NeuroToxicology 61 (2017) 143–187

Contents lists available at ScienceDirect

NeuroToxicology

Full Length Article

Risk factors associated with the onset and progression of Alzheimer’s

disease: A systematic review of the evidence
Mona Hersia,b,* , Brittany Irvinea , Pallavi Guptaa , James Gomesa,c,d , Nicholas Birketta,b ,
Daniel Krewskia,b,e
a
McLaughlin Centre for Population Health Risk Assessment, University of Ottawa, Ottawa, Ontario, Canada
b
School of Epidemiology, Public Health and Preventive Medicine, Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada
c
Environmental Health Research Unit, University of Ottawa, ON, Canada
d
Interdisciplinary School of Health Sciences, Faculty of Health Sciences, University of Ottawa, ON, Canada
e
Risk Sciences International, Ottawa, Ontario Canada

A R T I C L E I N F O A B S T R A C T

Article history:
Received 22 March 2017 A systematic review was conducted to identify risk factors associated with the onset and progression of
Accepted 22 March 2017 Alzheimer’s disease (AD). Moderate and high quality systematic reviews were eligible for inclusion.
Available online 29 March 2017 Primary studies reporting on non-genetic risk factors associated with neuropathologically or clinically
confirmed AD were considered. Eighty one systematic reviews reporting on AD onset and 12 reporting on
Keywords: progression satisfied the eligibility criteria. Four hundred and thirty-two relevant primary studies
Alzheimer’s disease reporting on onset were identified; however, only those published between 2010 and 2012 (n = 65) were
Dementia included in the qualitative synthesis. Several factors including statins, light-to-moderate alcohol
Etiology
consumption, compliance with a Mediterranean diet, higher educational attainment, physically and
Systematic review
cognitively stimulating activities, and APOE e2 appeared to be associated with a decreased risk of AD
Onset
Progression onset. The evidence was suggestive of an increased risk of AD associated with head injury in males, age,
Risk factors diabetes mellitus, conjugated equine estrogen use with medroxyprogesterone acetate, current smoking,
and lower social engagement. With respect to genetic factors, APOE e4 remained the strongest predictor
of AD. Physical and cognitive activities were associated with a beneficial effect on cognitive function and
other indicators of dementia progression while higher educational attainment was associated with faster
cognitive decline. Although suggestive of an association, the current evidence for a majority of the
identified putative factors for AD onset and progression was weak, at best due to conflicting findings
across studies or inadequate evidence. Further research is required to confirm the etiological or
protective role of a number of risk factors.
© 2017 Published by Elsevier B.V.

1. Introduction
Alzheimer’s disease (AD) is the most common form of dementia
(Lindsay and Anderson, 2003). Dementia is characteristic of a
number of conditions including, but not limited to, AD (Public
Health Agency of Canada, 2010; National Advisory Council on
Aging, 2004). AD involves irreparable neuronal degeneration
which gives rise to a myriad of clinically detectable neurological
impairments (Parihar and Hemnani, 2004; Gauthier et al., 2005).
* Corresponding author. Present address: Ottawa Hospital Research Institute, The
Ottawa Hospital, Ottawa, ON, Canada.
E-mail address: mhersi@ohri.ca (M. Hersi).
Short- and long-term memory loss represent the most pronounced
cognitive marker of AD, with non-cognitive indicators encompass-
ing declining physical capacity and alterations in behaviour
(Gauthier et al., 2005; Aglukkaq, 2011). Characterized by debili-
tating symptoms which can dramatically impact quality of life, AD
usually results in death within a decade post- diagnosis (Aglukkaq,
2011). Data suggests that one in ten elderly Canadians had
dementia in 2011 with 60–70% of these cases attributed to AD
(Aglukkaq, 2011). Compared to AD occurrence in 2008, a more than
two-fold increase in AD incidence and prevalence is anticipated by
2038 (Alzheimer Society, 2010).
Despite the abundance of AD research, the etiology of this
neurological disease is not yet understood (National Advisory
Council on Aging, 2004). The objective of the study was to

http://dx.doi.org/10.1016/j.neuro.2017.03.006
0161-813X/© 2017 Published by Elsevier B.V.
144 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
systematically search the current scientific literature to identify
risk factors associated with the onset and progression of AD.
2. Methods
This review was conducted in accordance to the methodological
details outlined in the accompanying methods paper (Hersi et al.,
2016). The systematic review of risk factors associated with disease
onset involved two phases: (1) an umbrella review of systematic
reviews (i.e., a systematic search of the literature to identify
existing systematic reviews); (2) a systematic search of the
literature to identify primary studies. For the systematic review of
factors associated with disease progression, only an umbrella
review of systematic reviews was conducted.
2.1. Search methods
Separate search strategies were designed to identify: (1)
systematic reviews reporting on risk factors for disease onset;
(2) systematic reviews reporting on risk factors for disease
progression; and (3) primary studies reporting on risk factors
for disease onset.
Systematic reviews reporting on onset-related factors were
identified through a search of electronic databases and grey
literature sources including MEDLINE and MEDLINE In-Process &
Other Non-Indexed Citations, PubMed, EMBASE, CINAHL, HuGE-
NET, PsychINFO, TOXNET Toxicology Data Network, AARP Ageline,
Cochrane Database of Systematic Reviews, Database of Abstracts of
Reviews of Effects (DARE), AlzGene, ProQuest Digital Dissertations
and Theses database, Google and GoogleScholar. Disease-specific
journals including the American Journal of Alzheimer’s Disease
and Other Dementias (1986-December 31, 2011) and the Interna-
tional Journal of Alzheimer’s Disease (2009-December 31, 2011)
were also searched. Where possible, database specific search
filters, such as those developed by InterTASC Information Special-
ists’ Sub-Group and McMaster University’s Health Information
Research Unit, were also applied to identify systematic reviews and
citations reporting on risk factors associated with onset or
progression. The reference lists of a sample of retrieved articles
were reviewed for potentially missed studies. The initial search
was conducted in September 2011. The OVID AutoAlerts feature
was enabled to ensure that newly indexed relevant articles were
identified. The AutoAlerts feature was disabled on December 31,
2011. Databases without an alerts feature were re-searched until
the end of December 2011. The search for systematic reviews on AD
progression was restricted to MEDLINE and EMBASE and was
conducted in March 2013.
Primary studies reporting on risk factors associated with AD
onset were identified through a search of electronic databases
including MEDLINE and MEDLINE In-Process & Other Non-Indexed
Citations, EMBASE, CINAHL and AgeLine. The search was conducted
in July 2012. The search for risk factors for AD progression was
limited to systematic reviews. All search strategies are provided in
Supplementary material I.
2.2. Eligibility criteria
For the first phase of the review, systematic reviews reporting
on risk factors associated with AD onset or progression were
eligible for inclusion. French and English language reviews were
eligible if a computerized database search was conducted and the
eligibility criteria of the review were explicitly stated. Reviews
which failed to distinguish AD from all-cause dementia or
dementia subtypes were excluded for the systematic review of
disease onset. An exception was made for reviews reporting on
progression. Systematic reviews examining non-pharmacological
factors impacting cognitive and non-cognitive (such as behaviour,
physical function, quality of life, mortality) indicators of AD
progression were considered.
Primary studies were eligible for inclusion if they reported on
non-genetic risk factors associated with the development of AD.
Only studies which diagnosed AD using neuropathologic exami-
nation or according to the Diagnostic and Statistical Manual of
Mental Disorders (DSM), the Neurological Disorders and Stroke-
Alzheimer Disease and Related Disorders Association (NINCDS-
ADRDA), or the International Classification of Diseases (ICD)
criteria were considered. Studies reporting solely on non-AD
dementias as well as those failing to distinguish AD from all-cause
dementia or dementia subtypes were excluded. Studies reporting
on the effect of psychological stress were excluded from the
present report and have been reviewed elsewhere (Hersi, 2015).
2.3. Study screening and data extraction
Studies were screened according to the process described by
Hersi et al. (2016). For systematic reviews reporting on disease
onset, data were extracted by one reviewer. A second reviewer
independently extracted data from a randomly selected 10%
sample of reviews. Disagreements were addressed through
consensus. Due to limited resources, data from primary studies
and systematic reviews reporting on AD progression was extracted
by a single reviewer only.
2.4. Quality appraisal
The quality of systematic reviews was assessed according to the
AMSTAR criteria (Shea et al., 2009). For systematic reviews
reporting on AD onset, quality assessment was conducted by
one reviewer. A second reviewer independently assessed the
quality of a randomly selected 10% sample of reviews. Quality
assessment of reviews reporting on AD progression was conducted
entirely by a single reviewer. Low quality reviews with an AMSTAR
score of 3 or less were excluded from the qualitative synthesis. The
quality of primary studies was not evaluated.
2.5. Literature search results
The study selection process is outlined in Figs. 1–3 . A total of
136 systematic reviews reporting on risk factors associated with
disease onset (n = 121) and progression (n = 15) were identified. Of
these, 93 were deemed to be of high quality and were included in
the qualitative syntheses for disease onset (n = 81) and progression
(n = 12). A list of excluded systematic reviews on onset and
progression, along with reasons for exclusion, can be found in
Supplementary material I. One moderate quality review (Daviglus
et al., 2011) was later excluded from the overview of AD onset as it
was a manuscript version of an identified Health Technology
Assessment (HTA) (Williams et al., 2010). Two reviews were also
later excluded from the overview of AD progression (Isaac et al.,
2008; Aguirre et al., 2013). The review by Isaac et al. (2008) was
excluded because an updated version (Farina et al., 2012) of the
review was identified. The review by Aguirre et al. (2013) was
excluded as it was a duplicate of a Cochrane review (Woods et al.,
2012). Details regarding the quality appraisal of each systematic
review on onset and progression are also provided in Supplemen-
tary material I.
A total of 432 relevant primary studies reporting on disease
onset were identified. The breadth of the primary study literature
precluded inclusion of all 432 studies. As such, only the most
recent studies (those published from 2010 to 2012) were selected
for inclusion in the qualitative synthesis (n = 65). A list of included
primary studies is provided in Supplementary material I. Five
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
studies, which examined the risk of AD associated with stress,
were excluded from the present review and were reviewed
separately (Hersi, 2015).
3. Results
3.1. Risk factors associated with AD onset
3.1.1. Diabetes mellitus
Eight moderate quality systematic reviews reported on the
association between diabetes mellitus (DM) and risk of AD
(Cukierman et al., 2005; Biessels et al., 2006; Weih et al., 2007;
Patterson et al., 2007; Kloppenborg et al., 2008; Lu et al., 2009;
Profenno et al., 2010; Williams et al., 2010). Table 1 in
Supplementary material II provides an overview of the character-
istics of these studies and their findings.
Our search for recent primary studies yielded eleven studies
which reported on diabetes or plasma glucose levels (Ronnemaa
et al., 2011; Cheng et al., 2011; Ohara et al., 2011; Kimm et al., 2011;
Song et al., 2011; Verdelho et al., 2010; N.C. Li et al., 2010a,b; Qiu
et al., 2010; Reitz et al., 2010b; Schrijvers et al., 2010; Ahtiluoto
et al., 2010). Table 90 in Supplementary material II provides an
overview of the characteristics of these studies and their findings.
3.1.1.1. Systematic reviews. The earliest systematic review
reporting on DM identified six longitudinal studies (Cukierman
et al., 2005). All six studies reported an increased risk of AD among
diabetics, however, only results from two studies reached
statistical significance. The reviewers concluded that diabetes
appears to increase risk of dementia and AD.
The systematic review reported by Biessels et al. (2006)
identified 13 longitudinal cohort studies, 11 of which examined the
association between late-life DM and risk of AD. Nine of the eleven
studies examining late-life DM reported risk estimates exceeding
1.00. Furthermore, results from six of these studies reached
statistical significance with reported risk estimates ranging from
1.4 to 2.4. Of the two studies reporting on midlife DM ( < 60 years),
one detected a statistically significant increased risk of AD (odds
ratio (OR) = 4.40; p < 0.01) while the other failed to detect an
association (relative risk (RR) = 1.00; 95% CI = 0.50, 2.00). Based on
these findings, the reviewers concluded that diabetes increases
risk of AD.
Seven relevant cohort studies were included in the review
reported by Weih et al. (2007). Three studies consistently reported
a statistically significant increased risk of AD among subjects with
diabetes. Four of the remaining studies reported risk estimates
exceeding 1.00, however, the results failed to achieve statistical
significance. The reviewers concluded that diabetes may increase
risk of AD.
Patterson et al. (2007) identified two pertinent studies. Both
studies reported risk estimates greater than 1.00, although, both
failed to reach statistical significance.
Six of the 11 longitudinal studies identified by Kloppenborg
et al. (2008) reported a significant or marginally significant
increased risk of AD associated with DM. Two studies reported on
the effect of impaired glucose metabolism. The studies reported
conflicting results with one reporting an increased risk of AD
attributed to impaired glucose metabolism while the other failed
to detect an association between abnormal glucose levels and AD.
The reviewers concluded that diabetes was associated with an
increased risk of AD, vascular dementia, and all-cause dementia.
A total of eight prospective cohort studies investigating the
association between diabetes and Alzheimer's disease were
eligible for inclusion in the systematic review and meta-analysis
reported by Lu et al. (2009). Seven of the eight identified studies
reported risk estimates exceeding 1.00 with results from two of
145
these studies reaching statistical significance. The pooled analysis
detected a statistically significant 39% increased risk of AD among
diabetics (RR = 1.39; 95% CI = 1.17, 1.66). Results of the review and
meta-analysis suggest that individuals with diabetes have a
significantly higher risk of developing Alzheimer's disease.
Profenno et al. (2010) identified eight longitudinal studies
which reported on the association between DM and risk of AD. Two
additional studies, which reported on the effect of hyperinsuline-
mia without diabetes and borderline diabetes, were also included.
Three of the eight studies reporting on DM detected a statistically
significant association. The remaining five studies also found that
risk of AD was higher among diabetics, however, the results failed
to reach statistical significance. The random effects meta-analysis
detected a statistically significant 54% increased risk of AD (effect
size (ES) = 1.54; 95% CI = 1.33, 1.79; I2 = 3.6%, p = 0.405) associated
with DM. When all ten studies were combined, DM, hyper-
insulinemia, and borderline diabetes were associated with a
statistically significant increased risk of AD (ES = 1.61; 95% CI = 1.40,
1.86). Results of the subgroup analysis restricted to studies (n = 5)
which adjusted for the potentially confounding effects of body
mass index (BMI), other vascular risk factors, and stroke yielded
consistent findings (ES = 1.64; 95% CI = 1.29, 2.09). Overall, results of
the review suggest that diabetes was associated with an increased
risk of AD.
Two systematic reviews (Biessels et al., 2006; Lu et al., 2009)
and two cohort studies were eligible for inclusion in the HTA
prepared for the US Agency for Healthcare Research and Quality
(Williams et al., 2010). Results from both systematic reviews
included in the report suggest that DM increases risk of AD. One of
the two identified cohort studies reported an increased risk of AD
associated with DM, however, results failed to reach statistical
significance (hazard ratio (HR) = 1.62; 95% CI = 0.98, 2.67). While
the second study failed to detect an association between diagnosed
diabetes and AD, both borderline (HR = 1.87; 95% CI = 1.11, 3.14) and
undiagnosed diabetes (HR = 3.29; 95% CI = 1.20, 9.01) were
associated with a statistically significant increased risk of AD.
Based on the available evidence, the reviewers concluded that DM
appears to increase the risk of AD.
3.1.1.2. Primary studies. Diabetes was not significantly associated
with risk of AD according to results of the LADIS prospective cohort
study (HR = 1.34; 95% CI = 0.35, 5.14; p = 0.668) and the Canadian
Study of Health and Aging (OR = 0.64; 95% CI = 0.39, 1.03) (Verdelho
et al., 2010; Song et al., 2011). Reitz et al. (2010b) also failed to
detect an association between DM and risk of late-onset AD
(LOAD).
Six studies detected a significant or marginally significant
association between DM and AD. Diabetics were at a significantly
increased risk of developing AD according to the Vantaa 85+ study
(HR = 2.45; 95% CI = 1.33, 4.53), Kungsholmen project (HR = 1.52;
95% CI = 1.03, 2.25), and the Hisayama Study (HR = 2.05; 95%
CI = 1.18, 3.57) (Ahtiluoto et al., 2010; Qiu et al., 2010; Ohara et al.,
2011). Participants of the Washington Heights-Inwood Columbia
Aging Project (Cheng et al., 2011) reporting a history of type 2
diabetes had an 80% increased risk of LOAD according to the
minimally adjusted model (HR = 1.80; 95% CI = 1.20, 2.70). Kimm
et al. (2011) detected an increased risk of AD among male
(HR = 1.60; 95% CI = 1.30, 2.00) and female (HR = 1.40; 95% CI = 1.10,
1.70) diabetics. Finally, diabetes was only marginally associated
with an increased risk of incident AD according to results reported
by Li et al. (2010a,b) (HR = 1.04; 95% CI = 1.00, 1.08; p = 0.049).
Ohara et al. (2011) utilized data from the Hisayama Study, a
longitudinal cohort study of Japanese subjects, to examine the
association between glucose tolerance and risk of AD and other
cognitive outcomes. The study detected an increased risk of AD
associated with higher levels of 2-h post-load glucose but not with
146 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Fig. 1. Screening process for systematic reviews reporting on risk factors for AD onset.
fasting plasma glucose or the remaining individual categories of
glucose tolerance (Table 90, Supplementary material II).
Higher levels of insulin and insulin resistance, but not glucose,
were associated with an increased risk of AD within three years of
follow-up in the Rotterdam Study (Schrijvers et al., 2010).
However, insulin and insulin resistance were not associated with
risk of AD after three years of follow-up. Fasting glucose remained
not associated with AD at subsequent follow-up periods.
The Uppsala Longitudinal Study of Adult Men (Ronnemaa et al.,
2011) failed to detect an association between higher levels of
fasting plasma glucose, ascertained in mid- and late-life, and
development of AD (Midlife
HR = 1.00; 95% CI = 0.80, 1.20. Late-life

HR = 1.00; 95% CI = 0.80, 1.30).
3.1.1.3. Discussion. Evidence summarized in eight moderate-
quality systematic reviews is suggestive of an association
between DM and risk of AD. Although a majority of the studies
reported an increased risk of AD associated with AD, results from
some studies failed to reach statistical significance. Inconsistent
findings across studies may be attributed to sources of
methodological heterogeneity with respect to exposure
ascertainment (Biessels et al., 2006; Williams et al., 2010). More
specifically, studies varied with respect to the methods used to
ascertain diabetes as well as the timing of the exposure
measurement (i.e., midlife vs. late-life) (Biessels et al., 2006;
Williams et al., 2010).
We identified nine primary studies which reported on diabetes
mellitus and three studies which reported on the effect of high
fasting glucose levels. Six (Ahtiluoto et al., 2010; Qiu et al., 2010;
Ohara et al., 2011; Cheng et al., 2011; Kimm et al., 2011; Li et al.,
2010a,b) of the nine studies reporting on DM detected a significant
or marginally significant association with risk of AD, although
results failed to reach statistical significance in the fully adjusted
model in one study (Cheng et al., 2011). Of the studies reporting on
higher fasting plasma glucose levels, all three failed to detect an
association with AD (Ohara et al., 2011; Schrijvers et al., 2010;
Ronnemaa et al., 2011). However, one study (Ohara et al., 2011)
detected an increased risk of AD associated with higher levels of 2-
h post-load glucose and one (Schrijvers et al., 2010) reported that
elevated insulin and insulin resistance were associated with an
increased short- but not long-term risk of AD.
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Fig. 2. Screening process for primary studies reporting on risk factors for AD onset.
3.1.2. Obesity, body mass index, body composition
Our search for systematic reviews yielded eight moderate
quality (Weih et al., 2007; Kloppenborg et al., 2008; Profenno et al.,
2010; Williams et al., 2010; Beydoun et al., 2008; Lee et al., 2010;
Anstey et al., 2011; Gorospe and Dave, 2007) and three low quality
reviews (Pitner, 2005; Chang, 2008; Barnes and Yaffe, 2011) that
reported on the association between obesity and body mass index
(BMI) with risk of AD.
Six primary studies were identified in the second phase of the
review (Chen et al., 2010; Reitz et al., 2010b; Ronnemaa et al., 2011;
Xu et al., 2011; Abellan van Kan et al., 2012; Gelber et al., 2012). The
study characteristics and findings are summarized in Tables 2 and
92 in Supplementary material II.
3.1.2.1. Systematic reviews. The evidence summarized in the
systematic review reported by Weih et al. (2007) is suggestive
of an increased risk of AD attributable to obesity, overweight BMI,
and increasing BMI. Two of three studies, which examined the risk
of AD associated with each unit increase in BMI, reported risk
estimates above 1.00. One study detected an increased risk of AD
associated with obesity (BMI >30 kg/m2) and being overweight
(BMI 25–30 kg/m2).
Gorospe and Dave (2007) identified a single study which
specifically reported on the risk of developing AD. The study
reported no association between AD and midlife BMI.
Kloppenborg et al. (2008) identified five relevant longitudinal
studies. There was no association between midlife and late-life
obesity with risk of developing AD according to two studies. Three
studies examined late-life BMI as a continuous variable, with
conflicting findings. One study found that each unit increase in BMI
was associated with an increased risk of AD, one reported that
increasing BMI was protective against the development of AD, and
one study failed to detect an association.
Lee et al. (2010) identified four longitudinal studies that
reported on the effect of obesity, overweight BMI, and increasing
BMI on AD risk. One of two studies, which reported on the effect of
high BMI, detected a statistically significant increased risk of AD
among obese and overweight subjects. Of the two studies
147
reporting on BMI treated as a continuous variable, only one
reported a statistically significant increased risk of AD with
increasing BMI.
The HTA reported by Williams et al. (2010) identified one
systematic review and three recent cohort studies. Results from
the systematic review (Beydoun et al., 2008) suggested an
increased risk of AD associated with obesity. The three cohort
studies reported a decreased risk of AD associated with late-life
obesity or overweight BMI. Considered cumulatively, however, the
reviewers concluded that obesity did not appear to be associated
with risk of AD.
Three systematic reviews reported a quantitative synthesis of
identified evidence. According to the pooled analysis (n = 4)
reported by Beydoun et al. (2008), obesity was associated with
a marginally significant increased risk of AD (RR = 1.80; 95%
CI = 1.00, 3.29). The meta-analysis of six longitudinal studies
reported by Profenno et al. (2010) yielded a risk estimate of 1.59
(95% CI = 1.02, 2.48; I2 = 65.5%, p = 0.005). The series of pooled
analyses reported by Anstey et al. (2011) detected an increased risk
of AD associated with low (RR = 1.96; 95% CI = 1.32, 2.92; I2 = 69.1%),
overweight (RR = 1.35; 95% CI = 1.19, 1.54; I2 = 92.0%), and obese
(RR = 2.04; 95% CI = 1.59, 2.62; I2 = 82.8%) midlife BMI. Analyses of
studies reporting on late-life BMI, treated categorically (RR = 1.46;
95% CI = 0.97, 2.21; I2 = 42.3%) and continuously (RR = 0.97; 95%
CI = 0.92, 1.02; I2 = 75.8%) failed to detect an association with AD.
Anstey et al. (2011) concluded that underweight, overweight, and
obese midlife BMI were all associated with an increased risk of AD.
In contrast, late-life BMI and obesity were not associated with AD.
3.1.2.2. Primary studies. Four studies reported on the effect of BMI
on risk of AD. While low (<18.5 kg/m2) BMI was consistently
associated with an increased risk of AD among males and females
across the lifespan, there was no evidence of an association
between overweight/obese (24 kg/m2) BMI and the risk of AD in
the case-control study reported by Chen et al. (2010). Similarly, the
Honolulu-Asia Aging Study and the Uppsala Longitudinal Study of
Adult Men failed to detect an association, between AD and higher
tertiles of midlife BMI and increasing mid- and late-life BMI
148 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Fig. 3. Screening process for systematic reviews reporting on risk factors for AD progression.
respectively (Gelber et al., 2012; Ronnemaa et al., 2011). The
remaining study, based on the Swedish Twin Registry (Xu et al.,
2011), reported an increased risk of AD associated with both
overweight (25–30 kg/m2 OR = 1.91; 95% CI = 1.30, 2.80) and obese
( > 30 kg/m2 OR = 3.43; 95% CI = 1.49, 7.90) midlife BMI. When
treated as a continuous variable, BMI was associated with an
increased risk of AD (OR = 1.06; 95% CI = 1.01, 1.10).
The EPIDOS-Toulouse (Abellan van Kan et al., 2012) cohort
study found no association between total lean mass (OR = 0.57; 95%
CI = 0.24, 1.34) or total fat mass (OR = 2.12; 95% CI = 0.90, 4.99) and
AD. High waist to hip ratio (WHR) was significantly associated with
the risk of probable and possible late-onset AD (LOAD) according to
results reported by Reitz et al. (2010b) (HR = 2.63; 95% CI = 1.39,
4.97). Results were similar when the analysis was restricted to
probable LOAD (HR = 3.12; 95% CI = 1.54, 6.33).
3.1.2.3. Discussion. Evidence summarized in eight moderate-
quality systematic reviews suggests that the association
between obesity or higher BMI and risk of AD is inconclusive.
While some reviews reported evidence of an increased risk of AD
attributed to obesity or higher BMI in midlife, others concluded
that there was no association. The evidence pertaining to late-life
obesity and risk of AD was also conflicting. While recent cohort
studies identified by Williams et al. (2010) were suggestive of a
decreased risk of AD associated with higher late-life BMI, the
pooled analysis reported in the most recent review (Anstey et al.,
2011) failed to detect an association. However, the meta-analysis
reported by Anstey et al. (2011) was limited by a small number of
identified studies.
Of the four observational studies reporting on BMI identified in
the second phase of our review, one examined BMI in early life
(Chen et al., 2010), four ascertained BMI in midlife (Ronnemaa
et al., 2011; Gelber et al., 2011; Xu et al., 2011; Chen et al., 2010) and
two examined late-life BMI (Ronnemaa et al., 2011; Chen et al.,
2010). Only one (Xu et al., 2011) of the four studies reporting on
midlife BMI detected an association between higher or increasing
BMI and increased risk of AD. Another study (Chen et al., 2010)
detected an increased risk of AD associated with low BMI in early,
mid-, and late-life. Both studies reporting on the effect of high or
increasing late-life BMI failed to detect an association. Finally,
while higher waist to hip ratio predicted risk of AD in one study
(Reitz et al., 2010b), late-life total lean mass and total fat mass were
not associated with risk of AD in another (Abellan van Kan et al.,
2012).
3.1.3. Hypertension
Six systematic reviews were identified. Two reviews were
excluded due to low methodological quality (Purnell et al., 2009;
Barnes and Yaffe, 2011). Results from one high-quality (Power et al.,
2011) and three moderate-quality reviews (Patterson et al., 2007;
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Kloppenborg et al., 2008; Williams et al., 2010) were included in
the qualitative synthesis. In the second phase of the review, we
identified nine cohort studies (Ronnemaa et al., 2011; Kimm et al.,
2011; Song et al., 2011; Verdelho et al., 2010; Qiu et al., 2010; Reitz
et al., 2010b; Yang et al., 2011; Joas et al., 2012; Chu et al., 2010). The
study characteristics and findings are summarized in Tables 3 and
91 in the Supplementary material II.
3.1.3.1. Systematic reviews. Five longitudinal cohort studies
reporting on the effect of blood pressure on AD risk were
eligible for inclusion in the review reported by Patterson et al.
(2007). Two studies reported an association between systolic
hypertension, defined as systolic blood pressure (SBP) greater than
180 mm Hg, and an increased risk of AD. Two studies detected an
increased risk of AD attributed to low (<140 mm Hg) or declining
systolic blood pressure. Low diastolic blood pressure (DBP;
<65 mm Hg) was also detrimental according to one study.
Finally, a single study reported an increased risk of AD
associated with both high (>84 mm Hg) and low (< 70 mm Hg)
pulse pressure (PP). Overall, results from this review suggest that
high and low systolic blood pressure, low diastolic pressure, and
both low and high pulse pressure may increase risk of AD.
Kloppenborg et al. (2008) identified twelve studies examining
the risk of AD associated with blood pressure. All four studies
reporting on the risk of AD associated with hypertension failed to
detect an association. Of the five studies reporting on the effect of
high SBP, one detected an increased risk, one reported a protective
effect, and three failed to detect an association. Four of five studies
reporting on high DBP failed to detect an association. The
remaining study reported a greater than four-fold increase in
the risk of AD among those with high DPB. Overall, four of 12
included longitudinal studies detected a statistically significant
association between measures of blood pressure and risk of AD.
Three of the four studies reported an increased risk of AD
associated with high or increasing blood pressure while the
remaining study detected a protective effect.
Of the ten cohort studies included in the HTA reported by
Williams et al. (2010), results from only three studies, which
examined two cohorts, detected a statistically significant increased
risk of AD associated with hypertension. More specifically, two of
seven studies detected an increased risk of AD associated with high
SBP. With respect to diastolic blood pressure, only one of seven
studies reported an increased risk of AD associated with high DBP.
The remaining studies failed to detect an association. The
reviewers concluded that the evidence was suggestive of a lack
of association between hypertension and AD.
Eighteen studies were included in the review reported by Power
et al. (2011). History of hypertension was not associated with the
risk of AD according to the random effects meta-analysis of ten
studies (RR = 0.98; 95% CI = 0.80, 1.19; I2 = 42%; 95% CI = 0%, 71%).
According to the pooled analysis of four studies, a 10 mm HG-
increase in SBP was associated with a marginally significant 5%
reduced risk of AD (RR = 0.95; 95% CI = 0.91, 1.00. I2 = 0%; 95% CI = 0%,
68%). A 10 mm HG-increase in DBP was not significantly associated
with risk of AD (RR = 0.94; 95% CI = 0.85, 1.04. I2 = 14%; 95% CI = 0%,
72%). When the analysis was restricted to the three studies that
reported on increasing late-life blood pressure, results remained
unchanged. The only study to report on increasing midlife blood
pressure failed to detect an association between AD and a 10 mm
HG increase in both SBP (RR = 1.03; 95% CI = 0.80, 1.32) and DBP
(RR = 1.16; 95% CI = 0.75; 1.81). Among studies that examined the
effect of blood pressure treated as a categorical variable, two of
three failed to detect an association between midlife SBP and risk
of AD. The remaining study detected an increased risk of AD
associated with the highest category of SBP (160 mm Hg)
(RR = 2.60; 95% CI = 1.10, 6.60). All three studies reported an
149
increased risk of AD associated with elevated midlife DBP,
however, results failed to reach statistical significance. Three
studies reported on the effect of late-life SBP and DBP. The
reviewers noted the lack of association between categories of SBP
measured in late-life and AD. There was some evidence of an
association between higher DBP and a reduced risk of AD. Overall,
results of the review suggested that the association between BP
and risk of AD remained inconclusive (Power et al., 2011).
3.1.3.2. Primary studies. According to results of the LADIS
prospective cohort study (HR = 0.71; 95% CI = 0.27, 1.90;
p = 0.499) and the Canadian Study of Health and Aging
(OR = 1.00; 95% CI = 0.75, 1.32), hypertension was not associated
with risk of AD (Verdelho et al., 2010; Song, Mitnitski, and
Rockwood, 2011). Similarly, Reitz et al. (2010b) failed to detect an
association between hypertension and risk of possible and
probable LOAD (HR = 1.16; 95% CI = 0.64, 2.11). Results remained
consistent when the analysis was restricted to probable LOAD
(HR = 1.03; 95% CI = 0.54, 1.95). Self-reported hypertension was not
associated with risk of AD according to results reported by Yang
et al. (2011) (HR = 0.82; 95% CI = 0.58, 1.15; P = 0.251). The only study
to detect an association between hypertension and risk of AD
reported a detrimental effect associated with stage 2 hypertension
(HR = 1.40; 95% CI = 1.10, 1.80) among Korean men (Kimm et al.,
2011). Both pre-hypertension (120–139 mmHg systolic and/or 80–
89 mmHg diastolic) and stage 1 hypertension (140–159 mmHg
systolic/90–99 mmHg diastolic), defined according to the Joint
National Committee (JNC 7) criteria, were not associated with AD
in men. Alternatively, all three categories of hypertension were not
associated with risk of AD in the sample of Korean women (see
Table 91 in the Supplementary material II).
Midlife and late-life SBP, treated as a continuous variable, were
not associated with risk of AD in the Uppsala Longitudinal Study of
Adult Men (Ronnemaa et al., 2011). Yang et al. (2011) also failed to
detect an association between SBP and risk of AD in a longitudinal
analysis (HR = 1.00; 95% CI = 0.99, 1.00; P = 0.294). In contrast, two
studies detected a detrimental effect attributed to higher SBP. Each
unit increase in SBP was associated with an increased risk of AD
(RR = 1.04; 95% CI = 1.01, 1.08) in an all-male sample of Chinese
subjects (Chu et al., 2010). Furthermore, the Kungsholmen project
(Qiu et al., 2010) detected an increased risk of AD attributed to high
SBP (140–159 mm Hg
HR = 1.47; 95% CI = 1.02, 2.12. 160 mm Hg

