Hypometabolism Exceeds Atrophy in
Presymptomatic Early-Onset Familial
Lisa Mosconi1,2, Sandro Sorbi2, Mony J. de Leon1, Yi Li1, Benedetta Nacmias2, Paul S. Myoung1, Wai Tsui1,
Andrea Ginestroni2, Valentina Bessi2, Mozghan Fayyazz2, Paolo Caffarra3, and Alberto Pupi2
1Department of Psychiatry, New York University School of Medicine, New York, New York;2Department of Clinical Physiopathology
and Neurology, University of Florence, Florence, Italy; and3Department of Psychiatry, University of Parma, Parma, Italy
The aim of the present study is to compare brain atrophy with
hypometabolism as preclinical markers of Alzheimer’s disease
(AD) by studying presymptomatic individuals from families with
risk FAD individuals (age, 35–49 y; 4 women; education $ 12 y)
and 7 matched healthy control subjects received complete clin-
ical, neuropsychologic, MRI, and18F-FDG PET examinations.
Regions of interest (ROIs: whole brain [WB], hippocampus
[Hip], entorhinal cortex [EC], posterior cingulate cortex [PCC], in-
ferior parietal lobule [IPL], and superior temporal gyrus (STG])
volumes on MRI and glucose metabolism (MRglc) from the MRI-
with controls and after correcting for head size, MRI volume
reductions in FAD subjects were restricted to the IPL (18%,
PET MRglc reductions were found in all FAD subjects compared
with controls in the WB (13%), bilaterally in the IPL (17%) and in
the STG (12%), and in the left EC (21%), PCC (20%), and Hip
tently less variable than MRI measures, yielding significantly
larger effect sizes in separating FAD from controls. Conclusion:
Presymptomatic FAD individuals show widespread MRglc re-
absence of structural brain atrophy. These data further suggest
that PET MRglc measures may serve as biomarkers for the pre-
clinical diagnosis of AD.
Key Words: familial Alzheimer’s disease;18F-FDG PET; brain
atrophy; Presenilin 1 gene; hypometabolism; presymptomatic
J Nucl Med 2006; 47:1778–1786
To develop prevention treatments for Alzheimer’s
disease (AD) it is necessary to identify early biologic markers
that are predictive of AD. To date, the best-recognized in
vivo markers of AD are measures of brain structure and
function as obtained with neuroimaging. Brain volume loss
(atrophy), as assessed on MRI, and reductions in the meta-
bolic rate of glucose (MRglc), as measured with18F-FDG
PET, are sensitive to AD-related brain changes. A crucial
question is whether these brain measures are capable of
preclinical detection of dementia. Understanding how AD
starts and progresses requires following subjects through
normal aging to the onset of clinical symptoms. However,
given the low incidence and slow progression of healthy
elderly to AD (1), such studies are hindered by the need for
very large samples and long follow-up intervals to observe
significant numbers of declines.
An alternative has been to examine individuals from fam-
ilies with early-onset AD with known genetic mutations.
Familial AD (FAD) is characterized by autosomal dominant
inheritance with nearly 100% penetrance and a specific age
of onset for a givenpedigree (2). Therefore, presymptomatic
at-risk individuals close to the expected age of onset may
provide unique information about preclinical AD-related
As these rare genetic mutations are found in 5% of the
AD cases in the general population (2), only a few neuro-
imaging studies on FAD exist, particularly in individuals at
a presymptomatic stage. Most FAD studies were done with
MRI and revealed that the onset of dementia is accompa-
nied by progressive structural atrophy in the hippocam-
pal formation and in the neocortical regions (3,4). The few
18F-FDG PET studies on FAD, including at-risk and sympo-
matic individuals, showed severe cortical MRglc reductions
in the parietotemporal and posterior cingulate cortices com-
pared with control subjects (5,6). However, to our knowl-
edge, no study has compared the MRI atrophy and PET
By using MRI and18F-FDG PET, the aim of the present
study was to examine brain atrophy and hypometabolism as
early indicators of future AD by studying presymptomatic
at-risk individuals from families with known early-onset
AD as caused by mutations in the Presenilin 1 (PS-1) gene.
Received May 3, 2006; revision accepted Jul. 28, 2006.
For correspondence or reprints contact: Lisa Mosconi, PhD, Center for
Brain Health, MHL-400, New York University School of Medicine, 560 First
Ave., New York, NY 10016.
1778THE JOURNAL OF NUCLEAR MEDICINE • Vol. 47 • No. 11 • November 2006
MATERIALS AND METHODS
Seven at-risk individuals from 3 unrelated Italian families with
known early-onset autosomal dominant FAD (7–9) carrying mu-
tations in the PS-1 gene were enrolled in an ongoing longitudinal
clinical/MRI/FDG PET study at the University of Florence (UoF),
Italy. These individuals were compared with 7 healthy, medica-
tion-free, age-, sex-, and education-matched control subjects. The
7 control subjects were 3 siblings from the same Italian families
who did not carry the genetic mutations (7–9) and 4 healthy sub-
jects derived from the clinical/MRI/FDG PET database of healthy
volunteers at the Center for Brain Health at the New York Univer-
sity (NYU) School of Medicine. Both centers administer several
identical neuropsychologic tests, use comparable criteria for the de-
finition of normality, follow the recommendations of the NINCDS–
ADRDAwork group (10) and the Diagnostic and Statistical Manual
of Mental Disorders (DSM-IV) (11) criteria for the diagnosis of
AD, and use common standardized imaging acquisition protocols.
