Content uploaded by Clifford R Jack
Author content
All content in this area was uploaded by Clifford R Jack on Jun 09, 2016
Content may be subject to copyright.
www.thelancet.com/neurology Vol 13 October 2014
997
Articles
Age-specifi c population frequencies of cerebral
β-amyloidosis and neurodegeneration among people with
normal cognitive function aged 50–89 years: a cross-
sectional study
Cliff ord R Jack Jr, Heather J Wiste, Stephen D Weigand, Walter A Rocca, David S Knopman, Michelle M Mielke, Val J Lowe, Matthew L Senjem,
Jeff rey L Gunter, Gregory M Preboske, Vernon S Pankratz, Prashanthi Vemuri, Ronald C Petersen
Summary
Background As preclinical Alzheimer’s disease becomes a target for therapeutic intervention, the overlap between
imaging abnormalities associated with typical ageing and those associated with Alzheimer’s disease needs to be
recognised. We aimed to characterise how typical ageing and preclinical Alzheimer’s disease overlap in terms of
β-amyloidosis and neurodegeneration.
Methods We measured age-specifi c frequencies of amyloidosis and neurodegeneration in individuals with normal
cognitive function aged 50–89 years. Potential participants were randomly selected from the Olmsted County (MN,
USA) population-based study of cognitive ageing and invited to participate in cognitive and imaging assessments. To
be eligible for inclusion, individuals must have been judged clinically to have no cognitive impairment and have
undergone amyloid PET,
¹⁸
F-fl uorodeoxyglucose (
¹⁸
F-FDG) PET, and MRI. Imaging results were obtained from
March 28, 2006, to Dec 3, 2013. Amyloid status (positive [A+] or negative [A–]) was determined by amyloid PET with ¹¹C
Pittsburgh compound B. Neurodegeneration status (positive [N+] or negative [N–]) was determined by an Alzheimer’s
disease signature
¹⁸
F-FDG PET or hippocampal volume on MRI. We determined age-specifi c frequencies of the four
groups (amyloid negative and neurodegeneration negative [A–N–], amyloid positive and neurodegeneration negative
[A+N–], amyloid negative and neurodegeneration positive [A–N+], or amyloid positive and neurodegeneration positive
[A+N+]) cross-sectionally using multinomial regression models. We also investigated associations of group frequencies
with APOE ε4 status (assessed with DNA extracted from blood) and sex by including these covariates in the
multinomial models.
Findings The study population consisted of 985 eligible participants. The population frequency of A–N– was 100%
(n=985) at age 50 years and fell to 17% (95% CI 11–24) by age 89 years. The frequency of A+N– increased to 28%
(24–32) at age 74 years, then decreased to 17% (11–25) by age 89 years. The frequency of A–N+ increased from age
60 years, reaching 24% (16–34) by age 89 years. The frequency of A+N+ increased from age 65 years, reaching 42%
(31–52) by age 89 years. The results from our multinomial models suggest that A+N– and A+N+ were more frequent in
APOE ε4 carriers than in non-carriers and that A+N+ was more, and A+N– less frequent in men than in women.
Interpretation Accumulation of amyloid and neurodegeneration are nearly inevitable by old age, but many people are
able to maintain normal cognitive function despite these imaging abnormalities. Changes in the frequency of
amyloidosis and neurodegeneration with age, which seem to be modifi ed by APOE ε4 and sex, suggest that
pathophysiological sequences might diff
er between individuals.
Funding US National Institute on Aging and Alexander Family Professorship of Alzheimer’s Disease Research.
Introduction
Recognition that biomarker evidence of Alzheimer’s
disease pathophysiology is present long before clinical
symptoms become apparent1 has motivated the
formulation of research criteria for preclinical
Alzheimer’s disease.2,3 In 2011, the authors of the National
Institute on Aging–Alzheimer’s Association (NIA–AA)
criteria described a method for defi ning and staging
preclinical Alzheimer’s disease, defi ning stage 1 as
cerebral amyloidosis, stage 2 as amyloidosis plus
neurodegeneration, and stage 3 as amyloidosis,
neurodegeneration, and subtle cognitive decline.2
Although the NIA–AA method probably accurately
refl ects the onset and staged progression of biomarkers
in the pathophysiology of Alzheimer’s disease,1,4–6
Alzheimer’s disease pathological changes do not typically
occur in isolation in elderly people, but rather co-occur
with other age-related degenerative processes.7 Structural
MRI and
¹⁸F-fl uorodeoxyglucose (¹⁸F-FDG) PET are
sensitive approaches to the measurement of
neurodegeneration or brain injury; however, even
signature Alzheimer’s disease topographic measures on
these modalities (eg, hippocampal atrophy on MRI) are
not specifi c for Alzheimer’s disease.8–10
Lancet Neurol 2014;
13: 997–1005
Published Online
September 5, 2014
http://dx.doi.org/10.1016/
S1474-4422(14)70194-2
See Comment page 965
Department of Radiology
(Prof C R Jack Jr MD,
Prof V J Lowe MD,
M L Senjem MS, J L Gunter PhD,
G M Preboske MS,
P Vemuri PhD), Department of
Health Sciences Research
(H J Wiste BA, S D Weigand MS,
Prof W A Rocca MD,
M M Mielke PhD,
V S Pankratz PhD), and
Department of Neurology
(W A Rocca,
Prof D S Knopman MD,
Prof R C Petersen MD), Mayo
Clinic and Foundation,
Rochester, MN, USA
Correspondence to:
Prof Cliff ord R Jack Jr,
Department of Radiology, Mayo
Clinic and Foundation,
Rochester, MN 55905, USA
jack.cliff ord@mayo.edu
Articles
998
www.thelancet.com/neurology Vol 13 October 2014
A two-feature biomarker classifi cation system based on
both β-amyloidosis and neurodegeneration, described
previously,11,12 expands the NIA–AA staging of preclinical
Alzheimer’s disease.2 This system can be used to classify
all individuals, rather than only those who are exclusively
in the Alzheimer’s disease pathophysiological pathway,
thereby accommodating the facts that Alzheimer’s
disease and non-Alzheimer’s disease degenerative
processes occur with ageing and that techniques for the
imaging of neurodegeneration are sensitive to various
degenerative processes. Every individual is assigned to
one of four groups in this scheme: amyloid negative and
neurodegeneration negative (A–N–), amyloid positive and
neurodegeneration negative (A+N–), amyloid negative
and neurodegeneration positive (A–N+), or amyloid
positive and neurodegeneration positive (A+N+).11,12 A
–N–
corresponds to NIA–AA stage 0, A+N– to NIA–AA stage 1,
A–N+ to suspected non-Alzheimer’s pathophysiology
(SNAP),13 and A+N+ to NIA–AA stages 2 and 3.