HR = 1.84; 95% CI = 1.06, 3.18).
The effect of DBP was examined in two studies. Using moderate
diastolic pressure (70–94 mm Hg) as a referent, low ( < 70 mm Hg

HR = 1.83; 95% CI = 1.26, 2.65) but not high (95 mm Hg

HR = 0.78; 95% CI = 0.54, 1.12) DBP was significantly associated with
AD risk in the Kungsholmen project (Qiu et al., 2010). Higher DBP in
late-life was marginally associated with risk of AD (HR = 0.98; 95%
CI = 0.97, 1.00; P = 0.030) according to the longitudinal analysis
reported in another study (Yang et al., 2011).
Results from the Kungsholmen project (Qiu et al., 2010)
suggested that low ( < 70 mm Hg
HR = 1.50; 95% CI = 1.07, 2.12)
but not high (85 mm Hg
HR = 1.19; 95% CI = 0.83, 1.69) pulse
pressure was significantly associated with AD when compared
with the middle tertile (70–84 mm Hg). The longitudinal analysis
presented by Yang et al. (2011) failed to detect an association
between pulse pressure (HR = 1.00; 95% CI = 0.99, 1.01; P = 0.919) or
mean arterial pressure (HR = 0.99; 95% CI = 0.98, 1.00; P = 0.060)
and risk of AD.
Using data from the Prospective Population Study of Women,
Joas et al. (2012) examined the association between blood pressure
“trajectories”, or changes in blood pressure from midlife to late-
life, and the development of dementia and AD. “All AD” included all
subjects diagnosed with AD according to the NINCDS-ADRDA
criteria while “pure AD” included subjects with “no
150 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
cerebrovascular or other disease that might contribute to
dementia” (Joas et al., 2012). There was no difference between
those who developed AD and those who did not with respect to
baseline SBP (Pure AD: 0.16 mm Hg; 95% CI =
3.82, 4.14; p = 0.94.
All AD: 2.69 mm Hg; 95% CI =
0.95, 6.33; p = 0.15). Compared to
those who did not develop dementia, there was a greater increase
in SBP between 1968 and 1992 for those who developed pure AD
(Pure AD: 0.32 mm Hg; 95% CI = 0.07, 0.56; p = 0.01. All AD: 0.20 mm
Hg (95% CI =
0.01, 0.42; p = 0.06) and a greater decrease from 1992
onwards for both “pure” and “all” AD.
3.1.3.3. Discussion. According to the results of four moderate/high
quality systematic reviews, the association between hypertension
and risk of AD is unclear. A majority of studies captured in the
reviews failed to detect an association between hypertension and
risk of AD. There is some evidence to suggest that the association
between hypertension and AD may be dependent on age (Power
et al., 2011). While some studies have reported an increased risk of
AD associated with midlife hypertension, results from other
studies have detected a protective effect of late-life hypertension
against the development of AD. However, the observed risk
reduction attributed to late-life hypertension is likely due to
sources of bias (Power et al., 2011). More specifically, selection bias
due to losses to follow-up, short lengths of follow-up, confounding,
and exposure/outcome misclassification error were all noted as
important limitations of primary studies included by Power et al.
(2011).
Of the five identified primary studies reporting on hyperten-
sion, only one (Kimm et al., 2011) detected an association with risk
of AD, namely, an increased risk among hypertensive males. Two
studies (Qiu et al., 2010; Chu et al., 2010) reported an increased risk
of AD associated with higher SBP, one (Yang et al., 2011) reported a
protective effect associated with higher DBP in late-life, one (Qiu
et al., 2010) reported a detrimental effect associated with low DBP,
one (Qiu et al., 2010) detected an increased risk associated with
low pulse pressure, and one (Joas et al., 2012) reported differences
in blood pressure trajectories from midlife to late-life among
subjects with AD as compared to those who did not develop
dementia.
3.1.4. Cholesterol, hypercholesterolemia
The effect of cholesterol levels was examined in seven
systematic reviews. Two reviews were of low methodological
quality and were excluded (Miller and Chacko, 2004; Purnell et al.,
2009). Findings from the remaining five moderate-quality reviews
were considered (Weih et al., 2007; Patterson et al., 2007;
Kloppenborg et al., 2008; Anstey et al., 2008; Williams et al.,
2010). The second phase of the review yielded five relevant cohort
studies (Reitz et al., 2010a,b; Mielke et al., 2010; Ronnemaa et al.,
2011; Kimm et al., 2011).
3.1.4.1. Systematic reviews. Eight cohort studies reporting on the
association between serum cholesterol and risk of AD were eligible
for inclusion in the review reported by Weih et al. (2007). Two
studies reported that serum cholesterol levels exceeding 6.5 mol/l
were associated with an approximate three-fold increased risk of
AD. However, conflicting results were reported in one study that
detected a decreased risk of AD associated with elevated serum
cholesterol. The remaining five studies failed to detect an
association between AD and elevated low-density lipoprotein
cholesterol (LDL-C) (n = 1) or total serum cholesterol (n = 4).
Both studies included in the review by Patterson et al. (2007)
detected an increased risk of AD attributed to high midlife serum
cholesterol. Based on these findings, the reviewers concluded that
elevated serum cholesterol in midlife may increase risk of AD.
A total of four longitudinal studies were eligible for inclusion in
the systematic review reported by Kloppenborg et al. (2008). The
three studies that reported risk estimates comparing high total
serum cholesterol to normal levels reported conflicting findings.
One study failed to detect an association, one study reported a
greater than two-fold increased risk of AD associated with high
total cholesterol, and the remaining study detected a reduced risk
of AD among those with high total cholesterol levels. Heterogene-
ity across studies, particularly pertaining to the timing of exposure
ascertainment, was cited as a likely source of the inconsistent
findings (Kloppenborg et al., 2008). The only study which
ascertained exposure at midlife reported an increased risk of AD
associated with high total cholesterol. Two studies reporting on
late-life total cholesterol failed to detect an association or reported
a decreased risk of AD associated with higher cholesterol levels.
The only study to report on total cholesterol considered continu-
ously, detected a 2% increased risk of AD associated with increasing
total cholesterol (OR = 1.02; p = 0.03). Lastly, both studies reporting
on triglycerides failed to detect an association.
Eight prospective cohort studies, four of which examined total
cholesterol measured at midlife and five of which examined the
effect of late-life total cholesterol (TC), were identified in the
review reported by Anstey et al. (2008). Of the four studies
examining midlife total cholesterol, two detected an increased risk
attributed to high (6.5 mmol/l) cholesterol, one reported a
detrimental effect of decreasing cholesterol levels from mid- to
late-life, and one failed to detect an association between increasing
(per 10 mg/dL) total cholesterol and risk of AD. Two of five studies
reporting on late-life TC failed to detect an association. Three of the
remaining studies, which examined the effect of increasing
quartiles of late-life TC, were combined in a fixed effects meta-
analysis. The pooled analysis failed to detect an association
between AD and increasing quartiles of late-life TC (Second
quartile vs. first quartile
RR = 0.85; 95% CI = 0.67, 1.10. Third
quartile vs. first quartile
RR = 1.03; 95% CI = 0.79, 1.35. Fourth
quartile vs. first quartile
RR = 0.85; 95% CI = 0.65, 1.12). Based on
the identified evidence, the reviewers concluded that while
elevated midlife TC was associated with increased risk of AD,
elevated late-life TC was not.
The HTA reported by Williams et al. (2010) relied solely on the
findings of the systematic review by Anstey et al. (2008). It was
concluded that elevated midlife TC, but not elevated late-life TC,
was associated with an increased risk of AD.
3.1.4.2. Primary studies. The prospective cohort study reported by
Reitz et al. (2010a) examined the risk of AD associated with total
cholesterol, high-density lipoprotein cholesterol (HDL-C), non-
high-density lipoprotein cholesterol (Non-HDL-C), and LDL-C. Risk
estimates were reported for probable AD and for the combined
outcome of possible and probable AD. According to the results of
the fully adjusted model, higher HDL-C was associated with a
decreased risk of probable and possible AD; however, only results
for the highest quartile reached statistical significance (Quartile 2