All subjects received an extensive screening and diagnostic bat-
tery that consisted of medical, neurologic, psychiatric, neuropsy-
chologic, MRI, and18F-FDG PET examinations. Criteria for the
diagnosis of normal (or asymptomatic in FAD individuals) were
(a) no evidence of functional impairment in the subject based on
intensive interviews and (b) a structured clinical interview result-
ing in a Clinical Dementia Rating (CDR) score 5 0 (12) or Global
Deterioration Scale (GDS) score 5 1 (13).
The clinical impression for 1 FAD subject (Table 1, FAD-3)
was less certain, and at the clinical consensus meeting this subject
was assigned a CDR score of 0.5, which indicates mild functional
impairments that do not reach a dementia level (12). However, this
subject was included in the present study, as the subject was con-
sidered to be asymptomatic by himself and the family.
None of the subjects had evidence of conditions affecting brain
structure or function—that is, stroke, clinically uncontrolled
diabetes, major head trauma, and depression as assessed on the
Hamilton Psychiatric Rating Scale for Depression (14) or use of
cognitively active medications. The participants provided informed
consent and were studied under guidelines approved by the Insti-
tutional Review Board, local ethics committees, and radiation-
protection authorities at both institutions.
The subjects completed a detailed battery of neuropsychologic
tests, including the Mini-Mental State Examination (MMSE),
immediate and delayed recall of paired associates, copy and recall
of the Rey’s complex figure, the designs test, the Token test, the
Stroop test, Trail making test, and Phonemic fluency tests (15).
For all of these tests, normative reference values exist (15).
early-onset FAD was extracted from peripheral blood samples with
the phenol–chloroform procedure, and all coding and 59 noncoding
exons of the PS-1 gene were analyzed with polymerase chain
reaction (PCR) amplification (7–9). Single-strand conformation
polymorphism (SSCP) was performed and, in the presence of
irregular SSCP patterns, PCR products were sequenced. Apolipo-
protein E (ApoE) genotyping was also performed using standard
PCR analysis. Of the 7 FAD individuals, 3 carried Leu392Val,
1 carried Met146Leu, and 3 carried Cys92Ser mutations in the
PS-1 gene (7–9) (Table 1). Description of codon changes and
phenotypicprofilesandinformation aboutthe expectedageatonset
for these 3 mutations are available from symptomatic FAD individ-
uals from the same families as well as other families as described
(7–9) (http://www.molgen.ua.ac.be/ADMutations). The mean age
at onset for the family carrying the Leu392Val mutation is 45 6 4 y
(range, 40–48 y) (8), for the family carrying the Met146Leu
mutation is 41 6 4 y (range, 38–45 y), and for the family carrying
the Cys92Ser mutation is 62 6 4 y (range, 56–67 y) (7,9).
MRI Study. All subjects received a standardized MRI scan
protocol, which included a clinical and a research MRI scan. The
clinical scan covered the entire brain with contiguous 3-mm axial
T2–weighted and proton density–weighted images. The research
(n 5 7)
Rey’s complex figure
Paired associates, delayed recall
Trail making test, part B (s)
Stroop test (interference, s)
Phonemic fluency test
Cys92Ser Leu392Val Met146Leu Cys92Ser Leu392Val Leu392Val Cys92Ser
5 e3/e3; 2 e3/e4
Values for control group are means (SD). FAD subjects are ordered by age.
PRECLINICAL HYPOMETABOLISM IN AD • Mosconi et al. 1779
MRI scan was acquired at both centers with a 3-dimensional (3D)
T1-weighted fast-gradient-echo sequence on a 1.5-T magnet (UoF
[Gyroscan ACS-NT; Philips Medical Systems]: repetition time
[TR] 5 25 ms, echo time [TE] 5 4.6 ms, flip angle 5 30?, matrix 5
256 · 160; NYU [Signa; GE Healthcare]: TR 5 35 ms, TE 5
9 ms, flip angle 5 60?, matrix 5 256 · 128), yielding identical
tissue contrast. Images were reconstructed into 124 contiguous
slices with 1.3-mm slice thickness in a coronal orientation perpen-
dicular to the long axis of the hippocampus (Hip).
18F-FDG PET Study. Within 1 mo of the MRI, all subjects
received a PET scan using18F-FDG as the tracer, using common
and standardized procedures and scanners with comparable spatial
resolution (UoF: Advance scanner [GE Healthcare], in-plane axial
resolution 5 4.6 mm, slice thickness 5 4.25 mm, axial field of
view [FOV] 5 154 mm; NYU: ECAT EXACT HR1 scanner
[Siemens], in-plane full width at half maximum [FWHM] 5
4.5 mm, slice thickness 5 4.23 mm, axial FOV 5 155 mm). Each
subject’s head was positioned using 2 orthogonal laser beams and
imaged with the scanner tilted 25? negative to the canthomeatal
plane, which runs approximately parallel to the long axis of the
Hip. To reduce head movement during scanning, a molded plastic
head holder was custom-made for each subject. Attenuation cor-
rection was obtained using68Ga/68Ge transmission scans. Subjects
received 110–370 MBq (;5.28 MBq/kg body weight) of18F-FDG
intravenously while laying supine in a dimly lit room. PET images
were acquired ;35 min after isotope injection and lasted for
20 min. Images were reconstructed using the Hanning filter with a
frequency cutoff of 0.5 cycle/pixel and resized using a bilinear
extrapolation scheme to a 256 · 256 matrix with pixel size 5
1.52 mm and slice thickness 5 4.25 mm.
