B-248 From Static to Dynamic: Establishing Next-Generation Reference Intervals for Cerebrospinal Fluid Biomarkers Using a SOM+refineR Framework

Abstract

Background Cerebrospinal fluid (CSF) biomarkers, including chloride (Cl), glucose (Glu), and total protein (Pro), play a vital role in diagnosing and monitoring central nervous system (CNS) disorders. However, ethical and practical challenges limit the direct establishment of reference intervals (RIs) from healthy populations. Methods Using real-world data from the Laboratory Information System of Peking Union Medical College Hospital, this study applied a combined self-organizing map (SOM) and refineR framework to identify presumptively healthy individuals. Age-overlapping subgrouping and penalized spline modeling were used to construct continuous, age-dependent next-generation RIs (NGRIs) for CSF Cl, Glu, and Pro. Sex- and age-related variations were assessed by standard deviation ratio (SDR) and visual analyses. Results The SOM?+?refineR approach effectively reduced abnormal data influence and generated robust, population-specific RIs. CSF-Cl remained stable across age and sex. CSF-Glu exhibited a mild, age-related increase consistent with systemic glucose trends. CSF-Pro concentrations were higher in males than in females and showed a progressive rise with aging, likely reflecting physiological changes in blood–brain barrier permeability and protein metabolism. Conclusion This study provides the first comprehensive, data-driven NGRIs for major CSF analytes using an integrated SOM + refineR approach. The resulting age- and sex-specific continuous RIs offer improved precision and clinical applicability, supporting individualized laboratory interpretation in neurological diagnostics.