A scalable federated learning solution for secondary care using low-cost microcomputing: privacy-preserving development and evaluation of a COVID-19 screening test in UK hospitals.

BACKGROUND: Multicentre training could reduce biases in medical artificial intelligence (AI); however, ethical, legal, and technical considerations can constrain the ability of hospitals to share data. Federated learning enables institutions to participate in algorithm development while retaining custody of their data but uptake in hospitals has been limited, possibly as deployment requires specialist software and technical expertise at each site. We previously developed an artificial intelligence-driven screening test for COVID-19 in emergency departments, known as CURIAL-Lab, which uses vital signs and blood tests that are routinely available within 1 h of a patient's arrival. Here we aimed to federate our COVID-19 screening test by developing an easy-to-use embedded system-which we introduce as full-stack federated learning-to train and evaluate machine learning models across four UK hospital groups without centralising patient data. METHODS: We supplied a Raspberry Pi 4 Model B preloaded with our federated learning software pipeline to four National Health Service (NHS) hospital groups in the UK: Oxford University Hospitals NHS Foundation Trust (OUH; through the locally linked research University, University of Oxford), University Hospitals Birmingham NHS Foundation Trust (UHB), Bedfordshire Hospitals NHS Foundation Trust (BH), and Portsmouth Hospitals University NHS Trust (PUH). OUH, PUH, and UHB participated in federated training, training a deep neural network and logistic regressor over 150 rounds to form and calibrate a global model to predict COVID-19 status, using clinical data from patients admitted before the pandemic (COVID-19-negative) and testing positive for COVID-19 during the first wave of the pandemic. We conducted a federated evaluation of the global model for admissions during the second wave of the pandemic at OUH, PUH, and externally at BH. For OUH and PUH, we additionally performed local fine-tuning of the global model using the sites' individual training data, forming a site-tuned model, and evaluated the resultant model for admissions during the second wave of the pandemic. This study included data collected between Dec 1, 2018, and March 1, 2021; the exact date ranges used varied by site. The primary outcome was overall model performance, measured as the area under the receiver operating characteristic curve (AUROC). Removable micro secure digital (microSD) storage was destroyed on study completion. FINDINGS: Clinical data from 130 941 patients (1772 COVID-19-positive), routinely collected across three hospital groups (OUH, PUH, and UHB), were included in federated training. The evaluation step included data from 32 986 patients (3549 COVID-19-positive) attending OUH, PUH, or BH during the second wave of the pandemic. Federated training of a global deep neural network classifier improved upon performance of models trained locally in terms of AUROC by a mean of 27·6% (SD 2·2): AUROC increased from 0·574 (95% CI 0·560-0·589) at OUH and 0·622 (0·608-0·637) at PUH using the locally trained models to 0·872 (0·862-0·882) at OUH and 0·876 (0·865-0·886) at PUH using the federated global model. Performance improvement was smaller for a logistic regression model, with a mean increase in AUROC of 13·9% (0·5%). During federated external evaluation at BH, AUROC for the global deep neural network model was 0·917 (0·893-0·942), with 89·7% sensitivity (83·6-93·6) and 76·6% specificity (73·9-79·1). Site-specific tuning of the global model did not significantly improve performance (change in AUROC <0·01). INTERPRETATION: We developed an embedded system for federated learning, using microcomputing to optimise for ease of deployment. We deployed full-stack federated learning across four UK hospital groups to develop a COVID-19 screening test without centralising patient data. Federation improved model performance, and the resultant global models were generalisable. Full-stack federated learning could enable hospitals to contribute to AI development at low cost and without specialist technical expertise at each site. FUNDING: The Wellcome Trust, University of Oxford Medical and Life Sciences Translational Fund.

Development of Asthma-Like Symptoms After COVID-19: A Cross-Sectional Study in Dubai, United Arab Emirates.

