Accurate risk prediction of adverse kidney-related outcomes in individuals with chronic kidney disease (CKD) is essential to guide personalized treatment. Plasma and urine metabolites may individually or jointly improve prediction beyond clinically established prognostic factors. We analyzed prospective German CKD (GCKD) cohort data from 5,217 individuals with predominantly CKD stage G3 at baseline and 6.5-year follow-up data. Baseline metabolite levels were measured via untargeted mass spectrometry in plasma (N=5,144; 1,096 metabolites) and urine (N=5,088; 1,129 metabolites). Outcomes were kidney failure and a composite kidney endpoint including kidney failure, a 40 percent estimated glomerular filtration rate decline, or estimated glomerular filtration rate <15 mL/min/1.73 m2. Time-to-event analyses used subdistribution hazard models with component-wise boosting for metabolite selection, and predictive performance was compared with benchmark models including established prognostic factors. Several individual metabolites improved kidney failure risk prediction beyond established prognostic factors. Multi-metabolite models for kidney failure showed good performance that declined for more distant time points, while overall improvements over clinical models were modest and only partly significant. Results for the composite kidney endpoint were similar. While certain metabolites improved prediction of adverse kidney-related outcomes, the added value was limited, although prognostic metabolites may reflect relevant CKD-related metabolic pathways.
Institute of Medical Biometry and Statistics,
Faculty of Medicine and Medical Center –
University of Freiburg
