Integrated HPLC–MS and NMR Data Fusion Improves Diagnostic Precision in Pediatric Chronic Kidney Disease

Introduction: The diagnostic strategies needed for chronic kidney disease and congenital uropathies in children should be minimally invasive, clinically feasible and detect early metabolic changes. This pilot study evaluated whether integration of targeted HILIC–MS/MS and ¹H NMR urine profiling improves the discrimination of children with renal impairment associated with congenital uropathies from controls.

Methods: A single-point prospective diagnostic pilot study included 47 pediatric urine samples collected at initial diagnosis: 22 children with vesicoureteral reflux grades II–V and 25 controls with no urinary tract disease. The mean age of the cohort was 2.5 ± 2.2 years. For HILIC–MS/MS analysis, urine samples were analysed using a Shimadzu LC-20AD system coupled to a Sciex 4500 QTRAP mass spectrometer in positive electrospray Scheduled MRM mode. For ¹H NMR analysis, urine samples were acquired on a 500 MHz Bruker AV III spectrometer using a 1D NOESY sequence. Metabolite concentrations were normalized to creatinine. Missing values did not exceed 5% and were imputed using Bayesian PCA. PCA, discriminant analysis, PLS-DA, leave-one-out cross-validation, and high-level data fusion were applied.

Results: Low sensitivity and high specificity for group separation obtained from the HILIC–MS/MS model, 57% and 91%, respectively. The greatest contribution was from the following variables: tryptophan, phenylalanine, and trimethylamine N-oxide. The ¹H NMR data additionally showed partial separation of the groups; the metabolites most responsible for this separation included anserine and homovanillate. Data fusion of HILIC–MS/MS and ¹H NMR outputs resulted in enhancement of the classification balance for 80% sensitivity and 92% specificity.

Conclusions: In this pediatric pilot cohort, urine metabolomics using combined HILIC–MS/MS and ¹H NMR was more informative for the diagnostic classification than single-platform analysis. The novelty of the work is the use of multimodal, creatinine-normalized urine metabolite profiling and data fusion in congenital-uropathy-associated renal impairment in children, resulting in an enhancement of sensitivity without loss of the high specificity.