Alzheimer-signature MRI biomarker predicts AD dementia in cognitively normal adults
Since Alzheimer disease (AD) neuropathology is thought to develop years before dementia, it may be possible to detect subtle AD-related atrophy in preclinical AD. Here we hypothesized that the "disease signature" of AD-related cortical thinning, previously identified in patients with mild AD dementia, would be useful as a biomarker to detect anatomic abnormalities consistent with AD in cognitively normal (CN) adults who develop AD dementia after longitudinal follow-up. We studied 2 independent samples of adults who were CN when scanned. In sample 1, 8 individuals developing AD dementia (CN-AD converters) after an average of 11.1 years were compared to 25 individuals who remained CN (CN-stable). In sample 2, 7 CN-AD converters (average follow-up 7.1 years) were compared to 25 CN-stable individuals. AD-signature cortical thinning in CN-AD converters in both samples was remarkably similar, about 0.2 mm (p < 0.05). Despite this small absolute difference, Cohen d effect sizes for these differences were very large (> 1). Of the 11 CN individuals with baseline low AD-signature thickness (≥ 1 SD below cohort mean), 55% developed AD dementia over nearly the next decade, while none of the 9 high AD-signature thickness individuals (≥ 1 SD above mean) developed dementia. This marker predicted time to diagnosis of dementia (hazard ratio = 3.4, p < 0.0005); 1 SD of thinning increased dementia risk by 3.4. By focusing on cortical regions known to be affected in AD dementia, subtle but reliable atrophy is identifiable in asymptomatic individuals nearly a decade before dementia, making this measure a potentially important imaging biomarker of early neurodegeneration.
Alzheimer-signature MRI biomarker
predicts AD dementia in cognitively
B.C. Dickerson, MD
T.R. Stoub, PhD
R.C. Shah, MD
R.A. Sperling, MD
R.J. Killiany, PhD
M.S. Albert, PhD
B.T. Hyman, MD, PhD
D. Blacker, MD, ScD
Objective: Since Alzheimer disease (AD) neuropathology is thought to develop years before dementia,
it may be possible to detect subtle AD-related atrophy in preclinical AD. Here we hypothesized that
the “disease signature” of AD-related cortical thinning, previously identified in patients with mild AD
dementia, would be useful as a biomarker to detect anatomic abnormalities consistent with AD in
cognitively normal (CN) adults who develop AD dementia after longitudinal follow-up.
Methods: We studied 2 independent samples of adults who were CN when scanned. In sample 1,
8 individuals developing AD dementia (CN-AD converters) after an average of 11.1 years were
compared to 25 individuals who remained CN (CN-stable). In sample 2, 7 CN-AD converters (av-
erage follow-up 7.1 years) were compared to 25 CN-stable individuals.
Results: AD-signature cortical thinning in CN-AD converters in both samples was remarkably sim-
ilar, about 0.2 mm (p ⬍ 0.05). Despite this small absolute difference, Cohen d effect sizes for
these differences were very large (⬎1). Of the 11 CN individuals with baseline low AD-signature
thickness (ⱖ1 SD below cohort mean), 55% developed AD dementia over nearly the next decade,
while none of the 9 high AD-signature thickness individuals (ⱖ1 SD above mean) developed de-
mentia. This marker predicted time to diagnosis of dementia (hazard ratio ⫽ 3.4, p ⬍ 0.0005); 1
SD of thinning increased dementia risk by 3.4.
Conclusions: By focusing on cortical regions known to be affected in AD dementia, subtle but
reliable atrophy is identifiable in asymptomatic individuals nearly a decade before dementia, mak-
ing this measure a potentially important imaging biomarker of early neurodegeneration.
AD ⫽ Alzheimer disease; CI ⫽ confidence interval; CN ⫽ cognitively normal; HR ⫽ hazard ratio; MCI ⫽ mild cognitive impair-
ment; MGH ⫽ Massachusetts General Hospital; MMSE ⫽ Mini-Mental State Examination; MTL ⫽ medial temporal lobe; ROI ⫽
region of interest.
The dementia of Alzheimer disease (AD) develops insidiously, initially emerging as mild cog-
nitive impairment before eventually progressing to rob the individual of independent function.
Ten or more years may elapse from first symptoms until clinical dementia.
The evolution of
AD neuropathology is longer, developing in cognitively normal (CN) adults and causing dys-
function and cell death in neuronal systems subserving cognition, eventually leading to the
Cognitive test measures may change 5–10 years before AD dementia,
and may be useful
for predicting dementia in CN cohorts.
A few imaging studies have reported subtle brain
abnormalities, particularly in medial temporal lobe and parietal regions, in groups of CN
e-Pub ahead of print on April 13, 2011, at www.neurology.org.
