrinology, Diabetes and Metabolic Disorders, Clinical Centre of Vojvodina, Medical Faculty, University of Novi Sad, Hajduk Veljkova 1, Novi Sad, 21000, Serbia, Tel +38163551606, Email pitstop021@gmail.com; djordje.popovic@mf.uns.ac.rsAbstract: Recreational diving with self-contained underwater breathing devices is gaining popularity worldwide as a sport and leisure activity. People living with type 1 diabetes mellitus (PLT1D) are no exception, although historically diabetes mellitus, especially insulin-treated, has been described as an absolute contra-indication for diving. However, based on observational data collected by the Divers Alert Network, the presence of background diabetes mellitus became only a relative contraindication for those without significant co-morbidities or long-term complications. Regarding diving activities among PLT1D, the primary concern is the risk of hypoglycaemia, especially in those with impaired awareness. Furthermore, symptoms consistent with hypoglycaemia could be confused with those originating from other factors related to diving. Although avoidance of hypoglycaemia is imperative among PLT1D practicing diving, the risk of severe hyperglycaemia should also be minimised. Continuous glucose monitoring (CGM) nowadays represents the standard of care for PLT1D, but its accuracy during diving activities is still a matter of debate. This commentary aims to summarize the existing data on accuracy, durability, and underwater performance of different CGM devices among PLT1D who engage in diving, and to call for additional research in the field. Based on available results, the application of real-time CGM still requires extreme caution since none of the existing systems has so far met the standards for accurate use in underwater conditions. Further improvements of contemporary CGM devices, validated through large-scale trials, are necessary before their widespread implementation among PLT1D practicing diving. Such advances should further enhance safety during this popular activity.Keywords: continuous glucose monitoring, diving, type 1 diabetes mellitus
Xenofon Baraliakos, Ye Eun Lee, Soyeon Park
et al.
Abstract Background CT-P13 SC, a new subcutaneous (SC) formulation of biosimilar infliximab (IFX), was approved by the European Medicines Agency in 2020 for the treatment of radiographic axial spondyloarthritis (axSpA) in adult patients. The present study aimed to assess the real-world outcomes of CT-P13 SC (SC IFX) as a treatment for both radiographic and non-radiographic axSpA. Methods Data were drawn from the Adelphi Real World axSpA Disease Specific Programme™, a cross-sectional survey with retrospective data collection in France, Germany, Italy, Spain, and the UK between June 2023 and June 2024. Rheumatologists and their patients with axSpA completed questionnaires on patient demographics, clinical characteristics, and treatment satisfaction. Outcomes for patients on SC IFX were analyzed, with additional evaluations based on baseline characteristics and treatment patterns. Such outcomes were also compared with patients receiving other advanced therapies, including biologics and Janus kinase inhibitors. Results In total, 191 patients were evaluated. The mean patient age was 44.5 years, and most were male (117 [61.3%]). Baseline characteristics were similar between radiographic and non-radiographic axSpA patients, with a higher proportion of patients in the radiographic group (116 [60.7%] vs. 75 [39.3%]). Significant improvements in disease severity were observed with SC IFX from treatment initiation to data collection (severe disease: 37.4% to 3.2%), along with significantly lower levels of pain and fatigue, and fewer tender entheseal points and affected joints (all p < 0.0001). SC IFX treatment was also associated with improvements in musculoskeletal, extra-articular, systemic, and functional symptoms, and inflammatory imaging features. Subgroup analysis showed that SC IFX was effective across various patient populations with differing characteristics such as age, prior experience with advanced therapies, or coexisting conditions. SC IFX showed high treatment satisfaction for both physicians and patients. No new safety signals were reported. Conclusions In this real-world study, SC IFX demonstrated clinical effectiveness in both radiographic and non-radiographic axSpA, with consistent results across diverse patient characteristics. Both physicians and patients reported high satisfaction with no new safety concerns. This analysis suggests that SC IFX can be an effective, convenient, and well-tolerated treatment option for diverse axSpA patient populations.
