Findings from this large multi-center study confirm the previously characterized pattern of regional hypometabolism in pAD and aMCI found in single- and multi-center studies and implicate additional brain regions. Furthermore, they confirm the previously characterized correlation between severity of clinical impairment in a combined group of patients and controls, but suggest that the hypometabolism observed in the frontal brain regions is associated with a more severe disease state and may be indicative of potential disease progression. Lastly, the results provide additional support for the comparability of images obtained from multi-site studies using different PET scanners (Herholz et al., 1993
; Herholz et al., 2002
; Mosconi et al., 2008b
; Silverman et al., 2001
), at least with the scanner qualification, real-time quality-assurance and quality control procedures, and pre-processing standardization steps used in ADNI.
The findings from the present study indicate that compared to normal controls, both the pAD and aMCI participants exhibited similar patterns of reduced CMRgl in posterior cingulate, precuneus, parietotemporal, frontal and occipital regions. Although the patterns of reductions were comparable, the magnitude and spatial extent of hypometabolism was greater among the pAD participants, and these patient group differences were not solely attributable to the effects of slightly older age in the AD group. These results, along with the relatively higher percentage of APOE ε4 carriers, add to the evidence that a significant proportion of patients meeting the aMCI criteria used in this study have early AD. Our ongoing longitudinal assessments will permit us to further characterize rates of clinical conversion to pAD, and determine the extent to which regional CMRgl reductions, alone or in combination with other measurements and risk factors, predicts clinical conversion to pAD.
Previous studies have reported significant correlations between clinical disease severity and lower MMSE scores and lower CMRgl in patients with pAD in posterior cingulate, precuneus, parietotemporal and frontal cortex (Bokde et al., 2005
; Chase et al., 1984
; Foster et al., 1983
; Kawano et al., 2001
; Mielke et al., 1992
; Minoshima et al., 1995
). Generally speaking, the results from the present study are in agreement with findings from previous studies and are noteworthy given the number of participants and the selection of pAD patients based on relatively mild clinical severity. In addition, we found a relationship between hypometabolism in the occipital cortex and greater disease severity, which, has been reported in some but not all FDG PET studies of AD (e.g., Alexander et al., 2002
), but which has also been suggested to distinguish dementia with Lewy bodies (DLB) from AD (Minoshima et al., 2001
). This finding might be partly attributable to our relatively large samples, permitting us to detect more subtle group differences than in some of the smaller single-center studies, and it may be partly related to the finding of preferential fibrillar amyloid-beta burden in AD patients (Klunk et al., 2004
), which has been suggested to reflect cerebral amyloid angiopathy (Greenberg et al., 2008
; Johnson et al., 2007
). It also is possible that some of these individuals may have unsuspected Lewy body pathology, although the presence of Parkinsonian signs was exclusionary at enrollment.
When the analysis was restricted to only participants with pAD, the significant correlations between lower MMSE scores and lower CMRgl were limited to left frontal and temporal brain regions, suggesting that these regions may be preferentially affected after the onset of dementia. Hypometabolism in the left frontal and temporal brains regions has previously been reported in a longitudinal PET study of AD and may be considered to be relevant to progressive decline after the onset of dementia (Alexander et al., 2002
; Choo et al., 2007
). While we did observe lateralized CMRgl reductions, the post-hoc hemisphere-wise comparisons did not find any asymmetric significance. Preliminary findings (unpublished) from the longitudinal ADNI data suggest that, for mild pAD individuals, the posterior regions may still exhibit CMRgl decline initially, but, over a larger range of clinical decline and longer duration, we postulate that the declines in the posterior regions may abate with stronger frontal declines.