From the Departments of Neurology (B.C.D., R.A.S., B.T.H.) and Psychiatry (D.B.), Massachusetts Alzheimer’s Disease Research Center (B.C.D.,
R.A.S., B.T.H., D.B.), and Athinoula A. Martinos Center for Biomedical Imaging (B.C.D., R.A.S.), Massachusetts General Hospital and Harvard
Medical School, Boston; Department of Neurological Sciences (T.R.S., L.d.-M.) and Department of Family Medicine, Rush Alzheimer’s Disease
Center (R.C.S.), Rush University Medical Center, Chicago, IL; Center for Alzheimer Research and Treatment (R.A.S.), Department of Neurology,
Brigham & Women’s Hospital, Boston; Department of Anatomy and Neurobiology (R.J.K.), Boston University School of Medicine, Boston, MA;
and Department of Neurology (M.S.A.), Johns Hopkins University School of Medicine, Baltimore, MD.
Study funding: Supported by the NIH (NIA R01-AG29411, R21-AG29840, P50-AG005134, P01-AG09466, P01-AG14449, P30-AG10161, R01-
AG17917, and R01-AG10688, and NCRR P41-RR14075 and U24-RR021382), the Alzheimer’s Association, the Mental Illness and Neuroscience
Discovery (MIND) Institute, and the Illinois Department of Public Health.
Disclosure: Author disclosures are provided at the end of the article.
Supplemental data at
Address correspondence and
reprint requests to Dr. Brad
Frontotemporal Dementia Unit
and Gerontology Research Unit,
149 13th St., Suite 2691,
Charlestown, MA 02129
Copyright © 2011 by AAN Enterprises, Inc. 1395
individuals who subsequently declined.
date, there has been very little investigation of
neuroimaging measures in CN subjects to
predict AD dementia.
We undertook this study hypothesizing that
neuroanatomic abnormalities consistent with
very early AD pathology could be detected in
CN individuals followed longitudinally and
eventually diagnosed with AD dementia. Work-
ing with 2 independent samples with nearly
identical demographic and cognitive characteris-
tics, we investigated whether these measures
could be translated into an AD–dementia risk
MRI biomarker applicable to CN individuals.
We employed a hypothesis-driven approach us-
ing regions of interest (ROIs) collectively called
the cortical signature of AD based on our earlier
comparing patients with mild AD de-
mentia to CN controls. After examining the 2
samples separately using these a priori ROIs, we
then pooled them to obtain estimates of this mea-
sure’s utility in risk assessment, including survival
models predicting time to AD dementia.
METHODS The 2 samples’ characteristics are described briefly;
details are provided in appendix e-1 on the Neurology
Web site at
www.neurology.org. Sample 1 (Massachusetts General Hospital
[MGH]) included community volunteers participating in a longitu-
dinal study. The participants in sample 2 (Rush) were recruited for a
longitudinal imaging project
from the community by the Rush
AD Center, the Religious Order Study, or the Rush Memory and
Aging Project. In both studies, subjects were excluded if they had
significant active medical, neurologic, or psychiatric illness, or major
vascular risk factors or disease (i.e., atrial fibrillation, insulin-
dependent diabetes mellitus, cerebral infarcts, cardiac bypass graft
surgery). At baseline, both samples underwent comprehensive clini-
cal evaluations, neuropsychological testing, and MRI scans. Com-
posite episodic memory Z scores were computed from multiple
neuropsychological memory measures. All subjects underwent an-
nual clinical evaluations and were determined to be CN, to be
mildly impaired, or to have dementia. For the present analyses, we
included all subjects in the original cohorts with a baseline CN sta-
tus and at least 4 annual follow-up visits (1 Rush subject had 3
follow-up visits) after MRI. Because our hypotheses were focused on
the presence of subtle neurodegenerative change in preclinical AD,
we restricted the sample to individuals who remained CN at most
recent follow-up (CN-stable) or those diagnosed with probable AD
dementia (CN-AD converter), excluding those diagnosed with mild
cognitive impairment (MCI, since their longer-term outcome is not
yet known) or non-AD dementia. This resulted in sample sizes of 33
(MGH: 25 CN-stable, 8 CN-AD converters) and 32 (Rush: 25
CN-stable, 7 CN-AD converters). Nearly all CN-AD converters in
both samples were evaluated at an intermediate time when their
clinical status was MCI; 2 individuals in the Rush sample changed
from CN to AD dementia without having been seen during that
MRI data were acquired very similarly for both samples us-
ing a General Electric (Milwaukee, WI) 1.5-T scanner and a
3-dimensional T1-weighted spoiled gradient recalled echo pulse
sequence. The MRI parameters and analysis methods are de-
scribed in previous publications and appendix e-1. We used an a
priori set of 9 ROIs derived from a previous analysis
e-1) and a primary visual cortex ROI as a control region. ROIs
were mapped to each individual and thickness values were ob-
tained; a single measure was derived from the average thickness
of all 9 ROIs, the AD-signature summary measure.
Group statistical comparisons were performed using analysis of
variance with post hoc pairwise comparisons for continuous mea-
for proportions (SPSS 16.0, Chicago, IL). Potential con-
founding factors, including age, gender, and education, were used as
covariates. Separate univariate Cox regression models were con-
structed to investigate the relationship of covariates, memory Z
scores, and AD-signature cortical thickness summary measures to
the likelihood of progression from CN to AD dementia. A multivar-
iate Cox regression model was then constructed including indepen-
dent variables that reached a trend-level effect ( p ⬍ 0.1) in the
univariate analyses ( p value-to-enter ⬍0.05).