HR = 0.80; 95% CI = 0.40, 1.50. Quartile 3
HR = 1.10; 95% CI = 0.60,
1.90. Quartile 4
HR = 0.40; 95% CI = 0.20, 0.90). Analyses of higher
total cholesterol, non-HDL-C, and LDL-C levels yielded risk
estimates below 1.00; however, results failed to achieve
statistical significance. Similar results were obtained when
probable AD was analyzed separately.
Low HDL-C was not significantly associated with possible and
probable LOAD (HR = 1.60; 95% CI = 0.77, 3.32) according to Reitz
et al. (2010b). Results were similar when the analysis was
restricted to probable LOAD (HR = 1.58; 95% CI = 0.71, 3.54).
Midlife serum cholesterol was not associated with risk of AD
after 32 years of follow-up according to the Prospective Population
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Study of Women (>6.5 mmol/l vs. less: HR = 1.48; 95% CI = 0.73,
2.96) (Mielke et al., 2010).
The study reported by Ronnemaa et al. (2011) examined the risk
of AD associated with vascular risk factors considered individually
and in combination. Mid- and late-life serum cholesterol, treated
as a continuous variable, was not associated with risk of AD
(Midlife HR = 1.00; 95% CI = 0.90, 1.20. Late-life HR = 1.10; 95%
CI = 0.90, 1.30).
The association between vascular risk factors and risk of AD
among Koreans was prospectively examined by Kimm et al. (2011).
Using normal (<5.17 mmol/l) cholesterol as the referent category,
both borderline (HR = 1.20; 95% CI = 1.00, 1.40) and high (HR = 1.20;
95% CI = 1.00, 1.50) cholesterol were not significantly associated
with risk of AD in men. Similarly, borderline and high cholesterol
did not increase risk of AD in women (Borderline cholesterol
HR = 1.00; 95% CI = 0.90, 1.20. High cholesterol HR = 1.10; 95%
CI = 0.90, 1.30).
3.1.4.3. Discussion. Five moderate-quality systematic reviews
examining the effect of high serum cholesterol on the risk of AD
were identified. While two earlier systematic reviews concluded
that the association between cholesterol and AD was inconclusive,
conflicting findings across studies was likely due to heterogeneity
with respect to the timing of exposure ascertainment (Weih et al.,
2007; Kloppenborg et al., 2008). The systematic review by Anstey
et al. (2008), which stratified studies based on timing of exposure
ascertainment, concluded that elevated midlife total cholesterol,
but not late-life cholesterol, was associated with an increased risk
of AD. Findings of the identified systematic reviews were limited
by sources of bias in the included observational studies. Firstly, the
observational studies failed to control for the potentially
confounding effect of lipid lowering medications (Anstey et al.,
2008). Secondly, although a majority of the included studies
adjusted for other potentially confounding variables, variability in
the degree of confounding adjustment across the included studies
may have given rise to methodological heterogeneity (Anstey et al.,
2008). Overall, evidence summarized in the identified systematic
reviews suggested that elevated midlife total serum cholesterol
was associated with an increased risk of AD. However, the analysis
of midlife cholesterol reported by Anstey et al. (2008) should be
interpreted cautiously due to the small number of included studies.
With respect to the primary studies identified in the second
phase of the review, three (Ronnemaa et al., 2011; Reitz et al.,
2010a,b) reported on late-life serum cholesterol and two
ascertained cholesterol in mid-life (Ronnemaa et al., 2011; Mielke
et al., 2010). All four cohort studies reporting on mid- and late-life
total serum cholesterol failed to detect a statistically significant
association. Only one (Reitz et al., 2010a) of two (Reitz et al., 2010a,
b) studies, which specifically reported on the effect of late-life
HDL–C, detected an association with AD. When evidence from
systematic reviews and primary studies were considered cumula-
tively, the association between elevated serum cholesterol and risk
of AD was unclear.
3.1.5. Homocysteine
Seven systematic reviews were identified (Reutens and
Sachdev, 2002; Kuo et al., 2005a; Weih et al., 2007; Patterson
et al., 2007; Williams et al., 2010; Ho et al., 2011; Dangour et al.,
2010). Two systematic reviews were of low methodological quality
and were excluded (Reutens and Sachdev, 2002; Kuo et al., 2005a).
The search for primary studies yielded one record (Hooshmand
et al., 2010).
3.1.5.1. Systematic reviews. The systematic review reported by
Weih et al. (2007) identified a single cohort study which examined
the association between homocysteine and risk of AD. Results of
151
the study suggest that each 1 standard deviation (SD) increase in
log transformed serum homocysteine at baseline was associated
with risk of AD (RR: 1.80; 95% CI = 1.30, 2.50).
The single cohort study included in the review by Patterson
et al. (2007) detected a greater than two-fold (RR = 2.11; 95%
CI = 1.19, 3.76) increased risk of AD among those with serum
homocysteine levels greater than 15 mmol/L.
Four community-based cohort studies were identified in the
HTA by Williams et al. (2010). Three of the four included cohort
studies reported a statistically significant increased risk of AD
associated with elevated homocysteine levels. The remaining
study reported an increased risk associated with elevated
homocysteine, although results failed to reach statistical signifi-
cance. Results from the three studies that reported on the risk of
AD associated with plasma homocysteine levels exceeding 14 or
15 mmol/L were pooled. According to the findings of the random
effects meta-analysis, elevated plasma homocysteine levels were
not associated with a statistically significant increased risk of AD
(HR = 1.53; 95% CI = 0.94, 2.49; Q-statistic: 6.378, p = 0.04,
I2 = 68.6%). Results of the pooled analysis were limited by the
small number of eligible studies as well as statistical heterogeneity
across studies (Williams et al., 2010). Based on these findings, the
reviewers concluded that the available evidence did not support an
increased risk of AD associated with elevated homocysteine
concentrations.
Four cohort studies were identified and included in the review
reported by Dangour et al. (2010). Three of the four included
studies reported an increased risk of AD associated with elevated
homocysteine levels. The remaining study failed to detect an
association between elevated plasma homocysteine concentra-
tions and risk of AD. Based on the findings of this review, the
reviewers concluded that there was some evidence to suggest that
there is an increased risk of AD associated with elevated serum/
plasma homocysteine concentrations.
A total of 14 studies identified by Ho et al. (2011) investigated
the association between homocysteine levels and AD. Data from 12
studies were combined in a random effects meta-analysis. The
quantitative analysis yielded a pooled standardized mean differ-
ence of 0.59 (95% CI = 0.38, 0.80; I2 = 79.1%) suggesting that subjects
with AD had significantly higher homocysteine levels compared to
controls. Four studies identified in the review examined the
association between elevated homocysteine and subsequent risk
of dementia and cognitive decline. However, only one of these
prospective studies focused on the outcome of AD. Results of this
study suggest that elevated homocysteine (>15 mmol/L) was not
associated with a statistically significant increased risk of AD
(OR = 1.60; 95% CI = 0.70, 3.50). Overall, results of this systematic
review and meta-analysis suggest that in comparison to controls,
subjects with AD had significantly higher homocysteine levels. The
only identified prospective study failed to detect an association.
3.1.5.2. Primary studies. Results from the Cardiovascular Risk
Factors, Aging, and Dementia (CAIDE) study suggest that both
increasing homocysteine (OR per 1 mmol/L = 1.19; 95% CI = 1.01,
1.39) and holotranscobalamin (OR per 1 pmol/l: 0.977 = 95%
CI = 0.958, 0.997) were marginally associated with AD
(Hooshmand et al., 2010). However, homocysteine was no
longer significantly associated with AD when the analysis was
adjusted for holotranscobalamin (OR = 1.10; 95% CI = 0.96, 1.25). The
authors concluded that the findings were suggestive of a possible
association between homocysteine and holotranscobalamin with
risk of AD.
3.1.5.3. Discussion. Five systematic reviews, collating evidence
from six cohort (Blasko et al., 2008; Luchsinger et al., 2004;
Ravaglia et al., 2005; Seshadri et al., 2002; Annerbo et al., 2006;
152 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Kim et al., 2008), six case-control (Clarke et al., 1998; Koseoglu and
Karaman, 2007; Bottiglieri et al., 2001; Religa et al., 2003;
Näggaet al., 2003; Hogervorst et al., 2002), and five cross-
sectional (Folin et al., 2005; Haan et al., 2007; Quadri et al.,
2004; Storey et al., 2003; Miller et al., 2002) studies were identified
in the present review. Four of the six cohort studies reported an
increased risk of AD associated with elevated homocysteine levels
(Blasko et al., 2008; Ravaglia et al., 2005; Seshadri et al., 2002;
Annerbo et al., 2006). The remaining two cohort studies also
reported risk estimates exceeding 1.00, although the results failed
to reach statistical significance (Luchsinger et al., 2004; Kim et al.,
2008). The only primary study (Hooshmand et al., 2010) identified
in the second phase of the review detected a marginally significant
association between homocysteine and risk of AD. However,
results failed to reach statistical significance upon adjustment for
holotranscobalamin.
Collectively, there is some evidence to suggest that elevated
homocysteine levels may increase the risk of AD. However,
conflicting findings across primary studies and reviews precludes
a definitive statement. The systematic reviews were limited by the
potential sources of bias affecting the included observational
studies. Firstly, a number of observational studies failed to adjust
for factors which may confound the association between
homocysteine and AD, namely, vitamin B supplementation,
smoking status, inactive lifestyle, increasing age, high coffee and
alcohol consumption, and medical conditions including hypothy-
roidism and rheumatoid arthritis (Williams et al., 2010; Ho et al.,
2011; Kuo et al., 2005a). Secondly, a majority of the identified
observational studies were case-control or cross-sectional studies,
which are susceptible to reverse causality (Ho et al., 2011).
Considered cumulatively, the association between elevated
homocysteine and risk of AD was inconclusive.
3.1.6. Metabolic syndrome
The effect of metabolic syndrome on risk of AD was examined in
two moderate-quality systematic reviews (Williams et al., 2010;
Hao et al., 2011a,b). The search for primary studies did not yield any
relevant reports.
3.1.6.1. Systematic reviews. Two prospective cohort studies were
identified and included in the HTA reported by Williams et al.
(2010). Both studies failed to detect an association. The reviewers
note that the results of the systematic review were limited by
heterogeneous criteria for exposure ascertainment and the small
number of studies available for analysis.
The systematic review by Hao et al. (2011a,b) examined the
effect of metabolic syndrome on the risk of dementia. A total of
nine relevant prospective cohort studies, three of which reported
on the risk of developing AD, were eligible for inclusion. All three
studies failed to detect an association between metabolic
syndrome and risk of AD. Overall, results of this review suggested
that metabolic syndrome was not associated with the AD risk.
3.1.6.2. Discussion. Based on findings from two moderate-quality
systematic reviews, which collated evidence from three cohort
studies, metabolic syndrome was not associated with risk of AD.
Further research to confirm the association between the metabolic
syndrome and risk of AD is warranted.
3.1.7. Ankle Brachial Index (ABI)
One moderate-quality systematic review examined the associ-
ation between low ankle brachial index (ABI), an indicator of
lower-extremities peripheral artery disease, and risk of dementia
(Guerchet et al., 2011). A total of 12 studies, four of which examined
risk of AD, were included in the review. One of the four studies was
cross-sectional while the remaining three studies employed a
longitudinal study design. All four studies reported an increased
risk of AD associated with low ABI (< 0.90); however, results from
two studies failed to reach statistical significance and one was only
marginally significant. Based on the available evidence, the
reviewers concluded that low ABI, an indicator of cardiovascular
disease, predicted risk of AD.
3.1.8. Stroke
Two systematic reviews (Patterson et al., 2007; Purnell et al.,
2009), one of which was excluded for low methodological quality,
reported on the effect of stroke on the risk of AD. We identified four
relevant primary studies in the second phase of the review (Song
et al., 2011; Verdelho et al., 2010; Li et al., 2010a,b; Qiu et al., 2010).
3.1.8.1. Systematic reviews. One relevant longitudinal cohort study
was identified by Patterson et al. (2007). According to the results of
the study, stroke was associated with an 83% increased risk of AD
(RR = 1.83; 95% CI = 1.14, 2.95).
3.1.8.2. Primary studies. According to results of the LADIS
prospective cohort study, previous stroke (HR = 0.22; 95%
CI = 0.05, 0.98; p = 0.051) and stroke during follow-up (HR = 1.35;
95% CI = 0.32, 5.70; p = 0.679) were not significantly associated with
risk of AD (Verdelho et al., 2010). Similarly, the Kungsholmen
project (HR = 1.20; 95% CI = 0.77, 1.89) and Canadian Study of Health
and Aging (OR = 0.69; 95% CI = 0.35, 1.36) failed to detect an
association between stroke and risk of AD (Qiu et al., 2010; Song
et al., 2011). The remaining study reported a two-fold increased
risk of incident AD associated with stroke (HR = 2.00; 95% CI = 1.93,
2.07; p < 0.001) (Li et al., 2010a,b).
3.1.8.3. Discussion. There was conflicting evidence pertaining to
the association between stroke and risk of AD. The single cohort
study identified in the systematic review by Patterson et al. (2007)
suggested an increased risk of AD associated with stroke. Three of
four primary studies identified in the second phase of the review
failed to detect an association between stroke and AD. Considered
cumulatively, the association between stoke and risk of AD was
inconclusive.
3.1.9. Atrial fibrillation
One systematic review (Purnell et al., 2009) was identified that
reported on the association between atrial fibrillation and risk of
AD. However, the review was excluded due to low methodological
quality. Two cohort studies reporting on the risk of AD associated
with atrial fibrillation were identified (Marengoni et al., 2011;
Dublin et al., 2011).
History of chronic atrial fibrillation did not increase risk of AD
according to findings from the Kungsholmen Project (Marengoni
et al., 2011), a prospective cohort study conducted in Sweden
(HR = 0.80; 95% CI = 0.40, 1.50).
Results from the Adult Changes in Thought study (Dublin et al.,
2011) yielded conflicting results, suggesting an increased risk of AD
associated with history of atrial fibrillation (HR = 1.50; 95% CI = 1.16,
1.94). The risk estimate was attenuated but remained statistically
significant based on the analysis with the most rigorous
adjustment for potential confounders (HR = 1.41; 95% CI = 1.07,
1.86).
3.1.10. Cardiovascular disease
Four observational studies were identified that reported on the
broad category of cardiovascular disease (CVD) and certain non-
stroke subtypes (Li et al., 2010a,b; Qiu et al., 2010; Eriksson et al.,
2010a; Song et al., 2011) (see Table 95 in Supplementary material
II.)
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
3.1.10.1. Primary studies. According to results reported by Li et al.
(2010a,b), cardiovascular disease was not associated with incident
AD (HR = 0.99; 95% CI = 0.96, 1.02; p = 0.505). Similarly, the
Canadian Study of Health and Aging failed to detect an
association between heart disease and AD (OR = 0.94; 95%
CI = 0.69, 1.28) (Song et al., 2011). The Kungsholmen project
detected an increased risk of AD associated with heart failure
(HR = 1.54; 95% CI = 1.05, 2.25) (Qiu et al., 2010).
Eriksson et al. (2010a) examined the effect of non-stroke
cardiovascular disease and myocardial infarction on AD risk.
According to the longitudinal analysis, there was no association
between non-stroke cardiovascular disease and risk of AD (First 3
years after CVD, HR = 1.48; 95% CI = 0.83, 2.64. More than 3 years
after CVD, HR = 0.67; 95% CI = 0.37, 1.21). However, the analysis
stratified by ApoE4 genotype detected an increased risk of AD
associated with CVD among ApoE4 carriers (HR = 2.39; 95%
CI = 1.15, 4.96) but not among non-carriers (HR = 0.76; 95%
CI = 0.24, 2.42). There was no association between myocardial
infarction (MI) alone and risk of AD (First 3 years after MI,
HR = 0.91; 95% CI = 0.35, 2.35. More than 3 years after MI, HR = 0.64;
95% CI = 0.27, 1.53). The co-twin control analysis failed to detect an
association between CVD and MI with AD (see Table 95 in
Supplementary material II).
3.1.10.2. Discussion. Of the four identified observational studies
reporting on cardiovascular disease and its non-stroke subtypes,
only two detected a positive association. More specifically, results
of the Kungsholmen project suggest an increased risk of AD
associated with heart failure (Qiu et al., 2010). The second study
provided preliminary evidence suggesting an increased risk of AD
associated with non-stroke CVD in ApoE4 carriers (Eriksson et al.,
2010a). The remaining two studies failed to detect an association
between CVD and risk of AD.
3.1.11. Multiple vascular risk factors
We did not identify any systematic reviews that examined the
effect of multiple, concurrent vascular risk factors. Three
(Ronnemaa et al., 2011; Qiu et al., 2010; Reitz et al., 2010b)
relevant cohort studies were identified in the second phase of the
review.
3.1.11.1. Primary studies. Using data from the Kungsholmen
project, Qiu et al. (2010) examined the risk of AD attributed to
systolic and diastolic blood pressure, pulse pressure, diabetes and
pre-diabetes, stroke, and heart failure. The cumulative effect of the
aforementioned vascular risk factors was also examined. The
overall vascular risk profile was derived based on the number of
present vascular risk factors, namely, “high systolic pressure
(160 mm Hg), low diastolic pressure ( < 70 mm Hg), low pulse
pressure (<70 mm Hg), diabetes/pre-diabetes, stroke, and heart
failure” (Qiu et al., 2010). Using absence of vascular risk factors as a
referent, the presence of two or more vascular risk factors
significantly increased the risk of AD (One risk factor, HR = 1.09;
95% CI = 0.75, 1.60. Two risk factors, HR = 1.77; 95% CI = 1.16, 2.71.
Three or more risk factors, HR = 2.66; 95% CI = 1.39, 5.08). An
atherosclerotic subtype profile was derived based on the presence
of high systolic pressure, diabetes/pre-diabetes, and stroke.
Presence of one of these factors marginally increased risk of AD
by 33% (HR = 1.33; 95% CI = 1.00, 1.78) while presence of two or
more factors significantly increased risk of AD by two-fold
(HR = 2.09; 95% CI = 1.31, 3.34). Finally, a hypoperfusion risk
profile was derived based on the presence of low diastolic
pressure, low pulse pressure, and heart failure. Presence of two
or more of these factors significantly increased risk of AD
(HR = 2.06; 95% CI = 1.35, 3.13), whereas presence of a single
factor had no significant effect (HR = 1.27; 95% CI = 0.96, 1.69). The p
153
for trend was statistically significant for all three risk profiles
indicating that increasing number of vascular factors were
associated with increasing AD risk.
Reitz et al. (2010b) developed a summary risk score, based on
vascular factors, to estimate the likelihood of developing LOAD.
Sex, age, diabetes, hypertension, current smoking, low HDL-C, high
waist to hip ratio, education, ethnicity, and APOE e4 allele were
considered. Using risk scores between 0 and 14 as a referent, higher
risk scores were associated with greater risk of probable and
possible LOAD (15–18, HR = 3.74; 95% CI = 1.42, 9.88. 19–22,
HR = 3.55; 95% CI = 1.31, 9.62. 23–28, HR = 12.57; 95% CI = 5.26,
30.08. >28, HR = 20.47; 95% CI = 8.38, 49.99). Similar results were
obtained when the analysis was restricted to risk of probable LOAD.
The study reported by Ronnemaa et al. (2011) examined the risk
of AD associated with vascular risk factors considered individually
and in combination. Participants enrolled in the Uppsala Longitu-
dinal Study of Adult Men were examined at two baseline
assessments. Participants were assessed at midlife (average
age = 49.6 years) and followed for an average of 29.0 years. Subjects
who did not develop dementia by age 70 (average age of late-life
assessment = 71.0 years) were enrolled in the late-life baseline
assessment and followed for an average of 13.0 years. Systolic
blood pressure, fasting plasma glucose, serum cholesterol, body
mass index, and current smoking, measured in midlife and late-
life, were not associated with risk of AD. Subjects possessing the
APOE e4 allele were at an increased risk of AD. To examine the
effect of multiple risk factors, subjects with one, two, three or more
existing risk factors, which included current smoking, systolic
blood pressure >140 mmHg, BMI >28, fasting serum glucose
>7.0 mmol/l, and serum cholesterol >7.0 mmol/l, were compared
to subjects free of these factors. According to the results of the
study, having one or more vascular risk factors at midlife and late-
life did not increase risk of AD.
3.1.11.2. Discussion. Two of the three identified studies reported
that the presence of multiple vascular risk factors (Qiu et al., 2010)
or a higher vascular risk score (Reitz et al., 2010b) could predict the
development of AD. The remaining study failed to detect an
association between multiple vascular factors and risk of
subsequent AD. Further research is required to clarify the effect
of multiple vascular risk factors on the risk of developing AD.
3.1.12. B-type Natriuretic Peptide (BNP)
A single primary study was identified which reported on the
association between B-type natriuretic peptide and risk of AD.
According to the results reported by Kerola et al. (2010) higher
levels of B-type natriuretic peptide were associated with risk of AD
(OR = 1.59; 95% CI = 1.09, 2.30; p = 0.015). Further research is
required to confirm the association.
3.1.13. Antihypertensive drugs
The effect of antihypertensive drugs on risk of AD was examined
in four systematic reviews (Standridge, 2004; Shah et al., 2009;
Williams et al., 2010; Chang-Quan et al., 2011), one of which was
excluded due to low methodological quality (Standridge, 2004)
(see Table 9 in the Supplementary material II). No recent primary
studies were identified.
3.1.13.1. Systematic reviews. Six studies investigating the
association between antihypertensive use and incidence or
progression of AD were eligible for inclusion in the systematic
review reported by Shah et al. (2009). Two studies reported solely
on progression of AD and will not be summarized herein. All four
studies consistently reported risk estimates below 1.00, however;
results from only two studies reached statistical significance.
154 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Eight cohort studies were identified and included in the HTA
reported by Williams et al. (2010). Only half of the included cohort
studies detected a statistically significant reduced risk of AD
associated with antihypertensive drug use. The reviewers also
identified a systematic review and meta-analysis restricted to
randomized controlled trials (RCTs) (McGuinness et al., 2009).
According to a pooled analysis of three RCTs, antihypertensive drug
use was not associated with risk of dementia (OR = 0.89; 95%
CI = 0.69, 1.16). The pooled risk estimate was not specific to AD
(Williams et al., 2010).
Two randomized controlled trials and five cohort studies were
eligible for inclusion in the systematic review reported by Chang-
Quan et al. (2011). Of the seven studies included, six reported a
non-significant reduced risk of AD among antihypertensive
medication users with risk estimates ranging from 0.55 to 0.89.
The remaining study reported a non-significant increased risk. In
the pooled analysis combining the results of all seven studies,
antihypertensive drug use was not associated with AD risk
(RR = 0.90; 95% CI = 0.79, 1.03). In the subgroup analysis stratified
by methodological study design, antihypertensives were not
associated with AD in the pooled analysis of cohort studies
(RR = 0.92; 95% CI = 0.79, 1.08) or RCTs (RR = 0.79; 95% CI = 0.53,
1.18).
3.1.13.2. Discussion. Available data from observational studies,
summarized in three moderate quality systematic reviews, suggest
that antihypertensive drugs do not significantly reduce risk of AD.
Although results from many of the observational studies failed to
reach statistical significance, a majority of studies reported risk
estimates below 1.00 (Shah et al., 2009; Chang-Quan et al., 2011). A
meta-analysis of three RCTs failed to detect an association between
antihypertensive drug use and risk of dementia (McGuinness et al.,
2009). Heterogeneity across the included studies with respect to
antihypertensive drug dosage, study methodology, patient
characteristics, and length of follow-up were cited as important
limitations (Shah et al., 2009). Collectively, there was a lack of
evidence to suggest that antihypertensive drug use was associated
with a protective effect against the development of AD.
3.1.14. Non-Steroidal Anti-Inflammatory Drugs (NSAIDs)
Six systematic reviews, which examined the association
between NSAID use and risk of AD, were identified (Etminan
et al., 2003; Standridge, 2004; Szekely et al., 2004; de Craen et al.,
2005; Patterson et al., 2007; Williams et al., 2010). One low quality
review was excluded (Standridge, 2004). A single primary study
was identified in the second phase of the review (Breitner et al.,
2011).
3.1.14.1. Systematic reviews. Six cohort and four case-control
studies investigating the association between NSAIDs and/or
aspirin use and AD were included in the systematic review
reported by Etminan et al. (2003). The pooled analysis of nine
studies suggested a 28% reduced risk of AD associated with NSAID
use (RR = 0.72; 95% CI = 0.56, 0.94; test for heterogeneity: p = 0.06).
Analyses stratified by methodological design detected an
association among case-control (RR = 0.62; 95% CI = 0.45, 0.82;
test for heterogeneity: p = 0.55) but not cohort (RR = 0.84; 95%
CI = 0.54, 1.05; test for heterogeneity: p = 0.04) studies. Four cohort
studies provided data pertaining to the length of NSAID use. Short-
and intermediate-term NSAID use was not associated with risk of
AD. Long term use, defined as greater than 12 months in one study
and greater or equal to 24 months in the remaining three studies,
was associated with a reduced risk of AD (RR = 0.27; 95% CI = 0.13,
0.58). No association was detected between aspirin use and risk of
AD according to the pooled analysis of eight studies (RR = 0.87; 95%
CI = 0.70, 1.07; test for heterogeneity: p = 0.79). When stratified by
methodological design, the meta-analysis of five cohort (RR = 0.85;
95% CI = 0.71, 1.03; test for heterogeneity: p = 0.76) and three case-
control (RR = 0.84; 95% CI = 0.61, 1.16; test for heterogeneity:
p = 0.77) studies yielded consistent findings suggesting a lack of
association between aspirin use and risk of AD. The reviewers
concluded that there was some evidence to suggest that NSAIDs,
particularly when used for longer durations, may exert a protective
effect against the development of AD.
Szekely et al. (2004) identified seven non-prospective and four
prospective studies. Of the seven non-prospective studies, four
studies were cross-sectional studies and three were case-control
studies. All four cross-sectional studies reported a statistically
significant reduced risk of AD among NSAID users, with odds ratios
ranging from 0.29 to 0.43. In contrast, only one case-control study
reported a significant reduced risk of AD associated with NSAID use
(OR = 0.55; 95% CI = 0.37, 0.82). In the pooled analysis of the seven
non-prospective studies representing 12979 subjects, lifetime use
of NSAIDs was associated with a statistically significant 49%
reduced odds of AD (OR = 0.51; 95% CI = 0.40, 0.66; Q = 5.55;
p = 0.48). NSAID use was associated with a protective effect against
the development of AD according to the pooled analysis of
prospective studies (RR = 0.74; 95% CI = 0.62, 0.89; Q = 4.27,
p = 0.23). Use of NSAIDs for two or more years was associated
with a greater risk reduction according to the pooled analysis of
three prospective studies (RR = 0.42; 95% CI = 0.26, 0.66; Q = 1.16,
p = 0.56). Overall, results of the review suggested that use of non-
aspirin NSAIDs was associated with a reduced risk of AD (Szekely
et al., 2004).
Eleven case-controls studies, which examined the association
between NSAID use and prevalent dementia, were identified by de
Craen et al. (2005). According to the random effects meta-analysis,
NSAID use was associated with a statistically significant 49%
reduced risk of dementia (RR = 0.51; 95% CI = 0.37, 0.70; test for
heterogeneity: p = 0.001). However, the results of the pooled
analysis were not specific to AD as data from three studies that
focused on all-cause dementia were included in the combined risk
estimate. Ten studies examined the association between NSAID use
and incident dementia. Eight of these studies specifically reported
on the risk of AD. All eight studies reported risk estimates below
1.00, however, only results from two studies reached statistical
significance. When data from the ten studies were pooled in a
random effects meta-analysis, NSAID use was associated with a
21% reduced risk of incident dementia (RR = 0.79; 95% CI = 0.68,
0.92). When stratified by methodological design, NSAID use was
associated with a statistically significant reduced risk of incident
dementia according to the pooled analysis of case-control studies
(RR = 0.77; 95% CI = 0.61, 0.96) and only a marginally significant
reduced risk among cohort studies (RR = 0.79; 95% CI = 0.61, 1.01).
Overall, results of this review suggested that NSAID use was
associated with a reduced risk of dementia and AD. However, the
reviewers cautioned that the observed protective effect may have
been attributed to sources of bias.
Patterson et al. (2007) identified a single study reporting on
NSAIDs. According to the results of the longitudinal cohort study,
NSAID use was associated with a 58% (RR = 0.42; 95% CI = 0.26, 0.66)
reduced risk of AD.
One systematic review (Szekely et al., 2004), four cohort
studies, and two RCTs were identified and included in the HTA
reported by Williams et al. (2010). Results of the pooled analysis
restricted to cohort studies, conducted by Szekely et al. (2004),
suggest that NSAID use was associated with a statistically
significant reduced risk of AD. Four additional prospective studies
not included in the review by Szekely et al. (2004), were identified
by Williams et al. Aspirin was not associated with AD in all three
cohort studies. Two of the four studies failed to detect an
association between NSAID use and AD, and one study reported
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
a statistically significant reduced risk of AD associated with NSAID
use among carriers of one or more APOE e4 alleles. The final study
failed to detect an association between moderate NSAID use and
risk of AD; however, heavy use was associated with a statistically
significant increased risk of AD (Williams et al., 2010). Findings
from the four cohort studies were combined with the four cohort
studies identified by Szekely et al. (2004). Six of the eight studies
reported risk estimates below 1.00, two of which reached
statistical significance. According to the pooled analysis, however,
NSAID use was not associated with risk of AD (HR = 0.83; 95%
CI = 0.63, 1.09; Q statistic = 40.84. p < 0.001, I2 = 83%). However, a
sensitivity analysis excluding one study yielded statistically
significant results (HR = 0.79; 95% CI = 0.69, 0.91). Findings from
two RCTs were also considered by Williams et al. According to one
study, rofecoxib (25 mg/day) increased the risk of AD (HR = 1.49;
95% CI = 1.08, 2.05). The second study, by the ADAPT Research
Group, reported consistent findings, suggesting an increased risk of
dementia associated with NSAID use when compared with placebo
(Celecoxib, HR = 4.11; 95% CI = 1.30, 13.0. Naproxen, HR = 3.57; 95%
CI = 1.09 to 11.7). Overall, relying on evidence from observational
studies would suggest that the association between NSAIDs and
risk of AD was inconclusive. However, results from two RCTs
suggest that use of certain NSAIDs, namely, rofecoxib, celecoxib,
and naproxen, were associated with a deleterious effect (Williams
et al., 2010). Based on the latter finding, the reviewers concluded
that NSAID use may increase risk of AD.
3.1.14.2. Primary Studies. One primary study was identified which
reported on the association between NSAID use and risk of AD
(Breitner et al., 2011). This study, reported by the Adapt Research
Group, was a follow-up to the RCT identified in the systematic
review by Williams et al. (2010). The study was an 18 to 24 month
follow-up of subjects previously randomized to receive naproxen,
celecoxib, or placebo for an average of two years. Results of the
follow-up were inconsistent with previous results, which
suggested an increased risk of AD associated with NSAID
treatment (Celecoxib, HR = 4.11; 95% CI = 1.30, 13.0. Naproxen,
HR = 3.57; 95% CI = 1.09 to 11.7). More specifically, results of the
follow-up failed to detect an increased risk of AD among subjects
randomized to naproxen (Rate ratio = 0.87; 95% CI = 0.43, 1.70) and
celecoxib (Rate ratio = 1.20; 95% CI = 0.72, 2.41).
3.1.14.3. Discussion. Five moderate quality systematic reviews
examined the effect of NSAID use and risk of AD. Results from
identified observational studies were meta-analyzed in four of the
five systematic reviews. Meta-analyses of cohort studies
conducted by Etminan et al. (2003) and de Craen et al. (2005)
failed to reach statistical significance; however, pooled risk
estimates were below 1.00. The HTA report by Williams et al.
(2010) yielded statistically significant results once the outlying
cohort study was excluded from the meta-analysis. Finally, the
pooled analysis of cohort studies by Szekely et al. (2004) suggested
a statistically significant protective effect associated with NSAID
use. As noted by Williams et al. (2010), sources of methodological
and clinical heterogeneity across observational studies may have
contributed to inconsistent findings across primary studies and the
reviews themselves. Overall, pooled analyses of cohort studies
consistently yielded risk estimates below 1.00. Considered
cumulatively, evidence from observational studies suggest that
NSAIDs, particularly when used for longer durations (Etminan
et al., 2003), may reduce the risk of AD.
Results from RCTs, which are generally susceptible to less bias,
present conflicting findings. Two RCTs identified by Williams et al.
(2010) detected an increased risk of AD and dementia associated
with NSAID treatment. However, a follow-up of one of the trials
failed to detect an increased risk (Breitner et al., 2011). In light of
155
the ADAPT trial update identified in the second phase of the review,
the association between NSAIDs and AD was inconclusive.
3.1.15. Cholinesterase inhibitors
Three moderate-quality systematic reviews examined the
association between cholinesterase inhibitors (ChEIs) and risk of
conversion from mild cognitive impairment (MCI) to AD (Raschetti
et al., 2007; Diniz et al., 2009; Williams et al., 2010). No relevant
primary studies were identified.
3.1.15.1. Systematic reviews. A total of three published and five
unpublished RCTs examining the effect of ChEIs on conversion
from MCI to AD were eligible for inclusion in the systematic review
reported by Raschetti et al. (2007). All four studies consistently
reported a higher conversion rate for the placebo group; however,
conversion rates were not significantly different between groups.
Moreover, the relative risks of conversion from MCI to AD, which
could only be calculated for one published (RR = 0.84; 95% CI = 0.57,
1.25) and one unpublished (RR = 0.85; 95% CI = 0.64, 1.12) study,
were not statistically significant. Based on the results of this
systematic review, use of ChEIs was not associated with a delay in
conversion from MCI to AD. Heterogeneous study populations,
ambiguity of MCI diagnosis, and unavailable data for two of the
studies included were all cited as important limitations to consider
when interpreting the results of the review (Raschetti et al., 2007).
Four randomized controlled trials were eligible for inclusion in
the review by Diniz et al. (2009). All four studies were also
identified and included in the earlier systematic review by
Raschetti et al. (2007). However, Diniz et al. (2009) excluded
two published and two unpublished studies that were included in
the earlier review, citing “conversion from MCI to AD [was] not the
main study outcome” as the basis for exclusion. According to meta-
analysis of the RCTs, 15.4% and 20.4% of subjects in the treated and
placebo groups converted to AD, respectively. Cholinesterase
inhibitor use was associated with a statistically significant 24%
(RR = 0.76; 95% CI = 0.65, 0.88; Q = 3.40, p = 0.33) decreased risk of
progression from MCI to AD. Overall, results of this review suggest
that cholinesterase inhibitor use, particularly for longer durations,
can reduce the risk of progression from MCI to AD (Diniz et al.,
2009).
Williams et al. (2010), examined the association between
cholinesterase inhibitors and risk of AD. No relevant observational
studies were identified. However, the reviewers identified and
included the systematic reviews by Raschetti et al. (2007) and
Diniz et al. (2009). The review by Raschetti et al. (2007) failed to
detect an association between ChEI use and risk of conversion from
MCI to AD. The meta-analysis of three RCTs by Diniz et al. (2009)
detected a 25% reduced risk of conversion from MCI to AD
associated with ChEIs. Due to the methodological limitations of the
aforementioned meta-analysis, Williams et al. (2010) concluded
that cholinesterase inhibitor use was not associated with conver-
sion from MCI to AD.
3.1.15.2. Discussion. Results from the earliest systematic review
suggest that use of ChEIs was not associated with a delay in
conversion from MCI to AD (Raschetti et al., 2007). A more recent
review and meta-analysis by Diniz et al. (2009) suggested that long
term ChEI use was associated with a statistically significant
reduced risk of progression. However, as noted by Williams et al.
(2010), the small number of studies and important sources of
methodological heterogeneity warrant cautious interpretation of
the pooled risk estimate derived by Diniz et al. (2009).
Consequently, when the available evidence was considered
cumulatively, the effect of ChEI use on AD risk among subjects
with MCI was inconclusive.
156 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
3.1.16. Hormone Replacement Therapy (HRT)
Three moderate systematic reviews examined the association
between hormone replacement therapy (HRT) and risk of AD (Yaffe
et al., 1998; LeBlanc et al., 2001; Williams et al., 2010). One low
quality review was excluded (Standridge, 2004). No primary
studies were identified.
3.1.16.1. Systematic reviews. Yaffe et al. (1998) identified two
prospective cohort and eight case-control studies that reported on
the association between HRT use and risk of AD. Results of the
pooled analysis suggest that postmenopausal estrogen therapy use
was associated with 29% reduced risk of AD (OR = 0.71; 95%
CI = 0.52, 0.98). Estrogen use reduced risk of AD, albeit only
marginally, according to the subgroup analysis restricted to studies
which used the widely accepted NINCDS-ADRDS criteria for
outcome ascertainment (OR = 0.69; 95% CI = 0.48, 1.01; test for
heterogeneity: p = 0.07). Estrogen use reduced risk of AD according
to the pooled analysis of cohort (OR = 0.48; 95% CI = 0.29, 0.81; test
for heterogeneity p = 0.86), but not case-control (OR = 0.80; 95%
CI = 0.56, 1.16; test for heterogeneity: p = 0.11) studies.
Two cohort and 10 case-control studies were eligible for
inclusion in the systematic review reported by LeBlanc et al.
(2001). The meta-analysis of all 12 studies suggested that HRT use
was associated with a statistically significant 34% reduced risk of
AD (RR = 0.66; 95% CI = 0.53, 0.82; test for heterogeneity: p > 0.10).
In the sensitivity analysis stratified by methodological design, the
protective effect of HRT was more pronounced in the pooled
analysis of cohort studies (Cohort, RR = 0.50; 95% CI = 0.30, 0.80.
Case-control, RR = 0.71; 95% CI = 0.56, 0.91). However, the protec-
tive effect of HRT did not remain statistically significant when the
analysis was restricted to methodologically superior studies
(RR = 0.64; 95% CI = 0.32, 1.06). Lastly, HRT use was associated
with a reduced risk of AD according to the subgroup analysis
excluding studies that ascertained HRT status using proxy
informant data (RR = 0.72; 95% CI = 0.55, 0.96).
The HTA reported by Williams et al. (2010) identified one
systematic review and two RCTs that examined the effect of
estrogen therapy on the incidence of dementia and AD. Results of
the identified review (LeBlanc et al., 2001) suggest that HRT use
may reduce the risk of AD. As noted by Williams et al. (2010),
findings of the included review must be interpreted with caution
due to variability across studies with respect to the types and
duration of HRT use. Both identified RCTs provided risk estimates
for all-cause dementia, which encompasses but is not restricted to
AD (Williams et al., 2010). According to the results of these trials,
conjugated equine estrogen use with (HR = 2.05; 95% CI = 1.21, 3.48)
but not without (HR = 1.77; 95%: CI = 0.74, 4.23) medroxyproges-
terone acetate was associated with a deleterious effect. Based on
these results, the reviewers concluded that estrogen combined
with progestin, namely, conjugated equine estrogen use with
medroxyprogesterone acetate, increased risk of dementia.
3.1.16.2. Discussion. Earlier systematic reviews, which relied
solely on observational studies, detected a protective effect of
HRT (Yaffe et al., 1998; LeBlanc et al., 2001). However, results from
two RCTs identified by Williams et al. (2010) suggest that estrogen
combined with progestin, namely, conjugated equine estrogen use
with medroxyprogesterone acetate, increases risk of AD/dementia.
In the absence of progestin, conjugated equine estrogen use did not
increase risk of AD/dementia (Williams et al., 2010).
3.1.17. Lipid lowering therapy/statins
Nine systematic reviews, which examined the effect of lipid
lowering therapies, were identified (Miller and Chacko, 2004;
Standridge, 2004; Patatanian and Gales, 2005; Patterson et al.,
2007; Zhou et al., 2007; Beri et al., 2009; Desilets et al., 2010;
Williams et al., 2010; Brinker, 2011). Three moderate quality
systematic reviews examined the association between lipid
lowering therapies, particularly statin use, and risk of AD
(Patterson et al., 2007; Zhou et al., 2007; Williams et al., 2010).
The remaining six low-quality studies were excluded. The search
for recent primary studies yielded one record (Li et al., 2010a,b)
(see Tables 13 and 100 in Supplementary material II).
3.1.17.1. Systematic reviews. Four cohort and two case-control
studies identified by Zhou et al. (2007) examined AD as an
outcome. According to a random effects meta-analysis, statin
therapy was not associated with risk of AD (OR = 0.81; 95%
CI = 0.56, 1.16; test for heterogeneity: p = 0.14).
Two longitudinal cohort studies identified by Patterson et al.
(2007) examined the effect of statin use. Both studies failed to
detect an association. One study examined the effect of all lipid
lowering drugs. Results from this study suggest that use of lipid
lowering drugs was associated with a reduced risk of AD
(OR = 0.26; 95% CI = 0.08, 0.88).
Six studies were eligible for inclusion in the HTA reported by
Williams et al. (2010). Findings from these six studies were
included in a random effects meta-analysis, which detected a
statistically significant reduced risk of AD associated with statin
use (HR = 0.73; 95% CI = 0.57, 0.94). There was no evidence of
statistical heterogeneity (Q statistic = 5.132, p-value = 0.40;
I2 = 2.58%) or publication bias. No relevant RCTs were identified.
Based on the available evidence from observational studies, the
reviewers concluded that statin use appeared to reduce the risk of
AD.
3.1.17.2. Primary studies. Li et al. (2010a,b) prospectively examined
the effect of statin therapy on risk of AD. According to the results of
the study, statin therapy was associated with a reduced risk of AD
(HR = 0.62; 95% CI = 0.40, 0.97). Stratified analysis by age detected a
protective effect in younger subjects and a detrimental non-
significant association in older subjects ( < 80 years of age,
HR = 0.44; 95% CI = 0.25, 0.78. 80 years of age, HR = 1.22; 95%
CI = 0.61, 2.42). Stratified analysis by APOE e4 status yielded risk
estimates below 1.00 for both strata, however, results failed to
reach statistical significance for non-carriers (Carriers, HR = 0.42;
95% CI = 0.20, 0.91. Non-carriers, HR = 0.77; 95% CI = 0.42, 1.40.
Unknown, HR = 0.84; 95% CI = 0.22, 3.15) (Li et al., 2010a,b).
3.1.17.3. Discussion. Two of the three identified systematic
reviews failed to detect a statistically significant protective
effect of statin use against the development of AD. However, the
most recent systematic review and meta-analysis by Williams et al.
(2010) suggests a reduced risk of AD attributed to statin therapy.
These inconsistent findings may be attributed to methodological
heterogeneity across the identified reviews with respect to the
utilized search strategies and inclusion criteria. A recent cohort
study (Li et al., 2010a,b), identified in the second phase of the
review, detected a reduced risk of AD attributed to statin therapy
use before the age of 80.
3.1.18. Benzodiazepines
One moderate systematic review examined the association
between benzodiazepines and risk of AD (Patterson et al., 2007).
However, the review did not identify any relevant studies.
Consequently, there was insufficient evidence to comment on
the association between benzodiazepines and risk of AD.
3.1.19. Memantine
One moderate systematic review examined the association
between memantine and risk of AD (Williams et al., 2010). As the
review did not identify any relevant primary studies, there is a lack
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
of evidence on which to evaluate the association between
memantine and risk of AD.
3.1.20. Vaccinations
One moderate systematic review examined the association
between vaccinations and risk of AD (Patterson et al., 2007). One of
the two identified longitudinal studies reported that having
received any vaccination was associated with a 60% reduced risk
of AD (RR = 0.40; 95% CI = 0.17, 0.96). In a second study, vaccination
against diphtheria or tetanus (RR = 0.41; 95% CI = 0.27, 0.62) and
poliomyelitis (RR = 0.60; 95% CI = 0.37, 0.99) was associated with a
decreased risk of AD. In the same study, vaccination against
influenza was not associated with risk of AD (RR = 0.75; 95%
CI = 0.54, 1.04).
3.1.21. Histamine-2 receptor antagonists
We did not identify any systematic reviews that reported on the
association between histamine-2 receptor antagonist (H2RA) use
and risk of AD. Our search for primary studies yielded one relevant
record (Gray et al., 2011).
The association between H2RA use and risk of AD was
prospectively examined by Gray et al. (2011). The study reported
on the effect of cumulative use, intensity of use, and cumulative
use relative to timing of exposure (recent and distant). There was
no significant association between cumulative H2RA use and risk
of AD. See Table 9. Categories delineating intensity of use were also
not associated with risk of AD (see Table 98 in Supplementary
material II). Only high distant and low recent H2RA use yielded
statistically significant findings (HR = 1.66; 95% CI = 1.07, 2.57). The
remaining categories of distant and recent cumulative use were
not associated with risk of AD. Overall, findings of the study
suggest that H2RA use was not associated with risk of AD.
3.1.22. Angiotensin receptor blockers
We did not identify any systematic reviews that reported on the
association between angiotensin receptor blocker use and risk of
AD. The second phase of the review yielded one relevant
observational study (Li et al., 2010a,b).
Using data from the US Veteran Affairs administrative database,
Li et al. (2010a,b) prospectively examined the association between
angiotensin receptor blocker use and risk of AD. Compared to
lisinopril, an angiotensin converting enzyme inhibitor, angiotensin
receptor blocker use was associated with a statistically significant
reduced risk of AD (HR = 0.81; 95% CI = 0.68, 0.96; p = 0.016).
Compared to the cardiovascular comparator (drugs “excluding
angiotensin receptor blockers, statins, and converting enzyme
inhibitors”), angiotensin receptor blockers marginally reduced risk
of incident AD (HR = 0.84; 95% CI = 0.71, 1.00; p = 0.045) (Li et al.,
2010a,b).
3.1.23. Lead
Four systematic reviews reported on the risk of AD associated
with occupational exposure to lead (Graves et al., 1991b;
Santibanez et al., 2007; Williams et al., 2010; Loef et al., 2011a,
b). One review was excluded due to low methodological quality
(Loef et al., 2011a,b). No relevant primary studies were identified.
The systematic review and re-analysis reported by Graves et al.
(1991b) identified four relevant case-control studies. All four
studies failed to detect a statistically significant association
between lead exposure and AD. However, two studies had only
one exposed case. The pooled analysis including 261 cases and 337
controls suggested that occupational lead exposure was not
associated with AD (RR = 0.71; 95% CI = 0.36, 1.41).
The HTA (Williams et al., 2010) examined the association
between lead and risk of AD. The reviewers relied on the
systematic review reported by Santibanez et al. (2007). All six
157
case-control studies included in the review failed to detect an
association between lead exposure and AD. Based on these
findings, the reviewers concluded that lead exposure was not
associated with AD risk.
3.1.24. Solvents
Three moderate-quality systematic reviews reported on the risk
of AD associated with occupational exposure to solvents (Graves
et al., 1991b; Santibanez et al., 2007; Williams et al., 2010). No
relevant primary studies were identified.
3.1.24.1. Systematic reviews. The systematic review and re-analysis
by Graves et al. (1991b) identified three case-control studies that
reported on the association between occupational exposure to
solvents and risk of AD. Two studies failed to detect an association.
A risk estimate could not be calculated for the third study as there
were no exposed cases. In the pooled analysis representing 221
cases and 287 controls, occupational exposure to solvents was not
associated with AD (RR = 0.76; 95% CI = 0.47, 1.23).
Santibanez et al. (2007) identified 10 case-control and one
cohort study. Only two of the eleven studies detected a statistically
significant increased risk of AD, attributed to occupational
exposure to solvents. The remaining nine studies failed to detect
an association. The reviewers suggest that the conflicting findings
may be attributed to heterogeneity with respect to the broad
category of solvents considered in the included studies.
Williams et al. (2010) relied on the systematic review by
Santibanez et al. (2007). The HTA report concluded that there was
insufficient evidence to support an association between solvents
and risk of AD.
3.1.24.2. Discussion. Only two of eleven observational studies
included in the identified systematic reviews detected an
association between solvent exposure and risk of AD. Further
research, by way of prospective cohort studies which focus on the
effect of specific solvents, is warranted (Santibanez et al., 2007;
Williams et al., 2010).
3.1.25. Electromagnetic Fields (EMF)
Four systematic reviews, which examined the effect of
occupational exposure to electromagnetic fields (EMF), were
identified (Ahlbom, 2001; Santibanez et al., 2007; García et al.,
2008; Kheifets et al., 2009). One low-quality review was omitted
(Kheifets et al., 2009) (see Table 19 in Supplementary material II).
No primary studies were identified.
3.1.25.1. Systematic reviews. Four case-control and one cohort
study were eligible for inclusion in the systematic review reported
by Ahlbom (2001). Two studies were clinical-based and three were
population-based studies. When the results of the two clinical-
based studies were combined in a meta-analysis, occupational
exposure to high or medium EMF levels was associated with an
increased risk of AD (RR = 3.20; 95% CI = 1.90, 5.40). According to
the pooled analysis of population-based studies, occupational EMF
exposure was not associated with risk of AD (RR = 1.20; 95%
CI = 0.70, 2.30). When results from all five studies were pooled in a
fixed effects meta-analysis, occupational EMF exposure was
associated with a greater than two-fold increased risk of AD
(RR = 2.20; 95% CI = 1.50, 3.20). Although results from two clinical
based studies reached statistical significance, three more recent
population-based studies, one of which employed a longitudinal
cohort design, failed to detect an association between EMF
exposure and AD (Ahlbom, 2001).
Seven studies included in the systematic review by Santibanez
et al. (2007) provided data pertaining to the association between
occupational EMF exposure and risk of AD. Only two studies by the
158 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
same author detected an increased risk of AD associated with EMF
exposure. The remaining five studies failed to detect an associa-
tion.
Nine case-control studies and five cohort studies reporting on
the effect of extremely low frequency electric and magnetic fields
(ELF-ELM) were eligible for inclusion in the review reported by
García et al. (2008). The meta-analysis combining data from all
case-control studies yielded a risk estimate of 2.03 (95% CI = 1.38,
3.00). The meta-analysis combining results from all five cohort
studies detected a 62% increased risk of AD associated with ELF-
EMF exposure (RR = 1.62; 95% CI = 1.16, 2.27; I2 = 54%; test for
heterogeneity: P value = 0.016).
3.1.25.2. Discussion. Evidence pertaining to the effect of ELF-EMF
exposure on risk of AD is relatively inconsistent. Results from the
most recent systematic review and meta-analysis (García et al.,
2008), which incorporated the largest number of studies, provide
some evidence of an increased risk of AD associated with ELF-EMF
exposure. However, these results must be interpreted with caution
due to statistical heterogeneity and publication bias (García et al.,
2008).
3.1.26. Aluminum
Three systematic reviews reported on the risk of AD associated
with exposure to aluminum (Santibanez et al., 2007; Ferreira et al.,
2008; Williams et al., 2010). One review was excluded due to low
methodological quality (Ferreira et al., 2008). No primary studies
were identified.
3.1.26.1. Systematic reviews. All three studies identified by
Santibanez et al. (2007) failed to detect an association between
occupational exposure to aluminum and risk of AD. Based on these
findings, the reviewers concluded that occupational aluminum
exposure was not associated with risk of AD.
The HTA reported by Williams et al. (2010) relied on the
findings reported by Santibanez et al. (2007), which failed to detect
an association between the variables of interest. The reviewers
identified one recent cohort study that reported an increased risk
of AD associated with aluminum from drinking water equal to or
exceeding 0.1 mg/day (RR = 1.34; 95% CI = 1.09, 1.65). The reviewers
concluded that there was insufficient evidence to comment on the
effect of aluminum.
3.1.26.2. Discussion. Based on the limited available evidence, the
association between aluminum and risk of AD was inconclusive.
3.1.27. Pesticides
Three moderate-quality systematic reviews examined the risk
of AD associated with exposure to pesticides (Patterson et al.,
2007; Santibanez, Bolumar, and Garcia 2007; Williams et al., 2010).
No relevant primary studies were identified.
3.1.27.1. Systematic reviews. The systematic review by Santibanez
et al. (2007) identified six studies that reported on the association
between occupational pesticide exposure and risk of AD. Of these
six studies, four case-control studies failed to detect a statistically
significant association. Results from two high quality cohort
studies provided conflicting results, suggesting a positive
association between occupational exposure to pesticides and
risk of AD. Based on findings of the cohort studies, the review offers
some evidence of an association between occupational pesticide
exposure and risk of AD (Santibanez et al., 2007).
One longitudinal cohort study identified by Patterson et al.
(2007) reported on the association between exposure to environ-
mental toxins and risk of AD. Results from this study suggest that
exposure to defoliants and fumigants was associated with a greater
than four-fold increased risk of AD (RR = 4.35; 95% CI = 1.05, 17.90).
The HTA reported by Williams et al. (2010) relied on the
systematic review reported by Santibanez et al. (2007). Results of
the identified review suggest that exposure to pesticides may
increase risk of AD.
3.1.27.2. Discussion. According to the evidence summarized in
three systematic reviews, exposure to pesticides may increase the
risk of AD. However, results of the identified reviews were limited
by the small number of available studies. Consequently, further
research is required to confirm the effect of pesticide exposure on
the development of AD.
3.1.28. Manual labour
One moderate quality systematic review reported on the risk of
AD associated with manual labour (Patterson et al., 2007). No
primary studies were identified.
According to the only relevant study identified by Patterson
et al. (2007), there was no significant association between manual
work and AD (RR = 1.40; 95% CI = 0.90, 2.10).
3.1.29. Mercury, mercury amalgam fillings
Risk of AD associated with environmental exposure to mercury
or mercury amalgam fillings was examined in two systematic
reviews. The review on mercury amalgam fillings was excluded
due to methodological quality (Bates, 2006).
The HTA reported by Williams et al. (2010) identified a single
cohort study that reported on the effect of environmental exposure
to mercury on risk of AD. The study detected a reduced risk of AD
associated with higher blood mercury concentrations (Third
quartile, OR = 0.41; 95% CI = 0.23, 0.74. Fourth quartile, OR = 0.56;
95% CI = 0.32, 0.99).
3.1.30. Copper, iron
One systematic review examined the effect of copper and iron
on the risk of AD (Loef and Walach, 2012a,b). The review was
deemed to be of low methodological quality and was excluded.
3.1.31. Genetic risk factors
Amyloid precursor protein (APP), presenilin 1 (PSEN1) or
presenilin 2 (PSEN2) mutations are genetic causes of autosomal-
dominant Alzheimer's disease (Bateman et al., 2011). We identified
a single review that reported on the 2/2 genotype of presenilin 1
(Rodriguez-Manotas et al., 2007). Several reviews reporting on
genetic risk factors for AD are also summarized herein.
3.1.31.1. Presenilin (PS)-1 2/2 genotype. A total of 38 case-control
studies examining the association between the presenilin-1 (PS-1)
2/2 genotype and late-onset AD were eligible for inclusion in the
review reported by Rodriguez-Manotas et al. (2007). When the
results of all 38 studies were combined in a random effects meta-
analysis, compared to the combined effect of the 1/1 and 1/2
genotypes, the 2/2 genotype was not associated with AD
(OR = 0.99; 95% CI = 0.84, 1.17; test for heterogeneity: x2 = 79.92,
p < 0.0001, I2 = 53.7%). Stratified analysis by ethnicity detected a
marginally significant increased risk among Europeans (OR = 1.19;
95% CI = 1.00, 1.41), a reduced risk among North Americans
(OR = 0.79; 95% CI = 0.64, 0.98), and no significant association in
Asians (OR = 0.79; 95% CI = 0.54, 1.17; test for heterogeneity: x
2
= 38.83, p = 0.0001, I2 = 69.1%).
3.1.31.2. Interleukin-1b. The systematic review reported by Di
Bona et al. (2008) examined the association between interleukin-
1b (IL-1b) SNPs
511 and +3953. For both polymorphisms, the TT
genotype was compared to the TC + CC genotypes. A total of five
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
studies reporting on IL-1b +3953 polymorphism were identified.
The pooled analysis yielded a risk estimate of 1.60 (95% CI = 1.16,
2.22; test for heterogeneity: x2 = 2.57, p = 0.63, I2 = 0%) suggesting a
60% increased odds of AD associated with the TT genotype of IL-1b
+3953. The meta-analysis restricted to Caucasian-only populations
yielded a risk estimate of 1.71 (95% CI = 1.21, 2.43). A total of 17
studies reported on IL-1b
511. According to the random effects
meta-analysis, the IL-1beta
511 TT genotype was not associated
with an increased odds of AD (OR = 1.01; 95% CI = 0.84, 1.21; test for
heterogeneity: x2 = 25.64, p = 0.06, I2 = 37.6%). When the analysis
was restricted to studies with the highest statistical power (n = 4),
the TT genotype of IL-1b
511 was significantly associated with
risk of AD (OR = 1.32; 95% CI = 1.03, 1.69; test for heterogeneity:
x2 = 0.01, p = 1.00, I2 = 0%).
3.1.31.3. Interleukin-1a. The systematic review by Rainero et al.
(2004) examined the association between the interleukin-1alpha

