Categorical and correlational analyses of baseline fluorodeoxyglucose positron emission tomography images from the Alzheimer's Disease Neuroimaging Initiative (ADNI)

Abstract

In mostly small single-center studies, Alzheimer's disease (AD) is associated with characteristic and progressive reductions in fluorodeoxyglucose positron emission tomography (PET) measurements of the regional cerebral metabolic rate for glucose (CMRgl). The AD Neuroimaging Initiative (ADNI) is acquiring FDG PET, volumetric magnetic resonance imaging, and other biomarker measurements in a large longitudinal multi-center study of initially mildly affected probable AD (pAD) patients, amnestic mild cognitive impairment (aMCI) patients, who are at increased AD risk, and cognitively normal controls (NC), and we are responsible for analyzing the PET images using statistical parametric mapping (SPM). Here we compare baseline CMRgl measurements from 74 pAD patients and 142 aMCI patients to those from 82 NC, we correlate CMRgl with categorical and continuous measures of clinical disease severity, and we compare apolipoprotein E (APOE) ɛ4 carriers to non-carriers in each of these subject groups. In comparison with NC, the pAD and aMCI groups each had significantly lower CMRgl bilaterally in posterior cingulate, precuneus, parietotemporal and frontal cortex. Similar reductions were observed when categories of disease severity or lower Mini-Mental State Exam (MMSE) scores were correlated with lower CMRgl. However, when analyses were restricted to the pAD patients, lower MMSE scores were significantly correlated with lower left frontal and temporal CMRgl. These findings from a large, multi-site study support previous single-site findings, supports the characteristic pattern of baseline CMRgl reductions in AD and aMCI patients, as well as preferential anterior CMRgl reductions after the onset of AD dementia.

Introduction

Fluorodeoxyglucose positron emission tomography (FDG PET) studies reveal characteristic and progressive reductions in the cerebral metabolic rate for glucose (CMRgl) in patients with clinically characterized and subsequently autopsy confirmed Alzheimer's disease (AD) (Hoffman et al., 2000, Jagust et al., 2007, Mielke et al., 1996, Minoshima et al., 2001, Silverman et al., 2001) and in patients with mild cognitive impairment (MCI), who have an increased risk of AD neuropathology and subsequent conversion to probable AD (pAD, the term commonly used for the clinical diagnosis of Alzheimer's dementia) (Arnaiz et al., 2001, Drzezga et al., 2003, Drzezga et al., 2005, Minoshima et al., 1997). Similar CMRgl reductions have been reported in cognitively normal people with one or two copies of the apolipoprotein E (APOE) ɛ4 allele, a common AD susceptibility gene, many years before the anticipated onset of symptoms (Reiman et al., 1996, Reiman et al., 2001, Reiman et al., 2004, Reiman et al., 2005). These and other studies raise the possibility that FDG PET, along with other brain imaging and biomarker measurements, could be used for earlier detection and tracking of Alzheimer's disease, in the differential diagnosis of AD in patients with dementia and MCI, in the enrichment or stratification of research subjects in clinical trials of putative AD-slowing therapies, and in the rapid evaluation of putative AD-slowing, risk-reducing and prevention therapies.

In numerous single-center (Alexander et al., 2002, Chase et al., 1984, de Leon et al., 1983, Duara et al., 1986, Foster et al., 1983, Foster et al., 1984, Haxby et al., 1990, Hoffman et al., 2000, Ibanez et al., 1998, Jagust et al., 1988, McGeer et al., 1990, Minoshima et al., 1994, Minoshima et al., 1995, Smith et al., 1992). multi-center (Herholz et al., 2002, Mosconi et al., 2008b, Silverman et al., 2001) FDG PET studies, AD has been associated with significantly lower CMRgl bilaterally in the precuneus and posterior cingulate, parietal and temporal cortex, and also with lower CMRgl in the frontal cortex and whole brain in more severely affected patients. Lower CMRgl has been correlated with dementia severity, either by assessing the correlation between performance on measures of overall cognitive function, such as the Mini-Mental State Exam [MMSE], and CMRgl reduction over all study participants, or by comparing subgroups based on measures of disease severity, such as the MMSE, Clinical Dementia Rating [CDR], or Global Deterioration Scale score (Chase et al., 1984, Choo et al., 2007, Foster et al., 1984, Minoshima et al., 1995, Mosconi, 2005, Silverman et al., 2001, Smith et al., 1992). The CMRgl reductions correspond to the spatial pattern of gray matter atrophy observed using volumetric magnetic resonance imaging (MRI) (Chételat et al., 2008) and they predict subsequent cognitive decline and the histopathological diagnosis of AD (Hoffman et al., 2000, Minoshima et al., 2001, Silverman et al., 2001). In longitudinal studies, the CMRgl reductions are progressive (Alexander et al., 2002, Haxby et al., 1990, Jagust et al., 1988, McGeer et al., 1990, Mosconi, 2005, Smith et al., 1992), and offer greater statistical power than clinical endpoints in the evaluation of putative AD-slowing treatments (Alexander et al., 2002, Reiman et al., in press).