Standard protocol approvals, registrations, and patient
Each participant gave written informed consent in
accordance with institutional Human Subjects Research Com-
Clinical characteristics. The 2 samples
were remarkably similar (table 1). Baseline Mini-
Mental State Examination (MMSE) was within 1 to 2
points of ceiling, and baseline episodic memory scores
were above or only slightly below the mean of large
normative groups. Over nearly a decade of follow-up,
CN-stable subjects remained clinically stable, an obser-
vation supported by unchanged MMSE and memory
scores. In contrast, CN-AD converters declined to AD
Table 1 Demographic and clinical characteristics of samples
(n ⴝ 25)
(n ⴝ 8)
(n ⴝ 25)
(n ⴝ 7)
Age, y 71.2 (4.0) 71.5 (2.1) 76.4 (6.0) 77.7 (4.6)
Male, n (%) 9 (36) 5 (63) 3 (12) 4 (57)
Education, y 14.9 (2.3) 14.4 (2.6) 15.6 (3.0) 15.3 (3.3)
APOE (% e4 carriers) 4 (16) 2 (25) 3 (12) 1 (14)
Baseline MMSE 29.3 (0.7) 28.9 (0.8) 29.1 (1.0) 28.0 (0.6)
Baseline episodic memory (Z) 0.29 (0.7) ⫺0.28 (1.1)
0.64 (0.5) 0.13 (0.5)
Follow-up, y 10.4 (3.1) 11.1 (2.5) 8.3 (3.1) 7.1 (1.1)
29.4 (0.8) 26.0 (2.9)
28.8 (2.1) 21.7 (2.7)
Follow-up episodic memory (Z)
0.12 (0.7) ⫺2.3 (1.4)
0.81 (0.5) ⫺1.27 (0.9)
Abbreviations: AD ⫽ Alzheimer disease; CN ⫽ cognitively normal; MGH ⫽ Massachusetts
General Hospital; MMSE ⫽ Mini-Mental State Examination.
Values are mean (SD) unless otherwise noted.
p ⬍ 0.01.
p ⬍ 0.1 (trend-level effect).
p ⬍ 0.05.
p ⬍ 0.001.
For CN-stable, these are the last available values; for CN-AD converter, these are the
values at the visit associated with time of diagnosis of AD dementia.
1396 Neurology 76 April 19, 2011
dementia, illustrated by substantial decrements in
MMSE and memory scores.
The trajectories of progressive decline are illustrated
in figure 1, which shows (top) memory performance
data, demonstrating the stability of CN-stable perfor-
mance compared to substantial declines in the CN-AD
converters, and (bottom) the trajectories of clinical mea-
sures for each individual CN-AD converter.
Regional cortical thinning in relation to future AD
In both samples, the CN-AD converters
showed thinner medial temporal lobe (MTL), tem-
poral pole, and superior frontal gyrus (p ⬍ 0.05)
than CN-stables. The MGH sample also demon-
strated thinning in the inferior parietal lobule (angu-
lar and supramarginal gyri) and middle frontal gyrus
(p ⬍ 0.05). Although not reaching statistical signifi-
cance in the Rush sample, these and other AD-
signature ROIs showed consistent effects in the
expected direction (table 2). In contrast, the primary
visual cortex did not show such a pattern (p ⬎ 0.1).
Finally, the AD signature thickness summary mea-
sure demonstrated strikingly similar effects, with the
CN-AD converters being 0.20 mm (MGH) or 0.19
mm (Rush) thinner than CN-Stable (figure 2, top). Co-
hen’s d effect sizes were 1.3 and 1.2, indicating very
large effects even though absolute magnitudes of group
Figure 1 Baseline and longitudinal clinical characteristics of the 2 samples
(A, B) Baseline and follow-up episodic memory Z scores from the 2 samples, illustrating the stability (or even slight improvement in sample 2) of perfor-
mance in the cognitively normal (CN)-stable groups over nearly a decade of follow-up. In contrast, the CN-Alzheimer disease (AD) converters performed
normally at baseline but declined substantially over nearly a decade of follow-up by the time of diagnosis of AD dementia. (C, D) Detailed illustrationof
cognitive decline in each of the CN-AD converter individuals. (C) Annual Clinical Dementia Rating (CDR)–sum of boxes scores increased from normal (0)at
baseline to mild dementia in each of the 8 individuals in the Massachusetts General Hospital sample. (D) Annual episodic memory Z scores declined from
normal to substantially impaired in each of the 7 individuals in the Rush sample. Breaks in lines indicate years for which data were not available.
Neurology 76 April 19, 2011 1397
differences were less than
millimeter. For compari-
son, the pooled-sample between-group effect size for
baseline Episodic Memory Z scores was 0.73. Table e-1
contains effect sizes for each ROI.