889C/T gene polymorphism and AD. A total of 11 case-control
studies were eligible for inclusion. According to random effects
meta-analysis, the TT genotype was associated with an increased
risk of AD when compared to the CT (OR = 1.51; 95% CI = 1.15, 1.99;
test for heterogeneity: x2 = 15.55, p = 0.11), CC (OR = 1.49; 95%
CI = 1.09, 2.03; test for heterogeneity: x2 = 19.32, p = 0.036), and CC/
CT (OR = 1.50; 95% CI = 1.12, 2.02) genotypes. Subgroup analysis
according to the fixed effects model detected an increased risk of
early onset AD (see Table 25 in Supplementary material II). Results
failed to reach statistical significance according to the random
effects model. No association was detected between the TT
genotype and LOAD according to both models. The pooled
analysis comparing the CT genotype to the CC genotype was not
statistically significant (OR = 0.98; 95% CI = 0.96, 1.14; test for
heterogeneity: p = 0.20).
3.1.31.4. Interleukin-10. The systematic review and meta-analysis
by Zhang et al. (2011) investigated the effect of the
1082G/A
polymorphism in the IL-10 gene on AD risk. Twelve studies were
eligible for inclusion. According to the preliminary random effects
meta-analysis, when compared to the GG genotype, the A allele
(AA + AG) was associated with an increased risk of AD (OR = 1.27;
95% CI = 1.02, 1.58; test for heterogeneity: x2 = 20.74, p = 0.04,
I2 = 47.0%). When the analysis was restricted to the 11 European
studies, the A allele remained significantly associated with an
increased risk of AD (OR = 1.27; 95% CI = 1.01, 1.59; test for
heterogeneity: x2 = 20.72, p = 0.02, I2 = 51.7%). The only Asian
study failed to detect a significant association (OR = 1.37; 95%
CI = 0.32, 5.88). Meta-analyses examining the remaining
comparisons failed to detect statistically significant results
(seeTable 26 in Supplementary material II)
3.1.31.5. Interleukin-6. Fourteen studies which examined the
association between IL-6
174G/C polymorphism and risk of AD
among Caucasians were included in the systematic review
reported by Han et al. (2011). According to random effects
meta-analysis, the GG (OR = 1.35; 95% CI = 1.06, 1.72; test for
heterogeneity: I2 = 52.5%, p = 0.01) and GG/GC (OR = 1.27; 95%
CI = 1.05, 1.53) genotypes were associated with risk of AD. However,
results failed to reach statistical significance when one study “that
reported a higher detection rate of C allele (57%)" was excluded
(Han et al., 2011). Results of the remaining meta-analytic
comparisons failed to reach statistical significance.
3.1.31.6. Apolipoprotein E (APOE) promoter
491A/T,
427T/C,
219T/
G polymorphisms. The association between APOE promoter

491A/T,
427T/C,
219T/G polymorphisms and risk of AD was
reviewed by Xin et al. (2010). A total of 32 studies, representing
5789 cases and 5962 controls, provided data pertaining to the
159

491A/T polymorphism. The meta-analysis for the allelic
comparison detected an increased risk of AD associated with
the A allele (OR = 1.45; 95% CI = 1.27, 1.65). The pooled analysis
comparing the AA genotype to the AT genotype yielded an OR of
1.47 (95% CI = 1.27, 1.70). Those carrying the AT genotype had a
marginally significant 25% (OR = 1.25; 95% CI = 1.00, 1.56) increased
odds of AD when compared to TT genotype carriers.
491A
homozygotes had a 72% (OR = 1.72; 95% CI = 1.39, 2.13) increased
odds of AD when compared to
491T homozygotes. When
491A
homozygotes were compared to the AT and TT (AT + TT) genotypes,