Acute myeloid leukemia (AML) requires strict monitoring of minimal residual disease (MRD). We aimed to explore if MRD monitoring on peripheral blood (PB) represents a valid and less invasive alternative to repeated bone marrow (BM) aspirates in adults with AML undergoing allogeneic hematopoietic stem cell transplantation (HSCT) at Federico II University Hospital in Naples. Molecular markers to follow on PB included NPM1, WT1, CBF-AML, BCR-ABL, and PML-RARA. MRD positivity or biochemical/clinical suspicion of disease relapse prompted to BM aspirate. Forty-one patients (17 males and 24 females) underwent HSCT for AML between July 2020 and November 2024. Median age at transplant was 58 (range 30-72) years. MRD monitoring post-transplant was performed using NPM1 in 18, WT1 in 18, CBF-AML in 2, BCR-ABL in 2 and PML-RARA in 1 patient. Disease status at transplant was CR1 in 25, CR2 in 10 (including 6 with MRD+ at transplant) and active disease in 6 patients. Conditioning regimen was myeloablative, reduced-intensity or sequential in 20, 14 and 7 patients, respectively. Stem cell source was BM in 12 and PB in 29 cases. Donors were HLA-identical in 11, matched unrelated in 21, haploidentical in 9 cases. Graft-versus-host disease (GVHD) prophylaxis consisted of cyclosporine associated to mycophenolate mofetil (n=29) or methotrexate (n=12). ATG or PTCY were also used in 27 and 11 patients, respectively, while 3 received both. All patients achieved neutrophil engraftment with a median time of 14 (range 8-23) days. All but one patient achieved platelet engraftment with a median time of 14 (range 8-34) days. Grade II-IV acute GVHD occurred in 6 while chronic GVHD in 8 patients (mild, n=5; moderate, n=2; extensive, n=1). MRD positivity during follow-up was detected in 10 cases, with 3 contextual overt relapses on BM and one occurring one month later. Preemptive treatments were introduced in 6 cases. Later on, 2 patients experienced overt hematological relapse 15 and 12 months from previous molecular relapse while one patient experienced extramedullary relapse. Preemptive treatments were used in 6 cases, prophylactic treatments were used in 4, while 7 patients were treated for overt relapse. With a median follow-up of 23 (range 5-57) months, 2-year PFS and OS were 67±9% and 85±6%. In our experience, MRD monitoring using PB represented a sensitive strategy to detect molecular or overt relapse avoiding repeated bone marrow aspirations during the follow-up.
Arti Virkud, Jessie K. Edwards, Michele Jonsson Funk
et al.
Identifying optimal medical treatments to improve survival has long been a critical goal of pharmacoepidemiology. Traditionally, we use an average treatment effect measure to compare outcomes between treatment plans. However, new methods leveraging advantages of machine learning combined with the foundational tenets of causal inference are offering an alternative to the average treatment effect. Here, we use three unique, precision medicine algorithms (random forests, residual weighted learning, efficient augmentation relaxed learning) to identify optimal treatment rules where patients receive the optimal treatment as indicated by their clinical history. First, we present a simple hypothetical example and a real-world application among heart failure patients using Medicare claims data. We next demonstrate how the optimal treatment rule improves the absolute risk in a hypothetical, three-modifier setting. Finally, we identify an optimal treatment rule that optimizes the time to outcome in a real-world heart failure setting. In both examples, we compare the average time to death under the optimized, tailored treatment rule with the average time to death under a universal treatment rule to show the benefit of precision medicine methods. The improvement under the optimal treatment rule in the real-world setting is greatest (additional ~9 days under the tailored rule) for survival time free of heart failure readmission.