PET/MRI Coregistration and Atrophy Correction. All image
processing and data analysis were performed at NYU. MRI and
PET scans were transferred to a Sun Sparc workstation (Sun
Microsystems), where each PET scan was coregistered with the
corresponding MR image by using a 3D method based on mini-
mizing the variance of the signal ratios implemented in the Multi-
modal Image Data Analysis System package (MIDAS, version
1.6) (16). The implementation calls for a preliminary spatial align-
ment, using intrinsic anatomic landmarks followed by a surface-
fitting algorithm (16). The coregistered PET/MR images consisted
of coronal sections perpendicular to the long axis of the Hip
(256 · 256 · 91 matrix, 1-mm2pixel size, 2-mm slice thickness).
After MRI coregistration, PET scans were corrected for the
partial-volume effects of cerebrospinal fluid (CSF) using a
2-segment model (i.e., brain tissue and CSF) (17–19). On the
basis of phantom-validated threshold techniques, MRI regions are
locally sampled to minimize radiofrequency coil inhomogeneity
(20) and converted to binary images, where pixels correspond-
ing to brain are given a value of 1 and those corresponding to
nonbrain—such as CSF and air spaces—are given a value of 0.
This binary image is convolved with the 3D point spread function
of the PET camera, resulting in a recovery coefficient (RC) image,
and the coregistered PET image is divided by the RC image to
yield PET images corrected for the partial volume of CSF. The
atrophy-corrected and noncorrected PET scans were sampled using
the MRI regions of interest (ROIs).
ROIs. The ROI technique was chosen for image analysis
because it allows one to directly compare hypometabolism with
atrophy within the same brain region and to examine small
structures such as the Hip with high anatomic precision. The
selected ROIs included 2 medial temporal lobe (MTL) structures
(Hip and the anterior portion of the parahippocampal gyrus, which
corresponds to the entorhinal cortex [EC]) and 3 cortical regions
(inferior parietal lobule [IPL], superior temporal gyrus [STG], and
posterior cingulate cortex [PCC]), which are typically affected in
AD (21,22). Figure 1 provides a visual depiction of the ROIs. One
on coronal (A and B) and left sagittal (C
and D) MR images of a representative
FAD subject (FAD-1). (A) Hippocampus
(1), entorhinal cortex (2), superior tempo-
ral gyrus (3). (B) Ventricles (4), whole
brain (5). (C) Posterior cingulate cortex
(6). (D) Inferior parietal lobe (7). Left
medial temporal lobe is shown magnified
in A to highlight hippocampal (1) and
entorhinal cortex (2) ROIs.
ROI drawings are displayed
1780THE JOURNAL OF NUCLEAR MEDICINE • Vol. 47 • No. 11 • November 2006
author who was unaware of subject diagnosis drew all ROIs in
both hemispheres using MIDAS. The intrarater reliability for these
ROIs, as measured using intraclass correlation coefficients (ICCs),
ranges from ICC 5 0.92 for the STG to ICC 5 0.99 for the Hip
(P values , 0.001) (18,19,23,24). The MRI ROIs were used to
sample MRglc from the MRI coregistered PET images for the
following brain regions:
EC. The anterior boundary was 4 mm posterior to the fronto-
temporal junction, and the posterior boundary was the anterior
margin of the lateral geniculate body. The superior boundary in
both anterior and posterior sections was the dorsal and most
medial aspect of the parahippocampal gyrus (PHG). The inferior
boundary was the rhinal or collateral sulcus (18,19,24).
Hip. Drawings were performed along the full anterior–posterior
extent of the pes hippocampus and included a portion of the
subiculum. The lateral border was the temporal horn of the lateral
ventricle, and the medial border was the ambient cistern. The
inferior border was the white matter (WM) of the PHG. Most
anterior, the Hip was distinguished from the amygdaloid body by
fibers of WM interposed between these regions (18,19,23,24).
IPL. This region was drawn on all slices between the posterior
commissure and the slice immediately caudal to the posterior limit
of the splenium of the corpus callosum (25). At the surface,
the superior boundary was the postcentral sulcus and the inferior
boundary was the most superior branch of the lateral sulcus. This
ROI includes most of Brodmann area (BA) 40.
PCC. This region is bounded by the cingulate and callosal sulci
and extends laterally to include all gray matter of the cingulate
gyrus. Its posterior boundary coincides with the occipitoparietal
sulcus. This region includes BA 23 and the retrosplenial cortex
(BA 29 and BA 30), which was shown to be affected in mild
cognitive impairment (MCI) and mild AD (26).
STG. This region spans from the anterior EC to the posterior
crux of the fornix, is bound superiorly by the sylvan fissure and
inferiorly by the superior temporal sulcus (18,19), and corre-
sponds to BA 22.
Whole Brain (WB). AWB ROI was drawn on the MRI scans by
excluding all nonbrain voxels—that is, scalp tissue, skull and dural
venous sinus, as well as CSF-containing voxels.
Ventricles. Ventricular volume was measured with a 3D ROI
constructed to span the ventricular CSF that excluded subarach-
noid CSF on MRI.
Reference Regions. The MRI ROI volumes were corrected for
between-subject variations in head size (HS) by using the volume
of the intradural supratentorial volume (19). The18F-FDG PET
ROI MRglc data were corrected for variations in the global MRglc
using pons MRglc, which was reported as the brain region least
metabolically affected in AD (27). Pons MRglc was sampled at
the center of a midpontine slice at the level of the middle cerebral
peduncles with an 8 · 8 mm box (19,23).
