JAMDA xxx (2017) 1.e1e1.

e13

JAMDA
journal homepage: www.jamda.com

Original Study

Memory Alterations and White Matter Hyperintensities in Elderly

Patients With Hypertension: The ADELAHYDE-2 Study
Joo Pedro Ferreira MD, PhD a, b, Anna Kearney Schwartz MD a, c,
Ghassan Watfa MD, PhD c, Lamiral Zohra MSc a, Jacques Felblinger MD, PhD a, d,
Jean-Marc Boivin MD, PhD a, Serge Bracard MD, PhD a, e, Gabriella Hossu MD, PhD a, f, g,
Antoine Verger MD h, Laure Joly MD, PhD c, f, Faiez Zannad MD, PhD a, f,
Patrick Rossignol MD, PhD a, f, Athanase Benetos MD, PhD a, c, f, *
a
INSERM, Centre dInvestigations Cliniques Plurithmatique 1433, INSERM U1116, Universit de Lorraine, CHRU de Nancy, F-CRIN INI-CRCT, Nancy,
France
b
Department of Physiology and Cardiothoracic Surgery, Cardiovascular Research and Development Unit, Faculty of Medicine, University of Porto, Porto,
Portugal
c
Service de Griatrie, CHRU de Nancy, Nancy, France
d
INSERM, U947, IADI, Nancy, F-54000, France, Dpartement de Neuroradiologie Diagnostique et Interventionnelle, Ple Imagerie, Centre Hospitalier
Universitaire de Nancy, Nancy, France
e
Dpartement de Neuroradiologie Diagnostique et Interventionnelle, Centre Hospitalier Universitaire de Nancy, Nancy, France
f
INSERM U1116, Universit de Lorraine, Nancy, France
g
U947, INSERM, Nancy, France
h
CHRU Nancy, Nuclear Medicine & Nancyclotep Experimental Imaging Platform, Nancy, France

a b s t r a c t

Keywords: Objectives: The longitudinal ADELAHYDE-2 study aims to identify the factors associated with cognitive
Memory alterations impairment/decline and white matter hyperintensities burden.
white matter hyperintensities Methods: Longitudinal single-center study comprising two visits separated by approximately 7 years. A
predictors and associations
total of 131 patients completed the two visits. The primary outcome was global memory composite scale,
risk factors
while the secondary outcome was white matter hyperintensities (WMH/Fazekas scale) load.
Results: Global memory at visit 2 (V2) was largely inuenced by age, smoking status, glycated hemo-
globin, and history of stroke already present at visit 1 (V1). These variables accounted for w51% of the
memory alterations at V2. WMH at V2 was likely inuenced by age, left ventricular hypertrophy, diabetes
mellitus, carotid intima-media thickness, and body mass index at V1. These ndings accounted for w37%
of the WMH changes at V2. Increase in pulse wave velocity from V1 to V2 showed a trend for association
with memory deterioration (adjusted estimates 0.06; P .067), whereas smoking and increase in
systolic blood pressure (trend) were associated with an increment in WMH (adjusted estimates 0.49;
P .047 and adjusted estimates 0.01; P .08, respectively). On the other hand, angiotensin-converting
enzyme inhibitor/angiotensin receptor blockers and statins (trend) were likely to be protective (adjusted
estimates for angiotensin-converting enzyme inhibitor/angiotensin receptor blockers 0.49; P .049,
and adjusted estimates for statins 0.46; P .055).
Conclusions: Several readily identiable factors are associated with memory deterioration and WMH,
many of which are potentially modiable. Interventions aimed to control these risk factors need to be
tested prospectively in order to assess their cognitive protective value.
2017 AMDA e The Society for Post-Acute and Long-Term Care Medicine.

This study was funded by grants from the Programme Hospitalier de Recherche
Clinique (PHRC national), INSERM, the Socit Francaise dHypertension artrielle,
the Bayer Research Foundation, and the Association de Recherche et dInformation
en Cardiologie (ARISC). The study was sponsored by Nancy CHRU.
The authors declare no conicts of interest.
* Address correspondence to Athanase Benetos, MD, PhD, Service de Griatrie,
Hpital Brabois, CHRU de Nancy, Vandoeuvre-ls-Nancy 54511, France.
E-mail address: a.benetos@chru-nancy.fr (A. Benetos).
Dementia has a high prevalence in developed countries,
affecting approximately 5% of individuals older than 65 years of age
and represents a substantial health care, social, and economic
burden.1 Determining the factors associated with cognitive
impairment and the ability to predict the evolution of cognitive
functions over time is thus crucial in gaining a better understanding

http://dx.doi.org/10.1016/j.jamda.2017.01.008
1525-8610/ 2017 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
1.e2 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13

