Our Leadership Team

DISCUSSIONIn addition to randomized studies, prior large observational studies have indicated a survival benefit for hemodiafiltration, as well as a possible reduction of hospitalizations. The target trial emulation study outlined in this protocol will expand this knowledge and provide generalizable insights on the effects of hemodiafiltration on outcomes by using real-world data representative of routine clinical practice while appropriately addressing sources of bias.BACKGROUNDRandomized studies have demonstrated that high-volume hemodiafiltration results in reduced mortality compared to conventional hemodialysis treatment. However, eligibility criteria in these trials may limit generalizability to routine clinical practice. Some of these trials reported a limited number of events, underscoring the need to further evaluate the effect of hemodiafiltration on mortality. We will conduct a target trial emulation study using data from routine clinical practice. The primary aim of this study is to evaluate whether high-volume hemodiafiltration reduces all-cause mortality. The secondary aim is to assess cause-specific mortality. Other aims include assessing all-cause and cause-specific hospitalizations, as well as cumulative length of hospital stay and the dose-response relationship between convection volume in hemodiafiltration and the outcomes.METHODSData will be obtained from the second version of ApolloDialDb (Apollo), an anonymized dialysis dataset capturing over 1000 variables from patients from all over the world. For this study, we will include adult patients from European countries with kidney failure who initiated with at least one treatment of high-flux hemodialysis or hemodiafiltration between 01 January 2018 and 30 June 2024, and who were prescribed a thrice-weekly dialysis schedule at the start. Patients starting with home dialysis will be excluded. We will use a target trial emulation approach with a clone-censor-weight design and marginal structural models, controlling for selection bias, survivor bias, and competing risk bias. Sub-analyses will be performed to investigate the effect of high-volume hemodiafiltration (≥ 23 L of convection volume). Inverse probability weighting will be applied to adjust for predefined confounders including sociodemographic, clinical, and anthropometric factors, as well as comorbidities to achieve balance between treatment groups.

PURPOSE OF REVIEWArtificial intelligence (AI) and machine learning (ML) are rapidly transforming healthcare, but their adoption in nephrology and dialysis remains relatively limited.SUMMARYAI in nephrology shows promise for personalized care and cost reduction, as demonstrated by tools like the Anemia Control Model. Yet, broad adoption requires rigorous validation, seamless workflow integration, regulatory clearance, and clinician trust. Future opportunities include digital twins, large language models, and multiomics integration, with AI poised to enhance both patient outcomes and system performance.RECENT FINDINGSThis review highlights key applications of AI in kidney disease, including prognostic modeling, imaging, personalized anemia and fluid management, patient engagement, and research acceleration. While numerous studies demonstrate improved prediction accuracy and clinical insights, translation into routine practice is rare. Examples such as the Anemia Control Model (ACM) demonstrate that AI can simultaneously improve clinical outcomes and reduce costs, though widespread adoption will require rigorous validation, seamless integration into clinical workflows, regulatory approval, and above all, clinician trust.