Search

Data sharing among different institutions represents one of the major challenges in developing distributed machine learning approaches, especially when data is sensitive, such as in medical applications. Federated learning is a possible solution, but requires fast communications and flawless security. Here, we propose SYNDSURV (SYNthetic Distributed SURVival), an alternative approach that simplifies the current state-of-the-art paradigm by allowing different centres to generate local simulated instances from real data and then gather them into a centralised hub, where an Artificial Intelligence (AI) model can learn in a standard way. The main advantage of this procedure is that it is model-agnostic, therefore prediction models can be directly applied in distributed applications without requiring particular adaptations as the current federated approaches do. To show the validity of our approach for medical applications, we tested it on a survival analysis task, offering a viable alternative to train AI models on distributed data. While federated learning has been mainly optimised for gradient-based approaches so far, our framework works with any predictive method, proving to be a comparable way of performing distributed learning without being too demanding towards each participating institute in terms of infrastructural requirements., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 The Authors. Published by Elsevier Ltd.. All rights reserved.)

Background & Aims: Nephrotoxicity of intravenous iodinated contrast media (ICM) in cirrhosis is still a debated issue, due to scarce, low-quality and conflicting evidence. This study aims to evaluate the incidence and predisposing factors of acute kidney injury (AKI) in patients with cirrhosis undergoing contrast-enhanced computed tomography (CECT)., Methods: We performed a prospective, multicenter, cohort study including 444 inpatients, 148 with cirrhosis (cohort 1) and 163 without cirrhosis (cohort 3) undergoing CECT and 133 with cirrhosis (cohort 2) unexposed to ICM. Kidney function parameters were assessed at T0, 48-72 h (T1), 5 and 7 days after CECT/enrollment. Urinary neutrophil gelatinase-associated lipocalin (U-NGAL) was measured in 50 consecutive patients from cohort 1 and 50 from cohort 2 as an early biomarker of tubular damage., Results: AKI incidence was not significantly increased in patients with cirrhosis undergoing CECT (4.8%, 1.5%, 2.5% in cohorts 1, 2, 3 respectively, p = n.s.). Most AKI cases were mild and transient. The presence of concomitant infections was the only independent predictive factor of contrast-induced AKI (odds ratio 22.18; 95% CI 2.87-171.22; p = 0.003). No significant modifications of U-NGAL between T0 and T1 were detected, neither in cohort 1 nor in cohort 2 (median ΔU-NGAL: +0.2 [-7.6 to +5.5] ng/ml, +0.0 [-6.8 to +9.5] ng/ml, respectively [p = 0.682])., Conclusions: AKI risk after CECT in cirrhosis is low and not significantly different from that of the general population or of the cirrhotic population unexposed to ICM. It mostly consists of mild and rapidly resolving episodes of renal dysfunction and it is not associated with tubular kidney injury. Patients with ongoing infections appear to be the only ones at higher risk of AKI., Impact and Implications: Nephrotoxicity due to intravenous iodinated contrast media (ICM) in patients with cirrhosis is still a debated issue, as the available evidence is limited and based on very heterogeneous studies, often conducted on small and retrospective cohorts. In this prospective three-cohort study we found that intravenous administration of ICM was associated with a low risk of AKI, similar to that of the general population and to that of patients with cirrhosis unexposed to ICM. Patients with ongoing infections were the only ones to have a significantly increased risk of contrast-induced AKI. Therefore, the actual recommendations of performing contrast imaging studies cautiously in cirrhosis do not seem to be reasonable anymore, with the exception of infected patients, who have a significantly higher risk of contrast-induced AKI., (Copyright © 2023 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved.)

Federated learning initiatives in healthcare are being developed to collaboratively train predictive models without the need to centralize sensitive personal data. GenoMed4All is one such project, with the goal of connecting European clinical and -omics data repositories on rare diseases through a federated learning platform. Currently, the consortium faces the challenge of a lack of well-established international datasets and interoperability standards for federated learning applications on rare diseases. This paper presents our practical approach to select and implement a Common Data Model (CDM) suitable for the federated training of predictive models applied to the medical domain, during the initial design phase of our federated learning platform. We describe our selection process, composed of identifying the consortium's needs, reviewing our functional and technical architecture specifications, and extracting a list of business requirements. We review the state of the art and evaluate three widely-used approaches (FHIR, OMOP and Phenopackets) based on a checklist of requirements and specifications. We discuss the pros and cons of each approach considering the use cases specific to our consortium as well as the generic issues of implementing a European federated learning healthcare platform. A list of lessons learned from the experience in our consortium is discussed, from the importance of establishing the proper communication channels for all stakeholders to technical aspects related to -omics data. For federated learning projects focused on secondary use of health data for predictive modeling, encompassing multiple data modalities, a phase of data model convergence is sorely needed to gather different data representations developed in the context of medical research, interoperability of clinical care software, imaging, and -omics analysis into a coherent, unified data model. Our work identifies this need and presents our experience and a list of actionable lessons learned for future work in this direction., Competing Interests: Declaration of Competing Interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: All authors reports financial support was provided by European Union., (Copyright © 2023 Elsevier Inc. All rights reserved.)

Background & Aims: Non-invasive scoring systems (NSS) are used to identify patients with non-alcoholic fatty liver disease (NAFLD) who are at risk of advanced fibrosis, but their reliability in predicting long-term outcomes for hepatic/extrahepatic complications or death and their concordance in cross-sectional and longitudinal risk stratification remain uncertain., Methods: The most common NSS (NFS, FIB-4, BARD, APRI) and the Hepamet fibrosis score (HFS) were assessed in 1,173 European patients with NAFLD from tertiary centres. Performance for fibrosis risk stratification and for the prediction of long-term hepatic/extrahepatic events, hepatocarcinoma (HCC) and overall mortality were evaluated in terms of AUC and Harrell's c-index. For longitudinal data, NSS-based Cox proportional hazard models were trained on the whole cohort with repeated 5-fold cross-validation, sampling for testing from the 607 patients with all NSS available., Results: Cross-sectional analysis revealed HFS as the best performer for the identification of significant (F0-1 vs. F2-4, AUC = 0.758) and advanced (F0-2 vs. F3-4, AUC = 0.805) fibrosis, while NFS and FIB-4 showed the best performance for detecting histological cirrhosis (range AUCs 0.85-0.88). Considering longitudinal data (follow-up between 62 and 110 months), NFS and FIB-4 were the best at predicting liver-related events (c-indices>0.7), NFS for HCC (c-index = 0.9 on average), and FIB-4 and HFS for overall mortality (c-indices >0.8). All NSS showed limited performance (c-indices <0.7) for extrahepatic events., Conclusions: Overall, NFS, HFS and FIB-4 outperformed APRI and BARD for both cross-sectional identification of fibrosis and prediction of long-term outcomes, confirming that they are useful tools for the clinical management of patients with NAFLD at increased risk of fibrosis and liver-related complications or death., Lay Summary: Non-invasive scoring systems are increasingly being used in patients with non-alcoholic fatty liver disease to identify those at risk of advanced fibrosis and hence clinical complications. Herein, we compared various non-invasive scoring systems and identified those that were best at identifying risk, as well as those that were best for the prediction of long-term outcomes, such as liver-related events, liver cancer and death., Competing Interests: Conflict of interest The authors declare no conflicts of interest that pertain to this work. Please refer to the accompanying ICMJE disclosure forms for further details., (Copyright © 2021 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved.)