of the characteristics of the population at risk of dementia as well as

ADELAHYDE 1
to determine the factors that may be modiable through external
intervention.
Alzheimer disease and vascular dementia are responsible for more
than 80% of dementia cases, with both conditions sharing common N=376
Withdrawal: 247
risk factors such as hypertension, smoking, and diabetes.2 The advent Death: 51
of detailed imaging methods and cognitive assessment has enabled Loss to follow-up: 60
assessing early dementia stages.3e5 In particular, impaired cognitive Refusal: 130
Other 6
performance has been associated with cardiovascular (CV) risk factors
and CV disease.6,7 White matter hyperintensities (WMH) assessed by
magnetic resonance imaging (MRI) T2-weighted images (and easily ADELAHYDE 2
reported as visual rating scales, such as the Fazekas scale),8 are N = 131
prevalent in the general population, affecting more than 25% of in-
dividuals over the age of 65, and have also been more commonly
described in patients with CV risk factors and CV disease.9 The WMH
Visit 1 Visit 2
burden can thus noninvasively quantify the degree of small vessel Psychometry: 131
Psychometry: 131
lesions (leukoaraiosis) and has been proposed as a risk marker for MRI: 126 MRI: 114
cognitive impairment and dementia of vascular origin to be used as an
intermediate endpoint in clinical trials.4,10,11
In the Vascular Alteration and Evolution of Cognitive Impairment
(ADELAHYDE)-1 study (results of the baseline visit 1 [V1]), we found
that after adjustment for the usual cardiovascular risk factors,
increased arterial stiffness (as assessed by pulse wave velocity [PWV])
was signicantly and independently associated with memory Both visits 1 and 2
impairment in men.12 The severity of leukoaraiosis as measured with Psychometry: 131
MRI: 113
WMH was independently associated with increased carotid intima
media thickness (cIMT) and stiffness (as assessed by augmentation
index),12 suggesting that arterial stiffness and thickness could repre-
Fig. 1. Study owchart. MRI, magnetic resonance imaging.
sent additional targets for dementia prevention.
The aims of the ADELAHYDE-2 were to study the potential role of
The study was approved by the local research ethics committee,
clinical, biological, and arterial parameters assessed during the base-
and all patients signed an informed consent form.
line visit on memory impairment (primary outcome) and leukoar-
Clinical evaluation included patient history and a general and
aiosis (secondary outcome) over a 7.7-year follow-up period in this
neurological examination. Education level was classied into two
population of older hypertensive patients with subjective memory
categories: low (ie, primary education) and intermediate/high (ie,
complaints.
intermediate/higher-level general education and university). Supine
SBP and DBP were measured after a minimum 10-minute rest period,
Material and Methods
on the left arm, using a validated electronic device (DINAMAP 400Pro).
Blood pressure (BP) was measured three times (both at V1 and V2),
The study description is available at ClinicalTrials.gov
and the average of these three measurements was used for statistical
(NCT01351961). The ADELAHYDE-2 study was a longitudinal, single-
analyses. Cognitive status was established by a set of psychometric
center study, comprising two visits separated by approximately
tests. First, the Cognitive Difculties Scale of McNair was performed by
7 years. Patients were recruited for V1 between September 2003 and
self-evaluation (inclusion criteria),14 after which the Geriatric
August 2005 by local press advertisements or referral by general
Depression Scale, a self-report scale, was completed.15 Global cogni-
practitioners from an investigator network of the Inserm Clinical
tive assessment was based on the Mini-Mental State Examination.16 A
Investigation Centre.12 Patients initially included in V1 were contacted
validated and comprehensive battery of neuropsychological tests was
by telephone to participate in visit 2 (V2) (conducted from April 2011
performed by trained psychologists, including immediate and delayed
to November 2013). Of the 376 patients included in V1, 247 did not
memory and language (free and cued recall) assessed by the Grober-
participate in V2 (51 patients died, 60 were lost to follow-up, 130
Buschke test17; visuoperceptual and visuospatial capacities by the
refused to participate, and 6 patients provided no specic reason).
Benton Visual Retention Test18; praxies by the Praxies scale,19 and
Hence, 131 patients participated in the ADELAHYDE-2 study, with all
executive function and long-term verbal memory by the Verbal
131 patients having full psychometric evaluation on both visits, and
Fluency Test.20 Composite scores were calculated for memory function
113 having an MRI on both visits (Figure 1).
(real memory score, including two components from the Grober-
Inclusion criteria at V1 were as follows: men and women aged 60
Buschke test: free [episodic memory] and cued [semantic memory]
to 85 years with a history of hypertension (lasting more than 1 year)
recalls), verbal uency (which describes language lexical uidity and
and treated at the time of enrollment for at least 1 month with at least
includes four tests), and visual memory capacity (visual memory
one antihypertensive agent and presenting with subjective memory
score, dening raw visual memory and errors with two tests). Each
complaints (SMCs) dened as a McNair scale >15 (revised validated
composite score was computed by transforming individual test scores
French version of 26 items).13 Exclusion criteria were known sec-
into standardized z-scores (z-score test score-mean test score/
ondary hypertension, severe hypertension (dened as systolic blood
standard deviation [SD]).6
pressure [SBP] > 200 mm Hg and/or diastolic blood pressure
[DBP] > 120 mm Hg) or malignant hypertension, severe hypotension
(dened as SBP < 90 mm Hg), serum creatinine > 25 mg/L, trans- Study Outcomes
aminase levels over twice the upper normal laboratory value, body
mass index < 40 kg/m2, history of heart failure, cancer. or systemic The primary prespecied outcome was the assessment of the
diseases, known dementia, and history of alcohol and/or drug abuse. global memory composite scale at V2.
 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13
The secondary prespecied outcome was the assessment of WMH
(Fazekas scale) at V2.
Magnetic Resonance Imaging
Brain MRI at 1.5 T (Signal General Electric; GEMS) was performed
in the axial and sagittal planes with a slice thickness of 5 mm.21 WMH
grading was conducted using T2-weighted axial images. The grading
was performed in a blinded manner by one experienced neuroradi-
ologist using the scale of Fazekas and Schmidt, which accounts
separately for periventricular corps and deep white matter corps
(subcortical). The Fazekas scale represents the sum of deep white
matter corps and periventricular corps divided into groups of
increasing severity (ie, 0 to 1, 2 to 3, 4 to 6).22 Among rating scales, the
Fazekas scale has the advantage of presenting the highest interrater
agreements and correlation with volumetric measurements.23,24
Blood and Urine Samples
Hemoglobin, sodium, potassium, total serum cholesterol, high-
density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL)
cholesterol, homocysteine, C-reactive protein, and serum creatinine
(with a glomerular ltration rate estimation by the Modication of
Diet in Renal Disease-4 [MDRD4] formula)25 were measured after an
overnight fast using a standardized venous blood sampling technique.
A 24-hour urine collection was performed to measure
microalbuminuria.
Subclinical Cardiovascular Disease
Ultrasonography of the right common carotid artery was performed
to measure carotid intima media thickness using a high-resolution B-
mode system with a 7.5-MHz linear array transducer (ATL Apogee
800).26 The arterial wall segments were assessed in a longitudinal view
with both walls clearly visible in order to achieve diameter measure-
ments. cIMT measurements were performed on the far wall along a
minimum 10-mm length of an arterial segment from automatic high-
quality image acquisitions using IOTEC software (IODP). Adventitia-
to-adventitia diameter and intraluminal diameter of the common
carotid artery were also measured. In our laboratory, intra- and inter-
observer reproducibility was 3.6  4% and 6.8  4.5%, respectively.
Common, internal, and/or external carotid plaque was considered
as a focal structure encroaching into the arterial lumen by at least
0.5 mm or when demonstrating a thickness of 1.5 mm as measured
from the mediaeadventitia interface to the intimaelumen interface.
Pulse wave velocity was measured noninvasively in the car-
otidefemoral segment using the automatic Complior device (Colson,
Paris, France). Information on the validation of this measurement
method and its reproducibility have been previously published.27
Augmentation index was assessed using high-delity applanation
tonometry (SphygmoCor Pulse Wave Analysis; PWV Medical Pty Ltd,
Ermington, Sidney, Australia) as previously described.28
Left ventricular hypertrophy (LVH) was dened by the presence of
Sokolow-Lyon (>38 mm) and/or Cornell product (>2.44 mm*ms)
indices on electrocardiography.29
Statistical Analysis
Continuous variables are presented as mean  SD if normally
distributed or as median (percentile 25th to 75th) if the distribution
was skewed. Categorical variables are presented as absolute numbers
and proportions in percentages.
Multivariate linear regression models were performed to assess the
associations between the dependent variables, namely memory score
(primary outcome) and WMH/Fazekas scale (secondary outcome), and
1.e3
the independent variables available in the dataset at V1 and V2. Linear
regression assumptions were veried. The independent variables
available at V1 were age, gender, education level, smoking status, body
mass index (BMI), SBP (both ofce measurements and 24-hour
ambulatory blood pressure measurement), heart rate, LVH, estimated
glomerular ltration rate (eGFR), microalbuminuria, LDL cholesterol
level, homocysteine, hypertension duration, diabetes status and gly-
cated hemoglobin (HbA1c), history of cardiovascular events, medica-
tions, cIMT, PWV, and augmentation index. At V2, sodium, potassium,
hemoglobin, and C-reactive protein levels were also included in the
models. Of note, DBP, pulse pressure, HDL and total cholesterol were
removed from all models due to the presence of collinearity. A back-
ward selection process was performed, and only the variables retained
in the nal model are shown herein. In order to verify the tness of the
models, a forward selection process with a sequential and progressive
manual elimination of factors with a less signicant P value (>.2) in
univariate testing was also performed with overlapping results. No
interaction was found between factors in the nal models.
In the rst instance, the predictors of memory and WMH/MRI
Fazekas at V2 were assessed using V1 variables. Second, the associa-
tions between memory and WMH/MRI Fazekas and the variables
available at V1 and V2 were assessed. The correlation between global
memory composite V1/V2 was 0.57; P < .001, and the correlation
between the WMH/MRI Fazekas scale V1/V2 was 0.84; P < .001, that
is, these parameters presented moderate to high correlation. Hence,
two models were constructed, model 1 without the V1 variable and a
model 2 with the V1 variable, thereby allowing the assessment of
model variance with and without the knowledge of memory testing
and WMH/MRI Fazekas at V1.
Logistic regression models were also performed on an exploratory
basis to assess the factors associated with worsening memory and
WMH/MRI Fazekas based on whether the latter had worsened or not
from V1 to V2. Multinomial regression models were also explored to
assess the factors associated with higher WMH/MRI Fazekas in the
various categories (0-1, 2-3, 4-6).
P < .05 was considered as statistically signicant, and P < .1 was
considered as trend toward statistical signicance.
All analyses were performed using version 22 of the SPSS software
(IBM Corp. Released 2013. IBM SPSS Statistics for Windows, Version
22.0. Armonk, NY: IBM Corp).
Results
Characteristics of the Study Population
At V2, the 131 patients included in both visits were 7.7  1.3 years
older than at V1, with a mean age of 75.5  5.6 years and 51.9% of
whom were women. BMI, pulse pressure, eGFR, microalbuminuria,
cIMT and PWV increased from V1 to V2 (1.0  2.2 kg/m2, 4.0
 15.0 mm Hg, 5.2  13.9 mL/min/1.73 m2, 3.6 (0.15 to 13.4), 0.09
 0.37 mm, and 1.8  3.7 m/s, respectively; P < .01 for all compari-
sons). The proportion of patients taking statins also increased from V1
to V2 (32.8% to 53.4%; P < .001), while total cholesterol decreased
(0.17  0.37 g/L). The endpoints of global memory composite score
and WMH assessed by the MRI Fazekas scale increased (ie, worsened)
from V1 to V2 (1.1  1.1 and 1.2  0.9, respectively; P < .001 for both
parameters) (Table 1).
Predictors of Global Memory and WMH at V2
Male gender, older age, smoker (or having smoked), presence of
LVH on ECG, higher glycated hemoglobin levels, and history of stroke
were positively associated with worse global memory (adjusted
estimates 0.58, 1.33, 0.69, 0.61, 0.25, and 1.71, respectively;
1.e4 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13

Table 1
Characteristics of the Study Population Who Completed Both V1 and V2 Visits of the ADELAHYDE Study