Qualitative Evaluations. All scans were visually inspected by 2
raters for the presence of significant atrophy on MRI and hypo-
18F-FDG PET using published protocols with
known intra- and interrater reliabilities (24,28,29). Each
FDG PET and MRI scan was independently rated by both
observers and the final diagnosis was made by joint agreement.
Quantitative Analysis. Statistical analyses were done using
SPSS 12.0 (SPSS Inc.). Ratios were created for all ROIs by
dividing the MRI volume measures by the HS and by dividing the
18F-FDG PET MRglc measures by pons MRglc.18F-FDG PET
analyses were done with and without atrophy correction. Descrip-
tive statistics included the arithmetic mean (SD) and the co-
efficient of variation (%CV 5 [SD/mean] ? 100). The statistical
significance of group differences was tested for each neuropsy-
chologic variable and ROI measure using the Mann–Whitney rank
sum test (a 5 0.05, 1-sided, exact inference) (30). Whereas the
MRI and18F-FDG PET scans of the FAD subjects were acquired
on the same MRI or PET scanner, those of the control group were
acquired using different scanners. Although the scanners had
comparable resolution, the Mann–Whitney test was used to
examine whether scanner effects were present within the control
group before other statistical analyses. Neither the MRI volumes
nor the MRglc values showed scanner effects in any ROIs (P . 1).
Cohen’s d tests (31) were used to calculate effect sizes (ES) for
all ROIs and to compare the capability of MRI volumes and PET
MRglc measures in detecting group differences. An ES of 0
indicates complete overlap of the 2 groups, whereas ES $ 1 are
considered significant, with an ES of 1 indicating 55.4% of non-
overlap between groups. Logistic regressions with cross-validation
(leave-one-out classification) were used to assess whether group
membership could be predicted using MRI volumes and18F-FDG
PET MRglc measures.
Clinical and neuropsychologic data of the individual
FAD subjects and the control group are found in Table 1.
Two FAD subjects from different families carrying
Leu392Val (FAD-2) and Met146Leu (FAD-3) mutations
were examined 1 y before the mean age of onset for each
respective pedigree, and the other 5 FAD subjects were
examined, on average, 13 6 9 y before the anticipated
mean age at onset (range, 2 y for Leu392Val mutation to
27 y for Cys92Ser mutation) (7,9). Two FAD subjects
(FAD-2 and FAD-5) were heterozygous for the ApoE E4
allele. Five FAD subjects had MMSE scores of 30 of 30,
and 2 individuals scored 27 of 30 (FAD-5) and 25 of 30
(FAD-6), respectively. The FAD subjects scored in the
normal range of all neuropsychologic tests administered
(Table 1) except on the delayed recall of Rey’s complex
figure (P 5 0.01).
Brain Imaging Data
Qualitative Analysis. We found 100% interrater agree-
ment with respect to the18F-FDG PET scans. With respect
to the MRI scans, the 2 raters disagreed only on the MRI of
subject FAD-6, whose MRI was deemed to be normal for
age, as no global atrophy or ventricular enlargement was
evident, whereas the other observer judged the MRI to show
mild atrophy restricted to the posterior cortical regions. The
subject was diagnosed as showing mild posterior brain
atrophy by joint agreement.
None of the subjects presented with vascular or signif-
icant WM lesions. On visual examination, none of the FAD
subjects showed cortical and MTL atrophy on MRI, except
for subject FAD-6—with the lowest MMSE scores and 49 y
of age—who showed mild atrophy localized in the superior
PRECLINICAL HYPOMETABOLISM IN AD • Mosconi et al.1781
parietal lobe/precuneus (Table 1; Fig. 2D). Interestingly, the
FAD subject with questionable impairment (FAD-3) did not
show cortical or MTL atrophy on MRI, whereas parieto-
temporal and MTL hypometabolism was evident on PET
(Fig. 2C). Cortical hypometabolism on18F-FDG PET was
observed in all FAD subjects, primarily involving the
parietal and temporal regions. All FAD subjects showed
mild-to-severe MTL hypometabolism (24). None of the
control subjects showed abnormalities on18F-FDG PET
visual inspection. Figure 2 shows the18F-FDG PET and
MRI scans of 3 FAD subjects who presented with regional
hypometabolism in the absence of structural atrophy.
Quantitative Analysis. MRI volume data are found in
Table 2]. There was no HS difference between FAD and
control subjects. The ROI volume study showed that, com-
pared with controls, the FAD subjects showed volume re-
ductions only in the IPL (left: 18%, P 5 0.017; right: 19%,
P 5 0.007). The FAD group did not show global atrophy
(P 5 0.26, not significant [NS]) and ventricular enlarge-
ment (P 5 0.46, NS) compared with control subjects.
Table 2 presents18F-FDG PET data. Atrophy correction
increased regional MRglc by 0%–10% in control subjects
and 2%–12% in FAD subjects, with the greatest adjustment
seen in the EC for both groups. Non-atrophy–corrected
MRglc data are found in Table 2. The following results are
restricted to the atrophy-corrected MRglc data. No differ-
ence was found for pons MRglc between FAD and control
subjects. Compared with controls, the FAD subjects showed
MRglc reductions in the WB (13%, P 5 0.017) and bi-
laterally in the IPL (left: 16%, P 5 0.026; right: 17%, P 5
0.038) and STG (left: 11%, P 5 0.044; right: 13%, P 5
0.039). Unilateral MRglc reductions were found in the left
hemisphere for the EC (21%, P 5 0.017), PCC (20%, P 5
0.007), and Hip (12%, P 5 0.039) (Table 2). As shown in
Figure 3, each FAD subject showed consistently reduced
MRglc values compared with the respective age-matched
MRI and18F-FDG PET effect sizes were examined. PET
MRglc measures tended to be less variable than MRI vol-
ume measures, yielding CV (%) values in the FAD group
of ,10% for all ROIs except left EC (12%) and left Hip
(11%). The CV (%) values for the MRI measures were .10%
in all ROIs.