Our classifi cation system11,12 also operationalises the 2014
International Working Group (IWG) research criteria (the
IWG-2 criteria) for the stage of asymptomatic at risk for
Alzheimer’s disease.3 Asymptomatic at risk for Alzheimer’s
disease is defi ned by the absence of a clinical phenotype
consistent with typical or atypical Alzheimer’s disease and
the presence of a pathophysiological biomarker consistent
with the presence of Alzheimer’s disease pathophysiology.
A positive amyloid PET scan is the only currently available
imaging fi nding that is diagnostic of Alzheimer’s disease
pathophysiology.11 Structural MRI and ¹⁸F-FDG PET
abnormalities in topographic areas characteristic of
Alzheimer’s disease are used to stage disease severity, but
not as diagnostic measures.3 Thus, framed in terms of the
IWG-2 criteria, A+N– and A+N+ individuals with normal
cognitive function would be designated as asymptomatic
at risk for Alzheimer’s disease, with A+N+ individuals at a
more advanced stage of the disease process. According to
the IWG-2 criteria, A–N– and A–N+ individuals would not be
regarded as having evidence of Alzheimer’s disease
pathophysiology.
From a clinical standpoint, typical ageing blends
imperceptibly with preclinical Alzheimer’s disease in the
population. Our objective was to characterise amyloidosis
and neurodegeneration in people with normal cognitive
function, a population that includes both typical ageing
and preclinical Alzheimer’s disease (asymptomatic at
risk for Alzheimer’s disease). We aimed to estimate age-
specifi c frequencies of the four groups based on
amyloidosis and neurodegeneration status in a large
sample of individuals with normal cognitive function
aged 50–89 years from a population-based cohort.
Methods
Study design and participants
We studied participants with normal cognitive function
in the Mayo Clinic Study of Aging (MCSA), a
population-based study of cognitive ageing among
residents of Olmsted County (MN, USA).14 The
Rochester Epidemiology Project15 medical records
linkage system was used to enumerate all Olmsted
County residents aged 50–89 years. All residents from
the population count were randomly ordered in lists
and stratifi ed by age and sex; we selected potential
participants from these ordered lists (by taking the fi rst
number on each list that had not already been selected)
until the target enrolment in each age and sex stratum
was achieved (age strata were defi ned as 5-year groups,
from 50–54 years to 85–89 years). Roughly half of the
randomly identifi ed individuals from Olmsted County
agree to a comprehensive, in-person clinical assessment
(as opposed to telephone-only participation), which was
necessary for inclusion in the present analysis because
only in-person participants are invited to participate in
the necessary imaging assessments. Men and women
are equally represented in each age stratum. All
individuals without a medical contraindication are
invited to participate in imaging assessments.
Since 2004, the MCSA investigators have enrolled
individuals without dementia aged 70–89 years, and in
2012 started to enrol people without dementia aged
50 years and older.
To be eligible for inclusion in the present analysis,
individuals must have been judged clinically to have no
cognitive impairment on the basis of a battery of nine
psychometric tests and assessments by a study
coordinator and a physician (only results from the
Auditory Verbal Learning Test are reported for
simplicity).14 Participants also had to have undergone
amyloid PET, ¹⁸F-FDG PET, and MRI within 7 months of
their index clinical visit. Imaging results were obtained
from March 28, 2006, to Dec 3, 2013. Amyloid PET,
18F-FDG PET, and MRI protocols were identical for all
participants. APOE genotype was assessed by use of
standard laboratory procedures with DNA extracted from
blood.16
This study was approved by institutional review boards
at the Mayo Clinic and Olmsted Medical Center (both
Rochester, MS, USA) and all participants provided
written informed consent.
Procedures
Amyloid PET imaging was done with
¹¹
C Pittsburgh
compound B (¹¹C-PIB) and consisted of four 5-min
dynamic frames acquired 40–60 min after injection.
¹⁸
F-FDG PET imaging was done on the same day as the
¹¹
C-PIB PET scan and consisted of four 2-min dynamic
frames acquired 30–38 min after injection. CT imaging
was done during the PET session for attenuation
correction. We analysed amyloid PET and
¹⁸
F-FDG PET
images using our in-house, fully automated image
processing pipeline,12 wherein image voxel values are
extracted from automatically labelled regions of interest.
An amyloid PET standardised uptake value ratio was
formed from the prefrontal, orbitofrontal, parietal,
Articles
www.thelancet.com/neurology Vol 13 October 2014
999
temporal, anterior cingulate, posterior cingulate, and
precuneus regions of interest, normalised to the whole
cerebellum. An Alzheimer’s disease signature ¹⁸F-FDG
PET standardised uptake value ratio was formed from
the angular gyrus, posterior cingulate, and inferior
temporal cortical regions of interest, normalised to pons
and vermis.17
MRI scanning was done on one of three 3-Tesla
scanners from the same manufacturer (General Electric,
Milwaukee, WI, USA). Hippocampal volume was
measured with FreeSurfer (version 5.3). Total intracranial
volume was measured by use of an in-house method.12
Tissue class segmentation maps were created by SPM12
with custom priors and passed through a series of
morphological opening, erosion, dilation, and
thresholding steps. The hippocampal masks and total
intracranial volume masks were manually inspected for
quality control by a trained analyst. Each participant’s
raw hippocampal volume was adjusted for total
intracranial volume to create an adjusted hippocampal
volume by calculation of the residual from a linear
regression of hippocampal volume versus total
intracranial volume among 133 participants with normal
cognitive function aged 30–59 years.12 Adjusted
hippocampal volume can be interpreted as the deviation
in cm³ in a participant’s hippocampal volume from what
is expected on the basis of their head size.