We did not find significant correlations between lower MMSE scores and lower CMRgl in the aMCI or NC groups. The absence of significant associations could be attributable to the modest range of MMSE scores in these groups, heterogeneity in the aMCI group, or a combination of these and other factors. It is possible that other neuropsychological measurements or clinical ratings could provide a more sensitive measure of cognitive decline and these subject groups, and that these measures be associated with lower CMRgl in brain regions preferentially affected by AD or normal aging. For instance, in a study with aMCI patients using annual change on the Mattis dementia rating scale as the measure of disease severity, the authors identified a significant positive relationship between CMRgl and disease severity in the right lateral temporo-parietal and bilateral medial frontal regions (Chételat et al., 2005
). Examining the association between CDR sum of boxes (CDR-SB) and CMRgl in patients with aMCI, another team reported a voxel-based inverse association in the right posterior cingulate gyrus, although it did not survive correction for multiple comparisons (Perneczky et al., 2007
Analyzing the effects of the APOE ε4 allele, cognitively normal ε4 carriers had lower CMRgl than non-carriers in several AD-affected precuneus and frontal regions, confirming findings from previous single-center studies (Reiman et al., 1996
; Reiman et al., 2004
; Reiman et al., 2005
; Small et al., 1995
). Still, these findings were less extensive than those reported in our own studies of cognitively normal late-middle-aged carriers and non-carriers, perhaps reflecting the combined effects of slightly older age and differential survivor bias, the inclusion of subjects who were less closely matched for age, gender and educational level, reduced sensitivity due to the inclusion of data from multiple scanners, or other factors. In aMCI patients, normal ε4 carriers had lower CMRgl than non-carriers in AD-affected lateral temporal regions, which could reflect a higher proportion of AD cases in the MCI patients with this susceptibility gene, more advanced disease in the APOE ε4 carriers, or a combination of these and other factors Additional analyses are needed to further clarify the extent to which APOE ε4 gene dose, the pattern of cerebral hypometabolism, or a combination of these and other factors predict subsequent rates of conversion to probable AD. Within the pAD group, we found very few differences in the carriers compared to the non-carriers, suggesting that the pattern of FDG hypometabolism in probable AD patients is not attributable to effects of the APOE ε4 allele independent of AD itself.
Among its important goals, ADNI is intended to help in the design and performance of multi-center clinical trials of putative AD-slowing treatments. ADNI has already developed and successfully implemented the protocols used to acquire 1.5T MRI, 3T MRI, FDG PET and PIB PET images in a standardized way, helping to account for differences among imaging systems, and demonstrating comparability in the quality of images on most--but not all--of the imaging systems assessed. Second, it has already developed and successfully implemented the site-qualification, real-time quality-assurance, image-preprocessing, and centralized data management procedures needed to provide high-quality data. Third, it has demonstrated the feasibility of collecting CSF samples in a high proportion of study participants. Fourth, it has already begun to provide data and specimens for the early detection and tracking of AD. Fifth, ADNI provides the means to evaluate promising new image analysis techniques and biomarker measurements, comparing them to existing methods and measurements in the early detection and tracking of AD. Perhaps most important, ADNI researchers have begun to characterize and directly compare different imaging methods, data analysis methods, and biomarkers in their ability to distinguish subject groups, their ability to predict rates of cognitive decline and clinical progression from MCI to probable AD, and their statistical power to evaluate putative β-amyloid-modifying and disease-slowing treatments in multi-center clinical trials, thereby facilitating the potential use of these methods for participant stratification or enrichment in clinical trials.
This study has several limitations. As previously mentioned, the MMSE is only one potential measure of disease severity and overall cognitive functioning. In addition, the ADNI study is not an epidemiological sample and may not be representative for all cognitively healthy older adults, aMCI, or mild pAD patients. Participants in ADNI were recruited primarily from subspecialty memory clinics and Alzheimer’s research centers, have unusually high levels of education, and included more males than females, and may have other factors that reflect their level of motivation required to participate in this demanding study. The study was not designed to test the diagnostic utility of imaging. Participants were highly selected and patients with concurrent medical, neurological, and psychiatric illness, or atypical features were excluded. Thus the study population does not reflect the diversity of patient characteristics seen in clinical practice. Furthermore, our MCI findings are restricted to the amnestic MCI subtype and may not be applicable to MCI patients with different cognitive profiles. Lastly, while every attempt was made to ensure that participants assigned to each subgroup were in fact cognitively healthy, had aMCI or pAD, there is always the possibility that some participants were misclassified when it comes to their underlying neuropathology. In our initial assessment of image quality, we found potentially confounding patterns of FDG uptake in the reconstructed images acquired on Siemens HRRT and BioGraph HiRez scanners, causing us to eliminate these scans from our primary analysis, though the findings were similar when these scans were included in our post-hoc analyses. (It remains to be determined what effect the inclusion of these or other scanners might have in terms of the power to detect or track AD-related changes in multi-center trials. ., Further, it remains possible that additional refinements in image reconstruction may reduce the variability initially associated with these scanners.
In conclusion, findings from this and other single-site and multi-center studies support the characteristic pattern of preferential posterior cingulate, precuneus, parietotemporal CMRgl reductions previously found in AD and aMCI patients, as well as preferential anterior CMRgl reductions after the onset of dementia.