Since the 2 samples’ characteristics were so similar,
they were pooled for subsequent analyses. To obtain a
preliminary sense of whether the thickness measure
might be a useful predictor, we divided the sample into
3 subgroups based on AD-signature thickness: 1) ⱖ1
SD below the entire cohort mean (low thickness sub-
group), 2) within 1 SD of the mean (average), and 3)
ⱖ1 SD above the mean (high). Of the 11 CN individ-
uals with baseline low AD-signature thickness, 55% de-
veloped AD dementia over nearly the next decade. In
contrast, 20% of the 45 average AD-signature thickness
individuals developed AD dementia, while none of the
9 high AD-signature thickness individuals developed
dementia, a highly significant difference (
⫽ 8.3, p ⬍
0.005; figure 2, bottom). Figure e-1 presents a scatter-
plot illustrating these data.
In the 2 samples analyzed separately, low/average/
high AD-signature thickness was associated with
conversion-to-dementia rates of 67%/17%/0%
(MGH) and 40%/24%/0% (Rush). This replication
shows that low/average/high AD-signature thickness
indicates high/middle/low risk for AD dementia
nearly a decade in the future, with the pooled analysis
above providing the best estimates of the actual like-
lihood for each level of risk.
Prediction of time to dementia in asymptomatic older
Univariate Cox regression models of covari-
ates indicated that gender was a predictor of time to
dementia (greater risk for men, p ⬍ 0.005) but age,
education, APOE status, and MMSE were not (p ⬎
0.1), at least in this small sample. Univariate and
multivariate (with gender) Cox regression models
indicated that the AD-signature thickness measure
was a strong predictor of time to dementia (uni-
variate model: hazard ratio [HR] ⫽ 3.5 for 1 SD
decrease in thickness; 95% confidence interval
[CI] 2.0–6.4; p ⬍ 0.00005; multivariate model:
HR ⫽ 3.4 for 1 SD thickness decrease; 95% CI
1.7–6.9; p ⬍ 0.0005). Figure 3 presents a survival
plot illustrating these results. The mean time to a
diagnosis of AD dementia in the pooled sample
was 8 years (SD ⫽ 1.9).
In comparison, episodic memory Z score was also
predictive of conversion but at a slightly weaker level
than AD-signature thickness (univariate model:
HR ⫽ 3.1 for 1 SD decrease in memory perfor-
mance; 95% CI 1.6 – 6.1; p ⬍ 0.001; multivariate
model: HR ⫽ 3.2 for 1 SD decrease in performance;
95% CI 1.5–6.6; p ⬍ 0.001).
DISCUSSION Two decades of investigation have
led to a mature understanding of the brain regions
that are consistently atrophic in patients with AD
dementia. Here we looked further back in the trajec-
tory of AD and demonstrated that subtle neuroanat-
omic abnormalities are detectable in asymptomatic
individuals nearly a decade before they are diagnosed
with AD dementia, and are useful not only for assess-
ing risk of AD dementia but also for predicting time
to onset. Despite the small sample size, the striking
consistency of these findings in 2 independent sam-
ples supports their generalizability.
Although sophisticated tools are now available for
investigating changes in brain structure, function, mo-
lecular profiles, and behavior, the challenges of longitu-
dinal research have made it difficult to study the full
course of AD from preclinical to prodromal to demen-
tia. Partly because they are time-, labor-, and cost-
intensive, very few longitudinal studies have been
completed of CN individuals who are followed until a
clinical diagnosis of AD dementia. In one of the earliest
studies of the decline of elderly individuals without de-
mentia to AD dementia, smaller baseline hippocampal
volume was present,
as also observed in a more recent
Another investigation of CN in-
dividuals demonstrated that ventricular volume was
predictive of cognitive decline (11 converted to MCI, 2
to AD), but hippocampal, entorhinal, or whole brain
volume were not.
A voxel-based morphometry study found reduced
MTL and parietal gray matter in 23 CN individuals
who later developed MCI,
and another analysis of
the same sample using manual ROIs demonstrated
Table 2 Mean (SD) region of interest measures by subject group (in mm)
(n ⴝ 25)
(n ⴝ 8)
(n ⴝ 25)
(n ⴝ 7)
MTL 3.28 (0.38) 2.94 (0.43)
3.37 (0.23) 2.83 (0.39)
Temporal pole 3.07 (0.20) 2.77 (0.30)
3.02 (0.29) 2.73 (0.31)
Inferior temporal 2.48 (0.20) 2.39 (0.30) 2.89 (0.26) 2.70 (0.28)
Angular gyrus 2.48 (0.22) 2.30 (0.25)
2.50 (0.28) 2.41 (0.23)
Supramarginal gyrus 2.54 (0.17) 2.38 (0.20)
2.54 (0.26) 2.38 (0.13)
Superior parietal 2.17 (0.21) 2.04 (0.19) 2.05 (0.21) 2.09 (0.25)
Precuneus 2.41 (0.17) 2.25 (0.13) 2.46 (0.22) 2.38 (0.22)
Middle frontal 2.40 (0.16) 2.24 (0.11)
2.30 (0.15) 2.20 (0.18)
Superior frontal 2.58 (0.23) 2.29 (0.19)
2.68 (0.34) 2.40 (0.19)
Primary visual 1.59 (0.12) 1.55 (0.08) 1.56 (0.17) 1.68 (0.25)
AD-signature summary measure 2.49 (0.14) 2.35 (0.17)
2.65 (0.19) 2.46 (0.12)
Abbreviations: AD ⫽ Alzheimer disease; CN ⫽ cognitively normal; MGH ⫽ Massachusetts
General Hospital; MTL ⫽ medial temporal lobe.
p ⬍ 0.05.
p ⬍ 0.001.
p ⬍ 0.005.
