CSF Tests for AD

abstract & commentary

Source: Hulstaert F, et al. Improved discrimination of AD patients using b-amyloid and tau levels in CSF. Neurology 1999;52:1555-1562.

Elevated levels of tau protein in the cerebrospinal fluid (CSF) and decreased levels of the 42 peptide form of beta-amyloid (Ab42) have been reported in Alzheimer’s disease (AD). Assays for these biomarkers have been developed and marketed as putative Alzheimer diagnostic tests. To further evaluate the value of these tests in identification and differential diagnosis of AD, a cooperative study was carried out across 10 centers in the United States and Europe.

In contrast to previous studies in which a single centralized laboratory performed the relevant CSF studies, each of the participating centers in this trial carried out the tau and Ab42 assays locally using ELISA kits designed for this purpose (Innogenetics, Ghent, Belgium). Cumulatively, a total of 150 samples from AD patients and 79 samples from patients with non-Alzheimer dementias were studied. An additional 184 samples were obtained from normal volunteers, patients with non-CNS pathology, and patients with other non-dementing neurologic disorders. All CSF from patients was derived from archival sources, while specimens from normal volunteers were obtained prospectively. Diagnosis was based on clinical judgment using published criteria. Optimal cutoff values for distinguishing AD patients from control groups were determined by applying exploratory statistical techniques.

The median value of Ab42 in AD patients was nearly half that of nondemented controls, while the CSF tau levels of AD patients were more than double those of the other groups. In a receiver operating characteristic (ROC) curve analysis, the combined use of CSF tau and Ab42 was reported to yield a specificity of 86% with a sensitivity of 85% in discriminating AD from non-demented controls. At a sensitivity of 85%, the combined test specificity was 58% for distinguishing patients with non-Alzheimer dementias from those in other groups. Neither test used in isolation achieved greater than 65% specificity in identifying AD patients. Hulstaert and colleagues conclude that the combined measurement of CSF tau and Ab42 meets requirements for clinical use in distinguishing AD from normal aging and specific neurologic disorders.


As we move away from the era of exclusionary diagnosis of AD toward one in which direct methods foster early diagnosis, biomarkers have the potential to play an increasingly important clinical role. A sizable body of data now indicates that high levels of tau protein and low Ab42 are more commonly observed in the CSF of AD patients than normals. Decreased Ab42 levels have been reported to correlate with possession of APOE-e4, while CSF tau levels appear to be independent of APOE status.

Nevertheless, although the multicenter study by Hulstaert et al provides another citation in support of the association between high CSF tau, low CSF Ab42 levels, and AD, it has only a limited clinical use of these markers for diagnostic purposes. It is misleading to suggest that the combined use of these tests "meets the requirements for clinical use" when the sensitivity and specificity values are assigned on the basis of ROC curve analysis rather than prospectively determined cutoffs. To improve diagnostic discrimination, use of these markers or the combined use of clinical criteria and these markers would have to bring us closer to the gold standard of autopsy diagnosis. The fact that this study cannot be used to define the positive predictive value is a strong indication that it is insufficient evidence to establish the clinical use of these measures.

Based on currently available information, CSF markers such as tau and Ab42 might be considered suitable for adjunct diagnostic use under special circumstances, but we cannot recommend them for routine clinical use. No laboratory test yet developed for the purpose of differential diagnosis of dementia can achieve the accuracy provided by careful evaluation and follow-up by a suitably trained and experienced clinician. —nrr

Alzheimer’s disease patients tend to have:

a. lower CSF tau and Ab42 levels than normal.

b. higher CSF tau and Ab42 levels than normal.

c. higher CSF tau and lower Ab42 levels than normal.

d. lower CSF tau and higher Ab42 levels than normal.