N 131 V1 V2 D V2eV1 P Value

Sociodemographics
Female gender, n (%) 68 (51.9) - - -
Age (years) 67.7  5.3 75.5  5.6 7.7  1.3 <.001
Intermediate/high education level, n (%) 97 (74.0) - - -
Risk factors
Past/current smoker, n (%) 50 (38.2) - - -
BMI (kg/m2) 27.8  4.1 28.8  4.7 1.0  2.2 <.001
SBP (mm Hg) 136.7  16.1 139.4  17.8 2.7  20.9 .147
DBP (mm Hg) 73.6  9.4 72.1  9.5 1.4  10.1 .112
Pulse pressure (mm Hg) 63.2  13.1 67.2  13.6 4.0  15.0 .003
Heart rate (bpm) 62.3  8.7 63.7  9.3 1.4  10.9 .137
LV hypertrophy on ECG, n (%) 17 (13.0) 21 (16.0) - .239
Laboratory
Hemoglobin (g/dL) - 12.0  1.5 - -
MDRD eGFR (mL/min/1.73 m2) 64.6  11.2 69.9  16.5 5.2  13.9 <.001
Microalbuminuria (mg/L)* 4.8 (3.0e17.1) 9.4 (3.8e77.1) 3.6 (0.15 to 13.4) <.001
Sodium (mmol/L) - 140.2  2.3 - -
Potassium (mmol/L) - 4.4  0.4 - -
Total cholesterol (g/L) 2.19  0.33 2.01  0.41 0.17  0.37 <.001
LDL cholesterol (g/L) 1.32  0.29 1.19  0.34 0.13  0.36 <.001
HDL cholesterol (g/L) 0.62  0.19 0.57  0.15 0.05  0.13 <.001
Homocysteine (mmol/L) 12.4  3.7 15.3  4.7 2.9  4.5 <.001
C-reactive protein (mg/L)* - 1.9 (1.0e8.6) - -
Medical history
Hypertension, n (%) 131 (100) 131 (100) - -
Hypertension duration (years)* 12 (6e17) 21 (15e26) 8 (7e9) <.001
Diabetes, n (%) 16 (12.2) 22 (16.8) - .083
HbA1c (%) 5.9  1.0 6.0  1.0 0.1  1.1 .123
History of vascular events
Cardiovascular 9 (6.9) 21 (16.0) - .005
Stroke 3 (2.3) 5 (3.8) - .317
Transient ischemic attack 4 (3.1) 7 (5.3) .366
Medications
ACEi/ARBs, n (%) 82 (62.6) 93 (71.0) - .132
b-blockers, n (%) 45 (34.4) 45 (34.4) - -
CCB, n (%) 42 (32.1) 58 (38.2) - .206
Diuretics, n (%) 60 (45.8) 70 (53.4) - .174
Statins, n (%) 43 (32.8) 70 (53.4) - <.001
Antithrombotic agents, n (%) 36 (27.3) 31 (23.7) - .384
Paraclinical Investigations and Psychometric Assessment
Grber-Buschke test:
Free recall (1) 29.8  5.9 25.4  8.7 4.4  6.6 <.001
Cued recall (2) 15.6  4.5 18.7  6.1 3.1  5.0 <.001
Total recall (1 2) 45.5  3.5 44.1  5.8 1.4  4.5 <.001
Composite score (global memory) 0.0  1.0 1.1  1.3 1.1  1.1 <.001
Morphology of the right carotid artery
Intima-media thickness (mm) 0.65  0.09 0.74  0.37 0.09  0.37 .008
Atheromatous plaque, n (%) 28 (21.4) 34 (26.0) - .237
Pulse wave velocity (m/s) 11.3  2.3 13.1  3.9 1.8  3.7 <.001
Augmentation index (%) 38.3  10.1 40.3  20.5 1.9  19.9 .386
Mean MRI Fazekas score 2.1  1.4 3.3  1.6 1.2  0.9 <.001
Fazekas scale in categories, n (%)
grade I (0-1) 43 (32.8) 10 (7.6) - <.001
grade II (2-3) 59 (45.0) 60 (45.8) -
grade III (4-6) 24 (18.3) 44 (33.6) -

ACEi/ARBs, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers; BMI, body mass index; CCB, calcium channel blockers; DBP, diastolic blood pressure;
ECG, electrocardiogram; eGFR, estimated glomerular ltration rate; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; LV, left
ventricular; MDRD, modied of diet in renal disease equation 4; MRI, magnetic resonance imaging; SBP, systolic blood pressure; V1, visit 1; V2, visit 2.
For continuous variables, results are presented as mean  standard deviation while *non-parametric test results are presented as median (percentile 25th to 75th); categorical
variables are presented as numbers (n) and proportions (%).

constant 10.68; P < .05 for all comparisons, except LVH where P for
trend 0.085) (Table 2).
Older age, presence of LVH on ECG, diabetes, higher cIMT and BMI
(trend) were positively associated with an increase in WMH, whereas
taking ACEi/ARBs was inversely associated with WMH (adjusted
estimates 0.69, 1.85, 1.06, 6.83, 0.08, 1.20, respectively;
constant 9.26; P < .05 for all comparisons, except BMI where P for
trend .063) (Table 2 and Figure 2).
Adding V1 global memory and WMH to the models (respectively,
model 2 in Table 2) improved the overall tness of the models (from
51% to 70% in global memory and from 37% to 72% in WMH) due to
strong correlations between V1 and V2 outcome variables (R 0.58
for global memory and R 0.85 for WMH). The addition of correlated
outcome variables to the models led to the exclusion of all biomarkers
of target organ damage.
Determinants of changes (D) in Global Memory and WMH Between
V1 and V2
Changes in global memory from V1 to V2 did not appear to be
signicantly inuenced by the variables in the model. Only w4% of the
variability in memory changes was explained by the model, and only
 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13 1.e5

Table 2
V1 Predictors of MRI Fazekas Score and Global Memory at V2

Adjusted R2 Standardized b Adjusted Estimate (95% CI) P Value

Global memory composite (V2)

Model 1: Excluding V1 Global memory composite
Overall Model t 0.51 - - <.001
Constant - - 10.68 (14.15 to 7.23) <.001
Gender (Male) - 0.22 0.58 (0.06e1.10) .030
Age (per 10-year increase) - 0.52 1.33 (0.88e1.78) <.001
Smoking (yes) - 0.26 0.69 (0.20e1.19) .007
LVH (yes) - 0.15 0.61 (0.09 to 1.30) .085
HbA1c (per 1% increase) - 0.19 0.25 (0.02e0.48) .034
Stroke (yes) - 0.26 1.71 (0.53e2.89) .005
Model 2: Including V1 Global memory composite
Overall Model t 0.70 - - <.001
Constant - - 7.10 (10.94 to 3.26) <.001
Global memory composite V1 - 0.58 0.73 (0.53e0.94) <.001
Gender (male) - 0.18 0.46 (0.00e0.92) .049
Age (per 10-year increase) - 0.37 0.96 (0.57e1.34) <.001
Stroke (yes) - 0.16 1.05 (0.07e2.04) .037
MRI Fazekas scale (V2)
Model 1: Excluding V1 MRI Fazekas scale
Overall model t 0.37 - - <.001
Constant - - 9.26 (14.72 to 3.78 .001
Age (per 10-year increase) - 0.23 0.69 (0.08e1.31) .028
LVH (yes) - 0.38 1.85 (0.83e2.85) .001
Diabetes (yes) - 0.23 1.06 (0.93e2.02) .032
cIMT (per 1 mm increase) - 0.35 6.83 (2.84e10.82) .001
ACEi/ARBs (yes) - 0.36 1.20 (1.89 to 0.51) .001
2
BMI (by 1 kg/m increase) - 0.20 0.08 (0.04 to 0.16) .063
Model 2: Including V1 MRI Fazekas scale
Overall model t 0.72 - - <.001
Constant - - 0.48 (1.06 to 2.01) .536
MRI Fazekas score V1 - 0.85 0.98 (0.81e1.14) <.001

ACEi/ARBs, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers; BMI, body mass index; cIMT, carotid intima-media thickness (right carotid artery); DBP,
diastolic blood pressure; ECG, electrocardiogram; eGFR, estimated glomerular ltration rate; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density
lipoprotein; LV, left ventricular; LVH, left ventricular hypertrophy; SBP, systolic blood pressure; V1, visit 1; V2, visit 2.
A stepwise backward conditional method was used in all models.
The predictor variables introduced in the model were age, gender, education level, smoking status, BMI, SBP, heart rate, LV hypertrophy on ECG, eGFR, microalbuminuria, LDL
cholesterol, homocysteine, hypertension duration, diabetes status and HbA1c, history of cardiovascular events, medications, intima-media thickness, pulse-wave velocity, and
augmentation indexdall at V1.
DBP, pulse pressure, HDL, and total cholesterol were removed from all models due to the presence of collinearity.
The correlation between global memory composite at V1 and V2 was 0.57; P < .001, and the correlation between MRI Fazekas scale at V1 and V2 was 0.84; P < .001 (ie,
these parameters have moderate to high correlation).
Both V1 ofce blood pressure and 24-hour ambulatory blood pressure monitoring were tested without difference in the nal models.

