Cross-Platform Analytical Assessment of Serum GFAP Quantification in Multiple Sclerosis: SIMOA Versus Two Automated Immunoassays

Jordi Tortosa Carreres^1,2Laura Cubas Núñez^2*Jéssica Castillo Villalba²Lorena Forés Toribio²Raquel Gasque Rubio²Carlos Quintanilla Bordás^2,3Carmen Alcalá Vicente^2,4Sara Carratalá Boscá²Ana Vaño Bellver¹Begoña Laiz Marro¹Francisco Carlos Pérez-Miralles^2,5Bonaventura Casanova Estruch^2,5

• ¹Laboratory Department, La Fe University and Polytechnic Hospital,, Valencia, Spain
• ²Neuroimmunology Research Group, La Fe Health Research Institute, Valencia, Spain
• ³Neurology Department, General University Hospital of Castellón, Castellón, Spain
• ⁴Neurology Department, University Hospital of La Ribera, Alzira, Spain
• ⁵Neuroimmunology Unit, La Fe University and Polytechnic Hospital, Valencia, Spain

Introduction: Serum glial fibrillary acidic protein (sGFAP) is a promising biomarker, but its quantification mainly relies on SIMOA, a technology not widely available in clinical practice. Objectives: To evaluate the analytical performance of two high-throughput automated platforms—Alinity® i (Abbott) and Lumipulse® G1200 (Fujirebio)—for sGFAP quantification. Methods: A retrospective longitudinal study included 107 serum samples from 23 MS patients. sGFAP was measured with SIMOA SR-X®, Lumipulse® G1200, and Alinity® i. Data were log-transformed. Agreement was assessed using Pearson correlations, Passing–Bablok regression, Bland–Altman analysis, and Δlog correlations between visits. Longitudinal differences across platforms were tested with a linear mixed-effects model (platform as fixed effect, SIMOA as reference). Moreover, ΔSIMOA was modeled against ΔLumipulse and ΔAlinity, adjusting for ΔEDSS, phenotype, relapses and new MRI lesions. Results: Passing–Bablok regression yielded slopes of 0.85 (SIMOA–Lumipulse), 0.81 (SIMOA–Alinity), and 0.95 (Lumipulse–Alinity), with intercepts of –0.32, –0.35, and – 0.05. Mean log-biases were –0.622, –0.733, and 0.109. Correlations between log-means and log-differences were r = 0.26 (P=0.006), 0.44 (P< 0.0001), and 0.15 (P=0.13). The mixed-effects model showed no significant Δlog differences relative to SIMOA (P > 0.1). When modeling ΔSIMOA, ΔLumipulse was a significant predictor (β=0.51; P=0.002), whereas ΔAlinity showed only a trend (β=0.31; P=0.051). No clinical covariates were significantly associated.