Xiuming Zhang1, Elizabeth C. Mormino2, Reisa A. Sperling2, Mert R. Sabuncu3,4, and B.T. Thomas Yeo1,3,5
1ASTAR-NUS Clinical Imaging Research Centre, Department of Electrical and Computer Engineering, Singapore Institute for Neurotechnology and Memory Networks Program, National University of Singapore, Singapore, Singapore, 2Department of Neurology, Massachusetts General Hospital/Harvard Medical School, Charlestown, MA, United States, 3Martinos Center for Biomedical Imaging, Massachusetts General Hospital/Harvard Medical School, Charlestown, MA, United States, 4Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, United States, 5Centre for Cognitive Neuroscience, Duke-NUS Graduate Medical School, Singapore, Singapore
Synopsis
Alzheimer's
disease (AD) is the most common form of dementia and greatly heterogeneous. Here we develop a model of the
heterogeneity of AD-related atrophy, demonstrating that most AD dementia
patients and at-risk nondemented participants express multiple latent atrophy
factors to varying degrees. Our study also demonstrates that these atrophy
factors are associated with distinct cognitive decline trajectories across the
preclinical and clinical stages. Our results provide a framework by which
biomarker readouts could potentially predict disease progression at the
individual level. Our analytic strategy is general and might be utilized to
discover subtypes within and across other heterogeneous brain disorders.Purpose
To discover the underlying factors governing
various atrophy patterns among Alzheimer's disease (AD) patients and
investigate how these latent factors relate to longitudinal cognitive decline
across the clinical spectrum.
Methods
Voxelwise atrophy of 188 AD dementia
patients was derived via voxel-based morphometry (VBM)
1 from their
structural MRI. A Bayesian model
2 (Fig. 1) was then applied to
estimate the probabilistic atrophy maps of latent atrophy factors and the
factor composition of each patient. Subsequent analyses proceeded in three
stages. First, the relationship between factors and longitudinal mini mental
state examination (MMSE)
decline was analyzed for the AD dementia patients with a regression model
3.
We then repeated similar analyses for 190 amyloid-positive nondemented
participants. Finally, the possibility that factor information predicts
MCI-to-AD progression was explored.
Results
The model revealed a “temporal” factor associated
with extensive atrophy in the temporal lobe and hippocampus (Fig. 2A), a
“subcortical” factor associated with atrophy in the cerebellum, striatum, and
thalamus (Fig. 2B), and a “cortical” factor associated with atrophy in the
frontal and parietal cerebral cortices (Fig. 2C).
The relationship
between atrophy factors and cognitive decline was explored by examining change
in MMSE while controlling for age, sex, education, and overall brain atrophy. As
illustrated in Fig. 3A, among AD dementia patients, MMSE decline rates were markedly
different across the three factors (overall p = 1.8e-8) with the cortical
factor associated with the fastest MMSE decline. There was no significant
difference between temporal and subcortical factors. As illustrated in Fig. 3B,
among amyloid-positive nondemented participants, MMSE decline rates were
markedly different across the three factors (overall p = 9.8e-6), with the
temporal factor associated with the fastest MMSE decline. The cortical factor
was associated with faster MMSE decline than the subcortical factor (p =
7.3e-3).
We further explored the possibility that atrophy factor expressions in MCI participants predicted progression to AD dementia within three years with logistic regression while controlling for age, sex, education, and overall brain atrophy. As illustrated in Fig. 4, MCI-to-AD progression rates were significantly different across the three latent factors (overall p = 3.7e-4), with the temporal factor associated with the highest probability of progressing to AD. There was no significant difference between subcortical and cortical factors (p = 0.68).
Discussion
The observed
atrophy patterns are consistent with the VBM analyses of pathologically-defined
subtypes4 – limbic-predominant and hippocampal-sparing subtypes5.
More specifically, the atrophy pattern of the hippocampal-sparing subtype4
is consistent with our cortical atrophy factor. On the other hand, the atrophy
pattern of the limbic-predominant subtype4 is consistent with our
temporal+subcortical atrophy factor.
Our finding of
elevated decline rate among the AD dementia patients expressing the cortical
atrophy pattern (Fig. 3A) is consistent with previous neuropathology work5
that showed the fastest longitudinal cognitive decline among AD dementia
patients with the hippocampal-sparing subtype. However, analyses of
amyloid-positive nondemented participants revealed that the temporal atrophy
factor was associated with the fastest longitudinal cognitive decline (Fig. 3B).
Overall, this pattern highlights that the associations between cognitive
decline and atrophy factors vary across the clinical spectrum and emerge during
the preclinical stage of AD (Fig. 5).
One key
advantage of our modeling strategy is that individuals can express multiple
latent atrophy factors (i.e., mixed membership) rather than belonging to a
single subtype. The use of mixed membership modeling provides us with stronger
statistical power to dissociate factor-dependent atrophy maps and cognitive
decline trajectories. Since each participant expressed his or her own factor
composition (e.g., 50% subcortical, 40% cortical, and 10% temporal), the factor
composition can be thought of as an individualized subtype diagnosis of the
participant, representing a small but crucial step towards precision medicine.
Conclusion
By utilizing a novel Bayesian modeling
framework, our study revealed at least three latent AD atrophy factors with
distinct cognitive trajectories across the clinical spectrum. Consistent with
previous work, the cortical factor was associated with the fastest cognitive
decline at the AD dementia stage. However, the temporal factor was associated
with the fastest cognitive decline at the nondementia stage and the highest
probability of progressing to AD dementia. These results suggest potential
implications for prevention and monitoring disease progression. Finally, our
methodological framework is general and can be utilized to discover subtypes in
other brain disorders.
Acknowledgements
This work was supported by NUS Tier 1, Singapore MOE Tier 2
(MOE2014-T2-2-016), NUS Strategic Research (DPRT/944/09/14), NUS SOM
Aspiration Fund (R185000271720), Singapore NMRC (CBRG14nov007,
NMRC/CG/013/2013) and NUS YIA. The research also utilized resources
provided by the Center for Functional Neuroimaging Technologies,
P41EB015896 and instruments supported by 1S10RR023401, 1S10RR019307,
and 1S10RR023043 from the Athinoula A. Martinos Center for Biomedical
Imaging at the Massachusetts General Hospital. Data collection and sharing for this project was funded by the Alzheimer's
Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD
ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National
Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous
contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery
Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.;
Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd
and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer
Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research &
Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx
Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging;
Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health
Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are
facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization
is the Northern California Institute for Research and Education, and the study is coordinated by the
Alzheimer's Disease Cooperative Study at the University of California, San Diego. ADNI data are
disseminated by the Laboratory for Neuro Imaging at the University of Southern California. References
1. Ashburner J, Friston KJ. Voxel-based morphometry—the
methods. NeuroImage. 2000;11(6):805-821.
2. Blei DM, Ng AY, Jordan MI. Latent Dirichlet allocation. J
Machine Learning Res. 2003;3:993-1022.
3. Bernal-Rusiel JL, Greve DN, Reuter M, et al. Statistical
analysis of longitudinal neuroimage data with linear mixed effects models.
NeuroImage. 2013;66:249-260.
4. Whitwell JL, Dickson DW, Murray ME, et al. Neuroimaging
correlates of pathologically defined subtypes of Alzheimer’s disease: a
case-control study. Lancet Neurol. 2012;11(10):868-877.
5. Murray ME, Graff-Radford NR, Ross OA, et al.
Neuropathologically defined subtypes of Alzheimer’s disease with distinct
clinical characteristics: a retrospective study. Lancet Neurol. 2011;10(9):785-796.