491A homozygotes had a 1.49 times (OR = 1.49; 95% CI = 1.29,
1.72; test for heterogeneity: x2 = 77.34, p < 0.001) increased odds
of AD. Finally, in comparison to the TT genotype, the AA and AT
(AA + AT) genotypes were associated with a 57% increased odds of
AD (OR = 1.57; 95% CI = 1.27, 1.94).
Fifteen studies provided data pertaining to the risk of AD
attributed to the
427T/C polymorphism. The series of allelic and
genotypic comparisons for the
427T/C polymorphism included
2962 cases and 2883 controls. All comparisons were non-
significant (see Table 28 in Supplementary material II)A total of
16 studies representing 6614 cases and 6363 controls provided
data on the association between the
219T/G polymorphism and
AD. Each meta-analytic comparison yielded statistically significant
findings with ORs ranging from 1.21 to 1.51.
3.1.31.7. Apolipoprotein E (APOE) e4. We identified two systematic
reviews (Elias-Sonnenschein et al., 2011; Hao et al., 2011a,b) and
five cohort studies (Reitz et al., 2010b; Chu et al., 2010; Blasko et al.,
2010; Ronnemaa et al., 2011; Yang et al., 2011) that examined the
association between APOE e4 and risk of AD.
3.1.31.7.1. Systematic reviews. Thirty-five studies, reporting on the
association between APOE e4 and risk of progression from MCI to
AD, were eligible for inclusion in the systematic review conducted
by Elias-Sonnenschein et al. (2011). According to the random
effects meta-analysis, possessing at least one APOE e4 allele was
associated with a 2.29 times (95% CI = 1.58, 4.42) increased risk of
progression from MCI to AD. When stratified by study type,
possessing at least one APOE e4 allele was associated with a 2.36,
1.83, and 2.64 times increased risk of progression from MCI to AD
according to 21 clinical studies, six population-based studies, and
eight ‘other’ studies, respectively.
The systematic review reported by Hao et al. (2011a,b)
examined the association between APOE e4 and risk of AD in
East Asians. Three studies were eligible for inclusion in a random
effects meta-analysis. According to the pooled analysis, APOE e4
carriers had a near 3-fold increased risk of AD (OR = 2.98; 95%
CI = 1.84, 4.84; test for heterogeneity: x2 = 6.05, p = 0.05, I2 = 67%).
3.1.31.7.2. Primary studies. Four studies detected an increased risk
of AD associated with the APOE e4. The longitudinal cohort study
conducted by Yang et al. (2011) detected an increased risk of AD
among those possessing one or more APOE e4 allele, with risk
estimates ranging from 1.65 to 1.69. In an all-male sample of
Chinese subjects (Chu et al., 2010), APOE e4 carriers had a five-fold
increased risk of AD compared to non-carriers (RR = 5.04; 95%
CI = 1.05, 24.10). Possession of an APOE e4 allele, ascertained in
mid- and late-life, was associated with an increased risk of AD
according to findings from the Uppsala Longitudinal Study of Adult
Men (Midlife, HR = 2.60; 95% CI = 1.70, 3.90. Late-life, HR = 2.50; 95%
CI = 1.60, 3.80) (Ronnemaa et al., 2011). Although Reitz et al.
(2010b) detected an increased risk of probable and possible LOAD
among carriers of one or more APOE e4 allele (HR = 1.83; 95%
CI = 1.02, 3.29), results failed to reach statistical significance when
the analysis was restricted to probable LOAD (HR = 1.85; 95%
160 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
CI = 0.96, 3.57). The remaining study (Blasko et al., 2010) failed to
detect an association (OR = 2.30; 95% CI = 0.90, 6.00).
3.1.31.8. Apolipoprotein E (APOE) e2. We did not identify any
systematic reviews which reported on the association between
APOE e2 and risk of AD. However, one relevant cohort study was
identified in the second phase of the review (Yang et al., 2011).
Results from the AlzGene database were also considered.
The longitudinal cohort study conducted by Yang et al. (2011)
examined the association between blood pressure, hypertension,
and risk of AD. The association between blood pressure measures,
namely, self-reported hypertension and objectively measured
systolic (SBP) and diastolic blood pressure (DBP), pulse pressure
(PP), and mean arterial pressure (MAP), were analyzed in separate
models. The association between APOE e2 status and risk of AD was
examined in each model. According to the risk estimates derived
from each model, APOE e2 was not associated with risk of AD.
Results from a pooled analysis (n = 37) summarized in the AlzGene
database (see “AlzGene Top Results” summary) was suggestive of a
protective effect associated with APOE e2 (OR = 0.62; 95% CI = 0.46,
0.85; I2 = 64%).
3.1.31.9. Aldehydedehydrogenase 2 gene (ALDH2) Glu504Lys
polymorphism. Hao et al. (2011a,b) reviewed the association
between the aldehyde dehydrogenase 2 gene (ALDH2) Glu504Lys
polymorphism and risk of AD in East Asians. Four studies were
eligible for inclusion. Random effects meta-analysis representing
821 cases and 1380 controls failed to detect an association
(OR = 1.35; 95% CI = 0.75, 2.42; test for heterogeneity: x2 = 23.01,
p < 0.0001, I2 = 87%). The presence of the ALDH2 genotype
increased the risk of AD in males (OR = 1.72; 95% CI = 1.10, 2.67)
but not females (OR = 0.91; 95% CI = 0.33, 2.49) according to
subgroup analysis of two studies. Results of the review support an
association between the GA/AA genotype of ALDH2 and an
increased risk of AD in East Asian men (Hao et al., 2011a,b).
3.1.31.10. Methylenetetrahydrofolatereductase C677T polymorphism
We identified two moderate-quality systematic reviews which
reported on the association between methylenetetrahydrofolater-
eductase C677T polymorphism and risk of AD (Zhang et al., 2010;
Hua et al., 2011).
Nineteen studies satisfied the eligibility criteria of the review
reported by Zhang et al. (2010). In the pooled analysis examining
the allelic comparison, the T allele of the MTHFR C677T
polymorphism was associated with a 15% (OR = 1.15; 95% CI = 1.06,
1.26) increased odds of AD. However, the subgroup analysis by
ethnicity suggested that the T allele increased risk of AD in East
Asians (OR = 1.22; 95% CI = 1.08, 1.39) and in “other populations”
(OR = 1.56; 95% CI = 1.14; 2.12) but not in Caucasians (OR = 1.05; 95%
CI = 0.93, 1.18). T allele carriers were at an increased risk of AD when
compared to those carrying the CC genotype (OR = 1.21; 95%
CI = 1.07, 1.36). TT genotype carriers had a 32% increased risk of AD
compared to CC genotype carriers (OR = 1.32; 95% CI = 1.11, 1.58).
There was no significant difference between TT and CT + CC
genotypes with respect to AD risk (OR = 0.86; 95% CI = 0.69, 1.06).
Twenty-seven studies, 13 of which were conducted in Asian
populations, were included in the systematic review reported by
Hua et al. (2011). According to random effects meta-analysis, the T
allele was associated with a 13% (OR = 1.13; 95% CI = 1.04, 1.24)
increased odds of AD when compared to the C allele. This
association was true for Asians (OR = 1.28; 95% CI = 1.07, 1.52; test
for heterogeneity: p = 0.002, I2 = 61.6%) but not Caucasians (OR =
1.02; 95% CI = 0.92, 1.12). Compared to the CT + CC genotypes, the
TT genotype was not associated with risk of AD (OR = 1.09; 95%
CI = 0.95, 1.26). In comparison with CC genotype carriers, those
carrying the T allele had an 18% (OR = 1.18; 95% CI = 1.06, 1.32)
increased risk of AD. However, an association was detected for
Asians (OR = 1.37; 95% CI = 1.12, 1.68) but not Caucasians (OR = 1.05;
95% CI = 0.92, 1.21).
3.1.31.11. Neuronal sortilin-related receptor gene (SORL1). We
identified two systematic reviews which reported on the
association between neuronal sortilin-related receptor gene
(SORL1) and risk of AD (Ning et al., 2010; Reitz et al., 2011).
Ning et al. (2010) examined the association between neuronal
SORL1 and LOAD. Four studies were eligible for inclusion. Random
effects meta-analysis of three studies detected an association
between SNP rs2070045 and LOAD (OR = 1.35; 95% CI = 1.11, 1.65). In
the pooled analysis of all four studies, representing 984 LOAD cases
and 1320 control subjects, rs3824968 (OR = 1.29; 95% CI = 1.14, 1.46)
and rs2282649 (OR = 1.34; 95% CI = 1.19, 1.52) were associated with
an increased risk of LOAD.
A total of 17 studies representing 20 data sets were eligible for
inclusion in the review reported by Reitz et al. (2011). A series of
meta-analyses were conducted to examine the association
between AD and SORL1 SNPs 4, 5, 8, 9, 10, 12, 19, 22, 23, 24, and
25. SNPs 4, 5, 8, 9, 10, 12, and 19 were significantly associated with
AD in the Caucasian data sets (SNPs 5, 8, 9, 10, 19 increased risk
while SNPs 4 and 12 reduced risk). SNPs 19, 23, 24, and 25 were
associated with risk of AD in Asians (SNPs 19 and 25 increased risk
while SNPs 23 and 24 were protective).
3.1.31.12. Cathepsin D (CTSD) Ala224Val gene polymorphism. Ntais
et al. (2004) reviewed the association between cathepsin D
Ala224Val gene polymorphism (CTSD) and risk of AD. Fourteen
studies were eligible for inclusion. The random effects meta-
analysis failed to detect a statistically significant difference in AD
risk between the alleles (OR = 1.17; 95% CI = 0.95, 1.44; test for
heterogeneity: p < 0.01). The remaining comparisons also failed to
yield statistically significant findings (see Table 34 in
Supplementary material II).
3.1.31.13. ACE D/I polymorphism. A total of 23 studies, which
examined the association between the I allele of the angiotensin-
converting enzyme (ACE) D/I polymorphism and LOAD, were
eligible for inclusion in the systematic review reported by Elkins
et al. (2004). When the results of all studies were combined in a
random effects meta-analysis, the I allele of the ACE D/I
polymorphism was associated with an increased risk of AD
(OR = 1.27; 95% CI = 1.10, 1.47; test for heterogeneity: p < 0.001).
When compared to carriers of the DD genotype, those with the II
(OR = 1.27; 95% CI = 1.04, 1.54) and I/D (OR = 1.27; 95% CI = 1.09, 1.47)
genotypes were also at an increased risk.
3.1.31.14. Tumor Necrosis Factor (TNF)
a polymorphisms. Di Bona
et al. (2009) systematically reviewed the literature to ascertain the
association between tumor necrosis factor (TNF)-alpha
polymorphisms and risk of AD. According to random effects
meta-analyses of six studies, TNF-alpha-308 was not associated
with risk of AD. When compared to the CC (OR = 1.63; 95% CI = 1.07,
2.49) and CT + CC genotypes (OR = 1.57; 95% CI = 1.08, 2.29), the TT
genotype of TNF-alpha-850 was significantly associated with risk
of AD. The remaining comparisons for TNF-alpha-850 failed to
reach statistical significance. When results from the four studies
examining TNF-alpha-863 SNP were combined in a random effects
meta-analysis, there was no statistically significant difference in
AD risk between the CC genotype and the AC + AA genotypes
(OR = 1.26; 95% CI = 0.86, 1.85). Similarly, results from the pooled
analysis comparing the AG + AA genotypes of the TNF-alpha-238
SNP to the GG genotype were not statistically significant
(OR = 0.78; 95% CI = 0.57, 1.08). Finally, the pooled analysis
examining the effect of the
1031 SNP failed to detect a
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
statistically significant difference in AD risk between the TT
genotype and the CT + CC genotypes (OR = 0.62; 95% CI = 0.36, 1.09).
Overall, results of the review provide no evidence to support an
association between AD and TNF-alpha SNPs
308,
863,
238,
and
1031. However, the TT genotype of the
850 SNP was
associated with an increased risk of AD.
3.1.31.15. PRNP M129V polymorphism. One systematic review
examined the effect of the PRNP M129 V polymorphism on the
risk of AD (Del Bo et al., 2006). The study was excluded due to low
methodological quality. Consequently, there was inadequate
evidence to comment on the effect of this polymorphism.
3.1.31.16. b-amyloid related genes. According to the series of meta-
analyses conducted by Llorca et al. (2008), only the C allele of the
OLR1 30 -UTR (+1073) polymorphism was associated with AD. The
overall pooled analyses for the remaining polymorphisms (BACE1,
CYP47, FE65, ACT Ala17Thr, BH 1443 V, OLR1 30 -UTR (+1071), VLDLR
50 -UTR (CGG-repeat)) failed to yield statistically significant
findings. However, according to subgroup analyses, the
association between the CC + CG genotype of the BACE1exon 5
(G/C) polymorphism was dependent on APOE status. More
specifically, the CC + CG genotype of the BACE1 exon 5 (G/C)
polymorphism was associated with a protective effect against AD
among APOE e4 carriers (OR = 0.57; 95% CI = 0.38, 0.88) and a
marginally significant increased risk of AD (OR = 1.33; 95% CI = 1.00,
1.78) among APOE e4 non-carriers. Similarly, the CC genotype of
the BACE1 exon 5 (G/C) polymorphism was associated with a
protective effect against AD among APOE e4 carriers (OR = 0.48;
95% CI = 0.26, 0.88) and an increased risk of AD among APOE e4
non-carriers (OR = 1.61; 95% CI = 1.11, 2.34). Although the VLDLR 50 -
UTR (CGG-repeat) polymorphism was not associated with AD in
the overall pooled analysis, subgroup analysis revealed that
genotype 2 was associated with a significant increase risk of AD
among Asians (OR = 1.70; 95% CI = 1.09, 2.63) and a reduced risk in
non-Asians (OR = 0.48; 95% CI = 0.26, 0.86).
3.1.31.17. AlzGene “Top Results”. Our systematic search of the
literature was supplemented with a review of the AlzGene
database, a frequently updated, publically available, online
database examining genetic variants associated with AD
(Khoury et al., 2009; Bertram et al., 2007). Studies included in
the database are obtained through a comprehensive, systematic
review of the literature. Pooled analyses are regularly updated with
newly emerging data (Bertram et al., 2007).
Currently, the AlzGene database summarizes evidence from
1395 primary studies covering 695 genes and 2973 polymor-
phisms. Moreover, 320 meta-analyses have been conducted to
date. Only evidence summarized in the “Top Results” section of the
database were reviewed. Briefly, this section presents polymor-
phisms which are “most strongly associated” with Alzheimer’s
disease according to a quantitative synthesis of the available
literature (Bertram et al., 2007).
The most strongly associated genes, listed in order of decreasing
strength of association according to p-value, were APOE e2/3/4,
BIN1, CLU, ABCA7, CR1, PICALM, MS4A6A, CD33, MS4A4E, and
CD2AP. Compared to the e3 allele, APOE e4 was associated with
3.69 increased risk of AD (95% CI = 3.30, 4.12, p-value = < 1E-50;
I2 = 30%). Polymorphism rs744373 of BIN1 was associated with a
1.17 times increased risk of AD (95% CI = 1.13, 1.20, p-value = 1.59E-
26; I2 = 0%). The rs3764650, rs670139, and rs9349407 polymor-
phisms of the ABCA7, MS4A4E, and CD2AP genes, respectively,
were all associated with an increased risk of AD (ABCA7, OR = 1.23;
95% CI = 1.18, 1.28, p-value = 8.17E-22; I2 = 0%. MS4A4E, OR = 1.08;
95% CI = 1.05, 1.11, p-value = 9.51E-10; I2 = 2%. CD2AP, OR = 1.12; 95%
CI = 1.08, 1.16, p-value = 2.75E-09; I 2 = 0%). APOE e2, rs610932
161
polymorphism of MS4A6A and rs3865444 polymorphism of CD33
were associated with a reduced risk of AD (APOE e2, OR = 0.62; 95%
CI = 0.46, 0.85; I2 = 64%. MS4A6A, OR = 0.90; 95% CI = 0.88, 0.93, p-
value = 1.81E-11; I2 = 23%. CD33, OR = 0.89; 95% CI = 0.86, 0.93, p-
value = 2.04E-10; I2 = 0%). Finally, the rs11136000, rs3818361, and
rs3851179 polymorphisms of CLU, CR1, and PICALM, respectively,
were associated with AD among Caucasians (CLU, OR = 0.88; 95%
CI = 0.86, 0.90, p-value = 3.37E-23; I2 = 0%. CR1, OR = 1.17; 95%
CI = 1.14, 1.21, p-value = 4.72E-21; I2 = 0%. PICALM, OR = 0.88; 95%
CI = 0.86, 0.90, p-value = 2.85E-20; I2 = 0%).
3.1.32. Alcohol
We identified ten systematic reviews reporting on the
association between alcohol consumption and risk of AD. Three
reviews were later excluded due to low methodological quality
(Standridge, 2004; Purnell et al., 2009; ESHRE Capri Workshop
Group, 2011). Results of the remaining seven moderate-quality
systematic reviews were used to evaluate the association between
alcohol consumption and risk of AD (Graves et al., 1991a; Weih
et al., 2007; Patterson et al., 2007; Peters et al., 2008b; Anstey et al.,
2009; Williams et al., 2010; Lee et al., 2010). Furthermore, two
primary studies were identified (Weyerer et al., 2011; Zhou et al.,
2011) (see Tables 38 and 107 in Supplementary material II).
3.1.32.1. Systematic reviews. The systematic review by Graves et al.
(1991a) examined the effect of smoking and alcohol consumption
on the risk of AD. Five case-control studies were eligible for
inclusion in the pooled analysis. Results of the pooled analysis,
adjusted for family history of dementia, suggested that when
compared to abstinence, mild (RR = 0.97; 95% CI = 0.59, 1.60),
moderate (RR = 0.75; 95% CI = 0.46, 1.20), and high (RR = 0.89; 95%
CI = 0.52, 1.51) alcohol consumption was not associated with a
reduced risk of AD. A more rigorous attempt to control for potential
confounders, namely head injury, education, and family history of
dementia, yielded consistent findings suggesting a lack of
association between alcohol consumption and risk of AD (Mild,
RR = 0.89; 95% CI = 0.59, 1.47. Moderate, RR = 0.82; 95% CI = 0.49,
1.39. Heavy, RR = 0.95; 95% CI = 0.51, 1.81). Although all risk
estimates failed to reach statistical significance, the reviewers
note that moderate alcohol consumption produced the lowest
relative risk. Finally, the sensitivity analysis excluding the
methodologically heterogeneous study also failed to reach
statistical significance for all three categories of alcohol
consumption (Mild, RR = 0.85; 95% CI = 0.48, 1.50. Moderate,
RR = 1.48; 95% CI = 0.67, 3.23. Heavy, RR = 0.68; 95% CI = 0.31,
1.48). Overall, results of the systematic review reported by
Graves et al. (1991a), suggested that alcohol consumption was
not associated with risk of AD
Eight studies were eligible for inclusion in the systematic
review reported by Weih et al. (2007). Two studies reported a
reduced risk of AD associated with mild and moderate alcohol
consumption. All three studies reporting on wine consumption
detected a protective effect. One study reported a detrimental
effect attributed to monthly beer consumption. The remaining
studies failed to detect an association. Overall, findings of the
review suggest that mild to moderate alcohol consumption may
reduce risk of AD, with a greater benefit associated with moderate
consumption.
Two longitudinal cohort studies were eligible for inclusion in
the systematic review reported by Patterson et al. (2007). Both
studies detected a reduced risk of AD associated with wine
consumption. One study detected a 47% (RR = 0.53; 95% CI = 0.30,
0.95) reduced risk of AD associated with moderate (250 to 500 ml/
day) wine consumption. The other reported consistent findings
(OR = 0.49; 95% CI = not reported).
162 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Eight longitudinal studies, which reported on risk of incident
AD, were eligible for inclusion in the pooled analysis reported by
Peters et al. (2008b). Results of the random effects meta-analysis
suggested that alcohol consumption was associated with a
protective effect against the development of AD (RR = 0.57; 95%
CI = 0.44, 0.74). However, there was evidence of heterogeneity
across studies (I2 = 46%, Cochran Q = 22.23, df = 12, P = 0.035) likely
attributed to the heterogeneous definitions of exposure. There was
no evidence of publication bias (Begg-Mazumdar test:
Kendall's t =
0.03, P = 0.86; Egger’s test: 0.27; 95% CI =
2.16,
2.71, P = 0.81).
A total of nine studies, identified by Anstey et al. (2009),
provided data on the risk of AD associated with alcohol
consumption. Six studies reported on the effect of light to
moderate alcohol consumption. The pooled analysis detected a
28% reduced risk of AD associated with light to moderate alcohol
consumption (RR = 0.72; 95% CI: 0.61, 0.86; Test for heterogeneity:
x2 = 11.43, p = 0.04). Four studies evaluated the effect of heavy/
excessive alcohol consumption. Heavy alcohol consumption was
not associated with risk of AD according to the pooled analysis
(RR = 0.92; 95% CI 0.59, 1.45; test for heterogeneity = not statisti-
cally significant). As noted by the reviewers, the failure to detect an
association may be attributed to sources of selection bias. Two
studies compared the risk of AD among drinkers compared to non-
drinkers. Results of the pooled analysis suggested a 34% decreased
risk (RR = 0.66; 95% CI = 0.47, 0.94) of AD associated with alcohol
consumption. The analysis, however, was limited by the small
number of studies suitable for inclusion (Anstey et al., 2009).
The HTA report by Williams et al. (2010) examined the effect of
alcohol consumption on risk of AD. The report identified one
systematic review (Anstey et al., 2009). According to findings of the
identified review, the report summarizes that light to moderate,
but not heavy, alcohol consumption protects against the develop-
ment of AD.
Lee et al. (2010) identified two studies which reported on the
association between alcohol consumption and risk of AD. Alcohol
consumption, defined as intake of sake one to two times per week,
was not associated with AD risk according to one of the two studies
(RR = not reported). The second study reported that in comparison
to abstinence, light (1 serving/month
6 servings/week) to
moderate (1–3 servings/day) wine consumption was associated
with a 45% (HR = 0.55; 95% CI = 0.34, 0.89) decreased risk of AD.
3.1.32.2. Primary studies. The effect of current alcohol
consumption in late-life was examined in the prospective cohort
study conducted by Weyerer et al. (2011). The effect of varying
quantities and types of alcoholic beverages were also examined.
Abstainers served as the referent category for all analyses. When
alcohol consumption was treated as a dichotomous variable,
current alcohol consumption was associated with a statistically
significant reduced risk of AD (HR = 0.58; 95% CI = 0.38, 0.89).
Stratification by quantity of alcohol revealed a reduced risk of AD
among those reporting consumption of 20 to 29 g/day (HR = 0.13;
95% CI = 0.02, 0.95). The remaining categories did not reach
statistical significance (1.9 g/day, HR = 0.61; 95% CI = 0.36, 1.01. 10–
19 g/day, HR = 0.72; 95% CI = 0.37, 1.39. 30–39 g/day, HR = 0.33; 95%
CI = 0.05, 2.41. 40 g/day, HR = 0.68; 95% CI = 0.16, 2.90). With
respect to the type of alcoholic beverage, only those reporting
consumption of a variety of drinks (wine, beer, spirits) were at a
reduced risk of AD (HR = 0.14; 95% CI = 0.03, 0.56). The “wine only”
and “beer only” categories failed to reach statistical significance
(Wine only, HR = 0.76; 95% CI = 0.46, 1.23. Beer only, HR = 0.60; 95%
CI = 0.30, 1.21). Based on these results, the authors concluded that
late-life alcohol consumption, in light-to-moderate quantities,
appeared to be associated with a protective effect against the
development of AD.
Compared to occasional drinking, only daily alcohol consump-
tion was associated with risk of AD (HR = 1.72; 95% CI = 1.07, 2.75)
according to the cohort study reported by Zhou et al. (2011).
3.1.32.3. Discussion. Evidence summarized in a majority of the
identified systematic reviews suggest a reduced risk of AD
associated with alcohol consumption. Results from two recent
cohort studies confirm the association between alcohol
consumption and risk of AD. One study detected a reduced risk
of AD associated with light-to-moderate alcohol consumption in
late-life (Weyerer et al., 2011). The other reported a detrimental
effect associated with daily alcohol consumption (Zhou et al.,
2011). Considered cumulatively, alcohol consumption appears to
reduce the risk of AD with the most convincing evidence for light-
to-moderate quantities of consumption.
3.1.33. Smoking
Twelve systematic reviews examined the association between
smoking and risk of Alzheimer’s disease (Graves et al., 1991a;
Almeida et al., 2002; Patterson et al., 2007; Weih et al., 2007;
Anstey et al., 2007; Peters et al., 2008c; Purnell et al., 2009;
Williams et al., 2010; Cataldo et al., 2010; Lee et al., 2010; ESHRE
Capri Workshop Group, 2011; Barnes and Yaffe, 2011). Three
reviews received a low quality score and were excluded (Purnell
et al., 2009; ESHRE Capri Workshop Group, 2011; Barnes and Yaffe,
2011). The remaining nine reviews were deemed to be of moderate
methodological quality with AMSTAR scores ranging from 4 to 7.
Our search for recent observational studies yielded six studies that
reported on the effect of smoking.
3.1.33.1. Systematic reviews. The systematic review by Graves et al.
(1991a) examined the effect of smoking and alcohol consumption
on the risk of AD. According to the pooled analysis, having ever
smoked was associated with a statistically significant 22%
(RR = 0.78; 95% CI = 0.62, 0.98) reduced risk of AD. Smoking less
than one pack per day was associated with a 0.84 times (95%
CI = 0.75, 0.92) reduced risk of AD while excessive smoking ( > 1
pack/day) was not protective (RR = 0.81; 95% CI = 0.53, 1.27). When
the outlying study was excluded from the analysis, however,
smoking less than one pack per day was no longer associated with
a reduced risk of AD (RR = 0.80; 95% CI = 0.60, 1.07) and excessive
smoking was associated with a marginally significant reduced risk
(RR = 0.66; 95% CI = 0.44, 1.00). Compared to never smokers, those
with a low pack-year smoking history (<15.5 pack-years) did not
have a statistically significant reduced risk of AD (RR = 0.73; 95%
CI = 0.50, 1.09). In contrast, moderate (15.5–37.0) and high (>37.0)
pack-year smoking histories were associated with a statistically
significant 0.63 (95% CI = 0.42, 0.95) and 0.50 (95% CI = 0.31, 0.80)
reduced risk of AD, respectively. The trend was statistically
significant (p = 0.0003) suggesting that risk of AD decreased
with increasing pack-year history. Based on an analysis of eight
case-control studies, the reviewers concluded that smoking
appeared to be associated with a protective effect against the
development of AD. The reviewers cautioned that prospective
studies were required to confirm these findings.
The systematic review and meta-analysis conducted by
Almeida et al. (2002) examined the effect of ever smoking on
risk of AD. A total of 21 case-control and eight cohort studies were
eligible for inclusion in the review. Results from the fixed effects
meta-analysis including all 21 case-control studies (n = 5323),
irrespective of study quality, suggested that ever smoking was
associated with a statistically significant decreased odds (OR =
0.74; 95% CI = 0.66, 0.84; test for heterogeneity: p = 0.246) of AD.
Results from the overall pooled analysis of case-control studies
suggest that history of smoking was associated with a protective
effect against AD. However, the series of sensitivity analyses
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
restricted to the most methodologically rigorous case-control
studies failed to detect an association, suggesting that the observed
protective effect may be due to sources of bias. Furthermore,
analyses restricted to cohort studies yielded risk estimates
exceeding 1.00, suggesting that smokers were at an increased
risk of AD. Although cohort studies are generally less susceptible to
bias, the analysis of cohort studies was based on findings from a
limited number of heterogeneous studies (Almeida et al., 2002).
Due to these inconsistent findings, the reviewers concluded that
the association between smoking and risk of AD was inconclusive.
The systematic review and meta-analysis by Anstey et al. (2007)
examined the association between smoking and risk of AD and
other cognitive outcomes. Ten studies, representing 13786
participants, provided data pertaining to the effect of smoking
on the development of AD. Current smokers had an increased risk
of AD when compared to never (RR = 1.79; 95% CI = 1.43, 2.23; test
for heterogeneity: p = 0.50) and former (RR = 1.70; 95% CI = 1.25,
2.31; test for heterogeneity: p = 0.60) smokers. The pooled risk
estimate did not reach statistical significance, however, when
current smoking was compared to the combined category of past
and never smoking (RR = 1.25; 95% CI = 0.49, 3.17; test for
heterogeneity: p = 0.03). The aforementioned analysis was limited
by the small number of studies (n = 2) reporting on this association
as well as obvious heterogeneity (Anstey et al., 2007). Compared to
never smoking, ever (RR = 1.21; 95% CI = 0.66, 2.22; test for
heterogeneity: p = 0.01) and former (RR = 1.01; 95% CI = 0.83,
1.23; test for heterogeneity: p = 0.26) smoking was not associated
with AD. According to the reviewers, the former and ever smoking
exposure categories were poorly defined and may include a group
of people with diverse smoking histories with respect to quantity
and duration of exposure (Anstey et al., 2007). These heteroge-
neous exposure categories serve as a limitation of the review and
may offer an explanation for the null findings for these exposure
groups (Anstey et al., 2007).
The systematic review conducted by Patterson et al. (2007)
examined the effect of several putative risk factors on the
development of AD and other cognitive outcomes, including
unspecified dementia and vascular dementia. One cohort study
and one systematic review were identified. The cohort study failed
to detect an association (RR = 1.10; 95% CI = 0.94, 1.29). The
identified systematic review and meta-analysis (Almeida et al.,
2002) restricted to methodologically sound prospective cohort
studies detected a two-fold increased risk of AD associated with
smoking (OR = 1.99; 95% CI = 1.33, 2.98).
Five longitudinal studies were identified by Weih et al. (2007).
All three studies that examined the effect of current smoking
reported an increased risk of AD. Smoking was not associated with
risk of AD according to the remaining two studies.
The effect of current and former smoking on the risk of AD and
other cognitive outcomes were reviewed by Peters et al. (2008c).
Eleven studies reporting on the association between smoking and
risk of AD were identified. Seven of the ten studies reported a
statistically significant increased risk of AD associated with current
smoking. According to the random effects meta-analysis, current
smoking was associated with an increased risk of AD (RR = 1.59;
95% CI = 1.15, 2.20; Cochran Q = 23.25, p = 0.0015). All eight studies
which examined the effect of former smoking on AD risk failed to
detect an association. In the random effects meta-analysis, former
smokers were not at an increased risk of developing AD when
compared to non-smokers (RR = 0.99; 95% CI = 0.81, 1.23; Cochran
Q = 11.27, p = 0.08).
The systematic review and meta-analysis by Cataldo et al.
(2010) examined the risk of Alzheimer's disease associated with
cigarette smoking. The reviewers adjusted for industry affiliation, a
factor which may influence study findings, introduce bias, and lead
to spurious results (Cataldo et al., 2010). A total of 43 studies (26
163
case-control and 17 cohort studies) satisfied the inclusion criteria
of the review. The pooled analysis of affiliated cohort studies failed
to detect an association (RR = 0.60; 95% CI: 0.27, 1.32), however,
industry affiliated case-control studies detected a reduced risk of
AD associated with cigarette smoke (OR = 0.86; 95% CI: 0.75, 0.98).
The pooled analysis of unaffiliated case-control (OR = 0.91; 95%
CI = 0.75, 1.10) failed to detect an association. Alternatively, when
unaffiliated cohort studies were quantitatively synthesized,
smoking was associated with a detrimental effect (RR = 1.45;
95% CI = 1.16, 1.80). Overall, when the analysis was restricted to
unaffiliated cohort studies, smoking appeared to be associated
with a detrimental effect (Cataldo et al., 2010).
Lee et al. (2010) systematically reviewed the literature for
longitudinal, community-cohort studies which examined the risk
of AD associated with smoking among elderly subjects (65 years).
Smoking was consistently associated with an increased risk of
AD in all four included cohort studies.
The HTA report by Williams et al. (2010) examined the effect of
smoking on risk of AD.
The report identified two relevant systematic reviews (Peters
et al., 2008c; Anstey et al., 2007), however, only findings reported
by Anstey et al. (2011) were considered. According to the results of
the identified review, current but not former smoking was
associated with an increased risk of AD. Two cohort studies were
also identified and included in the report. Both studies reported
risk estimates for current and former smoking and also presented
results stratified by APOE e4 status (Williams et al., 2010). Both
studies detected an increased risk of AD associated with current
but not former smokers. Both studies reported an increased risk of
AD associated with current smoking among APOE e4 non-carriers
but not among e4 allele carries (Williams et al., 2010). Based on the
available literature, the reviewers concluded that current smoking
was associated with an increased risk of AD.
3.1.33.2. Primary studies. Current (HR = 2.03; 95% CI = 1.34, 3.09)
but not past (HR = 1.01; 95% CI = 0.66, 1.55) smoking was associated
with an increased risk of AD according to the prospective cohort
study of Chinese subjects reported by Zhou et al. (2011).
Current smoking was significantly associated with risk of
probable LOAD according to results reported by Reitz et al. (2010b)
(HR = 2.52; 95% CI = 1.03, 6.19). Results failed to reach statistical
significance, however, when possible LOAD was included in the
analysis (HR = 1.98; 95% CI = 0.88, 4.49).
The nested case-control study by Gelber et al. (2012) examined
the effect of midlife lifestyle risk factors, namely, BMI, smoking
status, diet, and physical activity on the development of AD among
Japanese-American men enrolled in the Honolulu-Asia Aging
Study. Former (OR = 1.15; 95% CI = 0.71, 1.87) and current (OR = 0.97;
95% CI = 0.59, 1.60) smoking categories were not associated with
risk of AD when compared to never smoking.
The study reported by Ronnemaa et al. (2011) examined the risk
of AD associated with vascular risk factors considered individually
and in combination. Current smoking status, reported at midlife
and late-life, was not associated with risk of AD when compared to
non-smoking (Midlife, HR = 1.00; 95% CI = 0.70, 1.40. Late-life,
HR = 0.80; 95% CI = 0.40, 1.20).
The association between vascular risk factors and risk of AD
among Koreans was prospectively examined by Kimm et al. (2011).
Using never smoking as the referent category, both former
(HR = 1.00; 95% CI = 0.80, 1.20) and current (HR = 1.10; 95% CI =
0.90, 3.40) smoking were not associated with risk of AD among
men. Alternatively, current (HR = 1.30; 95% CI = 1.10, 1.50) but not
former (HR = 1.20; 95% CI = 0.90, 1.50) smoking increased risk of AD
in women. Stratified analysis detected a marginally significant
association between current smoking and risk of AD among
women 65 years of age and older (HR = 1.20; 95% CI = 1.00, 1.50) and
164 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
no association among those less than 65 years of age (HR = 1.40;
95% CI = 0.90, 2.10). Smoking was marginally associated with risk of
AD among men less than 65 years of age (HR = 1.30; 95% CI = 1.00,
1.70) but not among those 65 years or older (HR = 0.90; 95%
CI = 0.70, 1.10).
The prospective cohort study reported by Rusanen et al. (2011)
examined the effect of midlife smoking on the risk of developing
AD in late-life. Risk estimates were derived for former smoking and
for current smoking, which was stratified by quantity/pack(s)
smoked per day. Using non-smokers as the referent category,
former smokers were not at an increased risk of developing AD
(HR = 1.00; 95% CI = 0.89, 1.13). With respect to current smoking,
only the heaviest smokers (2 packs/day) were at a significantly
increased risk of AD (HR = 2.57; 95% CI = 1.63, 4.03). The remaining
current smoking categories were not significantly associated with
risk of AD, however, risk estimates exceeded 1.00 for the 0.5-1
pack/day and 1–2 pack/day categories. Overall, the authors
concluded that midlife smoking, particularly in heavy quantities,
increased risk of AD in late-life.
3.1.33.3. Discussion. Considered cumulatively, the identified
systematic reviews consistently concluded that smoking
increased risk of AD. The increased risk of AD, however, was
only apparent in current smokers. Four of the six observational
studies identified in the second phase of the review detected a
detrimental effect attributed to current smoking status.
Former smokers, or those reporting to have ever smoked, were
not at an increased risk of developing AD. As suggested, failure to
detect an association may be attributed to the fact that these
exposure categories were poorly defined (Anstey et al., 2007;
Peters et al., 2008c). Such heterogeneity may impede the detection
of an association between AD and former/ever smoking, if one
exists (Anstey et al., 2007; Peters et al., 2008c).
A number of case-control studies included in the identified
systematic reviews detected a reduced risk of AD attributed to
smoking. These findings were likely attributed to sources of bias
inherent in the methodological design of such studies (Cataldo
et al., 2010). This assumption was confirmed when findings of case-
control studies were compared to those derived from methodo-
logically rigorous prospective studies, which consistently yielded
conflicting findings (Cataldo et al., 2010). When the analysis was
restricted to cohort studies, there was consensus regarding the
detrimental effect of smoking.
3.1.34. Physical activity
We identified fifteen systematic reviews reporting on the effect
of physical activity (Podewils, 2003; Fratiglioni et al., 2004;
Penrose, 2005; Weih et al., 2007, 2010; Patterson et al., 2007;
Rolland et al., 2008; Hamer and Chida, 2009; Stern and Konno,
2009; Purnell et al., 2009; Williams et al., 2010; Lee et al., 2010;
Barnes and Yaffe, 2011; ESHRE Capri Workshop Group, 2011;
Hurley et al., 2011). Six reviews were deemed to be of low
methodological quality according to the AMSTAR criteria (Frati-
glioni et al., 2004; Penrose, 2005; Purnell et al., 2009; Hurley et al.,
2011; ESHRE Capri Workshop Group, 2011; Barnes and Yaffe, 2011).
Data was abstracted from the remaining nine systematic reviews
and results were qualitatively synthesized. One primary study was
identified in the second phase of the review (Gelber et al., 2012)
(see Tables 40 and 104 in Supplementary material II).
3.1.34.1. Systematic reviews. The effect of physical activity on risk
of all-cause dementia, AD, and vascular dementia was examined in
the systematic review and meta-analysis by Podewils et al. (2003).
Seven cohort and four case-control studies were eligible for
inclusion. According to the random effects meta-analysis, physical
activity was associated with a 41% (RR = 0.59; 95% CI = 0.42, 0.77)
and 76% (OR = 0.24; 95% CI = 0.09, 0.60) reduced risk of AD
according to analyses restricted to cohort and case-control
studies, respectively.
The systematic review by Weih et al. (2007) included seven
longitudinal studies. Physical activity had a protective effect
against the development of AD according to two studies. Four of
the five remaining studies reported risk estimates below 1.00,
however, results failed to reach statistical significance. Despite the
number of negative findings, the reviewers concluded that physical
activity may reduce risk of AD.
The systematic review conducted by Patterson et al. (2007)
examined the effect of several putative risk factors on the
development of AD and other cognitive outcomes, including
unspecified and vascular dementia. All three studies included in
the review consistently detected a reduced risk of AD attributed to
physical activity.
Ten studies identified by Rolland et al. (2008) reported on the
effect of physical activity on the risk of developing AD, although
risk estimates specific to AD were not specified for a majority of
studies. Seven of the ten studies detected an association between
physical activity, or at least one measure of activity, and risk of AD/
dementia. The remaining three studies failed to detect an
association. Based on the available evidence, the reviewers
concluded that physical activity may protect against the develop-
ment of AD and dementia.
The effect of physical activity on the development of AD, all-
cause dementia, and Parkinson’s disease was reviewed by Hamer
and Chida (2009). Six studies were eligible for inclusion in the
pooled analysis for the outcome of AD. Results of the random
effects meta-analysis, representing 13771 subjects, suggest that
physical activity was associated with a 45% reduced risk of AD
(RR = 0.55; 95% CI = 0.36, 0.84; test for heterogeneity: x2 = 29.12,
p< 0.001). Using subgroup analysis, the reviewers detected that
the protective association was more pronounced in methodologi-
cally rigorous studies. However, risk estimates for the subgroup
analysis were not reported for the outcome of AD.
The systematic review by Stern and Konno (2009) examined the
effect of physical leisure activities on the prevention of all-cause
dementia and AD. Two case-control and 11 cohort studies
reporting on the risk of AD were eligible for inclusion. The only
study that reported on activities during early adulthood deter-
mined that cases were less likely to engage in a range of activities
when compared to controls (OR = 2.67; 95% CI = 1.85, 3.85) (Stern
and Konno, 2009; Friedland et al., 2001). With respect to activities
engaged in during midlife, three of the four studies reporting on
the risk of AD detected a protective effect. Seven of the nine studies
reporting on the risk of AD associated with late-life engagement in
physical activities detected an association. Considered cumula-
tively, the reviewers concluded that the available evidence did not
consistently support a protective effect of physical activities for the
prevention of AD and all-cause dementia. When the analysis was
restricted to studies reporting on risk of AD specifically, three of
four and seven of nine studies reporting on midlife and late-life
physical activities, respectively, detected an association. The
protective effect was diluted when studies reporting solely on
all-cause dementia were also considered (3 of 5 for midlife
activities and 9 of 12 for late-life activities).
Lee et al. (2010) systematically reviewed the literature for
longitudinal, community-cohort studies which examined the risk
of AD associated with physical activity among elderly subjects
(65 years). Of the six relevant studies, four reported a statistically
significant association.
The HTA reviewed the available literature to examine the
association between physical activity and subsequent risk of AD
(Williams et al., 2010). Twelve cohort studies were identified, nine
of which were eligible for inclusion in a pooled analysis. The
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
random effects model yielded a risk estimate of 0.72 (95% CI = 0.53,
0.98; Q = 23.25, p = 0.003, I2 = 66%) suggesting a 28% reduced risk of
AD associated with engagement in physical activities. Based on the
available evidence, the reviewers concluded that physical activity
may reduce risk of AD. The reviewers caution, however, that the
observed protective effect may be attributed to confounding by
other putative protective factors that are associated with a
physically active lifestyle.
The effect of physical activity on risk of developing AD was
reviewed by Weih et al. (2010).
Six studies satisfied the inclusion criteria. All studies reported
risk estimates below 1.00 and all but one reached statistical
significance. According to the random effects meta-analysis,
compared to those with a sedentary lifestyle, physically active
subjects had a 41% reduced risk of AD (RR = 0.59; 95% CI = 0.50,
0.69). There was no evidence of statistical heterogeneity across
studies (test for heterogeneity: x2 = 1.84, df = 5, p = 0.87).
3.1.34.2. Primary studies. We identified a single observational
study reporting on the association between physical activity and
risk of AD. The nested case-control study by Gelber et al. (2012)
examined the effect of midlife lifestyle risk factors, namely, BMI,
smoking status, diet, and physical activity, on the development of
AD among Japanese-American men enrolled in the Honolulu-Asia
Aging Study. Physical activity, treated as a dichotomous variable
(high vs. low), was not significantly associated with risk of AD
(OR = 1.44; 95% CI = 0.94, 2.21).
3.1.34.3. Discussion. Eight of nine identified systematic reviews
consistently concluded that engagement in physical activities may
reduce risk of AD. The only identified observational study failed to
detect an association between physical activity and AD. Despite the
various limitations of observational studies, evidence collated in
several systematic reviews offer convincing evidence regarding the
protective effect of physical activity against the development of
AD.
3.1.35. Polyunsaturated fatty acids (PUFAs)
Nine reviews examining the association between polyunsatu-
rated fatty acids (PUFAs) and risk of AD were identified (Stand-
ridge, 2004; Issa et al., 2006; Lim et al., 2006; Weih et al., 2007;
Patterson et al., 2007; Fotuhi et al., 2009; Williams et al., 2010; Lee
et al., 2010; Dangour et al., 2010). Two of the reviews deemed to be
of low methodological quality were excluded (Standridge, 2004;
Fotuhi et al., 2009). One primary study was identified in the second
phase of the review (Lopez et al., 2011). Refer to
3.1.35.1. Systematic reviews. Issa et al. (2006) reviewed the effect
of omega-3 fatty acid intake on cognitive function and incidence of
dementia. Two of the three included studies reported statistically
significant findings suggesting a reduced risk of AD associated with
fish consumption. One study reported that total n-3 fatty acid
consumption and a high intake of DHA, but not EPA or ALA, was
associated with a reduced risk of AD.
Weih et al. (2007) identified three longitudinal cohort studies.
Two of the three included studies reported a statistically
significant reduced risk of dietary fish consumption and one
detected a protective effect associated with total n-3 fatty acids.
The remaining study reported a marginally significant protective
effect.
Patterson et al. (2007) identified a single study which reported
on the effect of fish consumption on incident AD. The study
detected a 70% decreased risk of AD (RR = 0.30; 95% CI = 0.10, 0.90)
associated with fish consumption exceeding 18.5 g/day.
The high-quality Cochrane systematic review conducted by Lim
et al. (2006) examined the association between omega-3 fatty acid
165
and the risk of AD. The review identified zero randomized
controlled trials.
Six prospective cohort studies reporting on the effect of PUFAs
were identified by Dangour et al. (2010). Only one of the six studies
reported a statistically significant reduced risk of AD associated
with fish consumption, omega-3 fatty acids, and a high intake of
DHA.
Both studies included in the review reported by Lee et al. (2010)
consistently demonstrated a reduced risk of AD associated with
PUFAs.
Williams et al. (2010) collated evidence pertaining to the
association between omega-3 fatty acids and risk of AD from
longitudinal cohort studies. Of the seven included studies
reporting on the effect of fish consumption, only three reported
a statistically significant reduced risk of AD. Similarly conflicting
findings were observed in the analysis examining the effect of
PUFAs derived from unspecified dietary sources or ascertained
according to plasma/serum levels.
3.1.35.2. Primary studies. The case-cohort study reported by Lopez
et al. (2011) examined the effect of plasma and dietary
docosahexaenoic acid (DHA) and fish consumption on the risk
of AD. Analyses treating plasma and dietary DHA exposure as a
continuous variable (per log SD increase) yielded risk estimates
below 1.00, however, results failed to reach statistical significance
(Plasma, OR = 0.76; 95% CI = 0.50, 1.14. Dietary, OR = 0.52; 95%
CI = 0.32, 0.84). Alternatively, when treated as a categorical
variable, low plasma (OR = 2.33; 95% CI = 1.04, 5.21) and dietary
(OR = 3.43; 95% CI = 1.42, 8.26) DHA significantly increased odds of
AD. High dietary (OR = 0.28; 95% CI = 0.09, 0.93) and plasma
(OR = 0.40; 95% CI = 0.15, 1.10) DHA were associated with a reduced
odds of AD, although results for plasma DHA failed to reach
statistical significance. The risk estimate for fish consumption was
below 1.00, however, results failed to reach statistical significance
(OR = 0.55; 95% CI = 0.20, 1.48). Considered cumulatively, results of
the study suggest that higher plasma and dietary DHA may reduce
risk of dementia and AD (Lopez et al., 2011).
3.1.35.3. Discussion. Seven higher quality systematic reviews,
collating evidence from 12 longitudinal studies reporting on the
association between PUFAs and risk of AD, were identified. Larger
systematic reviews by Williams et al. (2010) and Dangour et al.
(2010) identified a number of negative studies. One recent
observational study, which detected a protective effect
associated with higher plasma and dietary DHA, was identified
in the second phase of our review (Lopez et al., 2011). Considered
cumulatively, there was insufficient evidence to support a
beneficial effect of PUFAs against the development of AD.
3.1.36. Vitamin B12
Six systematic reviews examined the association between
vitamin B-12 and risk of AD (Standridge, 2004; Raman et al., 2007;
Purnell et al., 2009; Williams et al., 2010; Dangour et al., 2010; Lee
et al., 2010). Two reviews were deemed to be of low methodologi-
cal quality and were consequently excluded (Standridge, 2004;
Purnell et al., 2009). No primary studies were identified.
3.1.36.1. Systematic reviews. The effect of B-vitamins on the risk of
AD was examined by Raman et al. (2007). The reviewers identified
twelve studies which examined the association between blood
concentrations or dietary intake of B-vitamins and risk of AD. All
seven studies reporting on blood concentrations of vitamin B-12
failed to detect an association. The only study to report on the
effect of dietary intake of B-12 also failed to detect an association
(RR = not reported). Based on these findings, the reviewers
concluded that vitamin B-12 was not associated with AD.
166 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Five prospective cohort studies identified by Dangour et al.
(2010) examined the association between vitamin B-12 and AD
risk. Serum/plasma vitamin B-12 was not associated with risk of
AD according to all three studies reporting on this association.
However, one study, which examined the combined effect of serum
vitamin B-12 and folate, reported a greater than two-fold increased
risk of AD among those with low serum vitamin B-12 and low
serum folate (RR = 2.10; 95% CI = 1.20, 3.50). All three studies
reporting on dietary intake (food and supplements) of vitamin B-
12 failed to detect a statistically significant protective effect
associated with higher consumption. Considered cumulatively, the
reviewers concluded that the effect of vitamin B-12 on AD risk was
inconclusive.
Four cohort studies examining the effect of vitamin B-12 on AD
incidence were eligible for inclusion in the HTA reported by
Williams et al. (2010). All four studies failed to detect a statistically
significant association between AD and low dietary intake or low
serum concentrations of vitamin B12.
Lee et al. (2010) identified a single cohort study which reported
on the effect of vitamin B12. There was no significant association
between serum concentrations of vitamin B-12 and risk of AD
according to the study.
3.1.36.2. Discussion. Results from the identified moderate quality
systematic reviews suggest that vitamin B-12 was not associated
with a reduced risk of AD. Although studies that rely on self-
reported dietary intake collected by food frequency questionnaires
or food diaries are susceptible to exposure measurement error
(Kristal et al., 2005), primary studies that employed more objective
measures for exposure ascertainment, namely serum
concentrations, reported consistent findings suggesting a lack of
association between vitamin B-12 and risk of AD. However,
conclusions drawn from the reviews were limited by the small
number of available studies. Further research is required to
confirm the effect of vitamin B-12 on the risk of AD.
3.1.37. Vitamin B6
Five systematic reviews examined the association between
vitamin B-6 and risk of AD (Standridge, 2004; Raman et al., 2007;
Purnell et al., 2009; Dangour et al., 2010; Williams et al., 2010). Two
studies were excluded due to low methodological quality (Stand-
ridge, 2004; Purnell et al., 2009). No primary studies were
identified.
3.1.37.1. Systematic reviews. Raman et al. (2007) identified two
studies which reported on the effect of blood concentrations of
vitamin B-6 on risk of AD. Both failed to detect an association. With
respect to dietary intake, one study detected an association while
the other reported null findings. The reviewers concluded that the
available evidence did not implicate low vitamin B-6 as a risk factor
for the development of AD.
Both cohort studies included in the HTA reported by Williams
et al. (2010) failed to detect an association between vitamin B6 and
risk of AD. Based on the identified evidence, the reviewers
concluded that vitamin B-6 was not associated with risk of AD.
Of the four prospective cohort studies identified by Dangour
et al. (2010), two studies failed to detect an association between
dietary intake of vitamin B-6 and risk of AD and one study failed to
detect an association between plasma vitamin B-6 levels and risk
of AD. The remaining cohort study reported conflicting findings,
suggesting a protective effect attributed to high vitamin B-6
consumption. The reviewers concluded that vitamin B-6 did not
appear to be associated with risk of AD.
3.1.37.2. Discussion. All three moderate-quality systematic
reviews consistently failed to detect an association between
vitamin B-6 and risk of AD. Conflicting findings and the limited
number of available studies suggest that the association, if any, was
unclear (Raman et al., 2007).
3.1.38. Vitamin B3/Niacin
Two moderate quality systematic reviews, which reported on
the association between vitamin B-3/niacin and risk of AD, were
identified (Williams et al., 2010; Lee et al., 2010). The search for
primary studies did not yield any relevant records.
3.1.38.1. Systematic reviews. One study was identified by Williams
et al. (2010). The highest quintile of total niacin/vitamin B3 intake
was associated with a protective effect against the development of
AD (OR = 0.20; 95% CI = 0.01, 0.70). In the same study, tryptophan
(OR = 0.40; 95% CI = 0.10, 0.80), niacin obtained from food
(OR = 0.30; 95% CI = 0.10, 0.70), and “niacin equivalents”
(OR = 0.20; 95% CI = 0.10, 0.80) were all associated with a
statistically significant reduced risk of AD.
The review by Lee et al. (2010) examined the association
between a number of lifestyle-related risk factors and AD. One of
the included studies examined the association between niacin
intake and risk of AD. Higher intake of niacin was associated with a
statistically significant reduced risk of AD (RR = 0.30; 95% CI = 0.10,
0.70).
3.1.38.2. Discussion. Only one prospective cohort study examining
the effect of vitamin B-3/niacin on risk of AD was identified in two
systematic reviews. Results from one recent primary study
suggests that higher intake of niacin is associated with a
protective effect against the development of AD.
3.1.39. Folic Acid/Folate
Six systematic reviews reporting on the effect of folate were
identified (Standridge, 2004; Weih et al., 2007; Raman et al., 2007;
Dangour et al., 2010; Williams et al., 2010; Lee et al., 2010). One
review was of low methodological quality and was excluded
(Standridge, 2004). A single primary study was identified
(Hooshmand et al., 2010).
3.1.39.1. Systematic reviews. Nine studies identified by Raman
et al. (2007) examined the association between blood folate levels
and risk of AD. Four of the nine studies reported a detrimental
effect of low blood folate levels. The two studies that reported on
dietary folate intake reported conflicting findings. One detected an
association between low folate intake and AD while the other
reported null findings.
One study identified by Weih et al. (2007) reported on folate
intake. According to the study, low folate was associated with a
greater than two-fold increased risk of AD (RR = 2.10; 95% CI = 1.20,
3.50).
The systematic review by Dangour et al. (2010) examined the
association between several nutrients, including folate, and risk of
AD. A total of eight cohort studies investigating the association
between serum folate levels or folate intake were identified. Of the
five studies investigating the association between serum folate
levels and AD, only one detected an association (HR = 1.98; 95%
CI = 1.15, 3.40). Two of three studies reporting on dietary and
supplemental folate intake detected a protective effect, attributed
to higher intake. Overall, three of the eight included cohort studies
detected a reduced risk of AD associated with higher serum folate
concentrations and folate intake.
Lee et al. (2010) identified two cohort studies. One study
detected a near two-fold (HR = 1.98; 1.15, 3.40) increased risk of AD
associated with low serum folate concentrations (11.8 nmol/l
vs. >15.2 nmol/l). The remaining study failed to detect an
association.
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Williams et al. (2010) examined the effect of folate on the risk of
AD. Of the two included studies that examined serum folate, low
concentrations were associated with a near two-fold increased risk
of AD (HR = 1.98; 95% CI = 1.15, 3.40) in one study and only a
marginally significant 70% (HR = 1.70; 95% CI = 1.00, 3.40) increased
risk in the second study. High total folate intake was associated
with a statistically significant reduced risk of AD in one of two
studies. Findings from the review suggested that while low folate
serum concentrations were associated with an increased risk of
AD, the association between total folate intake and AD was
inconclusive.
3.1.39.2. Primary studies. We identified a single primary study
which reported on the association between serum folate and risk
of AD (Hooshmand et al., 2010). The Cardiovascular Risk Factors,
Aging, and Dementia (CAIDE) study failed to detect an association
between serum folate and risk of AD (OR = 1.01; 95% CI = 0.87, 1.18).
3.1.39.3. Discussion. A recent primary study failed to detect an
association between serum folate and risk of AD (Hooshmand
et al., 2010). Findings from four moderate systematic reviews
suggest that the effect of dietary, supplemental and serum folate
concentrations and risk of AD was inconclusive.
3.1.40. Vitamins C and E
Six systematic reviews examining the effect of vitamins C and/
or E on the risk of AD were identified (Standridge, 2004; Pham and
Plakogiannis, 2005; Weih et al., 2007; Patterson et al., 2007;
Williams et al., 2010,,2010). Two reviews were excluded due to low
methodological quality (Standridge, 2004; Pham and Plakogiannis,
2005). One relevant primary study was identified in the second
phase of the review (Devore et al., 2010).
3.1.40.1. Systematic reviews. Three cohort studies identified by
Weih et al. (2007) examined the effect of vitamin C, vitamin E or
the combined effect of these vitamins on the risk of AD. One of two
studies reporting on vitamin C detected an association. Two
studies reported a statistically significant (RR = 0.30; 95% CI = 0.10,
0.92) and a marginally significant (RR = 0.82; 95% CI = 0.66, 1.00)
reduced risk of AD associated with increasing dietary vitamin E
consumption. One study examined the combined effect of vitamin
C and E intake. The combined effect of vitamin C and E intake was
not associated with risk of AD (RR = 1.81; 95% CI = 0.91, 3.63). The
reviewers concluded that the effect of vitamins C and E on AD was
unclear.
One study identified by Patterson et al. (2007) examined the
effect of low serum vitamin E on AD risk. The lowest tertile of
serum vitamin E was not significantly associated with an increased
risk of AD (RR = 2.10; 95% CI: 0.81, 5.54).
Lee et al. (2010) identified four prospective cohort studies
examining the risk of AD associated with vitamin C and E intake.
Three studies failed to detect an association between dietary
intake of vitamin C and vitamin E and risk of AD. The remaining
study, however, reported a marginally significant 18% reduced risk
of AD associated with each 1 SD increase in vitamin C (RR = 0.82;
95% CI = 0.68, 0.99) and vitamin E (RR = 0.82; 95% CI = 0.66, 1.00)
consumption. The reviewers concluded that vitamin E may reduce
risk of AD.
Eleven studies were included in the HTA reported by Williams
et al. (2010). Six of the included studies failed to detect a
statistically significant association between dietary or supple-
mental intake of vitamins C and/or E and risk of AD. Of the
remaining five studies, three reported a marginally or statistically
significant reduced risk of AD associated with higher intake of
vitamin C or vitamin E. One reported a statistically significant
reduced risk of AD associated with combined supplementation of
167
vitamin C and E (HR = 0.36; 95% CI = 0.09, 0.99), however, isolated
supplementation of vitamin C or vitamin E was not protective
against the development of AD. The final study reported
conflicting findings, suggesting an increased risk of AD associated
with the second and fourth quartile of vitamin E intake. Based on
the available evidence, the reviewers concluded that there was
insufficient evidence to suggest that vitamins C and E were
related to risk of AD.
3.1.40.2. Primary studies. We identified one relevant primary
study which reported on the effect of vitamins C and E. Using data
from the Rotterdam Study, Devore et al. (2010) prospectively
examined the effect of dietary intake of vitamins C and E, beta-
carotene, and flavonoids on risk of AD. Using the lowest tertile of
consumption as a referent, higher tertiles of vitamin C intake were
not associated with risk of AD. The highest tertile of vitamin E
consumption, however, was associated with a statistically
significant reduced risk of AD. These findings suggest that
higher dietary consumption of vitamin E may confer a
protective effect against the development of AD.
3.1.40.3. Discussion. Based on four moderate quality systematic
reviews, the combined and isolated effect of vitamins C and E on
risk of AD was inconclusive. A majority of the primary studies
included in the identified systematic reviews reported null
findings. The only primary study which we identified in the
second phase of the review detected a protective effect attributed
to high consumption of vitamin E but not C (Devore et al., 2010).
Further research into the effect of dietary and supplemental intake
of vitamin C and E on the risk of AD is required (Williams et al.,
2010).
3.1.41. Multivitamins
Two systematic reviews reported on the effect of multivitamin
use on the risk of AD (Standridge, 2004; Williams et al., 2010). One
study was excluded due to low methodological quality (Standridge,
2004). No primary studies were identified.
The HTA reviewed the effect of supplemental multivitamin use
on the risk of incident AD (Williams et al., 2010). Both included
cohort studies failed to detect a protective effect attributed to
multivitamin use.
Multivitamin use did not appear to reduce the risk of AD. These
findings were limited by the very small number of available studies
which precludes a definitive conclusion.
3.1.42. Flavonoids
Three systematic reviews examined the association between
flavonoid intake and risk of AD (Standridge, 2004; Williams et al.,
2010; Lee et al., 2010). One review was excluded due to low
methodological quality (Standridge 2004). A single primary study
was identified in the second phase of the review (Devore et al.,
2010).
3.1.42.1. Systematic reviews. The systematic review by Lee et al.
(2010) examined the association between AD and a number of
dietary factors including flavonoids. Two of the three studies
identified in the HTA (Williams et al., 2010) were also identified by
Lee et al. (2010). Both cohort studies failed to detect a statistically
significant association between dietary flavonoid consumption
and risk of AD.
Three cohort studies were identified and included in the HTA
reported by Williams et al. (2010). Two of the three studies failed to
detect an association. One study reported that higher consumption
of dietary flavonoids was associated with a statistically significant
reduced risk of AD (HR = 0.49; 95% CI = 0.26, 0.92).
168 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
3.1.42.2. Primary studies. A single primary study was identified
which reported on the association between flavonoids and risk of
AD. According to the Rotterdam Study (Devore et al., 2010), there
was no association between flavonoid intake and risk of AD
(Middle tertile, HR = 1.11; 95% CI = 0.85, 1.44. Highest tertile,
HR = 1.09; 95% CI = 0.84, 1.42).
3.1.42.3. Discussion. One recent primary study failed to detect an
association between flavonoids and risk of AD. Results summarized
in two moderate quality systematic reviews indicate that there is
insufficient evidence to suggest that flavonoids reduce risk of AD.
Further research into the effect of flavonoids is required.
3.1.43. Fruit and vegetable intake
Two moderate-quality systematic reviews examined the effect
of fruit and vegetable intake and risk of incident AD (Williams
et al., 2010; Lee et al., 2010). No primary studies were identified.
3.1.43.1. Systematic reviews. Two cohort studies were eligible for
inclusion in the HTA reported by Williams et al. (2010). Both
studies detected a statistically significant association. One study
reported that consumption of fruit and vegetable juice three or
more times per week significantly reduced risk of AD (HR = 0.24;
95% CI = 0.09, 0.61). Moderate or high intake of fruits and
vegetables in midlife was associated with a reduced risk of AD
according to the second study (OR = 0.60; 95% CI = 0.41, 0.86). Based
on the available evidence, the reviewers concluded that high
consumption of fruits and vegetables may reduce the risk of AD.
However, these findings were limited by the small number of
available studies.
Lee et al. (2010) identified one cohort study which examined
the effect of fruit and vegetable intake on risk of AD. According to
the results of this study, consumption of fruit and vegetable juices
three or more times per week was associated with a 76%
(HR = 0.24; 95% CI = 0.09, 0.61) decreased risk of AD.
3.1.43.2. Discussion. The available evidence, summarized in two
moderate-quality systematic reviews, suggest that higher intake of
fruits and vegetables may decrease risk of AD. However, the small
number of available studies precludes a conclusive statement
regarding the effect of fruit and vegetable intake on the risk of
incident AD (Williams et al., 2010).
3.1.44. Beta-carotene
Four systematic reviews examined the association between
beta-carotene and risk of AD (Standridge, 2004; Weih et al., 2007;
Williams et al., 2010; Lee et al., 2010). One review was excluded
(Standridge, 2004). One primary study was identified in the second
phase of the review (Devore et al., 2010).
3.1.44.1. Systematic reviews. One cohort study identified in the
systematic review by Weih et al. (2007) examined the association
between beta-carotene intake and risk of AD. According to the
results of the study, low beta-carotene consumption was not
associated with reduced risk of AD (RR = 0.55; 95% CI = 0.22, 1.35).
The HTA identified four cohort studies that examined the effect
of carotenoids or beta-carotene on the risk of AD (Williams et al.,
2010). All four studies failed to detect an association. Findings of
the report suggest that beta-carotene and other carotenoids are not
associated with a reduced risk of AD (Williams et al., 2010).
The systematic review by Lee et al. (2010) identified four cohort
studies. All four studies failed to detect an association between
beta-carotene/carotenoids and AD.
3.1.44.2. Primary studies. We identified a single primary study
reporting on the effect of beta-carotene. Using data from the
Rotterdam Study, Devore et al. (2010) prospectively examined the
effect of dietary intake of vitamins C and E, beta-carotene, and
flavonoids on risk of AD. According to the results of the study, there
was no association between dietary beta carotene and risk of AD
(Middle tertile, HR = 1.02; 95% CI = 0.79, 1.31. Highest tertile,
HR = 1.07; 95% CI = 0.83, 1.39).
3.1.44.3. Discussion. Three moderate quality systematic reviews
summarizing evidence from five cohort studies suggest that beta-
carotene and other carotenoids are not associated with a beneficial
effect. A recent cohort study also failed to detect an association.
3.1.45. Caffeine consumption
Three moderate quality systematic reviews examined the effect
of caffeine consumption on the risk of AD (Barranco Quintana et al.,
2007; Santos et al., 2010; Lee et al., 2010). One relevant primary
study was identified in the second phase of the review (Gelber
et al., 2011).
3.1.45.1. Systematic reviews. Barranco Quintana et al. (2007)
identified two case-control and two cohort studies. Two of the
four identified studies detected a protective effect attributed to
coffee consumption. The random effects meta-analysis failed to
detect an association between coffee consumption and AD (Risk
estimate: 0.79; 95% CI = 0.46, 1.36). However, sensitivity analyses
restricted to cohort studies (Risk estimate: 0.73; 95%: CI 0.54, 0.99),
studies employing interview questionnaires (Risk estimate: 0.70;
95% CI = 0.55, 0.90), and studies that were less susceptible to bias
(Risk estimate: 0.58; 95% CI = 0.44, 0.77), all support the protective
effect of coffee consumption against the development of AD.
Because of the limited number of available studies and the
presence of heterogeneity across these studies, further research is
warranted (Barranco Quintana et al., 2007).
Two case-control and three cohort studies were included in the
systematic review reported by Santos et al. (2010). Among the five
included studies, three failed to detect a reduced risko of AD
associated with coffee consumption. The remaining two studies
reported a 31% and 60% reduced risk of AD among coffee
consumers. All three studies reporting tea consumption failed to
detect an association. Caffeine consumption was associated with a
38% (RR = 0.62; 95% CI = 0.45, 0.87; test for heterogeneity: p = 0.169;
I2 = 40.5%) reduced risk of AD according to the pooled analysis of
four studies. Findings from this review support the protective
effect of caffeine consumption. However, as noted by the
reviewers, methodological heterogeneity across studies serves as
an important limitation of the review.
The review by Lee et al. (2010) examined the association
between a number of lifestyle-related risk factors and AD. Tea
consumption was not associated with risk of AD according to the
only relevant study.
3.1.45.2. Primary studies. The nested case-control study conducted
by Gelber et al. (2011) examined the effect of coffee and caffeine
consumption on the risk of developing all-cause dementia, AD,
vascular dementia, and other cognitive outcomes. Higher intake of
coffee was not associated with risk of AD. Higher quintiles of
caffeine consumption, derived from self-reported coffee, tea, and
cola intake, were also not associated with risk of AD. When caffeine
intake was treated as a continuous variable, each 1 SD (223 mg)
increase in consumption was not associated with AD. Results of
this study suggest that coffee and caffeine consumption were not
associated with risk of AD.
3.1.45.3. Discussion. Three moderate quality systematic reviews
examined the effect of caffeine intake on the risk of AD. Among the
six primary studies included in the three reviews, only two studies
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
reported a statistically significant association between caffeine
consumption and risk of AD. Moreover, recent cohort studies,
including the nested case-control study identified in the second
phase of the review (Gelber et al., 2011), also failed to detect an
association. Although the pooled analysis reported by Santos et al.
(2010) and the subgroup analyses reported by Barranco Quintana
et al. (2007) were suggestive of a protective effect, these findings
were limited by the small number of available studies. Overall,
based on the conflicting findings of the observational studies, the
association between caffeine consumption and risk of AD remains
inconclusive.
3.1.46. Mediterranean diet
Six systematic reviews examined the association between
adherence to the Mediterranean diet and risk of AD (Weih et al.,
2007; Roman et al., 2008; Sofi et al., 2008, 2010; Williams et al.,
2010; ESHRE Capri Workshop Group, 2011). Two reviews were of
poor methodological quality and were consequently excluded. One
primary study was identified in the second phase of the review (Gu
et al., 2010a).
3.1.46.1. Systematic reviews. The effect of adherence to the
Mediterranean diet was examined in the review by Weih et al.
(2007). Evidence from the only identified study suggested that
high adherence to the Mediterranean was associated with a
reduced risk of AD (HR = 0.60; 95% CI = 0.40, 0.90).
The systematic review by Sofi et al. (2008) examined the effect
of the Mediterranean dietary pattern on the risk of numerous
outcomes including Alzheimer's disease, Parkinson's disease,
cancer, and mortality. The only study which focused on the
outcome of AD reported a protective effect attributed to higher
adherence (RR = 0.83; 95% CI = 0.70, 0.98).
The updated review reported by Sofi et al. (2010) identified one
newly published study focused on the risk of AD. The prospective
cohort study failed to detect an association between high
adherence to a Mediterranean diet and development of AD
(RR = 1.00; 95% CI = 0.71, 1.40).
Four cohort studies, three of which examined the same cohort,
were identified by Williams et al. (2010). Three of four identified
studies detected a reduced risk of AD associated with adherence to
the Mediterranean diet. Findings from this review suggest that the
Mediterranean diet may exert a protective effect against
incident AD, however, further research is required (Williams
et al., 2010).
3.1.46.2. Primary studies. According to results of the fully adjusted
analysis reported by Gu et al. (2010a), higher adherence to the
Mediterranean diet, treated as a continuous variable, was
associated with a significantly reduced risk of AD (HR = 0.87;
95% CI = 0.78, 0.97). Treated categorically, both moderate and high
adherence yielded risk estimates below 1.00, however, only the
moderate category reached statistical significance (Middle tertile,
HR = 0.56; 95% CI = 0.36, 0.86. Highest tertile, HR = 0.68; 95%
CI = 0.42, 1.08). Results of the study suggest that greater
adherence to the Mediterranean diet is associated with a
reduced risk of AD.
3.1.46.3. Discussion. Four systematic reviews, collating evidence
from four cohort studies, were identified. A recent cohort study
was also identified in the second phase of the review. Considered
cumulatively, the available evidence suggested that greater
adherence to the Mediterranean diet may be associated with a
reduced risk of AD. However, the limited number of available
studies precludes a conclusive statement regarding the effect of
this dietary pattern.
169
3.1.47. Other dietary patterns
We identified a single study that prospectively examined the
effect of multiple dietary patterns on the risk of developing AD (Gu
et al., 2010b).
Of the seven dietary patterns (DP) examined, only the second
DP was associated with risk of AD (HR = 0.54; 95% CI = 0.39, 0.75; p
< 0.001). This dietary pattern consisted of “higher intakes of salad
dressing, nuts, fish, tomatoes, poultry, cruciferous vegetables,
fruits, and dark and green leafy vegetables and a lower intake of
high-fat dairy products, red meat, organ meat, and butter” (Gu
et al., 2010b). Moderate and high compliance to the aforemen-
tioned dietary pattern conferred a protective effect against the
development of AD, however, results for moderate compliance
failed to reach statistical significance (Middle tertile, HR = 0.81;
95% CI = 0.59, 1.12. Highest tertile, HR = 0.62; 95% CI = 0.43, 0.89).
3.1.48. Fatintake, saturated fatty acids
Four moderate quality systematic reviews, which examined the
association between total fat, saturated fat, or transunsaturated fat
intake and risk of AD, were identified (Weih et al., 2007; Williams
et al., 2010; Lee et al., 2010; Crichton et al., 2010). No primary
studies were located.
3.1.48.1. Systematic reviews. Three cohort study identified in the
systematic review by Weih et al. (2007) examined the association
between fat intake and risk of AD. All three studies reported risk
estimates exceeding 1.00 for total or saturated fat intake, however,
results from only one study reached statistical significance.
Specifically, the study reported an increased risk of AD
associated with saturated and transunsaturated dietary fat but
not with total fat intake. The reviewers concluded that dietary fat
may be associated with an increased risk of AD.
Crichton et al. (2010) examined the association between fat
intake from dairy products and risk of AD. The reviewers identified
a single study which focused on the risk of developing AD.
The study reported that fat intake from milk or sour milk was
not associated with risk of AD. Moderate intake of saturated fatty
acids from spreads was associated with a near four-fold increased
risk of AD (OR = 3.82; 95% CI = 1.48, 9.87).
The review by Lee et al. (2010) examined the association
between a number of lifestyle-related risk factors and AD. Two
studies, reporting on the same cohort, examined the association
between saturated fatty acids obtained from spreads and risk of
AD. Higher saturated fatty acid intake from milk, sour milk, and
spreads was associated with an almost four fold increased odds of
AD (OR = 3.82; 95% CI = 1.48, 9.87) according to one study. The other
reported a greater than four-fold (OR = 4.34; 95% CI = 1.28, 14.68)
increased risk of AD among APOE e4 carriers associated with
higher saturated fatty acid consumption from spreads. Results of
the review suggest that higher saturated fatty acid consumption,
particularly obtained from spreads, may increase risk of AD.
Williams et al. (2010) identified two cohort studies that
examined the association between dietary fats (saturated,
monounsaturated, transunsaturated) and risk of AD. One study
detected an increased risk of AD associated with high saturated fat
consumption (RR = 2.20; 95% CI = 1.10, 4.70). High transunsaturated
fat consumption also increased risk of AD (Quintile 2, RR = 2.40;
95% CI = 1.10, 5.30. Quintile 3, RR = 2.90; 95% CI = 1.20, 7.20). The
second study detected an increased risk of AD associated with
saturated fat obtained from spreads, however, only the second
quartile reached statistical significance (OR = 3.82; 95% CI = 1.48,
9.87). Total and monounsaturated fats were not associated with
risk of AD according to both studies. Based on this evidence, the
reviewers concluded that saturated and transunsaturated fat
consumption may be associated with a deleterious effect (Williams
et al., 2010).
170 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
3.1.48.2. Discussion. Considered cumulatively, the available
evidence suggested that dietary saturated and transunsaturated
fats may be associated with an increased risk of AD. However, the
small number of available primary studies precludes any definitive
statement.
3.1.49. Caloric intake
Two moderate quality systematic reviews examined the
association between caloric intake and risk of AD (Weih et al.,
2007; Williams et al., 2010).
3.1.51.1. Systematic reviews. Weih et al. (2007) identified a single
cohort study that examined the association between caloric intake
and risk of AD. The study detected a marginally significant 50%
increased risk of AD associated with the highest quartile of caloric
intake (RR = 1.50; 95% CI = 1.00, 2.20).
The same primary study identified by Weih et al. (2007), which
suggested an increased risk of AD associated with high total daily
caloric intake, was also included in the HRT reported by Williams
et al. (2010).
3.1.49.2. Discussion. Results from a single cohort study suggest
that high caloric intake may increase risk of AD. However, as only
one primary study was identified in both reviews, further research
is needed to clarify this association.
3.1.50. Education
Three moderate-quality systematic reviews, which examined
the association between education and risk of AD, were identified
(Caamano-Isorna et al., 2006; Patterson et al., 2007; Williams et al.,
2010). Five observational studies were retrieved (Verdelho et al.,
2010; Reitz et al., 2010b; Song et al., 2011; Yang et al., 2011; Shah
et al., 2011).
3.1.50.1. Systematic reviews. The systematic review and meta-
analysis by Caamano-Isorna et al. (2006) examined the association
between educational attainment and risk of AD. Nine cohort and
five case-control studies were included. Low educational
attainment was associated with an 80% (RR = 1.80; 95% CI = 1.43,
2.27; test for heterogeneity: Ri statistic: 0.61; Cochrane Q statistic
p-value: <0.0001) increased risk of AD according to the pooled
analysis of all 14 studies. In the analysis restricted to cohort studies,
low educational attainment was associated with a 59% increased
risk of AD (RR = 1.59; 95% CI = 1.35, 1.86). The pooled analysis
restricted to case-control studies also yielded statistically
significant findings, however, the derived risk estimate was
greater in magnitude (RR = 2.40; 95% CI = 1.32, 4.38). Compared
to the highest level of education, the combined category of low and
moderate levels of education was associated with a significant 44%
increased risk of AD (RR = 1.44; 95% CI = 1.24, 1.67; Ri statistic: 0.47;
Cochrane Q statistic p-value: 0.073). Stratified analysis by
methodological design also yielded statistically significant
findings (Cohort studies, RR = 1.32; 95% CI = 1.09, 1.59. Case-
control studies, RR = 1.66; 95% CI = 1.30, 2.10). Based on the
available evidence, the reviewers concluded that low
educational attainment was associated with an increased risk of
AD (Caamano-Isorna et al., 2006).
The systematic review reported by Patterson et al. (2007)
identified a single cohort study which examined the effect of
educational attainment on the risk of developing AD. The study
detected a decreased risk of AD associated with lower educational
attainment (<12 years vs. >15 years) (RR = 0.48; 95% CI = 0.27 to
0.84).
One systematic review (Caamano-Isorna et al., 2006) and two
cohort studies were included in the HTA reported by Williams et al.
(2010). According to the findings of the systematic review by
Caamano-Isorna et al. (2006), low levels of education were
associated with an increased risk of AD. Findings from two cohort
studies, published after the review by Caamano-Isorna et al.
(2006), reported consistent findings. One of the studies reported a
statistically significant protective effect attributed to high levels of
education (9 years) (OR = 0.15; 95% CI = 0.05, 0.40). Moderate
levels of education (6–8 years) were associated with a marginally
significant (OR = 0.49; 95% CI = 0.24, 1.00) reduced risk of AD when
compared to low educational attainment. The second study
detected an increased risk of AD associated with lower educational
attainment ( high school) compared to high levels of education
(graduate school) (OR = 41.48; 95% CI = 4.00, 42.40). Moderate
education (undergraduate degree) was associated with an
increased risk of AD, however, results did not reach statistical
significance (OR = 2.07; 95% CI = 0.28, 15.10). Based on the evidence
provided by the included systematic review and cohort studies, the
reviewers concluded that higher educational attainment appeared
to be associated with a protective effect against the development of
AD (Williams et al., 2010).
3.1.50.2. Primary studies. According to results of the LADIS
prospective cohort study (Verdelho et al., 2010), educational
level, treated as a continuous variable, was not significantly
associated with risk of AD (HR = 0.96; 95% CI = 0.85, 1.09; p = 0.513).
When compared to more years of education ( > 9 years), fewer
years of education significantly increased risk of possible and
probable LOAD (7–9 years, HR = 3.01; 95% CI = 1.32, 6.87. 0–6 years,
HR = 4.51; 95% CI = 2.09, 9.74) according to the prospective study
reported by Reitz et al. (2010b). Results remained consistent when
the analysis was restricted to probable LOAD (7–9 years, HR = 2.93;
95% CI = 1.19, 7.23. 0–6 years, HR = 3.76; 95% CI = 1.61, 8.77).
The longitudinal cohort study conducted by Yang et al. (2011)
sought to examine the association between blood pressure,
hypertension, and risk of AD. The association between blood
pressure measures, namely, self-reported hypertension and
objectively measured systolic (SBP) and diastolic blood pressure
(DBP), pulse pressure (PP), and mean arterial pressure (MAP) were
analyzed in separate models. The association between education,
reported as a continuous variable (1 year increase), and risk of AD
was examined in each model. According to the risk estimates
derived from each model, increasing education was not associated
with risk of AD (Self-reported HTN model, HR = 1.01; 95% CI = 0.96,
1.06. SBP model, HR = 0.99; 95% CI = 0.94, 1.05. DBP model,
HR = 1.00; 95% CI = 0.95, 1.05. Pulse pressure model, HR = 1.00;
95% CI = 0.95, 1.05. Mean arterial pressure model, HR = 1.00; 95%
CI = 0.95, 1.05).
Results from the Rush Memory and Aging Project (Shah et al.,
2011) failed to detect an association between education and risk of
AD (HR = 0.98; 95% CI = 0.91, 1.05).
Lower education, defined as less than nine years, was associated
with an increased risk of AD according to results from the Canadian
Study of Health and Aging (OR = 1.37; 95% CI = 1.02, 1.83) (Song
et al., 2011).
3.1.50.3. Discussion. Only two of five observational studies
identified in the second phase of the review detected an
association between educational attainment and risk of AD.
However, findings from three moderate-quality systematic
reviews, collating evidence from 16 observational studies,
suggested that lower levels of education appeared to increase
risk of AD. Consequently, a majority of the available studies
supported the hypothesis that higher educational attainment is
associated with a reduced risk of AD. Whether educational
attainment is a true risk factor for AD or a factor involved in the
alleviation or postponement of clinical symptoms remains an issue
of continued debate (Caamano-Isorna et al., 2006; Stern, 2006).
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
3.1.51. Cognitive leisure activities
We identified three moderate- (Weih et al., 2007; Williams
et al., 2010; Stern and Munn, 2010) and two low- (Fratiglioni et al.,
2004; Barnes and Yaffe, 2011) quality systematic reviews that
examined the association between cognitively stimulating leisure
activities and subsequent risk of AD. No primary studies were
identified.
3.1.51.1. Systematic reviews. Weih et al. (2007) systematically
reviewed the association between cognitive stimulation, leisure
activities, and social contacts with risk of developing AD. All six
identified studies reported risk estimates below 1.00. Based on the
available evidence, the reviewers concluded that maintaining
social contacts and regular engagement in cognitively stimulating
and other leisurely activities appeared to protect against the
development of AD.
Stern and Munn (2010) examined the effect of early-, middle-,
and late-life participation in cognitive leisure activities on the risk
of AD. A total of five case-control and one cohort study examined
the effect of early and middle life (20 to 60 years of age)
participation. All six studies found a statistically significant or
marginally significant association between AD and at least one
domain of cognitive leisure activity or with cognitive leisure
activities considered as an all-encompassing category. Five cohort
studies, three of which specifically reported on risk of developing
AD, examined the effect of late-life (65 years of age) participation
in cognitive leisure activities. All three studies detected a
protective effect attributed to frequent engagement in late-life.
Based on the available evidence, the reviewers concluded that
middle and late-life participation in cognitive leisure activities may
protect against the development of AD.
Three of four cohort studies identified by Williams et al. (2010)
specifically reported on the risk of developing AD with respect to
engagement in cognitive leisure activities. All three studies
consistently reported a statistically significant reduced risk of
AD associated with frequent participation.
3.1.51.2. Discussion. According to evidence summarized in three
moderate-quality systematic reviews, increased participation in
cognitive leisure activities appears to be associated with a reduced
risk of AD.
3.1.52. Social Networks/Engagement
Two moderate- (Weih et al., 2007; Williams et al., 2010) and one
low- (Fratiglioni, Paillard-Borg, and Winblad 2004) quality
systematic reviews examined the association between social
engagement and risk of AD. Our search for primary studies
yielded one relevant record (Amieva et al., 2010).
3.1.52.1. Systematic reviews. Weih et al. (2007) identified six
relevant studies. Based on the available evidence, the reviewers
concluded that maintaining social contacts and regular
engagement in cognitively stimulating and other leisurely
activities were associated with a protective effect against the
development of AD.
Studies identified by Williams et al. (2010) suggest that low
social engagement, poor social networks, and loneliness were
associated with an increased risk of AD. Furthermore, never being
married, being single and living alone, as well as widowhood and
divorce in late-life were also associated with an increased risk.
3.1.52.2. Primary studies. Using data from the PAQUID cohort
study, Amieva et al. (2010) examined the effect of social networks
on risk of AD. Compared to married subjects, widowed (RR = 0.92;
95% CI = 0.70, 1.10; p = 0.53), single (RR = 1.36; 95% CI = 0.70, 2.30;
p = 0.27) and divorced/separated (RR = 0.88; 95% CI = 0.40, 1.70;
171
p = 0.72) subjects were not at a significantly increased risk of AD.
The size of social networks was not associated with AD risk (4–7
persons vs. 0–3 persons, RR = 1.24; 95% CI = 0.80, 1.70; p = 0.22.  8
persons vs. 0–3 persons, RR = 1.22; 95% CI = 0.80, 1.70; p = 0.25). The
ratio of family members to friends was also not significantly
associated with risk of AD (More family members than friends,
RR = 0.80; 95% CI = 0.60, 1.00; p = 0.12. More friends than family
members, RR = 0.92; 95% CI = 0.60, 1.30; p = 0.68). There was no
association between AD and satisfaction with social networks
(RR = 0.84; 95% CI = 0.30, 1.30; p = 0.21) and feeling misunderstood
(RR = 0.66; 95% CI = 0.30, 1.40; p = 0.24). Subjects reporting to
receive more than they give were at a marginally significant
reduced risk of AD (RR = 0.47; 95% CI = 0.20, 1.00; p = 0.05). Giving
more than receiving yielded a risk estimate above 1.00, however,
results failed to reach statistical significance (RR = 1.16; 95%
CI = 0.90, 1.40; p = 0.22).
3.1.52.3. Discussion. Results summarized in two moderate-quality
systematic reviews suggest that high social engagement and
maintaining social contacts were associated with a protective
effect against the development of AD (Weih et al., 2007; Williams
et al., 2010). There was evidence to suggest that marriage was
associated with a reduced risk of AD (Williams et al., 2010). Results
from the only primary study identified in the second phase of the
review detected a reduced risk of AD among those who reported
receiving more than they give (Amieva et al., 2010). However, the
study failed to detect an association between size and composition
(ratio of family to friends, marriage) of social networks with risk of
AD. Considered cumulatively, there was some evidence to suggest
that lower social engagement was associated with an increased
risk of AD.
3.1.53. Constricted life space
We identified a single prospective study reporting on the
association between life space and risk of AD (James et al., 2011).
Subjects were recruited from the Rush Memory and Aging Project
and the Minority Aging Research Study. The study detected an
increased risk of AD associated with constricted life space
(HR = 1.21; 95% CI = 1.08, 1.36). The risk estimate was slightly
attenuated, but remained statistically significant, when the
analysis was adjusted for additional variables, namely, physical
function, disability, depressive symptoms, social networks, vascu-
lar risk factors and diseases (HR = 1.15, 95% CI = 1.02, 1.31). Overall,
results of the study suggest that a higher level of life-space
constriction was associated with an increased risk of AD.
3.1.54. Lifestyle behaviour Patterns/Multiple lifestyle factors
We did not identify any systematic reviews reporting on the
effect of multiple, concurrent lifestyle risk factors. The search for
primary studies yielded two relevant studies (Norton et al., 2012;
Gelber et al., 2012).
3.1.54.1. Primary studies. Data from the Cache County Study was
used to examine the combined effect of diet, exercise, church
attendance, social interaction, alcohol consumption, and smoking
status on risk of developing AD (Norton et al., 2012). Using these
factors, four behavioural patterns or exposure categories were
developed: (a) unhealthy and not religious, (b) healthy and
moderately religious, (c) healthy and very religious, and (d)
unhealthy and religious, with the latter category serving as the
reference. Compared to the unhealthy and religious group, the
healthy and moderately religious group (HR = 0.54; 95% CI = 0.34,
0.86) as well as the healthy and very religious group (HR = 0.56;
95% CI = 0.36, 0.88) were at a reduced risk of AD. The risk estimate
for the unhealthy and not religious group did not reach statistical
significance (HR = 0.59; 95% CI = 0.31, 1.12). The association was not
172 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
modified by APOE e4 status. Results from this longitudinal cohort
study suggest that certain lifestyle behavioural patterns, which
take several risk factors into consideration, were associated with a
reduced risk of AD.
The nested case-control study by Gelber et al. (2012) examined
the effect of lifestyle risk factors, namely, BMI, smoking status, diet,
and physical activity on the development of AD among Japanese-
American men enrolled in the Honolulu-Asia Aging Study. The
combined effect of the aforementioned factors was also examined.
Lifestyle behavioural patterns were derived according to the
number of low-risk factors. The following patterns were examined:
one low-risk factor (non-smoking), two low-risk factors (non-
smoking and diet score in top 40%), three low-risk factors (non-
smoking, diet score in top 40%, BMI <25.0 kg/m2), and four low-risk
factors (non-smoking, diet score in top 40%, BMI <25.0 kg/m2, high
physical activity). According to the results of the study, the
aforementioned patterns were not associated with risk of AD (risk
estimates not reported). Overall, the study failed to detect an
association between putative lifestyle risk factors, considered in
combination, and risk of AD.
3.1.54.2. Discussion. The identified studies were heterogeneous
with respect to the risk factors considered in the analysis.
Methodological heterogeneity and the small number of studies
preclude a definitive statement regarding the effect of multiple,
concurrent lifestyle factors or behaviour patterns.
3.1.55. Head injury
Four moderate quality systematic reviews, which examined the
association between head injury and risk of AD, were identified
and eligible for inclusion in the present review.
3.1.55.1. Systematic reviews. Seven case-control studies were
eligible for inclusion in the systematic review reported by
Mortimer et al. (1991). The meta-analysis combining the results
of the seven included studies detected an increased risk of AD
associated with head trauma involving a loss of consciousness
(RR = 1.82; 95% CI = 1.26, 2.67). There was no evidence of statistical
heterogeneity. Several subgroup analyses were conducted. Cases
were stratified by family history of dementia, age of onset, and
gender. When cases were stratified by family history, the
association between head injury and AD was statistically
significant for those with a sporadic (RR = 2.31; 95% CI = 1.17,
4.84) but not familial (RR = 1.42; 95% CI = 0.76, 2.71) case of AD.
However, there was no statistical difference between groups
(x2 = 1.17, p = 0.30). When stratified by age of onset, the association
between AD and head trauma was similar in both groups
( < 70 years: RR = 1.95; 95% CI = 1.12, 3.48. 70 years: RR = 1.81;
95% CI = 1.03, 3.25). Lastly, according to the subgroup analysis by
gender, head trauma was associated with an increased risk of AD in
males (RR = 2.67; 95% CI = 1.64, 4.41) but not in females (RR = 0.85;
95% CI = 0.43, 1.70). Overall, this review suggested that head trauma
with loss of consciousness was associated with an increased risk of
AD in males.
The systematic review reported by Fleminger et al. (2003) was
aimed at updating the earlier review by Mortimer et al. (1991). A
total of 15 case-control studies, seven of which were published
after the Mortimer et al. review, were identified. Results of the
fixed effects meta-analysis (n = 15) suggested an increased odds of
AD associated with head trauma (OR = 1.58; 95% CI = 1.21, 2.06).
Studies published before the Mortimer et al. (1991) systematic
review were combined in a subgroup analysis. The pooled analysis
detected a 90% increased odds of AD associated with head injury
(OR = 1.90; 95% CI = 1.29, 2.81). In contrast, head trauma was not
associated with AD when the analysis was restricted to the case-
control studies published after the 1991 systematic review
(OR = 1.35; 95% CI = 0.94, 1.94). Head trauma was not associated
with risk of AD among females (OR = 0.91; 95% CI = 0.56, 1.47). In
contrast, head trauma increased risk of AD in males (OR = 2.29; 95%
CI = 1.47, 3.58). The increased risk of AD in males (OR = 2.26; 95%
CI = 1.13, 4.53) but not females (OR = 0.92; 95% CI = 0.53, 1.59) was
also observed when the analyses were restricted to post Mortimer
et al. (1991) studies. Overall, when the results of all 15 studies were
considered cumulatively, the review suggested that prior head
injury with loss of consciousness was associated with an increased
risk of AD. However, when the analysis was restricted to studies
published after the Mortimer et al. (1991) review, there was no
evidence of an association. The review was able to replicate
findings reported by Mortimer et al. (1991), which suggest an
association between head injury and risk of AD in males only.
Three longitudinal cohort studies, identified by Patterson et al.
(2007), reported on the association between head injury and AD.
Only one of the three studies detected an association between head
injury and AD. More specifically, the study reported that moderate
and severe head injury, experienced by U.S servicemen during
World War II, was associated with a greater than two-fold
(HR = 2.32; 95% CI = 1.04, 5.10) and greater than four-fold (HR =
4.51; 95% CI = 1.77, 11.47) increased risk of AD, respectively.
The HTA reported by Williams et al. (2010) included one
previously conducted systematic review by Fleminger et al. (2003)
and three additional cohort studies. The included systematic
review, which relied on findings from 15 case-control studies,
yielded results suggesting that head injury was associated with an
increased risk of AD in males but not females. Two recent
prospective cohort studies identified in the HTA failed to detect an
association between self-reported traumatic brain injury and risk
of AD. The third cohort study, which was retrospective in nature
and relied on an all-male sample, detected an increased risk of AD
associated with moderate (Risk estimate: 2.32; 95% CI = 1.04, 5.17)
and severe (Risk estimate: 4.51; 95% CI = 1.77, 11.47) traumatic
brain injury. Mild TBI, however, was not associated with risk of AD
(Risk estimate: 0.76; 95% CI = 0.18, 3.29) according to the same
study. The reviewers concluded that traumatic brain injury,
particularly in males, appeared to be associated with an increased
risk of AD.
3.1.55.2. Discussion. Based on the evidence presented in four
moderate quality systematic reviews, there appears to be an
association between head injury and risk of AD. Subgroup analysis
conducted in two reviews suggests that this association was true
for males but not females. One recent cohort study, which relied on
an all-male sample, corroborated this finding (Williams et al.,
2010).
3.1.56. Anaemia, haemoglobin
We identified one moderate-quality systematic review (Peters
et al., 2008a) and one cohort study (Shah et al., 2011) that
examined the association between anaemia and haemoglobin with
risk of AD.
3.1.56.1. Systematic reviews. A total of three studies investigating
the association between haemoglobin or anaemia and risk of
dementia, cognitive decline, and AD were eligible for inclusion in
the review reported by Peters et al. (2008a). AD was the outcome of
interest in only one of the three included publications. The
publication reported on two studies, one of which was a
retrospective case-control study while the other employed a
longitudinal cohort design. In the retrospective case-control
analysis, anaemia was associated with an increased odds of AD
(OR = 1.88; 95% CI = 1.17, 3.03). However, when stratified by sex,
anaemia was associated with a statistically significant increased
odds of AD in women (OR = 1.96; 95% CI = 1.11, 3.47) but not in men
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
(OR = 1.81; 95% CI = 0.75, 4.39). In the longitudinal cohort study by
the same author, there was no association between anaemia and
risk of AD (SIR: 0.98; 95% CI = 0.67, 1.37). Sex stratified analysis
failed to detect an association in females (SIR: 0.79; 95% CI = 0.49,
1.23) and males (SIR: 1.49; 95% CI = 0.79, 2.56). Overall, the
reviewers concluded that anaemia appeared to increase risk of
dementia and cognitive decline. However, when the analysis was
restricted to AD, the association was unclear.
3.1.56.2. Primary studies. The association between haemoglobin
levels and risk of AD was prospectively assessed by Shah et al.
(2011). Low haemoglobin or anaemia, defined as <12 g/dL for
women and <13 g/dL men, was associated with an increased risk of
AD (HR = 1.60; 95% CI = 1.02, 2.52). Similarly, when compared to
subjects with normal haemoglobin levels, those with high
haemoglobin (>15.5 g/dL for women, >17.5 g/dL for men) were at
an increased risk of AD (HR = 3.39; 95% CI = 1.25, 9.20). Overall,
results of the study suggest that both low and high haemoglobin
levels were associated with a detrimental effect.
3.1.56.3. Discussion. According to the identified systematic review,
the association between anaemia and AD was inconclusive.
Findings from a newly published cohort study implicated both
high and low haemoglobin levels as a risk factor for the
development of AD. The small number of available studies
precludes a conclusive statement regarding the effect of
anaemia and haemoglobin on risk of AD.
3.1.57. Depression
We identified three moderate-quality reviews that reported on
the association between depression and risk of AD (Patterson et al.,
2007; Ownby et al., 2006; Williams et al., 2010). Two low-quality
reviews were excluded (Jorm, 2001; Barnes and Yaffe, 2011). Seven
relevant primary studies were identified (Blasko et al., 2010;
Dotson et al., 2010; Saczynski et al., 2010; Kohler et al., 2011; Lenoir
et al., 2011; Li et al., 2011; Barnes et al., 2012).
3.1.57.1. Systematic reviews. Twenty studies, representing 102 172
subjects, were included in the random effects meta-analysis
reported by Ownby et al. (2006). According to the pooled analysis,
depressed subjects had a two-fold increased risk of AD (OR = 2.02;
95% CI = 1.80, 2.26; test for heterogeneity: Egger test, t = 2.75;
p = 0.01). Subgroup analyses restricted to case-control (n = 9) and
cohort studies (n = 11) yielded similar results (Case-control

