Abstract: FR-PO932
Evaluation of Novel Candidate Filtration Markers from a Global Metabolomics Discovery for Glomerular Filtration Rate Estimation
Session Information
- CKD Epidemiology, Risk Factors, Prevention - II
November 03, 2023 | Location: Exhibit Hall, Pennsylvania Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 2301 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Fino, Nora F., University of Utah Health, Salt Lake City, Utah, United States
- Adingwupu, Ogechi M., Tufts Medical Center, Boston, Massachusetts, United States
- Coresh, Josef, Johns Hopkins University Department of Epidemiology, Baltimore, Maryland, United States
- Greene, Tom, University of Utah Health, Salt Lake City, Utah, United States
- Haaland, Benjamin, University of Utah Health, Salt Lake City, Utah, United States
- Shlipak, Michael, San Francisco VA Health Care System, San Francisco, California, United States
- Costa e Silva, Veronica Torres, Serviço de Nefrologia, Instituto do Câncer do Estado de São Paulo, Faculdade de Medicina, Universidade de São Paulo, Sao Paulo, Brazil
- Kalil, Roberto S., University of Maryland School of Medicine, Baltimore, Maryland, United States
- Mindikoglu, Ayse Leyla, Baylor College of Medicine, Houston, Texas, United States
- Furth, Susan L., University of Pennsylvania Perelman School of Medicine, Philadelphia, Pennsylvania, United States
- Seegmiller, Jesse C., University of Minnesota Twin Cities, Minneapolis, Minnesota, United States
- Levey, Andrew S., Tufts Medical Center, Boston, Massachusetts, United States
- Inker, Lesley Ann, Tufts Medical Center, Boston, Massachusetts, United States
Group or Team Name
- CKD-EPI.
Background
Creatinine and cystatin-C are recommended for estimating glomerular filtration rate (eGFR) but accuracy is suboptimal. Using untargeted metabolomics data, we sought to identify candidate filtration markers using a novel approach based on their maximal joint association with measured GFR (mGFR) and with flexibility to consider their biological and chemical properties.
Methods
We analyzed metabolites measured in seven diverse studies of 2,851 participants on the Metabolon H4 platform that had Pearson correlations with log mGFR . We used a stepwise approach to develop models to estimate mGFR including two to 15 metabolites with and without inclusion of creatinine and demographics. We then selected candidate filtration markers from those metabolites found >20% in models that did not demonstrate substantial overfitting in cross-validation and with small (<0.1 in absolute value) coefficients for demographics.
Results
In total, 456 named metabolites were present in all studies, and 36 had correlations < -0.5 with mGFR. We developed 2,225 models including these metabolites; all had lower root mean square errors (RMSE) and smaller coefficients for demographic variables compared to estimates using untargeted creatinine. Cross-validated RMSEs (0.187-0.213) were similar to original RMSEs for models with ≤ 10 metabolites (Figure). Our criteria identified 17 metabolites, including 12 new candidate filtration markers.
Conclusion
We identified candidate metabolites with maximal joint association with mGFR and minimal association with demographic variables across varied clinical settings. Next, we will assess the selected metabolite biological and chemical characteristics to develop a panel eGFR that is more accurate and less reliant on demographic variables than current eGFR.
Funding
- NIDDK Support