DPWV showed a trend for association with changes in global memory
performance (adjusted estimate 0.06, P .067; constant 0.86)
(Table 3). The associations for global memory at both V1 and V2 are
represented in Supplemental Table S1. The associations for the indi-
vidual components of memory assessment are represented in
Supplemental Table S2. In this regard, DPWV was associated with a
worsening in cued recall (adjusted estimate 0.36; P .01;
constant 1.94).
Changes (ie, worsening) in WMH from V1 to V2 were partially
explained by smoking status (w19%, P .007). A similar trend was
observed between worsening in WMH and DSBP (increase), as well as
the absence of prescription of ACEi/ARBs and of statins (adjusted
estimates 0.49, 0.01, 0.49 and 0.46; constant 1.73; P < .05 for
smoking and ACEi/ARBs, P .080 for DSBP and P .055 for statin use,
respectively) (Table 3). The associations for WMH at both V1 and V2
are represented in Supplemental Table S1.
Logistic regression comparisons for global memory and changes in
leukoaraiosis from V1 to V2 (improvement vs worsening) is shown in
Supplemental Table S3. The multinomial regression for MRI Fazekas
scale categories (0-1, 2-3, 4-6) is shown in Supplemental Table S4. The
odds of having a higher-grade (4) Fazekas scale at V2 showed a trend
toward an increase according to age (odds ratio [95% condence in-
terval, CI] for a 10-year age increase 4.58 [0.98 to 21.36], P .053).
Comparison of the patients who completed the two study visits
versus those who were lost to follow-up and those who died is shown
in Supplemental Table S5. Patients who were lost to follow-up were
older and had a lower education level.

Fig. 2. Illustrative representation of visit 1 predictors of white matter lesions/MRI
Fazekas score and global memory at visit 2. BMI, body mass index; cIMT, carotid
intima-media thickness; DM/HbA1c, diabetes mellitus/glycated hemoglobin; LVH, left
ventricular hypertrophy; memory, composite memory score; WMH, white matter
hyperintensities.
Correlations Between WMH and Global Memory
Pearson correlations between WMH and global memory were
generally weak (0.17 at V1, P .061 and 0.24 at V2; P .012).
1.e6 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13

Table 3
Associations Between D (V1eV2) MRI Fazekas Score and Global Memory Assessed by Multiple Linear Regression Analysis

Adjusted R2 Standardized b Adjusted Estimate (95% CI) P Value

D Global memory (V1 to V2)

Overall model t 0.04 - - .067
Constant - - 0.86 (0.60e1.11) <.001
D PWV (per 1 m/s increase) - 0.23 0.06 (0.01 to 0.12) .067
D MRI Fazekas score (V1 to V2)
Overall model t 0.19 - - .007
Constant - - 1.73 (1.23e2.24) <.001
Smoking history (yes) - 0.26 0.49 (0.01e0.97) .047
D SBP (mm Hg) - 0.23 0.01 (0.00 to 0.23) .080
ACEi/ARBs (yes) - 0.26 0.49 (0.93 to 0.01) .049
Statins (yes) - 0.25 0.46 (0.94 to 0.01) .055

ACEi/ARBs, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers; BMI, body mass index; cIMT, carotid intima-media thickness (right carotid artery); DBP,
diastolic blood pressure; ECG, electrocardiogram; eGFR, estimated glomerular ltration rate; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density
lipoprotein; LV, left ventricular; LVH, left ventricular hypertrophy; PWV, pulse-wave velocity; SBP, systolic blood pressure; V1, visit 1; V2, visit 2.
A stepwise backward conditional method was used for all models.
In the D MRI Fazekas score and D global memory model from V1 to V2, the variables introduced in the initial model were D age, gender, education level, smoking status, D BMI,
D SBP, D heart rate, D eGFR, D microalbuminuria, D LDL cholesterol, D homocysteine, D hypertension duration, D HbA1c, history of cardiovascular events history, medications,
D intima-media thickness, D pulse-wave velocity, and D augmentation index.
DBP, pulse pressure, HDL, and total cholesterol were removed from all models due to the presence of collinearity.
The correlation between global memory composite at V1 and V2 was 0.57; P < .001, and the correlation between MRI Fazekas scale at V1 and V2 was 0.84; P < .001 (ie,
these parameters have moderate to high correlation).

