Comparison of cerebrospinal fluid levels of tau and Aβ 1-42 in Alzheimer disease and frontotemporal degeneration using 2 analytical platforms

Abstract

Objective: To use values of cerebrospinal fluid tau and β-amyloid obtained from 2 different analytical immunoassays to differentiate Alzheimer disease (AD) from frontotemporal lobar degeneration (FTLD).

Design: Cerebrospinal fluid values of total tau (T-tau) and β-amyloid 1-42 (Aβ 1-42) obtained using the Innotest enzyme-linked immunosorbent assay were transformed using a linear regression model to equivalent values obtained using the INNO-BIA AlzBio3 (xMAP; Luminex) assay. Cutoff values obtained from the xMAP assay were developed in a series of autopsy-confirmed cases and cross validated in another series of autopsy-confirmed samples using transformed enzyme-linked immunosorbent assay values to assess sensitivity and specificity for differentiating AD from FTLD.

Setting: Tertiary memory disorder clinics and neuropathologic and biomarker core centers.

Participants: Seventy-five samples from patients with cerebrospinal fluid data obtained from both assays were used for transformation of enzyme-linked immunosorbent assay values. Forty autopsy-confirmed cases (30 with AD and 10 with FTLD) were used to establish diagnostic cutoff values and then cross validated in a second sample set of 21 autopsy-confirmed cases (11 with AD and 10 with FTLD) with transformed enzyme-linked immunosorbent assay values.

Main outcome measure: Diagnostic accuracy using transformed biomarker values.

Results: Data obtained from both assays were highly correlated. The T-tau to Aβ 1-42 ratio had the highest correlation between measures (r = 0.928, P < .001) and high reliability of transformation (intraclass correlation coefficient= 0.89). A cutoff of 0.34 for the T-tau to Aβ 1-42 ratio had 90% and 100% sensitivity and 96.7% and 91% specificity to differentiate FTLD cases in the validation and cross-validation samples, respectively.

Conclusions: Values from 2 analytical platforms can be transformed into equivalent units, which can distinguish AD from FTLD more accurately than the clinical diagnosis.

Figure 1

Flow chart of the (A) transformation and (B) validation and cross-validation steps.

Figure 2

Transformation of CSF analytes into equivalent values between platforms. Shown are plots of raw and natural-log transformed values of (A) Aß_1-42, (B) t-tau, (C) p-tau₁₈₁, (D) t-tau:Aß_1-42 ratio, and (E) p-tau₁₈₁:Aß_1-42 ratio obtained with ELISA and xMAP. T-ELISA=transformed data.

Figure 3

Receiver operating characteristic curve analysis of xMAP analyte values in an autopsy-confirmed sample (neuropathological sample 1). The t-tau:Aß_1-42 ratio had the highest area under the curve at the optimal diagnostic cut-point of 0.34.

Figure 4

Diagnostic accuracy of t-tau:Aß_1-42 ratio cutoff in the validation and cross-validation data sets. Shown is a (A) box-plot of AD and FTLD values from both samples. (B) The t-tau:Aß_1-42 ratio was highly sensitive and specific for identifying AD and FTLD, with improved diagnostic accuracy compared with ante mortem clinical diagnosis.