Abstract: FR-PO0512
Nuclear Magnetic Resonance (NMR) Metabolomics Improves Prediction of CKD and Downstream Complications in Health-System and Population Biobanks
Session Information
- CKM: Clinical - Trials, Epidemiology, and Biomarkers
October 23, 2026 | Location: Exhibit Hall A, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Cardiovascular-Kidney-Metabolic Health
- 602 Cardiovascular-Kidney-Metabolic Health: Clinical
Authors
- Chacon-Barahona, Jonathan A., Weill Cornell Medicine, New York, New York, United States
- Rinetti-Vargas, Gina, Kaiser Permanente Bernard J. Tyson School of Medicine, Oakland, California, United States
- Absher, Devin M., Kaiser Permanente Bernard J. Tyson School of Medicine, Oakland, California, United States
- Rana, Jamal S., Kaiser Permanente Bernard J. Tyson School of Medicine, Oakland, California, United States
- Suhre, Karsten, Department of Systems and Computational Biomedicine, Weill Cornell Medical College, New York, New York, United States
- Krumsiek, Jan, Department of Systems and Computational Biomedicine, Weill Cornell Medical College, New York, New York, United States
Background
Targeted NMR metabolomics provides a scalable snapshot of circulating biochemical state and may capture early perturbations preceding chronic kidney disease (CKD), CKD progression, and kidney-adjacent cardiometabolic disease. We tested whether NMR biomarkers improve prediction of incident CKD, downstream complications in prevalent CKD, and related cardiometabolic, liver, cardiovascular, and pulmonary endpoints.
Methods
We profiled 250 NMR biomarkers in 54,933 Kaiser Permanente Research Biobank participants linked to longitudinal EHRs. Cox models tested incremental prediction beyond clinical predictors for incident CKD and kidney-adjacent cardiometabolic and cardiopulmonary diseases. In CKD patients, we evaluated CKD progression and cardiovascular complications – assessing discrimination, calibration, net benefit, genetic risk, subgroup performance, and UK Biobank transportability.
Results
NMR biomarkers improved incident CKD prediction beyond eGFR and other clinical predictors, with NMR-based CKD models having the strongest performance. Large gains were also observed for diabetes and liver disease, consistent with shared metabolic pathways linking cardiometabolic and liver disease to kidney risk. Improvements extended to cardiopulmonary endpoints relevant to CKD multimorbidity. Among participants with prevalent CKD, NMR metabolomics improved secondary-prevention risk stratification for end-stage renal disease, stroke or transient ischemic attack, acute coronary syndrome, and heart failure beyond comprehensive clinical biomarker panels. NMR-enhanced models retained value with inherited genetic risk, whereas polygenic risk scores added minimal discrimination after clinical and metabolomic predictors were included. Benefits were preserved across demographic, anthropometric, and genetic subgroups. Models were well calibrated and improved net benefit across most endpoints. UK Biobank-trained models transported directionally to Kaiser Permanente, although local training performed better.
Conclusion
Targeted NMR metabolomics improved incident CKD prediction beyond eGFR and clinical biomarkers and improved risk stratification for major complications among participants with established CKD. These findings support NMR metabolomics as a deployable platform for calibrated kidney and kidney-adjacent risk prediction in real-world healthcare cohorts.