To tap or not to tap: Cerebrospinal fluid biomarkers of Alzheimer's disease
J. Growdon
DOI: https://doi.org/10.1002/ana.410440105
IF: 11.2
1998-07-01
Annals of Neurology
Abstract:In this issue, Kanai and colleagues' report on the largest series to date on AP and tau measurements in cerebrospinal fluid (CSF) in patients with Alzheimer's disease (AD), non-Alzheimer's dementia, other degenerative neurological diseases, and control subjects. They confirm prior reports that AP42 is reduced in CSF of AD patients2 and tau is increased in CSF of AD pat i e n t ~ . ~ Further, they propose a new AD index, which is a recalculation of tau and the AP ratio (tau X (AP40/AB42) = AD index), and report that it is a more sensitive diagnostic measure for AD than either AP42 or tau alone. The authors emphasize the potential practical value of these findings and join a chorus of voices urging these measures as biomarkers for AD. The study of Kanai and colleagues is considered here in light of the consensus statement of a working group on molecular and biochemical markers of AD sponsored by the Reagan Research Institute of the Alzheimer's Association and the National Institute on Aging4 The consensus statement identified five different uses of a biomarker: epidemiological screening, predictive testing, confirming diagnosis, monitoring progression over time or in response to treatment, and studying brainbehavior relationships. Of these five, the results reported by Kanai and associates have potential relevance for three-aid to diagnosis, measure of disease progression, and brain-behavior correlations. The diagnostic accuracy of a test is generally expressed in terms of sensitivity and ~pecificity.~ Sensitivity refers to the capacity of a biomarker to identify a substantial percentage of patients with the disease; a sensitivity of 100% indicates that a marker can identify all of the patients with AD. Specificity refers to the capacity of a test to distinguish probable AD from normal aging and other causes of dementia. A test with 100% specificity would distinguish AD from other causes of dementia in every case. Since the introduction of formal criteria for the diagnosis of AD,6 the clinical diagnosis of AD, at least in specialized university-based clinics, is confirmed neuropathologically in 80% to 90% of the patients examined. For this reason, the consensus statement urged that for a biomarker to be useful as a diagnostic test, its sensitivity and specificity should approach and preferably exceed 80%. Furthermore, the diagnostic value of a biomarker will be greatest in the mildest cases, or in the cases with the most recent onset, in which the diagnostic uncertainty is greatest. After several years of dementia progression, the diagnosis is generally apparent to family and friends as well as the physician, and a confirmatory diagnostic test is rarely needed. At initial CSF examination, the results of Kanai and co-workers indicate that tau alone, AP,, alone, the AP ratio AP40/AP42 alone, and the tau and/or AP ratio fails to meet the workshop criteria for a useful diagnostic test. In each case where sensitivity was high, specificity was low, or vice versa. Only the AD index approached but did not meet the recommended standard. This report is the first to propose the AD index calculation and to show that it is superior to measures of tau and AP alone or in other calculated combinations. Additional studies from other investigators are necessary to confirm the diagnostic value of the AD index. Of the original 93 AD subjects, 32 had a second CSF examination 2 to 42 months (mean, 18.6 months) after the first examination. Mean values for all three measures-tau, the AP ratio, and the AD index-were greater in the second CSF sample than in the first. On the surface, these findings imply that any of the three measures would be good indices of AD progression, accepting the rough correlation between duration of illness and worsening of dementia. This conclusion is complicated, however, since mean values increased significantly only in those subjects whose initial values were below the set cutoff values for diagnosis. Subjects with high values for tau, AP ratio, and AD index did not have significantly higher values for these measures on repeated examination. This finding implies that there is a ceiling effect. Attention needs to be focused on the period of time during which these changes occur in order to define the dynamic range of the CSF measures. For the present, these observations detract somewhat from the goal of using CSF indices of tau and to follow the course of illness. As yet, there are no data on how these measures change in response to antidementia treatment. Dementia results from decreases in synaptic connections and neuronal loss, which far outstrip neurofibrillary tangle (NFT) acc~mulation.~ Both neuronal loss and NFT formation, however, increase with duration of illness' and are biological indices of dementia severity. At initial CSF examination, tau levels but not AP levels correlated significantly with estimates of dementia severity, as judged by the Mini-Mental State Examination (MMSE). This modest correlation improved somewhat over time, as the mean MMSE score declined and tau levels increased. Because the AD index is heavily influenced by the tau values, there was a significant correlation between this index and MMSE scores as well. These observations are consistent with some' but not all" previous reports on the relation between tau and the MMSE and suggest that tau levels