This study presents an analysis of cross-sectional clinical correlations of 2 core disease indicators in AD: MRI and CSF biomarkers. Our results suggest that of the 2 classes of disease indicators, structural MRI changes were more closely related to general cognitive and functional indices of disease stage in impaired subjects. These results are concordant with some CSF and MRI studies that have shown that CSF biomarkers do not correlate with cognitive measures cross-sectionally among patients with AD22
nor with plaque and tangle burden,23
whereas MRI biomarkers (such as hippocampal volume, STAND score) correlate with both degree of cognitive impairment as well as Braak NFT staging.24,25
All MRI/CSF biomarkers were found to be significant for intergroup discrimination of CN, aMCI, and AD. The combined predictor model with both MRI and t-tau/Aβ1-42
ratio performed better than any one biomarker alone and the contribution of both CSF and MRI was found to be significant. p-tau Epitopes are believed to be most useful for differentiating AD and non-AD dementias,26
which may explain why p-tau181
was not more sensitive than t-tau measurements in intergroup discrimination in subjects who lay along the normal to AD continuum.
The 3 disease markers examined in this article (MRI, CSF Aβ1-42
, and CSF t-tau) reflect different aspects of AD pathology. Low CSF Aβ1-42
is a marker of fibrillary amyloid deposition in plaques. Nearly complete concordance is present between individuals with positive Pittsburgh Compound B (PIB)–PET scans and those with low CSF Aβ1-42
Although correlations with Aβ1-42
were present in our study, well accepted reasons exist to explain why Aβ1-42
might not correlate highly with clinical indices of disease stage. Amyloid deposition is regarded to be an early event that occurs prior to clinical symptoms. In one proposed model of AD, a full complement of amyloid is deposited and then plateaus with little further deposition.28
Cognitive decline, as well as NFTs and synaptic loss, progressively worsen in the presence of a relatively static total load of amyloid.28
Animal data also indicate that amyloid plaque deposition precedes NFT.29
An additional possible explanation for our findings is that measurement of CSF Aβ1-42
appears to be inherently more variable than MRI ().
Increased CSF t-tau is a marker of neuronal injury which correlates well with NFT stage and NFT load.30,31
Atrophy on structural MRI also correlates with Braak NFT stage and NFT load24,25
but the most proximate histologic correlate of MRI volume loss is loss of neurons and synapses.3,32
It may at first be surprising to find that correlations with clinical disease stage are slightly stronger for MRI vs t-tau given that CSF t-tau is usually regarded as direct marker of neuronal injury. However, autopsy studies have shown that the appearance of NFT pathology in entorhinal cortex (EC) precedes the appearance of EC neuronal loss.33
Therefore, assuming CSF t-tau is a direct reflection of NFT pathology and atrophy on MRI is a direct result of neuron and synapse loss, one might expect slightly better correlation between clinical indices of disease stage (which themselves reflect neuron and synapse loss) and MRI than with CSF t-tau. The fact that the literature to date on CSF–clinical correlations contains some seemingly contradictory results in late onset AD (which is what our subject population represents) supports this notion. For example, one study reported no change in CSF t-tau levels over time in patients with AD and concluded that CSF t-tau does not predict either severity or rate of clinical decline in AD.34
Rates of change on MRI did correlate with change on MMSE scores, but change in CSF t-tau did not.35
Conversely, another study36
found that baseline CSF t-tau (and Aβ1-42
) predicted conversion to dementia and another22
found a direct correlation between increasing levels of CSF t-tau and severity of impairment in AD. The relationship between CSF t-tau and disease stage may therefore be complex. In contrast, the relationship between clinical disease stage and MRI seems to be a fairly straightforward direct correlation since MRI measures atrophy, which reflects cumulative damage. The literature on MRI is nearly unanimous in indicating close correlation between loss of cognitive function and loss of volume on MRI over time.37,38
A final possible explanation for our finding of better correlation between MRI and cognition than between CSF t-tau and cognition is that MRI may be a more stable indicator of neuronal injury. Brain volume quantification with MRI has nothing analogous to daily turnover of a soluble protein. Minimal physiologic variation in brain volume may translate into stronger correlations with cognition over many subjects.
Typically MRI and CSF biomarkers have been shown to have an accuracy of 80%–90% in discriminating AD and CN. In this study, we found the performance of MRI and CSF was slightly lower than the numbers seen in the literature. Despite rigorous standardization of procedures and processing, this could be attributed to the fact that ADNI is a multisite study with known site-to-site variation in methods of subject recruitment. We explored whether there was any evidence of site-specific differences that may require adjusting biomarker values and found that subject-to-subject variability was much greater than site-to-site variability for all biomarkers with intraclass correlations below 0.02 across all biomarkers. This suggests that less than 2% of the variability in biomarker values is due to site differences. In light of these findings, and because of the large number of sites and relatively few subjects per site, incorporation of site into our analyses was not required.
Advances in computational power and MRI technology in the last decade have enabled us to obtain automated MRI biomarkers such as STAND scores for assessing the disease state. Advantages of using MRI biomarkers are the noninvasive nature of the imaging modality, low processing time, and automation of biomarker estimation. This study also validates that MRI scans from different centers can be combined and STAND scores perform reasonably well for diagnostic purposes. There are two limitations of this study: 1) ADNI population is not generalizable to the general population. The recruitment mechanisms were those used for clinical trials in AD and included memory clinics, patient registries, public media campaigns, and other forms of public advertisements. 2) Since the gold standard is clinical diagnosis, which is based on the screening tests, we cannot directly evaluate the additive value of the biomarkers to clinical methods. This will require a different study design.