Large Language Models (LLMs) are transforming healthcare through the development of LLM-based agents that can understand, reason about, and assist with medical tasks. This survey provides a comprehensive review of LLM-based agents in medicine, examining their architectures, applications, and challenges. We analyze the key components of medical agent systems, including system profiles, clinical planning mechanisms, medical reasoning frameworks, and external capacity enhancement. The survey covers major application scenarios such as clinical decision support, medical documentation, training simulations, and healthcare service optimization. We discuss evaluation frameworks and metrics used to assess these agents' performance in healthcare settings. While LLM-based agents show promise in enhancing healthcare delivery, several challenges remain, including hallucination management, multimodal integration, implementation barriers, and ethical considerations. The survey concludes by highlighting future research directions, including advances in medical reasoning inspired by recent developments in LLM architectures, integration with physical systems, and improvements in training simulations. This work provides researchers and practitioners with a structured overview of the current state and future prospects of LLM-based agents in medicine.
Davide Belluomo, Tiziana Calamoneri, Giacomo Paesani
et al.
We present a new unified graph-based representation of medical data, combining genetic information and medical records of patients with medical knowledge via a unique knowledge graph. This approach allows us to infer meaningful information and explanations that would be unavailable by looking at each data set separately. The systematic use of different databases, managed throughout the built knowledge graph, gives new insights toward a better understanding of oncology medicine. Indeed, we reduce some useful medical tasks to well-known problems in theoretical computer science for which efficient algorithms exist.
While large language models (LLMs) have achieved state-of-the-art performance on a wide range of medical question answering (QA) tasks, they still face challenges with hallucinations and outdated knowledge. Retrieval-augmented generation (RAG) is a promising solution and has been widely adopted. However, a RAG system can involve multiple flexible components, and there is a lack of best practices regarding the optimal RAG setting for various medical purposes. To systematically evaluate such systems, we propose the Medical Information Retrieval-Augmented Generation Evaluation (MIRAGE), a first-of-its-kind benchmark including 7,663 questions from five medical QA datasets. Using MIRAGE, we conducted large-scale experiments with over 1.8 trillion prompt tokens on 41 combinations of different corpora, retrievers, and backbone LLMs through the MedRAG toolkit introduced in this work. Overall, MedRAG improves the accuracy of six different LLMs by up to 18% over chain-of-thought prompting, elevating the performance of GPT-3.5 and Mixtral to GPT-4-level. Our results show that the combination of various medical corpora and retrievers achieves the best performance. In addition, we discovered a log-linear scaling property and the "lost-in-the-middle" effects in medical RAG. We believe our comprehensive evaluations can serve as practical guidelines for implementing RAG systems for medicine.
AIAltMed is a cutting-edge platform designed for drug discovery and repurposing. It utilizes Tanimoto similarity to identify structurally similar non-medicinal compounds to known medicinal ones. This preprint introduces AIAltMed, discusses the concept of `AI-driven alternative medicine,' evaluates Tanimoto similarity's advantages and limitations, and details the system's architecture. Furthermore, it explores the benefits of extending the system to include PubChem and outlines a corresponding implementation strategy.
M. A. Shevchenko, D. G. Garbuz, A. I. Davletshin
et al.
Heat shock proteins 70 kDa (HSP70) protect intracellular proteins from the damaging effects of stress factors of various natures. Moreover, HSP70 play an important role in the vital activity of cells under normal physiological conditions, performing chaperone functions. These functions are realized in the intracellular space; however, in some cases, these proteins are also found on the cell surface and in the extracellular environment. The causes and mechanisms of HSP70 translocation to the cell surface and secretion into the extracellular space have not yet been well understood, but such an unusual localization of HSP70 activates the immune system. The surface HSP70 and their extracellular pool stimulate the cytotoxic activity of NK cells. However, direct experimental evidence for the internalization of HSP70 molecules by NK cells has not yet been demonstrated. This paper presents the results of the interaction of the extracellular HSP70 pool with NK cells from the peripheral blood. The results demonstrated the confirmation of the internalization of exogenous HSP70 molecules by NK cells. To this end, fluorescently labeled recombinant stress-inducible human HSP70 were obtained. The electrophoretic data indicated the absence of protein degradation during the labeling process, the purity and stability of the modified protein. To assess the interaction of HSP70 with NK cells, the fluorescently labeled HSP70 was added to an in vitro culture of NK cells isolated by magnetic separation from the peripheral blood mononuclear fraction and analyzed by confocal microscopy. This analysis indicated that living NK cells internalize extracellular HSP70 with localization both in lysosomes and in phagosomes. Our experiments illustrated for the first time the process of penetration of the extracellular form of HSP70 into these cells. The results suggest that the activation of NK cells under the action of exogenous HSP70 could be associated with the internalization of these protein molecules.