OR = 2.03; 95% CI = 1.73, 2.38; test for heterogeneity: p = 0.10.
Cohort
OR = 1.90; 95% CI = 1.55, 2.33; test for heterogeneity:
p = 0.02). The pooled analysis of retrospective cohort studies
(OR = 2.11; 95% CI = 1.82, 2.45) yielded a stronger risk estimate
when compared to that of prospective studies (OR = 1.78; 95%
CI = 1.16, 2.73). Results of the review suggest that depression was
associated with an increased risk of AD. This finding was confirmed
by subgroup analysis restricted to studies which employed widely
accepted criteria for diagnosis of both depression and AD (see
Table 62 in Supplementary material II).
Patterson et al. (2007) identified a single study which detected
an association between depression and risk of AD in males
(RR = 4.20; 95% CI = 2.10, 8.80) but not females (RR = 1.00; 95%
CI = 0.50, 1.90).
Williams et al. (2010) relied on findings from the systematic
review reported by Ownby et al. (2006). Five additional studies,
published after Ownby et al. (2006), were also considered. Ownby
et al. (2006) reported an increased risk of AD associated with
history of depression. All five included primary studies consistent-
ly detected a statistically significant increased risk of AD among
those with a history of depression. Based on these findings,
173
Williams et al. (2010) concluded that depression was associated
with an increased risk of AD.
3.1.57.2. Primary studies. Midlife symptoms of depression were
not associated with risk of AD (RR = 1.06; 95% CI = 0.85, 1.33)
according to the retrospective cohort study reported by Barnes
et al. (2012). Alternatively, symptoms experienced in late-life alone
(RR = 2.06; 95% CI = 1.67, 2.55) and in both mid- and late-life
(RR = 1.99; 95% CI = 1.47, 2.69) were associated with an increased
risk of AD. Findings from this study support the association
between depression, particularly in late-life, and risk of AD.
The prospective cohort study by Kohler et al. (2011) failed to
detect an association between depressive symptoms and risk of AD
among subjects enrolled in the Maastricht Ageing Study. More
specifically, depression, treated as a continuous (OR = 1.04; 95%
CI = 0.98, 1.09) and categorical (OR = 1.81; 95% CI = 0.78, 4.23)
variable, did not significantly increase risk of AD. As noted by the
authors, failure to detect a significant association may be
attributed to the small number of AD cases.
The effect of depressive symptoms at baseline, past and current
major depressive episodes, and lifetime treated depression on risk
of AD was prospectively examined in the Three-City study (Lenoir
et al., 2011). There was no association between depressive
symptoms at baseline and risk of AD (HR = 1.00; 95% CI, 0.70,
1.60). Similarly, lifetime treated depression and major depressive
episodes were not associated with AD. The study was limited by
the small sample size (Lenoir et al., 2011).
Using data obtained from the Adult Changes in Thought study,
Li et al. (2011) examined the association between depressive
symptoms, self-reported history of depression, and risk of AD, all-
cause dementia, and other cognitive outcomes. Depressive
symptoms increased risk of AD (HR = 1.43; 95% CI = 1.05, 1.94),
however, there was no association between self-reported history of
depression and risk of AD.
Using data from the Baltimore Longitudinal Study of Aging,
Dotson et al. (2010) examined the association between recurrent
elevated depressive symptoms (EDS) and risk of AD. Recurrence of
EDS was not associated with risk of AD (HR = 1.06; 95% CI = 0.90,
1.25). Compared to subjects without a history of EDS, those who
had one EDS were at a significantly increased risk of AD (HR = 1.83;
95% CI = 1.10, 3.02). Having two or more EDS was associated with an
increased risk of AD, however, results failed to reach statistical
significance (HR = 1.47; 95% CI = 0.73, 2.97).
Higher scores on the Geriatric Depression Scale were not
associated with risk of AD according to the VITA study (OR = 1.20;
95% CI = 1.00, 1.50) (Blasko et al., 2010).
The effect of depressive symptoms on risk of AD was examined
by Saczynski et al. (2010) using data from the Framingham Heart
Study. Depression was associated with an increased risk of AD
(HR = 1.76; 95% CI = 1.03, 3.01; p = 0.039). Treated continuously,
each 10 point increase in Center for Epidemiologic Studies
Depression Scale (CES-D) was associated with an increased risk
of AD (HR = 1.39; 95% CI = 1.11, 1.75; p = 0.005). Depression
remained significantly associated with AD following further
adjustments and in the subgroup analysis excluding subjects with
possible mild cognitive impairment. Based on these findings, the
authors concluded that there was evidence to suggest that late-life
depression was associated with an increased risk of AD.
3.1.57.3. Discussion. Evidence summarized in three moderate
quality systematic reviews was suggestive of an association
between history of depression and an increased risk of AD. All
but one observational study identified in the second phase of the
review reported risk estimates exceeding 1.00. Four of seven cohort
studies detected a statistically significant detrimental effect
associated with depression. Further research is required to
174 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187

