SUPPLEMENTAL MATERIAL ONLINE: Lorius et al. Vascular disease and risk factors
are associated with cognitive decline in the Alzheimer’s disease spectrum
METHODS
Subjects
The Alzheimer’s Disease Neuroimaging Initiative (ADNI) is a large multi-center,
longitudinal, observational trial taking place across North America, in which normal older
control (NC), amnestic mild cognitive impairment (MCI), and mild Alzheimer’s disease AD
dementia subjects are followed with periodic neuropsychological testing, multiple imaging
techniques, and fluid biomarkers. The goals of ADNI are to standardize brain imaging across
multiple sites, obtain a large longitudinal dataset for future research, and develop reliable
biomarker surrogates for treatment trials. ADNI is the result of efforts of many co-investigators
from a broad range of academic institutions and private corporations, and subjects have been
recruited from over 50 sites across the U.S. and Canada1.
At baseline, NC subjects had Mini-Mental State Exam (MMSE)2 scores between 25-30
(inclusive), a Clinical Dementia Rating (CDR)3 global score of 0, and no significant memory
impairment (subjects performed within 1.5 standard deviations of education adjusted cut-off
scores on the delayed recall portion of one Logical Memory story (LM-IIa) of the Wechsler
Memory Scale-Revised (WMS-R)4. Amnestic MCI subjects had MMSE scores between 24-30
(inclusive), a memory complaint, objective memory loss on the WMS-R LM-IIa, a CDR global
score of 0.5 and memory box score ≥ 0.5, essentially preserved activities of daily living, and were
not demented. Mild AD dementia subjects had MMSE scores between 20-26 (inclusive), CDR global
score of 0.5 or 1.0, and met the National Institute of Neurologic and Communicative Disorders
1
and Stroke and the AD and Related Disorders Association Work Group (NINCDS/ADRDA)
criteria for probable AD5 .
Clinical assessments
There are some differences between the vascular index score we used in the current study
and the Framingham Study Stroke Risk profile6. The Framingham profile includes a measure of
systolic blood pressure rather a history of hypertension; it includes left ventricular hypertrophy,
which our index does not include; it includes age and use of antihypertensive medications, which
we included separately as covariates in our analyses as described below; on the other hand our
index includes hyperlipidemia and stroke, which the Framingham profile does not include.
Missing values data for key predictors and dependent variables were less than 4% with
the exception of white matter hyperintensity (WMH) volume. 811 of the 812 subjects had Digit
Symbol scores; 812 had Rey Auditory Verbal Learning Test (RAVLT) Total Learning scores;
804 had Alzheimer Disease Assessment Scale Cognitive Subscale (ADAS-Cog) scores; 812 had
Vascular index scores; 810 had systolic blood pressure values; 784 had serum glucose values;
787 had serum total cholesterol values; 810 had Body Mass Index (BMI); 812 had
Apolipoprotein E ε4 (APOE4) carrier status; 632 had WMH volume values; and 808 had plasma
homocysteine level values.
Statistical Analyses
Cross-sectional analyses:
The inclusion of the interaction of vascular disease and risk factors with diagnosis
allowed us to test for any differential relation of any given vascular variable to a given cognitive
measure across diagnostic groups. Should the interaction be significant, a test of the same
2
relation could be followed up separately within each group, with these multiple p values
“protected” by the required significant preliminary omnibus interaction. On the other hand, if the
interaction was not significant, a homogeneous within group relation of the vascular variable to
the dependent variable, pooling strength from the data within each of the diagnostic groups,
could be performed.
Longitudinal analyses: Cox Proportional Hazards Models
The overwhelming majority of subjects who progressed from MCI to dementia had AD
dementia. The few subjects who progressed to non-AD dementia were included with the subjects
who progressed to AD dementia in the current analyses.
A p<0.01 cut-off was employed in the backward elimination algorithm for the crosssectional analyses rather than p<0.05 in order to protect against multiple test chance effects
owing to the sequence of significance tests performed in the backward elimination. A more
liberal cut-off of p<0.05 was used in the longitudinal backward elimination algorithms because
there were fewer number of terms in the initial predictor pool for the longitudinal analyses, as
well as our desire to be sensitive to important effects that may be missed in these less powerful
analyses.
