Speaking the same language: international cross-validation of emerging biomarkers for juvenile idiopathic arthritis and Still’s Disease

BACKGROUND: Protein biomarkers such as interleukin 18 (IL-18), CXCL9, and the S100A alarmin proteins are increasingly used in the diagnosis and treatment of juvenile idiopathic arthritis (JIA) and Still's Disease (SD). Reported values for these biomarkers vary considerably among different testing platforms at different centers, representing a barrier to their clinical application as well as international research collaborations. We undertook a systematic evaluation of measurement comparability across different platforms in Europe (EU) and North America (NoA). METHODS: Recombinant proteins including IL-18, CXCL9, S100A8/9, and S100A12 spiked into donor human serum were distributed in blinded fashion to participating centers in NoA and EU for determination of sample concentration on each center's measurement platform. Assay-specific mathematical correction formulas were calculated based on spike recovery data. Next, samples from the Childhood Arthritis and Rheumatology Research Alliance (CARRA) First-line Options for Systemic JIA Treatment (FROST) study were utilized for validation on selected platforms. Comparisons between measurement platforms before and after mathematical correction were analyzed. RESULTS: In the spiked samples, while overall strong correlation was observed for IL-18 measurements across different platforms, the percent recovery revealed significant analytical variation ranging between approximately 50-400%. Similar variation in recovery was also observed for CXCL9, S100A8/9, and S100A12 across different measurement platforms. Analysis of real-world samples from the FROST study revealed strong correlation in IL-18 and CXCL9 values when comparing the same bead-based platforms at two different centers. Significant differences and systematic bias in measurement of FROST samples were uncovered when comparing ELISA or Ella platforms to commercial Luminex platforms. The application of regression-based mathematical correction factors generated from the spike recovery assays could only partly resolve these inter-platform differences. CONCLUSIONS: This is the first systematic analysis of the comparability of biomarker measurements used in JIA and SD across different platforms including ELISA, Luminex, and Ella. We demonstrate substantial differences in the quantification of standardized biomarker concentrations both between assay types and across testing sites. In contrast, analyses of real-world samples from sJIA patients showed high concordance between two independent Ella platforms, indicating that this platform could improve clinical and research standardization.