In a smaller number of mostly small single-center FDG PET studies, amnestic mild cognitive impairment (aMCI) and non-amnestic MCI are associated with lower CMRgl in some of the same brain regions as pAD (de Leon et al., 2001, De et al., 2001, Drzezga et al., 2003, Mosconi et al., 2004, Mosconi, 2005, Mosconi et al., 2007), including direct comparisons between MCI and pAD (De Santi et al., 2001, Mosconi et al., 2005, Mosconi et al., 2008b, Nestor et al., 2003), and in an automatically characterized hippocampal region-of-interest (Mosconi et al., 2005, Mosconi et al., 2008a). A similar pattern of hypometabolism in the posterior cingulate cortex and hippocampus was reported in a large, multi-center study which included 114 patients with MCI (Mosconi et al., 2008b). In related studies, lower CMRgl in some of these regions predicted subsequent rates of clinical conversion to probable AD (Arnaiz et al., 2001, Drzezga et al., 2003, Drzezga et al., 2005, Minoshima et al., 1997) are progressive and able to distinguish those who subsequently converted to pAD from those who remained stable during the same time-frames (Anchisi et al., 2005, de Leon et al., 2001, De Santi et al., 2001, Drzezga et al., 2003, Mosconi et al., 2004). In one very small study that needs to be replicated, lower CMRgl in a preselected posterior cingulate region of interest (ROI) and APOE ɛ4 carrier or non-carrier status were used together to completely distinguish between aMCI patients who rapidly converted to pAD and those who remained stable during the same time frame (Drzezga et al., 2005).

The AD Neuroimaging Initiative (ADNI) was launched in 2003 by the National Institute on Aging (NIA), the National Institute of Biomedical Imaging and Bioengineering (NIBIB), the Food and Drug Administration (FDA), private pharmaceutical companies and non-profit organizations, as a $60 million, 5-year public-private partnership. ADNI is a large, multi-center, longitudinal study of 822 older adults, including 188 initially mildly affected pAD patients, 405 aMCI patients, and 229 cognitively normal controls (NC) who are being followed at 58 clinical sites in the United States (Mueller et al., 2005). All of the subjects have clinical ratings, neuropsychological tests, 1.5 T volumetric MRI, and blood and urine samples at every visit (most commonly every 6 months for 2–3 years depending on the subject group); half of the subjects also have FDG PET or 3 T MRI at every visit; more recently, a smaller number of subjects have fibrillar amyloid imaging measurements using Pittsburgh Compound B (PiB) PET; and more than half of the subjects have agreed to baseline and one-year cerebrospinal fluid (CSF) measurements.

ADNI is intended to provide a public resource to the entire community, providing privacy-protected data and images, selected biological samples, and findings from a limited number of analyses. It is intended to assist in the early detection and tracking of AD, and to provide information that could be used to help in estimating sample sizes and designing clinical trials of putative AD-slowing treatments in pAD and aMCI patients using brain imaging and other biomarker endpoints, as well as lessen the time and cost of clinical trials. ADNI provides a foundation to compare different brain imaging modalities and image-analysis techniques for each of these purposes. Along the way, it has begun to set standards for the selection of pAD and aMCI patients, the qualification of clinical and imaging sites in multi-center clinical trials, the acquisition and centralized pre-processing of MRI and PET images, use of MRI phantoms, and real-time quality-assurance and quality-control procedures in multi-center clinical trials. This study complements other multi-center FDG PET studies (Herholz et al., 2002, Mosconi et al., 2008b, Silverman et al., 2001), as well as other multi-center neuroimaging initiatives that have begun in other continents and countries following the ADNI example, including Europe, Japan, and Australia (Frisoni et al., 2008).

As members of the ADNI PET Coordinating Center, we are responsible for a small number of voxel-based FDG PET analyses using statistical parametric mapping (SPM). For this report, we provide a descriptive report of the comparisons between baseline CMRgl measurements from 74 pAD patients and 142 aMCI patients to those from 82 NC; we characterized correlations between three categorical measurements of clinical disease severity (i.e., NC, aMCI and pAD based on CDR scores), as well as correlations between a continuous measure of clinical disease severity (MMSE scores) and lower CMRgl using data from all 298 subjects; and we characterized correlations between the same continuous measure of clinical disease severity and lower CMRl within the pAD group. We predicted that pAD and aMCI patients would have significantly lower CMRgl than NC in precuneus and posterior cingulate, parietal, and temporal cortex; that the pAD group would also have lower CMRgl in frontal cortex; that clinical disease severity would be correlated with lower CMRgl in each of these regions when the analysis included all three subject groups, and that it would be correlated with lower CMRgl in frontal cortex when the analysis was restricted to those AD patients with dementia. Establishing that similar findings between previous single-site studies and large, multi-site studies can be obtained is critical as the field advances. Findings from the comparison of aMCI patients who subsequently converted to pAD and those who remain stabled during the same time frame will be described in a separate report.