As shown in Figure 4, the PET MRglc measures consis-
tently separated FAD and control subjects and resulted in
large ES for all ROIs (d values, 1.1–2.2). The largest ES
was observed in the PCC, reflecting the highest group
separation resulting from lower interindividual variability.
scans of 4 subjects. (A) Control subject,
male, age 5 40 y, MMSE 5 30, CDR 5 0.
(B) FAD, female, age 5 35 y, MMSE 5
30, CDR 5 0 (Table 1, FAD-1). Bilateral
hypometabolism of parietal cortex and
MTL is evident on PET in absence of
atrophy on MRI. Hypometabolism is
more severe in left hemisphere. (C)
FAD, male, age 5 41 y, MMSE 5 30,
CDR 5 0.5 (Table 1, FAD-3). Bilateral
hypometabolism of parietal and temporal
cortices and of MTL is evident on PET in
absence of atrophy on MRI. MTL hypo-
metabolism is more severe in left hemi-
sphere. (D) FAD, female, age 5 47 y,
MMSE 5 25, CDR 5 0 (Table 1, FAD-6).
Hypometabolism of parietal regions and
MTL (bilaterally) and of left temporal lobe
is evident on PET. Mild atrophy is present
on MRI in parietal/precuneus regions.
Within these regions, atrophy correction
increased the MRglc measures by 10%,
which were still 1 SD below control sub-
jects’ mean. Coregistered PET and MRI
scans are shown at high (top row), middle
(middle row), and low (bottom row) trans-
axial levels. PET scans are displayed in a
blue-to-red color-coded scale, with in-
tensity in each pixel representing radio-
active counts per second.18F-FDG PET
images are shown using the same color
scale. Areas of regional hypometabolism
on PET are indicated by arrows on first
slice showing abnormalities.
1782THE JOURNAL OF NUCLEAR MEDICINE • Vol. 47 • No. 11 • November 2006
Significant ES for the MRI measures were restricted to the
IPL volumes (left: d 5 1.9, right: d 5 2.0), which were
comparable with the PET IPL MRglc data.
MRI and18F-FDG PET classification of sensitivity, spec-
ificity, and accuracy estimates is given in Table 3.
(STG, P 5 0.85, NS) to 57% (PCC and EC, P 5 0.59, NS),
clinical groups with 86% sensitivity (6/7 FAD correctly
identified) and 86% specificity (6/7 controls correctly iden-
tified) (86% accuracy, x125 8.20, P 5 0.004).
The PET MRglc measures, mostly in the left hemisphere,
proved to be sensitive group discriminators. The best group
separation was achieved by the left PCC MRglc, which
averaged across hemispheres. Compared with respective con-
trol subjects, all FAD subjects show reduced MRglc values. Two
FAD subjectswith MMSE scoresof 27 of 30 (FAD-5) and 25 of 30
(FAD-6) are marked with a white circle and a white diamond,
respectively, and 1 FAD subject with questionable impairment
(FAD-3) is marked with a white triangle. Note that these subjects
were not driving statistical effects. HIP 5 hippocampus.
18F-FDG PET MRglc data for age-, sex-, and
volumes (white) with PET MRglc measures (hatched).
ES were examined for all ROIs, comparing MRI
MRI and18F-FDG PET Data
Parameter Control FADControlFAD
Ventricular CSF (mL)
1.31 (0.22) [1.45 (0.18)]
1.23 (0.14) [1.30 (0.06)]
1.02 (0.14) [1.14 (0.14)]*
1.13 (0.14) [1.23 (0.05)]
1.36 (0.06) [1.38 (0.06)]
1.32 (0.07) [1.33 (0.07)]
1.22 (0.13) [1.24 (0.13)]*
1.19 (0.14) [1.21 (0.15)]
1.84 (0.31) [1.99 (0.25)]
1.87 (0.32) [1.99 (0.31)]
1.56 (0.13) [1.67 (0.13)]*
1.55 (0.12) [1.66 (0.12)]*
2.26 (0.23) [2.28 (0.23)]
2.18 (0.12) [2.22 (0.34)]
1.75 (0.16) [1.83 (0.18)]y
1.79 (0.19) [1.86 (0.18)]
1.64 (0.19) [1.76 (0.20)]
1.69 (0.18) [1.83 (0.21)]
1.66 (0.14) [1.78 (0.15)]
1,121 (427) [1,137 (439)]
1.43 (0.13) [1.57 (0.16)]*
1.47 (0.09) [1.62 (0.11)]*
1.43 (0.12) [1.55 (0.13)]*
1,239 (420) [1,245 (342)]
*P # 0.05, FAD , control subjects.
yP # 0.01, FAD , control subjects.
MRI measures 5 (ROI/HS volumes) · 100;18F-FDG PET measures 5 ROI/pons MRglc; MRglc 5 glucose metabolism (overall count rate).
Values are means (SD) [atrophy-corrected MRglc values].
PRECLINICAL HYPOMETABOLISM IN AD • Mosconi et al.1783
yielded 100% classification accuracy, reflecting no overlap
between FAD and control subjects (x125 19.5, P , 0.001).
The other ROIs had accuracies ranging from 64% in the left
STG (x125 5.28, P 5 0.022) to 86% in the left IPL (x125
5.53, P 5 0.019).
