Diagnostic accuracy of CSF Ab42 and florbetapir PET for Alzheimer's disease

BACKGROUND: Reduced cerebrospinal fluid (CSF) β-amyloid42 (Aβ42) and increased florbetapir positron emission tomography (PET) uptake reflects brain Aβ accumulation. These biomarkers are correlated with each other and altered in Alzheimer's disease (AD), but no study has directly compared their diagnostic performance. METHODS: We examined healthy controls (CN, N = 169) versus AD dementia patients (N = 118), and stable (sMCI; no dementia, followed up for at least 2 years, N = 165) versus progressive MCI (pMCI; conversion to AD dementia, N = 59). All subjects had florbetapir PET (global and regional; temporal, frontal, parietal, and cingulate) and CSF Aβ42 measurements at baseline. We compared area under the curve (AUC), sensitivity, and specificity (testing a priori and optimized cutoffs). Clinical diagnosis was the reference standard. RESULTS: CSF Aβ42 and (global or regional) PET florbetapir did not differ in AUC (CN vs. AD, CSF 84.4%; global PET 86.9%; difference [95% confidence interval] −6.7 to 1.5). CSF Aβ42 and global PET florbetapir did not differ in sensitivity, but PET had greater specificity than CSF in most comparisons. Sixteen CN progressed to MCI and AD (six Aβ negative, seven Aβ positive, and three PET positive but CSF negative). INTERPRETATION: The overall diagnostic accuracies of CSF Aβ42 and PET florbetapir were similar, but PET had greater specificity. This was because some CN and sMCI subjects appear pathological using CSF but not using PET, suggesting that low CSF Aβ42 not always translates to cognitive decline or brain Aβ accumulation. Other factors, including costs and side effects, may also be considered when determining the optimal modality for different applications.