We defi ned the negative and positive threshold for
amyloid PET,
¹⁸
F-FDG PET, and adjusted hippocampal
volume such that 90% of a group of 75 individuals with
clinically diagnosed Alzheimer’s disease dementia from
the Mayo Clinic Alzheimer Disease Research Center
and MCSA were classifi ed as having abnormal values,
using the same approach described previously.12
Abnormal amyloid PET was defi ned as a standardised
uptake value ratio of 1·40 or higher. Neurodegeneration
was defi ned as either an adjusted hippocampal volume
of –2·40 cm³ or less or an
¹⁸
F-FDG PET standardised
uptake value ratio of 1·32 or less. We assigned
participants into one of four groups defi ned by the
combination of abnormality for amyloid and neuro-
degeneration: A–N–, A+N–, A–N+, or, A+N+.12
Statistical analysis
We assessed pairwise diff erences in characteristics
among the four groups using Wilcoxon rank-sum tests or
χ² tests, as appropriate. We used multinomial regression
models that included terms for both age and sex to
estimate frequencies (percentages) of each group for
amyloid and neurodegeneration abnormality as a
function of age. As a generalisation of binary logistic
regression, multinomial regression is appropriate
because each participant could be classifi ed into one of
four unordered categories. To allow for fl exible age-
dependent trends in these frequencies, we modelled age
trends using restricted cubic splines with knots at ages
60, 70, and 80 years.
We calculated 95% CIs on the probability scale using a
parametric bootstrap. We used this procedure because
linear approximations that used the delta method were
inadequate. To do the parametric bootstrap, we sampled
10 000 multivariate normal deviates with means equal to
the parameter estimates and variance structure equal to
the variance-covariance matrix of the fi tted model. These
samples represent plausible realisations of the model
coeffi cients while allowing for statistical uncertainty in
their estimated values. Each realisation was used as the
set of parameter estimates in the multinomial regression
equation and predictions from each resampled model
were used to calculate frequency estimates as a function
of age for each group while controlling for sex diff erences.
We also used this same procedure to calculate 95% CIs
for the diff erences in frequency between groups by age.
These 95% CIs were defi ned as the 2·5th and 97·5th
quantiles of the resampled distribution. We interpreted
95% CIs that did not include zero as signifi cant.
We also examined how the age-dependent group
frequencies varied by combinations of sex and APOE ε4
carrier status. We used likelihood ratio tests to assess the
signifi cance of additive eff ects for each of these patient
characteristic groupings, as well as two-way and three-
way interactions between age, sex, and APOE ε4 status.
Our analysis examines age, sex, and APOE ε4 status
associations across the four groups, which arguably
raises the issue of multiple comparisons. However, since
our inference is mainly based on only two models, one
with 12 parameters and one with 15 parameters—not
many parameters for our sample size—we do not believe
that classic multiple-testing problems, or the related
issue of spare-data bias, are applicable.
We used SAS version 9.3 (SAS Institute, Cary, NC,
USA) and R version 3.0.2 (R Foundation for Statistical
Computing, Vienna, Austria) with the “multinom”
function from the “nnet” add-on package for the
statistical analyses.
Role of the funding source
The funders of the study had no role in study design,
data collection, data analysis, data interpretation, or
writing of the report. All authors had full access to all the
data in the study. The corresponding author had fi nal
responsibility for the decision to submit for publication.
Results
985 individuals met the criteria for inclusion in the
present analysis. The median age increased by group in
the order A–N–, A+N–, A–N+, then A+N+ (p<0·0001 for all
pairwise comparisons apart from A–N+ vs A+N+ [p=0·013];
table 1). The proportion of men was higher in the A–N+
group than in the A+N– group (p=0·006). The proportion
of men was higher in the A+N+ group than in the A–N–
group (p=0·010) or the A+N– group (p=0·0007). The
proportion of APOE ε4 carriers was higher in the A+N– and
A+N+ groups than in the A–N– and A–N+ groups (p≤0·0001
Articles
1000
www.thelancet.com/neurology Vol 13 October 2014
for all comparisons). Of the 269 neurodegeneration-
positive individuals, 170 (63%) were classifi ed as such
because of abnormalities on
¹⁸
F-FDG PET alone, 51 (19%)
because of abnormal adjusted hippocampal volume alone,
and 48 (18%) because of both.
Frequencies of the diff erent amyloid and
neurodegeneration groups varied substantially by 5-year
age stratum (appendix p 1, table 2). Abnormal imaging
results fi rst appeared at age 60–64 years. The model-
based estimates and 95% CIs adjusted for sex (fi gure 1)
were largely in agreement with the empirical values
(appendix p 1, table 2). The model generated smoothed or
denoised estimated frequencies, which are more realistic,
because jumps between successive 5-year age brackets
are smoothed. We therefore based our inferences on the
model-based estimates (fi gure 1).
At age 50 years, all individuals were A–N–. The estimated
frequency of A–N– fell monotonically with age, reaching
17% (95% CI 11–24) by age 89 years. The estimated
frequency of A+N– increased to a maximum of
28% (24–32) by age 74 years, then fell to 17% (11–25) by
age 89 years. The estimated frequency of A–N+ was close
to zero before age 60 years and increased monotonically
thereafter, reaching a frequency of 24% (16–34) by age
89 years. The estimated frequency of A+N+ was close to
zero before age 65 years and increased monotonically
thereafter, reaching 42% (31–52) by age 89 years.