This study shows metabolic deficits on18F-FDG PET in
a group of presymptomatic individuals from early-onset
FAD families carrying mutations in the PS-1 gene, which
were evident in the absence of structural abnormalities on
MRI and remained significant after atrophy correction.
Compared with age-matched healthy control subjects, FAD
individuals showed MRglc reductions in the WB and in all
brain regions examined, which were more consistent in
the left hemisphere. These data suggest that the global
MRglc reduction usually observed in AD patients (21,22)
occurs early in the course of the disease, with the regional
temporoparietal, PCC, and MTL hypometabolism being
present on a background of widespread global MRglc
impairment. These findings are consistent with previous
18F-FDG PET reports of widespread cortical hypometabo-
lism in FAD members of a British pedigree with mutations
in the amyloid precursor protein (APP) gene or in the PS-
1 gene (5,6). Parietotemporal hypometabolism was ob-
served in all FAD subjects who also showed moderate
atrophy within the same brain regions, possibly reflecting
more advanced disease (5,6). Atrophy correction of the18F-
FDG PET data was not performed in these studies (5,6).
To clarify the brain changes implicated in disease onset,
it is important to determine the extent to which hypome-
tabolism measured with
atrophy. The presence of brain atrophy artificially lowers
PET measures because of the partial-volume effects of
CSF, and the resulting MRglc measures reflect the com-
bined effects of hypometabolism and atrophy. The present
study shows that measures of MRglc per unit brain volume
(i.e., as obtained with atrophy correction) are reduced in
FAD, providing realistic evidence for tissue MRglc impair-
ment. These data are consistent with PET studies showing a
relative independence of hypometabolism from atrophy in
MCI and AD patients (17,19,23).
We used a 2-segment (brain tissue vs. CSF) partial-
volume correction (PVC) technique (17). This method cor-
rects PET measures for the partial-volume effects of the
CSF pool but does not take into account possible differ-
ences in18F-FDG uptake between gray and WM regions.
Alternatively, a 3-segment (gray matter vs. WM vs. CSF)
PVC (32) may be used to correct for ‘‘spill-out’’ of activity
from gray matter as well as ‘‘spill-in’’ of WM activity into
gray matter pixels, which is the preferred choice when tracer
uptake is heterogeneous depending on the brain tissue
involved, such as in receptor studies. Although this method
might be of value also in18F-FDG studies, we chose to
use the 2-segment PVC because our subjects showed only
minimal atrophy, and the 3-segment procedure is highly sus-
ceptible to segmentation and coregistration inaccuracies—
particularly in small brain structures, where the likelihood
of tissue misclassification is higher.
In contrast to previous studies of FAD (3–6), our FAD
subjects showed minimal, if not absent, atrophy on MRI, as
underlined by the lack of ventricular enlargement compared
with control subjects. At-risk FAD subjects included in pre-
vious publications were examined 1–3 y before they devel-
oped symptoms of AD (3–6), whereas our FAD subjects
were examined over a broader range of 1–27 y before the
anticipated age at onset. The fact that hypometabolism was
found in all subjects, irrespective of age, suggests that
MRglc reductions may be an early sign of future AD. It is
well established that substantial hypometabolism is evident
18F-FDG PET by the time a patient presents with
symptoms of dementia (29,33). The presence of severe
hypometabolism in our FAD subjects, who were clinically
asymptomatic, suggests that18F-FDG PET evaluations may
support the clinical diagnosis when cognitive deficits are
still minimal and difficult to interpret.
Diagnostically, we found that PCC MRglc provided
complete group separation and was positively correlated
with the time to the expected age of decline in the FAD
group (r 5 0.78, P 5 0.005; Fig. 5), with those subjects
closer to the expected age of AD onset having lower
MRglc. Although the anticipated age of onset for a given
pedigree is known, it does not predict exactly when FAD
subjects will develop the clinical symptoms of AD, which
may vary according to the individual compliance against
advancing pathology. Nonetheless, these data show a po-
tential for using18F-FDG PET measures in the preclinical
18F-FDG PET is due to brain
Diagnostic Value of MRI Volumes and18F-FDG PET
MRglc Measures in Discriminating Between FAD and
MeasureA SS SPPA SSSPP
A 5 % accuracy; SS 5 % sensitivity; SP 5 % specificity.
1784THE JOURNAL OF NUCLEAR MEDICINE • Vol. 47 • No. 11 • November 2006
diagnosis of AD to identify individuals at risk for future
decline based on the extent of the metabolic reductions.
The pattern of hypometabolism observed in our FAD
subjects is consistent with reports on presymptomatic in-
dividual carriers of the ApoE E4 allele, a major suscepti-
bility gene for late-onset AD. Compared with noncarriers,
E4 carriers typically have reduced MRglc in the same
regions as clinically affected AD patients, with the PCC
showing the most prominent reductions (34,35). It has been
suggested that the AD-like regional pattern of hypometab-
olism found in E4 carriers may explain the added genetic
risk for AD. Likewise, longitudinal18F-FDG PET studies
show that excess hypometabolism in the PCC and parieto-
temporal regions in MCI patients predicts future decline to
AD (22). In addition, the present data show MRglc reduc-
tions in the Hip and EC in presymptomatic FAD subjects.
These data underline the importance of MTL evaluations
with PET in the early detection of AD and are consistent
with previous18F-FDG PET and MRI studies in sporadic
AD showing that excess MTL hypometabolism and atrophy
are risk factors for declining from normal aging to MCI
The larger ES found for PET compared with MRI mea-
sures show that18F-FDG PET may have stronger discrim-
ination capacity than MRI in the preclinical AD stages.