Discussion vascular thickness (assessed by cIMT) was positively associated with

To the best of our knowledge, this is the rst longitudinal study in
older hypertensive subjects with subjective memory complaints in
which a complete biochemical and vascular workup is performed.
Other cohorts have longitudinally evaluated cognitive function and
leukoaraiosis, however, such evaluations were performed in patients
with established cognitive impairment or dementia and without a
thorough vascular workup,30,31 whereas other studies performed a
vascular workup but without an extended longitudinal evaluation.6,32
On top of the previous reports (including the transversal ADELAHYDE-
1 study),12 the present longitudinal study comprised a global memory
and leukoaraiosis assessment including integrative association and
prediction models with risk factors (both modiable and non-
modiable), vascular explorations, and cardiac evaluations and
treatments, thereby allowing a holistic interpretation of memory
impairment and brain lesions in elderly hypertensives without
established dementia. In this regard, the present study shows that
global memory is moderately determined by nonmodiable factors
such as increasing age and male gender, but also by modiable factors
such as smoking, LVH, and glycemic control. In addition, the present
analysis conrms that arterial thickness (assessed by cIMT) is an in-
dependent predictor of increased leukoaraiosis, which is also
(partially) determined by modiable factors such as BMI, diabetes
mellitus and LVH, as well as nonmodiable risk factors such as age.
Memory deterioration from V1 to V2 is possibly (trend) inuenced by
an increment in PWV, while an increase in leukoaraiosis from V1 to V2
is likely inuenced by smoking (deleterious) and by ACEi/ARBs and
statins (protective). Increasing SBP is also possibly (trend) associated
with an increment in leukoaraiosis. These factors are relevant beyond
previous memory status or WMH. The interpretation of the underly-
ing conditions able to inuence memory deterioration and leukoar-
aiosis increment is of utmost relevance for prognostic purposes and to
identify potential preventive strategies to be tested in future clinical
trials using memory assessment and leukoaraiosis as intermediate
endpoints.
Composite memory assessment and the Fazekas scale are weakly
correlated; hence, it is expected that they may be inuenced by a
variety of different factors. Herein, both leukoaraiosis and global
memory were inuenced by age, whereas male gender was likely to be
associated with worse memory (but not WMH) independently of age.
Both diabetes/glycemic control and end-organ damage (ie, LVH)
negatively inuenced both memory and WMH. Of note, increased
leukoaraiosis (but not memory deterioration), suggesting that leu-
koaraiosis is inuenced by early alterations in the vascular tree.
In the ADELAHYDE-1 transversal study, PWV was associated with
memory impairment in men,12 whereas this nding was not present
in the patients who completed both visits (possibly as a consequence
of the loss of statistical power due to the sample size reduction from
V1 to V2); however, the increase in PWV from V1 to V2 displayed a
trend for association with memory impairment. A higher PWV has
previously been associated with worse cognitive function in obser-
vational cross-sectional studies in various population settings.33e37
Furthermore, among young healthy adults, a higher PWV has been
shown to be associated with measures of reduced white and gray
matter integrity in areas implicated in cognitive decline.38 Herein, we
demonstrate that PWV increment (ie, arterial stiffness) may be asso-
ciated with cognitive decline over time. To date, there are no studies or
trials evaluating how or whether interventions on arterial stiffness can
reduce or prevent cognitive decline and the risk of developing de-
mentia. Moreover, additional research is required to establish a causal
link between PWV and cognition, and to identify whether potential
interventions aimed to reduce arterial stiffness may benet cognitive
performance in the elderly.39 In addition, in the present cohort, a
higher cIMT was independently associated with increased leukoar-
aiosis. Carotid IMT has been suggested as a promising method to
measure atherosclerosis, arterial hypertrophy, or hyperplasia induced
by pressure overload and age-related sclerosis directly at the vessel
level,40,41 and it has also been associated with an increased leukoar-
aiosis burden in other cohorts.42 Altogether, these ndings suggest
that alterations in vascular structure and function are crucial for brain
performance. However, PWV/arterial stiffness may be preferentially
associated with cerebral perfusion efciency, whereas cIMT may be
more associated with small/deep vascular lesions.43,44 Nonetheless,
both are associated (without established causality) with cognitive
impairment, stroke risk, and death.40,45,46
An additional relevant nding in our study is that a simple, readily
available and inexpensive LVH determination on ECG is associated
with and able to predict both cognitive impairment and WMH. This
nding is consistent with other studies,47,48 in which both echocar-
diographic and electrocardiographic LVH have been associated with
asymptomatic cerebral lesions and dementia.
In the present study, several modiable risk factors (eg, blood
pressure, diabetes mellitus/glucose control, obesity, and smoking)
were associated with both memory decline and increased
 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13
leukoaraiosis. These are the same factors that have been associated
with higher arterial stiffness and cIMT,49e52 and their adequate con-
trol has been associated with reductions in arterial wall thickness and
stiffness,53e56 suggesting that these measurements can be used as
intermediate markers of CV complications in trials. Moreover, in pa-
tients with severe WMH burden, statin therapy is likely to delay
leukoaraiosis progression57; however, their effect on dementia pre-
vention and treatment is yet to be determined.58 Similarly, effective
hypertension treatment may reduce leukoaraiosis,59,60 and, to some
extent, may help to slow cognitive decline.61,62 These results are
supportive of the inverse associations of statins/ACEi/ARBs and WMH
observed herein. Notwithstanding, recent large RCTs support the use
of statins and strict blood pressure control in high cardiovascular-risk
populations.63,64
Clinical and Research Perspective
Our study reinforces the need for action in regard to primary and
secondary prevention, because many factors associated with memory
impairment and leukoaraiosis can be modied: blood pressure
reduction (possibly with ACEi/ARBs in rst-line), statin use, and
glucose and weight control are likely attainable with adequate
interventions.
From a research perspective, randomized controlled trials using
statins and/or ACEi/ARBs should be performed in populations at high
risk for dementia in order to assess their impact on intermediate
endpoints such as memory and leukoaraiosis.
Limitations
Several limitations should be acknowledged in this study. First, this
is an observational study, hence, causality cannot be ascertained.
However, its longitudinal nature allows testing the inuence of risk
factors over time. Second, our study does not include nonhypertensive
patients or a control population; therefore, the associations found
herein cannot be generalized to other populations limiting the
external validity of our ndings. Third, all patients were taking anti-
hypertensive medication (of different classes), and the testing of a
particular drug class was not prespecied, nor was randomization or
concealed allocation performed increasing the probability of chance
ndings. Fourth, other secondary causes of dementia (eg, thyroid
dysfunction, nutritional deciencies, infections) were not assessed.
Finally, the high drop-out rate may have limited the power and ac-
curacy to determine the associations, increasing the probability of
type I error occurrence. However, the clinical and statistical consis-
tency of the ndings reinforces the internal validity of our results.
Conclusions
Several readily identiable factors are associated with memory and
white matter deterioration, many of which (including smoking, dia-
betes, hypertension, and obesity) are potentially modiable. In-
terventions aimed to control these risk factors need to be tested
prospectively in order to assess their cognitive protective value.
References
1. Ritchie K, Lovestone S. The dementias. Lancet 2002;360:1759e1766.
2. OBrien JT, Erkinjuntti T, Reisberg B, et al. Vascular cognitive impairment.
Lancet Neurol 2003;2:89e98.
3. Appelros P, Samuelsson M, Lindell D. Lacunar infarcts: Functional and cognitive
outcomes at ve years in relation to MRI ndings. Cerebrovasc Dis 2005;20:
34e40.
4. Schmidt R, Scheltens P, Erkinjuntti T, et al. White matter lesion progression: A
surrogate endpoint for trials in cerebral small-vessel disease. Neurology 2004;
63:139e144.
1.e7
5. Al-Qazzaz NK, Ali SH, Ahmad SA, Islam S. Cognitive assessments for the early
diagnosis of dementia after stroke. Neuropsychiatr Dis Treat 2014;10:
1743e1751.
6. Muller M, Grobbee DE, Aleman A, et al. Cardiovascular disease and cognitive
performance in middle-aged and elderly men. Atherosclerosis 2007;190:
143e149.
7. Rafnsson SB, Deary IJ, Smith FB, et al. Cardiovascular diseases and decline in
cognitive function in an elderly community population: The Edinburgh Artery
Study. Psychosom Med 2007;69:425e434.
8. Fazekas G, Fazekas F, Schmidt R, et al. Brain MRI ndings and cognitive
impairment in patients undergoing chronic hemodialysis treatment. J Neurol
Sci 1995;134:83e88.
9. Launer LJ. Epidemiology of white matter lesions. Top Magn Reson Imaging
2004;15:365e367.
10. Debette S, Markus HS. The clinical importance of white matter hyperintensities
on brain magnetic resonance imaging: Systematic review and meta-analysis.
BMJ 2010;341:c3666.
11. Prins ND, Scheltens P. White matter hyperintensities, cognitive impairment
and dementia: An update. Nat Rev Neurol 2015;11:157e165.
12. Kearney-Schwartz A, Rossignol P, Bracard S, et al. Vascular structure and
function is correlated to cognitive performance and white matter hyper-
intensities in older hypertensive patients with subjective memory complaints.
Stroke 2009;40:1229e1236.
13. Derouesne C [Dementia. Diagnostic problems]. Therapiewoche 1993;48:
189e193.
14. Verdelho A, Henon H, Lebert F, et al. Depressive symptoms after stroke and
relationship with dementia: A three-year follow-up study. Neurology 2004;62:
905e911.
15. Yesavage JA, Tinklenberg JR [Computerized tests of attention, memory and
motor performance applied to geriatric research]. Presse Med 1983;12:
3170e3172.
16. 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;
12:189e198.
17. Grober E, Buschke H, Crystal H, et al. Screening for dementia by memory
testing. Neurology 1988;38:900e903.
18. Spreen O, Benton AL. Simulation of mental deciency on a visual memory test.
Am J Ment Dec 1963;67:909e913.
19. Peigneux P, Van Der Linden M, Andres-Benito P, et al [A neuropsychological
and functional brain imaging study of visuo-imitative apraxia]. Rev Neurol
(Paris) 2000;156:459e472.
20. Spreen O. Prognosis of learning disability. J Consult Clin Psychol 1988;56:
836e842.
21. Kerber KA, Whitman GT, Brown DL, Baloh RW. Increased risk of death in
community-dwelling older people with white matter hyperintensities on MRI.
J Neurol Sci 2006;250:33e38.
22. Fazekas F, Chawluk JB, Alavi A, et al. MR signal abnormalities at 1.5 T in Alz-
heimers dementia and normal aging. AJR Am J Roentgenol 1987;149:351e356.
23. Wahlund LO, Barkhof F, Fazekas F, et al. A new rating scale for age-related
white matter changes applicable to MRI and CT. Stroke 2001;32:1318e1322.
24. Kapeller P, Barber R, Vermeulen RJ, et al. Visual rating of age-related white
matter changes on magnetic resonance imaging: Scale comparison, interrater
agreement, and correlations with quantitative measurements. Stroke 2003;34:
441e445.
25. Levey AS, Bosch JP, Lewis JB, et al. A more accurate method to estimate
glomerular ltration rate from serum creatinine: A new prediction equation.
Modication of Diet in Renal Disease Study Group. Ann Intern Med 1999;130:
461e470.
26. Touboul PJ, Hennerici MG, Meairs S, et al. Mannheim carotid intima-media
thickness consensus (2004-2006). An update on behalf of the Advisory Board
of the 3rd and 4th Watching the Risk Symposium, 13th and 15th European
Stroke Conferences, Mannheim, Germany, 2004, and Brussels, Belgium, 2006.
Cerebrovasc Dis 2007;23:75e80.
27. Asmar R, Benetos A, Topouchian J, et al. Assessment of arterial distensibility by
automatic pulse wave velocity measurement. Validation and clinical applica-
tion studies. Hypertension 1995;26:485e490.
28. ORourke MF, Gallagher DE. Pulse wave analysis. J Hypertens Suppl 1996;14:
S147eS157.
29. Okin PM, Devereux RB, Jern S, et al. Regression of electrocardiographic left
ventricular hypertrophy during antihypertensive treatment and the prediction
of major cardiovascular events. JAMA 2004;292:2343e2349.
30. Carmichael O, Schwarz C, Drucker D, et al. Longitudinal changes in white
matter disease and cognition in the rst year of the Alzheimer disease neu-
roimaging initiative. Arch Neurol 2010;67:1370e1378.
31. Tosto G, Zimmerman ME, Carmichael OT, Brickman AM. Predicting aggressive
decline in mild cognitive impairment: The importance of white matter
hyperintensities. JAMA eNeurol 2014;71:872e877.
32. Hanon O, Haulon S, Lenoir H, et al. Relationship between arterial stiffness and
cognitive function in elderly subjects with complaints of memory loss. Stroke
2005;36:2193e2197.
33. Fukuhara M, Matsumura K, Ansai T, et al. Prediction of cognitive function by
arterial stiffness in the very elderly. Circ J 2006;70:756e761.
34. Triantafyllidi H, Arvaniti C, Lekakis J, et al. Cognitive impairment is related to
increased arterial stiffness and microvascular damage in patients with never-
treated essential hypertension. Am J Hypertens 2009;22:525e530.
1.e8 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13
35. Poels MM, van Oijen M, Mattace-Raso FU, et al. Arterial stiffness, cognitive
decline, and risk of dementia: The Rotterdam study. Stroke 2007;38:888e892.
36. Watson NL, Sutton-Tyrrell K, Rosano C, et al. Arterial stiffness and cognitive
decline in well-functioning older adults. J Gerontol A Biol Sci Med Sci 2011;66:
1336e1342.
37. Zhong W, Cruickshanks KJ, Schubert CR, et al. Pulse wave velocity and cognitive
function in older adults. Alzheimer Dis Assoc Disord 2014;28:44e49.
38. Maillard P, Mitchell GF, Himali JJ, et al. Effects of arterial stiffness on brain
integrity in young adults from the Framingham Heart Study. Stroke 2016;47:
1030e1036.
39. Zeki Al Hazzouri A, Yaffe K. Arterial stiffness and cognitive function in the
elderly. J Alzheimers Dis 2014;42:S503eS5014.
40. Bots ML, Hoes AW, Koudstaal PJ, et al. Common carotid intima-media thickness
and risk of stroke and myocardial infarction: The Rotterdam Study. Circulation
1997;96:1432e1437.
41. Di Bello V, Carerj S, Perticone F, et al. Carotid intima-media thickness in
asymptomatic patients with arterial hypertension without clinical cardiovas-
cular disease: Relation with left ventricular geometry and mass and coexisting
risk factors. Angiology 2009;60:705e713.
42. Ueda K, Watanabe Y, Katsumata T, et al. Carotid intima-media thickness and
cerebral white matter lesions are more advanced in acute ischemic stroke
patients with renal dysfunction. Clin Nephrol 2011;76:290e295.
43. Isa K, Sakima A, Sakima H, et al. Association between the intima-media
thickness of the brachiocephalic trunk and white matter hyperintensity in
brain MRI. Hypertens Res 2013;36:980e984.
44. Tarumi T, Shah F, Tanaka H, Haley AP. Association between central elastic ar-
tery stiffness and cerebral perfusion in deep subcortical gray and white matter.
Am J Hypertens 2011;24:1108e1113.
45. Tsivgoulis G, Vemmos K, Papamichael C, et al. Common carotid artery intima-
media thickness and the risk of stroke recurrence. Stroke 2006;37:1913e1916.
46. Mattace-Raso FU, van der Cammen TJ, Hofman A, et al. Arterial stiffness and
risk of coronary heart disease and stroke: The Rotterdam Study. Circulation
2006;113:657e663.
47. Kahonen-Vare M, Brunni-Hakala S, Lindroos M, et al. Left ventricular hyper-
trophy and blood pressure as predictors of cognitive decline in old age. Aging
Clin Exp Res 2004;16:147e152.
48. Selvetella G, Notte A, Maffei A, et al. Left ventricular hypertrophy is associated
with asymptomatic cerebral damage in hypertensive patients. Stroke 2003;34:
1766e1770.
49. The Reference Values for Arterial Stiffness Collaboration. Determinants of
pulse wave velocity in healthy people and in the presence of cardiovascular
risk factors: Establishing normal and reference values. Eur Heart J 2010;31:
2338e2350.
50. Ferreira JP, Girerd N, Bozec E, et al. Intima-media thickness is linearly and
continuously associated with systolic blood pressure in a population-based
cohort (STANISLAS Cohort Study). J Am Heart Assoc 2016;5:e003529.
51. Blacher J, Asmar R, Djane S, London GM, et al. Aortic pulse wave velocity as a
marker of cardiovascular risk in hypertensive patients. Hypertension 1999;
1933:1111e1117.
52. Salonen R, Salonen JT. Determinants of carotid intima-media thickness: A
population-based ultrasonography study in eastern Finnish men. J Intern Med
1991;229:225e231.
53. Yokoyama H, Katakami N, Yamasaki Y. Recent advances of intervention to
inhibit progression of carotid intima-media thickness in patients with type 2
diabetes mellitus. Stroke 2006;37:2420e2427.
54. Wang JG, Staessen JA, Li Y, et al. Carotid intima-media thickness and antihy-
pertensive treatment: A meta-analysis of randomized controlled trials. Stroke
2006;37:1933e1940.
55. Riccioni G. Statins and carotid intima-media thickness reduction: An up-to-
date review. Curr Med Chem 2009;16:1799e1805.
56. Asmar R. Effect of antihypertensive agents on arterial stiffness as evaluated by
pulse wave velocity: Clinical implications. Am J Cardiovasc Drugs 2001;1:
387e397.
57. Mok VC, Lam WW, Fan YH, et al. Effects of statins on the progression of cerebral
white matter lesion: Post hoc analysis of the ROCAS (Regression of Cerebral
Artery Stenosis) study. J Neurol 2009;256:750e757.
58. Gauthier JM, Massicotte A. Statins and their effect on cognition: Lets clear up
the confusion. Can Pharm J (Ott) 2015;148:150e155.
59. de Leeuw FE, de Groot JC, Oudkerk M, et al. Hypertension and cerebral white
matter lesions in a prospective cohort study. Brain 2002;125:765e772.
60. Liao D, Cooper L, Cai J, et al. Presence and severity of cerebral white
matter lesions and hypertension, its treatment, and its control. The ARIC
Study. Atherosclerosis Risk in Communities Study. Stroke 1996;27:
2262e2270.
61. Duron E, Hanon O. Antihypertensive treatments, cognitive decline, and de-
mentia. J Alzheimers Dis 2010;20:903e914.
62. Rouch L, Cestac P, Hanon O, et al. Antihypertensive drugs, prevention of
cognitive decline and dementia: A systematic review of observational studies,
randomized controlled trials and meta-analyses, with discussion of potential
mechanisms. CNS Drugs 2015;29:113e130.
63. Wright JT Jr, Williamson JD, Whelton PK, et al. A randomized trial of
intensive versus standard blood-pressure control. N Engl J Med 2015;373:
2103e2116.
64. Yusuf S, Lonn E, Pais P, et al. Blood-pressure and cholesterol lowering in
persons without cardiovascular disease. N Engl J Med 2016;374:
2032e2043.
 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13 1.e9

