THE RELATIONSHIP OF BODY MASS INDEX WITH BEHAVIOR,
BRAIN STRUCTURE AND LONGITUDINAL CHANGES IN
MILD COGNITIVE IMPAIRMENT
By
Ashley H. Sanderlin

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirement
for the degree of
Neuroscience – Doctor of Philosophy
2017

ABSTRACT
THE RELATIONSHIP OF BODY MASS INDEX WITH BEHAVIOR, BRAIN STRUCTURE
AND LONGITUDINAL CHANGES IN MILD COGNITIVE IMPAIRMENT
By
Ashley H. Sanderlin
Mild Cognitive Impairment (MCI) is a syndrome characterized by cognitive deficits that
lie on the spectrum between normal aging and dementia. The clinical course of MCI at diagnosis
is not easily predictable, because it represents a heterogeneous population. Neuropsychiatric
symptoms (NPS) and midlife obesity increase the likelihood of developing Alzheimer’s disease;
yet, these two risk factors have not been studied together in MCI. The goal of this dissertation is
to examine the relationships of weight measured by body mass index (BMI), with behavior, brain
structure, and longitudinal changes in MCI. First, we examined the relationship of obesity and
NPS in MCI. It is unknown whether obesity or related health conditions modify the risk of NPS
or severity of cognitive impairment in MCI. We found that in MCI, obese subjects were younger
and had a higher frequency and severity of affective (depression and anxiety) symptoms near the
time of diagnosis. In addition, we examined a number of obesity-related disorders to determine if
the relationship between BMI and NPS was more strongly mediated by these secondary factors
than BMI itself. We found that type-2-diabetes mellitus (T2DM) and obstructive sleep apnea,
also exhibited a specific frequency and severity of NPS. While there were no effects of obesity
on cognition, T2D subjects had lower cognitive scores and nearly double the NPS burden. Next,
we wanted to determine whether BMI had an effect on brain structure. We selected 36 regional
brain volumes related to MCI or weight from the Alzheimer’s disease Neuroimaging Initiative
(ADNI) dataset. The ADNI sample provided over 600 MCI subjects and we found a main effect
of BMI on brain volume in 14 out of 36 regions. Surprisingly, normal weight subjects had lower

brain volumes. Since normal weight subjects were significantly older we separated the sample
by middle age (55-65 years) and Seniors (>65 years) to determine if age group mediated the
effects on brain structure and found that Seniors had lower brain volumes and there was no
difference in brain structure for middle-aged subjects. Finally, we measured the relationship of
BMI on longitudinal behavioral and cognitive changes over two years and measured the survival
distributions of BMI, age, and NPS groups. Over two years NW subjects had greater cognitive
deficits. Senior subjects with low baseline NPS showed a faster progression to Alzheimer’s
dementia. These findings indicate that in MCI obese subjects may have a higher likelihood of
NPS and those that have T2D may be at risk for cognitive impairment. In addition, NW MCI
subjects may be at an increased risk for brain atrophy and lower cognitive scores. This research
may inform lifestyle interventions in regards to obesity, and clinical treatment for NPS prior to
the establishment of irreversible cognitive impairments. Further, low body weight should be
monitored in old age for progressive gray matter atrophy and cognitive decline.

For my grandmother Cornelia (Cooka) Hannah,
who showed me that “He’s able” in life and in death.

ACKNOWLEDGEMENTS

The creation of this dissertation has been a journey. I have to acknowledge those that
have played a part in guiding me to this point. I started the Neuroscience program in July of 2010
with an Early-Start Fellowship from the graduate school. Throughout my time, I have found the
graduate school at MSU to be an amazing resource for navigating the PhD, with helpful staff and
administrators. I would like to specifically thank Dr. Tony Nunez and Dr. Pero Dagbovie,
Associate Deans of the graduate school for their unwavering support.
Having a quality support system to help me focus has been a critical factor as I have
navigated my time during my program. Working with the Alliance for Graduate Education in the
Professoriate (AGEP) program has provided an intentional space for me to further connect
within the graduate school and foster meaningful relationships that proved critical to my success
during my time at MSU. Through engagements within the AGEP program, and specially
working with the program manager, Steven Thomas, I have been able to maximize my PhD
experience. Additionally, there have been many friends that have shaped my time as a graduate
student that were not in my program, but provided resources, and ultimately became examples
for me as I worked to complete my degree. Thank you, Tiffeny, Sakeena, Paula, Kamarah and
Carmel.
I have to also thank my committee who has provided guidance by critiquing the
development of my research questions and has served as a sounding board for numerous
technical, theoretical, statistical and clinical questions. First, my advisor and committee chair Dr.
Bozoki has been a tremendous resource in providing me with first-hand experience working with
MCI and Alzheimer’s patients, as well as giving critical feedback on clinical methodology and
scientific writing. Dr. Todem has been a great help in developing statistical models and
v

reviewing analyses procedures for many of the dissertation chapters. Dr. Zhu provided guidance
on learning neuroimaging procedures, which included study design and analyses. This support
ranged from scanning patients in an MRI machine to group analyses of those images with the use
of multiple imaging modalities and software. Finally, Dr. Symonds has been a resource in
developing neuroscience aims and methods for cognitive decline and dementia as it relates to
neuroscience. I am thankful for the individual and collective support of my committee over the
years.
During graduate school, I had the privilege of mentoring undergraduate research
assistants.. These students volunteered their time to gain research experience as they developed
into future researchers and professionals. In addition to their work on my projects, I required that
they ask questions of their own and present on their independent research projects. My
experience as a mentor has been one I have enjoyed immensely, as I had the opportunity to
utilize my work as a platform to assist others develope and grow. With that, I would like to thank
Cort, Jill, Andrea, Michelle and Mehma for their support.
Last, but certainly not least, I must acknowledge and thank my amazing family. My
husband, Marcus Sanderlin Sr., my ‘Michigan blessing’ has guided me to the end of this season
with great love and patience. In addition to his active career and responsibilities as husband and
father, he has been my sounding board, coach, practice audience and greatest motivator. My son,
Marcus Sanderlin Jr., has been my inspiration in the final year to accomplish my goals with a
positive spirit. His light, energy and charm have given me a boost and serve as an example for
him, that despite numerous obstacles, through faith and focus, all things are possible. This faith
began in my home with my parents, Kevin and Kim Hannah and brother Kevin Hannah II. They
have been a boundless support throughout my entire educational journey; from the home daycare

vi

my mother ran to completing the PhD. My family has instilled many values in me, one of which
was a respect of education and love of learning. Each step they have supported and believed that
I could accomplish anything that I set out to do. The demonstration of their love has made me
believe that I can truly do anything I put my mind to.
Finally, I would like to thank my grandmother Cornelia Hannah and aunt Sheila
Thompson. They were a mother-daughter team of educators and entrepreneurs that I had the
privilege of seeing daily, work hard, create, design, implement, and dream. Together they
attained a Master’s in Education degree and made a great impact on their family, community,
students, employees, and especially me. Before #BlackGirlMagic was a popular saying or I
understood the true meaning of “Girl Power,” I had them as an example of educated and
innovative Black women. To my family listed and my numerous supportive aunts, uncles and
cousins that I could not, thank you.

vii

TABLE OF CONTENTS
LIST OF TABLES

xi

LIST OF FIGURES

xiii

KEY TO ABBREVIATIONS

xv

INTRODUCTION
Mild Cognitive Impairment
Neuroimaging in Mild Cognitive Impairment
Neuropsychiatric symptoms in Mild Cognitive Impairment
Neuroimaging of neuropsychiatric symptoms
Obesity and cognitive deficits
Neuroimaging of obesity
Comorbidity of obesity and neuropsychiatric symptoms
Summary

1
2
4
5
6
7
9
10
11

CHAPTER 1
13
OBESITY AND CO-MORBID CONDITIONS ARE ASSOCIATED WITH SPECIFIC 13
NEUROPSYCHIATRIC SYMPTOMS IN MILD COGNITIVE IMAPAIRMENT
Introduction
14
Methods
15
MCI diagnosis
15
BMI groups
16
BMI-related disorders
16
Neuropsychiatric symptoms
17
NPI-Q clusters
17
MCI severity
18
Statistical analysis
18
Results
19
Demographics
19
BMI-related disorders
21
NPI-Q clusters
22
Discussion
24
Limitations
27
Conclusions
28
CHAPTER 2
THE EFFECT OF BODY MASS INDEX ON BRAIN STRUCTURE IN
MILD COGNITIVE IMPAIRMENT
Introduction
Methods
Participants
BMI groups

viii

30
30
31
33
33
34

ADNI imaging data acquisition
ADNI FreeSurfer methods
FreeSurfer region of interest analysis
Statistical analysis
Results
Participants
Brain volume
Discussion
Limitations
Conclusions
APPENDIX

35
35
36
37
37
37
38
43
46
47
48

CHAPTER 3
THE INTERACTION OF BODY MASS INDEX, NEUROPSYCHIATRIC
SYMPTOMS AND AGE IN MILD COGNITIVE IMAPIRMENT: A
LONGITUDINAL STUDY
Introduction
Methods
Participants
Factors
Longitudinal analysis methods and statistical analyses
Survival analysis methods and statistical analyses
Results
Longitudinal analysis
Survival analysis
Discussion
Limitations
Conclusions
APPENDIX

50

CHAPTER 4
ALTERNATE METHODS
THE EFFECT OF OBESITY ON BRAIN WHITE MATTER IN MILD
COGNITIVE IMPAIRMENT
Introduction
Methods – MSU COGENT
Participants
Data collection overview
MRI acquisition of COGENT data
DTI analyses
MRI acquisition in 6 diffusion weighted directions
Tract Based Spatial Statistics (TBSS)
Statistical analysis
Results – MSU COGENT
Methods – ADNI

73
73

ix

50
51
53
53
54
55
56
57
58
60
65
66
67
68

74
74
76
76
76
76
77
78
78
79
80
81

Participants
MRI acquisition of ADNI data
DTI analyses
Statistical analysis
Results – ADNI
Summary and conclusions – MSU COGENT & ADNI DTI analysis
THE EFFECT OF OBESITY ON BRAIN CORTICAL THICKNESS
IN MILD COGNITVE IMPAIRMENT
Introduction
Methods
Participants
Cortical thickness analysis
FreeSurfer region of interest analysis
Statistical analysis
Results
Brain cortical thickness
Summary and conclusions – Thickness averages
APPENDIX

81
82
82
83
83
85

CHAPTER 5
CONCLUSIONS AND FUTURE DIRECTIONS

97
97

REFERENCES

101

x

88
88
89
89
89
90
90
90
91
92
94

LIST OF TABLES
TABLE 1.1.
Demographic, cognitive and behavioral measures of the MCI sample grouped by BMI

20

TABLE 1.2.
The frequency of BMI-related disorders within BMI groups

21

TABLE 1.3.
Demographic, cognitive and behavioral measures of T2D and OSA groups

23

TABLE 2.1.
39
Demographic, cognitive, functional and behavioral measures for all MCI subjects and for each
BMI group
TABLE 2.2.
40
Significant cortical and subcortical volume differences across BMI groups of MCI subjects
TABLE 2A.
Selected FreeSurfer brain regions for volume analysis across BMI groups

49

TABLE 3.1.
58
Demographic, cognitive and behavioral characteristics for all ADNI MCI subjects and across
BMI groups
TABLE 3.2.
59
Longitudinal changes in behavior, cognitive and functional test scores over two years for all MCI
subjects and across BMI groups
TABLE 3A.
69
Demographic, cognitive, cardiovascular, and behavioral characteristics comparing Middle-Age
and Senior MCI subjects
TABLE 3B.
70
Demographic, cognitive, cardiovascular, and behavioral characteristics comparing male and
female MCI subjects
TABLE 3C.
71
Demographic, cognitive, cardiovascular, and behavioral characteristics comparing MCI subjects
with low and high NPI-Q scores
TABLE 3D.
72
Demographic, cognitive, cardiovascular, and behavioral characteristics comparing MCI subjects
with low and high GDS scores

xi

TABLE 4.1.
84
White matter tracts that significantly differed across BMI groups for fractional anisotropy and
mean diffusivity measures
TABLE 4.2.
91
Brain regions that significantly differed in cortical thickness average measures across BMI
groups of MCI subjects
TABLE 4A.
95
Regions of interest from the JHU ‘EVE’ atlas white matter tract list generated in FreeSurfer
TABLE 4B.
Brain regions of interest for the cortical thickness analysis generated in FreeSurfer

xii

96

LIST OF FIGURES
FIGURE 1.1.
29
The NPI-Q cluster frequency and severity of BMI, T2D and OSA MCI subject groups.
The 12 NPI-Q symptoms domains are clustered into 4 groups of, Hyperactivity (agitation,
disinhibition, irritability, motor disturbances and euphoria), Apathy (apathy, appetite), Affective
(depression, anxiety) and Psychosis (delusions, hallucinations, night-time behaviors). A) The
frequency of each NPI-Q cluster is plotted for BMI, T2D and OSA groups. Cluster frequency
statistics were conducted using the chi-square test of independence (2 degrees of freedom for
BMI and 1 degree of freedom for T2D and OSA). B) The mean (SE) severity of NPI-Q clusters
for BMI, T2D and OSA groups. Mean differences in cluster severity were compared using the
analysis of variance (ANOVA) model for BMI and a student’s t-test for T2D and OSA.
Significant associations are marked as follow, * p < 0.05, ** p < 0.01, *** p < 0.001.
Abbreviations: NPI-Q, Neuropsychiatric Inventory Questionnaire; MCI, mild cognitive
impairment; BMI, body mass index; T2D, type 2 diabetes; OSA, obstructive sleep apnea.
FIGURE 2.1.
42
Cortical brain regions that significantly differed by BMI. Subcortical regions that also differed
by BMI are not shown: the hippocampus (F = 5.1, p = .006) and amygdala (F = 4.9, p = .008).
The effect of BMI on brain volumes were analyzed using a MANOVA model correcting for age
and education. Regions range in their significance with light yellow regions meeting statistical
significance of p< 0.05 and dark red having a p value of p < 0.001.
FIGURE 2.2.
43
The correlation of Precuneus volume with BMI and age. BMI was positively associated with
precuneus volume whereas age was negatively associated. The associations were measured using
a Pearson’s correlation with significance set at p < 0.05.
FIGURE 3.1.
61
Kaplan-Meier curves comparing the rate of survival of BMI groups from baseline MCI status to
the diagnosis of Alzheimer’s type dementia. The cumulative survival of three BMI groups were
compared over 24 months or 4 study visits. Crosses indicate censored events.
Abbreviations: NW, normal weight; OW, overweight; OB, obese.
FIGURE 3.2.
62
Kaplan-Meier curves comparing the rate of survival of age groups from baseline MCI status to
the diagnosis of Alzheimer’s type dementia. The cumulative survival of Middle Age and Senior
groups were compared over 24 months or 4 study visits. Crosses indicate censored events.
FIGURE 3.3.
63
Kaplan-Meier curves comparing the rate of survival of BMI groups factored by age group from
baseline MCI status to the diagnosis of Alzheimer’s type dementia. The cumulative survival of
Middle Age and Senior subjects within three BMI groups of normal weight, overweight and
obese were compared over 24 months or 4 study visits. Crosses indicate censored events.

xiii

FIGURE 3.4.
64
Kaplan-Meier curves comparing the rate of survival of age groups factored by high and low NPS
groups from baseline MCI status to the diagnosis of Alzheimer’s type dementia. (A) NPI-Q
groups of high and low symptom burden compare the cumulative survival of Middle Age and
Senior MCI subjects. (B) GDS groups of high and low symptom burden compare the cumulative
survival of Middle Age and Senior MCI subjects. The low group represents total test scores
between 0 and 3 and the high scores are ≥ 4. The cumulative survival of age factored by NPS
group were compared over 24 months or 4 study visits. Crosses indicate censored events.
Abbreviations: NPS, neuropsychiatric symptoms; NPI-Q, Neuropsychiatric Inventory
Questionnaire; GDS, Geriatric Depression Scale.
FIGURE 4.1.
80
The 25-direction MSU-COGENT FA results comparing NW/OW and OB groups. Multiple
corrections were computed per voxel and overlaid on the mean FA skeleton. The green trace
indicates the mean FA comparison between groups and is non-significant shown in a coronal
section on the left and a sagittal section on the right. Abbreviations: S, superior; R, right; L, left;
I, inferior.

xiv

KEY TO ABBREVIATIONS
MCI

Mild Cognitive Impairment

AD

Alzheimer’s Disease

BMI

Body Mass Index

NPS

Neuropsychiatric Symptoms

NPI-Q

Neuropsychiatric Inventory Questionnaire

GDS

Geriatric Depression Scale

NW

Normal Weight

OW

Over-Weight

OB

Obese

T2D

Type-2-Diabetes

OSA

Obstructive Sleep Apnea

MSU

Michigan State University

COGENT

Cognitive and Geriatric Neurology Team

ADNI

Alzheimer’s Disease Neuroimaging Initiative

MRI

Magnetic Resonance Imaging

DTI

Diffusion Tensor Imaging

JHU

John’s Hopkins University

FSL

FMRIB’s Software Library

TBSS

Tract Based Spatial Statistics

DMN

Default Mode Network

CCN

Cognitive Control Network

U.S.

United States
xv

MMSE

Mini Mental Status Examination

ADAS-cog 13

Alzheimer’s Disease Assessment Scale – Cognitive 13

CDR

Clinical Dementia Rating Scale

ANOVA

Analysis of Variance

MANOVA

Multivariate Analysis of Variance

SD

Standard deviation

SE

Standard error

xvi

INTRODUCTION

1

Mild Cognitive Impairment
Mild Cognitive Impairment (MCI) is a syndrome characterized by cognitive deficits that
lie on the spectrum between normal aging and dementia. The clinical course of MCI at diagnosis
is not easily predictable, because as a whole, MCI represents a heterogeneous population.
However, MCI is associated with an increased likelihood of developing Alzheimer’s disease
(AD), with an average conversion rate to dementia of 10-15% per annum over five years. There
is an increased risk of AD for amnestic (memory-impaired) MCI subtypes 1–4. The high risk of
developing AD makes the study of MCI a priority for better understanding prodromal states of
dementia. This is largely based on the theory that early cognitive deficits are the result of an
underlying neuropathology.
The diagnostic criterion for MCI were originally established by Ron Petersen and
colleagues in 19991. A few years later these criteria were revised to capture the heterogeneity in
the cognitive deficits presented. MCI is defined by the following criteria; (a) the subject does not
have normal cognition and is not demented (b) cognitive deterioration is evident as reported by
the subject and an informant or objectively measured over time (c) activities of daily living are
preserved and complex instrumental functions are either intact or minimally impaired5. Potential
outcomes of MCI over time include; (a) stable MCI, never progressing to AD or other forms of
dementia (b) AD; (c) other dementias (i.e. Frontotemporal dementia (FTD), vascular dementia
(VaD), Lewy body dementia (LBD)); and lastly, in some cases, (d) reversion to normal
cognition.
Due to the heterogeneity of MCI, identifying factors within this population that increase
the risk for dementia are essential. These factors or biomarkers, measured characteristics that are
indicative of an underlying biological state, condition or disease process, are an important area of

2

dementia research. Biological, behavioral and neuroimaging biomarkers of an underlying AD
pathology have been previously reported in MCI and include decreased cortical thickness and
volume of the hippocampus and medial temporal lobe6, hypometabolism of the posterior
cingulate cortex (PCC)7, amyloid beta deposition8,9, Apolipoprotein e4 allele status10,11 and
depression.
All individuals who develop AD pass through the transient state of MCI and express
varying degrees of behavioral, cognitive and brain structure changes. Clinical symptoms include
mild impaired function in daily activities, behavior and mood. In addition to these factors
epidemiological studies have indicated that age, being female and decreased educational
attainment increase the risk for the development of AD and conversion from MCI12,13 . It is now
known that the changes that occur in the transition from normal cognition to AD begin decades
prior to clinical symptoms of dementia14,15.
Cognitive subtypes are the only recognized subgroups of MCI. In MCI, subtypes that
have prominent memory deficits, amnestic MCI (aMCI) are at an increased risk for dementia16,17,
compared to non-amnestic MCI and multi-domain MCI subtypes. However, patients from all
subtypes of MCI can convert to AD despite their initial cognitive impairments. Many research
studies and clinical trials focus predominantly on amnestic MCI patients in order to study early
stages of AD, yet non-cognitive factors that are highly prevalent in MCI are beginning to gain
more attention. Non-cognitive risk factors such as behavioral changes and metabolic disorders
may provide additional understanding of the underlying pathology of MCI and allow for the
construct of a profile of specific subgroups that may be at an increased risk for AD.

3

Neuroimaging in Mild Cognitive Impairment
Neuroimaging has been an important tool in understanding the pathophysiology of MCI and
AD18–22. AD is considered to be a disease of the limbic system23,24. Primary regions altered due
to AD include the cingulate gyrus, hippocampus, parahippocampal gyrus (PHG), entorhinal
cortex and fornix. Changes within these regions have been quantified using various
neuroimaging techniques to gauge structural and functional deficits. Structural neuroimaging
techniques include volumetric MRI (vMRI) which assess the integrity of grey matter regions,
and diffusion tensor imaging (DTI) which allows for the estimation of microstructural white
matter fiber tract integrity18,21,25. Brain atrophy in limbic regions, the bilateral hippocampus,
amygdala, fornix, and parahippocampal gyri on vMRI have been correlated with cognitive
deficits of MCI subjects25,26. Most prominently, volumetric measures of the hippocampus
correlate with decreased Mini Mental Status Exam (MMSE)27 scores and positively predict
conversion to AD28. In addition, diffusion imaging has allowed for better understanding of the
structural changes of white matter tracts and neuronal health. The disease process of AD and
MCI encompass significant white matter changes including decreased volume, lesions and
hyperintensities in fornix, cingulum and frontal white matter24,29–31.
Functional imaging techniques include measures of cerebral glucose metabolism, task
dependent activation of brain regions and functional connectivity of brain networks at rest,
through the use of flourodeoxyglucose – Positron Emission Tomography (fdg-PET), stimulationbased functional MRI (fMRI) and resting state fMRI (rsfMRI) respectively. A reduction in
glucose metabolism in the posterior cingulate cortex (PCC) is a reliable early clinical biomarker
for AD pathology in MCI32,33. Many studies have analyzed the reductions in task-dependent
activation in brain regions affected by AD in memory retrieval, attention, and executive

4

processing tasks 34–36. Neurodegenerative diseases such as AD also have altered functional
connectivity of the brain at rest, with reductions of region and network based connectivity among
temporal, parietal and frontal regions compared to healthy controls37,38. However, in some
regions increased connectivity is seen and speculated to be a compensation for loss of function in
regions affected by AD pathology. Resting state fMRI measures blood oxygen level dependent
(BOLD) signal fluctuations of the brain at rest 36.
In theory, regions that are functionally connected exhibit similar or correlated activity at
rest. The most widely researched resting state network is the Default Mode Network (DMN) 39.
Brain activity related to internal thoughts such as autobiographical memory and thinking of the
future utilize DMN brain regions. The DMN shows increased activity at rest, is deactivated
during tasks, and is anchored in the posterior cingulate cortex (PCC) and medial prefrontal
cortex (mPFC) with prominent brain nodes of connectivity in the medial temporal lobe (MTL)
and angular gyrus. The network connectivity of the DMN is altered in normal aging and illness,
including MCI19,40,41. Discovery of the DMN has launched the search for other resting state
networks involved in numerous cognitive and behavioral disorders42. Among these is the
cognitive control network (CCN)43. The CCN is involved in attention, working memory, and self
– control. The CCN has major nodes within the dorsal anterior cingulate cortex (ACC),
dorsolateral prefrontal cortex (dlPFC), and portions of the parietal lobe. In functional imaging
studies the DMN is deactivated in the presence of a task where the CCN is active in related tasks
making them anti-correlated networks37. Network and seed based analysis of CCN regions
correlate with the degree of executive dysfunction, more prominent in naMCI21.

Neuropsychiatric symptoms in Mild Cognitive Impairment
Research studies indicate that individuals with MCI have an increased prevalence of

5

neuropsychiatric symptoms (NPS) compared to the normal population44–46. Apathy, depression
and anxiety are the most prevalent NPS in MCI4,44,47,48. In cognitively normal older adults, NPS
such as depression and apathy correlate with cognitive decline and the development of MCI at a
higher rate than cognitively normal subjects without NPS25,49,50. In MCI, apathy symptoms are
correlated with conversion to AD by almost 7 times, the presence of multiple symptoms on the
neuropsychiatric inventory-questionnaire (NPI-Q) have an additive effect on progression to AD4.
There is also evidence that NPS in general may be a marker of MCI severity, decreasing the time
to progression to dementia by almost 2.5 times44,51. Neuropsychiatric symptoms in MCI
measured by the NPI-Q with cut off scores of 0-3 and ≥4 are a reliable indicator of group
differences; scores ≥4 are more likely to be aMCI with increased medical comorbidities and
functional impairments52. As cognitive scores decline over time, NPS and functional
impairments increase from MCI to AD and appear to be a result of brain damage. Further, the
most frequently reported NPS in AD also overlap with those in MCI including, apathy and
depression for both disorders50,53,54. There may be a link in the transition to dementia between
cognitive decline and NPS prevalence and severity. Thus, behavioral symptoms lie on the
continuum of AD pathology, making their presence in early cognitive impairments a feature of
important research investigation. Proper identification of NPS syndromes in MCI may aid in
discerning the etiology of MCI subtypes, and provide a more reliable prognosis at the time of
diagnosis.
Neuroimaging of neuropsychiatric symptoms
Neuroimaging has aided in the quantification of brain changes in the presence of
psychiatric symptoms for many years. Many studies have sought to identify specific brain
regions and neural networks important in emotional regulation. Brain imaging of depression with

6

and without MCI has been widely researched and has provided evidence for structural changes in
grey and white matter related to affect. For example, white matter microstructural changes in
depression include decreased fractional anisotropy (FA) in white matter sub-adjacent to the
ACC, superior frontal gyri, left middle frontal gyrus55,56. In a direct investigation of NPS in MCI,
low FA values in the anterior cingulate were most related to NPI-Q symptoms of irritability,
agitation, depression, apathy and nighttime behaviors. Another study using DTI in anxiety
disorders found decreased FA values of the uncinate fasciculus (UF) and the inferior longitudinal
fasciculus57. The UF tract connects the amygdala and orbitofrontal cortex, important affective
brain regions that demonstrate disrupted brain connectivity in the presence of NPS26,58. In
addition cortical brain atrophy of subjects with high NPS is seen in the ACC, orbitofrontal and
dorsal parietal lobe regions20. Psychopathology also attenuates the functional connectivity (FC)
of the brain. Interestingly, FC in depressed patients increases within the thalamus and subgenual
cingulate59. An example of this has been demonstrated in AD where, Balthazaar and colleagues
found increased FC in mild AD patients between the ACC and anterior insula54, which was
related to NPS symptoms of hyperactivity (agitation, irritability, aberrant motor behavior,
euphoria and disinhibition)60.

