Risk Prediction In Long Term Kidney Transplant Recipients - Model Development Using Apelinergic Markers And Machine Learning Tools

IntroductionLimited tools exist for predicting kidney function in long-term kidney transplant recipients (KTRs). Elabela and apelin are APJ receptor agonists that constitute the apelinergic axis, which is a recently discovered system regulating vascular and cardiac tissue, in opposition to renin-angiotensin-aldosterone. MethodsLongitudinal, observational cohort of 102 KTRs who maintained graft function [≥]24 months, with no acute rejection history or current active or chronic infection. Serum apelin, elabela, fibroblast growth factor 23 (FGF-23) and -Klotho were tested using enzyme-linked immunoassay and compared with a control group of 32 healthy volunteers. ResultsMedian (IQR) follow-up time was 83 (42, 85) months. Higher serum FGF-23 and elabela, but lower Klotho concentrations were observed in KTRs. Most KTRs had stable trajectories of renal function. All candidate markers were significantly associated with mean two-year eGFR over follow-up, which itself was validated respective to death with functioning graft censored dialysis requirement. Using a cross-validation approach, we demonstrated eGFR at initial visit as the most salient predictor of future renal function. Machine learning models incorporating both clinical and biochemical (candidate markers) assessments were estimated to explain 15% of variance in future eGFR when considering eGFR-independent predictions. ConclusionsUtilization of machine learning tools that incorporate clinical information and biochemical assessments, including serum amrkers of the apelinergic axis, may help stratify risk and aid decision making in the care of long term KTRs.