Appendix

Supplemental Table S1
Associations Between MRI Fazekas Score and Global Memory at V1 Assessed by Multiple Linear Regression Analysis

Adjusted R2 Standardized b Adjusted Estimate (95% CI) P Value

Global memory (V1)

Overall model t 0.30 - - <.001
Constant - - 3.65 (6.56 to .0.74) .015
Gender (male) - 0.37 0.80 (0.33e1.26) .001
Age (per 10-year increase) - 0.21 0.43 (0.01e0.86) .045
Smoking (yes) - 0.24 0.52 (0.04e0.99) .035
LVH (yes) - 0.19 0.60 (0.05 to 1.25) .069
Stroke (yes) - 0.24 1.24 (0.18e2.30) .023
Global memory (V2)
Model 1: Excluding global memory score at V1
Overall model t 0.11 - - .001
Constant - - 2.32 (0.64e4.00) .007
Gender (male) - 0.32 0.85 (0.36e1.33) .001
Model 2: Including global memory score at V1
Overall model t 0.21 - - <.001
Constant - - 1.91 (0.31e3.51) .020
Global memory composite at V1 - 0.34 0.90 (0.41e1.38) <.001
Gender (male) - 0.21 0.54 (0.05e1.02) .031
MRI Fazekas scale (V1)
Overall model t 0.38 - - <.001
Constant - - 10.10 (14.94 to 5.26) <.001
Age (per 10-year increase) - 0.22 0.59 (0.05e1.13) .032
LVH (yes) - 0.22 0.93 (0.11e1.75) .026
HbA1c (per 1% increase) - 0.21 0.28 (0.00e0.56) .048
cIMT (per 1 mm increase) - 0.37 6.32 (2.77e9.87) .001
ACEi/ARBs (yes) - 0.20 0.61 (1.20 to 0.03) .041
MRI Fazekas scale (V2)
Model 1: Excluding MRI Fazekas scale at V1
Overall model t 0.14 - - .004
Constant - - 1.28 (6.09 to 3.54) .600
Age (per 10-year increase) 0.73 0.73 (0.14e1.31) .016
Model 2: Including MRI Fazekas scale at V1
Overall model t 0.73 - - <.001
Constant - - 3.31 (5.43 to 1.19) .003
MRI Fazekas scale V1 - 0.85 1.06 (0.92e1.21) <.001
SBP (per 10 mm Hg increase) - 0.19 0.17 (0.07e0.27) .001
HbA1c (per 1% increase) - 0.18 0.31 (0.10e0.52) .005
ACEi/ARBs (yes) - 0.15 0.52 (0.91 to 0.14) .008

ACEi/ARBs, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers; CCB, calcium channel blockers; CI, condence interval; cIMT, carotid intima-media
thickness (right carotid artery); HbA1c, glycated hemoglobin; MRI, magnetic resonance imaging; PWV, pulse wave velocity; SBP, systolic blood pressure; V1, visit 1; V2, visit 2.
A stepwise backward conditional method was used for all models.
In the MRI Fazekas scale and global memory model at V1, the variables introduced in the initial model were age, gender, education level, smoking status, BMI, SBP, heart rate,
LV hypertrophy on ECG, eGFR, microalbuminuria, LDL cholesterol, homocysteine, hypertension duration, diabetes status and HbA1c, history of cardiovascular events, med-
ications, intima-media thickness, pulse-wave velocity and augmentation indexdall at V1.
In the MRI Fazekas scale and global memory model at V2, the variables introduced in the initial model were MRI Fazekas scale and global memory at V1 (only in model 2), age,
gender, education level, smoking status, BMI, SBP, heart rate, LV hypertrophy on ECG, eGFR, microalbuminuria, sodium, potassium, hemoglobin, LDL cholesterol, homocys-
teine, C-reactive protein, hypertension duration, diabetes status and HbA1c, history of cardiovascular events, medications, intima-media thickness, pulse-wave velocity,
augmentation indexdall at V2.
1.e10 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13

Supplemental Table S2
Multiple Linear Regression Analysis of Memory Assessment for Free Recall, Cued Recall, and Total (Free plus Cued) Recall at V1, V2, and D (V2 e V1)

Adjusted R2 Standardized b Adjusted Estimate (95% CI) P Value

Free recall (V1)

Overall model t 0.41 - - <.001
Constant - - 58.50 (45.46e71.56) <.001
Gender (male) - 0.54 7.08 (9.68 to 4.48) <.001
LDL cholesterol (per 1 g/L increase) - 0.22 5.09 (9.82 to 0.37) .035
HbA1c (per 1% increase) - 0.23 1.44 (2.69 to 0.19) .025
Myocardial infarction (yes) - 0.25 6.22 (11.34 to 1.11) .018
Augmentation index (per 1% increase) - 0.33 0.21 (0.33 to 0.08) .002
Free recall (V2)
Overall model t 0.45 - - <.001
Constant - - 3.64 (12.45 to 5.18) .414
Free recall V1 - 0.62 0.95 (0.71e1.20) <.001
Intermediate/high education level (yes) - 0.17 3.41 (0.16e6.65) .040
Hypertension (years) - 0.19 0.21 (0.38 to 0.04) .018
ACEi/ARBs (yes) - 0.16 3.04 (0.04 to 6.12) .053
D Free recall (V1 to V2)
Overall model t 0.04 - - .027
Constant - - 6.48 (8.56 to 4.04) <.001
ACEi/ARBs (yes) - 0.16 2.28 (0.24 to 4.80) .075
Cued recall (V1)
Overall model t 0.11 - - .009
Constant - - 14.05 (12.74e15.35) <.001
Gender (male) - 0.26 2.38 (0.42e4.45) .018
Cued recall (V2)
Overall model t 0.40 - - <.001
Constant - - 8.64 (22.08 to 4.79) .205
Cued recall V1 - 0.56 0.75 (0.55e0.94) <.001
D Cued recall (V1 to V2)
Overall model t 0.12 - - .006
Constant - - 1.94 (0.68e3.18) .003
D PWV (per 1 m/s increase) - 0.30 0.36 (0.09e0.63) .010
Atheroma plaque (yes) - 0.27 3.14 (0.51e5.77) .020
Total (free cued) recall (V1)
Overall model t 0.23 - - <.001
Constant - - 45.50 (39.32e51.67) <.001
Gender (male) - 0.34 2.70 (4.38 to 1.02) .002
Body mass index (per 1 kg/m2 increase) - 0.24 0.25 (0.04e0.45) .019
MDRD (per 10 mL/min/1.73 m2 increase) - 0.27 0.09 (0.16 to 0.02) .020
Total (free cued) recall (V2)
Overall model t 0.21 - - <.001
Constant - - 54.95 (49.45e65.44) <.001
Total recall V1 - 0.32 0.40 (0.14e0.46) <.001
Age (per 10-year increase) - 0.20 1.47 (2.80 to 0.15) .030
D Total (free cued) recall (V1 to V2)
Overall model t 0.15 - - .001
Constant - - 1.69 (2.53 to 0.85) <.001
D MDRD (mL/min/1.73 m2) - 0.18 0.06 (0.01e0.12) .045

ACEi/ARBs, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers; BMI, body mass index; CI, condence interval; DBP, diastolic blood pressure; ECG,
electrocardiogram; eGFR, estimated glomerular ltration rate; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; LV, left ventricular;
MDRD; modied of diet in renal disease equation 4; PWV, pulse wave velocity; SBP, systolic blood pressure; V1, visit 1; V2, visit 2.
A stepwise backward conditional method was used.
Variables included in both V1 and V2 models were age, gender, education level, smoking status, BMI, SBP, DBP, pulse pressure, heart rate, LV hypertrophy on ECG, eGFR,
microalbuminuria, total cholesterol, LDL cholesterol, HDL cholesterol, homocysteine, hypertension duration, diabetes status and HbA1c, history of cardiovascular events,
medications, intima-media thickness, pulse-wave velocity, and augmentation index.
In the D models, the variables introduced were D age, gender, education level, smoking status, D BMI, D SBP, D heart rate, D eGFR, D microalbuminuria, D LDL cholesterol, D
homocysteine, D hypertension duration, D HbA1c, history of cardiovascular events, medications, D intima-media thickness, D pulse-wave velocity, and D augmentation index.
DBP, pulse pressure, HDL and total cholesterol were removed from all models due to collinearity.
 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13 1.e11

Supplemental Table S3
Multiple Logistic Regression Analysis of Dichotomic MRI Fazekas Score and Global
Memory at V1, V2, and D (V2-V1)

OR (95% CI) P Value

Better global memory (reference) vs worse global memory at V1

Smoking (yes) 2.78 (1.26e6.13) .011
LVH (yes) 3.42 (1.10e10.61) .033
cIMT (per 1 mm increase) 163 (2e15860) .029
Better global memory (reference) vs worse global memory at V2
Age (per 10-year increase) 1.86 (0.93e3.74) .081
Gender (male) 4.59 (2.13e9.89) <.001
Intermediate/high education level (yes) 0.46 (0.19e1.10) .079
Worsening in global memory composite from V1 to V2
D Age (per 10-year increase) 1.48 (0.99e2.21) .055
MRI Fazekas score 0 or 1 (reference) vs Fazekas score 2 at V1
Age (per 10-year increase) 3.47 (1.05e11.53) .042
Augmentation index (per 1% increase) 1.09 (1.02e1.16) .008
HbA1c (per 1% increase) 2.23 (0.98e5.05) .056
LVH (yes) 6.40 (0.71e57.84) .099
MRI Fazekas score 0 or 1 (reference) vs Fazekas score 2 at V2
Age (per 10-year increase) 1.13 (0.96e1.32) .146
2 point increase in MRI Fazekas score from V1 to V2
D SBP (per 1 mm Hg increase) 1.02 (0.99e1.04) .08
Atheroma plaque (yes) 2.50 (0.97e6.43) .057

