Comparison of the analytical and clinical performance of three immunoassay platforms for plasma glial fibrillary acidic protein in Alzheimer's disease.

OBJECTIVES

This study aimed to compare the analytical and clinical performance of plasma glial fibrillary acidic protein (GFAP) across three immunoassay platforms.

METHODS

Plasma GFAP was measured on three immunoassay platforms (Simoa HD-X, Maccura i1000, MS-Fast Pro 160) in 302 participants from the Peking Union Medical College Hospital dementia cohort (139 Alzheimer's disease dementia [ADD], 116 non-AD dementia [NADD]). Inter-platform agreement was assessed using Passing-Bablok regression, Bland-Altman analysis, and Spearman correlation. ROC analyses and multimarker models on the Simoa platform were used to evaluate GFAP, NfL, a core plasma panel (Aβ1-42, p-tau181, p-tau217), and their combinations.

RESULTS

Plasma GFAP levels were significantly higher in ADD than in NADD across all three platforms. Inter-platform correlations were strong (Spearman's r=0.874-0.932), but Passing-Bablok regression showed substantial proportional bias and systematically higher concentrations on the Simoa platform. ROC-based discrimination between ADD and NADD was comparable across platforms (AUC 0.732-0.740), whereas assay-specific optimal cut-offs differed markedly. On the Simoa platform, GFAP alone achieved an AUC of 0.731. The core plasma panel (Aβ1-42, p-tau181, p-tau217) achieved an AUC of 0.898, which increased to 0.924 after adding GFAP and NfL.

CONCLUSIONS

This study provides a comparison of plasma GFAP measurements across three immunoassay platforms, revealing strong correlations but substantial differences in absolute values and decision cut-offs. The clinical analyses show that similar discriminative performance can coexist with markedly different platform-specific cut-offs, underscoring the need for platform-specific cut-offs and further harmonization of GFAP measurements.