BackgroundAnoikis is a programmed cell death process that was proven to be associated with cancer. Uroepithelial carcinoma of the bladder (BLCA) is a malignant disease of the urinary tract and has a strong metastatic potential. To determine whether anoikis-associated genes can predict the prognosis of BLCA accurately, we evaluated the prognostic value of anoikis-associated genes in BLCA and constructed the best model to predict prognosis.MethodThe BLCA transcriptome data were downloaded from TCGA and GEO databases, and genes with differential expression were selected and then clustered using non-negative matrix factorization (NMF). The genes with the most correlation with anoikis were screened and identified using univariate Cox regression, lasso regression, and multivariate Cox regression. The GEO dataset was used for external validation. Nomograms were created based on risk characteristics in combination with clinical variants and the performance of the model was validated with receiver operating characteristic (ROC) curves. The immunotherapeutic significance of this risk score was assessed using the immune phenomenon score (IPS). IC50 values of predictive chemotherapeutic agents were calculated. Finally, we used RT-qPCR to determine the mRNA expression of four genes, CALR, FASN, CASP6, and RAD9A.ResultWe screened 406 tumor samples and 19 normal tissue samples from the TCGA database. Based on anoikis-associated genes, we classified patients into two subtypes (C1 and C2) using NMF method. Subsequently, nine core genes were screened by multiple methods after analysis, which were used to construct risk profiles. The design of nomograms based on risk profiles and clinical variables, ROC, and calibration curves confirmed that the model could well have the ability to predict the survival of BLCA patients at 1, 3, and 5 years. By predicting the IC50 values of chemotherapeutic drugs, it was learned that the high-risk group (HRG) was more susceptible to paclitaxel, gemcitabine, and cisplatin, and the low-risk group (LRG) was more susceptible to veriparib and afatinib.ConclusionIn summary, the risk score of anoikis-associated genes can be applied as a predictor to predict the prognosis of BLCA in clinical practice.
Mansour S Aljabry, Fahad Alabbas, Ghaleb Elyamany
et al.
BACKGROUND: Rare bleeding disorder (RBDs) encompasses a deficiency of one or more of FXIII, FXI, FX, FVII, FV, FII, and FI clotting factors, leading to bleeding disorders with variable presentations and outcomes ranging from none or minimal to life-threatening events. RBDs are still underdiagnosed and underreported, especially in Saudi population with a high prevalence of consanguinity.
OBJECTIVES: The study aimed to determine the frequency of RBDs, grading of their bleeding severity, and assessment of clinical manifestations and management of RBDs in tertiary Saudi Arabian hospitals.
DESIGN AND SETTINGS: This retrospective study of RBDs describes the clinicopathological features of refereed cases to both Prince Sultan Military Medical City and King Khaled University Hospital in Riyadh, Saudi Arabia, from September 2018 to September 2021. Any patient who had already been diagnosed or suspected to have RBDs was enrolled in the study.
PATIENTS AND METHODS: Patient's medical records were reviewed for demographic data, clinical presentations, bleeding and family history, consanguinity, treatment outcomes, and molecular testing. Samples were run in specialized coagulation laboratories. Patients with liver dysfunction or acquired factor deficiency were excluded. Patients were categorized into four groups according to the severity of bleeding episodes: asymptomatic, Grade I, Grade II, and Grade III.