RESULTS
As noted in Table 1, there was a greater proportion of males among MCI subjects
compared to NC and AD dementia subjects. AD dementia subjects had less years of education
compared to NC and MCI subjects. The American National Adult Reading Test (AMNART IQ),
CDR sum of boxes, MMSE, ADAS-Cog, Digit Symbol, RAVLT Total Learning, and APOE4
3
carrier status were different among all three diagnostic groups in the expected directions. BMI
was greater in NC subjects compared to AD dementia subjects. WMH volume was greater in AD
dementia subjects compared to MCI subjects. Homocysteine level was lower in NC subjects
compared to MCI and AD dementia subjects.
Longitudinal Analyses: Cox Proportional Hazards Models
MCI to AD dementia progression: Post hoc Tukey tests indicated that having one or two
APOE4 alleles significantly (p<0.01) increased the hazard rate over having none, whereas there
was no significant difference when comparing having one vs. two APOE4 alleles.
ACKNOWLEDGEMENTS
Data collection and sharing for this project was funded by the Alzheimer's Disease
Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904). ADNI is
funded by the National Institute on Aging, the National Institute of Biomedical Imaging and
Bioengineering, and through generous contributions from the following: Abbott; Alzheimer’s
Association; Alzheimer’s Drug Discovery Foundation; Amorfix Life Sciences Ltd.;
AstraZeneca; Bayer HealthCare; BioClinica, Inc.; Biogen Idec Inc.; Bristol-Myers Squibb
Company; Eisai Inc.; Elan Pharmaceuticals Inc.; Eli Lilly and Company; F. Hoffmann-La Roche
Ltd and its affiliated company Genentech, Inc.; GE Healthcare; Innogenetics, N.V.; IXICO Ltd.;
Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson
Pharmaceutical Research & Development LLC.; Medpace, Inc.; Merck & Co., Inc.; Meso Scale
Diagnostics, LLC.; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Servier; Synarc Inc.; and
Takeda Pharmaceutical Company. The Canadian Institutes of Health Research is providing funds
4
to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the
Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the
Northern California Institute for Research and Education, and the study is coordinated by the
Alzheimer's Disease Cooperative Study at the University of California, San Diego. ADNI
data are disseminated by the Laboratory for Neuro Imaging at the University of California, Los
Angeles. This research was also supported by NIH grants P30 AG010129 and K01 AG030514.
5
REFERENCES
1. Weiner MW, Veitch DP, Aisen PS et al. The Alzheimer's Disease Neuroimaging Initiative: a
review of papers published since its inception. Alzheimers Dement. 2012;8:S1-68.
2. Folstein MF, Folstein SE, McHugh PR. "Mini-mental state". A practical method for grading
the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189-198.
3. Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules.
Neurology. 1993;43:2412-2414.
4. Wechsler D. WMS-R Wechsler Memory Scale Revised Manual. New York: The Psychological
Corporation, Harcourt Brace Jovanovich, Inc, 1987.
5. McKhann G, Drachman D, Folstein M et al. Clinical diagnosis of Alzheimer's disease: report
of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human
Services Task Force on Alzheimer's Disease. Neurology. 1984;34:939-944.
6. D'Agostino RB, Wolf PA, Belanger AJ et al. Stroke risk profile: adjustment for
antihypertensive medication. The Framingham Study. Stroke. 1994;25:40-43.
6
Figure e-1. Predicted values from fixed effects of best fitting longitudinal model of RAVLT
Total Learning score across time in the study by Serum Total Cholesterol at selected baselines by
diagnostic group. APOE4 alleles was set to equal presence of 1 allele. NC (Top), MCI (Middle),
and AD dementia (Bottom). AD (Alzheimer’s disease), APOE4 (Apolipoprotein E 4), MCI
(mild cognitive impairment), NC (normal older control), RAVLT (Rey Auditory Verbal
Learning Test), SD (standard deviation).
7
Figure e-1. Lorius et al.
Serum Total Cholesterol:
-----: mean
___:
1 SD < mean
RAVLT Total Learning score
….. : 1 SD > mean
8
Table e-1. Cross-sectional general linear model of association of vascular disease and risk
factors and Digit Symbol score, displaying predictors retained in the final model.