1398 Neurology 76 April 19, 2011
Two other inves-
tigations found subtle shape abnormalities in hip-
pocampal subregions in CN individuals who
progressed to MCI.
Very recently, a group of 9 CN individuals har-
boring brain amyloid who progressed to very mild
AD dementia were shown to have lower baseline
parahippocampal volumes compared to a stable CN
We previously demonstrated that CN adults
with brain amyloid have subtle AD-signature cortical
thinning compared to CN adults without brain amy-
loid, but these individuals had not yet been followed
Finally, a few studies have identified MTL and
whole brain atrophy in the presymptomatic stage of
genetically determined early-onset familial AD, again
supporting the concept that atrophy, while subtle,
may be present for years before AD symptoms.
From the perspective of quantitative imaging bio-
markers, the present study adds to prior literature in
several ways. First, most prior work has employed
manual tracings of one or a few ROIs. While neuro-
made MTL structures a logical
place to focus when investigating brain regions in
which the earliest atrophy occurs in AD, recent imag-
ing research has highlighted ventrolateral temporal,
lateral parietal, and the posterior cingulate/precuneus
as also involved early in the disease course.
Here a relatively new technique for quantitative
neuroanatomic measurement (cortical thickness
analysis)—employed using a novel a priori approach
focusing on brain regions known to be consistently
affected early in AD— enabled the detection of sta-
tistically robust (large effect sizes due to small mea-
surement error), replicable group differences of very
subtle absolute magnitude (⬃
mm) in MRI scans
now considered previous generation technology.
Such observations highlight the value of returning to
older MRI datasets when new measurement technol-
ogy is developed.
More importantly, these measures can be effi-
ciently, reliably obtained from single individuals. A
determination of whether AD-signature thickness is
low, average, or high relative to similarly aged indi-
viduals appears valuable in assessing risk of future
AD dementia, a finding replicated across both sam-
ples. Furthermore, the Cox regression model results
are powerful in that they augment the risk assessment
with an estimate of the potential time to AD demen-
tia. Whether such an approach will prove useful in a
generalizable manner is an important research topic.
A critical element of studies aiming to identify
biomarkers of asymptomatic AD is the cognitive as-
sessment methodology. Contemporary study designs
include the use of questionnaires or careful struc-
tured interviews to ascertain the presence and sever-
ity of symptoms of declining memory and cognition
in daily life. Questions are asked of both the individ-
ual and usually a knowledgeable informant with re-
gard to whether the individual has experienced a
decline in his or her ability to remember the details of
recent events or conversations, items to purchase
while shopping, or how to navigate to familiar places.
Changes of this sort, particularly when corroborated
by an informant, are in some cases associated with
atrophy in neural circuits subserving memory
may potentially be predictors of future decline and
although reports are inconsistent.
Some individuals who do not endorse symptoms
of memory impairment in daily life may exhibit signs
Figure 2 Thinner cortex in regions associated with Alzheimer disease (AD) in
asymptomatic adults who eventually develop dementia
(A) For each of the 2 samples, the average cortical thickness of the AD-signature regions
was slightly thinner in the cognitively normal (CN)-AD converter group than the CN-stable
group at the time of the baseline MRI scan. The absolute difference in the means of the
groups for each sample is less than
mm, but because variance is small the Cohen d effect
size between the groups is very large for each sample (⬎1.2). Inset illustrates AD-signature
regions of interest. (B) Illustration of the risk of AD dementia as a function of AD-signature
thickness from baseline MRI scans in the pooled group of all participants in this study. For
the group of 9 individuals with high AD-signature thickness, none developed AD dementia in
the follow-up period, while for the group of 45 individuals with average AD-signature thick-
ness, 20% developed AD dementia in the follow-up period. A striking 55% of the 11 indi-
viduals with low AD-signature thickness developed AD dementia in the follow-up period.
Neurology 76 April 19, 2011 1399
of early impairment on detailed neuropsychological
performance-based testing. Verbal list learning,
paired associates, and similar challenging memory
tests may provide objective evidence of mild impair-
ments in individuals who otherwise appear normal
when tested but who ultimately develop cognitive
decline and AD dementia after follow-up.
gitudinal cognitive testing, evidence of decline (or in
some cases absence of a practice effect) can be seen
for 5–10 years or more prior to AD dementia.
