Evaluation of the Risk of Heart Failure in Patients With Chronic Kidney Disease and Type 2 Diabetes Mellitus

Overview

The goal of this study, performed entirely with computer programs, is to learn if using a combination of two drugs works better than using one alone, in the treatment of people with chronic kidney disease and type 2 diabetes mellitus. The main questions it aims to answer are: * How many events of heart failure, heart failure hospitalizations, and cardiovascular death happen in each group? * How do serious kidney problems progress in each of the groups? Researchers will compare reference therapy (finerenone) with a combination of finerenone and empagliflozin. There will not be human participants in this study. With the help of artificial intelligence, researchers will recreate a large population of people with chronic kidney disease and type 2 diabetes mellitus. Data on their disease, treatment, and laboratory parameters will be based on and simulated using real data from previous clinical trials conducted with real participants.

Traditional risk stratification tools, such as the PREVENT (Predicting Risk of cardiovascular disease EVENTs) equations, estimate 10-year risk using static baseline variables. Simply adjusting the UACR input in the PREVENT calculator offers only a crude approximation and does not capture the dynamic, nonlinear effects of ongoing pharmacological treatment, biological hysteresis, and competing risks. To address these limitations, we will develop a high-fidelity digital twin simulation framework powered by the Aeterna Deep computational engine. This in silico trial will project 10-year cardio-renal outcomes for finerenone, empagliflozin, and their combination, while quantifying the mechanistic contribution of UACR reduction relative to other cardio-renal factors. The reference population will be defined according to the CONFIDENCE trial criteria, including adults with chronic kidney disease (eGFR 30-90 ml/min/1.73 m²), type 2 diabetes, and persistent albuminuria (UACR 100-5000 mg/g), receiving maximally tolerated renin-angiotensin system inhibitors. From this base, a synthetic cohort of 100,000 digital twins will be constructed using a four phase instantiation framework designed to ensure biological plausibility. First, marginal distributions for key demographic and clinical variables (age, sex, eGFR, UACR, blood pressure, HbA1c) will be parameterized from validated real-world datasets. Second, nonlinear physiological interdependencies will be reconstructed through multivariate coupling functions to preserve the in vivo covariance structure. Third, individual phenotypes will be generated via high-dimensional stochastic sampling to capture the full range of cardio renal metabolic trajectories. Finally, all instantiated profiles will undergo thermodynamic and biophysical truncation, whereby parameter combinations violating conservation principles or exceeding human homeostatic limits will be excluded and regenerated, ensuring that only physiologically viable digital entities will be retained for downstream analyses.