Obesity and cognitive deficits
Obesity is a prevalent health condition in the United States affecting 36.5% of adults and
17% of youth (ages 2 – 19)61 and is becoming more prevalent worldwide. Obesity is disorder
characterized by excess body fat and low energy expenditure that is associated with an increased
risk for health problems. The prevalence of obesity in the US has dramatically increased over the
last 30 years and is highest among middle age (aged 40-65) and older adults (aged 66 and older).
Obesity often occurs co-morbidly with conditions such as, Type 2 Diabetes, vascular diseases,

7

heart disease and sleep apnea, and it is the leading cause of work disability62. Additional side
effects include decreased global brain volume, a high risk for the metabolic syndrome, a high
likelihood of comorbid NPS and premature death63. Obesity is considered a modifiable health
condition with the cause related to a combination of excessive food intact, lack of physical
activity and genetic susceptibility64. A common measure of obesity is body mass index (BMI),
which calculates body fat by taking into account a person’s weight divided by the square of their
height. There are 3 BMI groups, normal weight, overweight and obese. Body mass index is not
as accurate as measure of waist to hip ratio or detailed body fat assessments65, but it is widely
used in research studies on obesity as it allows for a quick assessment of body fat using standard
clinical measures of height and weight.
Multiple lines of evidence indicate a link between obesity and the development of
neurodegenerative diseases66–70. Most importantly, obesity in midlife is associated with an
increased likelihood of developing AD 54,67. Obesity is association with changes to the central
nervous system, including changes in appetite-regulating hormones, cortical and subcortical
brain volumes reductions and increased deficits across age groups71,72. These may be a mediating
factor in the development of cognitive impairments and eventually dementia as a result. Altered
hormone signaling of leptin and ghrelin acting on the hypothalamus often occur in obesity73,74.
Further, studies have shown decreased total brain volume and altered processing within brain
regions involved in cognitive control, such as the CCN. Most studies examining the relationship
between obesity and cognition have examined brain differences within weight groups of healthy
middle aged and older adults or are retrospective studies that link obesity in middle age with an
ultimate conversion to dementia later in life. However, few studies have investigated obesity
within the transitional cognitive state of MCI.

8

Neuroimaging of obesity
There is evidence that chronic obesity affects the structure and function of the brain.
Brain regions associated with obesity include those within networks of cognitive control. This
association is hypothesized to be a cause of overeating due to a lack of inhibition which
ultimately leads to obesity75–77. Primary brain regions of the CCN include the dlPFC, anterior
cingulate cortex, orbitofrontal cortex, amygdala, hippocampus and nucleus accumbens78,79.
Obesity in elderly adults is associated with decreased total grey matter volume80. In
neuroimaging studies on obesity across age groups obese adults have lower brain volumes in
many AD related regions, such as the temporal lobe, hippocampus, cingulate cortex, dorsolateral
prefrontal cortex and posterior parietal cortex75,81. Thus, the correlation of brain volume decline
and obesity may be a sensitive measure in those that develop a cognitive impairment. For
example, adults obese in midlife exhibit significantly decreased brain volume in predominantly
frontal and temporal lobes and in measures of global atrophy as they age80,81. Further, a study
measuring the cortical thickness of regions in the CCN showed that between obese, non-obese
and successful weight losers (maintenance of significant weight loss for a minimum 3 years),
obese subjects had significant cortical thinning in the following regions: anterior insula, ventral
striatum, rostral ACC and ventromedial PFC. In cognitively normal older adults, obesity has also
been shown to decrease grey matter volume within the orbitofrontal cortex, anterior cingulate
gyrus, hippocampus and basal ganglia. These results remained even after controlling for diabetes
status, hypertension, and white matter hyperintensities80. To date, only one article has discussed
the effect of obesity on brain structure specifically within MCI subjects. Ho et. al. examined the
effects of obesity on brain volume in MCI and early AD subjects and found that BMI was

9

directly correlated with increased brain atrophy in frontal, temporal and parietal regions. Further,
every point increase in BMI was associated with a 0.5% - 1.5% decrease in brain volume82.
Diffusion tensor imaging of white matter structural integrity depict difference in white
matter tracts between normal weight and obese individuals. Most prominently, measures of
decreased FA of the fornix and corpus callosum in obese subjects has been seen in multiple
studies83–85. In addition a recent study using DTI showed white matter atrophy and decreased FA
values among obese subjects within the inferior frontal gyrus, temporal gyri, insular cortex,
occipital gyri and amygdala85. Overall there was a negative relationship between body fat and
WM volume.
Literature on the functional connectivity (FC) of the obese brain in adults is limited and
currently has not been investigated in MCI. However in adolescents, fMRI studies have shown
activation of the insula and portions of the operculum in regard to the anticipation of food77. In a
similar procedure in young adults, food anticipation (cravings) resulted in increased brain
activation in the hippocampus, insula and caudate86.
Comorbidity of obesity and neuropsychiatric symptoms
Despite the difference in factors that contribute to the onset of obesity or NPS
independently, they often occur as comorbid conditions across age groups87. Compared to
normal weight adults, obese subjects are more likely to exhibit symptoms of depression, apathy
and anxiety. Adults with severe psychiatric disorders exhibit a prevalence of obesity of
approximately 50 percent, compared to the national average of 30%88. Further, obesity and NPS
can be predictive of one another. For example, with depression are more likely to be obese as
adults89 and obese adolescents are more like to develop psychiatric symptoms comorbid with
obesity in adulthood90. A five-year longitudinal study of adults 50 and older measured the

10

relationship between obesity and depression and found that obese subjects were twice as likely to
be depressed at the end of 5 years and the depressed patients at baseline were more likely to be
obese after 5 years91. However, depression did not predict obesity at follow up. The relationship
between NPS and obesity in the literature is unclear and it is unknown how they interact in MCI.
The proposed neurocircuitry of both NPS and obese involve dominant pathways from the
prefrontal cortex with relays through the cingulate gyrus79,92–94, yet the interactive effects of
obesity and NPS on brain volumes have rarely been studied, especially in older adults, and not at
all in MCI. In MCI, NPS and obesity have different times of onset and possible course of
pathology: where as obesity is likely to cause changes in brain structure, NPS may be a symptom
of such changes. A link between obesity and NPS has been hypothesized as low-grade
inflammation95. Chronic inflammation is believed to lead to neurodegeneration by affecting the
expression of amyloid-beta precursor protein96. While both obesity and NPS increase the risk of
AD they have not been studied together in MCI. This dissertation will investigate the
relationship of obesity, measured by body mass index, and NPS across clinical, neuroimaging
and longitudinal measures in MCI subjects.
Summary
Since the original description of MCI in 1999, research in the field has focused on
longitudinal follow-up of MCI samples, and refining the psychometric and eligibility criteria for
MCI and its subtypes in order to maximize predictive sensitivity and specificity for the disorder
over time. Despite the heterogeneity of MCI, subtypes (aMCI/naMCI) are based solely on
cognitive criteria. The majority of research investigates aMCI/naMCI-related cognitive changes
of MCI and their relationship to disease progression, neuroimaging markers and behavior, but
this only captures a piece of the puzzle. Obesity affects the central nervous system and is also

11

associated with cognitive and behavioral changes. Despite the high prevalence of obesity across
age groups this factor is relatively absent from the body of work that seeks to characterize and
provide an understanding for MCI and relationships between the brain and behavior. This
problem is exacerbated by the fact that there is no cure or effective treatment for either MCI or
AD. The most popular and widely distributed drug for AD, Aricept (donepezil) produces small
improvements in the cognitive symptoms of AD, and is relatively ineffective in MCI97,98.
Identifying non-cognitive markers or risk factors for AD may provide clinicians with an avenue
for concentrated interventions for particularly high-risk individuals within the larger MCI
population.
Thus, the goal of this dissertation is to bring together non-cognitive modifiable risk
factors for AD that may further characterize behavioral, brain, and longitudinal changes in MCI.
This research may inform lifestyle interventions in regards to obesity, and clinical treatment for
NPS prior to the establishment of irreversible cognitive impairments. Prior to this collection of
research, obesity and NPS have not been studied together in MCI. There is a lack of knowledge
regarding the comorbidity of obesity and NPS in MCI, the interaction of obesity and NPS on
brain structure and white matter integrity, and disease progression related to these factors. This
dissertation will address these gaps in knowledge by using body mass index as a measure of
adiposity in MCI subjects. The following research aims will,: 1) identify the prevalence of
obesity in MCI and its comorbidity with NPS; 2) assess volumetric and white matter brain
changes related to obesity; and 3) examine longitudinal changes in cognitive, functional and
behavioral measures and progression from MCI to AD.

12

CHAPTER 1
OBESITY AND CO-MORBID CONDITIONS ARE ASSOCIATED WITH SPECIFIC
NEUROPSYCHIATRIC SYMPTOMS IN MILD COGNITIVE IMPAIRMENT

This chapter has been published in manuscript form: Sanderlin AH, Todem D and Bozoki
AC (2017) Obesity and Co-morbid Conditions Are Associated with Specific Neuropsychiatric
Symptoms in Mild Cognitive Impairment. Front. Aging Neurosci. 9:164.
doi:10.3389/fnagi.2017.00164
13

Introduction
Behavioral changes or neuropsychiatric symptoms are prevalent in Mild Cognitive
Impairment (MCI) and are associated with an increased likelihood of conversion to dementia99.
MCI is a transitional state between normal cognition and dementia and the presence of
neuropsychiatric symptoms (NPS) predict the progression of MCI to AD, decreasing the time of
progression to dementia by 2.5 fold 4,50. Neuropsychiatric symptoms such as depression, anxiety
and apathy are a hallmark of Alzheimer’s disease (AD) 48,100. As high as 80% of AD patients
have at least one symptom on the Neuropsychiatric inventory with affective and apathy
symptoms having the highest prevalence53,101. In MCI, depression is one of the most prevalent
symptoms and has been directly related to cognitive decline and the development of
dementia102,103.
Obesity is a common disorder characterized by excess adipose tissue and is associated
with cognitive deficits and an increased likelihood of developing dementia when present at
midlife104. The prevalence of obesity in the U.S. has nearly tripled over the last 30 years and is
highest among middle age and older adults 105. Side effects of chronic obesity include lower
global brain volume, a high risk for metabolic syndrome, and premature death 63. Further, obesity
affects cognition 66 and often occurs co-morbidly with NPS across age groups 87. Additional
conditions, consequences of obesity such as type 2 diabetes, sleep apnea and other vascular
disorders are also associated with cognitive decline and increased neuropsychiatric symptoms.
Multiple lines of evidence demonstrate a link between midlife obesity and the development of
dementia 66–68. However, the relationship between NPS and obese subjects and cognitive decline
within early MCI has not been studied.
Neuropsychiatric symptoms and obesity have not been measured together to determine
their co-morbidity in MCI and interactions with cognition. In the present study, our hypothesis is

14

that in MCI, obesity is associated with higher total NPS scores and a higher prevalence and
severity of affective symptoms (depression and anxiety), as well as more extensive cognitive loss
as measured by MCI severity compared to normal weight subjects. We sought to first identify the
prevalence of obesity, obesity-related health conditions, and NPS within MCI, examining their
relationship and their effect on the severity of cognitive impairment. We then clustered similar
NPS together, and examined the frequency and severity of behavioral clusters across weight
groups and BMI-related health conditions.
Methods
All study data came from medical records dating between 2004 and 2014 from a tertiary
geriatric neurology clinic at Michigan State University serving the mid-Michigan area. Clinical
and behavioral data were taken at the time of diagnosis of MCI. This study involved minimal
risk to human subjects and a waiver of consent was requested and approved by the Michigan
State University Institutional Review Board.
MCI diagnosis
The diagnosis of MCI was determined according to Petersen’s Criteria 1 by an expert
neurologist (A. Bozoki). The diagnostic process included an initial clinical evaluation by the
neurologist followed by a neuropsychological assessment battery, MRI (head CT if MRI was
contraindicated) and serologic testing for metabolic profile, thyroid function and vitamin B12
level. The neuropsychological assessment battery (a modified CERAD battery 106, described in
further detail in the MCI Severity section), assessed memory, verbal and visual delayed recall,
language, visuospatial and executive functions, and was administered to all subjects. Subjects
scoring ≥ -1.5 standard deviations (SD) below the education and age-adjusted mean in one or
more cognitive domains were classified as MCI. The MCI sample represented a heterogeneous
population consisting of amnestic MCI, non-amnestic MCI and multi-domain MCI subtypes.
15

Inclusion criteria for this study were as follows: subjects were between the ages of 50-95
years, able to speak, comprehend and read English with at least 8 years of education, and a Mini
Mental Status Examination (MMSE) 27 score between 24 - 30. Subjects were excluded if they
had a history of a coexisting central nervous system disorder or uncontrolled depression that
could account for the cognitive impairment, any uncontrolled or unstable medical condition, and
alcohol or substance abuse within the last two years. Exclusion criteria were determined based
on medical records review. Over the 10-year period, there were 667 subjects with
neuropsychometric data. Of the 667 subjects examined, 117 were diagnosed with MCI. A total of
4 subjects were excluded from the study due to a history of major depression (n = 3) and stroke
(n = 1). Our final sample consisted of 113 subjects that met the inclusion criteria.
BMI groups
The MCI sample was grouped by traditional BMI criteria: normal weight (NW; BMI
18.5- 24.9), overweight (OW; BMI 25 - 29.9) or obese (OB; BMI ≥ 30). Height (in inches) and
weight (in pounds) measurements were taken at the time of clinical diagnosis of MCI. BMI was
converted to the unit kg/m2 using the follow calculation, [(Weight (lb.) / Height2 (in.)) x 703].
BMI-related disorders
A clinical history of BMI-related disorders was recorded in order to account for
conditions that may be comorbid with increased weight 96 but pose an independent risk factor for
cognitive decline 107, or have an increased prevalence of neuropsychiatric symptoms 108. These
included, type 2 diabetes (T2D), hypertension (HTN), hyperlipidemia (HL), gastroesophageal
reflux disease (GERD) and obstructive sleep apnea (OSA). The presence or absence of each of
these conditions was recorded for each subject at the time of MCI diagnosis. In addition, blood
pressure recordings at the time of diagnosis were used to calculate a mean arterial pressure
(MAP) value for each subject as a measure of cardiovascular health.
16

Neuropsychiatric symptoms
Neuropsychiatric symptom were measured using the Neuropsychiatric Inventory
Questionnaire (NPI-Q) 109 and the Geriatric Depression Scale –short form (GDS) 110. The NPI-Q
is a validated measure for assessing behavioral disturbances across 12 different domains in a
brief caregiver-reported questionnaire 109. These include; delusions, hallucinations,
agitation/aggression, depression/dysphoria, anxiety, elation/euphoria, apathy/indifference,
disinhibition, irritability, aberrant motor behavior, sleep and nighttime behavioral changes, and
appetite and eating disorders. An informant familiar with the subject reported NPI-Q symptoms,
by rating each symptom first for their presence (yes / no), and then severity (range of 1 - 3) with
a total of 36 possible points. Behavioral changes reported on the NPI-Q reflect symptoms present
within one month of testing. The self-reported 15-point GDS scale was used for further
quantification of depressive symptoms. Mild NPS was designated as a total score ≥ 1 and
moderate NPS as ≥ 4 for each test. Quantification of NPS burden was measured based on the
total symptom score for each test and the prevalence of mild and moderate symptoms groups.
NPI-Q clusters
Neuropsychiatric Inventory- Questionnaire symptoms were grouped into clusters based
on a prior research study demonstrating that specific neuropsychiatric inventory (NPI; original
full test version) symptoms tend to cluster together in their prevalence and severity when they
emerge as part of a dementia 60. The benefits of assessing NPI/NPI-Q clusters instead of
individual symptoms include both examination of underlying similarities in prevalence,
progression of symptoms and biological correlates111. Thus, in the present study the 12 NPI-Q
symptoms were grouped into 4 clusters: Hyperactivity (agitation, disinhibition, irritability, motor
disturbances and euphoria), Psychosis (delusions, hallucinations, night-time behaviors), Apathy
(apathy, appetite), and Affective (depression, anxiety). The presence of a symptom cluster
17

required the presence of at least one symptom within each cluster. The cluster severity was the
average of the total score (0-3) across each symptom within a cluster for each subject.
MCI severity
To determine whether BMI groups were associated with an increase in MCI severity
(MCI-SV), a z-score was computed for each cognitive test in the neuropsychological test battery,
then averaged to obtain a mean overall z score for each subject. Included test measures evaluated
global cognition (MMSE; Modified Mini Mental Exam (3ME) 112), memory (CERAD Word
List, immediate/delayed/recognition 113), language (CERAD 15-item Boston Naming Test 114;
categorical and phonemic verbal fluency 115), executive function (Trail Making Test 116; Stroop
117

), and visuospatial tests (CERAD Constructional Praxis, immediate/delayed 118).

Statistical analysis
The analysis of variance (ANOVA) model was used to compare NPI-Q total score, GDS
score, NPI-Q cluster severity and MCI severity scores across BMI groups. These comparisons
were further adjusted for age and education using the analysis of covariance (ANCOVA) model.
Specific BMI-related disorders that had a high prevalence of obesity were also used as
independent variables. A chi-square test of independence was conducted to compare the
frequency of NPI-Q clusters across BMI groups and BMI-related disorders. A Fisher’s exact test
was used to compare frequencies of NPI-Q clusters between groups when cell sample sizes were
small. A direct examination of NPI-Q score differences between NW and OB were measured
using a Student’s t-test. And the relationship of BMI as a continuous variable with NPI-Q was
measured using a Pearson’s correlation. Statistical analysis was conducted using SPSS software
(Hewlett Packard; Palo Alto, CA). A two-sided p-value less than 5% (p<0.05) was used for
statistical significance.

18

Results
Demographics
A description of the study sample is reported in Table 1.1. Of the 113 MCI subjects
included in the study, 110 had available BMI data and roughly 1/3 each were NW, OW and OB.
Overall, the average BMI, mean age and MMSE of the sample was 27.4 kg/m2, 74.1 years and
26.5 respectively. Over 90% of the sample was Caucasian with an average educational
attainment of 14.6 years. Surprisingly, NW subjects were significantly older than OW and OB (p
< 0.001). Overall, 78.6 % of subjects had at least one symptom on the NPI-Q and 87.3% had one
symptom on the GDS. BMI was positively correlated with NPI-Q score (Pearson’s r = 0.225; p =
0.04), and a direct comparison of NW and OB groups revealed a significantly higher prevalence
of NPI-Q symptoms (Student’s t = 2.05; p = 0.045, unadjusted). However, there was not an
effect of BMI on NPS or cognitive measures in the ANOVA model.

19

TABLE 1.1. Demographic, cognitive and behavioral measures of the MCI sample grouped
by BMI
Entire
Characteristic
NW
OW
OB
Statistic
Group
N = 113
N = 38
N = 39
N = 33
Χ2 or F P value
74.3 (0.72)

78.74 (1.00)

72.4 (1.33)

71.2 (1.12)

11.97

<0.01b,c

53 (47)

20 (53)

20 (51)

11 (33)

3.23

0.20

Education
(yrs.)
MMSE

14.6 (0.31)

15.6 (0.49)

14.2 (0.49)

13.8 (0.64)

2.86

0.06

26.5 (0.17)

26.0 (0.24)

26.8 (0.29)

26.7 (0.33)

2.24

0.11

MCI-Severity a

-0.89(0.06)

-1.04 (0.09)

-0.80 (0.09)

-0.82 (0.14)

1.78

0.18

NPI-Q score a

5.2 (0.56)

4.0 (0.72)

5.4 (1.02)

6.7 (1.27)

0.47

0.63

≥ 1, n (%)

68 (79)

25 (74)

26 (87)

17 (77)

1.72

0.42

≥ 4, n (%)

41 (48)

14 (41)

13 (43)

14 (64)

3.05

0.22

3.0 (0.26)

3.0 (0.35)

2.7 (0.35)

3.3 (0.73)

0.63

0.53

≥ 1, n (%)

91 (88)

34 (90)

33 (92)

22 (79)

2.70

0.26

≥ 4, n (%)

32 (31)

14 (37)

9 (25)

8 (29)

1.29

0.53

Age (yrs.)
Female, n (%)

GDS score a

Abbreviations: MCI, Mild Cognitive Impairment; BMI, body mass index; MCI SV, MCI
severity; MMSE, Mini-Mental State Examination; MCI-SV, MCI severity; GDS, Geriatric
Depression Scale; NPI-Q, Neuropsychiatric Inventory Questionnaire; NW, normal weight; OW,
overweight; OB, obese.
NOTE. Values are presented as mean (standard error) for continuous variables and n (%) for
categorical variables. The statistic is chi-square test of independence for categorical variables (2
degrees of freedom) and an ANOVA F statistic for continuous variables. Significance was set as
p < 0.05.Available behavioral data were as follows: NPI-Q, n = 86 and GDS, n = 102.
Subscript a indicates adjustment for covariates age and education; b, indicates significant
difference between NW and OB, c indicates significant difference between NW and OW.

20

BMI-related disorders
The frequency of all examined BMI-related disorders are displayed in Table 1.2. There
was no difference in HTN, HL, GERD and MAP across BMI groups. However, a significantly
higher proportion of T2D and OSA subjects were OB compared to NW and OW. Thus, T2D and
OSA were used as independent variables in further analysis of individual NPI-Q cluster
frequency and severity.

TABLE 1.2. The frequency of BMI-related disorders within BMI groups
BMI-related
Entire
NW
OW
OB
Χ2 or F
disorder
Group
GERD
14%
16%
13%
27%
0.24

P value
0.89

HP

43%

34%

44%

52%

2.18

0.34

HTN

53%

53%

49%

58%

0.56

0.76

OSA

21%

5%

15%

46%

18.37

<0.001b,c

T2D

21%

11%

13%

42%

13.26

0.001 a

MAP

95.4 (1.1)

94.2 (2.2)

95.0 (1.8)

97.2 (1.9)

0.55

0.58

Abbreviations: MCI, Mild Cognitive Impairment; BMI, body mass index; T2D, Type 2
diabetes; HTN, hypertension; HLD, hyperlipidemia; OSA, obstructive sleep apnea; GERD,
gastroesophageal reflux disease; MAP, mean arterial pressure; NW, normal weight; OW,
overweight; OB, obese. NOTE. Values reported as percentages for categorical variables and
mean (SE) for continuous variables. The statistic is chi-square test of independence (2 degrees
of freedom) for categorical variables and an ANOVA F statistic for continuous variables.
Subscript b, indicates significant difference between NW and OB, c indicates significant
difference between NW and OW.
The demographics data of subjects with and without T2D and OSA are presented in
Table 1.3. Age and MCI-SV were similar between groups although education and MMSE score
were lower in subjects with T2D. The NPI-Q mean total score was significantly higher in
subjects with T2D and OSA. Further, the prevalence of moderate level NPI-Q symptoms differed
based on the presence of T2D and OSA. Depression scores measured by the GDS were also
21

significantly higher for T2D subjects. There was no difference in age, education, MMSE, MCISV or GDS across OSA groups.
NPI-Q clusters
The prevalence and severity of specific NPI-Q clusters differed with respect to MCI
subjects who were obese, and had T2D or OSA. The Hyperactivity cluster was the most frequent
with 56% of subjects having at least one symptom. Figure 1.1A shows the frequency of each
symptom cluster across groups. Affective symptoms significantly differed between OB and NW
groups (X2 2 = 6.76, p = 0.03). Subjects with sleep apnea also showed a significantly higher
frequency of solely Affective symptoms (X21 = 5.39, p = 0.02). Diabetic subjects had a
significantly higher frequency across 3 clusters, Affective (X2 1 = 8.85, p = 0.003), Hyperactivity
(X21 = 14.19, p <0.001) and Psychosis (X21 = 3.74, p = 0.05) symptoms. A posterior power
analysis was then conducted to measure the strength of the association between obesity, T2DM,
and OSA with each NPI-Q clusters. For our significant comparisons of OB and OSA with
Affective symptoms had a power of 60% and 56% respectively. In addition, for each significant
association of T2DM, as the cluster significance decreased, the power of the association
increased: Psychosis, power = 50%, Affective, power = 86% and Hyperactivity, power = 98%.
The mean severity of NPI-Q clusters across groups is shown in Figure 1.1B. In OB
subjects, Affective symptoms were also more severe (F = 3.30, p = .04) along with Psychosis
cluster (F = 4.55, p = 0.03). Subjects with OSA also had a higher severity of Psychosis
symptoms (Student’s t = 2.50, p = 0.02) as well as Apathy (Student’s t = 2.17, p = 0.03)
compared to those without a sleep disorder. Two NPS clusters were more severe in diabetic
subjects: Affective (Student’s t = 2.11, p = 0.04), and Psychosis (Student’s t = 2.52, p = 0.02);
Apathy and Hyperactivity was unrelated to this condition.

22

TABLE 1.3. Demographic, cognitive and behavioral measures of T2D and OSA groups
BMIDisorder
Disorder
Statistic
Characteristic
related
P value
Present
Absent
Χ2 or F
disorder
T2D
73.0 (1.25)
74.6 (0.87)
1.06
0.29
Age (yrs.)
OSA
72.7 (1.38)
74.7 (0.86)
1.07
0.287
T2D
12 (52.2)
40 (45.5)
0.33
0.57
Female, n (%)
OSA
7 (30.4)
45 (51.1)
3.14
0.076
T2D
13.3 (0.77)
14.9 (0.33)
2.01
0.047
Education
(yrs.)
OSA
15.0 (0.86)
14.4 (0.33)
0.64
0.45
T2D
29.9
(1.08)
26.7
(0.47)
3
0.003
BMI (kg/m2)
OSA
30.65 (0.92)
26.52 (0.48)
3.95
< 0.001
T2D
25.8 (0.36)
26.7 (0.19)
2.13
0.035
MMSE
OSA
26.74 (0.37)
26.40 (0.19)
0.83
0.411
T2D
-1.04 (0.15)
-0.86 (0.06)
1.21
0.23
MCI-Severity
OSA
-0.73 (0.18)
-0.93 (0.06)
1.27
0.208
T2D
7.63 (1.18)
4.48 (0.62)
2.38
0.019
NPI-Q score
OSA
7.75 (1.37)
4.59 (0.60)
2.23
0.028
T2D
17 (89.5)
51 (76.1)
1.6
0.338
≥ 1, n (%)
OSA
14 (87.5)
54 (77.1)
0.84
0.505
T2D
14 (73.7)
27 (40.3)
6.61
0.01
≥ 4, n (%)
OSA
11 (68.8)
30 (42.9)
3.5
0.061
T2D
4.26 (0.87)
2.67 (0.26)
2.36
0.02
GDS score
OSA
3.76 (0.85)
2.77 (0.25)
1.12
0.274
T2D
18 (94.7)
71 (85.5)
1.18
0.453
≥ 1, n (%)
OSA
18 (85.7)
71 (87.7)
0.06
0.812
T2D
9 (47.4)
22 (26.5)
3.18
0.075
≥ 4, n (%)
OSA
8 (38.1)
23 (28.4)
0.74
0.389
Abbreviations: BMI, body mass index; MMSE, Mini-Mental State Examination; MCI-SV, MCI
severity; GDS, Geriatric Depression Scale; NPIQ, Neuropsychiatric Inventory Questionnaire;
T2D - type 2 diabetes; OSA, obstructive sleep apnea. NOTE. Results are presented as mean (SE)
for continuous variables and n (%) for categorical variables. The statistic is chi-square test of
independence (1 degree of freedom) for categorical variables and an independent samples t test, t
statistic for continuous variables.