Background Coronaviruses are viral agents that commonly infect animals, but have the ability to cause respiratory illness in humans, exemplified by the ongoing novel coronavirus outbreak (COVID-19). Due to the sparse literature on the effects of COVID-19 on the respiratory system, and the possible development of persistent asthma-like symptoms after infection, this cross-sectional analysis was performed in order to compare the clinical and investigative parameters between post-COVID patients and asthmatic patients. Methods A retrospective cross-sectional study was conducted on patients with prior history of COVID-19 infection that presented to the pulmonology or respiratory outpatient clinics with asthma-like symptoms and were subsequently compared to known asthmatic patients with absent history of prior COVID-19 infection, in order to evaluate the degree of similarity between both cohorts. In this study, asthma-like symptoms were defined as: (i) cough, (ii) wheezing, (iii) chest tightness, and (iv) shortness of breath. Moreover, comparisons of investigative parameters were also performed, including (i) fractional exhaled nitric oxide (FeNO), (ii) serum immunoglobulin E (IgE), (iii) absolute eosinophil counts, and (iv) qualitative spirometry results. All statistical analyses were conducted via chi-squared testing for categorical variables, and independent t-test for continuous variables. Results In this study, there were a total of 76 patients included that conformed to the eligibility criteria, including 39 patients with post-COVID symptoms with absent history of asthma or other respiratory illnesses, and 37 patients with known asthma with absent history of prior COVID-19 infection or other respiratory illnesses. Overall, this study revealed the similarities between both cohorts with respect to the incidence of cough, chest tightness, and shortness of breath. Moreover, there were similarities between the serum IgE and spirometry results. However, there were differences within the complaint of wheeze, FeNO values, and eosinophil counts between both cohorts. The placement of post-COVID patients on bronchodilator therapy involving inhaled corticosteroids and long-acting beta-agonists revealed improvement in all follow-up patients. Conclusion In conclusion, there was considerable similarity in the complaint of asthma-like symptoms after COVID-19 infection, associated with an improvement after the use of bronchodilator therapy, indicating the potential role of anti-asthma therapy (e.g., bronchodilator therapy) in managing post-COVID asthma-like symptoms. In order to validate our conclusion, further comprehensive studies with robust methodologies and larger sample populations are encouraged.

Immunomodulatory properties of morphine and the hypothesised role of long-term opioid use in the immunopathogenesis of tuberculosis.

Epidemiological studies have shown high tuberculosis (TB) prevalence among chronic opioid users. Opioid receptors are found on multiple immune cells and immunomodulatory properties of opioids could be a contributory factor for ensuing immunosuppression and development or reactivation of TB. Toll-like receptors (TLR) mediate an immune response against microbial pathogens, including Mycobacterium tuberculosis. Mycobacterial antigens and opioids co-stimulate TLRs 2/4/9 in immune cells, with resulting receptor cross-talk via multiple cytosolic secondary messengers, leading to significant immunomodulatory downstream effects. Blockade of specific immune pathways involved in the host defence against TB by morphine may play a critical role in causing tuberculosis among chronic morphine users despite multiple confounding factors such as socioeconomic deprivation, Human immunodeficiency virus co-infection and malnutrition. In this review, we map out immune pathways involved when immune cells are co-stimulated with mycobacterial antigens and morphine to explore a potential immunopathological basis for TB amongst long-term opioid users.

Clinical severity classes in COVID-19 pneumonia have distinct immunological profiles, facilitating risk stratification by machine learning.

