Kidney Week

Abstract: TH-OR068

A Nuclear Magnetic Resonance-Based Method for Accurate Assessment of Glomerular Filtration Rate

Session Information

• Harnessing Molecular, Machine-Learning, and Genomic Innovations in Pathology
  October 25, 2018 | Location: 24A, San Diego Convention Center
  Abstract Time: 05:54 PM - 06:06 PM

Category: Pathology and Lab Medicine

• 1502 Pathology and Lab Medicine: Clinical

Authors

• Ehrich, Jochen H.h., Hannover Medical School, Hannover, Germany

Jochen H.h. Ehrich, MD

• Dubourg, Laurence, Hopital E Herriot, Lyon, France

Laurence Dubourg,

• Hansson, Sverker, Sahlgrenska University Hospital, Göteborg, Sweden

Sverker Hansson,

• Steinle, Tobias, numares AG, Regensburg, Germany

Tobias Steinle,

• Fruth, Jana, numares AG, Regensburg, Germany

Jana Fruth,

• Höckner, Sebastian, numares AG, Regensburg, Germany

Sebastian Höckner,

• Schiffer, Eric, numares AG, Regensburg, Germany

Eric Schiffer,

Background

Measuring glomerular filtration rate (mGFR) using renal or plasma clearance of an exogenous filtration marker (tracer) is the gold standard for assessing kidney function, but this procedure is time consuming and associated with a high burden for patients. Therefore, GFR is commonly estimated from serum creatinine (eGFR_creat) or serum cystatin C (eGFR_cysC). However, there exist different estimating equations for adult and paediatric patients and these equations often perform only moderately outside the cohort in which they were developed.

Methods

We used metabolomics based on nuclear magnetic resonance (NMR) spectroscopy and biostatistical modeling to identify metabolites correlated with mGFR and to combine these metabolites to biomarker networks for accurate GFR prediction. The biomarker networks were established and tested in two separate cohorts from two European centres. These cohorts comprised serum samples from both paediatric and adult patients with various nephrological conditions, covering the whole GFR range from hypo- to hyperfiltration.

Results

By combining creatinine with the uremic toxin myo-inositol, valine as indicator of metabolic acidosis, and a marker of oxidative stress into a metabolic network, we were able to generate reliable results over the whole GFR range. Compared to eGFR_creat and eGFR_cysC, the marker network showed a higher correlation with mGFR (Pearson correlation coefficient r = 0.880 vs. 0.848 and 0.636) and a 31% and 41% reduction, respectively, in overall root mean square error (RMSE 19.2 vs 28.0 and 32.8) in an independent test cohort. Especially in the “creatinine blind spot” between 60-90 ml/min/1.73m², it more than halved the variation from 24.5 to 11.0 compared to eGFR_creat. The marker network increased the percentage of estimated GFR values within 30% of mGFR (P₃₀) from 72% (66%) observed for eGFR_creat (eGFR_cysC) to 82%, which is within the performance range of plasma clearance methods, depending on the tracer used for mGFR measurement.

Conclusion

We developed a metabolomics-based serum test for accurate prediction of GFR in both adult and paediatric patients. It combines the precision of plasma clearance with the convenience of creatinine-based eGFR and obviates the need of invasive tracer application.

Funding

• Commercial Support –