23

Discussion
In this study, we assessed the relationship of weight with specific NPS and the severity of
cognitive impairment in MCI subjects. Our hypothesis was supported, in part, in that the
frequency and severity of affective NPS were significantly higher in obese MCI subjects. To our
knowledge, a direct examination of the relationship between BMI and NPS in MCI has not been
reported, although the prevalence of each BMI group in our sample is similar to national
averages of overweight and obese individuals in the adult US population 119. While there is a
growing body of literature on the effects of obesity on behavioral symptoms and cognition we
sought to include in our analysis obesity related conditions which often occur co-morbidly104 and
are believed to share neuropathological commonalities. Interestingly, HTN, HL and GERD
proved not to be significantly related to obesity in our sample; they were present in relatively
equal proportions in all 3 BMI groups (although HL showed a definite trend toward increase).
This likely speaks to the multifactorial nature of these conditions, such that the contribution of
obesity is only one of several driving factors. Our results indicate that in MCI the combination of
increased weight with T2D showed the greatest differences in behavioral disturbances in regards
to total scores, symptom cluster frequency and severity as well as changes in global cognition.

In our sample of early stage MCI subjects, BMI and related health conditions
demonstrated a significantly higher prevalence and severity of specific NPS. Previous studies
have reported depression, anxiety and apathy symptoms as the most frequent NPS seen in MCI,
among obese persons 87,120 as well as in subjects with T2D 121 and OSA 122. Our study supports
these findings in that the Affective cluster (depression, anxiety) was more frequent in subjects
with OB, T2D and OSA compared to those that were NW/OW or without T2D and OSA. The
Affective cluster was also rated with greater severity for OB and T2D subjects, which leads to

24

our main finding that that there is a relationship between depression, anxiety and obesity in MCI.
For all groups the Psychosis cluster had a significant difference in mean severity although it was
only more prevalent in the T2D group. A possible explanation may be that despite delusions and
hallucinations constituting the least frequent symptoms in MCI 46,48, their presence in the early
stage of cognitive impairment is perceived more severely by the informant. Further, the presence
of nighttime behaviors in this cluster were most likely the driving factor: nighttime behaviors
were the most frequent and severe individual NPI-Q symptom present in 42% of subjects. The
Apathy cluster did not differ in frequency in any of the group comparisons; however it did have a
higher severity only in OSA subjects. Higher apathy in relationship to daytime sleepiness has
been shown in OSA subjects 123, which may explain the heightened severity rating when present
in this group. Thus, when assessing MCI subjects with behavioral disturbances, consideration
should be given to higher BMI and BMI-related health conditions, specifically T2D and OSA, as
possible contributors to the presentation of NPS. Future research will be necessary to determine
whether lifestyle interventions and treatment of weight related disorders affect the persistence
and severity of NPS over time.

The link between weight-related health conditions and NPS is not well understood. As
with similar findings between these health conditions and cognition, current research has begun
to identify central inflammation as a possible mechanism. One theory postulates that weight gain
modulates adipocyte function resulting in a higher secretion of pro-inflammatory markers that
reach the brain and alter neuronal function, ultimately leading to alterations in neurocircuitry and
neural plasticity. These changes affect brain regions such as the prefrontal cortex and cingulate
gyrus, resulting in the presentation of neuropsychiatric symptoms 74. Moreover, a recent animal
study showed a possibly direct effect of obesity on dopamine receptor function resulting in

25

depression-like behaviors and alterations in reward circuitry 124. In MCI, obesity is considered to
be a chronic condition yet the time course of the indirect changes described by the mechanism of
central inflammation is unclear. Further research is needed to understand whether weight-related
brain changes present in conjunction with the onset of cognitive impairment or whether they act
separately to promote the presentation of neuropsychiatric symptoms.

In contrast to our hypothesis, MCI severity was not associated with BMI or T2D and
OSA groups. One explanation may be that MCI is defined by cognitive impairment and
represents a transitional state with a narrow range of deficits. There is a cut-off to the severity
that reflects mild cognitive impairments before one achieves psychometric criteria for dementia.
Moreover our study subjects are diagnosed as MCI by a stringent criterion of -1.5 SD in at least
one cognitive domain, which in other studies has been broader, (e.g., -1.0 SD in 2 cognitive
domains). This difference in criteria may provide a more uniform assessment of overall MCI
severity. Another possibility is that an overall severity score is not a sufficiently nuanced
measure of cognitive status. Diabetes and OSA show greater cognitive deficits in executive
function than memory. It may be more effective to measure individual cognitive domain severity
in order to detect differences in the effects of disorders such as T2D, OSA and even OB. Finally,
overall MCI severity may differentiate groups later in the disease course, which cannot be
examined in a cross-sectional design. However, one research study showed that MCI subjects
with at least one symptom on the NPI-Q or GDS, and lower initial cognitive status resulted in a
more rapid development of dementia 102. In our T2D subjects, general cognition measured by the
MMSE was significantly lower (p = .03) while NPI-Q and GDS total scores were nearly doubled
compared to subjects without T2D. This may indicate that MCI subjects with T2D and NPS ≥ 4
are at an increased risk for conversion to dementia.

26

A surprising finding of this study was the lower mean age of overweight and obese MCI
subjects at the time of diagnosis by nearly 7 years. Middle age obesity promotes a higher risk of
conversion to AD 125. Since our data come from newly diagnosed MCI patients, this suggests
that higher adiposity may cause an earlier emergence of cognitive impairment, likely through the
burden of additional physiologic stressors. Further, an early onset of cognitive impairment in the
obese may create a group of individuals susceptible to the onset of AD at an earlier age
compared to those of normal weight. Despite many OB subjects having T2D and OSA, there was
not a difference in the age at diagnosis based on these conditions. Isolating risk factors associated
with weight may unmask features of the underlying pathological changes associated with
prodromal AD.
Limitations
There are some limitations that must be taken into account in interpreting our results.
First, this study is cross sectional and therefore does not assess cognitive and NPS status over
time in relationship to BMI groups. For the same reason, it also cannot evaluate the direction of
the association or capture the initiation and persistence of NPS. Second, due to the sample being
a specialty referral clinical, genetic testing was not routinely done to establish apolipoprotein
allele status, chronicity of overweight and obesity, or effectively capture socio economic status.
In addition, our examination of NPI-Q symptomology was based on a score of ≥ 1, which is a
very mild disturbance. However, recent studies have shown that even the measurement of the
presence or absence of symptoms can predict disease progression 126. In this regard, it is notable
that large differences were seen in the frequency and severity of NPI-Q clusters with respect to
BMI-related disorders T2D and OSA.
Lastly, in our analysis we did not have adequate power to detect some of our associations
between OB and OSA with affective symptoms. The results provided in this article allow for the
27

generation of hypothesis for future work. This initial investigation provides new information
about possible co-morbidities in MCI that can be replicated in a larger sample such as the
Alzheimer’s Disease Neuroimaging Initiative127. The current focus of this research group is to
further analyze the relationships of NPS and OB in MCI using a more rigorous epidemiologic
approach with subjects from the ADNI dataset. Future studies will focus on longitudinal followup to examine whether there is a relationship between weight, NPS prevalence and MCI severity
at later stages of MCI and early AD, and also to establish whether a higher BMI produces a
greater incidence of NPS over time.
Conclusions
This study demonstrates that within MCI, BMI and related disorders T2D and OSA
showed a higher rate of psychopathologic changes, most particularly in the Affective,
Hyperactivity and Psychosis clusters. Further, increased late life adiposity, which represented
over 65% of subjects, was associated with a lower mean age at the onset of cognitive symptoms.
Future research should focus on better understanding the intersection of NPS and OB in MCI, as
well as the combined effect of these disorders and BMI-related disorders on the brain and
clinical progression of MCI. In clinical settings, diabetic patients with MCI should be monitored
for behavioral changes.

28

FIGURE 1.1. The NPI-Q cluster frequency and severity of BMI, T2D and OSA MCI subject
groups. The 12 NPI-Q symptoms domains are clustered into 4 groups of, Hyperactivity
(agitation, disinhibition, irritability, motor disturbances and euphoria), Apathy (apathy, appetite),
Affective (depression, anxiety) and Psychosis (delusions, hallucinations, night-time behaviors).
A) The frequency of each NPI-Q cluster is plotted for BMI, T2D and OSA groups. Cluster
frequency statistics were conducted using the chi-square test of independence (2 degrees of
freedom for BMI and 1 degree of freedom for T2D and OSA). B) The mean (SE) severity of
NPI-Q clusters for BMI, T2D and OSA groups. Mean differences in cluster severity were
compared using the analysis of variance (ANOVA) model for BMI and a student’s t-test for T2D
and OSA. Significant associations are marked as follow, * p < 0.05, ** p < 0.01, *** p < 0.001.
Abbreviations: NPI-Q, Neuropsychiatric Inventory Questionnaire; MCI, mild cognitive
impairment; BMI, body mass index; T2D, type 2 diabetes; OSA, obstructive sleep apnea.

29

CHAPTER 2

THE EFFECT OF BODY MASS INDEX ON BRAIN STRUCTURE IN
MILD COGNITIVE IMPAIRMENT

30

Introduction
Obesity is a growing epidemic affecting 30 million people worldwide and 38% of adults
in the United States128. In the U.S., the rate of obesity is growing fastest among older adults ages
40 – 80. Obesity affects the peripheral and central nervous system resulting in an increased risk
of metabolic syndrome, cognitive deficits, behavioral disturbances, and Alzheimer’s disease
(AD)68,129. Obesity also affects brain structure. Differences in brain volume have been shown in
obese (OB) compared to lean controls72,80,83. A few studies have shown lower regional brain
volume or decreasing volume over time for OB subjects across age groups82,130,131 yet cognitive
deficits in OB are not consistently present across young, middle and old age adults132,133.
Previous research in aging studies have demonstrated a high body mass index (BMI) in midlife
(45-65 years) correlating with increased cognitive impairment and a higher likelihood of
developing dementia later in life68,134, but high BMI in late life (>65 years) correlating with no
changes in cognition and a slower progression to dementia135,136. This relates to the obesity
paradox, which is defined as a counterintuitive relationship of obesity with a seemingly positive
health outcome135,136. These findings indicate that there is an interaction of weight and age on
cognition, and cognitive decline.
In mild cognitive impairment (MCI), an intermediate state between normal cognition and
dementia1, the effect of age and weight on brain structure has yet to be explored. All MCI
subjects are at an increased risk for dementia5. Research on MCI subjects seeks to identify
individuals at highest risk for conversion by assessing cognitive factors, biomarkers, and
behavioral and metabolic influences. The combination of these factors makes MCI a
heterogeneous group of individuals far beyond the three primary cognitive subtypes of amnestic
(memory dominant), non-amnestic (non-memory, i.e. language) and multi-domain MCI. A

31

multipronged approach to understanding prodromal dementia states is necessary due to the
possibility of multiple outcomes that reflect different underlying pathologies.
Research focusing on non-cognitive symptoms such as metabolic disorders (ie. obesity,
type-2-diabetes, and hypertension) and general behavioral changes has grown rapidly over the
last 10 years. Further, age and weight each affect the susceptibility to cognitive and structural
brain changes. The age of onset of MCI ranges from 55-90 years and spans middle age to old
age. There is nearly a 3-fold increase (Hazard ratio = 2.7) in the risk of dementia for middle age
obese subjects with normal cognition68,125,134. However, in studies on patients later in life with
MCI, low BMI results in greater cognitive decline over time. The apparent susceptibility of
normal and underweight individuals to faster decline in MCI is not well understood. Based on
previous research this study seeks to measure the relationship of weight and age on brain
structure in order to find biological differences that might explain this relationship.
Mild cognitive impairment and obesity affect brain volume in specific brain regions. The
majority of atrophic changes related to amnestic MCI are localized to the medial temporal lobe.
Specifically, entorhinal cortex and hippocampus atrophy are sensitive measures of early amnestic
impairment and are predictive of AD pathology82. Only one study has measured the effects of
BMI on brain volume in a medium sized sample of MCI subjects. Ho et al measured the
relationship of BMI and brain volume using a voxel based analysis of the whole brain. Their
results indicated that as BMI increased, brain volume decreased within frontal, temporal and the
parietal lobes. This complements research that has also found a negative correlation between
BMI and brain volume in early adulthood and in older adults without cognitive impairments80,137.
These studies found significant changes in whole brain volume, frontal lobe, occipital lobe and
the temporal lobe (specifically the hippocampus) volume. Since decreased brain volume is a

32

biomarker for AD, lower brain volumes related to weight in MCI may make individuals more
vulnerable to dementia.
The Alzheimer’s disease Neuroimaging Initiative (ADNI) is a multisite longitudinal
study that has collected research data from people with normal cognition, MCI and AD to better
understand the development of AD138. The power of the ADNI lies in the thorough clinical,
neuroimaging and biomarker data collected on now over 1,000 study volunteers, allowing for the
analysis of critical questions related to understanding MCI and AD through the use of a large
sample. More information on ADNI can be found in the Methods section.
For this study, the ADNI’s processed volumetric MRI data, analyzed by ADNI
researchers via the software FreeSurfer allowed for the comparison of region-specific brain
volumes across BMI groups of MCI subjects139. Our main research question was: how does BMI
relate to brain structure, if at all, in MCI subjects? Based on previous research in older adults and
MCI subjects72,80–82,137 we predicted that in MCI a higher BMI will correlate with lower brain
volume.
Methods
Participants
Behavioral, height, weight, and MRI data were obtained from The Alzheimer’s Disease
Neuroimaging Initiative (ADNI) database138 (adni.loni.ucla.edu). The ADNI was launched in
2003 by the National Institute on Aging (NIA), the National Institute of Biomedical Imaging and
Bioengineering (NIBIB), the Food and Drug Administration (FDA), private pharmaceutical
companies and non-profit organizations, as a $60 million, 5-year public-private partnership. The
primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI),
positron emission tomography (PET), other biological markers, and clinical and
neuropsychological assessment can be combined to measure the progression of MCI and early
33

AD. Determination of sensitive and specific markers of very early AD progression is intended to
aid researchers and clinicians to develop new treatments and monitor their effectiveness, as well
as lessen the time and cost of clinical trials.
ADNI subjects were recruited from over 50 sites across the U.S. and Canada. The initial
goal of ADNI was to recruit 800 subjects but ADNI has been followed by ADNI-GO and ADNI2. To date these protocols have recruited over 1500 adults ages 55 to 90, to participate in the
research, consisting of cognitively normal older individuals, people with early or late MCI, and
people with early AD. Clinical, behavioral and neuroimaging data are collected for each patient
with up to 54 months of follow up per patient. MRI methods, procedures and preprocessing by
ADNI have been previously described140 and can also be accessed at the following website: afniinfo.org
Inclusion criteria for MCI set forth by ADNI, included; a MMSE score between 24 – 30,
a subjective memory complaint by the patient or caregiver, objective memory loss measured by
the Wechsler Memory Scale Logical Memory II, a global Clinical Dementia Rating (CDR) of
0.5, preserved activities of daily living, and the absence of dementia.

BMI groups

The MCI sample was grouped by traditional BMI criteria: normal weight (NW; BMI
18.5- 24.9 kg/m2), overweight (OW; BMI 25 - 29.9 kg/m2) or obese (OB; BMI ≥ 30 kg/m2).
Height (in inches) and weight (in pounds) measurements were taken at the time of clinical
diagnosis of MCI. BMI was converted to the unit kg/m2 using the follow calculation, [(Weight
(lb.) / Height2 (in.)) x 703].

34

ADNI imaging data acquisition
The ADNI MRI method protocol has been previously published139. Briefly, all subjects
underwent whole-brain MRI scanning on 3-Tesla GE Medical Systems scanners, on at least one
of two occasions (baseline and 6 months). T1-weighted IR-FSPGR (spoiled gradient echo)
sequences (256×256 matrix; voxel size = 1.2×1.0×1.0 mm3; TI=400 ms; TR = 6.98 ms; TE =
2.85 ms; flip angle = 11°), were collected as well as diffusion-weighted images (DWI; 35 cm
field of view, 128×128 acquired matrix, reconstructed to a 256×256 matrix; voxel size:
2.7×2.7×2.7 mm3; scan time = 9 min; more imaging details may be found at
http://adni.loni.usc.edu/wpcontent/uploads/2010/05/ADNI2_GE_3T_22.0_T2.pdf).

ADNI FreeSurfer methods
Cortical reconstruction and volumetric segmentation was performed with the Freesurfer
image analysis suite, which is documented and freely available for download online
(http://surfer.nmr.mgh.harvard.edu/). The technical details of these procedures are described in
prior publications141–154. Briefly, this processing includes motion correction and averaging 153 of
multiple volumetric T1 weighted images (when more than one is available), removal of nonbrain tissue using a hybrid watershed/surface deformation procedure152, automated Talairach
transformation, segmentation of the subcortical white matter and deep gray matter volumetric
structures (including hippocampus, amygdala, caudate, putamen, ventricles)145,146 intensity
normalization155, tessellation of the gray matter white matter boundary, automated topology
correction144,156, and surface deformation following intensity gradients to optimally place the
gray/white and gray/cerebrospinal fluid borders at the location where the greatest shift in
intensity defines the transition to the other tissue class141–143. Once the cortical models are
complete, a number of deformable procedures can be performed for further data processing and

35

analysis including surface inflation147, registration to a spherical atlas which is based on
individual cortical folding patterns to match cortical geometry across subjects148, parcellation of
the cerebral cortex into units with respect to gyral and sulcal structure149,157, and creation of a
variety of surface based data including maps of curvature and sulcal depth. This method uses
both intensity and continuity information from the entire three dimensional MR volume in
segmentation and deformation procedures to produce representations of cortical thickness,
calculated as the closest distance from the gray/white boundary to the gray/CSF boundary at each
vertex on the tessellated surface143. The maps are created using spatial intensity gradients across
tissue classes and are therefore not simply reliant on absolute signal intensity. The maps
produced are not restricted to the voxel resolution of the original data thus are capable of
detecting sub millimeter differences between groups. Procedures for the measurement of cortical
thickness have been validated against histological analysis158 and manual measurements159,160.
FreeSurfer morphometric procedures have been demonstrated to show good test-retest reliability
across scanner manufacturers and across field strengths150,154.
FreeSurfer region of interest analysis
The regions of interest (ROI) output by FreeSurfer and made available on the ADNI
website was downloaded for analysis of regional volumes across weight and age groups. Data
was obtained from the UCSFFX spreadsheet of FreeSurfer Version 5.1. Subjects that had a nonaccelerated T1 screening MRI, whose status was “complete” and had an overall quality check
(QC) of “Pass” for all QC regions were included in the analyses. Brain regions that related to
MCI and BMI were selected for analysis, resulting in 36 regions. Each brain region was
corrected for the total intracranial volume to eliminate subject specific differences in brain region
volumes that are related to total brain size. The corrected value for each brain region per subject
was then used for statistical analysis.
36

Statistical analysis
Regional brain volumes, behavioral, and clinical data of MCI subjects at screening or
baseline was downloaded from the ADNI database. The main comparison groups were BMI
categorized as normal weight (NW, BMI: <25 kg/m2), overweight (OW, BMI: 25 – 29.9 kg/m2)
and obese (OB, BMI: > 30 kg/m2). Thirty-six cortical and subcortical regional brain volumes
were selected due to previously reported relationships with either BMI or MCI and averaged
between hemispheres. The raw brain volumes were corrected for head size by dividing by the
total intracranial volume (ICV). An analysis of variance (ANOVA) model was used in each
experiment to compare cognitive and behavioral scores, and regional brain volumes across BMI
groups. A multivariate ANOVA model corrected for age and education. Lastly, a correlation
analysis measured the relationship of brain volume with BMI, age, MMSE, CDR-sum of boxes
(CDR-SB) scores.

Results
Participants
The MCI sample consisted of 635 subjects with baseline regional brain volume measures.
The BMI groups consisted of 216 (34%) NW, 282 (44%) OW and 137 (22%) OB. Two
underweight subjects with a BMI of 17.64 and 17.85 were included in the normal weight group.
Overall, the MCI cohort had a mean BMI of 27.1, age of 71.9 years, education of 15.9 years, and
43% were female. Obese subjects were significantly younger than NW, with lower educational
attainment and a higher mean GDS score. Age and BMI values were negatively correlated (r = 0.144, P < 0.001). There were 140 Middle Age (age between 55 and 65) subjects and 494
Seniors (age greater than 65). The demographic characteristics of the MCI subjects as whole and
by BMI groups are reported in Table 2.1.

37

Brain volume
Baseline regional brain volumes were compared across BMI groups to determine whether
there was a difference in mean brain volumes. First, a MANOVA model measured the main
effect of BMI on regional brain volumes. There was a statistically significant difference in brain
volume across BMI group, F (72, 1194) = 1.36, p = 0.026; Wilks’Λ = 0.854, partial η2 = 0.08.
Next, age and education were added as covariates, based on previous research indicating their
relationship with brain volume. In this full factorial model, BMI, age and education were all
independent significant contributors to volume differences in the overall multivariate tests (BMI:
F (72, 1186) = 1.31, p = 0.048; Wilks’Λ = 0.858, partial η2 = 0.07; Age: F (36, 593) = 9.17, p <
.005; Wilks’Λ = 0.642, partial η2 = 0.36; Education: F (36, 593) = 2.15, p < .005; Wilks’Λ =
0.885, partial η2 = 0.12). In the full factorial model, 14 (out of 36) regions significantly differed
by BMI. Each of the regions that significantly differed are displayed in Figure 2.1.

38

TABLE 2.1. Demographic, cognitive, functional and behavioral measures for all MCI subjects
and for each BMI group
All MCI

NW

OW

OB

Statistic
F or Χ2

P value

N = 635

N = 216

N = 282

N = 137

71.89 (7.46)

73.03 (7.22)

71.94 (7.41)

69.95 (7.60)

7.34

0.001a, b

272 (43%)

107 (50%)

108 (38%)

57 (42%)

6.52

0.038b

15.89 (2.88)

16.43 (2.79)

15.73 (2.85)

15.38 (2.94)

6.52

0.002a, b

MMSE score

27.62 (1.82)

27.52 (1.84)

27.60 (1.84)

27.80 (1.75)

0.983

0.375

CDR-SB

1.52 (0.86)

1.49 (0.82)

1.53 (0.89)

1.54 (0.89)

0.162

0.850

NPI-Q score

2.33 (3.23)

2.05 (3.26)

2.24 (2.96)

2.97 (3.62)

3.32

0.037c

GDS score

1.69 (1.47)

1.65 (1.46)

1.57 (1.34)

2.02 (1.58)

4.54

0.011d

Characteristic
Age (yrs)
Female, n
(%)
Education
(yrs.)

Values are presented as mean (SD) for continuous variables and number (%) for categorical
variables. Data are analyzed via analysis of variance (ANOVA) model. Superscripts indicate the
direction of the differences after Bonferroni method correction: a = NW > OB, b = NW > OW, c
= OB > NW, d = OB > OW. Statistical significance is set at p < 0.05.

For all significant comparisons, brain volume was lowest for NW subjects. Table 2.2.
lists the regions that significantly differed and the mean raw volume measures for each BMI
group. A follow-up correlation analysis confirmed a positive association of BMI with brain
volume. As expected, age was negatively associated with volume in each region. The correlation
between precuneus volume and BMI (Pearson’s r = .159, p < .001) and age (Pearson’s r = -.391,
p < .001) are displayed in Figure 2.2. In addition, while neither MMSE nor CDR-SB differed by
BMI in the ANOVA, the brain volumes of all 14 significant regions were positively correlated
with the MMSE and 12 out of 14 were negatively correlated with the CDR-SB. This indicates
the expected cognitive and functional relationship, in that, as brain volume increases MMSE

39

scores increase and CDR-SB scores decrease (moving toward normal cognitive and functional
abilities).
TABLE 2.2. Significant cortical and subcortical volume differences across BMI groups of MCI
subjects
Volume (mm3)
Normal
Weight
n = 216

Overweight
n = 282

Obese
n = 137

F

P value

Precuneus

8005 (1217)

8455 (1197)

8596 (1339)

7.33

0.001a,b

Middle temporal gyrus

9487 (1485)

9946 (1486) 101743 (1409)

5.26

0.005a,b

Hippocampus

3226 (537)

3432 (554)

3565 (619)

5.14

0.006a,b

Lingual gyrus

5662 (924)

5976 (989)

6084 (894)

5.06

0.007b

4.90

0.008b

Brain Region

Lateral occipital cortex

10099 (1602) 10554 (1523) 10678 (1674)

Amygdala

1254 (247)

1331 (243)

1363 (233)

4.89

0.008a,b

Rostral anterior
cingulate gyrus

2048 (404)

2164 (390)

2160 (425)

4.04

0.018b

Superior parietal lobule 11052 (1606) 11549 (1709) 11943 (1913)

3.69

0.026a

Isthmus cingulate gyrus

2148 (380)

2250 (392)

2317 (381)

3.63

0.027

Insular cortex

5965 (847)

6226 (830)

6279 (851)

3.56

0.029b

Pericalcarine cortex

1837 (343)

1938 (370)

1958 (317)

3.47

0.032a,b

Medial orbitofrontal
gyrus

4103 (698)

4325 (727)

4424 (713)

3.46

0.032b

3.30

0.038b

Inferior parietal lobule

11329 (1822) 11852 (1876) 12061 (1858)

Banks of the superior
2109 (342)
2219 (369)
2219 (386)
3.26 0.039b
temporal sulcus
Raw brain volumes, mean (SD), for regions that significantly differed by BMI in a MANOVA
model- BMI; NW (n = 216), OW (n = 281), OB (n = 137). The statistic is a multivariate general
linear Model (GLM), factored by BMI group, with covariates age, and education. Superscripts
indicate the direction of the differences after pairwise comparison correction using the
Bonferroni method: a = NW < OB, b = NW <OW (analysis was done using brain volumes
corrected for total intracranial volume). Significant was set at p < 0.05.