Figure 2 shows pairwise diff erences in group
frequencies by age and suggests estimates of the ages at
which signifi cant diff erences in frequencies of the
groups were present (we interpret diff erences as
signifi cant at ages for which the 95% CI around the
estimated mean diff erence does not include the
horizontal line indicating zero diff erence). A–N– was
more frequent than any of the other three groups from
age 50 years until about age 80 years. The frequency of
A+N– exceeded that of A–N+ and A+N+ from about age
60 years to between ages 75 and 80 years. A+N+ was more
frequent than A+N– and A–N– at age 85 years and older.
The frequencies of A+N+ and A–N+ did not diff er
signifi cantly at any age.
Age (p<0·0001), sex (p=0·004), and APOE ε4 status
(p<0·0001) were each independently associated with
group frequencies. We did not identify evidence of an
interaction between age and APOE ε4 status (p=0·63), an
interaction between sex and APOE ε4 status (p=0·78), or a
three-way interaction (p=0·47) on group frequencies. The
age-by-sex interaction was not signifi cant (p=0·06). A
model that incorporated this interaction provided
estimates of group frequency that were similar (but with
wider CIs between ages 50 and 60 years and beyond age
85 years)to those in which age, sex, and APOE ε4 status
were treated as additive eff ects (ie, no interaction). We
therefore chose to report group frequencies from the
model with additive age, sex, and APOE eff ects.
The biomarker group frequencies by age among
groups defi ned by sex and APOE ε4 status are shown in
fi gure 3, and diff erences in frequencies comparing men
with women within APOE ε4 status and ε4 carriers with
non-carriers within sex are shown in the appendix (p 2).
See Online for appendix
Overall
(n=985)
Amyloid
negative,
neuro-
degeneration
negative
(AN; n=503)
Amyloid
positive,
neuro-
degeneration
negative
(A⁺N; n=213)
Amyloid
negative,
neuro-
degeneration
positive
(AN⁺; n=130)
Amyloid
positive,
neuro-
degeneration
positive
(A⁺N⁺; n=139)
Age (years) 74 (67–80) 70 (63–76) 74 (70–80) 77 (74–83) 80 (77–83)
Men 540 (55%) 268 (53%) 100 (47%) 81 (62%) 91 (65%)
Years in education 14 (12–16) 15 (12–17) 14 (12–16) 14 (12–16) 14 (12–16)
APOE ε4 carriers 255 (26%) 98 (20%) 79 (37%) 23 (18%) 55 (40%)
AVLT* 59 (47–70) 62 (52–73) 59 (50–70) 51 (42–64) 50 (39–61)
Data are median (IQR) or number (%). AVLT=Auditory Verbal Learning Test. *Sum of trials 1–5 plus the immediate and
delayed recall trials (possible total score of 105).
Table 1: Participant characteristics
Amyloid negative,
neuro degeneration
negative
(AN; n=503)
Amyloid positive,
neuro degeneration
negative
(A⁺N; n=213)
Amyloid negative,
neuro degeneration
positive
(AN⁺; n=130)
Amyloid positive,
neuro degeneration
positive
(A⁺N⁺; n=139)
50–54 years 35 (100%) 0 0 0
55–59 years 37 (100%) 0 0 0
60–64 years 93 (85%) 13 (12%) 3 (3%) 1 (1%)
65–69 years 80 (61%) 37 (28%) 10 (8%) 5 (4%)
70–74 years 110 (52%) 58 (27%) 28 (13%) 16 (8%)
75–79 years 91 (43%) 44 (21%) 36 (17%) 40 (19%)
80–84 years 40 (24%) 45 (27%) 33 (20%) 50 (30%)
85–89 years 17 (21%) 16 (20%) 20 (25%) 27 (34%)
Data are n (%).
Table 2: Numbers of participants in each biomarker group by 5-year age stratum
Figure 1: Estimated frequency (percentage) of participants in each biomarker
group, by age
Estimates are from a multinomial model adjusted for sex. Non-linearity in age
was allowed in the model by fi tting age as a spline with knots at ages 60, 70, and
80 years. Shaded areas are 95% parametric bootstrap CIs.
50 60 70 80 90
0
10
20
30
40
50
60
70
80
90
100
Frequency (%)
Age (years)
A–N–
A+N–
A–N+
A+N+
Articles
www.thelancet.com/neurology Vol 13 October 2014
1001
A–N– was more frequent in APOE ε4 non-carriers than in
carriers at all ages. A+N– was more frequent in APOE ε4
carriers than in non-carriers until about age 80 years.
A+N– was more frequent in women than in men from
about age 75 years onwards. A–N+ was less frequent in
APOE ε4 carriers than in non-carriers from about age
75 years onwards. A+N+ was more frequent in APOE ε4
carriers than in non-carriers from about age 65 years
onwards. A+N+ was more frequent in men than in women
from about age 65–70 years onwards.
Discussion
In our cross-sectional study among individuals with
normal cognitive function aged 50–89 years, the
frequency of A–N– fell monotonically with age from 100%
at age 50 years to 17% at age 89 years. The frequency of
A–N+ and A+N– increased monotonically with age. The
pattern for A+N– was diff erent, with the frequency
increasing until age 74 years and decreasing thereafter.
Sex and APOE4 ε4 status seemed to modify these age
trends.
Various processes that can be detrimental to brain
structure and function become more prevalent in the
adult population with increasing age. These include
Alzheimer’s disease pathological changes, non-
Alzheimer’s disease pathological changes, and ageing
without specifi c pathological changes.9,18–21 An idealised
system can be envisioned wherein individuals are
classifi ed on the basis of biomarkers of all relevant age-
related processes. Although such a classifi cation system
might be developed in the future if biomarkers for all
major processes associated with cognitive ageing become
available, for now a four-class system11,12 based on
amyloidosis and neurodegeneration could be useful.