The diagnostic advantage of PET MRglc over MRI volume
measurements found in this study derives from the higher
percentage of MRglc reductions and the smaller inter-
subject variability obtained with PET. As reduced MRglc
reflects synaptic dysfunction, neuronal damage is also likely
to be under way, yet not sufficient to result in gross atrophy
detectable with MRI.
From a methodologic point of view, we chose to use
manually defined MRI-guided ROIs instead of automated
voxel-based analysis (VBA) techniques. Several VBA tools
have been developed for analysis of brain images to provide
examination of statistical effects on a voxel-by-voxel basis
(38,39), enabling automated and time-saving assessment of
statistical effects. Nonetheless, the MRI-guided ROI data
remain the gold standard for PET sampling because of the
superior anatomic precision, especially in aging and neu-
rodegenerative diseases, where the brains undergo major
structural changes (23). Our primary goal was to compare
atrophy with hypometabolism within the same brain region,
with the purpose of clarifying whether MRglc reductions
can be detected before overt atrophy on MRI. This cannot
be done with VBA, where the brain regions showing sta-
tistical effects are a posteriori and functionally, instead of
anatomically, defined. Moreover, the use of preprocessing
procedures (i.e., spatial normalization and smoothing) may
obscure the detection of abnormalities in small brain struc-
tures such as the Hip and EC (23). On the other hand, VBA
techniques have proven to be sensitive to detect subtle struc-
tural brain changes (4,36)—particularly longitudinally—
and these analyses would be of interest to monitor the
transition from a presymptomatic to a symptomatic stage.
Follow-up evaluations of our FAD subjects are neces-
sary to confirm that the observed MRglc abnormalities are
forerunning subsequent development of symptoms. How-
ever, the observed pattern of hypometabolism is consistent
with AD, which makes it reasonable to hypothesize that
the observed brain changes relate to a presymptomatic AD
stage. Second, findings from FAD subjects may not reflect
the natural history of the more common late-onset sporadic
AD. Although there is evidence that FAD subjects present
with brain pathology similar to that of sporadic AD (2),
FAD patients become symptomatic at younger ages than
sporadic AD (onset age, usually .70 y), suggesting that
brain deterioration may be more aggressive in FAD. Lastly,
FAD individuals belonged to different families and, al-
though all expressed mutations in the PS-1 gene, they
carried mutations at 3 different codons. Despite genotypic
variability, all FAD subjects shared the same pattern of
metabolic abnormalities on18F-FDG PET in the relative
absence of structural damage on MRI, revealing a relatively
homogeneous brain profile.
Presymptomatic FAD individuals show widespread
MRglc reductions consistent with the expected AD PET
pattern in the absence of severe atrophy on MRI. These
results further suggest that PET MRglc measures have a
potential as preclinical biomarkers for dementia and for
tracking disease progression.
This work was supported by National Institutes of
Health, National Institute on Aging grants AG12101,
AG13613, AG08051, and AG022374 and National Cancer
for Research Resources grant MO1RR0096. The authors
have reported no conflicts of interest.
expected age at AD onset in FAD group (r 5 0.78, P 5 0.005).
Correlation between PCC MRglc and time to
PRECLINICAL HYPOMETABOLISM IN AD • Mosconi et al.1785
REFERENCES Download full-text
1. Petersen RC, Smith GE, Waring SC, Ivnik RJ, Tangalos EG, Kokmen E. Mild
cognitive impairment: clinical characterization and outcome. Arch Neurol.
2. Tanzi R, Bertram L. New frontiers in Alzheimer’s disease genetics. Neuron.
3. Fox NC, Warrington EK, Rossor MN. Serial magnetic resonance imaging of
cerebral atrophy in preclinical Alzheimer’s disease. Lancet. 1999;353:2125.
4. Fox NC, Crum WR, Scahill RI, Stevens JM, Janssen JC, Rossor MN. Imaging of
onset and progression of Alzheimer’s disease with voxel-compression mapping
of serial magnetic resonance images. Lancet. 2001;358:201–205.
5. Kennedy AM, Newman SK, Frackowiak RS, et al. Chromosome 14 linked
familial Alzheimer’s disease: a clinico-pathological study of a single pedigree.
6. Kennedy AM, Frackowiak RSJ, Newman SK, et al. Deficits in cerebral glucose
metabolism demonstrated by positron emission tomography in individuals at risk
of familial Alzheimer’s disease. Neurosci Lett. 1995;186:17–20.
7. Sorbi S, Nacmias B, Forleo P, et al. Missense mutation of S182 gene in Italian
families with early-onset Alzheimer’s disease. Lancet. 1995;346:439–440.
8. Rogaev EI, Sherrington R, Rogaeva EA, et al. Familial Alzheimer’s disease in
kindreds with missense mutations in a gene on chromosome 1 related to the
Alzheimer’s disease type 3 gene. Nature. 1995;376:775–778.
9. Tedde A, Nacmias B, Ciantelli M, et al. Identification of new Presinilin gene muta-
10. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM.
Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA work
group under the auspices of Department of Health & Human Services Task Force
on Alzheimer’s disease. Neurology. 1984;34:939–944.
11. American Psychiatric Association. Diagnostic and Statistical Manual of Mental
Disorders. 4th ed. Washington, DC: American Psychiatric Association; 1994.
12. Morris JC, Ernesto C, Shaefer K, et.al. Clinical dementia rating (CDR) training
and reliability protocol: The Alzheimer Disease Cooperative Study Unit expe-
rience. Neurology. 1997;48:1508–1510.
