Association of Antioxidant Supplement Use and Dementia in the
Prevention of Alzheimer’s Disease by Vitamin E and Selenium
Trial (PREADViSE)
Richard J. Kryscio, Ph.D.
1,2,3,4
, Erin L. Abner, Ph.D.
1,2,3,5
, Allison Caban-Holt, Ph.D.
1,2
, Mark
Lovell, Ph.D.
1,2,6
, Phyllis Goodman, M.S.
7
, Amy K. Darke, M.S.
7
, Monica Yee, B.A.
8
, John
Crowley, Ph.D.
8
, and Frederick A. Schmitt, Ph.D.
1,2,9
1
Sanders-Brown Center on Aging, University of Kentucky, Lexington, KY 40536, USA
2
Alzheimer’s Disease Center, University of Kentucky, Lexington, KY 40536, USA
3
Department of Biostatistics, University of Kentucky, Lexington, KY 40536, USA
4
Department of Statistics, University of Kentucky, Lexington, KY 40536, USA
5
Department of Epidemiology, University of Kentucky, Lexington, KY 40536, USA
6
Department of Chemistry, University of Kentucky, Lexington, KY 40536, USA
7
SWOG Statistical Center, Seattle, Fred Hutchinson Cancer Research Center, WA, USA
8
SWOG Statistical Center, Seattle, Cancer Research and Biostatistics, WA, USA
9
Department of Neurology, College of Medicine, University of Kentucky, Lexington, KY 40536,
USA
Abstract
Importance—Oxidative stress is an established dementia pathway, but it is unknown if the use of
antioxidant supplements can prevent dementia.
Objective—To determine if antioxidant supplements (vitamin E or selenium) used alone or in
combination can prevent dementia in asymptomatic older men.
Design—PREADVISE, designed to answer this question, began as a double-blind, randomized
controlled trial in 2002 which transformed into a cohort study from 2009–2015.
Setting—PREADVISE was ancillary to SELECT, a randomized controlled trial of the same anti-
oxidant supplements for preventing prostate cancer. SELECT closed in 2009 due to a futility
analysis.
Corresponding author: Richard J. Kryscio, Department of Statistics and Biostatistics and Alzheimer’s Disease Center, Sanders Brown
Center on Aging, University of Kentucky, Lexington, KY 40536, Phone: 859-257-4064 and FAX: 859-257-4665,
Kryscio@email.uky.edu.
Data Access: RJK and ELA had full access to the data and take responsibility for the integrity of the data and the accuracy of the data
analysis.
Conflict of interest: All authors declare no conflict of interest.
Ethical standards: The study protocol was approved by the University of Kentucky Institutional Review Board in addition to
Institutional Review Boards at each study site. All participants provided written informed consent.
HHS Public Access
Author manuscript
JAMA Neurol
. Author manuscript; available in PMC 2017 July 12.
Published in final edited form as:
JAMA Neurol
. 2017 May 01; 74(5): 567–573. doi:10.1001/jamaneurol.2016.5778.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Participants—PREADVISE recruited 7,540 men, of whom 3,786 continued into the cohort
study. Participants were at least 60 years old at study entry and were enrolled at one of 130
SELECT sites.
Intervention—Participants were randomized to vitamin E, selenium, vitamin E + selenium, or
placebo. While taking study supplements, PREADVISE men visited their SELECT site and were
evaluated for dementia using a two-stage screen. During the cohort study, men were contacted by
telephone and assessed using an enhanced two-stage cognitive screen. In both phases, men were
encouraged to visit their doctor if the screen indicated possible cognitive impairment.
Main Outcome—Dementia case ascertainment relied on a consensus review of the cognitive
screens and medical records for those suspected cases that visited their doctor for an evaluation, or
by review of all supplemental information provided by SELECT, including a functional
assessment screen.
Results—Under a modified intent-to-treat analysis, dementia incidence (4.43%) was not
different among the four study arms. A Cox model, which adjusted incidence for participant
demographics and baseline self-reported co-morbidities, yielded hazard ratios of 0.88 (95% CI:
0.64–1.20) for vitamin E, 1.00 (0.75–1.35) for the combination, and 0.83 (0.60–1.13) for selenium
compared to placebo.
Conclusions and Relevance—Neither supplement prevented dementia. This is the first study
to investigate the long term effect of anti-oxidant supplement use on dementia incidence among
asymptomatic men.
Trial Registration—ClinicalTrials.gov: NCT00040378.
Introduction
In the United States (US) an estimated 5 million elderly have Alzheimer’s disease (AD), and
AD cases are expected to increase substantially by 2050.
1
A recent review of clinical trials
from 1984–2014 shows a focus on enrolling mild to moderate dementia cases in many trials
with no real progress on identifying disease modifying treatments
2
. As a result, there has
been a shift in focus to clinical trials emphasizing the prevention of cognitive decline,
especially through secondary prevention trials targeting high risk groups
3,4
and large trials
that promote lifestyle changes to address modifiable risk factors for AD
5,6
. The usual
primary endpoints of these trials are cognitive decline or composites of biomarkers and
cognitive measures
7
. The gold standard of prevention is disease incidence, but few current
trials have this as their primary endpoint because of the time required to observe reductions
in disease incidence.