cIMT, carotid intima-media thickness (right carotid artery); BMI, body mass index;
ECG, electrocardiogram; eGFR, estimated glomerular ltration rate; HbA1c, glycated
hemoglobin; HDL, high-density lipoprotein; LV, left ventricular; MRI, magnetic
resonance imaging; SBP, systolic blood pressure; V1, visit 1; V2, visit 2.
V1: Fazekas score 2, n (%) 83 (63.4).
V2: Fazekas score 2, n (%) 104 (79.4).
2 point increase in Fazekas score from V1 to V2, n (%) 41 (31.3).
V1: Global memory above the median (worse), n (%) 65 (49.6).
V2: Global memory above the median ( worse), n (%) 65 (49.6).
Increase ( worsening) global memory from V1 to V2, n (%) 16 (12.2).
Variables were initially tested in univariable analysis, and only variables with P < .15
were retained in the model, after which a backward selection process was per-
formed.
In the MRI Fazekas score and global memory model at V1, the variables introduced
in the initial model were age, gender, education level, smoking status, BMI, SBP,
heart rate, LV hypertrophy on ECG, eGFR, microalbuminuria, LDL cholesterol, ho-
mocysteine, hypertension duration, diabetes status and HbA1c, history of cardio-
vascular events, medications, intima-media thickness, pulse-wave velocity, and
augmentation indexdall at V1.
In the MRI Fazekas score and global memory model at V2, the variables introduced
in the initial model were MRI Fazekas score and global memory at V1 (only in model
2), age, gender, education level, smoking status, BMI, SBP, heart rate, LV hypertrophy
on ECG, eGFR, microalbuminuria, sodium, potassium, hemoglobin, LDL cholesterol,
homocysteine, C-reactive protein, hypertension duration, diabetes status and
HbA1c, history of cardiovascular events, medications, intima-media thickness,
pulse-wave velocity and augmentation indexdall at V2.
In the D MRI Fazekas score and global memory change model from V1 to V2, the
variables introduced in the initial model were D age, gender, education level,
smoking status, D BMI, D SBP, D heart rate, D eGFR, D microalbuminuria, D LDL
cholesterol, D homocysteine, D hypertension duration, D HbA1c, history of cardio-
vascular events, medications, D intima-media thickness, D pulse-wave velocity, and
D augmentation index.
DBP, pulse pressure, HDL, and total cholesterol were removed from all models due
to the presence of collinearity.
1.e12 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13

Supplemental Table S4
Multinomial Logistic Regression for MRI Fazekas Score at V1 and V2 According to
Grade Categories

Grade 0 or 1 Variables OR (95% CI) P Value

(ref.)

MRI Fazekas score at V1

Grade 2 or 3 Age (per 10-year increase) 2.94 (1.20e7.23) .019
Grade  4 Age (per 10-year increase) 3.81 (1.25e11.63) .019
cIMT (per 1 mm increase) 4514 (6e>100,000) .014
MRI Fazekas score at V2
Grade 2 or 3 Age (per 10-year increase) 2.09 (0.46e9.45) .335
Grade  4 Age (per 10-year increase) 4.58 (0.98e21.36) .053

CI, condence interval; cIMT, carotid intima-media thickness (right carotid artery);
BMI, body mass index; DBP, diastolic blood pressure; ECG, electrocardiogram; eGFR,
estimated glomerular ltration rate; HbA1c, glycated hemoglobin; HDL, high-
density lipoprotein; LDL, low-density lipoprotein; LV, left ventricular; MRI, mag-
netic resonance imaging; OR, odds ratio; SBP, systolic blood pressure; V1, visit 1; V2,
visit 2.
Variables were initially tested in univariable analysis, and only variables with P < .15
were retained in the model, after which a backward selection process was per-
formed.
In the MRI Fazekas score model at V1, the variables introduced in the initial model
were age, gender, education level, smoking status, BMI, SBP, heart rate, LV hyper-
trophy on ECG, eGFR, microalbuminuria, LDL cholesterol, homocysteine, hyperten-
sion duration, diabetes status and HbA1c, history of cardiovascular events,
medications, intima-media thickness, pulse-wave velocity, and augmentation
indexdall at V1.
In the MRI Fazekas score model at V2, the variables introduced in the initial model
were MRI Fazekas score and global memory at V1 (only in model 2), age, gender,
education level, smoking status, BMI, SBP, heart rate, LV hypertrophy on ECG, eGFR,
microalbuminuria, sodium, potassium, hemoglobin, LDL cholesterol, homocysteine,
c-reactive protein, hypertension duration, diabetes status and HbA1c, history of
cardiovascular events, medications, intima-media thickness, pulse-wave velocity,
and augmentation indexdall at V2.
DBP, pulse pressure, HDL, and total cholesterol were removed from all models due
to the presence of collinearity.
 J.P. Ferreira et al. / JAMDA xxx (2017) 1.e1e1.e13 1.e13

Supplemental Table S5
Characteristics of Patients Who Completed the Two Visits Versus Those Who Died or Were Lost to Follow-Up

V1 and V2 (n 131) Deceased (n 51) Lost to Follow-Up/Refusal (n 196) P Value

Female gender, n (%) 68 (51.9) 25 (49.0) 112 (57.1) .47

Age (years) 67.7  5.3 74.1  6.3 71.2  6.3 <.0001
Intermediate/high education level, n (%) 97 (74.6) 28 (54.9) 115 (58.7) .005
Past/current smoker, n (%) 50 (38.2) 32 (62.7) 70 (35.7) .0006
BMI (kg/m2) 27.8  4.1 27.1  4.1 28.4  4.5 .2
SBP (mm Hg) 136.7  16.1 142.0  17.9 139.7  18.3 .13
DBP (mm Hg) 73.6  9.4 74.5  10.5 73.8  9.3 .81
Pulse pressure (mm Hg) 63.2  13.2 67.6  15.9 66.0  14.7 .22
Heart rate (bpm) 62.3  8.7 68.0  11.9 63.3  9.7 .007
LV hypertrophy on ECG, n (%) 17 (13.3) 13 (26.0) 36 (18.5) .12
MDRD eGFR (mL/min/1.73 m2) 64.7  11.2 60.7  13.9 63.4  12.5 .13
Microalbuminuria (mg/L)* 4.80 (3.00e17.1) 8.30 (3.40e16.30) 4.15 (3.00e8.50) .004
Total cholesterol (g/L) 2.19  0.33 2.21  0.37 2.16  0.35 .53
HDL cholesterol (g/L) 0.6  0.2 0.7  0.2 0.6  0.2 .9
LDL cholesterol (g/L) 1.32  0.29 1.31  0.34 1.29  0.33 .64
Homocysteine (mmol/L) 12.4  3.7 14.8  6.3 12.6  5.0 .027
Hypertension, n (%) 131 (100) 51 (100) 192 (100) .005
Hypertension duration (years)* 11 (5e17) 12 (5e18) 11 (6e19) .72
Diabetes, n (%) 12 (9.4) 8 (16.0) 20 (10.3) .43
HbA1c (%) 6.0  0.8 6.1  0.9 6.0  0.7 .34
Cardiovascular 9 (6.9) 7 (13.7) 18 (9.2) .35
Stroke 2 (1.5) 2 (3.9) 3 (1.5) .5
Transient ischemic attack 4 (3.1) 7 (14.0) 12 (6.2) .026
ACEi/ARBs, n (%) 82 (62.6) 34 (66.7) 131 (66.8) .8
b-blockers, n (%) 45 (34.4) 15 (29.4) 63 (32.1) .8
CCB, n (%) 42 (32.1) 17 (33.3) 68 (34.7) .88
Diuretics, n (%) 60 (45.8) 28 (54.9) 94 (48.0) .54
Statins, n (%) 43 (32.8) 15 (29.4) 61 (31.1) .89
Antithrombotic agents, n (%) 36 (27.5) 13 (25.5) 58 (29.6) .82
Free recall (1) 29.8  5.9 26.2  7.7 28.2  6.5 .006
Cued recall (2) 15.6  4.5 17.0  5.7 16.8  5.0 .056
Total recall (1 2) 45.5  3.5 43.1  6.8 45.0  3.3 .05
Cued recall (2) 0.0  1.0 0.1  1.1 0.0  1.0 .4

ACEi/ARBs, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers; BMI, body mass index; bpm, beats per minute; CCB, calcium channel blockers; DBP,
diastolic blood pressure; ECG, electrocardiogram; eGFR, estimated glomerular ltration rate; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density
lipoprotein; LV, left ventricular; MDRD, modied of diet in renal disease equation 4; MRI, magnetic resonance imaging; SBP, systolic blood pressure; V1, visit 1; V2, visit 2.
For continuous variables, results are presented as mean standard deviation while *non-parametric test results are presented as median (percentile 25th to 75th); categorical
variables are presented as numbers (n) and proportions (%).