Next, we further examined the relationship of age and BMI with brain volume in our
sample by conducting a series of post-hoc analyses to determine whether age might be a

40

confounding variable to our results. For both the Middle Age and Senior groups brain volume
increased as BMI increased. Body mass index and age values had a significant yet weak
association (Pearson’s r = -0.146, p < .001). Further, a two-way ANOVA tested the interaction of
BMI (NW, OW, OB) and age (Middle-aged, Senior), was not significant, F (72, 1180) = 0.75, p
= 0.939; Wilks’Λ = 0.914, partialη2 = 0.044. Further, we separated the Middle-age and Senior
groups and ran a MANOVA controlling for age and education to examine the effect of BMI on
brain volume within these two distinct life stages. In the model of only middle-aged subjects
there was no difference in any of the 36 brain regions examined based on BMI group. However,
we found robust effects for brain volume differences by BMI in the Senior group. In addition,
despite the large age range for the Senior group (range = 55 – 89 years) there was not a
difference in age across BMI groups. For Seniors, all of the original 14 regions remained
significant, plus three additional regions; the fusiform gyrus (F= 3.51, p = 0.034),
parahippocampal gyrus (F = 3.37, p = 0.035), and the superior temporal gyrus (F = 4.5, p =
0.012). Additional post hoc analyses included BMI, age and brain volumes were also entered
into a multiple regression analysis with the goal of assessing the relationship of BMI and brain
volumes with and without age added into the model. In preparation for this analysis the DurbinWatson test was run to determine whether there was independence between residuals in the
model. This test did not meet the threshold of approximately 2.0 and further analyses could not
be completed due to the higher correlation among our variables (age, BMI, brain volume).
Finally, one third of the population was randomly removed to test whether the large size of our
sample may have generated false positive data. Within the reduced sample, with covariates
added, there was not an effect of BMI on brain volumes, F (72, 770) = 1.09, p = 0.295; Wilks’Λ
= 0.824, partial η2 = 0.09. The reduction of the sample did not affect the observed power of the
analysis, power = 0.998.
41

FIGURE 2.1. Cortical brain regions that significantly differed in volume by BMI. Subcortical
regions that also differed by BMI are not shown: the hippocampus (F = 5.1, p = .006) and
amygdala (F = 4.9, p = .008). The effects of BMI on brain volumes were analyzed using a
MANOVA model correcting for age and education. Regions range in their significance with light
yellow regions meeting statistical significance of p< 0.05 and dark red having a p value of p <
0.001.

42

FIGURE 2.2. The correlation of precuneus volume with BMI and age. BMI was positively
associated with precuneus volume whereas age was negatively associated. The associations were
measured using a Pearson’s correlation with significance set at p < 0.05.

Discussion
This study is the first to provide neurological evidence for the possible association of a
normal BMI with lower brain volumes in MCI. We predicted that as BMI increased, brain
volume would decrease. Yet, in our sample, the NW subjects showed statistically lower brain
volumes. Obese subjects had less education, were younger, with higher mean NPI-Q and GDS
scores. There were 14 regions (out of 36 preselected cortical and subcortical regions) that
showed a medium effect of BMI on brain volume and, in all regions, raw and corrected volumes
were lowest for NW subjects compared to OW and OB subjects. This is the largest MCI study on
the relationship of weight and brain structure, with over 600 MCI subjects.
Previous studies investigating the effect of BMI on brain volume in cognitively normal
adults have found obesity to result in lower brain volumes and worse cognitive performance161.
One study in older women found frontal, posterior parietal and occipital regions to be
specifically vulnerable ROIs in obesity; along with cognitive deficits in executive function that
were associated with smaller left orbitofrontal gyrus volume72. In elderly men and women, BMI
was negatively correlated with brain volume in the orbitofrontal cortex, anterior cingulate gyrus

43

and the medial temporal lobe80. Across these studies whole brain volume also decreased as BMI
increased80,82,162 . Possible mediators between a high BMI and brain atrophy include
inflammation163, Type 2 diabetes164, and hypercortisolemia165. While the link between BMI and
brain volume is not likely to be direct there may be one or more mediators that play a role in this
relationship and in MCI the underlying pathology may initiate the opposite responses seen in
older obese adults with normal cognition.
The only other study on BMI and brain volume in MCI found that brain volume
decreased by 1% for every 1-point increase in BMI82. The difference between this study’s
findings and our own is most likely due to different methodologies. In our study, subjects were
grouped by BMI, and ROI mean volumes were compared, whereas Ho et al., used a whole brain
voxel based analysis of ADNI MCI subjects (ADNI-1, n = 399) with the addition of 77 subjects
from the Pittsburgh Cardiovascular Health Study-Cognition Study (CHS-CS). Our study had a
larger ADNI sample using all 3 phases of only amnestic MCI subjects, and specifically
compared regional brain volumes related to both MCI and obesity. The hypothesis of this study
was created in tandem with the available evidence for the effects of BMI on brain volume
outlined above, yet our findings provide new insight into the relationship of weight and BMI in
MCI.
Our findings were similar in the brain regions affected by BMI however; we found the
opposite relationship of BMI and brain volume (volume decreased as BMI decreased). The brain
regions affected by a low BMI primarily constituted the parietal and occipital lobes of the brain.
The parietal lobe functions primarily in the integration of sensory information and the occipital
lobe in vision. Specifically, the parietal lobe functions in somatosensory and tactile function
(postcentral gyrus), motility (precuneus and superior parietal lobule), spatially directed attention
(R-inferior parietal lobe), symbolic thought and memory (L-inferior parietal lobe) and memory
44

vision and proprioception (inferior parietal lobe, posterior parietal gyrus)166. The occipital lobe
functions in visual perception, including object motion, face recognition, and visual control of
skilled actions. The pathological stages of Alzheimer’s disease described by Braak and
colleagues, indicate involvement of occipital and parietal lobe regions in Stages II – IV of the
disorder (for amyloid), or early to moderate Alzheimer’s disease.
Our findings suggest that the pattern of regional brain volume reductions in NW MCI
may indicate a more severe state of MCI with early deficits in brain regions that reflect middle to
late stage Alzheimer’s disease. Normal weight subjects may have smaller brain volumes
compared to OW and OB in our sample, yet, the most likely explanation is due to the older age
of NW subjects. Additional post hoc analyses were conducted including the separate analysis of
middle-age and Senior groups, multiple regression, random removal of one third of the sample to
address the possible confounding of age that may have persisted despite correction in the
MANOVA model. These post-hoc analyses showed a loss in the effect of BMI on brain volume.
Future studies that aim to assess the relationship of BMI and brain volume in MCI should be
designed so that there is an equal representation of middle-age and Senior subjects. In addition,
study designs should focus on specific age groups with a limited age range of 5 years; this will
more effectively address the possible confounding of age on brain volume, and difference in age
related brain atrophy that occurs in middle age to Senior stages (increasing from 0.5 to 1%
atrophy per year after age 70).
While our initial findings contradict studies on regional brain volumes in normal
cognition and MCI obese subjects, they align with many longitudinal studies that demonstrate a
low BMI resulting in shorter progression time from MCI to AD dementia81,167–170. Current
research demonstrates an increased risk of developing dementia in normal or underweight MCI
subjects and a protective effect or decreased risk of dementia in overweight and obese subjects.
45

A longitudinal study of brain volume trajectories of older adults who were obese at middle age
showed no change in whole brain volume over time and significant regional change only in the
cingulate gyrus81. Tolppanen and colleagues also found that a decrease in BMI from midlife to
late life increased the risk of AD (HR = 1.14)170. Our study is the first to provide neurological
evidence for the findings listed above. The increased risk of dementia and shorter progression
time of NW-MCI subjects may be due to lower regional brain volumes that extend beyond
regions initially affected by MCI but of those that represent AD as shown in the present study.
Limitations
Some limitations must be taking into account when interpreting the results of this study.
The main analysis examined the effect of BMI on brain volumes in brain regions related to high
adiposity and MCI. There was a strong association between age and brain volume and in our
sample, as BMI groups significantly differed by age. We controlled for age in our MANOVA
model to address any potential confounding effects of age on our results171. When we did this the
p-value for the model increased, from p = .027 to p = 0.048 and the effect decreased from a
partial η2 = 0.08 to partial η2 = 0.07. Despite this, our analysis had adequate power (observed
power = 1.0) and remained statistically significant, which may indicate a true effect of BMI on
brain volumes. However, the interpretation of our results must be made with caution due to the
high correlation of our variables (BMI, age, and regional brain volumes) within our analysis and
the loss of a BMI effect in post hoc analyses.

46

Conclusions
Normal weight MCI subjects were unexpectedly older with lower regional brain volumes
compared to OW and OB subjects. This study may provide neurological evidence for recent
findings of a protective effect of a high BMI on MCI progression, and an increased likelihood of
NW individuals progressing to dementia. However, further research is needed to elucidate the
effects of age on brain volume separate from that of BMI on brain volume in MCI subjects. As
the population of older adults increase, the number of overweight and obese people is expected
to grow proportionally. Many seniors are surviving to older ages despite their weight; this is
likely due to evolving treatments for conditions often comorbid with increased weight such as
type-2-diabetes, cardiovascular disease and hypertension. The survivorship of overweight and
obese adults increases the susceptibility of MCI and AD to a wide range of BMIs. It will be
necessary to understand how weight effects or alters the pathophysiology of AD. Understanding
the interactions of weight, age and brain structure may be important in assessing neurologic
vulnerability and dementia risk in individuals with MCI.

47

APPENDIX

48

APPENDIX

TABLE 2A. Selected FreeSurfer brain regions for volume analysis across BMI groups
No.

Label

Region Name

No.

Label

1 ParOrbital

Pars orbital cortex

19 PostCing

2 ParOperc

20 SupPariet

7 ParaCent

Pars opercularis
cortex
Pars triangularis
cortex
Parahippocampal
gyrus
Middle temporal
gyrus
Inferior temporal
gyrus
Paracentral lobule

8 PostCent

Postcentral gyrus

9 RostAntCing

3 ParTriang
4 ParaHipp
5 MidTemp
6 InfTemp

10 CaudAntCing
11 Insula
12 CaudMidFront
13 RostMidFront
14 Precuneus
15 LatOccip
16 PreCent

Region Name

22 Supramarg

Posterior cingulate
gyrus
Superior parietal
lobule
Superior temporal
gyrus
Supramarginal cortex

23 TempPole

Temporal pole

24 TransTemp

26 Cuneus

Transverse temporal
cortex
Banks of the superior
temporal Sulcus
Cuneus

Rostral anterior
cingulate gyrus
Caudal anterior
cingulate gyrus
Insular cortex

27 Entorhinal

Entorhinal gyrus

28 FrontPole

Frontal pole

29 Fusiform

Fusiform gyrus

Caudal middle
frontal gyrus
Rostral middle
frontal gyrus
Precuneus

30 InfPariet

Inferior parietal lobule

31 IsthCing

Lateral occipital
cortex
Precentral gyrus

33 Lingual

Isthmus cingulate
gyrus
Lateral orbitofrontal
gyrus
Lingual gyrus

21 SupTemp

25 Bankss

32 LatOrbFront

34 MedOrbFront

17 SupFront

Medial orbitofrontal
gyrus
Amygdala

Superior frontal
35 Amygdala
gyrus
18 Pericalcar
Pericalcarine
36 Hipp
Hippocampus
cortex
The full list of selected regions from the FreeSurfer software are indicated in the table above by
their label name in FreeSurfer and the full anatomical name of each region.

49

CHAPTER 3

THE INTERACTION OF BODY MASS INDEX, NEUROPSYCHIATRIC SYMPTOMS
AND AGE IN MILD COGNITIVE IMPAIRMENT:
A LONGITUDINAL STUDY

50

Introduction
A growing number of Americans have Alzheimer’s disease, and the prevalence will
dramatically increase over the next 30 years. The pathological process of Alzheimer’s disease is
a long trajectory of physiological, behavioral and cognitive changes that begin many years prior
to clinical symptoms of dementia172. Prior to meeting clinical criteria for dementia, patients
experience an intermediate stage of mild cognitive impairment (MCI) that is characterized by
mild cognitive and behavioral deficits5. Many factors predict the conversion from mild cognitive
impairments to dementia. The majority of studies on these risk factors focus on cognitive
changes and their relationship with progression from MCI to dementia. However, there are noncognitive factors that predict the development of dementia50,164. These non-cognitive conditions
are of great interest because they may provide an indication of early changes that are not directly
related to the irreversible pathological stream that predates cognitive dysfunction. Of even
further interest are those that are ‘modifiable’, which include conditions that involve lifestyle
factors such as, diet and exercise. Two non-cognitive factors that fit these criteria and are
prevalent in MCI are neuropsychiatric symptoms (NPS) and obesity. When MCI subjects have
high NPS (>4) in the early stages of MCI, they show greater cognitive deficits and a higher rate
of conversion to dementia101,102. An increased risk for dementia is also true for individuals that
are obese in middle age (45 – 65 years old) with normal cognition68,134. Obesity and NPS often
occur co-morbidly across age groups in individuals with normal cognition74,173; however, the
association of these two factors with MCI progression to AD has not been measured.
Neuropsychiatric symptoms are highly prevalent in AD (70-90%) and in MCI (40 –
60%)52,102. MCI subjects with high NPS scores are 2.5 times more likely to develop dementia
and also show a faster rate of cognitive decline47. Further, when NPS burden worsens over time
individuals demonstrate faster cognitive and functional decline, and progression to AD over 2
51

years174. Depression, anxiety and apathy are highly prevalent in obese individuals with normal
cognition87,90. MCI subject who are obese with high NPS may be at an increased risk for
cognitive decline.
The number of longitudinal studies investigating BMI in MCI continues to grow with a
primary focus on cognition function175 and the risk of conversion to dementia168,169 over one and
two year intervals. Research findings indicate that a high body mass index (BMI) in MCI does
not result in increased cognitive deficits or a higher conversion rate to dementia. Many findings
identify an increased risk of cognitive decline and the development of dementia in normal weight
MCI subjects 125,169. In these studies normal weight subjects have greater cognitive decline than
over-weight and obese subjects, demonstrated by tests of global cognition, the mini mental status
exam (MMSE) and Alzheimer’s disease assessment scale – cognitive (ADAS –cog scores)176.
The risk of developing dementia is increased 2.5 times in MCI subjects with a low baseline BMI
whereas being overweight reduces the risk of developing dementia over 2 years177 Further,
weight loss also increases the risk of cognitive changes. Cognitively normal elderly are more
susceptible to MCI after weight loss178, and in MCI the risk of dementia is increased by 3.4 fold
and AD specifically by 3.2 fold with weight loss179. The findings of these studies may seem
surprising at first in that, one would expect obesity to lead to more cognitive and functional
dysfunction; yet, being overweight is protective against the onset of dementia. Because the link
between obesity and AD has only been demonstrated when obesity is present at middle age, the
relationship between increased weight and cognitive decline represents a paradox with age as its
nexus135,136. Previous studies have measured cognitive changes and progression to dementia
across BMI groups. This study will add to previous research by measuring the interaction of BMI
and NPS within age groups of Middle Age and Senior subjects.

52

Recent work in our lab on MCI subjects near the time of diagnosis has shown that those
who were obese had a higher frequency of affective neuropsychiatric symptoms (NPS) than
normal weight subjects. Further, increased adiposity measured by BMI was associated with a
younger age at onset of MCI by 7 years on average compared to normal weight subjects.
Currently, no studies have investigated longitudinal changes in cognitive, functional and
behavioral scores. This study seeks to address this gap in knowledge and provide evidence for
specific cognitive and functional changes that may be related to BMI, age, and NPS in MCI.
While we expect to see greater cognitive changes in the NW groups similar to previous reports,
we sought to validate the obesity paradox in this context by directly comparing Middle Aged and
Senior groups who have MCI. We hypothesize that the Middle Age obese group will show a
faster progression to Alzheimer’s type dementia than normal and over-weight middle-age
subjects.
Methods
Participants
Demographic and behavioral data of MCI subjects were obtained from The Alzheimer’s
Disease Neuroimaging Initiative (ADNI) database (adni.loni.ucla.edu)127. The ADNI was
launched in 2005 by the National Institute on Aging (NIA), the National Institute of Biomedical
Imaging and Bioengineering (NIBIB), the Food and Drug Administration (FDA), private
pharmaceutical companies and non-profit organizations, as a $60 million, 5-year public-private
partnership. The primary goal of ADNI has been to test whether serial magnetic resonance
imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and
neuropsychological assessment can be combined to measure the progression of MCI and early
AD. Determination of sensitive and specific markers of early AD progression is intended to aid

53

researchers and clinicians to develop new treatments and monitor their effectiveness, as well as
lessen the time and cost of clinical trials.
ADNI subjects were recruited from over 50 sites across the U.S. and Canada. The initial
goal of ADNI was to recruit 800 subjects but ADNI has been followed by ADNI-GO and ADNI2. To date these protocols have recruited over 1500 adults ages 55 to 90, consisting of
cognitively normal older individuals, people with early or late MCI, and people with early AD.
Clinical, behavioral and neuroimaging data are collected for each patient for up to 54 months.
MRI methods, procedures and preprocessing by ADNI have been previously described140 and
can also be accessed at: www.afni-info.org. Inclusion criteria for MCI set forth by ADNI
include; an MMSE score between 24 – 30, a subjective memory complaint by the patient or
caregiver, objective memory loss measured by the Wechsler Memory Scale Logical Memory II,
a global Clinical Dementia Rating (CDR) of 0.5, preserved activities of daily living, and the
absence of dementia.
For our study, we abstracted data on individuals who met ADNI-defined MCI criteria
from all 3-phases of ADNI. Overall, the analysis was undertaken in two parts, first by conducting
a longitudinal analysis that analyzed baseline, 2-year and change scores (2-yr – bl) on the NPI-Q,
GDS, MMSE, ADAS-cog 13 and CDR-sb. Second, by determining the number of subjects that
progressed to AD and examining the survival of subjects based on group membership in BMI
(NW, OW, OB) or age (Middle-Age and Senior) group. Only MCI subjects with cognitive,
behavioral and functional measures at both baseline and 2-year time points were included in the
longitudinal analysis. We also analyzed whether there was an interaction of age with BMI groups
and how NPS affected progression over 2 years.
Factors

54

We sub-divided our sample into a Middle Age group, aged 55 – 65 and a Senior group
who were 66 years of age or older. We also created BMI groups as follows: normal weight (NW;
BMI 18.5- 24.9 kg/m2), overweight (OW; BMI 25 - 29.9 kg/m2) or obese (OB; BMI ≥ 30 kg/m2).
Height (in inches) and weight (in pounds) measurements were taken at the time of clinical
diagnosis of MCI. BMI was converted to the unit kg/m2 using the follow calculation, [(Weight
(lb.) / Height2 (in.)) x 703].
Longitudinal analysis methods and statistical analyses
The longitudinal analysis assessed changes in cognitive, functional and behavioral
measures over a 2-year interval. Global cognitive measures included the MMSE, and ADAS-cog
13. The MMSE assesses changes in 5 categories: orientation (i.e. date, place), registration or
immediate recall, delayed recall, attention and calculation (i.e. subtraction) and language27. There
are 30 questions worth 1 point each with scores ranging from normal cognition (>25) to severely
impaired (≤ 10). In addition, the ADAS cognitive 13-item subscale tests cognitive domains of
memory, language, praxis, attention, and other cognitive abilities with a total scoring range of 0
– 70 points118. Higher scores indicate a higher degree of impairment. Clinical Dementia Rating
Scale – sum of boxes (CDR-SB)180 scores were compared across groups as an additional measure
of combined cognitive and functional change. The CDR-SB includes five categories, memory,
orientation, judgment, community affairs, home and hobbies, and personal care that are assessed
with item scores ranging from no impairment (0), mild impairment (0.5) up to severe impairment
(3). The sum of each category score resulted in the total sum of boxes score, which could range
from 0.5 to 15.
Neuropsychiatric symptoms were measured using the Neuropsychiatric Inventory
Questionnaire (NPI-Q) 109 and the Geriatric Depression Scale –short form (GDS) 110. The NPI-Q
is a validated measure for assessing behavioral disturbances across 12 different domains in a
55

brief caregiver-reported questionnaire 109. These include; delusions, hallucinations,
agitation/aggression, depression/dysphoria, anxiety, elation/euphoria, apathy/indifference,
disinhibition, irritability, aberrant motor behavior, sleep and nighttime behavioral changes, and
appetite and eating disorders. An informant familiar with the subject reports NPI-Q symptoms,
by rating each symptom for their presence (yes/no), and severity (range of 0 - 3) for each item.
Behavioral changes reported on the NPI-Q reflect symptoms present within one month of testing,
with 36 possible points. The self-reported 15-point GDS scale further quantified depressive
symptoms. Low NPS was designated as a total score between 0-3 symptoms and high NPS as ≥ 4
on either test.
The total score for each test and the prevalence of mild and moderate symptoms were
measured across BMI and age groups. A difference score was computed between the baseline
and two-year visit for each test and analyzed across all subjects and BMI groups. The difference
in baseline, 2 year, and change scores were compared across BMI groups using an ANOVA
model. A Bonferroni post-hoc test was used to determine which groups differed from each other.
Significance was set at p <0.05.
Survival analysis methods and statistical analyses
A survival analysis181 was used to measure the survival distribution of BMI, age and NPS groups
in order to identify differences in cumulative survival over two years related to these conditions
independently and when factored together. The survival analysis included the following
parameters: the time of origin was the baseline visit, the event was a dementia (probable AD)
diagnosis, and the comparison groups were BMI (NW, OW, OB), age group (Middle-Age,
Senior) and NPS (High/ Low baseline NPI-Q and GDS score). The survival distributions were
measured for each group independently and factored together. Time was measured in 6-month
intervals from the baseline visit to the 2-year visit. Censored events included loss to follow-up,
56

death, and another diagnosis. All ADNI MCI subjects that met the inclusion criteria were
included in the survival analysis.
The cumulative survival of MCI subjects was calculated using Kaplan-Meier plots in
SPSS version 22.0 (Chicago, IL). Significance was set at p <0.05.

Results
Baseline measures for all ADNI MCI subjects from each phase were combined into one
dataset. Selected demographic information for the combined dataset is displayed in Table 1.
Overall, the sample was predominantly Caucasian, with nearly 16 years of education and over
66% were overweight or obese. The obese group was younger, less educated, with higher mean
arterial pressure (MAP) values and a higher proportion of T2D, HTN, HLP and OSA co-morbid
conditions compared to normal weight. Obese subjects also had higher mean NPI-Q scores and
symptom prevalence compared to NW subjects. Age group differences in demographic,
cognitive and behavioral variables are reported in the Appendix, Table 3A.

57

TABLE 3.1. Demographic, cognitive and behavioral characteristics for all ADNI MCI subjects
and across BMI groups
All MCI

NW

OW

OB

F
or
X2
23.3

Characteristic

(n =956)

(n = 318 )

(n = 430)

(n = 205)

p value

Female, n (%)
Age, yrs
MidAge, n (%)

396 (41.5%)
72.51 (7.9)
199 (20.8%)

162 (50.9%)
73.47 (7.8)
51 (16%)

144 (33.6%)
72.68 (7.7)
88 (20.5%)

90 (43.9%)
70.68 (8.1)
59 (28.8%)

8.1
12.3

< 0.01
0.002c

Senior, n (%)
Education, yrs
MMSE
CDR-SB
BMI (kg/m2)
NPI-Q score

756 (79.2%)
15.93 (2.8)
27.57 (1.8)
1.54 (0.9)
27.06 (4.7)
2.26 (3.1)

267 (84%)
16.49 (2.8)
27.46 (1.8)
1.55 (0.9)
22.59 (1.8)
2.0 (3.1)

341 (79.5%)
15.84 (2.72)
27.57 (1.8)
1.53 (0.9)
27.09 (1.4)
2.16 (2.9)

146 (71.2%)
15.22(2.9)
27.75 (1.7)
1.56 (0.9)
33.9 (4.0)
2.26 (3.2)

12.3

0.002c

13.4
1.7
0.1
5.6

< 0.01
0.19
0.88
< 0.01
0.010b

≥1

494 (62.2%)

157 (58.4%)

223 (62.5%)

113 (69.8%)

5.6

0.021c

≥4

183 (23.2%)

53 (19.7%)

82 (23%)

47 (29%)

4.9

0.034c

1.69 (1.5)

1.64 (1.5)

1.6 (1.42)

1.97 (1.8)

4.4

0.059b

≥1

730 (76.4%)

242 (76.1%)

326 (75.8%)

161 (78.9%)

0.8

0.053c

≥4

128 (13.4%)

40 (12.6%)

50 (11.6%)

38 (18.6%)

6.1

0.122c

GDS score

0.015c

Mean comparisons across BMI groups were conducted using an ANOVA model for continuous
variables, mean (SD) and the chi-square test of independence for categorical variables, n (%).
Behavioral tests are reported as the mean score and total symptom categories of ≥1 or ≥4. b Nonparametric Kruskal-Wallis Test C Gamma approx p value, ordinal by ordinal. Significance was
set at p <0.05.
Abbreviations: NPI-Q, neuropsychiatric inventory questionnaire; GDS, geriatric depression
scale; MMSE, Mini Mental Status Examination; CDR-SB, Clinical Dementia Rating Scale –
Sum of Boxes; BMI, body mass index; NW, normal weight; OW, overweight; OB, obese;
MidAge, Middle-Age (<66 years), Senior, seniors (>65 years); yrs, years.

Longitudinal analysis
A total of 634 subjects had both baseline and 2 year behavioral, cognitive and functional
data measurements. The mean test scores for all subjects and by BMI are displayed in Table 2.
At the 2-year follow-up, average NPS and CDR-SB scores were increased and MMSE scores
decreased. The baseline and 2 year ADAS-cog scores significantly differed across BMI groups:

58

normal weight subjects had higher mean ADAS-cog 13 scores than obese. Normal weight
subjects also had a greater degree of cognitive change than obese over two years measured by the
ADAS-cog 13, nearly three times the mean change of OB subjects. The CDR-SB change scores
were also greatest for NW subjects indicating more cognitive and functional impairments than
the OB group. There was no difference in change scores for the NPI-Q, GDS, or the MMSE.