Model without WMH volume: R2=0.29, p<0.0001
Model with WMH volume: R2=0.32, p<0.0001
β
Predictor
95% CI for β
p
% Variance Accounted for
Total
Partial
Intercept
23.14
14.66, 31.62
<0.0001
Homocysteine
-0.33
-0.59, -0.06
0.02
0.4
0.6
WMH volume
-0.71
-1.00, -0.42
<0.0001
2.4
3.4
Diagnosis*
AD
27.2
0.83
<0.0001
20.8
22.7
MCI
37.01
0.57
NC
45.04
0.75
0.19
0.11, 0.25
<0.0001
2.3
3.1
level
AMNART IQ
AD (Alzheimer’s disease), AMNART IQ (American National Adult Reading Test intelligence
quotient), β (partial unstandardized regression coefficient estimate), CI (confidence interval),
MCI (mild cognitive impairment), NC (normal older control), WMH (white matter
hyperintensity).
“%Variance Total” represents percent of total variance of Digit Symbol uniquely associated with
the indicated predictor (unbiased population estimate). “%Variance Partial” represents percent of
variance of Digit Symbol with portion associated with other predictors pre-removed, which is
uniquely associated with the indicated predictor (unbiased population estimate).
9
* For Diagnosis and sex, covariate adjusted means and standard errors were reported instead of β
and 95% CI for β, respectively.
10
Table e-2. Cross-sectional general linear model of association of vascular disease and risk
factors and RAVLT Total Learning, displaying predictors retained in the final model.
Model: R2=0.45, p<0.0001
β
Predictor
95% CI for β
p
% Variance Accounted for
Total
Partial
Intercept
35.89 26.88, 44.91
<0.0001
Vascular Index score
-0.86
0.007
0.4
0.8
Diagnosis*
AD
23.81 0.63
<0.0001
33.7
37.8
MCI
31.21 0.44
NC
42.59 0.57
<0.0001
1.4
2.4
Sex*
-1.48, -0.24
Female 33.95 0.47
Male
31.12 0.42
Age
-0.13
-0.22, -0.04
0.004
0.5
0.9
AMNART IQ
0.14
0.09, 0.19
<0.0001
1.8
3.2
AD (Alzheimer’s disease), AMNART IQ (American National Adult Reading Test intelligence
quotient), β (partial unstandardized regression coefficient estimate), CI (confidence interval),
MCI (mild cognitive impairment), NC (normal older control), RAVLT (Rey Auditory Verbal
Learning Test).
“%Variance Total” represents percent of total variance of RAVLT Total Learning uniquely
associated with the indicated predictor (unbiased population estimate). “%Variance Partial”
represents percent of variance of RAVLT Total Learning with portion associated with other
11
predictors pre-removed, which is uniquely associated with the indicated predictor (unbiased
population estimate).
* For Diagnosis and sex, covariate adjusted means and standard errors were reported instead of β
and 95% CI for β, respectively.
12
Table e-3. Cross-sectional general linear model of association of vascular disease and risk
factors and ADAS-Cog score, displaying predictors retained in the final model.
Model: R2=0.59, p<0.0001
β
Predictor
95% CI for β
p
% Variance Accounted for
Total
Partial
Intercept
19.96
15.43, 24.49
<0.0001
APOE4
0.89
0.25, 1.54
0.007
0.3
0.8
AD
26.75
0.67
<0.0001
22.0
34.8
MCI
18.96
0.33
NC
10.62
0.43
0.67
0.32, 1.02
0.0002
0.7
1.6
-0.09
-0.13, -0.05
<0.0001
1.1
2.6
Diagnosis*
AD dementia
symptom duration
(years)
AMNART IQ
AD (Alzheimer’s disease), ADAS-Cog (Alzheimer Disease Assessment Scale Cognitive
Subscale), AMNART IQ (American National Adult Reading Test intelligence quotient), APOE4
(Apolipoprotein E 4), β (partial unstandardized regression coefficient estimate), CI (confidence
interval), MCI (mild cognitive impairment), NC (normal older control).
“%Variance Total” represents percent of total variance of ADAS-Cog uniquely associated with
the indicated predictor (unbiased population estimate). “%Variance Partial” represents percent of
variance of ADAS-Cog with portion associated with other predictors pre-removed, which is
uniquely associated with the indicated predictor (unbiased population estimate).
13
* For Diagnosis, covariate adjusted means and standard errors were reported instead of β and
95% CI for β, respectively.
