Variability of CSF Alzheimer's disease biomarkers: implications for clinical practice

Abstract

Background: Cerebrospinal fluid (CSF) biomarkers are increasingly being used for diagnosis of Alzheimer's disease (AD).

Objective: We investigated the influence of CSF intralaboratory and interlaboratory variability on diagnostic CSF-based AD classification of subjects and identified causes of this variation.

Methods: We measured CSF amyloid-β (Aβ) 1-42, total tau (t-tau), and phosphorylated tau (p-tau) by INNOTEST enzyme-linked-immunosorbent assays (ELISA) in a memory clinic population (n = 126). Samples were measured twice in a single or two laboratories that served as reference labs for CSF analyses in the Netherlands. Predefined cut-offs were used to classify CSF biomarkers as normal or abnormal/AD pattern.

Results: CSF intralaboratory variability was higher for Aβ1-42 than for t-tau and p-tau. Reanalysis led to a change in biomarker classification (normal vs. abnormal) of 26% of the subjects based on Aβ1-42, 10% based on t-tau, and 29% based on p-tau. The changes in absolute biomarker concentrations were paralleled by a similar change in levels of internal control samples between different assay lots. CSF interlaboratory variability was higher for p-tau than for Aβ1-42 and t-tau, and reanalysis led to a change in biomarker classification of 12% of the subjects based on Aβ1-42, 1% based on t-tau, and 22% based on p-tau.

Conclusions: Intralaboratory and interlaboratory CSF variability frequently led to change in diagnostic CSF-based AD classification for Aβ1-42 and p-tau. Lot-to-lot variation was a major cause of intralaboratory variability. This will have implications for the use of these biomarkers in clinical practice.

Figure 1. CSF levels by analysis and marker.

Results are frequencies and mean (SD) for each CSF marker on the left (A) for CSF intralaboratory analyses and on the right (B) for CSF interlaboratory analyses. The solid line represents the mean CSF levels. Analysis 1 is routine practice and analysis 2 is performed as part of the LeARN study. CSF = cerebrospinal fluid, Aβ = amyloid beta, t-tau = total tau, p-tau = phosphorylated tau. **P<0.001, *p<0.05 compared to CSF analysis 1 or lab 1.

Figure 2. Bland-Altman plots of variability between CSF analyses.

The average of CSF analysis 1 and 2 is plotted against the difference between both analyses, on the left (A) for CSF intralaboratory analyses and on the right (B) for CSF interlaboratory analyses. The solid line represents the mean and the dotted lines the upper and lower 1.95 SD. CSF = cerebrospinal fluid, Aβ = amyloid beta, t-tau = total tau, p-tau = phosphorylated tau.

Figure 3. Internal control data of the Amsterdam laboratory.

Results are changes in high and low CSF biomarker levels for intralaboratory reanalysis in two lots (routine lot and LeARN lot) as part of internal control of data at the Amsterdam laboratory. A) Aβ1-42: Lot 1 used from February 2010 to February 2011 (n = 18) and lot 2 used from March to October 2012 (n = 17). B) T-tau: Lot 1 used from February 2010 to April 2011 (n = 24) and lot 2 used from March to June 2012 (n = 11). C) P-tau: Lot 1 used from February 2010 to April 2011 (n = 23) and lot 2 used from March to October 2012 (n = 11). CSF = cerebrospinal fluid, Aβ = amyloid beta, t-tau = total tau, p-tau = phosphorylated tau. *P<0.001 for differences between lot of analysis 1 (routine lot) and lot of analysis 2 (LeARN lot).

Figure 4. Intralaboratory lot-to-lot variation.

Results in this graph are based on deming regression and show the CSF levels of lot 1 and lot 3 for Aβ1-42 (A), t-tau (B), and p-tau (C). The slope is different for Aβ1-42 levels between lot 1 and lot 3. The mean difference in CSF levels between lot 1 and 3 was significantly different for Aβ1-42 and t-tau. Lot 1 = clinical routine lot, lot 3 = LeARN lot, Aβ = amyloid beta, t-tau = total tau, p-tau = phosphorylated tau.

Figure 5. Change in CSF marker classification according to cut-offs.

Results are CSF levels only for subjects in which reanalysis let to a different biomarker classification using the cut-offs from Amsterdam to define abnormal CSF values. On the left (A) are changes in biomarker levels for intralaboratory reanalysis and on the right (B) for interlaboratory reanalysis. The arrow represents the applied Amsterdam CSF cut-off. Analysis 1 is routine practice and analysis 2 is performed as part of the LeARN study. CSF = cerebrospinal fluid, Aβ = amyloid beta, t-tau = total tau, p-tau = phosphorylated tau.