The frequency of abnormal amyloid PET scans by age
in our study is similar to those in a previous report.22
However, classifi cation of individuals by both amyloidosis
and neurodegeneration adds an important dimension to
classifi cation by amyloidosis alone (panel). Our results
suggest that from age 50 to 60 years the frequency of
preclinical Alzheimer’s disease (asymptomatic at risk for
Alzheimer’s disease (ie, A+N– or A+N+) is close to zero,
and we estimate that by age 89 years the proportion is
more than half (59%) of people with normal cognitive
function in the general population will meet these
criteria.2,3 More specifi cally, we estimate that by age
89 years the frequency of A+N– (asymptomatic at risk for
Alzheimer’s disease without neurodegeneration, or
preclinical Alzheimer’s disease stage 1) is 17%, whereas
the frequency of A+N+ (asymptomatic at risk for
Alzheimer’s disease with neurodegeneration, or
preclinical Alzheimer’s disease stage 2 and 3) is 42%.
A–N+ (or SNAP) represents an increasingly large
proportion of people with normal cognitive function in
the general population older than 60 years. In our initial
description of SNAP,11 we suggested that imaging
evidence of Alzheimer’s disease-like neurodegeneration
without amyloidosis probably represents any
combination of non-Alzheimer’s disease processes
including medial temporal tauopathy, cerebrovascular
disease, Lewy body disease, grain disease, hippocampal
sclerosis, TDP-43 proteinopathies, or ageing changes
without specifi c pathological causes.7,19 Although both
our ¹⁸F-FDG PET and MRI measures capture
characteristic topographic patterns of Alzheimer’s
disease, neither is specifi c for Alzheimer’s disease
pathophysiology and thus abnormal results could be
present in people at risk of disorders other than
Alzheimer’s disease. Additionally, because ¹⁸F-FDG
PET was not corrected for partial volume eff ects, this
measure captures both the eff ects of decreased ¹⁸F-FDG
uptake and brain atrophy.
Figure 2: Diff erences in estimated frequencies (percentages) of participants in each biomarker group, by age
Estimates are from a multinomial model adjusted for sex. Non-linearity in age was allowed in the model by fi tting
age as a spline with knots at ages 60, 70, and 80 years. Diff erences in estimated frequencies are plotted with
95% parametric bootstrap CIs (shaded areas).
–100
–50
0
50
Difference (%)
A+N– vs A–N–A–N+ vs A–N–
–100
–50
0
50
Difference (%)
A+N+ vs A–N–A–N+ vs A+N–
50 60 70 80 90
–100
–50
0
50
Difference (%)
A
g
e (years)
A+N+ vs A+N–
50 60 70 80 90
A
g
e (years)
A+N+ vs A–N+
Articles
1002
www.thelancet.com/neurology Vol 13 October 2014
Frequencies of amyloidosis and neurodegeneration in
the population changed substantially with age in our
study. Reasons for this fi nding fall into two main
categories: selective censoring of individuals in one or
more groups relative to the others because of death, non-
participation, or progression to cognitive impairment; or
transition from one group to another while remaining
cognitively unimpaired (thus remaining in the eligible
study pool). We acknowledge that without longitudinal
data for every participant the relative contributions of each
of these eff ects cannot be disentangled. However, an
obvious major overall trend in people aged 60–70 years is
a monotonic decrease in the frequency of A–N– with age
coupled with increases in the frequencies of the other
groups. For selective elimination of participants aged
60–70 years from the cognitively normal population to be
able to fully account for these trends is highly improbable.
The better explanation almost certainly is that individuals
transition from A–N– to more advanced stages of
amyloidosis and neurodegeneration while maintaining
normal cognitive function. Beyond age 70 years the most
likely cause of selective elimination of participants from
the study pool is progression to cognitive impairment.
However, this outcome is most likely for individuals in the
A+N+ group,23,24 which would decrease the frequency of
A+N+ with age. Yet we see the opposite: a substantial
increase in the frequency of A+N+ beyond age 65 years.
Therefore transitions from less to more severe groups
while individuals maintain normal cognitive function
must be a dominant explanation for the changing cross-
sectional frequencies with age that we have identifi ed.
Interpretive parallels might therefore be drawn between
our data and studies by Braak and colleagues25 and
Duckyaerts and Hauw,26 who infer the natural history of
progression of tauopathy and amyloidosis within
individuals, based on the observation that population
frequencies of less advanced pathological stages decrease
and those of more advanced pathological stages increase
with advancing age.
Results from our theoretical modelling studies27,28
provide a possible integrated explanation for the patterns
of change in the frequencies of amyloidosis and
neurodegeneration with age seen in the present study.
Those models predict that individuals with normal
cognitive function might follow diff erent patho physio-
logical sequences denoted by amyloid and neuro-
degenerative biomarkers. The fi rst potential sequence is
A–N– to A+N– to A+N+ (fi gure 4A). This is the Alzheimer’s
disease biomarker sequence of pre clinical Alzheimer’s
disease without major comorbid non-Alzheimer’s disease
pathological changes.2,29 The second potential patho-
physiological sequence is A–N– to A–N+ to A+N+ (fi gure 4B).
We have reported the A–N+ to A+N+ transition in an earlier
study of incident amyloid PET positivity among individuals
with normal cognitive function13 and proposed that it
indicates someone who fi rst develops SNAP neuro-
degeneration, and later enters the Alzheimer’s disease
pathophysiological pathway, denoted by a positive amyloid
PET fi nding. The assumption that A–N+ (SNAP) represents
non-Alzheimer’s disease neurodegenerative pathological
change is lent support by the fact that the proportion of
APOE ε4 carriers in A–N+ was low (18%) compared with
that in A+N– (37%) and A+N+ (40%) in our study. The third
sequence is A–N– to A–N+ (fi gure 4C), which we propose
represents someone who develops SNAP neuro-
degeneration without progressing into the Alzheimer’s
disease pathophysiological pathway.