TABLE 3.2. Longitudinal changes in behavior, cognitive and functional test scores over two
years for all MCI subjects and across BMI groups
All MCI
F
Test
NW
OW
OB
p value
N = 634
value
NPI-Q
Baseline
1.83
1.84
1.67
2.17
1.13
0.326
2 Years
2.53 (.15)
2.40 (.23)
2.43 (.23)
3.01 (.42)
1.13
0.324
Change
0.71 (.14)
0.55 (.24)
0.79 (.21)
0.84 (.33)
0.40
0.669
score
GDS
Baseline
1.69
1.64
1.60
1.97
1.66
0.191
2 Years
1.93
1.86
1.86
2.19
0.12
0.884
Change
0.30 (.07)
0.28 (.12)
0.31 (.10)
0.33 (.18)
0.03
0.969
score
MMSE
Baseline
27.57
27.46
27.57
27.75
1.16
0.314
2 Years
26.36
26.04
26.41
26.78
1.43
0.240
Change
-1.29 (.12) -1.44 (.19) -1.28 (.19) -1.08 (.26)
0.61
0.543
score
ADAS-cog 13
Baseline
16.16 (.27) 16.95 (.45) 16.04 (.38) 15.11 (.62)
3.23
0.040
2 Years
18.63 (.41) 20.48 (.74) 18.29 (.58) 16.36 (.87)
6.95
0.001
Change
2.46 (.26)
3.52 (.48)
2.21 (.37)
1.25 (.53)
5.33
0.005
score
CDR - SB
Baseline
1.54
1.54
1.52
1.56
1.92
0.148
2 Years
2.53
2.74
2.52
2.22
0.12
0.884
Change
0.99 (.07)
1.19 (.13)
0.99 (.11)
0.66 (.15)
3.29
0.038
score
Mean comparisons across BMI groups were conducted using an ANOVA model and reported as
the mean value and standard deviation. Significance was set at p < 0.05.
Abbreviations: 2-yr, 2 year score; NPI-Q, neuropsychiatric inventory questionnaire; GDS,
geriatric depression scale; MMSE, Mini Mental Status Examination; CDR-SB, Clinical
Dementia Rating Scale – Sum of Boxes; NW, normal weight; OW, overweight; OB, obese.

59

Survival analysis
Nine hundred and fifty-six MCI subjects met the inclusion criteria at baseline and were
included in the survival analysis. Over 2 years, 172 (27%) subjects converted to probable AD
dementia, 23 (3.6%) reverted to normal cognition and 322 (34%) were lost to follow-up. Kaplan
Meier plots display the cumulative survival of MCI subjects over 2 years. The survival
distribution of our primary factors of interest, BMI (NW, OW, OB), NPS (high/low NPI-Q and
GDS) and age (Middle-Aged, Senior) were analyzed first. A log rank test was used to determine
if there were differences in the overall survival distributions for the BMI, NPI-Q, GDS and age
groups. In the analyses, the cumulative survival did not fall below 50% for any comparisons;
therefore median survival estimates were not generated. However, a similar percentage of
censorship was present across all group, seen by in the NW (75%), OW (79%), and OB (83%)
BMI groups, as well as in the middle age (84%) and Senior (77%) groups. The survival
distributions by BMI (Figure 3.1) were not significantly different, Χ2 (2) = 4.81, p = 0.090. The
high and low NPS symptom group survival distributions also did not differ for the NPI-Q (Χ2 (2)
= 0.93, p = 0.336) and GDS (Χ2 (2) = 0.21, p = 0.64). However, the survival distributions by age
group (Figure 3.2) were significantly different between Middle Age and Senior subjects, Χ2 (1) =
4.86, p = 0.027. The mean time to conversion was 22 months (95% CI, 21.7 to 22.4) in Senior
MCI subjects compared to 23 months (95% CI, 22.0 to 23.2 months) for Middle-aged subjects.
After analyzing the primary factors we then tested the interactions of age with BMI and
NPS. The log rank test for the survival distribution factored by age and adjusted for BMI was
statistically significant, Χ2 (1) = 4.05, p = 0.044 (Figure 3.3). However, further post hoc analysis
using pairwise comparisons of age distributions within BMI groups was not able to determine the
where the differences were: NW, Χ2 (1) = 2.81, p = 0.094, OW, Χ2 (1) = 0.706, p = 0.401, OB Χ2

60

(1) = 1.08, p = 0.298. This is likely due to the conservative nature of the test when correcting for
multiple comparisons.

FIGURE 3.1. Kaplan-Meier curves comparing the rate of survival of BMI groups from baseline
MCI status to the diagnosis of Alzheimer’s type dementia. The cumulative survival of three BMI
groups were compared over 24 months or 4 study visits. Crosses indicate censored events.
Abbreviations: NW, normal weight; OW, overweight; OB, obese.

Finally, we were interested in the interaction of BMI and age with behavioral scoring on
the NPIQ and GDS. Since there was no difference in survival by BMI, the NPS test scores,
grouped as high or low, were stratified across age groups (Figure 3.4). These interactions
demonstrated an overall significant relationship between age group, and NPI-Q and GDS
high/low symptom groups. The survival distributions for Middle-age and Senior groups that had
low NPIQ score compared to high NPIQ scores at baseline significantly differed, Χ2 (1) = 4.66, p
61

= 0.031 (Fig. 3.4A). There was also a significant difference in survival distributions for MiddleAge and Seniors that had low GDS score compared to high GDS scores at baseline, Χ2 (1) = 4.68,
p = 0.031 (Fig. 3.4B). Pairwise comparisons of the low and high NPI-Q groups indicated a
significant difference in survival for MCI subjects with low baseline NPI-Q scores, Χ2 (1) = 5.0,
p = 0.025, indicating a faster progression to AD for Seniors with Low NPI-Q scores at baseline.
The difference in survival was not significant for MCI subjects with high baseline NPIQ scores
after correction, Χ2 (1) = 0.427, p = 0.514. For GDS groups, there was also no difference in the
survival distribution for baseline low, Χ2 (1) = 3.29, p = 0.070, or high, Χ2 (1) = 1.52, p = 0.218,
scores after correcting for multiple comparisons.

FIGURE 3.2. Kaplan-Meier curves comparing the rate of survival of age groups from baseline
MCI status to the diagnosis of Alzheimer’s type dementia. The cumulative survival of Middle
Age and Senior groups were compared over 24 months or 4 study visits. Crosses indicate
censored events.
62

FIGURE 3.3. Kaplan-Meier curves comparing the rate of survival of BMI groups factored by age
group from baseline MCI status to the diagnosis of Alzheimer’s type dementia. The cumulative
survival of Middle Age and Senior subjects within three BMI groups of normal weight,
overweight and obese were compared over 24 months or 4 study visits. Crosses indicate
censored events.

63

A.

B.

FIGURE 3.4. Kaplan-Meier curves comparing the rate of survival of age groups factored by high
and low NPS groups from baseline MCI status to the diagnosis of Alzheimer’s type dementia.
(A) NPI-Q groups of high and low symptom burden compare the cumulative survival of Middle
Age and Senior MCI subjects. (B) GDS groups of high and low symptom burden compare the
cumulative survival of Middle Age and Senior MCI subjects. The low group represents total test
scores between 0 and 3 and the high scores are ≥ 4. The cumulative survival of age factored by
NPS group were compared over 24 months or 4 study visits. Crosses indicate censored events.
Abbreviations: NPS, neuropsychiatric symptoms; NPI-Q, Neuropsychiatric Inventory
Questionnaire; GDS, Geriatric Depression Scale.

64

Discussion
This study of over 600 MCI subjects investigated how weight and age influenced the
progression of MCI by investigating its cognitive, functional, and behavioral features over two
years. We found that NW subjects had greater cognitive changes over 2 years, similar to
previous reports169,176,177,182. The Kaplan-Meier curves for BMI did not reach significance to
show a difference in survival but our sample did show significant, if small, differences in
survival based on age group. New findings include the interaction of BMI and age resulting in a
change in the survival distribution. Further, the interaction of age and NPS also affected survival
in MCI subjects. We were not able to support our hypothesis that middle aged OB have a faster
progression time to AD. In our sample, OB individuals had less education, higher NPS scores
and multiple metabolic co-morbidities, but equivalent levels of cognitive impairment compared
to NW subjects at baseline. We showed that in MCI, a higher proportion of Middle Age subjects
were obese compared to Seniors. However, the duration of our study period did not allow for the
identification of longitudinal effects of obesity in middle age subjects.
Interestingly, MCI subjects with 4 or more symptoms on the NPI-Q had higher CDR-SB
scores (Appendices, Table 3C.) The mean BMI of the NPI-Q high group was significantly higher
than the low group, and a higher proportion of OB subject had symptoms ≥ 4. While NW
subjects had a lower NPS burden compared to OB they had greater cognitive deficits at each visit
(ADAS-cog 13) and in their overall cognitive change score (ADAS-cog 13 and CDR-SB). A
possible explanation could be that metabolic and or pathologic changes specific to low body
weight MCI subjects are added to by even a low NPS, affecting cognition. Other studies have
found that even a mild NPS burden produces significant cognitive changes over time99,183.
Moreover the younger age of obese subjects may have a protective effect on cognition.

65

Senior MCI subjects with low NPS at baseline had a decreased time of progression to
dementia compared to middle age subjects. The NPI-Q and GDS are common measures of
behavioral disturbance in geriatric populations, providing insight into changes in mood and
behavior while being easy and quick to administer. Previous work has shown that even one
symptom on the NPI-Q can predict future changes in cognitive status. Our average scores for the
NPI-Q and GDS were low overall and the difference between groups was in the range of 1 point.
While these values are low as far as indicating significant behavioral disturbances, the difference
between groups provides evidence for a possible increased risk of dementia to Senior MCI
subjects when at least one NPS is present.
Limitations
While this study included a large number of MCI subjects, some limitations should be
taken into account. First, the age groups within our sample were unequal with over 3 times the
number of Seniors compared to middle aged subjects. A more diverse group in regards to race,
ethnicity and educational attainment may provide further insight into how BMI and age influence
the progression from MCI to AD over 2 years. Second, the sample was primarily Caucasian and
well educated having an average of almost 16 years of education. A wider range of demographic
factors may allow for the analysis of patient sub-groups not seen in this study, such as the
survival distributions of obese subjects with only a high school education. Third, our KaplanMeier analyses did not reach a cumulative survival of at least 50 percent to allow for traditional
reporting of median difference. Extending the follow-up time of the analysis may provide for a
more sensitive measurement of the survival distributions.

66

Conclusions
Mild Cognitive Impairment includes a heterogeneous group of individuals and therefore
requires the assessment of multiple risk factors for their contributions to disease progression. In
this study, we provide insight into how age and weight interact with each other and NPS
symptoms to lower the survival of Senior NW individuals. This study is the first of its kind to
assess age effects on disease progression within the ADNI cohort, with a focus on middle-aged
MCI subjects. Weight and NPS are modifiable risk factors for Alzheimer’s type dementia and
their relationship with age may indicate groups at highest risk for the conversion to dementia.
Further, a targeted approach to recruiting more middle-aged subjects with MCI may provide
greater insight to the cognitive, functional and behavioral changes over time specific to this
group. Future studies may add to these findings by assessing BMI groups over a longer time
period. Further, assessing brain structure changes related to BMI and age after two years may
provide additional insight into cognitive changes and overall progression rate to dementia.

67

APPENDIX

68

APPENDIX

TABLE 3A. Demographic, cognitive, cardiovascular, and behavioral characteristics comparing
Middle-Age and Senior MCI subjects
Middle Age
Senior
Statistic
Characteristic
(n = 199)
(n = 756)
t or X2
p value
292
Female, n (%)
104 (52.3)
12.07
0.001
(38.6%)
15.85
Education, years
16.25 (2.6)
3.13
0.077
(2.86)
27.43
MMSE
28.09 (1.68)
22.07
< 0.001
(1.78)
CDR sum of boxes 1.51 (0.93)
1.55 (0.89)
0.261
0.609
26.76
BMI (kg/m2)
28.19 (5.41)
14.84
< 0.001
(4.46)
NW, n (%)
51 (25.8%)
267 (35.4)
6.29
0.012
OW, n (%)
88 (44.4%)
341 (45.2)
0.011
0.918
OB, n (%)
59 (29.8%)
146 (19.4)
10.11
0.001
94.94
95.35
MAP
0.257
0.612
(10.69)
(10.18)
T2D
23 (11.9%)
72 (9.8%)
0.67
0.413
369
HTN
87 (44.8%)
1.95
0.16
(50.5%)
352
HLP
68 (35.1%)
10.62
0.001
(48.2%)
OSA
24 (12.4%)
70 (9.6%)
1.31
0.25
NPI-Q score
2.42 (3.39)
2.23 (3.08)
0.48
0.49
396
≥1
98 (65.8%)
0.78
0.376
(61.9%)
146
≥4
37 (24.8%)
0.28
0.6
(22.8%)
GDS score
2.07 (1.66)
1.59 (1.49)
15.49
< 0.001
560
≥1
169 (85.4%)
11.08
0.001
(74.1%)
≥4
42 (21.2%)
86 (11.4%)
13.07
< 0.001
Values are presented as mean (SD) for continuous variables and n (%) for categorical variables.
The statistic is chi-square test of independence for categorical variables and a t statistic for
continuous variables. Behavioral tests are reported as the mean score and two total symptom
categories. The samples for behavioral tests differed for the NPI-Q (n = 789) and GDS (n = 954).
Significance was set as p < 0.05.
Abbreviations: MCI, Mild Cognitive Impairment; BMI, body mass index; MCI SV, MCI
severity; MMSE, Mini-Mental State Examination; GDS, Geriatric Depression Scale; CDR,
clinical dementia rating scale; NPI-Q, Neuropsychiatric Inventory Questionnaire; NW, normal
weight; OW, overweight; OB, obese; MAP, mean arterial pressure; T2D, type-2-diabetes; HTN,
hypertension; HLP, hyperlipidemia; OSA, obstructive sleep apnea.
69

TABLE 3B. Demographic, cognitive, cardiovascular, and behavioral characteristics comparing
male and female MCI subjects
Male
Female
Statistic
Characteristic
(n = 559)
(n = 396)
F or X2
p value
Age, years
73.33 (7.68)
71.34 (7.95)
15.17
< 0.001
Middle Age, n (%)
95 (17%)
104 (26.3%)
12.07
0.001
Senior, n (%)
464 (83%)
292 (73.7%)
12.07
0.001
Education, years
16.26 (2.79)
15.48 (2.78)
18.11
< 0.001
MMSE
27.49 (1.77)
27.68 (1.79)
2.5
0.114
CDR sum of boxes
1.56 (0.9)
1.51 (0.87)
0.58
0.445
BMI (kg/m2)
27.33 (4.07)
26.67 (5.46)
4.46
< 0.001a
NW, n (%)
155 (27.9%) 161 (40.8%)
17.27
< 0.001
OW, n (%)
287 (51.6%) 144 (36.4%)
21.72
< 0.001
OB, n (%)
115 (20.7%)
90 (22.7%)
0.57
0.45
95.54
94.88
MAP
0.95
0.33
(10.11%)
(10.53%)
T2D
65 (12%)
30 (7.8%)
4.3
0.038
HTN
278 (51.4%) 178 (46.4%)
2.28
0.13
HLP
263 (48.6%) 157 (40.9%)
5.41
0.02
OSA
69 (12.8%)
25 (6.5%)
9.59
0.002
NPI-Q score
2.45 (3.3)
1.99 (2.87)
4.18
0.041a
≥1
299 (64%)
195 (60.6)
0.98
0.32
≥4
117 (25.1%)
66 (20.5%)
2.22
0.14
GDS score
1.6 (1.51)
1.82 (1.57)
4.48
0.034
≥1
425 (76%)
304 (77%)
0.11
0.74
≥4
63 (11.3%)
65 (16.5%)
5.36
0.021
Values are presented as mean (SD) for continuous variables and n (%) for categorical variables.
The statistic is chi-square test of independence for categorical variables and a t statistic for
continuous variables. Behavioral tests are reported as the mean score and two total symptom
categories. The samples for behavioral tests differed for the NPI-Q (n = 789) and GDS (n = 954).
Significance was set as p < 0.05.
Abbreviations: MCI, Mild Cognitive Impairment; BMI, body mass index; MCI SV, MCI
severity; MMSE, Mini-Mental State Examination; GDS, Geriatric Depression Scale; CDR,
clinical dementia rating scale; NPI-Q, Neuropsychiatric Inventory Questionnaire; NW, normal
weight; OW, overweight; OB, obese; MAP, mean arterial pressure; T2D, type-2-diabetes; HTN,
hypertension; HLP, hyperlipidemia; OSA, obstructive sleep apnea.
a
Non-parametric Mann Whitney U test p-value

70

TABLE 3C. Demographic, cognitive, cardiovascular, and behavioral characteristics comparing
MCI subjects with low and high NPI-Q scores
NPI-Q low
NPI-Q high Statistic
Characteristic
(n = 606)
(n = 183)
t or X2
p value
Age
73.29 (7.75)
71.66 (7.43)
2.52
0.012
Middle Age, < 66, n
112 (18.5%)
37 (20.2%)
0.277
0.599
(%)
Senior, > 66, n (%)
494 (81.5%)
146 (79.8%) 0.277
0.599
Female, n (%)
256 (42.2%)
66 (36.1%)
2.22
0.136
Education
15.93 (2.89)
15.98 (2.7)
0.206
0.831
MMSE
27.56 (1.82)
27.54 (1.77)
0.142
0.887
CDR sum of boxes
1.41 (0.81)
1.9 (0.96)
7.68
<0.001
BMI (kg/m2)
26.75 (4.53)
27.76 (5.09)
2.57
0.01
NW
215 (35.5%)
52 (28.6%)
3.028
0.082
OW
276 (45.5%)
83 (45.6%)
0.12
0.989
OB
115 (19%)
47 (25.8%)
4.018
0.045
MAP
95.1 (10.3)
95.2 (11.09)
0.118
0.906
T2D
50 (8.4%)
22 (12.4%)
2.59
0.107
HTN
292 (49.2%)
90 (50.8%)
0.156
0.693
HLP
268 (45.1%)
82 (46.3%)
0.081
0.777
OSA
56 (9.4%)
21 (11.9%)
0.901
0.343
GDS score
1.54 (1.37)
2.12 (1.54)
4.55
<0.001
≥1
454 (75%)
155 (85.2%)
8.19
0.004
≥4
64 (10.6%)
37 (20.3%)
11.89
0.001
Values are presented as mean (SD) for continuous variables and n (%) for categorical variables.
The statistic is chi-square test of independence for categorical variables and a t statistic for
continuous variables. Behavioral tests are reported as the mean score and two total symptom
categories. The samples for behavioral tests differed for the NPI-Q (n = 789) and GDS (n = 954).
Significance was set as p < 0.05.
Abbreviations: MCI, Mild Cognitive Impairment; BMI, body mass index; MCI SV, MCI
severity; MMSE, Mini-Mental State Examination; GDS, Geriatric Depression Scale; CDR,
clinical dementia rating scale; NPI-Q, Neuropsychiatric Inventory Questionnaire; NW, normal
weight; OW, overweight; OB, obese; MAP, mean arterial pressure; T2D, type-2-diabetes; HTN,
hypertension; HLP, hyperlipidemia; OSA, obstructive sleep apnea.

71

TABLE 3D. Demographic, cognitive, cardiovascular, and behavioral characteristics
comparing MCI subjects with low and high GDS scores

GDS low
Characteristic (n = 827)
Age, yrs
72.81 (7.73)
Mid Age, n (%) 156 (19%)
Senior, n (%) 670 (81%)
Female, n (%) 330 (40%)
Education, yrs 15.94 (2.81)
MMSE
27.58 (1.78)
CDR-SB
1.53 (0.89)
BMI (kg/m2 ) 26.93 (4.59)
NW, n (%) 277 (33.7%)
OW, n (%) 381 (46.2%)
OB, n (%) 166 (20.1%)
MAP
95.12 (10.2)
T2D
79 (9.9%)
HTN
401 (50.2%)
HLP
365 (45.7%)
OSA
82 (10.3%)
NPI-Q score 2.05 (2.85)
≥1
420 (61.2%)
≥4
145 (21.1%)

GDS high
(n = 126)
70.6 (8.34)
41 (32.5%)
85 (67.5%)
65 (51.6%)
15.93 (2.82)
27.51 (1.77)
1.59 (0.90)
27.86 (5.38)
38 (30.2%)
50 (39.7%)
38 (30.2%)
96.15 (10.96)
15 (12.2%)
55 (44.7%)
55 (44.7%)
11 (8.9%)
3.67 (4.46)
72 (71.3%)
37 (36.6%)

Statistic
2

t or X
2.96
12.47
12.21
6.10
0.04
0.43
0.64
1.85
0.60
1.90
6.50
1.04
0.62
1.28
0.04
0.21
3.55
3.80
11.89

p value
0.003
<0.001
<0.001
0.014
0.971
0.665
0.526
0.67
0.437
0.169
0.011
0.30
0.431
0.258
0.841
0.651
0.001
0.051
0.001

Values are presented as mean (SD) for continuous variables and n (%) for categorical variables.
The statistic is chi-square test of independence for categorical variables and a t statistic for
continuous variables. Behavioral tests are reported as the mean score and two total symptom
categories. The samples for behavioral tests differed for the NPI-Q (n = 789) and GDS (n = 954).
Significance was set as p < 0.05.
Abbreviations: MCI, Mild Cognitive Impairment; BMI, body mass index; MCI SV, MCI
severity; MMSE, Mini-Mental State Examination; GDS, Geriatric Depression Scale; CDR-SB,
clinical dementia rating scale- sum of boxes; NPI-Q, Neuropsychiatric Inventory Questionnaire;
NW, normal weight; OW, overweight; OB, obese; MAP, mean arterial pressure; T2D, type-2diabetes; HTN, hypertension; HLP, hyperlipidemia; OSA, obstructive sleep apnea.

72

CHAPTER 4

ALTERNATE METHODS

73

THE EFFECT OF OBESITY ON BRAIN WHITE MATTER IN MILD COGNITIVE
IMPAIRMENT
Introduction
Changes in white matter microstructure result in altered brain connectivity184. A MRI
diffusion-weighted imaging sequence quantifies possible changes in axonal integrity via
measures of diffusion rate and directionality. When brain white matter is intact, water diffuses
along an axon in one direction. However, when there is damage diffusion is altered, suggesting a
loss in the myelin sheath that insulates the axon and helps propagate neural impulses. The most
commonly used measures include fractional anisotropy (FA), mean diffusivity (MD), radial
diffusivity (RD) and axial diffusivity (AxD). These measures provide estimates of the
microstructural integrity of the axon bundles that structurally connect brain regions185. Fractional
anisotropy is an estimate of fiber integrity reflecting the coherence of the orientation of water
diffusion independent of the rate of diffusion, whereas MD, RD, and AxD measure the rate of
diffusion along an axis. Mean diffusivity measures the average rate of diffusion; RD measures
diffusion perpendicular to the major axis of water diffusion and AxD is the rate of diffusion
along the major axis. Each measure is thought to assess different components of white matter
structure: MD may increase with decreasing myelination, RD decreases with diminished myelin
integrity and AxD may decrease with axonal damage.
Brain white matter is susceptible to microstructural changes associated with MCI and
obesity30,83,186. Research studies of MCI subjects have demonstrated decreased FA in limbic
white matter24 with values decreasing from normal cognition to MCI to AD. Another study
found that diffusivity measures do a better job of distinguishing groups of NC, MCI and AD
compared to FA184. These studies suggest that as AD pathology progresses clinically, axonal
fiber integrity diminishes. Since MCI is an intermediate stage, early changes in white matter

74

structure are evident and identifying additional factors that influence axonal integrity may
highlight features of early pathological changes.
Studies investigating the relationship of weight and brain white matter across a wide age
range of adults demonstrate an inverse relationship between BMI and FA values. In limbic
system tracts with connections to the temporal and frontal lobe, FA decreases as BMI
increases83,187. A recent study measuring the association of BMI with all 4 DTI-based measures
(FA, MD, AxD, RD) found a negative association with FA and RD in the right middle cerebellar
peduncle, and MD and AxD in the bilateral corticospinal tract and anterior thalamic radiation. In
the same study, BMI was also positively associated with MD and AxD in the right superior
longitudinal fasciculus186. As BMI increased so did MD and AxD values suggesting decreased
fiber integrity (MD). To date there have not been any studies that investigate the relationship of
brain white matter structure and BMI in MCI. The hypothesis for the following experiments is
that in MCI, obese subjects will have lower mean FA values and higher MD values indicating
deficits in axonal fiber structure and myelination.
There is limited information regarding white matter integrity in obese individuals and
none in MCI. The following sections will outline two studies that investigated whether
differences in brain white matter existed between BMI groups of MCI subjects. The first study
uses a single site research dataset from the MSU Cognitive and Geriatric Neurology Team
(COGENT) and the second uses subjects from a large multi-site database, the Alzheimer’s
disease neuroimaging initiative (ADNI). The methods and results for each study are reported
independently with a summary and conclusion on the effect of obesity on MCI brain white
matter at the end.