Objective: Clinical triage in coronavirus disease 2019 (COVID-19) places a heavy burden on senior clinicians during a pandemic situation. However, risk stratification based on serum biomarker bioprofiling could be implemented by a larger, nonspecialist workforce. Method: Measures of Complement Activation and inflammation in patientS with CoronAvirus DisEase 2019 (CASCADE) patients (n = 72), (clinicaltrials.gov: NCT04453527), classified as mild, moderate, or severe (by support needed to maintain SpO2 > 93%), and healthy controls (HC, n = 20), were bioprofiled using 76 immunological biomarkers and compared using ANOVA. Spearman correlation analysis on biomarker pairs was visualised via heatmaps. Linear Discriminant Analysis (LDA) models were generated to identify patients likely to deteriorate. An X-Gradient-boost (XGB) model trained on CASCADE data to triage patients as mild, moderate, and severe was retrospectively employed to classify COROnavirus Nomacopan Emergency Treatment for covid 19 infected patients with early signs of respiratory distress (CORONET) patients (n = 7) treated with nomacopan. Results: The LDA models distinctly discriminated between deteriorators, nondeteriorators, and HC, with IL-27, IP-10, MDC, ferritin, C5, and sC5b-9 among the key predictor variables during deterioration. C3a and C5 were elevated in all severity classes vs. HC (p < 0.05). sC5b-9 was elevated in the "moderate" and "severe" categories vs. HC (p < 0.001). Heatmap analysis shows a pairwise increase of negatively correlated pairs with IL-27. The XGB model indicated sC5b-9, IL-8, MCP1, and prothrombin F1 and F2 were key discriminators in nomacopan-treated patients (CORONET study). Conclusion: Distinct immunological fingerprints from serum biomarkers exist within different severity classes of COVID-19, and harnessing them using machine learning enabled the development of clinically useful triage and prognostic tools. Complement-mediated lung injury plays a key role in COVID-19 pneumonia, and preliminary results hint at the usefulness of a C5 inhibitor in COVID-19 recovery.

Capability, Opportunity, and Motivation Model for Behavior Change in People With Asthma: Protocol for a Cross-Sectional Study.

BACKGROUND: Asthma is a common lung condition that cannot be cured, but it can usually be effectively managed using available treatments. Despite this, it is widely acknowledged that 70% of patients do not adhere to their asthma treatment. Personalizing treatment by providing the most appropriate interventions based on the patient's psychological or behavioral needs produces successful behavior change. However, health care providers have limited available resources to deliver a patient-centered approach for their psychological or behavioral needs, resulting in a current one-size-fits-all strategy due to the nonfeasible nature of existing surveys. The solution would be to provide health care professionals with a clinically feasible questionnaire that identifies the patient's personal psychological and behavioral factors related to adherence. OBJECTIVE: We aim to apply the capability, opportunity, and motivation model of behavior change (COM-B) questionnaire to detect a patient's perceived psychological and behavioral barriers to adherence. Additionally, we aim to explore the key psychological and behavioral barriers indicated by the COM-B questionnaire and adherence to treatment in patients with confirmed asthma with heterogeneous severity. Exploratory objectives will include a focus on the associations between the COM-B questionnaire responses and asthma phenotype, including clinical, biological, psychosocial, and behavioral components. METHODS: In a single visit, participants visiting Portsmouth Hospital's asthma clinic with a diagnosis of asthma will be asked to complete a 20-minute questionnaire on an iPad about their psychological and behavioral barriers following the theoretical domains framework and capability, opportunity, and motivation model. Participants' data are routinely collected, including demographics, asthma characteristics, asthma control, asthma quality of life, and medication regime, which will be recorded on an electronic data capture form. RESULTS: The study is already underway, and it is anticipated that the results will be available by early 2023. CONCLUSIONS: The COM-B asthma study will investigate an easily accessible theory-based tool (a questionnaire) for identifying psychological and behavioral barriers in patients with asthma who are not adhering to their treatment. This will provide useful information on the behavioral barriers to asthma adherence and whether or not a questionnaire can be used to identify these needs. The highlighted barriers will improve health care professionals' knowledge of this important subject, and participants will benefit from the study by removing their barriers. Overall, this will enable health care professionals to use effective individualized interventions to support improved medication adherence while also recognizing and meeting the psychological needs of patients with asthma. TRIAL REGISTRATION: ClinicalTrials.gov NCT05643924; https://clinicaltrials.gov/ct2/show/NCT05643924. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/44710.

Combining viral genomics and clinical data to assess risk factors for severe COVID-19 (mortality, ICU admission, or intubation) amongst hospital patients in a large acute UK NHS hospital Trust.

