Clinic-based real-world analysis of amyloid PET positivity and MRI markers across subjective cognitive decline, mild cognitive impairment, and early Alzheimer's disease
Heii Arai, Satoko Aiba, Hideo Yamamoto, Yuki Akiba, Junpei Takayama, Reiko AraiBackground
Alzheimer's disease (AD) is increasingly defined biologically. Amyloid positron emission tomography (PET) is a validated biomarker; however, evidence from truly consecutive, real-world memory-clinic cohorts spanning subjective cognitive decline (SCD), mild cognitive impairment (MCI), and early AD remains limited. Structural MRI markers such as the voxel-based specific regional analysis system for AD (VSRAD) are widely used, yet their incremental value for predicting amyloid positivity in real-world settings is uncertain.
Objective
To quantify amyloid PET positivity across SCD, MCI, and early AD in a real-world cohort and test whether MRI indices improve prediction beyond clinical diagnosis.
Methods
We retrospectively analyzed 194 consecutive patients undergoing 18 F-florbetapir amyloid PET (SCD n = 93; MCI n = 47; early AD n = 54). Positivity was determined by certified visual reads. Quantitation used Centiloid (CL) values retained on the original continuous scale (including negative values) and regional standardized uptake value ratios (SUVR). MRI/VSRAD indices were available in 153 patients. Logistic regression models were evaluated by cross-validated area under the curve (AUC).
Results
Amyloid PET positivity increased stepwise (SCD 40.9%, MCI 66.0%, early AD 91.7%; p < 0.0001). CL and regional SUVRs increased in parallel, most prominently in the temporal cortex and precuneus. Although VSRAD indices correlated with amyloid burden, adding MRI to the base model did not improve discrimination (AUC 0.752 versus 0.751).
Conclusions
Clinical utility of amyloid PET was confirmed in a real-world, consecutive memory-clinic cohort, supporting the relevance of amyloid assessment for routine practice and future clinical application.