75

Methods – MSU COGENT
Participants
Nineteen participants (16 MCI and 3 cognitively normal) were recruited from the MSU
Neurology clinic and through community advertisements between 2012 - 2014. Participants were
diagnosed with MCI based on Petersen criteria1,5 by an expert neurologist prior to study
participation. Inclusion criteria were as follows; subjects were between the ages of 50-95, able to
speak, comprehend and read English with at least 8 years of education, and a Mini Mental Status
Examination (MMSE) 27 score between 24 - 30. Subjects were excluded from the study if they
had a history of a coexisting central nervous system disorder or uncontrolled depression that
could account for the cognitive impairment, any uncontrolled or unstable medical condition, and
alcohol or substance abuse within the last two years. Exclusion criteria were determined based
on medical records review.
Data collection overview
Informed consent was obtained directly from each subject. Subjects were then screened
to confirm eligibility (see above inclusion/exclusion criteria), and then underwent magnetic
resonance imaging (MRI) scanning followed by neuropsychological and behavioral testing. All
study procedures were reviewed and approved by the MSU Institutional Review Board.
MRI acquisition of COGENT data
MRI whole-brain imaging procedures were conducted on a GE 3T Signa HDx MR
scanner (GE Healthcare, Waukesha, WI) equipped with an 8-channel head coil in the Radiology
Department of MSU. The MRI protocol lasted approximately 40 minutes. During scanning
sessions patients were asked to lie still with their eyes open. First and higher-order shimming
procedures were carried out to improve magnetic field homogeneity. The scanning procedure
included three MRI sequences: resting state fMRI, DTI, and 3D magnetization-prepared rapid
76

acquisition gradient echo (MPRAGE). The resting state-fMRI data collection involved a 7minute functional scan with the following parameters: 38 contiguous 3-mm axial slices in an
interleaved order, time of echo (TE) = 27.7 ms, time of repetition (TR) = 2500 ms, flip angle =
80o; field of view (FOV) = 22 cm x 22cm, matrix size = 64 x 64, ramp sampling with the first
four data points discarded. The first four data points were excluded from analysis due to
enhanced longitudinal magnetization in the first few scans. Each volume of slices was acquired
164 times. Next, diffusion-weighted images were acquired using a dual spin-echo echo-planar
imaging sequence for 12 minutes and 6 seconds with the following parameters: 48 contiguous
2.4-mm axial slices in an interleaved order, FOV =22cm x 22cm, matrix size=128 x 128, number
of excitation (NEX) = 2, echo time (TE) = 77.5 ms, repetition time (TR) = 13.7 s, 25 diffusionweighted volumes (one per gradient direction) with b=1000 s/mm2, one volume with b = 0, and
parallel imaging acceleration factor = 2. Finally, 180 T-1 weighted 1-mm3 isotropic volumetric
inversion recovery fast spoiled gradient-recalled images were acquired (10 minute scan time).
The whole brain was covered with the following parameters: TE=3.8 ms, TR of acquisition
=8.6ms, time of inversion (T1) = 831ms, TR of inversion 2332ms, flip angle=8°, FOV =25.6
cm×25.6 cm, matrix size=256 × 256, slice thickness =1 mm, and receiver bandwidth= ± 20.8
kHz.
DTI analyses
DTI data were manually preprocessed using the basic processing steam for FSL (the
FMRIB Software Library)188 . MR images from the scanner were converted from dicom to nifti
format, for registration and brain extraction followed by eddy- current distortion and motion
correction. Finally, DTIFIT was run to generate diffusion weighted maps, including the FA map.
The tract based spatial statistics (TBSS) program was run after DTIFIT in FSL to compute group
level statistics of FA (procedure outlined below).
77

Analyses of DTI images were done in two parts after all images were pre-processed. The
first included the analysis of newly recruited MCI subjects (n=12) with a 25-direction DTI
sequence. Imaging data from MCI subjects previously collected in our lab with a 25-direction
DTI scan (n = 6) were added to the sample for a total group of 18 subjects. The second included
the analysis of 19 MCI subjects whose data was previously collected within the COGENT lab
and used a 6-diffusion-weighted direction DTI sequence (methods outline below). The two
groups were analyzed separately due to differences in the average eigenvalues computed when
using 6 compared to 25 directions; the accuracy of estimation may improve at a voxel level as
diffusion directions increase,189 therefore the two groups were not combined . Subjects who met
the inclusion criteria and had BMI data were included in the following analyses of white matter
tract integrity using TBSS.
MRI acquisition in 6 diffusion weighted directions
MCI subjects previously collected were analyzed using the methods of Bozoki et al 2012.
Briefly, scan time was shorter at 4 minutes and 50 sec with 40 axial slices collected using a spin
echo EPI pulse sequence with TE = 69.3 ms and TR = 10,000ms. The in-plane resolution = 3
mm, slick thickness = 3mm, interslice gap = 0 mm, 240 mm FOV (80 x 80 matrix), and NEX =
4. For this study the diffusion encoding was collected in six non-collinear directions with b-value
of 1,000 s mm-2. DTI images were interpolated on the scanner to a voxel size of 0.9375 x 0.9375
x 3mm3.
Tract Based Spatial Statistics (TBSS)
Technique adapted from Smith et al 2006. Voxel-wise statistics were performed by tractbased spatial statistics (TBSS, version 1.2)190, a part of the FSL program188. All subjects’ FA
data were first aligned into a common space using the FMRIB’s nonlinear image registration tool
(FNIRT), which uses a b-spline representation of the registration warp field191. We then
78

registered the FA images to the JHU-DTI atlas as a common space in order to correspond the
results from voxel wise and ROI analyses. Next, the mean FA image was created and thinned to
generate a mean FA skeleton that represents the centers of all tracts common to all subjects. Each
subject’s aligned FA data was projected onto this skeleton and the resulting data were fed into
voxel wise and ROI-based cross-subject statistics. ROI analysis that is limited to the TBSS
skeleton was performed using the deep WM atlas (ICBM-DTI-81 white-matter atlas) developed
by the Johns Hopkins University (JHU)192. Mean values of a diffusion metric for selected ROI
segmentations were extracted from each participant.
BMI groups
For this analysis we used MCI subjects with available BMI data. Between groups t-tests
were used to compare FA values between a combined normal weight/over-weight group
(NW/OW; BMI < 30) and obese group (OB; BMI ≥ 30).
Statistical analysis
Voxel wise statistics were performed by general linear model, a part of the FSLrandomise program. Diffusion metrics (FA) were compared between groups in the TBSS
program. Threshold-free cluster enhancement corrected for multiple comparisons with 5,000
permutations. These corrected maps were further thresholded by P<0.05. Twenty-six white
matter tracts were assessed voxel by voxel as surviving multiple comparisons and P-value
thresholding on the WM skeletons. The final statistics included thresholding the mean_FA
skeleton, performing a t-test between groups followed by permutation testing (500 times) and
test fully corrected for multiple comparisons across time.

79

Results – MSU COGENT
DTI measures were compared across BMI groups to assess whether the structural
integrity of brain white matter differed between groups. After processing the two groups (25direction and 6-direction sequences) through the TBSS processing stream independently, the
mean FA skeletons showed no significant differences between the NW/OW (n = 9) and OB (n =
9) groups. The raw (unthresholded) and multiple correction images for FA in both the 25 and 6direction scans were non-significant (25 direction, Figure 4.1).

FIGURE 4.1. The 25-direction MSU-COGENT FA results comparing NW/OW and OB groups.
Multiple corrections were computed per voxel and overlaid on the mean FA skeleton. The green
trace indicates the mean FA comparison between groups and is non-significant shown in a
coronal section on the left and a sagittal section on the right. Abbreviations: S, superior; R, right;
L, left; I, inferior.

80

METHODS - ADNI
Participants
MRI data were obtained from the ADNI database (adni.loni.ucla.edu). The ADNI was
launched in 2003 by the National Institute on Aging (NIA), the National Institute of Biomedical
Imaging and Bioengineering (NIBIB), the Food and Drug Administration (FDA), private
pharmaceutical companies and non-profit organizations, as a $60 million, 5-year public-private
partnership. The primary goal of ADNI has been to test whether serial magnetic resonance
imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and
neuropsychological assessment can be combined to measure the progression of MCI and early
AD. Determination of sensitive and specific markers of very early AD progression is intended to
aid researchers and clinicians to develop new treatments and monitor their effectiveness, as well
as lessen the time and cost of clinical trials.
ADNI subjects were recruited from over 50 sites across the U.S. and Canada. The initial
goal of ADNI was to recruit 800 subjects but ADNI has been followed by ADNI-GO and ADNI2. To date these protocols have recruited over 1500 adults ages 55 to 90, to participate in the
research, consisting of cognitively normal older individuals, people with early or late MCI, and
people with early AD. Clinical, behavioral and neuroimaging data are collected for each patient
with up to 54 months of follow up per patient. MRI methods, procedures and preprocessing by
ADNI have been previously described140 and can also be accessed at the following website: afniinfo.org
Inclusion criteria for MCI set forth by ADNI, include; a MMSE score between 24 – 30, a
subjective memory complaint by the patient or caregiver, objective memory loss measured by the
Wechsler Memory Scale Logical Memory II, a global Clinical Dementia Rating (CDR) of 0.5,
preserved activities of daily living, and the absence of dementia.
81

MRI acquisition of ADNI data
The MRI acquisition of ADNI data has been previously described (Chapter 2-Methods).
In brief, all subjects underwent whole-brain MRI scanning on 3-Tesla GE Medical Systems
scanner, on at least one of two occasions (baseline and 6 months). T1-weighted IR-FSPGR
(spoiled gradient echo) sequences (256×256 matrix; voxel size = 1.2×1.0×1.0 mm3; TI=400 ms;
TR = 6.98 ms; TE = 2.85 ms; flip angle = 11°), were collected as well as diffusion-weighted
images (DWI; 35 cm field of view, 128×128 acquired matrix, reconstructed to a 256×256 matrix;
voxel size: 2.7×2.7×2.7 mm3; scan time = 9 min; more imaging details may be found at,
http://adni.loni.usc.edu/wp-content/uploads/2010/05/ADNI2_GE_3T_22.0_T2.pdf. Forty-six
separate images were acquired for each DTI scan: 5 T2-weighted images with no dedicated
diffusion sensitization (b0 images) and 41 diffusion-weighted images (b=1000 s/mm2).
DTI and inversion-recovery spoiled gradient recalled (IR-SPGR) T1-weighted imaging data were
acquired on several General Electric 3 T scanners using scanner specific protocols. Briefly, DTI
data were acquired with a voxel size of 1.372 × 2.70 mm3, 41 diffusion gradients and a b-value
of 1000 s/mm2. In order to increase data uniformity, the data underwent a standardized
preprocessing procedure at the ADNI project. All imaging protocols and preprocessing
procedures are available at the ADNI website (http://adni.loni.usc.edu/methods/)
DTI analyses
FSL and TBSS procedures were the same as stated above [COGNENT-DTI analyses].
The same JHU ‘EVE’ atlas was used to calculate FA and MD measures across the 26 DTI tracts.
These analyses were done by the Laboratory of Neuro Imaging at USC and made available on
the ADNI research website. The DTI-template, with corresponding white matter tract atlas, was
registered to each individual subject using the JHU ‘EVE’ DTI atlas192, which included a total of

82

26 regions and 5 summary regions. Summary measures of each tract were computed based off of
a subject’s mean FA skeleton.
Statistical analysis
To determine whether BMI affected brain connectivity we examined differences in DTI
measures across BMI groups. We categorized subjects BMI into three groups of, normal weight
(NW, BMI: <25 kg/m2), overweight (OW, BMI: 25 – 29.9 kg/m2) and obese (OB, BMI: > 30
kg/m2). White matter tracts from the JHU ‘EVE’ DTI atlas192, a total of 26, were averaged
between hemispheres. An analysis of variance (ANOVA) model was used in each experiment to
compare the summary ROI measures (FA, MD) of white matter tracts across BMI groups,
corrected for age and gender. Statistical significance was set at p<0.05. Statistical analyses were
performed using SPSS 22.0 (SPSS inc., Chicago, IL).
Results - ADNI
MCI subjects with a MRI diffusion sequence at screening (available in ADNI-GO and
ADNI-2 only) totaled 102. We measured the effect of BMI on DTI measures of anisotropy (FA)
and diffusivity (MD) in all 26 white matter tracts using an ANOVA model. A large percentage of
the sample were overweight (49%) and obese (24.5%). Results indicated differences in BMI in 5
out of 26 tracts for FA, and 4 out of 26 for MD. (Table 4.1.). For each of these regions FA values
were lower and MD values were higher for NW subjects. This indicates adverse microstructural
changes of lower fiber integrity and myelination for NW subjects.
Based on this initial summary analysis we then analyzed group differences using a
MANOVA model in order to control for age and gender, two variables that affect white matter
structure. First, we measured the main effect of BMI on FA values and found there were no
longer significant differences, F (52, 148) = 1.17, p = 0.23; Wilks’Λ = 0.502, partial η 2 = 0.29,
nor were was there an effect in the full factorial MANOVA model with covariates, F (52, 144) =
83

1.18, p = 0.22; Wilks’Λ = 0.49, partial η 2 = 0.30. There was also not a main effect of BMI on
MD values, F (52, 148) = 1.08, p = 0.28; Wilks’Λ = 0.525, partial η 2 = 0.28, nor in the full
factorial model, F (52, 144) = 1.10, p = 0.33; Wilks’Λ = 0.512, partial η 2 = 0.28.

TABLE 4.1. White matter tracts that significantly differed across BMI groups for fractional
anisotropy and mean diffusivity measures
Tract
NW
OW
OB
F value P value
Tract Name
label
FA
ICP Inferior cerebellar
0.035
0.308 (.01) 0.328 (.04) 0.329 (.03)
3.46
peduncle
SCP Superior cerebellar
peduncle

0.411 (.03)

0.426 (.02)

0.415 (.03)

3.38

0.038

PTR Posterior thalamic
radiation

0.365 (.04)

0.387 (.03)

0.380 (.04)

3.01

0.054

0.254 (.05)

0.277 (.03)

0.279 (.04)

3.03

0.053

0.253 (.07)

0.299 (.06)

0.286 (.08)

4.24

0.017

MD
PTR Posterior thalamic
radiation

1.0x10-3
(1.3x10-4)

9.4x10-4
(1.0x10-4)

9.5x10-4
(1.1x10-4)

3.72

0.028

FX_ST Fornix (cres) / Stria
terminalis

1.4x10-3
(3.0x10-4)

1.3x10-3
(2.6x10-4)

1.2x10-3
(2.8x10-4)

3.58

0.032

FX-ST Fornix (cres) / Stria
terminalis
TAP
Tapetum

SCC Splenium of corpus
1.2x10-3
1.1x10-3
1.1x10-3
3.45
0.036
callosum
(1.8x10-4) (1.2x10-4) (1.7x10-4)
TAP Tapetum
1.8x10-3
1.5x10-3
1.5x10-3
3.57
0.032
)))(3.9x10(3.2x10-4)
(4.3x10-4)
Mean fractional anisotropy (FA) and
mean diffusivity
(MD) values
that significantly differed in
4) as mean (standard error). The statistic is an ANOVA F
an ANOVA model. Values are presented
statistic for continuous variables. Significance was set as p < 0.05.
Abbreviations: NW, normal weight; OW, overweight; OB, obese.

84

Summary and conclusions – MSU COGENT & ADNI DTI analysis
We used two different samples to analyze the relationship between BMI and white
matter microstructure. First, the MSU COGENT dataset, a single site, small sample of MCI
subjects that were recruited from the mid-Michigan area compared white matter microstructure
values of FA across two weight groups (NW/OW and OB). In this study the administration of all
neuropsychological, behavioral scales, and DTI processing were completed locally. Due to the
limitation of a small sample size in the COGENT study, data from the ADNI, a large, multi-site
study, was also examined across three BMI weight groups (NW, OW, OB). The DTI preprocessing and TBSS analysis of FA and MD values in this dataset were completed by ADNI
investigators and available for download. In both, BMI was used as a grouping variable to
compare white matter microstructure.
In our analyses of the samples, we did not find significant differences in white matter
structural measures across BMI groups. In the COGENT study, there were no between group
differences in the raw and corrected FA values. In the ADNI study, specific white matter tracts
showed microstructural changes of decreased FA and increased MD values, in the brain stem,
corpus callosum and temporal lobes, but this effect was lost in the MANOVA model that took
into account covariates of age and gender. The initial differences in ADNI all indicated poorer
white matter structural integrity for NW MCI subjects. This was similar to our brain volume
results (Chapter 2) on regional gray matter volumes.
In older adults with normal cognition, DTI studies indicate an opposite relationship than
we found, regarding white matter integrity and weight83,85,186. Mild cognitive impairment specific
white matter changes included the posterior cingulate cortex and hippocampus193 while obesity
prominently effects the midbrain and brain stem nuclei194. A possible explanation for our
findings of white matter measures being negatively altered in NW subjects may be that these
85

subjects are more vulnerable to neural changes associated with their age and likely underlying
AD pathology. Fractional anisotropy measures of the fornix and corpus callosum are altered in
OB subjects, and their changes in NW may indicate that these regions are related in some way to
body weight independent of MCI. Interestingly, our volumetric results (Chapter 2.) indicated
decreased volume of the amygdala and hippocampus, which make connections with the fornix
and stria terminalis. Further, the tapetum, a bundle of axons branching off the corpus callosum
laterally toward the inferior temporal lobe, correspond with the decreased volume in the banks of
superior temporal sulcus and inferior middle temporal gyrus also seen in NW subjects. White
matter changes of these tracts may suggest early microstructural changes that precede decreases
in brain volume and eventual cognitive dysfunction as the severity of MCI progresses over time.
Some limitations may be responsible for the null findings of these analyses. First, the
sample size of the MSU COGENT dataset was very small which likely affected the ability to
measure differences between even two groups. The ADNI sample was larger with 102 subjects
but this analysis was still underpowered to measure differences across three BMI groups. In our
ADNI analysis of brain volume across BMI groups we had over 600 subjects and were able to
show a medium effect of BMI (partial η 2 = 0.08) on brain volume. Second, the ADNI data in
this study were processed by another lab, not manually computed from the original DTI images.
Because of this, the statistics could not be computed within the TBSS procedure thus preventing
group level statistics to be drawn directly from the group mean FA and MD skeletons. With the
knowledge of differences in brain volume, with regard to BMI group, hypotheses can be
generated that outline a targeted approach for understanding the effects of weight on brain white
matter a priori, that expand on these analyses and measure changes prospectively. Third, BMI is
not the most accurate measure of adiposity and may not definitively establish groups that reflect
the true metabolic effects of adiposity on the brain. A study of MCI subjects that uses waist
86

circumference measures or calculates body fat percentages for each participant may provide a
more accurate measure of over-weight and obesity status. Moreover, co-morbid metabolic
conditions, such as type 2 diabetes mellitus, hypertension, and serum inflammation markers,
must also be taken into account to assess their influence on brain structure compared to BMI
alone. Many metabolic conditions are associated with specific white matter changes such as
lesions and white matter hyperintensities195,196. Finally, utilizing weight groups with cut-offs in
values that do not border each other may demonstrate more clearly the differences in brain
structure related to adiposity. Overall, increasing the sample sizes of these groups and manually
processing the DTI scans may offer a more sensitive approach.
Modern neuroimaging allows us to visualize disease pathology in vivo. In MCI, there
may be altered brain structure related to adiposity that affects both neuronal cell bodies and
impulse transmission via axon bundles. It is not possible to determine a direction of change:
whether white matter damage results in grey matter atrophy or vice-versa. Further investigation
into white matter changes of these tracts in MCI subjects is necessary in both normal and overweight to determine what types of microstructural changes occur that may precede decreases in
brain volume and eventual cognitive dysfunction as the severity of MCI progresses over time.

87

THE EFFECT OF OBESITY ON BRAIN CORTICAL THICKNESS IN MILD
COGNITIVE IMPAIRMENT
Introduction
Cortical thickness in neuroimaging is measured as the space between the pial surface and
the beginning of brain white matter. The thickness of the cortex is mostly determined genetically
but changes throughout life can occur as a result of diseases197,198. In MCI, cortical thinning
present in frontal brain regions can reliably differentiate progressive compared to stable MCI
subjects over time199. This can even be more reliable than cognitive test scores. In obese young
and older adults with normal cognition, decreased cortical thickness is also evident compared to
normal weight adults137. Some studies suggest that cortical thinning precedes changes in volume,
making cortical thickness measurement a possible early indicator of pathologic influences.
Cortical thickness changes related to obesity are located in specific brain regions. One
study found that a high BMI and visceral adipose tissue were independently connected to
reduced levels of cortical thickness within the lateral occipital area, inferior temporal lobule, the
precentral gyrus and the inferior parietal brain region200. Hassenstab et al. examined the cortical
thickness of the cognitive control network (CCN, described further in Introduction –
Neuroimaging of Obesity) between three groups; successful weight loss maintainers (SWLM),
never obese lean (NOL), and obese (OB) individuals. They found that SWLM had a thicker
cortex compared to OB and more prefrontal and temporal brain activation when shown pictures
of foods high in calories. Obese individuals also had cortical thinning within the anterior
cingulate and posterior parietal cortices. They found structural differences within CCN regions
between OB and NOLs, yet these regions in SWLM did not significantly differ from the OB
group 75,201. These findings suggest cortical changes are plastic in regards to obesity. The brain

88

may even alter its structure in relation to positive metabolic changes, or in this case attaining a
healthier body weight.
Currently, no studies have investigated cortical thickness changes related to obesity in
MCI. Based on previous finding on cortical changes related to obesity and thickness described
above, we hypothesized that OB MCI subjects will have decreased cortical thickness averages
compared to normal weight (NW) and overweight (OW) groups. This study focused primarily on
region directly related to either MCI or obesity and included 17 regions from the frontal, parietal
and temporal lobes. The full list of regions is located in Table 4B.

Methods
Participants
The analysis for this project began in 2015 with the collection of ADNI MCI subject data
that met the inclusion criteria for MCI with a baseline MRI scan and FreeSurfer analysis. The
study procedures, MCI inclusion criteria and MRI acquisition protocol are documented in
Chapter 4 [MRI acquisition of ADNI data], and FreeSurfer analysis methods are found in
Chapter 2 [ADNI FreeSurfer methods].
Cortical thickness analysis
Alzheimer’s disease neuroimaging initiative MCI subjects with a screening MRI scan and
FreeSurfer analysis were included in the data analysis of cortical thickness measurements across
BMI groups. All three phases of ADNI were used and merged into one dataset. Subjects with a
non-accelerated T1 image, that passed the ‘Overall QC’ a quality check for accurate cortical
parcellation by FreeSurfer of the frontal, parietal, occipital and temporal lobes were included in
this study. From this sample, demographic (age, sex, BMI), cognitive (MMSE) and behavioral
(NPI-Q and GDS) measures were then matched to each MCI subject. A total of 76 brain regions

89

were included in the FreeSurfer analysis, from this list we selected regions previously identified
as related to obesity. The left and right hemisphere data was combined for each region, resulting
in a total of 17 regions. These regions were analyzed preliminarily to our cortical volume
measures within chapter 2. The few papers that highlight cortical thickness changes related to
weight, showed significant changes in these specific regions and the full list is available in Table
4B.
FreeSurfer region of interest analysis
The ROI thickness average (TA) values were generated using FreeSurfer and made
available on the ADNI website was for download. Data was obtained from the UCSFFX
spreadsheet of FreeSurfer Version 5.1. Subjects that had a non-accelerated T1 screening MRI,
whose status was complete and had an overall quality check (QC) of Pass for all QC regions
were included in the analyses.
Statistical analysis
An ANOVA model was used to compare average thickness measures for each region
across BMI groups. First, the main of effect of BMI was calculated across all 17 regions. Then, a
full factorial model included age and education as covariates. The analysis was conducted using
SPSS version 22. Statistical significance was set at p < 0.05.
Results
There were 635 MCI subjects that met the inclusion criteria and had MRI data that passed
the overall quality check for FreeSurfer indicating successful cortical parcellation. The
demographic, behavioral and cognitive measures across BMI groups were the same as those
reported in Table 2.1.

90

Brain cortical thickness
Baseline cortical thickness averages were compared across BMI groups to determine
whether there was a difference in mean thickness related to the subject’s BMI. First, a
MANOVA model measured the main effect of BMI on regional brain volumes. There was a
statistically significant difference in brain thickness averages based on BMI group, F (34, 1234)
= 1.61, p = 0.015; Wilks’Λ = 0.917, partial η2 = 0.042. Next, age and education were added as
covariates based on previous research indicating their relationship with the cortical thickness of
the brain. In this full factorial model, BMI, and age were independent significant contributors to
volume differences in the overall multivariate tests, although education was not: BMI: F
(34,1226) = 1.60, p = 0.017; Wilks’Λ = 0.917, partial η2 = 0.042; Age: F (17, 613) = 15.171, p <
.005; Wilks’Λ = 0.704, partial η2 = 0.296; Education: F (17, 613) = 1.468, p = 0.10; Wilks’Λ =
0.961, partial η2 = 0.039. In the full factorial model, 3 (out of 17) regions significantly differed
by BMI, the Precuneus, lateral occipital cortex and the post central gyrus. Table 4.2 shows the
brain regions that differed in cortical thickness measures across BMI groups. For all significant
comparisons cortical thickness was lower in NW subjects compared to OW and/or OB.

TABLE 4.2. Brain regions that significantly differed in cortical thickness average measures
across BMI groups of MCI subjects
Brain Region
NW
OW
OB
F
p value
Precuneus

2.11 (0.18)

2.17 (0.19)

2.21 (0.19)

7.36

0.001a,b

Lateral Occipital
Cortex
Post Central gyrus

2.03 (0.18)

2.08 (0.17)

2.10 (0.17)

5.06

0.007a,b

1.82 (0.16)

1.86 (0.17)

1.87 (0.15)

3.11

0.045b

Values are reported as the mean (standard deviation (SD)). The MANOVA model corrected for
age and education. Superscripts indicate the direction of the differences after Bonferroni method
correction: a = NW < OB, b = NW < OW. Statistical significance is set at p < 0.05.
Abbreviations: NW, normal weight; OW, overweight; OB, obese.

91

Summary and conclusions – Thickness averages
In this sub-study of ADNI FreeSurfer data, we sought to identify whether BMI was
related to decreased cortical thickness in MCI subjects. The available data on the effect of weight
on cortical thickness measures is limited and has not been studied in MCI. We again combined
the 3 phases of ADNI and found small yet significant effects of BMI on cortical thickness in a
MANOVA model that included age and educations covariates. In our analysis, NW subjects had
significantly lower cortical thickness measurements in 3 out of 17 brain regions compared to OW
and OB subjects
In our study, only two regions overlapped with the volume reductions we saw in NW
subjects, the precuneus and the lateral occipital cortex. An additional region, the post central
gyrus, also had a lower thickness average in NW compared to OB subjects, yet there were no
volume differences for this region. To better compare our volumetric and cortical thickness
analyses, we also analyzed cortical brain volumes using only the 17 regions within the cortical
thickness analyses. The MANOVA model for this comparison showed a significant effect of
BMI on brain volumes, F (34, 1224) = 1.86, p = 0.002; Wilks’Λ = 0.904, partial η2 = 0.05. All of
the regions that were significant in the original volume analysis and included in this smaller
analysis remained significant; middle temporal gyrus (F = 5.26, p = 0.005), rostral anterior
cingulate gyrus (F = 4.04, p = 0.018), insula (F = 3.56, p = 0.029), precuneus (F = 7.33, p =
0.001), and the lateral occipital gyrus (F = 4.90, p = 0.008).
Initially, this study was designed to provide a complete assessment on brain structural
changes with the hope of complementing the investigation of grey matter volume and white
matter microstructure. However, there were not consistent changes in brain regions across our
multiple neuroimaging modalities. This inconsistency may be due to the wide range of ages
included in our analyses. While reduced cortical thickness has been hypothesized as an early
92

indicator of cortical change199 this may only be true when investigating discreet age ranges. A
recent study has shown that as a person ages there is a dynamic relationship in the direction and
magnitude of changes to cortical thickness, surface area and the total volume of brain regions197.
For example, this study demonstrated that in adults with normal cognition ranging in age from 23
– 87 years, as age increased there was accelerated changes in temporal and occipital brain
regions, while frontal and anterior cingulate regions decelerated in overall volume, surface area
and thickness. These finding may help explain why only two regions overlapped between our
cortical thickness and cortical volume analyses; due to the significant differences in age across
BMI groups (Table 2.1).
Our study had a large age range of over 30 years (age range 55 – 89 years) which may
require targeted analysis of Middle-Age and Senior groups as well as investigating discreet age
ranges within these two larger age categories. More work needs to be done to better understand
the effect obesity has on brain structure in older adults with normal cognition and in those with
pathological changes that are a signature of MCI or Alzheimer’s disease. Future studies should
assess volume and cortical thickness measures together, with a specific focus on measuring the
relationship of BMI and brain structure within discreet age groups of MCI subjects.