RESULTS: A total of 26 cases with RBDs were identified during the study period. Most of the included patients are males and pediatrics (<14 years) representing 15 (57.7%) and 14 (53.8%), respectively. FVII was the most common factor deficiency encountered in 9 (35%) patients, followed by FXIII in 5 (19%), FXI in 4 (15%), FX in 3 (11.5%), FV in 3 (11.5%), and combined factor deficiency in 2 (8%) patients. 17 (65.4%) RBD patients presented with bleeding manifestation either with Grade I (9%), Grade II (39%), or Grade III (15%), whereas 47% were asymptomatic.
CONCLUSION: The study emphasizes on importance of establishing a national registry of RBDs in Saudi Arabia and the need for further genetic studies to clarify the genotype/phenotype relationships.
Diseases of the circulatory (Cardiovascular) system
Bedros Taslakian, Larry E. Miller, Tarub S. Mabud
et al.
Objective: Genicular artery embolization (GAE) is a novel, minimally invasive procedure for treatment of knee osteoarthritis (OA). This meta-analysis investigated the safety and effectiveness of this procedure. Design: Outcomes of this systematic review with meta-analysis were technical success, knee pain visual analog scale (VAS; 0–100 scale), WOMAC Total Score (0–100 scale), retreatment rate, and adverse events. Continuous outcomes were calculated as the weighted mean difference (WMD) versus baseline. Minimal clinically important difference (MCID) and substantial clinical benefit (SCB) rates were estimated in Monte Carlo simulations. Rates of total knee replacement and repeat GAE were calculated using life-table methods. Results: In 10 groups (9 studies; 270 patients; 339 knees), GAE technical success was 99.7%. Over 12 months, the WMD ranged from −34 to −39 at each follow-up for VAS score and −28 to −34 for WOMAC Total score (all p < 0.001). At 12 months, 78% met the MCID for VAS score; 92% met the MCID for WOMAC Total score, and 78% met the SCB for WOMAC Total score. Higher baseline knee pain severity was associated with greater improvements in knee pain. Over 2 years, 5.2% of patients underwent total knee replacement and 8.3% received repeat GAE. Adverse events were minor, with transient skin discoloration as the most common (11.6%). Conclusions: Limited evidence suggests that GAE is a safe procedure that confers improvement in knee OA symptoms at established MCID thresholds. Patients with greater knee pain severity may be more responsive to GAE.
Pre-training and fine-tuning have emerged as a promising paradigm across various natural language processing (NLP) tasks. The effectiveness of pretrained large language models (LLM) has witnessed further enhancement, holding potential for applications in the field of medicine, particularly in the context of Traditional Chinese Medicine (TCM). However, the application of these general models to specific domains often yields suboptimal results, primarily due to challenges like lack of domain knowledge, unique objectives, and computational efficiency. Furthermore, their effectiveness in specialized domains, such as Traditional Chinese Medicine, requires comprehensive evaluation. To address the above issues, we propose a novel domain specific TCMDA (TCM Domain Adaptation) approach, efficient pre-training with domain-specific corpus. Specifically, we first construct a large TCM-specific corpus, TCM-Corpus-1B, by identifying domain keywords and retreving from general corpus. Then, our TCMDA leverages the LoRA which freezes the pretrained model's weights and uses rank decomposition matrices to efficiently train specific dense layers for pre-training and fine-tuning, efficiently aligning the model with TCM-related tasks, namely TCM-GPT-7B. We further conducted extensive experiments on two TCM tasks, including TCM examination and TCM diagnosis. TCM-GPT-7B archived the best performance across both datasets, outperforming other models by relative increments of 17% and 12% in accuracy, respectively. To the best of our knowledge, our study represents the pioneering validation of domain adaptation of a large language model with 7 billion parameters in TCM domain. We will release both TCMCorpus-1B and TCM-GPT-7B model once accepted to facilitate interdisciplinary development in TCM and NLP, serving as the foundation for further study.
Jennifer H. Therkorn, Sean Hu, Anays M. Sotolongo
et al.