14
Table e-4. Longitudinal mixed effects model of association of baseline vascular disease and risk
factors and Digit Symbol score over time, displaying predictors retained in the final model.
Model: R2=0.76 for fixed effects, p<0.0001; R2=0.92 including random terms, p<0.0001
Predictor
β
95% CI for β
p
Intercept
5.02
2.83, 7.20
<0.0001
Time
0.13
-0.52, 0.77
<0.0001
APOE4
-0.003
-0.80, 0.80
0.99
APOE4 x Time
-0.84
-1.35, -0.34
0.001
WMH volume
0.10
-0.08, 0.28
0.26
WMH volume x Time
-0.10
-0.21, 0.004
0.06
AD
-2.24
-4.00, -0.48
0.03
MCI
-0.45
-1.72, 0.82
NC
0
AD
-3.65
-4.74, -2.56
MCI
-1.58
-2.31, -0.86
NC
0
Baseline Diagnosis
Baseline Diagnosis x Time
0.91
Baseline Digit Symbol
0.88, 0.95
<0.0001
<0.0001
AD (Alzheimer’s disease), APOE4 (Apolipoprotein E 4), β (partial regression coefficient
estimate), CI (confidence interval), MCI (mild cognitive impairment), NC (normal older
control), WMH (white matter hyperintensity). x indicates an interaction.
15
Table e-5. Longitudinal mixed effects model of association of baseline vascular disease and risk
factors and RAVLT Total Learning score over time, displaying predictors retained in the final
model.
Model: R2=0.69 for fixed effects, p<0.0001; R2=0.90 including random terms, p<0.0001
Predictor
β
95% CI for β
p
Intercept
7.53
4.52, 10.54
<0.0001
Time
0.61
-0.72, 1.94
0.55
Serum Total Cholesterol
0.01
-0.0003, 0.02
0.06
Serum Total Cholesterol x Time
-0.006
-0.01, 0.0002
0.06
APOE4
-0.72
-1.37, -0.08
0.03
RAVLT Total Learning
0.78
0.73, 0.83
<0.0001
AD
-5.30
-7.06, -3.54
<0.0001
MCI
-3.35
-4.64, -2.06
NC
0
AD
-2.21
-3.10, -1.31
MCI
-0.82
-1.42, -0.22
NC
0
Baseline Diagnosis
Baseline Diagnosis x Time
<0.0001
AD (Alzheimer’s disease), APOE4 (Apolipoprotein E 4), β (partial unstandardized regression
coefficient estimate), CI (confidence interval), MCI (mild cognitive impairment), NC (normal
older control), RAVLT (Rey Auditory Verbal Learning Test). x indicates an interaction.
16
Table e-6. Longitudinal mixed effects model of association of baseline vascular disease and risk
factors and ADAS-Cog score over time, displaying predictors retained in the final model.
Model: R2=0.75 for fixed effects, p<0.0001; R2=0.95 including random terms, p<0.0001
Predictor
β
95% CI for β
P
Intercept
-3.40
-8.12, 1.31
0.16
Time
-2.92
-4.23, -1.60
0.02
APOE4
-0.22
-0.81, 0.37
0.47
APOE4 x Time
0.65
0.24, 1.06
0.002
Serum Total Cholesterol
0.006
-0.002, 0.01
0.14
Serum Total Cholesterol x Time
0.007
0.001, 0.01
0.02
Baseline ADAS-Cog
0.79
0.72, 0.85
<0.0001
Baseline ADAS-Cog x Time
0.15
0.10, 0.20
<0.0001
AD
3.62
1.95, 5.30
0.0001
MCI
1.55
0.50, 2.60
NC
0
AD
2.68
1.47, 3.89
MCI
0.84
0.13, 1.55
NC
0
Baseline Diagnosis
Baseline Diagnosis x Time
<0.0001
Baseline WMH volume
0.12
-0.004, 0.25
0.06
Serum Glucose
-0.01
-0.03, -0.00002
0.05
Baseline Age
0.07
0.01, 0.12
0.02
17
AD (Alzheimer’s disease), ADAS-Cog (Alzheimer Disease Assessment Scale Cognitive
Subscale), APOE4 (Apolipoprotein E 4), β (partial regression coefficient estimate), CI
(confidence interval), MCI (mild cognitive impairment), NC (normal older control), WMH
(white matter hyperintensity). x indicates an interaction.
18