93

APPENDIX

94

APPENDIX
TABLE 4A. Regions of interest from the JHU ‘EVE’ atlas white matter tract list generated in
FreeSurfer
ROI Label
ROI definition
ACR
Anterior corona radiata*
ALIC
Anterior limb of internal capsule *
BCC
Body of corpus callosum *
CGC
Cingulum*
CGH
Cingulum (hippocampus)*
CP
Cerebral peduncle
CST
Corticospinal tract
EC
External capsule*
FX
Fornix *
FX_ST
Fornix (cres) / Stria terminalis*
GCC
Genu of corpus callosum*
ICP
Inferior cerebellar peduncle
IFO
Inferior fronto-occipital fasciculus*
ML
Medial lemniscus
PCR
Posterior corona radiata *
PLIC
Posterior limb of internal capsule*
PTR
Posterior thalamic radiation
RLIC
Retrolenticular part of internal capsule*
SCC
Splenium of corpus callosum*
SCP
Superior cerebellar peduncle
SCR
Superior corona radiate*
SFO
Superior fronto-occipital fasciculus*
SLF
Superior longitudinal fasciculus*
SS
Sagittal stratum*
TAP
Tapetum *
UNC
Uncinate fasciculus*
SUMBCC
Bilateral body of the corpus callosum*
SUMCC
Bilateral full corpus callosum*
SUMFX
Bilateral fornix*
SUMGCC
Bilateral genu of the corpus callosum*
SUMSCC
Bilateral splenium of the corpus callosum*
White matter tract list from the JHU “EVE” atlas. There are a total of 26
unique tracts and 5 summary measures. The ROI label lists the atlas label and
the ROI definition indicates the full name of the white matter tract or adjacent
brain region. *White matter commissural and association fibers.

95

TABLE 4B. Selected brain regions of interest for the cortical thickness analysis generated in
FreeSurfer
ROI Label

Region Name

Precuneus

Precueus

LateralOccipital

Lateral occipital gyrus

PostCentral

Postcentral gyrus

ParsOrbital

Pars orbitalis

ParOperc

Pars opercularis

ParsTriang

Pars triangularis

ParaHipp

Parahippocampal gyrus

MiddleTemp

Middle temporal gyrus

InfTemporal

Inferior temporal gyrus

ParaCentral

Paracentral gyrus

RostralAntCing

Rostral anterior cingulate gyrus

CaudalAntCing

Caudal anterior cingulate gyrus

Insula

Insular cortex

CaudalMidFront

Caudal middle frontal gyrus

RostralMidFront

Rostral middle frontal gyrus

SuperiorFrontal

Superior frontal gyrus

PreCentral

Precentral gyrus

Seventeen brain regions were selected based on literature that indicated a relationship of the
region and structural brain changes related to weight or BMI. The ROI label indicated the
FreeSurfer label and the Region Name gives the full anatomical name for each region.

96

CHAPTER 5

CONCLUSIONS AND FUTURE DIRECTIONS

97

Prior to this work the intersections of obesity, NPS, age and cognitive decline in MCI had
not been explored. This dissertation has begun to explore these relationships. Each of the
previously mentioned elements are an independent risk factor for the ultimate development of
dementia and may or may not have a significant effect on the time to conversion additively.
While we have provided new insight into these relationships more research is necessary. It is not
well understood how NPS and obesity structurally and functionally alter the brain in the presence
of cognitive deficits, or whether those factors are additive. In addition, the direction of the
relationship linking NPS and obesity with AD and with each other is not well understood. And
results in more questions such as, does obesity cause a relatively toxic environment to neurons
thereby producing brain damage over time, or is obesity a marker for certain pre-existing brain
“weaknesses” (e.g. as seen in the cognitive control network) that enable the emergence of
cognitive impairment more readily at a later age? We do know that NPS tends to emerge de
novo at the time MCI onset, whereas obesity is highly likely to have persisted throughout
adolescence, adulthood, middle age and later life. Thus, NPS is more likely to be a marker for
emerging brain pathology while obesity is more likely to indicate either a pre-existing or an
ongoing state of independent loss of connectivity.
This research has provided new information on the effects of obesity in MCI. First, it was
demonstrated that in MCI obese subjects have a higher prevalence of NPS measured by the NPIQ. Specifically, OB had a higher percentage of affective symptoms of depression and anxiety
and when these symptoms were present, they were most severe. In Chapter 1. type-2-diabetes
mellitus and obstructive sleep apnea also have specific NPS associated with these conditions.
Throughout each of the preceding chapters, when measured, obese subjects had significantly
higher mean NPI-Q scores compared to NW subjects. However, we were not able to draw a
direct connection between obesity, NPS and deficits in cognition.
98

Many factors emerged in Chapters 2 – 4 that may influence or mediate the effects of
weight on either cognition, or brain structure. The effects of obesity on cognitive decline are
demonstrated when OB are middle age. Studies that have shown this relationship sampled
individuals with normal cognition. In our studies, we did not see a cross sectional difference in
cognitive scoring related to obesity. In our cross-sectional imaging study, we found that normal
weight MCI subjects had lower brain volumes and similar cognitive scores as OW and OB
subjects. While we expected to see deficits primarily related to obesity Chapters 2 and 3 of this
dissertation identified that difference related to BMI were driven by the normal weight group. In
Chapter 2, NW subjects had lower brain volumes and in Chapter 3 lower baseline cognitive
scores (ADAS-cog) followed by greater cognitive changes over two years. These studies suggest
that in regards to increased risk for cognitive decline and progression to dementia, normal weight
MCI subjects may be a high-risk group. Despite obese subjects having higher NPS and more
metabolic comorbidities, NW subjects were more vulnerable to structural brain changes, which
likely preceded the cognitive changes that occurred after two years.
Another well-known factor that we simply grouped was age. This was done to highlight
that MCI comprises two very different life stages, an approach that has not been taken in current
research. In our samples, the NW group was consistently older than the OB group. The age of
NW subjects is likely a large driving factor that plays a role in the expression of dementia related
pathologic changes in a way that is more dominate than mood and physiologic changes related to
metabolic conditions. So, where does that leave future research on obese MCI subjects? One
could argue that the young age of OB subjects should not be overlooked. Even if their clinical
course to dementia is slower, the development of MCI at a young age could be related an early
development of AD. With a large prevalence of obese MCI subjects this groups needs better
characterization in future work, specifically, middle age obese persons with MCI.
99

The goal of this dissertation was to introduce obesity as a measurable risk factor for
Alzheimer’s disease within MCI. The ultimate goal of nearly all MCI research involves
understanding the development of Alzheimer’s disease. Thus, research studies that identify BMI
groups in MCI have done so only in cases where progression to Alzheimer’s disease is being
studied directly, such as longitudinal and survival analysis research. The focus of this research
took a new direction, to study and characterize obesity in MCI directly. With the high prevalence
of obesity on the rise from adolescences to older adulthood, the effects of chronic obesity on
brain and behavior over a lifetime are relatively unknown. The addition of a pathological
disorder to the aging process, such as MCI, introduces an avenue where there is no knowledge
regarding the effect that increased weight has on the presentation of the disease, or what
symptoms reflect metabolic changes and not the disease pathology. In this research, we were not
able to take into account some key features that may answer such complex questions regarding
the interaction between chronic obesity and Alzheimer’s disease pathology. Future directions of
this research involve tracking the chronicity of obesity and using that time variable as a control
measure for brain structure changes. In addition, designing studies prospectively where changes
in BMI are tracked over time, along with cognitive, brain imaging and behavioral measures.
Further, many of the studies in this dissertation categorized BMI into three categories of NW,
OW and OB, but a study design that uses regression analyses may identify relationships of
weight and brain structure that provide interpretations on associative relationship that could
change under different conditions.

100

REFERENCES

101

REFERENCES

1.

Petersen R, Smith G, Waring S, Ivnik R, Tangalos E, Kokmen E. Mild Cognitive
Impairment: Clinical Characterization and Outcome. Arch Neurol. 1999;56:303-309.

2.

Petersen R, Doody R, Kurz A, et al. Current concepts in mild cognitive impairment. Arch
Neurol. 2001;58(12):1985-1992.

3.

Petersen R. Mild cognitive impairment as a diagnostic entity. J Intern Med. 2004;256:183194.

4.

Palmer K, Berger A, Monastero R, Winblad B, Bäckman L, Fratiglioni L. Predictors of
progression from mild cognitive impairment to Alzheimer disease. Neurology.
2007;68(19):1596-1602.

5.

Winblad B, Palmer K, Kivipelto M, et al. Mild cognitive impairment-beyond
controversies, towards a consensus: report of the International Working Group on Mild
Cognitive Impairment. J Intern Med. 2004;256(3):240-246. doi:10.1111/j.13652796.2004.01380.x.

6.

Jack C, Petersen R, Xu Y, et al. Prediction of AD with MRI-based hippocampal volume in
mild cognitive impairment. Neurology. 1999;52(7):1397.

7.

Leon MJ De, Convit A, Wolf OT, et al. Prediction of cognitive decline in normal elderly
subjects with 2- [18 F] fluoro-2-deoxy- D-glucose/ positron-emission tomography (
FDG/PET ). PNAS. 2001;98(19):10966-10971.

8.

Hardy J, Selkoe D. The amyloid hypothesis of Alzheimer’s disease: progress and
problems on the road to therapeutics. Science (80- ). 2002;297:353-356.
doi:10.1126/science.1072994.

9.

Crews L, Masliah E. Molecular mechanisms of neurodegeneration in Alzheimer’s disease.
Hum Mol Genet. 2010;19(R1):R12-20. doi:10.1093/hmg/ddq160.

10.

Schuff N, Woerner N, Boreta L, et al. MRI of hippocampal volume loss in early
Alzheimer’s disease in relation to ApoE genotype and biomarkers. Brain.
2009;132(4):1067-1077.

11.

Roses A. Apolipoprotein E alleles as risk factors in Alzheimer’s disease. Annu Rev Med.
1996;47:387-400.

12.

Stern Y, Gurland B, Tatemichi T, Tang M, Wilder D, Mayeux R. Influence of education
and occupation on the incidence of Alzheimer’s disease. JAMA. 1994;271(13):1004-1010.

13.

Ott A, Breteler M, van Harskamp F, et al. Prevalence of Alzheimer’s disease and vascular
dementia: association with education. The Rotterdam study. BMJ. 1995;310:970-973.
102

14.

Amieva H, Jacqmin-Gadda H, Orgogozo J, et al. The 9 year cognitive decline before
dementia of the Alzheimer type: a prospective population-based study. Brain.
2005;128(5):1093-1101.

15.

Sperling R, Aisen P, Beckett L, et al. Toward defining the preclinical stages of
Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s
Association workgroups on diagnostic guidelines for Alzheimer’s. Alzheimer’s Dement.
2011;7(3):280-292. doi:10.1016/j.jalz.2011.03.003.Toward.

16.

Bozoki A, Giordani B, Heidebrink J. Mild cognitive impairments predict dementia in
nondemented elderly patients with memory loss. Arch Neurol. 2001;58(3):411-416.

17.

Fischer P, Jungwirth S, Zehetmayer S, et al. Conversion from subtypes of mild cognitive
impairment to Alzheimer dementia. Neurology. 2007;68(4):288-291.

18.

McDonald CR, Gharapetian L, McEvoy LK, et al. Relationship between regional atrophy
rates and cognitive decline in mild cognitive impairment. Neurobiol Aging.
2012;33(2):242-253. doi:10.1016/j.neurobiolaging.2010.03.015.

19.

Tartaglia MC, Rosen HJ, Miller BL. Neuroimaging in dementia. Neurotherapeutics.
2011;8(1):82-92. doi:10.1007/s13311-010-0012-2.

20.

Bruen PD, McGeown WJ, Shanks MF, Venneri A. Neuroanatomical correlates of
neuropsychiatric symptoms in Alzheimer’s disease. Brain. 2008;131(Pt 9):2455-2463.
doi:10.1093/brain/awn151.

21.

Duara R, Loewenstein D a, Greig MT, et al. Pre-MCI and MCI: neuropsychological,
clinical, and imaging features and progression rates. Am J Geriatr Psychiatry.
2011;19(11):951-960. doi:10.1097/JGP.0b013e3182107c69.

22.

Petrovic P, Kalso E, Petersson K, Ingvar M. Placebo and opioid analgesia--imaging a
shared neuronal network. Science (80- ). 2002;295:1737-1740.

23.

Dickerson BC, Stoub TR, Shah RC, et al. Alzheimer-signature MRI biomarker predicts
AD dementia in cognitively normal adults. Neurology. 2011;76(16):1395-1402.
doi:10.1212/WNL.0b013e3182166e96.

24.

Bozoki AC, Korolev IO, Davis NC, Hoisington L a, Berger KL. Disruption of limbic
white matter pathways in mild cognitive impairment and Alzheimer’s disease: a
DTI/FDG-PET study. Hum Brain Mapp. 2012;33(8):1792-1802. doi:10.1002/hbm.21320.

25.

Steffens DC, McQuoid DR, Payne ME, Potter GG. Change in hippocampal volume on
magnetic resonance imaging and cognitive decline among older depressed and
nondepressed subjects in the neurocognitive outcomes of depression in the elderly study.
Am J Geriatr psychiatry. 2011;19(1):4-12. doi:10.1097/JGP.0b013e3181d6c245.

103

26.

Laakso M, Soininen H, Partanen K, et al. Volumes of hippocampus, amygdala and frontal
lobes in the MRI-based diagnosis of early Alzheimer’s disease: correlation with memory
functions. J Neural Transm. 1995;9(1):73-86.

27.

Folstein M, Folstein S, McHugh P. Mini-Mental State: a practical method for grading the
cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189-198.

28.

Visser PJ, Verhey FRJ, Hofman PAM, Scheltens P, Jolles J. Medial temporal lobe atrophy
predicts Alzheimer’s disease in patients with minor cognitive impairment. J Neurol
Neurosurg Psychiatry. 2002;72:491-497.

29.

Delano-Wood L, Bondi MW, Sacco J, et al. Heterogeneity in mild cognitive impairment:
differences in neuropsychological profile and associated white matter lesion pathology. J
Int Neuropsychol Soc. 2009;15(6):906-914. doi:10.1017/S1355617709990257.

30.

Delano-Wood L, Abeles N, Sacco JM, Wierenga CE, Horne NR, Bozoki A. Regional
white matter pathology in mild cognitive impairment: differential influence of lesion type
on neuropsychological functioning. Stroke. 2008;39(3):794-799.
doi:10.1161/STROKEAHA.107.502534.

31.

Zhang Y, Schuff N, Jahng G, et al. Diffusion tensor imaging of cingulum fibers in mild
cognitive impairment and Alzheimer disease. Neurology. 2007;68(1):13-19.

32.

Mosconi L. Brain glucose metabolism in the early and specific diagnosis of Alzheimer’s
disease. Eur J Nucl Med Mol Imaging. 2005;32:486-510. doi:10.1007/s00259-005-1762-7.

33.

Mosconi L, Mistur R, Switalski R, et al. FDG-PET changes in brain glucose metabolism
from normal cognition to pathologically verified Alzheimer’s disease. Eur J Nucl Med
Mol Imaging. 2009;36:811-822. doi:10.1007/s00259-008-1039-z.

34.

Wee C, Yap P-T, Denny K, et al. Resting-state multi-spectrum functional connectivity
networks for identification of MCI patients. PLoS One. 2012;7(5):e37828.
doi:10.1371/journal.pone.0037828.

35.

Xie C, Bai F, Yu H, et al. Abnormal insula functional network is associated with episodic
memory decline in amnestic mild cognitive impairment. Neuroimage. 2012;63(1):320327. doi:10.1016/j.neuroimage.2012.06.062.

36.

Fox MD, Raichle ME. Spontaneous fluctuations in brain activity observed with functional
magnetic resonance imaging. Nat Rev Neurosci. 2007;8(9):700-711. doi:10.1038/nrn2201.

37.

Alexopoulos GS, Hoptman MJ, Kanellopoulos D, Murphy CF, Lim KO, Gunning FM.
Functional connectivity in the cognitive control network and the default mode network in
late-life depression. J Affect Disord. 2012;139(1):56-65. doi:10.1016/j.jad.2011.12.002.

38.

Zhu DCD, Majumdar S, Korolev IIO, Berger KL, Bozoki AC. Alzheimer’s Disease and
Amnestic Mild Cognitive Impairment Weaken Connections Within the Default-Mode
104

Network: A Multi-Modal Imaging Study. J Alzheimer’s Dis. 2013;34(4):969-984.
doi:10.3233/JAD-121879.
39.

Buckner RL, Andrews-Hanna JR, Schacter DL. The brain’s default network: anatomy,
function, and relevance to disease. Ann N Y Acad Sci. 2008;1124:1-38.
doi:10.1196/annals.1440.011.

40.

Binnewijzend M, Schoonheim MM, Sanz-Arigita E, et al. Resting-state fMRI changes in
Alzheimer’s disease and mild cognitive impairment. Neurobiol Aging. 2012;33(9):20182028.

41.

Damoiseaux JS, Beckmann CF, Arigita EJS, et al. Reduced resting-state brain activity in
the “default network” in normal aging. Cereb Cortex. 2008;18(8):1856-1864.

42.

Menon V. Large-scale brain networks and psychopathology: a unifying triple network
model. Trends Cogn Sci. 2011;15(10):483-506.

43.

Cole M, Schneider W. The cognitive control network: Integrated cortical regions with
dissociable functions. Neuroimage. 2007;37(1):343-360.

44.

Feldman H, Scheltens P, Scarpini E, et al. Behavioral symptoms in mild cognitive
impairment. Neurology. 2004;62(7):1199-1201.

45.

Lyketsos C, Lopez O. Prevalence of neuropsychiatric symptoms in dementia and mild
cognitive impairment. JAMA. 2002;288(12):1475-1483.

46.

Apostolova LGL, Cummings JLJ. Neuropsychiatric manifestations in mild cognitive
impairment: a systematic review of the literature. Dement Geriatr Cogn Disord.
2008;25:115-126. doi:10.1159/000112509.

47.

Palmer K, di Iulio F, Varsi AE, et al. Neuropsychiatric predictors of progression from
amnestic-mild cognitive impairment to Alzheimer’s disease: the role of depression and
apathy. J Alzheimers Dis. 2010;20(1):175-183. doi:10.3233/JAD-2010-1352.

48.

Geda YE, Roberts RO, Knopman DS, et al. The prevalence of neuropsychiatric symptoms
in mild cognitive impairment and normal cognitive aging: A population-based study. Arch
Gen Psychiatry. 2008;65(10):1193-1198. doi:10.1001/archpsyc.65.10.1193.

49.

Rozzini L, Vicini Chilovi B, Conti M, et al. Neuropsychiatric symptoms in amnestic and
nonamnestic mild cognitive impairment. Dement Geriatr Cogn Disord. 2008;25:32-36.
doi:10.1159/000111133.

50.

Teng W, Lu PH, Cummings JL. Neuropsychiatric symptoms are associated with
progression from mild cognitive impairment to Alzheimer’s disease. Dement Geriatr
Cogn Disord. 2007;24(4):253-259. doi:10.1159/000107100.

51.

Rose S, McMahon K, Janke A, et al. Diffusion indices on magnetic resonance imaging
105

and neuropsychological performance in amnestic mild cognitive impairment. J Neurol
Neurosurg Psychiatry. 2006;77:1122-1128.
52.

Edwards E, Spira A, Barnes D, Yaffe K. Neuropsychiatric symptoms in mild cognitive
impairment: differences by subtype and progression to dementia. Int J Geriatr Psychiatry.
2009;24(7):716-722.

53.

Lyketsos C, Carrillo M, Ryan J, et al. Neuropsychiatric symptoms in Alzheimer’s disease.
Alzheimers Dement. 2011;7(5):532-539.
doi:10.1016/j.jalz.2011.05.2410.Neuropsychiatric.

54.

Balthazar M, Pereira F. Neuropsychiatric symptoms in Alzheimer’s disease are related to
functional connectivity alterations in the salience network. Hum Brain Mapp. 2013.

55.

Bae JN, MacFall JR, Krishnan KRR, Payne ME, Steffens DC, Taylor WD. Dorsolateral
prefrontal cortex and anterior cingulate cortex white matter alterations in late-life
depression. Biol Psychiatry. 2006;60(12):1356-1363. doi:10.1016/j.biopsych.2006.03.052.

56.

Tighe SK, Oishi K, Mori S, et al. Diffusion Tensor Imaging of Neuropsychiatric
Symptoms in Mild Cognitive Impairment and Alzheimer’s Dementia. 2012;24(4):484488. doi:10.1176/appi.neuropsych.11120375.Diffusion.

57.

Hettema JJM, Kettenmann B, Ahluwalia V, et al. Pilot multimodal twin imaging study of
generalized anxiety disorder. Depress Anxietynd anxiety. 2012;29(3):202-209.
doi:10.1002/da.20901.

58.

Norris CJ, Chen EE, Zhu DC, Small SL, Cacioppo JT. The interaction of social and
emotional processes in the brain. J Cogn Neurosci. 2004;16(10):1818-1829.
doi:10.1162/0898929042947847.

59.

Greicius M. Resting-state functional connectivity in neuropsychiatric disorders. Curr Opin
Neurol. 2008;24(4):424-430. doi:10.1097/WCO.0b013e328306f2c5.

60.

Aalten P, Verhey F, Boziki M, et al. Neuropsychiatric syndromes in dementia. Dement
Geriatr Cogn Disord. 2007;24:457-463.

61.

Ogden CCL, Carroll M, Fryar C, Flegal K. Prevalence of Obesity Among Adults and
Youth: United States, 2011-2014. Med Benefits. 2015;33(1):5.

62.

Must A, Spadano J, Coakley EH, Field A, Colditz G, Dietz W. The Disease Burden
Associated With Overweight and Obesity. JAMA. 1999;282(16):1523-1529.
doi:10.1001/jama.282.16.1523.

63.

Han T, Tajar A, Lean M. Obesity and weight management in the elderly. Br Med Bull.
2011;97:169-196.

64.

Sharma A, Padwal R. Obesity is a sign - over-eating is a symptom: an aetiological
106

framework for the assessment and management of obesity. Obes Rev. 2010;11(5):362370. doi:10.1111/j.1467-789X.2009.00689.x.
65.

Papachristou E, Ramsay SE, Lennon LT, et al. The relationships between body
composition characteristics and cognitive functioning in a population-based sample of
older British men. BMC Geriatr. 2015;15:172. doi:10.1186/s12877-015-0169-y.

66.

Gustafson D. Adiposity indices and dementia. Lancet Neurol. 2006;5(8):713-720.
doi:10.1016/S1474-4422(06)70526-9.

67.

Beydoun M, Beydoun H, Wang Y. Obesity and central obesity as risk factors for incident
dementia and its subtypes: a systematic review and meta-analysis. Obes Rev.
2008;9(3):204-218. doi:10.1111/j.1467-789X.2008.00473.x.

68.

Whitmer R, Gustafson DR, Barrett-Connor E, Haan MN, Gunderson EP, Yaffe K. Central
obesity and increased risk of dementia more than three decades later. Neurology.
2008;71(14):1057-1064. doi:10.1212/01.wnl.0000306313.89165.ef.

69.

Hildreth KL, Van Pelt RE, Schwartz RS. Obesity, insulin resistance, and Alzheimer’s
disease. Obesity. 2012;20:1549-1557. doi:10.1038/oby.2012.19.

70.

Gregor MF, Hotamisligil GS. Inflammatory mechanisms in obesity. Annu Rev Immunol.
2011;29:415-445. doi:10.1146/annurev-immunol-031210-101322.

71.

Flier JS. Obesity Wars: Molecular Progress Confronts an Expanding Epidemic. Cell.
2004;116:337-350. doi:10.1016/S0092-8674(03)01081-X.

72.

Walther K, Birdsill A, Glisky E, Ryan L. Structural brain differences and cognitive
functioning related to body mass index in older females. Hum Brain Mapp.
2010;31(7):1052-1064. doi:10.1002/hbm.20916.

73.

Zeki Al Hazzouri A, Stone KL, Haan MN, Yaffe K. Leptin, mild cognitive impairment,
and dementia among elderly women. J Gerontol A Biol Sci Med Sci. 2013;68(2):175-180.
doi:10.1093/gerona/gls155.

74.

Castanon N, Lasselin J, Capuron L. Neuropsychiatric comorbidity in obesity: role of
inflammatory processes. Front Endocrinol (Lausanne). 2014;5(74):1-9.
doi:10.3389/fendo.2014.00074.

75.

Hassenstab J, Sweet L, Parigi A, et al. Cortical thickness of the cognitive control network
in obesity and successful weight loss maintenance: A preliminary MRI study. Psychiatry
Res Neuroimaging. 2012;202(1):77-79.

76.

Stoeckel L, Kim J, Weller R, Cox J, Cook W, Horwitz B. Effective connectivity of a
reward network in obese women. Brain Res Bull. 2009;79(6):388-395.

77.

Stice E, Spoor S, Bohon C, Veldhuizen M, Small D. Relation of Reward from Food Intake
107

and Anticipated Food Intake to Obesity: A Functional Magnetic Resonance Imaging
Study. J Abnorm Psychol. 2008;117(4):924-935. doi:10.1037/a0013600.Relation.
78.

Lee N, Carter A, Owen N, Hall W. The neurobiology of overeating. EMBO Rep.
2012;13:785-790.

79.

Pannacciulli N, Parigi A Del, Chen K, Le D, Reiman E, Tataranni P. Brain abnormalities
in human obesity: a voxel-based morphometric study. Neuroimage. 2006;31:1419-1425.

80.

Raji CA, Ho AJ, Parikshak N, et al. Brain Structure and Obesity. Hum Brain Mapp.
2011;31(3):353-364. doi:10.1002/hbm.20870.Brain.