Abstract Background Service member exposure to explosive blast overpressure waves is common with considerable attention to traumatic brain injury (TBI) and neuropsychological sequalae. Less is known about the impacts on the respiratory system, particularly long-term effects, despite vulnerability to overpressure. Using a national registry, we previously observed an independent relationship between self-reported blast exposure and respiratory symptoms; however, the impact on objective measures of pulmonary function is poorly understood. Methods 307 Veterans referred to our national specialty center for post-deployment health concerns underwent a comprehensive multi-day evaluation that included complete pulmonary function testing (PFT), occupational and environmental medicine history, neuropsychological or psychological evaluation. We developed an a priori chart abstraction process and template to classify Veterans into blast exposure groups: (1) none, (2) single-mild, or (3) multiple-mild. This template focused primarily on clinician documented notes of blast related TBI that were used as proxy for blast overpressure injury to thorax. PFT variables characterizing flow (FEV1%; %∆FEV1), volume (TLC%), diffusion (DLCO%) and respiratory mechanics (forced oscillometry) were selected for analysis. Results Veterans (40.5 ± 9.7 years; 16.3% female) were referred 8.6 ± 3.6 years after their last deployment and presented with considerable comorbid conditions and health problems (e.g., 62% post-traumatic stress, 55% dyspnea). After chart abstraction, Veterans were assigned to none (n = 208), single mild (n = 52) and multiple mild (n = 47) blast exposure groups. Among the blast exposed, clinicians documented 73.7% were < 50 m from the blast and 40.4% were physically moved by blast. PFT outcome measures were similar across all groups (p value range: 0.10–0.99). Conclusions In this referred sample of deployed Veterans, PFT measures of flow, volume, diffusion, and respiratory mechanics were not associated with clinician documented blast exposure per the retrospective chart abstraction methodology applied. Yet, these clinical findings suggest future research should determine and assess distinction between Veteran recollections of perceived blast experiences versus overpressure wave exposure to the respiratory system.
Public health is the most recent of the biomedical sciences to be seduced by the trendy moniker "precision." Advocates for "precision public health" (PPH) call for a data-driven, computational approach to public health, leveraging swaths of genomic "big data" to inform public health decision-making. Yet, like precision medicine, PPH oversells the value of genomic data to determine health outcomes, but on a population-level. A large historical literature has shown that over-emphasizing heredity tends to disproportionately harm underserved minorities and disadvantaged communities. By comparing and contrasting PPH with an earlier attempt at using big data and genetics, in the Progressive era (1890-1920), we highlight some potential risks of a genotype-driven preventive public health. We conclude by suggesting that such risks may be avoided by prioritizing data integration across many levels of analysis, from the molecular to the social.
Background: Despite increasing representation of women in medicine, gender bias remains pervasive. The authors sought to evaluate speaker introductions by gender in the grand rounds of multiple specialties at a large academic institution to understand the cultural context of this behavior and identify predictors of formality. Materials and Methods: The authors reviewed grand rounds recordings of speakers with doctorates presenting to the departments of family medicine, general surgery, internal medicine, obstetrics and gynecology, and pediatrics at one institution from 2014 to 2019. The primary outcome was whether a speaker's professional title was used as the first form of address. The authors assessed factors correlated with professional introduction using multivariable logistic regression. Results: Speakers were introduced professionally in 346/615 recordings (56.3%). Female introducers were more likely to introduce speakers professionally (odds ratio [OR]: 2.52). A significant interaction existed between speaker gender and home institution: female speakers visiting from an external institution were less likely than male external speakers to be introduced professionally (OR: 0.49), whereas female speakers internal to the institution were more likely to be introduced professionally than male internal speakers (OR: 1.75). Use of professional titles varied by specialty and was higher than average for family medicine (83.2%), surgery (75.8%), and pediatrics (64.0%) and lower for internal medicine (37.5%) and obstetrics and gynecology (50.7%). Conclusions: These findings suggest a complex relationship between gender and formality of introduction that merits further investigation. Understanding differences in culture across specialties is important to inform efforts to promote equity.