BACKGROUND: Cerebrospinal fluid (CSF) biomarker studies have shown variable accuracy for diagnosis of Alzheimer's disease (AD); therefore, internal validation is recommended.

OBJECTIVE: To investigate the correlation between CSF biomarkers and cerebral 18-Florbetapir positron emission tomography (Amyloid-PET) and calculate their sensitivity and specificity to obtain the optimal clinical cut-off points to diagnose the etiology of cognitive impairment.

METHODS: We performed Amyloid-PET scans and CSF biomarker levels analyses in 68 subjects (50 with mild cognitive impairment, 11 with AD dementia, and 7 with non-AD dementia). Visual examination of Amyloid-PET scans was performed. CSF analyses were performed using standard sandwich ELISA.

RESULTS: Amyloid-PET was positive in 36 subjects, negative in 26, and inconclusive in 6. Optimal clinical cut-off points for CSF markers were the following: amyloid-β 1-42 (Aβ42) = 629 pg/ml, total tau (t-tau) = 532 pg/ml, phosphorylated tau (p-tau) = 88 pg/ml, and t-tau/Aβ42 ratio = 0.58. T-tau/Aβ42 ratio showed the best sensitivity and specificity (92 and 84%, respectively). T-tau and p-tau CSF levels (r2 = 0.867) followed by the t-tau and t-tau/Aβ42 CSF ratio (r2 = 0.666) showed the strongest inter-marker correlation. Interestingly, subjects with inconclusive Amyloid-PET showed intermediate values for all CSF markers between negative and positive Amyloid-PET groups.

CONCLUSIONS: CSF t-tau/Aβ42 ratio appears to be the most accurate AD CSF marker. The presence of intermediate values for CSF markers among the subjects with inconclusive Amyloid-PET suggests the presence of other dementias associated with AD pathology or intermediate phenotypes.