81.

Driscoll I, Beydoun M a, An Y, et al. Midlife obesity and trajectories of brain volume
changes in older adults. Hum Brain Mapp. 2012;33(9):2204-2210.
doi:10.1002/hbm.21353.

82.

Ho A, Raji C, Becker J, et al. Obesity is linked with lower brain volume in 700 AD and
MCI patients. Neurobiol Aging. 2010;31(8):1326-1339.
doi:10.1016/j.neurobiolaging.2010.04.006.

83.

Stanek KKM, Grieve SSM, Brickman AMA, et al. Obesity is associated with reduced
white matter integrity in otherwise healthy adults. Obesity. 2011;19:500-504.
doi:10.1038/oby.2010.312.

84.

Marks B, Katz L, Styner M, Smith J. Aerobic fitness and obesity: relationship to cerebral
white matter integrity in the brain of active and sedentary older adults. Br J Sports Med.
2011;45:1208-1215.

85.

Karlsson HHK, Tuulari JJJ, Hirvonen J, et al. Obesity is associated with white matter
atrophy: A combined diffusion tensor imaging and voxel‐based morphometric study.
Obesity. 2013;21(12):2530-2537. doi:10.1002/oby.20386.

86.

Pelchat M, Johnson A, Chan R, Valdez J, Ragland J. Images of desire: food-craving
activation during fMRI. Neuroimage. 2004;23:1486-1493.

87.

Simon GE, Von Korff M, Saunders K, et al. Association between obesity and psychiatric
disorders in the US adult population. Arch Gen Psychiatry. 2006;63(7):824-830.
doi:10.1001/archpsyc.63.7.824.

88.

Allison DB, Newcomer JW, Dunn AL, et al. Obesity among those with mental disorders:
a National Institute of Mental Health meeting report. Am J Prev Med. 2009;36(4):341350. doi:10.1016/j.amepre.2008.11.020.

89.

Barefoot J, Heitmann B, Helms M, Williams R, Surwit R, Siegler I. Symptoms of
depression and changes in body weight from adolescence to mid-life. Int J Obes.
1998;22:688-694.

108

90.

Boutelle KKN, Hannan P, Fulkerson JA, Crow SJ, Stice E. Obesity as a prospective
predictor of depression in adolescent females. Heal Psychol. 2010;29(3):293-298.
doi:10.1037/a0018645.

91.

Roberts R, Deleger S, Strawbridge W, Kaplan G. Prospective association between obesity
and depression: evidence from the Alameda County Study. Int J Obes. 2003;27:514-521.

92.

Levitan RD, Davis C. Emotions and Eating Behaviour: Implications for the Current
Obesity Epidemic. Univ Toronto Q. 2010;79(2):783-799. doi:10.3138/utq.79.2.783.

93.

Vago DR, Epstein J, Catenaccio E, Stern E. Identification of Neural Targets for the
Treatment of Psychiatric Disorders: The Role of Functional Neuroimaging. Vol 22.; 2011.
doi:10.1016/j.nec.2011.01.003.

94.

Tekin S, Cummings JL. Frontal-subcortical neuronal circuits and clinical neuropsychiatry:
an update. J Psychosom Res. 2002;53(2):647-654.

95.

Soczynska JK, Kennedy SH, Woldeyohannes HO, et al. Mood disorders and obesity:
understanding inflammation as a pathophysiological nexus. Neuromolecular Med.
2011;13(2):93-116. doi:10.1007/s12017-010-8140-8.

96.

Misiak B, Leszek J, Kiejna A. Metabolic syndrome, mild cognitive impairment and
Alzheimer’s disease--the emerging role of systemic low-grade inflammation and
adiposity. Brain Res Bull. 2012;89(3-4):144-149. doi:10.1016/j.brainresbull.2012.08.003.

97.

Petersen R, Thomas R, Grundman M, et al. Vitamin E and donepezil for the treatment of
mild cognitive impairment. N Engl J Med. 2005;352:1-10.

98.

Rogers S, Doody R. Donepezil improves cognition and global function in Alzheimer
disease: a 15-week, double-blind, placebo-controlled study. Arch Intern Med.
1998;158(9):1021-1031.

99.

Ismail Z, Smith EE, Geda Y, et al. Neuropsychiatric symptoms as early manifestations of
emergent dementia: Provisional diagnostic criteria for mild behavioral impairment.
Alzheimer’s Dement. 2016;12(2):195-202. doi:10.1016/j.jalz.2015.05.017.

100. Lyketsos C, Lopez O, Jones B, Fitzpatrick A, Breitner J, DeKosky S. Prevalence of
neuropsychiatric symptoms in dementia and mild cognitive impairment: results from the
cardiovascular health study. JAMA. 2002;288(12):1475-1483.
101. Zhao Q-F, Tan LL, Wang H-F, et al. The prevalence of neuropsychiatric symptoms in
Alzheimer’s disease: Systematic review and meta-analysis. J Affect Disord.
2016;190:264-271. doi:10.1016/j.jad.2015.09.069.
102. Rosenberg PB, Mielke MM, Appleby BS, Oh ES, Geda YE, Lyketsos CG. The
association of neuropsychiatric symptoms in MCI with incident dementia and Alzheimer
disease. Am J Geriatr Psychiatry. 2012;21(7):685-695. doi:10.1016/j.jagp.2013.01.006.
109

103. Gorska-Ciebiada M, Saryusz-Wolska M, Ciebiada M, Loba J. Mild cognitive impairment
and depressive symptoms in elderly patients with diabetes: prevalence, risk factors, and
comorbidity. J Diabetes Res. 2014;2014:179648. doi:10.1155/2014/179648.
104. Kandimalla R, Thirumala V, Reddy PH. Is Alzheimer’s disease a Type 3 Diabetes? A
critical appraisal. Biochim Biophys Acta. 2016;1863(5):1078-1089.
doi:10.1016/j.bbadis.2016.08.018.
105. Flegal KM, Carroll MD, Ogden CL, Curtin LR, Johnson CL. Prevalence and trends in
obesity among US adults, 1999-2000. JAMA. 2010;303(3):1723-1727.
106. Moms J, Heyman A, Mohs R, et al. The Consortium to Establish a Registry for
Alzheimer’s Disease (CERAD). Part I. Clinical and neuropsychological assesment of
Alzheimer’s disease. Neurology. 1989;39(9):1159.
107. Li J, Wang Y, Zhang M, et al. Vascular risk factors promote conversion from mild
cognitive impairment to Alzheimer disease. Neurology. 2011;76(17):1485-1491.
108. Steinberg M, Hess K, Corcoran C, et al. Vascular risk factors and neuropsychiatric
symptoms in Alzheimer’s disease: the Cache County Study. Int J Geriatr Psychiatry.
2014;29(2):153-159. doi:10.1002/gps.3980.
109. Kaufer DI, Cummings JL, Ketchel P, et al. Validation of the NPI-Q, a brief clinical form
of the Neuropsychiatric Inventory. J Neuropsychiatry Clin Neurosci. 2000;12(2):233-239.
110. Herrman N, Mittmann N, Silver IL, et al. A validation study of the Geriatric Depression
Scale short form. Int J Geriatr Psychiatry. 1996;11:457-460.
111. Aalten P, de Vugt ME, Jaspers N, Jolles J, Verhey FRJ. The course of neuropsychiatric
symptoms in dementia. Part II: relationships among behavioural sub-syndromes and the
influence of clinical variables. Int J Geriatr Psychiatry. 2005;20(6):531-536.
doi:10.1002/gps.1317.
112. Teng E, Chui H. The Modified Mini-Mental State (3MS) examination. J Clin Psychiatry.
1987;48:314-318.
113. Atkinson R, Shiffrin R. The control of short-term memory. Sci Am. 1971;225:82-90.
114. Kaplan E, Goodglass H, Weintraub S. The Boston naming test, 2nd. Philadelphia Lea
Febiger. 1983.
115. Borkowski J, Benton A, Spreen O. Word fluency and brain damage. Neuropsychologia.
1967;5(2):135-140.
116. Tombaugh T. Trail Making Test A and B: Normative data stratified by age and education.
Arch Clin Neuropsychol. 2004;10(2):203-214.
110

117. Stroop J. Studies of interference in serial verbal reactions. J Exp Psychol. 1935;18:643662.
118. Rosen W, Mohs R, Davis K. A new rating scale for Alzheimer’s disease. Am J Psychiatry.
1984;141:1356-1364.
119. Ogden CL, Carroll MD, Kit BK, Flegal KM. Prevalence of Childhood and Adult Obesity
in the United States, 2011-2012. JAMA. 2014;311(8):806. doi:10.1001/jama.2014.732.
120. Luppino F, Wit L de, Bouvy P. Overweight, obesity, and depression: a systematic review
and meta-analysis of longitudinal studies. Arch Gen Psychiatry. 2010;67(3):220-229.
121. Katon W, Lyles C, Parker MM, Karter AJ, Huang ES, Whitmer RA. Association of
depression with increased risk of dementia in patients with type 2 diabetes: the Diabetes
and Aging Study. Arch Gen Psychiatry. 2012;69(4):410-417.
122. Akmal MK, Ezat M, Raafat O, Hamed H, Bediwy A. Association of depression, anxiety,
and impairment in executive functions in patients with obstructive sleep apnea. Middle
East Curr Psychiatry. 2013;20(1):30-34. doi:10.1097/01.XME.0000422808.09000.59.
123. Vernet C, Redolfi S, Attali V, et al. Residual sleepiness in obstructive sleep apnoea:
Phenotype and related symptoms. Eur Respir J. 2011;38(1):98-105.
doi:10.1183/09031936.00040410.
124. Sharma S, Fernandes M, Fulton S. Adaptations in brain reward circuitry underlie palatable
food cravings and anxiety induced by high-fat diet withdrawal. Int J Obes. 2013;37:11831191.
125. Gustafson D. A life course of adiposity and dementia. Eur J Pharmacol. 2008;585(1):163175. doi:10.1016/j.ejphar.2008.01.052.
126. Rosenberg PB, Mielke MM, Appleby B, et al. Neuropsychiatric symptoms in MCI
subtypes: the importance of executive dysfunction. Int J Geriatr Psychiatry.
2012;26(4):364-372. doi:10.1002/gps.2535.Neuropsychiatric.
127. Petersen RC, Aisen PS, Beckett LA, et al. Alzheimer’s disease Neuroimaging Initiative
(ADNI) clinical characterization. Neurology. 2010;74(3):201-209.
doi:10.1212/WNL.0b013e3181cb3e25.
128. Flegal K, Kruszon-Moran D, Carroll M, Fryar C, Ogden C. Trends in obesity among
adults in the United States, 2005 to 2014. JAMA. 2016;315(21):2284-2291.
129. Bischof GN, Park DC. Obesity and Aging : Consequences for Cognition, Brain Structure,
and Brain Function. Psychosom Med. 2015;77:697-709.
doi:10.1097/PSY.0000000000000212.

111

130. Yokum S, Ng J, Stice E. Relation of regional gray and white matter volumes to current
BMI and future increases in BMI: a prospective MRI study. Int J Obes. 2011;36(10):656664. doi:10.1038/ijo.2011.175.
131. Albanese E, Davis B, Jonsson P V, et al. Overweight and Obesity in Midlife and Brain
Structure and Dementia 26 Years Later: The AGES-Reykjavik Study. Am J Epidemiol.
2015;181(9):672-679. doi:10.1093/aje/kwu331.
132. Smith E, Hay P, Campbell L, Trollor JN. A review of the association between obesity and
cognitive function across the lifespan : implications for novel approaches to prevention
and treatment. Obes Rev. 2011;12:740-755. doi:10.1111/j.1467-789X.2011.00920.x.
133. Elias MF, Goodell AL, Waldstein SR. Obesity, cognitive functioning and dementia: back
to the future. J Alzheimer’s Dis. 2012;30:S113-S125. doi:10.3233/JAD-2011-111175.
134. Whitmer R, Gunderson E, Barrett-Connor E, Quesenberry C, Yaffe K. Obesity in middle
age and future risk of dementia: a 27 year longitudinal population based study. BMJ.
2005;330:1360.
135. Banack H, Kaufman J. The “Obesity Paradox” Explained. Epidemiology. 2013;24(3):461462. doi:10.1097/EDE.0b013e31828c776c.
136. Fielding RA, Gunstad J, Gustafson DR, et al. The paradox of overnutrition in aging and
cognition. Ann N Y Acad Sci. 2013;1287(1):31-43. doi:10.1111/nyas.12138.
137. Marqués-Iturria I, Pueyo R, Garolera M, et al. Frontal cortical thinning and subcortical
volume reductions in early adulthood obesity. Psychiatry Res Neuroimaging.
2013;214(2):109-115. doi:10.1016/j.pscychresns.2013.06.004.
138. Mueller SG, Weiner MW, Thal LJ, et al. Ways toward an early diagnosis in Alzheimer’s
disease: The Alzheimer’s Disease Neuroimaging Initiative (ADNI). Alzheimer’s Dement.
2005;1(1):55-66.
139. Jack CCR, Bernstein MA, Fox NC, et al. The Alzheimer’s disease neuroimaging initiative
(ADNI): MRI methods. J Magn Reson Imaging. 2008;27(4):685-691.
doi:10.1002/jmri.21049.
140. Weiner M, Veitch D, Aisen P, et al. The Alzhiemer’s Disease Neuroimaging Initiative: A
review of papers published since its inception. Alzheimer’s Dement. 2012;8(10):S1-68.
doi:10.1016/j.jalz.2011.09.172.The.
141. Dale A, Fischl B, Sereno M. Cortical surface-based analysis: I. Segmentation and surface
reconstruction. Neuroimage. 1999;9:179-194.
142. Dale A, Sereno M. Improved localization of cortical activity by combining EEG and MEG
with MRI cortical surface reconstruction: a linear approach. J Cogn Neurosci. 1993;5:162176.
112

143. Fischl B, Dale A. Measuring the thickness of the human cerebral cortex from magnetic
resonance images. Proc Natl Acad Sci U S A. 2000;97:11050-11055.
144. Fischl B, Liu A, Dale A. Automated manifold surgery: constructing geometrically
accurate and topologically correct models of the human cerebral cortex. IEEE Trans Med
Imaging. 2001;20:70-80.
145. Fischl B, Salat D, Busa E, Albert M, Dieterich M. Whole brain segmentation: automated
labeling of neuroanatomical structures in the human brain. Neuron. 2002;33:341-355.
146. Fischl B, Salat D, Kouwe A van der, Makris N. Sequence-independent segmentation of
magnetic resonance images. Neuroimage. 2004;23(Suppl 1):S69-84.
147. Fischl B, Sereno M, Dale A. Cortical surface-based analysis: II: inflation, flattening, and a
surface-based coordinate system. Neuroimage. 1999;9:195-207.
148. Fischl B, Sereno M, Tootell R, Dale A. High-resolution intersubject averaging and a
coordinate system for the cortical surface. Hum Brain Mapp. 1999;8:272-284.
149. Fischl B, van der Kouwe A, Destrieux C. Automatically parcellating the human cerebral
cortex. Cereb Cortex. 2004;14:11-22.
150. Han X, Jovicich J, Salat D, et al. Reliability of MRI-derived measurements of human
cerebral cortical thickness: the effects of field strength, scanner upgrade and manufacturer.
Neuroimage. 2006;32:180-194.
151. Jovicich J, Czanner S, Greve D, et al. Reliability in multi-site structural MRI studies:
effects of gradient non-linearity correction on phantom and human data. Neuroimage.
2006;30:436-443.
152. Ségonne F, Dale A, Busa E, et al. A hybrid approach to the skull stripping problem in
MRI. Neuroimage. 2004;22:1060-1075.
153. Reuter M, Rosas H, Fischl B. Highly accurate inverse consistent registration: a robust
approach. Neuroimage. 2010;53:1181-1196.
154. Reuter M, Schmansky N, Rosas H, Fischl B. Within-subject template estimation for
unbiased longitudinal image analysis. Neuroimage. 2012;61(4):1402-1418.
155. Sled J, Zijdenbos A, Evans A. A nonparametric method for automatic correction of
intensity nonuniformity in MRI data. IEEE Trans Med Imaging. 1998;17:87-97.
156. Ségonne F, Pacheco J, Fischl B. Geometrically accurate topology-correction of cortical
surfaces using nonseparating loops. IEEE Trans Med Imaging. 2007;26:518-529.
157. Desikan R, Ségonne F, Fischl B, Quinn B. An automated labeling system for subdividing
113

the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage.
2006;31:968-980.
158. Rosas H, Liu A, Hersch S, et al. Regional and progressive thinning of the cortical ribbon
in Huntington’s disease. Neurology. 2002;58:695-701.
159. Kuperberg G, Broome M, McGuire P, et al. Regionally localized thinning of the cerebral
cortex in schizophrenia. Arch Gen Psychiatry. 2003;60:878-888.
160. Salat D, Buckner R, Snyder A, Greve D. Thinning of the cerebral cortex in aging. Cereb
Cortex. 2004;14:721-730.
161. Kharabian Masouleh S, Arélin K, Horstmann A, et al. Higher body mass index in older
adults is associated with lower gray matter volume: Implications for memory
performance. Neurobiol Aging. 2016;40:1-10. doi:10.1016/j.neurobiolaging.2015.12.020.
162. Boyle CP, Raji CA, Erickson KI, et al. Physical activity, body mass index, and brain
atrophy in Alzheimer’s disease. Neurobiol Aging. 2015;36(1):S194-202.
doi:10.1016/j.neurobiolaging.2014.05.036.
163. Miller AA, Spencer SJ. Obesity and neuroinflammation: A pathway to cognitive
impairment. Brain Behav Immun. 2014;42:10-21. doi:10.1016/j.bbi.2014.04.001.
164. Xu W, Caracciolo B, Wang H-X, et al. Accelerated Progression From Mild Cognitive
Impairment to Dementia in People With Diabetes. Diabetes. 2010;59(11):2928-2935.
doi:10.2337/db10-0539.
165. Lupien SJ, de Leon M, de Santi S, et al. Cortisol levels during human aging predict
hippocampal atrophy and memory deficits. Nat Neurosci. 1998;1(1):69-73.
doi:10.1038/271.
166. Catani M, de Schotten MT. Atlas of Human Brain Connections. 1st ed. Oxford University
Press; 2012.
167. Fitzpatrick A, Kuller L, Lopez O, et al. Midlife and late-life obesity and the risk of
dementia: cardiovascular health study. Arch Neurol. 2009;66(3):336-342.
doi:10.1001/archneurol.2008.582.Mid-.
168. Doruk H, Naharci MI, Bozoglu E, Isik a. T, Kilic S. The relationship between body mass
index and incidental mild cognitive impairment, Alzheimer’s disease, and Vascular
Dementia in elderly. J Nutr Health Aging. 2010;14(10):834-838. doi:10.1007/s12603-0100113-y.
169. Sobów T, Fendler W, Magierski R. Body mass index and mild cognitive impairment-todementia progression in 24 months: a prospective study. Eur J Clin Nutr.
2014;68(11):1216-1219. doi:10.1038/ejcn.2014.167.

114

170. Tolppanen AM, Ngandu T, Kåreholt I, et al. Midlife and late-life body mass index and
late-life dementia: Results from a prospective population-based cohort. J Alzheimer’s Dis.
2014;38(1):201-209. doi:10.3233/JAD-130698.
171. Pourhoseingholi MA, Baghestani AR, Vahedi M. How to control confounding effects by
statistical analysis. Gastroenterol Hepatol Bed Bench. 2012;5(2):79-83.
172. Savva G, Wharton S, Ince P, Forster G, Matthews F, Brayne C. Age, Neuropathology, and
Dementia. N Engl J Med. 2009;360(22):2302-2309.
173. Hryhorczuk C, Sharma S, Fulton S. Metabolic disturbances connecting obesity and
depression. Front Neurosci. 2013;7:177.
174. David N, Lin F, Porsteinsson A. Trajectories of neuropsychiatric symptoms and cognitive
decline in mild cognitive impairment. Am J Geriatr Psychiatry. 2015;24(1):70-80.
doi:10.1016/j.jagp.2015.06.001.
175. Hsu CL, Voss MW, Best J, et al. Elevated body mass index and maintenance of cognitive
function in late life: Exploring underlying neural mechanisms. Front Aging Neurosci.
2015;7(JUL):1-14. doi:10.3389/fnagi.2015.00155.
176. Cronk B, Johnson D, Burns J. Body Mass Index and Cognitive Decline in Mild Cognitive
Impairment. Alzheimer Dis Assoc Disord. 2010;24(2):126-130.
doi:10.1097/WAD.0b013e3181a6bf3f.Body.
177. Cova I, Clerici F, Maggiore L, et al. Body Mass Index Predicts Progression of Mild
Cognitive Impairment to Dementia. Dement Geriatr Cogn Disord. 2016;41(3-4):172-180.
doi:10.1159/000444216.
178. Alhurani RE, Vassilaki M, Aakre JA, et al. Decline in Weight and Incident Mild
Cognitive Impairment. JAMA Neurol. 2016;73(4):439.
doi:10.1001/jamaneurol.2015.4756.
179. Cova I, Clerici F, Rossi A, et al. Weight loss predicts progression of mild cognitive
impairment to Alzheimer’s disease. PLoS One. 2016;11(3):e0151710.
doi:10.1371/journal.pone.0151710.
180. Hughes CP, Berg L, Danziger WL, Coben LA, Martin RL. A new clinical scale for the
staging of dementia. Br J Psychiatry. 1982;140(6):566-572. doi:10.1192/BJP.140.6.566.
181. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat
Assoc. 1958;53(282):457-481. doi:10.2307/2281868.
182. Ye BS, Jang EY, Kim SY, et al. Unstable Body Mass Index and Progression to Probable
Alzheimer’s Disease Dementia in Patients with Amnestic Mild Cognitive Impairment. J
Alzheimer’s Dis. 2015;49(2):1-9. doi:10.3233/JAD-150556.

115

183. Pietrzak RH, Maruff P, Woodward M, et al. Mild Worry Symptoms Predict Decline in
Learning and Memory in Healthy Older Adults: A 2-Year Prospective Cohort Study. Am J
Geriatr Psychiatry. 2012;20(3):266-275. doi:10.1097/JGP.0b013e3182107e24.
184. Nir TM, Jahanshad N, Villalon-Reina JE, et al. Effectiveness of regional DTI measures in
distinguishing Alzheimer’s disease, MCI, and normal aging. NeuroImage Clin.
2013;3:180-195. doi:10.1016/j.nicl.2013.07.006.
185. Charlton R, Schiavone F, Barrick T, Morris R, Markus H. Diffusion tensor imaging
detects age related white matter change over a 2 year follow-up which is associated with
working memory decline. J Neurol Neurosurg Psychiatry. 2010;81(1):13-19.
doi:10.1136/jnnp.2008.167288.
186. Kullmann S, Callaghan MF, Heni M, et al. Specific white matter tissue microstructure
changes associated with obesity. Neuroimage. 2016;125:36-44.
doi:10.1016/j.neuroimage.2015.10.006.
187. Kullmann S, Schweizer F, Veit R, Fritsche a, Preissl H. Compromised white matter
integrity in obesity. Obes Rev. 2015;16(4):273-281. doi:10.1111/obr.12248.
188. Smith S, Jenkinson M, Woolrich M, et al. Advances in functional and structural MR
image analysis and implementation as FSL. Neuroimage. 2004;23:S208-S219.
189. Ni H, Kavcic V, Zhu T, Ekholm S, Zhong J. Effects of number of diffusion gradient
directions on derived diffusion tensor imaging indices in human brain. Am J Neuroradiol.
2006;27(8):1776-1781. doi:27/8/1776 [pii].
190. Smith SM, Jenkinson M, Johansen-Berg H, et al. Tract-based spatial statistics: Voxelwise
analysis of multi-subject diffusion data. Neuroimage. 2006;31(4):1487-1505.
doi:10.1016/j.neuroimage.2006.02.024.
191. Rueckert D, Sonoda LI, Hayes C, Hill DLG, Leach MO, Hawkes DJ. Nonrigid
registration using free-form deformations: application to breast MR images. IEEE Trans
Med Imaging. 1999;18(8):712-721. doi:10.1109/42.796284.
192. Mori S, Oishi K, Jiang H, et al. Stereotaxic white matter atlas based on diffusion tensor
imaging in an ICBM template. Neuroimage. 2008;40(2):570-582.
doi:10.1016/j.neuroimage.2007.12.035.
193. Zhou Y, Dougherty J, Hubner K, Bai B, Cannon R, Hutson R. Abnormal connectivity in
the posterior cingulate and hippocampus in early Alzheimer’s disease and mild cognitive
impairment. Alzheimer’s Dement. 2008;4(4):265-270.
194. Papageorgiou I, Astrakas LG, Xydis V, et al. Abnormalities of brain neural circuits related
to obesity : A Diffusion Tensor Imaging study. Magn Reson Imaging. 2017;37:116-121.
doi:10.1016/j.mri.2016.11.018.

116

195. Moran C, Phan TG, Chen J, et al. Brain atrophy in type 2 diabetes: regional distribution
and influence on cognition. Diabetes Care. 2013;36(12):4036-4042. doi:10.2337/dc130143.
196. van Bloemendaal L, Ijzerman RG, ten Kulve JS, et al. Alterations in white matter volume
and integrity in obesity and type 2 diabetes. Metab Brain Dis. 2016;31(3):621-629.
doi:10.1007/s11011-016-9792-3.
197. Storsve AB, Fjell AM, Tamnes CK, et al. Differential Longitudinal Changes in Cortical
Thickness, Surface Area and Volume across the Adult Life Span: Regions of Accelerating
and Decelerating Change. J Neurosci. 2014;34(25):8488-8498.
doi:10.1523/JNEUROSCI.0391-14.2014.
198. Brundel M, Kappelle LJ, Biessels GJ. Brain imaging in type 2 diabetes. Eur
Neuropsychopharmacol. 2014;24(12):1967-1981. doi:10.1016/j.euroneuro.2014.01.023.
199. Querbes O, Aubry F, Pariente J, et al. Early diagnosis of Alzheimer’s disease using
cortical thickness: impact of cognitive reserve. Brain. 2009;132(8):2036-2047.
doi:10.1093/brain/awp105.
200. Veit R, Kullmann S, Heni M, et al. Reduced cortical thickness associated with visceral fat
and BMI. NeuroImage Clin. 2014;6:307-311. doi:10.1016/j.nicl.2014.09.013.
201. Sweet L, Hassenstab J, McCaffery J, et al. Brain response to food stimulation in obese,
normal weight, and successful weight loss maintainers. Obesity. 2012;20(11):2220-2225.

117