determine whether depression is a risk factor for AD or an early 3.1.62. Hearing loss, central auditory dysfunction
symptom of the disease (Barnes et al., 2012; Li et al., 2011). No systematic reviews reporting on the association between
hearing loss or impairment were identified. However, our search
for observational studies yielded two relevant cohort studies
3.1.58. Mild Cognitive Impairment (MCI)
(Gates et al., 2011; Lin et al., 2011).
One moderate-quality systematic review examined the risk of
The association between central auditory dysfunction (CAD)
conversion from MCI to AD (Mitchell and Shiri-Feshki, 2009). Two
and risk of AD was examined using data obtained from the Adult
low-quality reviews were excluded from the qualitative synthesis
Changes in Thought (ACT) Study (Gates et al., 2011). CAD was
(Bruscoli and Lovestone, 2004; Mitchell and Shiri-Feshki, 2008).
ascertained according to three auditory tests (SSI-ICM, DDT, DSI)
No recent primary studies were identified.
with moderate and severe CAD defined as achieving scores less
The systematic review by Mitchell and Shiri-Feshki (2009)
than 50% and 80% correct on each test, respectively. Moderate CAD
examined the annual conversion rate from mild cognitive
was associated with an increased risk of AD according to results
impairment (MCI) to AD. A meta-analysis examining the risk of
from the DSI (HR = 6.80; 95% CI = 1.90, 24.10) and DDT (HR = 7.00;
AD among subjects with MCI, in comparison to healthy elderly
95% CI = 1.60, 31.00) tests. While the SSI-CCM produced a risk
controls, was also conducted. Among studies employing the Mayo
estimate above 1.00, the results failed to reach statistical
clinic definition for MCI diagnosis, the annual conversion rate
significance (HR = 2.50; 95% CI = 0.90, 7.50). Severe CAD was
(ACR) was 8.10% (95 CI: 6.30%, 13.40%) in “specialist settings” and
associated with a significant increased risk of AD according to
6.80% (95% CI = 1.90%, 14.50%) in “community settings”. Six studies
the DSI (HR = 9.90; 95% CI = 3.60, 26.70) but not the DDT (HR = 2.20;
examined the risk of AD among individuals diagnosed with mild
95% CI = 0.70, 7.00) or SSI-CCM (HR = 2.10; 95% CI = 0.70, 6.60).
cognitive impairment (MCI) in comparison to healthy elderly
Considered cumulatively, the authors concluded that CAD
controls. According to the results from the random effects meta-
appeared to be a risk factor for AD.
analysis, those with MCI had a near nine-fold increased risk of AD
The association between hearing loss and risk of AD and all-
in comparison to healthy elderly controls (RR = 8.93; 95% CI = 4.18,
cause dementia was prospectively examined by Lin et al. (2011).
19.07).
Treated as a continuous variable, each 10 dB of hearing loss was
associated with an increased risk of dementia, however, results did
3.1.59. Sporadic cerebral amyloid angiopathy not reach statistical significance for AD (HR = 1.20; 95% CI = 0.94,
One identified systematic review (Keage et al., 2009) reported 1.53). As noted by the authors, the analysis was limited by the small
on the effect of sporadic cerebral amyloid angiopathy. However, number of incident AD cases.
the review was of low methodological quality and was subse-
quently excluded. 3.1.63. Declining health and self-Rated health
We identified one cohort study (Song et al., 2011) that reported
on the effect of declining health and another (Montlahuc et al.,
3.1.60. Essential tremor
2011) that examined the association between self-rated health and
One systematic review reported on the association between
risk of AD.
essential tremor and risk of AD (LaRoia and Louis, 2011). Due to low
Using data from the Canadian Study of Health and Aging, Song
methodological quality, the review was excluded.
et al. (2011) examined the risk of AD associated with “health
deficits not known to predict dementia”. A frailty index (FI-NTRF),
3.1.61. Cancer
which considered 19 deficit indicators, was developed to measure
We identified no systematic reviews which reported on the
the exposure of interest. Results from the minimally adjusted
association between cancer and risk of AD. Two cohort studies
model suggest that the FI-NTRF was associated with an increased
were identified in the second phase of the review (Roe et al., 2010;
risk of AD (OR = 1.03; 95% CI = 1.01, 1.05). However, results
Driver et al., 2012).
approached the null when putative risk factors for AD were
According to the Cognition Sub-study of the Cardiovascular
adjusted for in the analysis (OR = 1.01; 95% CI = 1.00, 1.03). Based on
Health Study (CHS), history of cancer among whites was associated
the results of the study, the authors concluded that health deficits
with a statistically significant reduced risk of AD (HR = 0.57; 95%
or declining health may predict risk of AD.
CI = 0.36, 0.90) (Roe et al., 2010). In contrast, history of cancer was
The Three City prospective cohort study detected an increased
associated with an increased risk among minority subjects
risk of AD among subjects who rated their health as poor
(HR = 5.04; 95% CI = 1.99, 12.80). As noted by the author, the latter
(HR = 1.48; 95% CI = 1.00, 2.24) when compared to subjects
finding may be attributed to the small sample size of minority
reporting to be in good health (Montlahuc et al., 2011).
subjects (n = 29) as well as racial differences with respect to cancer
type (Roe et al., 2010).
3.1.64. Mild Parkinsonian Signs
Results reported by Driver et al. (2012) suggest that cancer was
We did not find any systematic reviews which reported on the
associated with a reduced risk of probable AD among subjects
association between mild Parkinsonian signs (MPS) and risk of AD.
enrolled in the Framingham Heart Study (HR = 0.67; 95% CI = 0.47,
One relevant primary study was identified (Louis et al., 2010).
0.97). When results were stratified by type of cancer, only subjects
Louis et al. (2010) prospectively examined the association
with a history of smoking related cancers were at a significantly
between MPS and risk of AD. According to the results of the study,
reduced risk of probable AD (Smoking related cancers, HR = 0.26;
MPSs were significantly associated with risk of AD (HR = 1.78; 95%
95% CI = 0.08, 0.82. Non-smoking related cancers, HR = 0.82; 95%
CI = 1.20, 2.65; p = 0.004).
CI = 0.57, 1.19). However, the association failed to reach statistical
significance when the analysis was adjusted for APOE e4 status,
3.1.65. Spine surgery under general anesthesia
education, and homocysteine level (All cancers, HR = 0.76; 95%
No systematic review was identified that reported on the
CI = 0.52, 1.12. Smoking related cancers, HR = 0.34; 95% CI = 0.11,
association between spine surgery under general anesthesia and
1.08. Non-smoking related cancers, HR = 0.91; 95% CI = 0.61, 1.35).
risk of AD. One relevant primary study was identified (Zuo and Zuo,
No association between cancer and possible AD was detected.
2010).
Based on the results of the study, the authors concluded that
Zuo and Zuo (2010) retrospectively examined the association
cancer appeared to reduce risk of AD.
between spine surgery under general anesthesia and risk of AD.
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
The study failed to detect an association between volatile
anesthesia and risk of AD (OR = 1.27; 95% CI = 0.52, 3.09; p-
value = 0.603). There was also no association between AD and
length of hospital stay (OR = 0.99; 95% CI = 0.93, 1.05; p-value =
0.742) or length of operating room time (OR = 1.00; 95% CI = 0.99,
1.01; p-value = 0.960).
3.1.66. Various medical factors
Three moderate quality systematic reviews examined the
association between various medical factors and risk of AD
(Breteler et al., 1991; Patterson et al., 2007; Williams et al.,
2010). Medical factors analyzed in these reviews include thyroid
disorders, goitre, atopy, hay fever, rheumatoid arthritis, osteoar-
thritis, poliomyelitis, herpes zoster, herpes simplex, encephalitis/
meningitis, epilepsy, severe headaches/migraines, blood trans-
fusions, general anesthesia, and sleep apnea.
3.1.66.1. Systematic reviews. The systematic review and
“collaborative re-analysis” by Breteler et al. (1991) examined the
association between several exposures pertaining to medical
history and risk of AD. According to the re-analysis and pooled risk
estimates of eight case-control studies, there was no association
between AD and a majority of the examined medical conditions
and procedures. More specifically, thyroid disease (unspecified),
hyperthyroidism, goitre, atopy, hay fever, rheumatoid arthritis,
poliomyelitis, herpes zoster, herpes simplex, encephalitis/
meningitis, epilepsy, and general anesthesia were not associated
with AD. However, the reviewers suggest that failure to detect a
significant association may be attributed to inadequate statistical
power. While hypothyroidism was associated with a marginally
significant increased risk of AD, osteoarthritis and severe
headaches and migraines were associated with a marginally
significant decreased risk of AD. Finally, blood transfusions were
associated with a statistically significant 40% decreased risk of AD
(95% CI = 0.40, 0.90). As noted by the reviewers, however, this
finding was likely attributed to selection bias.
Patterson et al. (2007) identified a single cohort study which
examined the association between AD and serum thyroid
stimulating hormone. According to the study, a low level of
thyroid stimulating hormone was associated with an increased risk
of AD (RR = 3.50; 95% CI = 1.10, 11.50).
The HTA (Williams et al., 2010) examined the association
between sleep apnea and subsequent risk of AD. However, there
were no studies eligible for inclusion.
3.1.66.2. Primary studies. Our search of the literature identified
one prospective cohort study (Forti et al., 2012) that reported on
the association between serum thyroid stimulating hormone
(TSH), hypothyroidism and risk of AD. According to the fully
adjusted model, hypothyroidism (OR = 1.13; 95% CI = 0.53, 2.38)
was not associated with risk of AD. Compared to the lowest tertile
of serum thyroid stimulating hormone (<2.20 mIU/L), both middle
(2.20-3.59 mIU/L) and high (>3.59 mIU/L) tertiles of TSH were not
associated with risk of AD (Middle, OR = 1.14; 95% CI = 0.50, 2.57.
Highest, OR = 1.30; 95% CI = 0.61, 2.77). Treated continuously, each 1
SD increase in log transformed TSH was also not associated with
risk of AD (OR = 1.13; 95% CI = 0.86, 1.52).
3.1.66.3. Discussion. Evidence summarized in one moderate
quality systematic review suggested that: (a) goitre, atopy, hay
fever, rheumatoid arthritis, poliomyelitis, herpes zoster, herpes
simplex, encephalitis/meningitis, epilepsy, and general anesthesia
were not significantly associated with AD, (b) osteoarthritis and
severe headaches and migraines were associated with a marginally
significant decreased risk of AD, and (c) blood transfusions were
associated with a statistically significant decreased risk of AD
175
(Breteler et al., 1991). These findings must be interpreted with
caution due to the small number of available studies and reliance
on case-control studies. Furthermore, the review relied on older
studies published in the late 1980s and early 1990s. Current
observational studies are required prior to arriving at a definitive
conclusion regarding the effect of these factors on the risk of AD.
The association between thyroid function and risk of AD was
unclear. One review found a marginally significant increased risk of
AD associated with hypothyroidism but failed to detect an
association with thyroid disease (unspecified) and hyperthyroid-
ism (Breteler et al., 1991). Results from a single study identified by
Patterson et al. (2007) detected a greater than three-fold increased
risk of AD associated with low levels of thyroid stimulating
hormone. However, conflicting evidence was reported in the
recent prospective study identified in the second phase of the
review (Forti et al., 2012). More specifically, the study failed to
detect an association between levels of thyroid stimulating
hormone, hypothyroidism and risk of AD.
3.1.67. C-reactive protein
One moderate-quality systematic review reported on the
association between c-reactive protein and risk of AD (Kuo
et al., 2005b).
The systematic review conducted by Kuo et al. (2005b)
identified two nested case-control studies. One study detected
an increased risk of AD among men with higher concentrations of
CRP. However, the risk estimate derived for the greatest quartile of
concentration was only marginally significant (Second quartile,
OR = 2.80; 95% CI = 1.30, 6.30. Third quartile, OR = 5.40; 95%
CI = 2.50, 11.70. Fourth quartile, OR = 2.30; 95% CI = 1.00, 5.40).
Increasing CRP concentration was not associated with risk of AD
according to the second study (RR = 1.10; 95% CI = 0.96, 1.26).
Overall, results of the review suggest that CRP concentrations were
associated with risk of cognitive decline and dementia. However,
when the analysis was restricted to AD, the association was less
clear. Although both studies reported risk estimates greater than
1.00, results from only one reached statistical significance.
3.1.68. Plasma clusterin levels
No systematic review examined the effect of plasma clusterin
levels and risk of AD. One relevant primary study was identified.
The case-cohort study reported by Schrijvers et al. (2011)
examined the association between plasma clusterin levels, treated
continuously and categorically, and risk of prevalent and incident
AD. Treated as a continuous variable, each 1 SD increase in plasma
clusterin levels was associated with prevalent (OR = 1.63; 95%
CI = 1.21, 2.20) but not incident (HR = 1.00; 95% CI = 0.85, 1.17) AD.
When treated as a categorical variable (47.2- 99.5 mg/mL as
referent), only the highest quartile (132.5
198.0 mg/mL) of
plasma clusterin was associated with prevalent AD (OR = 2.99; 95%
CI = 1.25, 7.16). The test for trend was statistically significant,
however (p = 0.004). Alternatively, there was no association
between increasing quartiles of plasma clusterin and risk of
incident AD (see Table 79 in Supplementary material II).
3.1.69. White matter hyperintensities
One moderate-quality systematic review reported on the
association between white matter hyperintensities (WMH) and
risk of AD (Debette and Markus, 2010). One primary study, which
reported on the association between white matter changes and risk
of AD, was also identified (Verdelho et al., 2010).
3.1.69.1. Systematic reviews. A total of six studies examining the
association between white matter hyperintensities, deep white
matter hypertensities, and/or periventricular hyperintensities
with AD risk were eligible for inclusion in the review reported
176 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187

