Predictive Molecular Biomarkers in Pediatric Steroid-Resistant Nephrotic Syndrome: A Systematic Review

Nephrotic syndrome is the leading cause of pediatric chronic kidney disease, particularly the steroid-resistant subtype (SRNS). The current diagnostic paradigm relies on prolonged steroid trials or invasive biopsies, often delaying optimal treatment. This study systematically synthesizes evidence from cohort studies on non-invasive molecular biomarkers predicting SRNS at onset. We searched PubMed, Scopus, ScienceDirect, and MEDLINE Ultimate via EBSCO for cohort studies published between January 2015 and September 2025. The study protocol was registered with PROSPERO (CRD420251147997), and quality was assessed using the Newcastle-Ottawa Scale. Eleven studies comprising 718 study-level participant observations were included; as two studies originated from an overlapping cohort within a single research group, the number of unique children is likely lower. Candidate molecular signatures were identified across plasma proteins, cytokines, metabolites, urinary markers, transcriptomic markers, and a genetic/epigenetic model. Multi-marker panels showed high predictive utility, including a plasma protein panel (VDB+ADIPOQ+MMP2; AUC 0.780; P = 0.003) and a cytokine panel (IL-7+IL-9+MCP-1; AUC 0.846; sensitivity 64.3%; specificity 84.6%). Among individual markers, TGF-β2 mRNA showed the highest predictive accuracy (AUC 0.972; P < 0.001) in a small single cohort, followed by plasma glutamine in an age-combined model (AUC 0.810; OR 1.01), while reduced plasma malonate showed predictive value in children older than 3 years (AUC 0.84; OR 0.94). Conversely, urinary VDBP failed to distinguish SRNS from SSNS (P = 0.417), suggesting possible confounding by proteinuria severity. These exploratory findings suggest that several molecular biomarkers, particularly composite panels, may hold promise for future early risk stratification; however, given the small number and size of cohorts, participant overlap, and methodological heterogeneity, these results should be interpreted as hypothesis-generating rather than evidence of superiority of any single biomarker class, requiring validation in larger, independent, prospective cohorts before clinical application.