We believe it is biologically meaningful that, of the
three groups with abnormal values (A+N–, A–N+, and
A+N+), only A+N– falls in frequency with age. Although the
decrease in A+N– frequency after age 74 years could be
due to a higher frequency of progression to cognitive
impairment than in other groups, the fact that people
classed as A+N+ are most likely to become impaired23,24 yet
increase in frequency monotonically with age argues
against this possibility. A possible explanation for our
fi nding is that A+N– is an inherently unstable state and
that from age 74 onwards transition out of A+N– while
maintaining normal cognitive function for some period
of time is likely. Because we believe that A+N– is not a
natural biomarker end state, it is not included as a
possible pathway endpoint in fi gure 4. We do not discuss
transition directly from A–N– to A+N+ because we believe
Figure 3: Estimated frequency (percentage) of participants in each biomarker group by age, sex, and
APOE ε4 status
Estimates are from a multinomial model with age, sex, and APOE ε4 status. Non-linearity in age was allowed in
the model by fi tting age as a spline with knots at ages 60, 70, and 80 years. Shaded areas are 95% parametric
bootstrap CIs.
0
10
20
30
40
50
60
70
80
90
100
Frequency (%)
APOE ε4-negative women APOE ε4-negative men
50 60 70 80 90
0
10
20
30
40
50
60
70
80
90
100
Frequency (%)
Age (years)
APOE ε4-positive women
50 60 70 80 90
Age (years)
APOE ε4-positive men
A–N–
A+N–
A–N+
A+N+
Articles
www.thelancet.com/neurology Vol 13 October 2014
1003
an individual would be unlikely to progress to A+ and to
N+ simultaneously if it were possible to sample these
imaging fi ndings in real time.
By age 89 years, about 83% of individuals with normal
cognitive function in our study had levels of amyloidosis,
neurodegeneration, or both that were similar to those
seen in mild Alzheimer’s disease. Thus, these
abnormalities seem almost an inevitable consequence of
ageing, and people are able to maintain normal cognitive
function despite these abnormalities. Typical cognitive
ageing, defi ned as remaining free of dementia, is
therefore most often characterised by the presence rather
than the absence of these imaging abnormalities. The
fact that some elderly individuals retain normal cognitive
function in the presence of substantial pathological
changes in the brain while others do not has become the
focus of much research interest.30,31
We identifi ed APOE-specifi c and sex-specifi c variations
in the frequencies of the four groups. Notably, APOE ε4
is overrepresented in A+N– and A+N+; this fi nding suggests
that APOE ε4 selectively increases risk for amyloidosis
among individuals with normal cognitive function,
which is consistent with previous fi ndings.32,33 Men are
overrepresented in A+N+ and to a lesser degree in A–N+,
and underrepresented in A+N–. This fi nding can be
interpreted as men being better able to tolerate
neurodegeneration and still retain normal cognitive
function. Alternatively, men might be at increased risk
for neurodegeneration, perhaps because of greater
lifestyle risk exposures for cerebrovascular disease34 or
greater risk of Lewy body disease,35 compared with
women. The eff ect of male sex on neurodegeneration
appears from age 65–70 years onwards, which is
consistent with an acquired rather than a developmental
eff ect. Because eff ects of both sex and APOE ε4 status
were seen on group frequencies in our study, sex and
Panel: Research in context
Systematic review
We searched PubMed for reports published in English up to
May 29, 2014, using the search terms “aging AND brain
volume”, “amyloid PET”, “aging AND amyloid PET”, and
“aging AND FDG PET”. We also checked the reference lists of
identifi ed reports for relevant publications. Previous studies
in which the investigators acquired MRI,
18F-fl uorodeoxyglucose PET, and amyloid PET in all
participants have included few individuals younger than
60 years and were done in selected volunteers rather than
population-based samples. No previous studies have assessed
the frequency of biomarker groups defi ned by β-amyloidosis
and neurodegeneration as a function of age.
Interpretation
In this cross-sectional study, we used a system that classifi es
all individuals with normal cognitive function on the basis of
both amyloidosis and neurodegeneration to generate
estimates of the frequencies of four groups (amyloid
negative and neurodegeneration negative [A–N–], amyloid
positive and neurodegeneration negative [A+N–], amyloid
negative and neurodegeneration positive [A–N+], or amyloid
positive and neurodegeneration positive [A+N+]) by age from
age 50 to 89 years in a population-based sample. Our results
showed that the frequency of A–N– decreased with age, the
frequencies of both A–N+ and A+N+ increased with age, and
the frequency of A+N– fi rst increased and then decreased with
age. These age trends are modifi ed by APOE ε4, which
increases the risk for amyloidosis, and male sex, which
increases the risk for neurodegeneration. This classifi cation
system can be used to operationalise the new International
Working Group3 and National Institute on Aging–Alzheimer’s
Association2 criteria for preclinical Alzheimer’s disease or
asymptomatic at risk for Alzheimer’s disease. The high
frequency of amyloid and neurodegeneration imaging
abnormalities in old age among individuals with normal
cognitive function shows that typical cognitive ageing,
defi ned as remaining free of dementia, is more often
characterised by the presence than by the absence of these
imaging abnormalities. We suggest a theoretical framework
to interpret changing frequencies of amyloidosis and
neurodegeneration based on the idea that people might
follow any of several diff erent possible pathophysiological
sequences while remaining cognitively normal.
Figure 4: Possible transitions from one biomarker group to another
assuming normal cognitive function is maintained
Red lines represent Alzheimer’s disease pathophysiological pathways and blue
lines non-Alzheimer’s disease pathways (ie, SNAP). (A) Pathophysiological
sequence of preclinical Alzheimer’s disease without major comorbid
non-Alzheimer’s disease pathological changes. (B) Pathophysiological sequence
of someone who fi rst develops suspected non-Alzheimer’s pathophysiology
(SNAP) neurodegeneration, then later enters the Alzheimer’s disease
pathophysiological pathway, denoted by positive amyloid PET fi ndings.
(C) Pathophysiological sequence of someone who develops SNAP
neurodegeneration without progressing into the Alzheimer’s disease
pathophysiological pathway.
AA+N–
A–N+
A–N–A+N+
BA+N–
A–N+
A–N–A+N+
CA+N–
A–N+
A–N–A+N+
Articles
1004
www.thelancet.com/neurology Vol 13 October 2014
APOE interactions, as have been previously reported,36,37
might well be expected. However, we were unable to
detect any interactions between sex and APOE ε4 status
with respect to group frequencies by age.
