Back to 2026 Abstracts
Urinary Biomarkers of Tubular Stress and Fibrosis for Detecting Upper Tract Injury in Children with Neurogenic Bladder- A Supervised Machine Learning Risk Stratification Study
Sachit Anand, MBBS, MCh, Deepthi Ramya Ravindran, M.Sc, Ajay Verma, MBBS, MCh, Jitendra Kumar Meena, MBBS, MD, DM, Himalaya Kumar, MSc, Pankaj Hari, MD, FIAP, FAMS, Vineet Ahuja, MBBS, MD, DM.
All India Institute of Medical Sciences, New Delhi, India.
BACKGROUND: Children with neurogenic bladder (NB) are at risk for silent upper tract deterioration, while conventional surveillance tools may detect injury only after functional decline or established renal scarring. We evaluated urinary biomarkers of tubular stress and fibrosis, integrated with supervised machine-learning models, for detecting and stratifying upper tract injury in children with NB.
METHODS: This cross-sectional study included 42 children with NB and 43 age- and sex-matched healthy controls. Urinary connective tissue growth factor (CTGF), periostin, insulin-like growth factor binding protein-7 (IGFBP7), and tissue inhibitor of metalloproteinases-2 (TIMP2) were measured by ELISA and normalized to urinary creatinine. Renal function was assessed using DTPA-derived glomerular filtration rate (GFR), and renal scarring was assessed by DMSA scintigraphy. Biomarker levels were compared between groups and analyzed against GFR and scarring severity. Supervised machine-learning models, including logistic regression, random forest, support vector machine, and elastic net, were trained using functional variables, biomarker variables, and combined datasets.
RESULTS: Compared with controls, NB patients had significantly higher median urinary CTGF (4.25 vs 1.59 pg/gCr; p<0.0001), periostin (0.030 vs 0.004 ng/gCr; p<0.0001), IGFBP7 (34.43 vs 26.44 pg/gCr; p=0.048), and TIMP2 (0.18 vs 0.07 ng/gCr; p<0.0001). All biomarkers correlated inversely with GFR: CTGF (r=−0.3454; p=0.0251), periostin (r=−0.3357; p=0.0297), IGFBP7 (r=−0.3086; p=0.0467), and TIMP2 (r=−0.3400; p=0.0276). Biomarkers increased progressively with scarring severity. Jonckheere-Terpstra trend analysis was significant for CTGF, periostin, and TIMP2 (p=0.0001 each) and IGFBP7 (p=0.0146), with significance retained after false discovery rate correction. Individual biomarkers did not show independent predictive significance on logistic regression. In machine-learning analysis, combined functional-biomarker models outperformed functional-only or biomarker-only models. Random forest achieved the highest discrimination (AUC 0.889), followed by support vector machine (AUC 0.852), elastic net (AUC 0.778), and logistic regression (AUC 0.741). Net reclassification analysis demonstrated improved risk stratification with combined models, with bootstrap validation supporting significant reclassification improvement for the SVM combined model (NRI ≈0.67; 95% CI excluding zero).
CONCLUSIONS: Urinary CTGF, periostin, IGFBP7, and TIMP2 are significantly elevated in children with NB and are associated with both reduced GFR and increasing DMSA-defined renal scarring severity. Although individual biomarkers had limited standalone predictive value, their integration with functional parameters improved supervised machine-learning risk stratification. These findings support urinary tubular stress and fibrosis biomarkers as non-invasive adjuncts for identifying upper tract injury in pediatric NB, warranting longitudinal validation.
Back to 2026 Abstracts