Throughout the COVID-19 pandemic, valuable datasets have been collected on the effects of the virus SARS-CoV-2. In this study, we combined whole genome sequencing data with clinical data (including clinical outcomes, demographics, comorbidity, treatment information) for 929 patient cases seen at a large UK hospital Trust between March 2020 and May 2021. We identified associations between acute physiological status and three measures of disease severity; admission to the intensive care unit (ICU), requirement for intubation, and mortality. Whilst the maximum National Early Warning Score (NEWS2) was moderately associated with severe COVID-19 (A = 0.48), the admission NEWS2 was only weakly associated (A = 0.17), suggesting it is ineffective as an early predictor of severity. Patient outcome was weakly associated with myriad factors linked to acute physiological status and human genetics, including age, sex and pre-existing conditions. Overall, we found no significant links between viral genomics and severe outcomes, but saw evidence that variant subtype may impact relative risk for certain sub-populations. Specific mutations of SARS-CoV-2 appear to have little impact on overall severity risk in these data, suggesting that emerging SARS-CoV-2 variants do not result in more severe patient outcomes. However, our results show that determining a causal relationship between mutations and severe COVID-19 in the viral genome is challenging. Whilst improved understanding of the evolution of SARS-CoV-2 has been achieved through genomics, few studies on how these evolutionary changes impact on clinical outcomes have been seen due to complexities associated with data linkage. By combining viral genomics with patient records in a large acute UK hospital, this study represents a significant resource for understanding risk factors associated with COVID-19 severity. However, further understanding will likely arise from studies of the role of host genetics on disease progression.

Immuno-informatics analysis predicts B and T cell consensus epitopes for designing peptide vaccine against SARS-CoV-2 with 99.82% global population coverage.

The current global pandemic due to Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has taken a substantial number of lives across the world. Although few vaccines have been rolled-out, a number of vaccine candidates are still under clinical trials at various pharmaceutical companies and laboratories around the world. Considering the intrinsic nature of viruses in mutating and evolving over time, persistent efforts are needed to develop better vaccine candidates. In this study, various immuno-informatics tools and bioinformatics databases were deployed to derive consensus B-cell and T-cell epitope sequences of SARS-CoV-2 spike glycoprotein. This approach has identified four potential epitopes which have the capability to initiate both antibody and cell-mediated immune responses, are non-allergenic and do not trigger autoimmunity. These peptide sequences were also evaluated to show 99.82% of global population coverage based on the genotypic frequencies of HLA binding alleles for both MHC class-I and class-II and are unique for SARS-CoV-2 isolated from human as a host species. Epitope number 2 alone had a global population coverage of 98.2%. Therefore, we further validated binding and interaction of its constituent T-cell epitopes with their corresponding HLA proteins using molecular docking and molecular dynamics simulation experiments, followed by binding free energy calculations with molecular mechanics Poisson-Boltzmann surface area, essential dynamics analysis and free energy landscape analysis. The immuno-informatics pipeline described and the candidate epitopes discovered herein could have significant impact upon efforts to develop globally effective SARS-CoV-2 vaccines.

Review of hygiene adaptations among UK doctors in controlling the spread of SARS-CoV-2 infection.

BACKGROUND: A significant number of healthcare workers around the world have contracted COVID-19 from their workplace, thus there is a need to investigate common hygiene practices. AIM: We aimed to describe the common hygiene adaptations of doctors in the hospital and at home. METHODS: A survey of 110 doctors in UK was carried out to determine their hygiene adaptations and practices. Data were collected on demographics and personal protective equipment compliance. RESULTS: The key findings were frequent hand washing (51%), change of clothing entering and leaving hospital (88% and 92%, respectively), taking a shower upon returning home (85%) and washing work attire at temperatures of 60-80°C (50%). A higher proportion of junior doctors washed their scrubs (p=0.004) and stethoscopes (p=0.014) compared with consultants and seniors. Female doctors cleaned their mobile phones (p=0.022) and work belongings (p=0.01) more frequently. CONCLUSION: This study signified that junior doctors were more meticulous in hygiene adaptations and female doctors were more fastidious in personal hygiene. The observations of this study may be beneficial in preventing transmission of infection to families of healthcare professionals and are important to implement in the case of a second wave of COVID-19.

Assessment of a dried blood spot C-reactive protein method to identify disease flares in rheumatoid arthritis patients.