Our study had several limitations. We grouped
participants by imaging abnormalities that were of
suffi cient severity to be on par with the mild end (tenth
percentile) of those seen in people with Alzheimer’s
disease dementia;11 therefore, we did not capture subtle
amyloidosis or neurodegeneration below this threshold.
Amyloidosis approaches a plateau by moderate
Alzheimer’s disease dementia, whereas 18F-FDG PET and
MRI abnormalities continue to progress.1,38,39 As we have
discussed previously,28 selecting cutoff points in an
identical way for all imaging measures, as we did here,
seems rational. However, this approach will place the
cutoff point for amyloid PET at a more advanced stage in
its full dynamic range than the cutoff points for 18F-FDG
PET or MRI. Unfortunately, it is not feasible to scan
individuals in end-stage dementia and thus the
maximum abnormal values for 18F-FDG PET and MRI
cannot realistically be ascertained.
Another caveat is that our sample includes only
individuals with normal cognitive function. Certainly for
ages older than 80 years, the frequency of A–N– would be
lower and that of A+N+ higher if our sample included the
entire cognitive spectrum. Finally, our data are cross-
sectional and a more complete understanding of how
frequencies of amyloidosis and neurodegeneration
change with age, including how individuals transition
between the diff erent biomarker groups, will require
longitudinal data acquired uniformly across the entire
age spectrum.
Contributors
CRJ contributed to conceptualisation of the study, analysis and
interpretation of data, and drafting and revision of the report. HJW and
SDW contributed to conceptualisation of the study, analysis and
interpretation of data, drafting and revision of the report, and the
statistical analysis. WAR, DSK, MMM, VJL, PV, and RCP contributed to
drafting and revision of the report. MLS and JLG contributed to drafting
and revision of the report, and provided technical support. GMP
provided technical support. VSP contributed to drafting and revision of
the report and to the statistical analysis.
Declaration of interests
CRJ has provided consulting services for Janssen Research and
Development. He receives research funding from the US National
Institutes of Health (NIH; R01-AG011378, U01-HL096917, U01-
AG024904, RO1 AG041851, R01 AG37551, R01AG043392, and U01-
AG06786) and the Alexander Family Alzheimer’s Disease Research
Professorship of the Mayo Foundation. WAR receives research support
from the NIH (R01 AG034676, U01 AG006786, and P50 AG044170). DSK
serves on a data safety monitoring board for Eli Lilly; is an investigator in
a clinical trial sponsored by Janssen Pharmaceuticals; and receives
research support from the NIH (R01-AG11378, P50 AG16574, U01 AG
06786, AG 29550, AG32306, and U01 96917). VJL is a consultant for
Bayer Schering Pharma and Piramal Imaging and receives research
support from GE Healthcare, Siemens Molecular Imaging, AVID
Radiopharmaceuticals, the NIH, the Elsie and Marvin Dekelboum
Family Foundation, the MN Partnership for Biotechnology and Medical
Genomics, and the Leukemia and Lymphoma Society. VSP is funded by
the NIH (R01AG040042, U01AG06786, P50AG16574/Core C, and
R01AG32990). RCP serves on scientifi c advisory boards for Elan
Pharmaceuticals, Wyeth Pharmaceuticals, and GE Healthcare, and
receives research support from the NIH (P50-AG16574, U01-AG06786,
R01-AG11378, and U01–24904). HJW, SDW, MMM, MLS, JLG, GMP, and
PV declare no competing interests.
Acknowledgments
This study was supported by the US National Institute on Aging and the
Alexander Family Professorship of Alzheimer’s Disease Research.
References
1 Villemagne VL, Burnham S, Bourgeat P, et al. Amyloid β
deposition, neurodegeneration, and cognitive decline in sporadic
Alzheimer’s disease: a prospective cohort study. Lancet Neurol 2013;
12: 357–67.
2 Sperling RA, Aisen PS, Beckett LA, et al. Toward defi ning the
preclinical stages of Alzheimer’s disease: recommendations from
the National Institute on Aging–Alzheimer’s Association
workgroups on diagnostic guidelines for Alzheimer’s disease.
Alzheimers Dement 2011; 7: 280–92.
3 Dubois B, Feldman HH, Jacova C, et al. Advancing research
diagnostic criteria for Alzheimer’s disease: the IWG-2 criteria.
Lancet Neurol 2014; 13: 614–29.
4 Fagan AM, Xiong C, Jasielec MS, et al. Longitudinal change in CSF
biomarkers in autosomal-dominant Alzheimer’s disease.
Sci Transl Med 2014; 6: 226ra30.
5 Benzinger TL, Blazey T, Jack CR Jr, et al. Regional variability of
imaging biomarkers in autosomal dominant Alzheimer’s disease.
Proc Natl Acad Sci USA 2013; 110: E4502–09.
6 Jack CR Jr, Vemuri P, Wiste HJ, et al. Evidence for ordering of
Alzheimer disease biomarkers. Arch Neurol 2011; 68: 1526–35.
7 Nelson PT, Head E, Schmitt FA, et al. Alzheimer’s disease is not
“brain aging”: neuropathological, genetic, and epidemiological
human studies. Acta Neuropathol 2011; 121: 571–87.
8 Jagust WJ, Zheng L, Harvey DJ, et al. Neuropathological basis of
magnetic resonance images in aging and dementia. Ann Neurol
2008; 63: 72–80.
9 Fjell AM, McEvoy L, Holland D, Dale AM, Walhovd KB. Brain
changes in older adults at very low risk for Alzheimer’s disease.
J Neurosci 2013; 33: 8237–42.
10 Wirth M, Villeneuve S, Haase CM, et al. Associations between
Alzheimer disease biomarkers, neurodegeneration, and cognition
in cognitively normal older people. JAMA Neurol 2013; 70: 1512–19.