Multiple mechanisms associated with disease onset and progression have been described,
8
and one key mechanism implicated in AD is oxidative stress,
9
which may be modifiable
through diet and/or antioxidant supplements
10
. Antioxidant use as a potential treatment for
cognitive impairment or dementia has been of interest for many years. Vitamin E has had
mixed results in treatment trials: in moderately demented patients treated for two years, its
use slowed disease progression,
11
but more recently, when used in an antioxidant cocktail,
its use failed to improve cognition in AD patients with mild to moderate dementia
12
. It also
Kryscio et al.
Page 2
JAMA Neurol
. Author manuscript; available in PMC 2017 July 12.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
failed as a preventive agent for dementia progression in subjects with mild cognitive
impairment (MCI),
13
although early in the trial its use did somewhat improve cognition. A
review of controlled trials and case-control studies on the use of selenium in arresting the
progression of AD also yields mixed results
14
. Observational studies correlate cognitive
decline with decreased plasma selenium over time
15,16
. However, nothing is known about
long-term use of these supplements as preventative agents in asymptomatic individuals. The
purpose of this manuscript is to address this gap in the literature.
Specifically, this manuscript reports the main results of the Prevention of Alzheimer’s
Disease with Vitamin E and Selenium (PREADVISE) primary prevention trial.
PREADVISE, the largest primary prevention trial in AD to date, began in 2002 as an
ancillary trial within the Selenium and Vitamin E Cancer Prevention Trial (SELECT), a
double-blind, randomized controlled prostate cancer prevention trial
17
. When SELECT
ended prematurely in 2009 due to a futility analysis, PREADVISE continued as a cohort
study for a subset of its enrollees
18
. Extended follow-up over a seven-year period (blinded)
allowed for case ascertainment, yielding a comparison of the study arms for effectiveness in
preventing dementia. We present the results of this large primary prevention study of
antioxidant supplements as a method to modify oxidative stress, one mechanism in the
evolution of AD
9
.
Methods
Design
Study design
19
, participant recruitment
20
, and the conversion of the trial into a cohort
study
18
are described in earlier publications. Briefly, the parent study, SELECT, was
designed as a double-blind, four-arm randomized controlled trial (RCT) and initiated
enrollment in 2001
17
. SELECT’s primary aim was to determine the effectiveness of the
antioxidant supplements vitamin E (400 IU/day) and selenium (200 μg/day) alone or in
combination in preventing prostate cancer. PREADVISE recruited a subsample of SELECT
participants age 62 and over (age 60 if black) at 130 participating clinical sites in the US,
Canada, and Puerto Rico. PREADVISE enrolled 7,540 non-demented SELECT men
between 2002 and 2008 (Appendix, CONSORT diagram). The RCT was powered to detect a
hazard rate of 0.60 with 80% power assuming a targeted enrollment of 10,400 men; this was
then lowered to a detectable hazard rate 0.5 once the actual enrollment finished at 7,540
men.
Study eligibility required SELECT enrollment at a participating site and absence of:
dementia, active neurologic and/or neuropsychiatric conditions that affect cognition, as well
as history of serious head injury (>30-minute loss of consciousness within the last five years
prior to enrollment) and substance abuse. The RCT was scheduled to continue supplements
until 2012 but in September 2008, the SELECT Data and Safety Monitoring Committee
recommended that supplements be discontinued due to lack of efficacy on prostate cancer
incidence. Study sites closed over the next two years, during which time both PREADVISE
and SELECT transitioned into cohort studies
18,21
. RCT participants were asked to continue
in the cohort study, and 4,271 of the original 7,540 PREADVISE men consented to continue
annual memory screenings, which were then conducted by telephone. All study activities
Kryscio et al.
Page 3
JAMA Neurol
. Author manuscript; available in PMC 2017 July 12.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
were approved by Institutional Review Boards at the University of Kentucky and at each
SELECT site. All participants provided written informed consent.
Screening
The Memory Impairment Screen (MIS)
22
was used as the primary screening instrument in
both the RCT and cohort study. Participants who scored below cutoffs for intact cognition
received a secondary screening instrument. The modified Telephone Interview for Cognitive
Status (TICS-m)
23
was used during the cohort study, replacing the expanded Consortium to
Establish a Registry in Alzheimer’s Disease (CERADe) neuropsychological battery
24
used
during the RCT. Annual screenings were completed in May 2014, and case ascertainment
follow-up continued until fall 2015. All PREADVISE participants who completed at least
one follow-up visit were included in the current analyses (intent-to-treat analysis, ITT),
whether they participated in just the RCT or both the RCT and cohort studies.