Given the challenges in these types of assessments,
the clinical designation of “cognitively normal”
should not be used lightly and the methods for mak-
ing this determination deserve further investigation
Three points regarding the cognitive test perfor-
mance characteristics of the present participants are
worth discussion. First, coloring the interpretation of
all of these tests, the participants were relatively well-
educated with all but 3 individuals having at least a
high-school education and more than half having at
least a college degree. Second, no individual per-
formed below 28 on the MMSE at baseline. Third,
although the preclinical AD dementia groups per-
formed slightly lower on baseline memory testing
than those remaining stable, performance would
have been considered within the normal range for
nearly all individuals. Only 3 individuals performed
more than 1 SD below the mean (scores were 1.0 –
1.5 SD below the mean), and one remained clinically
stable as CN nearly a decade later. Nevertheless, epi-
sodic memory measures were useful in prediction of
AD dementia, with similar predictive power in the
Cox model as the MRI measure. We interpret this to
mean that the CN individuals with preclinical AD
may perform within the low-normal range on mem-
ory testing (presumably lower than they would have
performed previously) because of the early neurode-
generative change being detected by this MRI tech-
nique. Longitudinal studies support the idea that a
causal link can be detected between subtle cortical
atrophy at baseline in CN individuals and subse-
quent decline in memory performance.
Studies going back to the 1960s have highlighted
the presence of AD pathology in individuals who die
within a relatively short time after having been as-
sessed as CN,
and extensive investigations have
demonstrated the dissociations that can be present
between the quantitative burden of AD pathology
and the degree of cognitive impairment.
and other observations illustrate the distinction be-
tween the neuropathology of AD as a biological dis-
ease as opposed to the clinical syndrome of AD
dementia, which, although strongly coupled, are by
no means synonymous. Factors such as cognitive re-
serve have been invoked to explain why some indi-
viduals may remain CN despite the presence of
substantial AD neuropathology.
is becoming clear that the presence of AD pathology,
as ascertained via amyloid imaging or CSF analysis,
in CN individuals appears to elevate the probability
of subsequent cognitive decline and dementia.
forts are increasing to elucidate relationships between
the presence and severity of these surrogate molecular
markers of AD neuropathology and the timing of the
development of symptoms. The present data demon-
strate that in CN older adults a thinner cerebral cor-
tex in regions typically affected by AD is predictive of
time to onset of AD dementia, providing an addi-
tional tool to link the biological changes of the dis-
ease with the symptoms of the illness. We
conceptualize this measure as a marker of early neu-
rodegenerative change which appears to be of poten-
tial value among the candidate biomarkers being
proposed as part of the new research criteria for the
diagnosis of preclinical AD.
The major limitation of this study is the relatively
small sample size. Yet most prior studies that resem-
ble the present one also include small samples. The
field is now primed for larger, prospectively designed
studies of this sort, although many challenges remain
in making such cumbersome investigations as effi-
cient as possible. Such studies will require as much as
Figure 3 Alzheimer disease (AD)-signature MRI biomarker predicts time to
dementia in people who were cognitively normal when scanned
Univariate survival plot of predicted time to AD dementia for hypothetical average study
participants with AD-signature cortical thickness in the lowest (smallest) tertile, middle ter-
tile, and highest tertile. The displayed survival curves are therefore model predictions and
do not directly represent subject results. AD-signature thickness was predictive of pro-
gression from normal cognition to AD dementia (hazard ratio 3.5 for 1 SD decrease in thick-
ness; 95% confidence interval 2.0–6.4; p ⬍ 0.00005) in the model. The mean thickness of
the lowest tertile was 1.1 standard deviations thinner than the mean of the entire group of
all participants, the mean of the middle tertile was at approximately the mean of the entire
group, and the mean of the highest tertile was 1.1 standard deviations thicker than the
1400 Neurology 76 April 19, 2011
10 years of follow-up and may need to employ new
methodologic advances, including not only molecu-
lar imaging and CSF analysis, but also telephone-
based cognitive screening and imaging-guided
neuropathologic assessment, to make them at once
cost-effective and informative. Extending these in-
vestigations to more representative samples would
also be a valuable goal, since prior cohorts including
the present ones include highly selected, generally
healthy individuals who are motivated to participate
in this type of research. Another limitation is that the
computational analysis technology required to pro-
duce these measurements, while publicly available,
requires nontrivial expertise and computer systems.
Finally, although prior work has demonstrated that
cortical thickness measures are highly reliable across
different scanner platforms and field strengths,
not clear whether the differences between the mea-
sures in the 2 samples reported here are attributable
to true biological variability, instrument variability,
or other factors. Further investigation of these topics
will be required.
Statistical analysis was conducted by Dr. Dickerson.
The authors thank the faculty and staff of the Massachusetts ADRC and the
Rush ADC for their expertise in coordinating and evaluating participants; and
the participants in this study and their families for their contributions.
Dr. Dickerson serves on the editorial board of Hippocampus and receives
research support from the NIH and the Alzheimer’s Association. Dr.
Stoub reports no disclosures. Dr. Shah receives research support from
Ceregene, Danone Research B.V., Eisai Inc., Elan Corporation, Merck
Serono, Pamlab, L.L.C., Orasi, Inc., Pfizer Inc, the NIH, and the Illinois
Department of Public Aid Alzheimer’s Disease Assistance Center. Dr.
Sperling served on the editorial board of Alzheimer’s Disease and Associated
Disorders; has served as a consultant for Elan Corporation, Wyeth, Bristol-
Myers Squibb, Pfizer Inc, and Bayer Schering Pharma; and receives re-
search support from Elan Corporation, Janssen, Bristol-Myers-Squibb,
NIH/NIA, the Alzheimer’s Association, the Anonymous Foundation, and
the American Health Assistance Foundation. Dr. Killiany receives re-
search support from the NIH. Dr. Albert serves as a consultant for Genen-
tech Inc. and Eli Lilly and Company and receives research support from
the NIH/NIA. Dr. Hyman serves on a scientific advisory board for
NeuroPhage; serves as a consultant for FoldRx Pharmaceuticals, Pfizer
Inc, EMD Serono, Inc., Janssen, Takeda Pharmaceutical Company Lim-
ited, Bristol-Myers Squibb, NeuroPhage, Campbell Alliance, Quanterix,
VBI Belgium, and Schlesinger Associates & The Research House; receives
research support from Fidelity Biosciences, the NIH and the Alzheimer’s
Association; and holds stock in Novartis. Dr. Blacker receives research
support from the NIH and the Alzheimer’s Association. Dr. deToledo-
Morrell serves on the editorial board of Neurobiology of Aging and receives
research support from the NIH/NIA.
Received November 16, 2010. Accepted in final form January 11, 2011.
1. Amieva H, Le Goff M, Millet X, et al. Prodromal Alzhei-
mer’s disease: successive emergence of the clinical symp-
toms. Ann Neurol 2008;64:492–498.
2. Rubin EH, Storandt M, Miller JP, et al. A prospective
study of cognitive function and onset of dementia in cog-
nitively healthy elders. Arch Neurol 1998;55:395–401.
3. Grober E, Hall CB, Lipton RB, Zonderman AB, Resnick
SM, Kawas C. Memory impairment, executive dysfunc-
tion, and intellectual decline in preclinical Alzheimer’s dis-
ease. J Int Neuropsychol Soc 2008;14:266–278.
4. Hall CB, Lipton RB, Sliwinski M, Stewart WF. A change
point model for estimating the onset of cognitive decline
in preclinical Alzheimer’s disease. Stat Med 2000;19:
5. Howieson DB, Dame A, Camicioli R, Sexton G, Payami
H, Kaye JA. Cognitive markers preceding Alzheimer’s de-
mentia in the healthy oldest old. J Am Geriatr Soc 1997;
6. Johnson DK, Storandt M, Morris JC, Galvin JE. Longitu-
dinal study of the transition from healthy aging to Alzhei-
mer disease. Arch Neurol 2009;66:1254–1259.
7. Blacker D, Lee H, Muzikansky A, et al. Neuropsychologi-
cal measures in normal individuals that predict subsequent
cognitive decline. Arch Neurol 2007;64:862–871.
8. Morris JC, Roe CM, Grant EA, et al. Pittsburgh com-
pound B imaging and prediction of progression from cog-
nitive normality to symptomatic Alzheimer disease. Arch
9. Fox NC, Warrington EK, Freeborough PA, et al. Pres-
ymptomatic hippocampal atrophy in Alzheimer’s disease: a
longitudinal MRI study. Brain 1996;119:2001–2007.
10. Kaye JA, Swihart T, Howieson D, et al. Volume loss of the
hippocampus and temporal lobe in healthy elderly persons
destined to develop dementia. Neurology 1997;48:1297–
11. Fox NC, Crum WR, Scahill RI, Stevens JM, Janssen JC,
Rossor MN. Imaging of onset and progression of Alzhei-
mer’s disease with voxel-compression mapping of serial
magnetic resonance images. Lancet 2001;358:201–205.
12. Schott JM, Fox NC, Frost C, et al. Assessing the onset of
structural change in familial Alzheimer’s disease. Ann Neu-
13. Smith CD, Chebrolu H, Wekstein DR, et al. Brain struc-
tural alterations before mild cognitive impairment. Neu-
14. den Heijer T, Geerlings MI, Hoebeek FE, Hofman A,
Koudstaal PJ, Breteler MM. Use of hippocampal and
amygdalar volumes on magnetic resonance imaging to pre-
dict dementia in cognitively intact elderly people. Arch
Gen Psychiatry 2006;63:57–62.
15. Martin SB, Smith CD, Collins HR, Schmitt FA, Gold BT.
Evidence that volume of anterior medial temporal lobe is
reduced in seniors destined for mild cognitive impairment.
Neurobiol Aging 2011;31:1099–1106.
16. Apostolova LG, Mosconi L, Thompson PM, et al. Subre-
gional hippocampal atrophy predicts Alzheimer’s dementia
in the cognitively normal. Neurobiol Aging 2010;31:
17. Csernansky JG, Wang L, Swank J, et al. Preclinical detec-
tion of Alzheimer’s disease: hippocampal shape and vol-
ume predict dementia onset in the elderly. Neuroimage
18. Dickerson BC, Bakkour A, Salat DH, et al. The cortical
signature of Alzheimer’s disease: regionally specific cortical
thinning relates to symptom severity in very mild to mild
AD dementia and is detectable in asymptomatic amyloid-
positive individuals. Cereb Cortex 2009;19:497–510.