11 Jack CR Jr, Knopman DS, Weigand SD, et al. An operational approach
to National Institute on Aging–Alzheimer’s Association criteria for
preclinical Alzheimer disease. Ann Neurol 2012; 71: 765–75.
12 Jack CR Jr, Wiste HJ, Knopman DS, et al. Rates of β-amyloid
accumulation are independent of hippocampal neurodegeneration.
Neurology 2014; 82: 1605–12.
13 Jack CR Jr, Wiste HJ, Weigand SD, et al. Amyloid-fi rst and
neurodegeneration-fi rst profi les characterize incident amyloid PET
positivity. Neurology 2013; 81: 1732–40.
14 Roberts RO, Geda YE, Knopman DS, et al. The Mayo Clinic Study
of Aging: design and sampling, participation, baseline measures
and sample characteristics. Neuroepidemiology 2008; 30: 58–69.
15 St Sauver JL, Grossardt BR, Leibson CL, Yawn BP, Melton LJ 3rd,
Rocca WA. Generalizability of epidemiological fi ndings and public
health decisions: an illustration from the Rochester Epidemiology
Project. Mayo Clin Proc 2012; 87: 151–60.
16 Hixson JE, Vernier DT. Restriction isotyping of human
apolipoprotein E by gene amplifi cation and cleavage with HhaI.
J Lipid Res 1990; 31: 545–48.
17 Landau SM, Harvey D, Madison CM, et al. Associations between
cognitive, functional, and FDG-PET measures of decline in AD and
MCI. Neurobiol Aging 2011; 32: 1207–18.
18 Wilson RS, Leurgans SE, Boyle PA, Schneider JA, Bennett DA.
Neurodegenerative basis of age-related cognitive decline. Neurology
2010; 75: 1070–78.
19 Jagust W. Vulnerable neural systems and the borderland of brain
aging and neurodegeneration. Neuron 2013; 77: 219–34.
20 Chetelat G. Alzheimer disease: Aβ-independent processes-
rethinking preclinical AD. Nat Rev Neurol 2013; 9: 123–24.
21 Knopman DS, Jack CR Jr, Wiste HJ, et al. Brain injury biomarkers
are not dependent on β-amyloid in normal elderly. Ann Neurol 2013;
73: 472–80.
Articles
www.thelancet.com/neurology Vol 13 October 2014
1005
22 Rowe CC, Ellis KA, Rimajova M, et al. Amyloid imaging results
from the Australian Imaging, Biomarkers and Lifestyle (AIBL)
study of aging. Neurobiol Aging 2010; 31: 1275–83.
23 Knopman DS, Jack CR Jr, Wiste HJ, et al. Short-term clinical
outcomes for stages of NIA–AA preclinical Alzheimer disease.
Neurology 2012; 78: 1576–82.
24 Vos SJB, Xiong C, Visser PJ, et al. Preclinical Alzheimer’s disease
and its outcome: a longitudinal cohort study. Lancet Neurol 2013;
12: 957–65.
25 Braak H, Thal DR, Ghebremedhin E, Del Tredici K. Stages of the
pathologic process in Alzheimer disease: age categories from 1 to
100 years. J Neuropathol Exp Neurol 2011; 70: 960–69.
26 Duyckaerts C, Hauw JJ. Prevalence, incidence and duration of
Braak’s stages in the general population: can we know?
Neurobiol Aging 1997; 18: 362–69.
27 Jack CR Jr, Holtzman DM. Biomarker modeling of Alzheimer’s
Disease. Neuron 2013; 80: 1347–58.
28 Jack CR Jr, Knopman DS, Jagust WJ, et al. Tracking
pathophysiological processes in Alzheimer’s disease: an updated
hypothetical model of dynamic biomarkers. Lancet Neurol 2013;
12: 207–16.
29 Jack CR Jr, Knopman DS, Jagust WJ, et al. Hypothetical model of
dynamic biomarkers of the Alzheimer’s pathological cascade.
Lancet Neurol 2010; 9: 119–28.
30 Arnold SE, Louneva N, Cao K, et al. Cellular, synaptic, and
biochemical features of resilient cognition in Alzheimer’s disease.
Neurobiol Aging 2013; 34: 157–68.
31 Perez-Nievas BG, Stein TD, Tai HC, et al. Dissecting phenotypic
traits linked to human resilience to Alzheimer’s pathology. Brain
2013; 136: 2510–26.
32 Morris JC, Roe CM, Xiong C, et al. APOE predicts amyloid-beta but
not tau Alzheimer pathology in cognitively normal aging.
Ann Neurol 2010; 67: 122–31.
33 Villemagne VL, Pike KE, Chetelat G, et al. Longitudinal assessment
of Aβ and cognition in aging and Alzheimer disease. Ann Neurol
2011; 69: 181–92.
34 DeCarli C, Massaro J, Harvey D, et al. Measures of brain
morphology and infarction in the Framingham Heart Study:
establishing what is normal. Neurobiol Aging 2005; 26: 491–510.
35 Savica R, Grossardt BR, Bower JH, Boeve BF, Ahlskog JE,
Rocca WA. Incidence of dementia with Lewy bodies and Parkinson
disease dementia. JAMA Neurol 2013; 70: 1396–402.
36 Farrer LA, Cupples LA, Haines JL, et al. Eff ects of age, sex, and
ethnicity on the association between apolipoprotein E genotype and
Alzheimer disease: a meta-analysis. JAMA 1997; 278: 1349–56.
37 Altmann A, Tian L, Henderson VW, Greicius MD. Sex modifi es the
APOE-related risk of developing Alzheimer disease. Ann Neurol
2014; 75: 563–73.
38 Villain N, Chetelat G, Grassiot B, et al. Regional dynamics of
amyloid-β deposition in healthy elderly, mild cognitive impairment
and Alzheimer’s disease: a voxelwise PiB-PET longitudinal study.
Brain 2012; 135: 2126–39.
39 Jack CR Jr, Wiste HJ, Lesnick TG, et al. Brain β-amyloid load
approaches a plateau. Neurology 2013; 80: 890–96.