Rheumatoid arthritis (RA) is characterised by painful, stiff and swollen joints. RA features sporadic 'flares' or inflammatory episodes-mostly occurring outside clinics-where symptoms worsen and plasma C-reactive protein (CRP) becomes elevated. Poor control of inflammation results in higher rates of irreversible joint damage, increased disability, and poorer quality of life. Flares need to be accurately identified and managed. A method comparison study was designed to assess agreement between CRP values obtained by dried blood spot (DBS) versus conventional venepuncture sampling. The ability of a weekly DBS sampling and CRP test regime to detect flare outside the clinic was also assessed. Matched venepuncture and finger lancet DBS samples were collected from n = 100 RA patients with active disease at baseline and 6 weeks (NCT02809547). A subset of n = 30 RA patients submitted weekly DBS samples over the study period. Patient demographics, including self-reported flares were recorded. DBS sample CRP measurements were made by enzyme-linked immunosorbent assay, and venepuncture samples by a reference immunoturbometric assay. Data was compared between sample types by Bland-Altman and weighted Deming regression analyses. Flare detection sensitivity and specificity were compared between 'minimal' baseline and 6 week sample CRP data and the 'continuous' weekly CRP data. Baseline DBS ELISA assay CRP measures yielded a mean positive bias of 2.693 ± 8.640 (95% limits of agreement - 14.24 to 19.63%), when compared to reference assay data. Deming regression revealed good agreement between the DBS ELISA method and reference assay data, with baseline data slope of 0.978 and intercept -0.153. The specificity of 'continuous' area under the curve (AUC) CRP data (72.7%) to identify flares, was greater than 'minimal' AUC CRP data (54.5%). This study indicates reasonable agreement between DBS and the reference method, especially at low to mid-range CRP values. Importantly, longitudinal CRP measurement in RA patients helps to clearly identify flare and thus could assist in remote monitoring strategies and may facilitate timely therapeutic intervention.Trial registration: The Remote Arthritis Disease Activity MonitoR (RADAR) study was registered on 22/06/2016 at ClinicalTrials.gov Identifier: NCT02809547. https://clinicaltrials.gov/ct2/show/NCT02809547 .

Clinical and Laboratory Associations with Methotrexate Metabolism Gene Polymorphisms in Rheumatoid Arthritis.

Rheumatoid arthritis (RA) is a chronic systemic autoimmune disease that causes loss of joint function and significantly reduces quality of life. Plasma metabolite concentrations of disease-modifying anti-rheumatic drugs (DMARDs) can influence treatment efficacy and toxicity. This study explored the relationship between DMARD-metabolising gene variants and plasma metabolite levels in RA patients. DMARD metabolite concentrations were determined by tandem mass-spectrometry in plasma samples from 100 RA patients with actively flaring disease collected at two intervals. Taqman probes were used to discriminate single-nucleotide polymorphism (SNP) genotypes in cohort genomic DNA: rs246240 (ABCC1), rs1476413 (MTHFR), rs2231142 (ABCG2), rs3740065 (ABCC2), rs4149081 (SLCO1B1), rs4846051 (MTHFR), rs10280623 (ABCB1), rs16853826 (ATIC), rs17421511 (MTHFR) and rs717620 (ABCC2). Mean plasma concentrations of methotrexate (MTX) and MTX-7-OH metabolites were higher (p < 0.05) at baseline in rs4149081 GA genotype patients. Patients with rs1476413 SNP TT or CT alleles have significantly higher (p < 0.001) plasma poly-glutamate metabolites at both study time points and correspondingly elevated disease activity scores. Patients with the rs17421511 SNP AA allele reported significantly lower pain scores (p < 0.05) at both study intervals. Genotyping strategies could help prioritise treatments to RA patients most likely to gain clinical benefit whilst minimizing toxicity.

Favourable Prognosis when Lung-Cancer Patients with Superior Vena Cava Obstruction (SVCO) are Referred Promptly to EBUS-TBNA Prior to Medical or Surgical Management.

BACKGROUND: Primary care patients with superior vena cava obstruction (SVCO) syndrome are usually referred to emergency departments for urgent medical management (high-dose corticosteroids to reduce inflammation), pre-biopsy radiotherapy and/or stent placements to restore patency to the vessel. Biopsy, diagnosis and staging of the mediastinal mass is often postponed until resolution of SVCO symptoms. However, lung cancers metastasise rapidly and delays can influence the eventual outcome of patients. An additional merit in treating SVCO symptoms post-biopsy is that high-dose corticosteroids and pre-biopsy radiotherapy will degrade the quality of biopsy specimens, complicating diagnosis and subsequent management. AIMS: To determine if direct referrals of SVCO patients from primary care to the respiratory department for Endobronchial ultrasound (EBUS)-transbronchial needle-aspiration (TBNA) resulted in better outcomes. METHODS: Direct referrals to the respiratory department from primary care physicians were sought. A total of 8 patients with symptoms of SVCO were rapidly diagnosed via EBUS-TBNA and ROSE, radiotherapy and specific chemotherapy was initiated following communication with oncology colleagues. High-dose corticosteroids were administered post-EBUS. RESULTS: Rapid resolution of symptoms for SVCO were noted, without need for surgical intervention. In particular, one patient with small-cell lung cancer (the most aggressive type of lung cancer) remains well and cancer-free 14 months from diagnosis. DISCUSSION: EBUS-TBNA is a safe modality for biopsy in SVCO as there is no risk of further compression of the vessel. We need a paradigm shift in referral and a guideline of SVCO patients in primary care, an urgent biopsy is important in mediastinal cancers which have high metastatic potentials.

ATP interacts with the CPVT mutation-associated central domain of the cardiac ryanodine receptor.

BACKGROUND: This study was designed to determine whether the cardiac ryanodine receptor (RyR2) central domain, a region associated with catecholamine polymorphic ventricular tachycardia (CPVT) mutations, interacts with the RyR2 regulators, ATP and the FK506-binding protein 12.6 (FKBP12.6). METHODS: Wild-type (WT) RyR2 central domain constructs (G(2236)to G(2491)) and those containing the CPVT mutations P2328S and N2386I, were expressed as recombinant proteins. Folding and stability of the proteins were examined by circular dichroism (CD) spectroscopy and guanidine hydrochloride chemical denaturation. RESULTS: The far-UV CD spectra showed a soluble stably-folded protein with WT and mutant proteins exhibiting a similar secondary structure. Chemical denaturation analysis also confirmed a stable protein for both WT and mutant constructs with similar two-state unfolding. ATP and caffeine binding was measured by fluorescence spectroscopy. Both ATP and caffeine bound with an EC50 of ~200-400µM, and the affinity was the same for WT and mutant constructs. Sequence alignment with other ATP binding proteins indicated the RyR2 central domain contains the signature of an ATP binding pocket. Interaction of the central domain with FKBP12.6 was tested by glutaraldehyde cross-linking and no association was found. CONCLUSIONS: The RyR2 central domain, expressed as a 'correctly' folded recombinant protein, bound ATP in accord with bioinformatics evidence of conserved ATP binding sequence motifs. An interaction with FKBP12.6 was not evident. CPVT mutations did not disrupt the secondary structure nor binding to ATP. GENERAL SIGNIFICANCE: Part of the RyR2 central domain CPVT mutation cluster, can be expressed independently with retention of ATP binding.

Ryanodine receptor arrays: not just a pretty pattern?

Ryanodine receptors (RyRs) are colossal membrane protein complexes that reside in the endoplasmic reticulum of skeletal and cardiac muscle myocytes and neurons, in addition to many non-excitable cells. They comprise high-conductance ion channels that mediate the massive release of Ca2+ ions from the endoplasmic reticulum into the cytoplasm. This is the trigger for contraction during each muscle excitation-contraction coupling cycle. Individual RyRs are believed to network with other RyRs indirectly, through diffusion of released Ca2+ ions, namely the Ca2+-induced Ca2+ release phenomenon. However, RyRs can intrinsically organize into a regular array resembling a distinctive checkerboard pattern, with each square-shaped receptor appearing to abut four neighbours at each corner. In this opinion article, we describe recent data showing structural interactions between RyR oligomers in reconstituted arrays, and we suggest that this provides strong evidence for direct inter-RyR communication through a novel, allosteric regulatory mechanism.