Abstract
BACKGROUND: The AD Resemblance Atrophy Index (AD‐RAI) has demonstrated superior performance compared to other neurodegeneration biomarkers in diagnosis and prognosis of Alzheimer's Disease (AD). Based on the 2024 NIA‐AA criteria, in addition to Amyloid (A), Tau (T), and Neurodegeneration (N, NfL, atrophy) biomarkers, Inflammation (I, GFAP) and Vascular (V, white matter hyperintensity [WMH]) biomarkers are recognized as co‐pathology biomarkers contributing to AD progression. These biomarkers have previously been reported to be associated with cognitive decline, however, there is currently a lack of studies exploring the performance of a combination of biomarkers in predicting early cognitive decline. METHOD: We recruited 314 non‐demented participants from a Singapore memory clinic who had either no cognitive impairment or mild cognitive impairment and were followed up for 2 years. Plasma biomarkers (GFAP, NfL, p‐tau181, Aβ42) and MRI measures (AD‐RAI, hippocampal volume, WMH) were collected at baseline. Cognitive decline was defined as a one‐point increase in the CDR‐SB score within 2 years. RESULT: Of the 314 participants, 77 (24.5%) showed cognitive decline within 2 years. Baseline CDR‐SB, AD‐RAI, plasma GFAP, APOE4 carrier status, and hippocampal volume were independently associated with cognitive decline after adjustments for age, gender, and education. Among the individual biomarkers analyzed using ROC, plasma GFAP demonstrated the best performance in predicting cognitive decline with an AUC of 0.73 (sensitivity: 85.7%; specificity: 51.9%). AD‐RAI out‐performed plasma NfL and all chronic brain change biomarkers, achieving an AUC of 0.72 (sensitivity: 74.0%; specificity: 63.7%). A parsimonious model, which included demographic factors, CDR‐SB, plasma biomarkers (GFAP, Aβ42), AD‐RAI, and WMH, achieved an optimal balance between model complexity and performance (AUC: 0.83; sensitivity: 85.7%; specificity: 66.2%). CONCLUSION: Incorporating demographic factors, CDR‐SB, plasma biomarkers (GFAP, Aβ42), AD‐RAI, and WMH demonstrated good predictive performance for cognitive decline. Combining co‐pathology biomarkers with AD‐RAI and plasma based Aβ biomarkers enhanced model performance, highlighting the potential of multi‐biomarker approaches in predicting cognitive decline. Future studies with larger, multi‐ethnic cohorts are needed to validate these findings and assess their generalizability