Neurology 76 April 19, 2011 1401
19. DeToledo-Morrell L, Stoub TR, Bulgakova M, et al.
MRI-derived entorhinal volume is a good predictor of con-
version from MCI to AD. Neurobiol Aging 2004;25:
20. Jack CR Jr, Shiung MM, Weigand SD, et al. Brain atrophy
rates predict subsequent clinical conversion in normal el-
derly and amnestic MCI. Neurology 2005;65:1227–1231.
21. Braak H, Braak E. Neuropathological stageing of
Alzheimer-related changes. Acta Neuropathol 1991;82:
22. Price JL, Davis PB, Morris JC, White DL. The distribu-
tion of tangles, plaques and related immunohistochemical
markers in healthy aging and Alzheimer’s disease. Neuro-
biol Aging 1991;12:295–312.
23. Arriagada PV, Growdon JH, Hedley-Whyte ET, Hyman
BT. Neurofibrillary tangles but not senile plaques parallel
duration and severity of Alzheimer’s disease. Neurology
24. Thompson PM, Mega MS, Woods RP, et al. Cortical
change in Alzheimer’s disease detected with a disease-
specific population-based brain atlas. Cereb Cortex 2001;
25. Good CD, Scahill RI, Fox NC, et al. Automatic differenti-
ation of anatomical patterns in the human brain: valida-
tion with studies of degenerative dementias. Neuroimage
26. Saykin AJ, Wishart HA, Rabin LA, et al. Older adults with
cognitive complaints show brain atrophy similar to that of
amnestic MCI. Neurology 2006;67:834– 842.
27. Schmand B, Jonker C, Hooijer C, Lindeboom J. Subjec-
tive memory complaints may announce dementia. Neurol-
28. Jorm AF, Masaki KH, Davis DG, et al. Memory com-
plaints in nondemented men predict future pathologic
diagnosis of Alzheimer disease. Neurology 2004;63:1960–
29. Murphy EA, Holland D, Donohue M, et al. Six-month
atrophy in MTL structures is associated with subsequent
memory decline in elderly controls. Neuroimage 2010;53:
30. Tomlinson BE, Blessed G, Roth M. Observations on the
brains of non-demented old people. J Neurol Sci 1968;7:
31. Price JL, McKeel DW Jr, Buckles VD, et al. Neuropathol-
ogy of nondemented aging: presumptive evidence for pre-
clinical Alzheimer disease. Neurobiol Aging 2009;30:
32. Crystal H, Dickson D, Fuld P, et al. Clinico-pathologic
studies in dementia: nondemented subjects with patholog-
ically confirmed Alzheimer’s disease. Neurology 1988;38:
33. Katzman R, Terry R, DeTeresa R, et al. Clinical, patholog-
ical, and neurochemical changes in dementia: a subgroup
with preserved mental status and numerous neocortical
plaques. Ann Neurol 1988;23:138–144.
34. Dickson DW, Crystal HA, Mattiace LA, et al. Identifica-
tion of normal and pathological aging in prospectively
studied nondemented elderly humans. Neurobiol Aging
35. Hyman BT, Marzloff K, Arriagada PV. The lack of accu-
mulation of senile plaques or amyloid burden in Alzhei-
mer’s disease suggests a dynamic balance between amyloid
deposition and resolution. J Neuropathol Exp Neurol
36. Rentz DM, Locascio JJ, Becker JA, et al. Cognition, re-
serve, and amyloid deposition in normal aging. Ann Neu-
37. Roe CM, Mintun MA, D’Angelo G, Xiong C, Grant EA,
Morris JC. Alzheimer disease and cognitive reserve: varia-
tion of education effect with carbon 11-labeled Pittsburgh
Compound B uptake. Arch Neurol 2008;65:1467–1471.
38. Stern Y. Cognitive reserve and Alzheimer disease. Alzhei-
mer Dis Assoc Disord 2006;20:S69–S74.
39. Fagan AM, Roe CM, Xiong C, Mintun MA, Morris JC,
Holtzman DM. Cerebrospinal fluid tau/beta-amyloid(42)
ratio as a prediction of cognitive decline in nondemented
older adults. Arch Neurol 2007;64:343–349.
40. Sperling RA, Aisen PS, Beckett LA, et al. Towards defining
the preclinical stages of Alzheimer’s disease: recommenda-
tions from the National Institute on Aging and the Alzhei-
mer Association Workgroup. Alzheimers Dement (in press
41. Dickerson BC, Fenstermacher E, Salat DH, et al. Detec-
tion of cortical thickness correlates of cognitive perfor-
mance: Reliability across MRI scan sessions, scanners, and
field strengths. Neuroimage 2008;39:10–18.
Career Moves Begin at Neurology Career Center
Job seekers: The AAN’s Neurology Career Center is a one-stop shop for qualified candidates
looking to make a career move in neurology. Search for opportunities in your state and area of
interest and create a profile that you can share with only those employers in whom you’re interested.
Advertising for a position in neurology? Reach more than 22,500 Academy members online, in
print, and at special career events.
Make your next career connection with the Neurology Career Center!
1402 Neurology 76 April 19, 2011