Case Ascertainment
Dementia incidence, the primary endpoint, was determined by one of two methods. First, if a
man failed both the first tier of the screen (MIS ≤5 out of 8) and the second tier (T Score ≤
35 on the CERAD battery, total score ≤ 35 on the TICS-m), then he was encouraged to
obtain a memory work-up from his local clinician and share the medical records with
PREADVISE. Medical records were reviewed by a team of 2–3 expert neurologists and 2–3
expert neuropsychologists to determine a consensus diagnosis. Participants who did not
obtain the work-up were assessed by additional longitudinal measures collected during the
study. These included the AD8 Dementia Screening Interview
26
, self-reported medical
history, self-reported medication use, and cognitive scores including the MIS, CERAD T
Score, NYU Paragraph Delayed Recall
27
, and TICS-m. An AD8 ≥1 (at any time during
follow-up) plus a self-reported dementia diagnosis, use of a memory enhancing prescription
drug (i.e., donepezil, rivastigmine, galantamine, memantine), or cognitive score ≥1.5
standard deviations below expected performance yielded a dementia diagnosis. The
diagnosis date was assigned to the earliest event.
Statistical Methods
Groups were compared using chi-square statistics for categorical variables and two-sample t-
tests or analysis of variance for interval level variables. Cox proportional hazards models
were used perform a modified intent-to-treat (mITT) analysis to compare hazard rates
among the study arms. Hazard ratios were adjusted for the following covariates: baseline
age, black race,
APOE
ε4 carrier status (present or absent), college education, baseline MIS
score, and the presence/absence of the following self-reported co-morbidities at
PREADVISE baseline: coronary artery bypass graft (CABG), congestive heart failure
(CHF), diabetes, hypertension, stroke, sleep apnea, and memory change or problem. In this
analysis, survival time is the time from PREADVISE baseline to ascertainment of a case
(event) or last cognitive assessment (right censored). All dropouts due to death, personal
withdrawals from the study, and administrative withdrawals (e.g., SELECT sites not re-
consenting men for the cohort study) were right censored. The proportional hazards
assumption required by the Cox model was checked using martingale residuals and
Schoenfeld residuals. The latter analysis indicated that proportionality was met for the
Kryscio et al.
Page 4
JAMA Neurol
. Author manuscript; available in PMC 2017 July 12.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
indicators of the treatment arms and all covariates except for the indicators of baseline
memory problems and baseline MIS score less than 7 or 8. A stratified analysis using the
combination of these latter two variables yielded the same results for the effect of the
treatment arms. Therefore, the original Cox model is reported here. Missing
APOE
ε4
carrier status values (4.8% of the subjects) were imputed; a re-analysis limited to known
APOE
ε4 yielded virtually the same results as those reported here.
Three additional analyses were conducted. First, the observations were weighted by
treatment compliance. This weight equaled the length of time the man could take the
supplements (time between enrollment and the date the supplements were stopped)
multiplied by the proportion of visits at which the man was compliant with assigned
supplement one (vitamin E or placebo) or compliant with supplement two (selenium or
placebo), whichever was greater. Compliance information was based on returned pill counts.
For example, suppose a randomly selected study participant had three years on treatment,
and he was found to be 50% compliant with assigned treatment. The weight assigned to that
man’s record was 1.5 (3 years × 50% compliance, normed to make the sum of all weights
equal the sample size for this analysis given the 7,289 men with compliance information).
Second, the Cox models were re-fitted using only medical records-based dementia cases as
the outcome (cases identified without medical records were right censored).
Results
Table 1 shows that participants evaluated with at least one memory screen in the cohort
study (n=3,786) were similar to the complete PREADVISE enrollment, except possibly for
less education at the college level or higher (52.2% versus 60%), fewer black participants
(8.4% versus 10.0%), and fewer Hispanic participants (2.5% versus 6.9%). Table 2 lists
participant characteristics at the time of PREADVISE enrollment by study arm. Based on
the means and percentages, despite randomization occurring at SELECT baseline rather than
PREADVISE enrollment, there were no perceivable differences between study arms in terms
of medical history,
APOE
ε4 genotype, or initial MIS.
Incident dementia cases were defined as above. There were 325/7,540 incident dementia
cases in the study (4.31%) and, of these, 121/325 (37.2%) provided medical records (Table
3). Unadjusted cumulative incidence varied among the study arms: 3.95% in the vitamin E
arm, 4.15% in the selenium arm, 4.62% in the placebo arm, and 4.96% in the combination
arm (Table 3).
To compare the hazard rates for dementia among the four study arms, mITT analysis based
on a Cox proportional hazards model was fitted to the data. The mITT analysis excluded 201
participants who had only a baseline visit. Hence, the Cox model included 7,338 men with at
least one follow-up visit. None of the active study arms had a significantly lower adjusted
hazard rate for incident dementia when compared to the placebo arm; the selenium arm had
a hazard ratio (HR) of 0.83 vs. placebo (95% CI: 0.61–1.13), while the vitamin E arm had
HR=0.88 (95% CI: 0.64–1.20) (Table 4). The combined supplements arm was
indistinguishable from the placebo arm had HR=1.00 (95% CI: 0.74–1.35) (Figure 1).
Kryscio et al.
Page 5
JAMA Neurol
. Author manuscript; available in PMC 2017 July 12.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript