Background: Longitudinal studies including both structural and amyloid neuroimaging as well as cognitive,
genetic and biochemical biomarkers are allowing a better understanding of the role of brain Aβ deposition in
cognitive decline. The purpose of this study was to compare the accuracy of different biomarkers to predict
cognitive decline and conversion to Alzheimer's disease (AD).
Methods: Follow-up was obtained 20±3 months after biochemical (plasma Aβ42/Aβ40), genetic (ApoE), cognitive
(memory and non-memory scores) and neuroimaging (3D MRI, FDG and PiB-PET) evaluation in 57 subjects with
mild cognitive impairment (MCI) (79% amnestic) and 97 age-matched healthy controls (HC) (73±7 years of age).
Results: At follow-up, progression to AD occurred in 47% of MCI, while 4% were re-classified as HC. Comparison
of converters to non-converters showed a significant difference in episodic memory scores, prevalence of
ApoE-ε4 allele, PiB retention, hippocampal volume (HV) and posterior cingulate glucose metabolism. There were
no differences in plasma Aβ42/Aβ40 and non-memory scores. The accuracy in predicting conversion from MCI to
AD based on cut-off values was 81% for PiB, 78% for memory, 77% for ApoE-ε4, 72% for HV, 68% for FDG, and
61% for plasma Aβ. Combining PiB and memory, the predictive accuracy increased to 87%. Of the high PiB HC,
14% developed MCI or AD by 20 months and at least 21% by 3 years. One (2%) low PiB HC developed MCI. Of
PiB, HV, FDG, age, memory score and non-memory cognitive score, only PiB SUVR (p<0.001) and memory score
(p=0.003) survived as significant predictors of decline in MMSE in the combined MCI and HC cohort using stepwise
regression.
Conclusions: PiB binding and cognitive measures were the best predictors of cognitive decline and conversion
to AD over 20 months. These findings may have application in early diagnosis of AD and subject selection for
therapeutic trials.