13. Reisberg B, Ferris SH, de Leon MJ, Crook T. The global deterioration scale for as-
sessmentofprimary degenerative dementia. AmJ Psychiatry. 1982;139:1136–1139.
14. Hamilton M. Development of a rating scale for primary depression illness. Br J
Soc Clin Psychol. 1967;6:278–296.
15. Lezak MD. Neuropsychological Assessment. 3rd ed. New York, NY: Oxford
University Press; 1995.
16. Woods RP, Mazziotta JC, Cherry SR. MRI-PET registration with automated
algorithm. J Comput Assist Tomogr. 1993;17:536–546.
17. Meltzer CC, Zubieta JK, Brandt J, Tune LE, Mayberg HS, Frost JJ. Regional
hypometabolism in Alzheimer’s disease as measured by positron emission
tomography after correction for effects of partial volume averaging. Neurology.
18. de Leon MJ, Convit A, Wolf OT, et al. Prediction of cognitive decline in normal
elderly subjects with 2-[18F]fluoro-2-deoxy-D-glucose/positron-emission tomog-
raphy (FDG/PET). Proc Natl Acad Sci U S A. 2001;98:10966–10971.
19. De Santi S, de Leon MJ, Rusinek H, et al. Hippocampal formation glucose
20. Rusinek H, Chandra R. Accuracy of tissue volume determination from MRI: a
phantom study. Invest Radiol. 1993;28:890–895.
21. Silverman DHS. Brain18F-FDG PET in the diagnosis of neurodegenerative
dementias: comparison with perfusion SPECT and with clinical evaluations
lacking nuclear imaging. J Nucl Med. 2004;45:594–607.
22. Mosconi L. Brain glucose metabolism in the early and specific diagnosis of
Alzheimer’s disease. Eur J Nucl Med. 2005;32:486–510.
23. Mosconi L, Tsui WH, De Santi S, et al. Reduced hippocampal metabolism in
mild cognitive impairment and Alzheimer’s disease: automated FDG-PET image
analysis. Neurology. 2005;64:1860–1867.
24. Mosconi L, De Santi S, Li Y, et al. Visual rating of medial temporal lobe
metabolism in mild cognitive impairment and Alzheimer’s disease using FDG-
PET. Eur J Nucl Med Mol Imaging. 2006;33:210–221.
25. Raz N, Gunning FM, Head D, et al. Selective aging of the human cerebral cortex
observed in vivo: differential vulnerability of the prefrontal gray matter. Cereb
26. Nestor PJ, Fryer TD, Smielewski P, Hodges JR. Limbic hypometabolism in
Alzheimer’s disease and mild cognitive impairment. Ann Neurol. 2003;54:
27. Minoshima S, Frey KA, Foster NL, Kuhl DE. Preserved pontine glucose
metabolism in Alzheimer’s disease: a reference region for functional brain image
(PET) analysis. J Comput Assist Tomogr. 1995;19:541–547.
28. Scheltens P, Barkhof F, Leys D, et al. A semiquantative rating scale for the
assessment of signal hyperintensities on magnetic resonance imaging. J Neurol
29. Silverman DHS, Small GW, Chang CY, et al. Positron emission tomography in
evaluation of dementia: regional brain metabolism and long-term outcome.
30. Hollander M, Wolfe D. Nonparametric Statistical Methods. New York, NY: John
Wiley and Sons; 1988.
31. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed.
Hillsdale, NJ: Lawrence Erlbaum Associates; 1988.
32. Muller-Gartner HW, Links JM, Prince JL, et al. Measurement of radiotracer
concentration in brain gray matter using positron emission tomography: MRI-
based correction for partial volume effects. J Cereb Blood Flow Metab.
33. Silverman DHS, Truong CT, Kim SK, et al. Prognostic value of regional cerebral
metabolism in patients undergoing dementia evaluation: comparison to a quan-
tifying parameter of subsequent cognitive performance and to prognostic assess-
ment without PET. Mol Genet Metab. 2003;80:350–355.
34. Small GW, Mazziotta JC, Collins MT, et al. Apolipoprotein E type 4 allele and
cerebral glucose metabolism in relatives at risk for familial Alzheimer disease.
35. Reiman EM, Caselli RJ, Chen K, Alexander GE, Bandy D, Frost J. Declining
brain activity in cognitively normal apolipoprotein E epsilon 4 heterozygotes:
a foundation for using positron emission tomography to efficiently test
treatments to prevent Alzheimer’s disease. Proc Natl Acad Sci U S A. 2001;98:
36. Rusinek H, De Santi S, Frid D, et al. Regional brain atrophy rate predicts future
cognitive decline: 6-year longitudinal MR imaging study of normal aging.
37. Jack CR Jr, Shiung MM, Gunter JL, et al. Comparison of different MRI brain
atrophy rate measures with clinical disease progression in AD. Neurology. 2004;
38. Ashburner J, Csernansky JG, Davatzikos C, Fox NC, Frisoni GB, Thompson PM.
Computer-assisted imaging to assess brain structure in healthy and diseased
brains. Lancet Neurol. 2003;2:79–88.
39. Minoshima S, Frey KA, Koeppe RA, Foster NL, Kuhl DE. A diagnostic ap-
proach in Alzheimer’s disease using three-dimensional stereotactic surface
projections of fluorine-18-FDG PET. J Nucl Med. 1995;36:1238–1248.
1786THE JOURNAL OF NUCLEAR MEDICINE • Vol. 47 • No. 11 • November 2006