by Debette and Markus (2010). All of the included studies were
prospective population-based studies or prospective studies
conducted in high-risk populations. One population-based study
detected an association between white matter hyperintensities
and risk of AD (WMH >3, HR = 1.50; 95% CI = 1.17, 1.99). The
remaining three studies, all of which examined subjects with MCI,
failed to detect an association. Results from three studies were
eligible for inclusion in a meta-analysis. The pooled analysis
detected an increased risk of AD associated with WMH (HR = 1.50;
95% CI = 1.20, 1.80; p for heterogeneity = 0.74; I2 = 0%). All four
studies, which reported on the effect of deep white matter
hyperintensities (DWMH), failed to detect an association. Two of
three studies, one of which examined subjects with amnestic MCI,
detected an association between periventricular hyperintensities
(PVH) and risk of AD. Overall, results of this review suggest that
WMH were associated with an increased risk of incident AD.
However, this association was not observed in high risk
populations, namely, those with MCI. PVH appeared to increase
risk of AD in high-risk and general populations, while DWMH were
not associated with risk of AD.
3.1.69.2. Primary studies. According to results of the LADIS
prospective cohort study (Verdelho et al., 2010), white matter
change severity was not significantly associated with risk of AD
(Moderate, HR = 0.29; 95% CI = 0.06, 1.41; p = 0.126. Severe,
HR = 1.29; 95% CI = 0.46, 3.65; p = 0.632). Medial temporal
atrophy (MTA), however, was significantly associated with AD
(p = 0.046). A MTA score of 4, indicating most severe atrophy, was
significantly associated with risk of AD (HR = 15.62; 95% CI = 1.95,
124.77; p = 0.010). Lower scores were not significantly associated
with AD (see Table 87 in Supplementary material II).
3.1.70. Age
One moderate-quality systematic review reported that age was
a “significant [predictor] of incidence of AD (F42 = 15.6,
p = 0.0003 . . . )” (Gao et al., 1998). Seven primary studies reporting
on the effect of age were also identified (Song et al., 2011; Verdelho
et al., 2010; Li et al., 2010a,b; Reitz et al., 2010b; Yang et al., 2011;
Shah et al., 2011; Zuo and Zuo, 2010). Six studies detected an
increased risk of AD associated with age. The remaining study
failed to detect an association (Verdelho et al., 2010).
contrast, both the Rush Memory and Aging Project (Shah et al.,
2011) (HR = 1.49; 95% CI = 0.95, 2.35) and the Canadian Study
of Health and Aging (Song et al., 2011) (OR = 1.13; 95% CI = 0.85,
1.49) failed to detect an association between male sex and risk of
AD.
3.1.71.3. Discussion. A pooled analysis (n = 7) reported in a
moderate-quality systematic review detected a statistically
significant increased risk of AD among females. However, all but
one of the six recent primary studies failed to detect an association
between sex and AD.
3.1.72. Sex hormones and sex hormone binding globulin
One moderate-quality systematic review reported on the
association between sex hormones and risk of AD (Patterson
et al., 2007). We also identified two cohort studies which
reported on the association between AD and sex hormones or
sex hormone binding globulin (SHBG) (Muller et al., 2010; Chu
et al., 2010).
3.1.72.1. Systematic reviews. The systematic review by Patterson
et al. (2007) examined the association between sex hormones and
risk of AD. One of two included studies failed to detect an
association between high estrogen levels and risk of AD in females
(RR = 1.30; 95% CI = 0.88, 1.99). The second study reported that
higher free testosterone index in males was protective against the
development of AD (RR = 0.74; 95% CI = 0.57 to 0.96).
3.1.72.2. Primary studies. A prospective cohort study identified in
the second phase of the review examined the association between
testosterone and risk of AD in an all-male sample of Chinese
subjects (Chu et al., 2010). The study detected an association
between late-life baseline serum bioavailable testosterone (BT)
level and risk of AD. More specifically, each unit (one nmol/L)
increase of serum BT level was associated with a reduced risk of AD
(RR = 0.22; 95% CI = 0.07, 0.69). Serum total testosterone (TT) and
SHBG levels were not predictors of AD.
Results reported by Muller et al. (2010) detected an increased
risk of AD associated with SHBG (HR = 1.30; 95% CI = 1.10, 1.60).
Higher levels of SHBG increased risk in men (HR = 1.30; 95%
CI = 1.00, 2.00) and women (HR = 1.40; 95% CI = 1.10, 1.90).

3.1.71. Sex
3.1.73. Ethnicity, race
One moderate-quality systematic review examined the associ-
One systematic review, which reported on the effect of
ation between sex and AD (Gao et al., 1998). We also identified six
ethnicity/race, was identified (Shadlen et al., 2000). However,
relevant primary studies.
the study was deemed to be of low methodological quality and was
excluded from the review. Two relevant cohort studies were
3.1.71.1. Systematic reviews. The systematic review and meta-
identified in the second phase of the review (Reitz et al., 2010b;
analysis by Gao et al. (1998) examined the association between
Yang et al., 2011).
age, sex, and risk of AD. According to the pooled analysis of the
Compared to white subjects, black (HR = 1.94; 95% CI = 0.84,
included studies, females had a 56% increased risk of AD (OR = 1.56;
4.52) and Hispanic (HR = 1.69; 95% CI = 0.67, 4.22) subjects were not
95% CI = 1.16, 2.10).
at a significantly increased risk of developing possible and
probable LOAD according to Reitz et al. (2010b). Similar results
3.1.71.2. Primary studies. Zuo and Zuo (2010) retrospectively
were derived when the analysis was restricted to probable LOAD
examined the association between spine surgery under general
(Black, HR = 1.82; 95% CI = 0.69, 4.82. Hispanic, HR = 2.45; 95%
anesthesia and risk of AD. According to the results of the study, sex
CI = 0.90, 6.72).
was not significantly associated with risk of AD post-surgery
The longitudinal cohort study conducted by Yang et al. (2011)
(OR = 1.17; 95% CI = 0.51, 2.69; p-value = 0.702). Similarly, Yang et al.
examined the association between blood pressure, hypertension,
(2011) failed to detect an association between sex and risk of AD.
and risk of AD. The association between blood pressure measures
Female sex was not significantly associated with risk of possible
were analyzed in separate models. The association between race,
and probable LOAD according to results reported by Reitz et al.
treated as a dichotomous variable (Caucasian vs. African-Ameri-
(2010b) (HR = 1.15; 95% CI = 0.61, 2.21). Similar results were derived
can), and risk of AD was examined in each model. According to the
when the analysis was restricted to probable LOAD (HR = 1.14; 95%
risk estimates derived from each model, Caucasians had a reduced
CI = 0.51, 2.54).
risk of AD when compared to African-Americans (see Table 85 in
The VITA study detected an association between male sex and
Supplementary material II).
risk of AD (OR = 3.30; 95% CI = 1.30, 8.40) (Blasko et al., 2010). In
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
3.1.74. Maternal age
The systematic review and re-analysis of case-control studies
reported by Rocca et al. (1991) examined the association between
maternal age and risk of AD. The pooled analysis for early maternal
age yielded a risk estimate of 1.50, however, it failed to reach
statistical significance (RR = 1.50; 95% CI = 0.80, 3.00). The pooled
analysis detected a marginally significant 70% increased risk of AD
associated with a maternal age of 40 years and above (RR = 1.70;
95% CI = 1.00, 2.90). Maternal age between 20 and 24 years
(RR = 0.90; 95% CI = 0.60, 1.40), 30 to 34 years (RR = 0.90; 95%
CI = 0.60, 1.30), and 35 to 39 years (RR = 1.00; 95% CI = 0.60, 1.60)
were not associated with AD. Overall, results of the review suggest
that late maternal age may increase risk of AD. The pooled analysis
yielded a risk estimate above 1.00 for early maternal age, however,
results did not reach statistical significance.
3.1.75. Parental longevity
We did not identify any systematic reviews that reported on the
association between parental longevity and risk of AD. One cohort
study was identified (Lipton et al., 2010). When compared to the
offspring of parents with usual survival, having a parent with
exceptional longevity (85 years) was associated with a protective
effect against AD according to results of the fully adjusted model
(HR = 0.58; 95% CI = 0.35, 0.96).
3.1.76. Bonemineral density
No systematic reviews examined the association between bone
mineral density (BMD) and risk of AD. One relevant cohort study
was identified. The prospective cohort study reported by Zhou
et al. (2011) detected an increased risk of AD associated with lower
quartiles of BMD among Chinese subjects (Quartile 1, HR = 2.68;
95% CI = 1.53, 4.71. Quartile 2, HR = 2.11; 95% CI = 1.17, 3.80. Quartile
3, HR = 1.67; 95% CI = 0.90, 3.12). Using the lowest quartile as a
reference, the study detected that higher rates of bone loss were
also detrimental. While adiponectin was not associated with AD
risk (HR = 0.98; 0.96, 1.01), the study detected a reduced risk of AD
associated with higher levels of plasma leptin (HR = 0.93; 95%
CI = 0.91, 0.95).
3.1.77. Retinal vascular calibers
No systematic reviews reported on the association between
retinal vascular calibers and risk of AD. However, one prospective
cohort study was identified in the second phase of the review (de
Jong et al., 2011). According to the fully adjusted model, each SD
decrease in arteriolar caliber was not associated with risk of AD
without cardiovascular disease (CVD) (HR = 1.02; 95% CI = 0.91,
1.14). Although the risk estimate was greater in magnitude, results
also failed to reach statistical significance for AD with CVD
(HR = 1.27; 95% CI = 0.89, 1.82). The study also failed to detect an
association between venular caliber (per 1 SD increase) and risk of
AD with (HR = 1.16; 95% CI = 0.82, 1.64) and without (HR = 1.06; 95%
CI = 0.95, 1.19) CVD. According to the results of the study, there was
no significant association between AD and smaller arteriolar
calibers or larger venular calibers.
3.1.78. Beta-amyloid, t , cystatin C
We did not identify any systematic reviews that examined the
effect of beta-amyloid, t, or cystatin C and risk of AD. Four
observational studies were identified in the second phase of the
review (Blasko et al., 2010; Leinonen et al., 2010; Sundelöf et al.,
2012; Shah et al., 2012).
3.1.78.1. Primary studies. Presence of amyloid beta (Ab) alone
(OR = 10.8; 95% CI = 4.90, 23.8; p < 0.001) and with
hyperphosphorylated t (OR = 68.2; 95% CI = 22.1, 210; p < 0.001)
in right frontal cortical biopsy samples was associated with
177
increased risk of AD according to results reported by Leinonen et al.
(2010).
According to the results reported by Sundelöf et al. (2012) there
was no association between cystatin C levels in cerebrospinal fluid
(CSF) and risk of AD (OR = 0.83; 95% CI = 0.57, 1.19; p = 0.32). Lower
Ab42 (OR = 0.09; 95% CI = 0.04, 0.21; p < 0.001) and higher total
(OR = 6.50; 95% CI = 3.31, 12.7; p < 0.001) and phosphorylated
(OR = 4.06; 95% CI = 2.23, 7.39; p < 0.001) t in CSF were significantly
associated with AD.
Using data from the VITA study, Blasko et al. (2010) detected
that higher baseline levels of plasma Ab42 was associated with
risk of AD among subjects who were cognitively healthy at baseline
(OR = 2.00; 95% CI = 1.20, 3.30).
The association between beta-amyloid, midlife blood pressure,
and risk of AD was prospectively examined by Shah et al. (2012)
using data from an all-male sample of Japanese-American men.
When treated as a continuous variable, decreasing (per 1 SD)
plasma Ab1-40 and Ab1-42 levels were associated with an
increased risk of “all” and “pure” AD (see Table 80 in Supplemen-
tary material II). Treated categorically, the lowest quartiles of Ab1-
40 (All AD, HR = 4.30; 95% CI = 1.90, 9.80. Pure AD, HR = 4.00; 95%
CI = 1.50, 10.70) and Ab1-42 (All AD, HR = 3.10; 95% CI = 1.20, 8.00.
Pure AD, HR = 4.30; 95% CI = 1.20, 15.90) were associated with an
increased risk of both pure and all AD. Additional analyses detected
that the presence of high diastolic blood pressure strengthened the
association between lower Ab1-40, Ab1-42 and risk of AD.
Considered cumulatively, the study suggested that lower plasma
Ab was associated with an increased risk of AD.
3.1.78.2. Discussion. Cortical amyloid beta (Leinonen et al., 2010)
and lower Ab42 in CSF (Sundelöf et al., 2012) were associated with
AD according to two studies. Higher plasma levels of Ab42
predicted AD risk in the VITA study (Blasko et al., 2010). The second
study detected a detrimental effect associated with lower plasma
Ab42 and Ab40 (Shah et al., 2012). With respect to t, one study
(Leinonen et al., 2010) detected an association between presence of
hyperphosphorylated t and amyloid beta in right frontal cortical
biopsy samples and risk of AD, while another (Sundelöf et al., 2012)
reported an association between AD and higher total and
phosphorylated t in CSF. The only study to report on the effect
of cystatin C failed to detect an association with risk of AD
(Sundelöf et al., 2012).
3.1.79. Family History of Dementia, Down’s Syndrome and Parkinson’s
disease
One moderate-quality systematic review examined the effect of
family history of dementia, Down’s syndrome, and Parkinson’s
disease on the risk of AD (van Duijn et al., 1991).
3.1.79.1. Systematic reviews. A total of 7 case-control studies
identified by van Duijn et al. (1991) examined the association
between family history of dementia and AD. The meta-analysis
combining the results of all seven studies detected a greater than
three-fold increased risk of AD among those with a family history
of AD (RR = 3.50; 95% CI = 2.60, 4.60). The association remained
statistically significant when data from an outlying study was
excluded (RR = 3.60; 95% CI = 2.70, 4.90). Sensitivity analyses by
gender detected an increased risk in both males (RR = 3.90; 95%
CI = 2.50, 6.50) and females (RR = 3.30; 95% CI = 2.30, 4.60) with a
first degree relative with dementia. Family history was associated
with risk of AD irrespective of age of onset.
Five of the included studies examined the association between
family history of Down's syndrome and AD. Risk estimates could
only be derived for two studies. The pooled analysis yielded a
statistically significant risk estimate (RR = 2.70; 95% CI = 1.20, 5.70).
178 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
The pooled analysis of two studies detected a marginally
significant increased risk of AD among those with a family history
of Parkinson’s disease (RR = 2.40; 95% CI = 1.00, 5.80). The pooled
analysis stratified by sex failed to detect an association in both
males and females (Males
RR = 4.40; 95% CI = 0.90, 20.90. Females

RR = 1.60; 95% CI = 0.50, 4.90).
3.1.79.2. Discussion. The available evidence suggested that a
positive history of dementia and Down’s syndrome was
associated with an increased risk of AD. Parkinson’s disease was
associated with only a marginally significant increased risk of AD.
These findings must be interpreted with caution due to the small
number of available studies and reliance on case-control studies.
Furthermore, the review relies on older studies published in the
late 1980s and early 1990s. Current observational studies are
required prior to arriving at a definitive conclusion regarding the
effect of these factors on the risk of AD.
3.1.80. IgM antibodies against PC
One primary study was identified that reported on the effect of
IgM antibodies against phosphorylcholine (anti-PC). The associa-
tion between low levels of IgM antibodies against PC (anti-PC)
and both incident and prevalent AD was examined by Eriksson
et al. (2010b) using a nested case-control and case-control design,
respectively. The lowest 25th and 10th percentile of serum anti-
PC were not associated with incident AD. When treated as tertiles,
lower tertiles of serum anti-PC were not associated with incident
AD. With respect to prevalent AD, those in the lowest 25th
( > 31.6 U/ml vs 31.6 U/ml, OR = 2.70; 95% CI = 1.45, 4.99) and 10th
( > 19.6 U/ml vs. 19.6 U/ml, OR = 3.98; 95% CI = 1.81, 8.74)
percentile were at a significantly increased risk. Compared to
the highest tertile, those in the lowest tertile of serum anti-PC
were at a significantly increased risk of prevalent AD (OR = 2.36;
95% CI = 1.14, 4.90). Results of the study suggest low levels of anti-
PC were associated with prevalent but not incident AD.
3.1.81. Gait speed
A single cohort study reported on the association between
gait speed and risk of dementia or AD. According to the results of
the EPIDOS-Toulouse cohort study, slow gait speed (OR = 3.38; 95%
CI = 1.80, 6.33) was associated with a statistically significant
increased risk of AD (Abellan van Kan et al., 2012). No
association was detected between AD and total lean mass
(OR = 0.57; 95% CI = 0.24, 1.34) or total fat mass (OR = 2.12; 95%
CI = 0.90, 4.99).
3.1.82. Propositional density
A single primary study was identified that reported on the
association between propositional density and risk of AD.
Propositional density, defined as a measure of “complexity of
written and spoken language”, was associated with risk of AD
according to the nested case-control study reported by
Engelman et al. (2010). Subjects with higher scores of proportional
density, measured in early life, were at a reduced risk of AD
according to the matched analysis for age and sex (OR = 0.16; 95%
CI = 0.03, 0.90; p = 0.02). The association was slightly attenuated
after additional adjustments for age (OR = 0.22; 95% CI = 0.03, 1.36;
p = 0.05).
3.1.83. Purpose in life
We identified a single primary study which reported on the
effect of purpose in life. According to the Rush Memory and Aging
Project (Boyle et al., 2010), greater purpose in life was significantly
associated with a reduced risk of AD according to the minimally
(HR = 0.48; 95% CI = 0.33, 0.69; p < 0.001) and fully (HR = 0.60; 95%
CI = 0.39, 0.92; p = 0.02) adjusted models.
3.2. Risk factors associated with AD progression
Twelve moderate or high quality systematic reviews satisfied
the inclusion criteria of the present review. Reviews included
examined the effect of education (Paradise et al., 2009; Meng and
D'Arcy, 2012), APoE4 (Allan and Ebmeier, 2011), folic acid (Dangour
et al., 2010; Malouf et al., 2003; Malouf and Grimley, 2008),vitamin
B-12 (Malouf and Areosa, 2003), fatty acids (Issa et al., 2006;
Dangour et al., 2010), vitamin E (Farina et al., 2012), lecithin
(Higgins and Flicker, 2000), physical activity (Heyn et al., 2004;
Littbrand et al., 2011), and cognitive stimulation (Woods et al.,
2012) on the progression of AD. Methodological characteristics and
results of the included systematic reviews are summarized (see
Table 122 in Supplementary material II).
3.2.1. Education
The systematic review by Paradise et al. (2009) examined the
effect of educational attainment on survival in AD. A total of 22
cohort studies were identified. Using the Oxford Centre for
Evidence Based Medicine quality appraisal criteria, 11 of the
cohort studies were deemed to be of good methodological quality.
Of the good quality reviews, only one study detected an association
between educational attainment and survival. More specifically,
the study detected shorter survival in subjects with higher
educational attainment (HR = 1.10; p = 0.01). Of the lower quality
studies, only one study detected an association. This study
detected shorter survival in subjects with higher educational
attainment (HR = 1.06; 95% CI = 1.01, 1.11). Collectively, results of
the systematic review do not support an association between
educational attainment and survival in AD (Paradise et al., 2009).
Only two of 22 cohort studies detected an association between
higher educational attainment and shorter survival in subjects
with AD (Paradise et al., 2009).
The systematic review by Meng and D'Arcy (2012) examined
the effect of education on the onset and progression of AD. The
reviewers identified primary studies which used cognitive decline
(n = 20) and mortality (n = 8) as indicators of clinical progression.
With respect to cognitive decline, all but six studies detected an
association between higher educational attainment and faster
cognitive decline. The reviewers concluded that higher educational
attainment was associated with faster cognitive decline among
subjects with dementia/AD. As noted by the reviewers, this finding
was in support of the cognitive reserve hypothesis which posits
that higher educational attainment delays the clinical detection of
the disease and is correlated with worse brain damage upon
diagnosis and faster cognitive decline. Of the eight studies that
reported on the effect of education on mortality, all but three failed
to detect an association. Two studies detected reduced survival
among subjects with higher educational attainment and one study
detected reduced survival among AD/dementia subjects with
lower educational attainment. Overall, results of the review
suggest that higher educational attainment was associated with
faster cognitive decline but not with mortality (Meng and D'Arcy,
2012).
3.2.2. APoE4
The systematic review and meta-analysis conducted by Allan
and Ebmeier (2011) examined the effect of APoE4 status on
progression of AD. Progression was defined according to cognitive
decline and mortality. In the random effects meta-analysis
including 16 studies and a total of 1618 subjects, APOE e4 status
was not associated with cognitive decline (effect size: 0.025; 95%
CI =
0.09, 0.14; Q = 16.4; p = 0.36; I2 = 8.41). There was also no
association between APoE4 and mortality according to the random
effects meta-analysis including four studies (effect size: 0.74; 95%
CI = 0.36, 1.53; Q = 8.83; p = 0.03; I2 = 66.0%). Results of the review
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
suggest that APoE4 status does not predict clinical progression
of AD. More specifically, there was no difference in cognitive
decline or mortality when carriers and non-carriers were
compared.
3.2.3. Folic acid
The Cochrane Systematic Review reported by Malouf et al.
(2003) examined the effect of folic acid with or without vitamin B-
12 on the progression of dementia or onset of cognitive
impairment. The review included four randomized controlled
trials, two of which reported on progression of dementia/AD. One
study reported that participants randomized to receive 10 mg of
folic acid per day (duration unknown) did not significantly differ
from the placebo group with respect to cognitive tasks. The second
study also failed to detect a significant difference between groups
(2 mg folic acid plus 1 mg vitamin B12/day vs. placebo for 12
weeks) with respect to cognitive decline, measured according to
Mini Mental State Examination (MMSE) (weighted mean differ-
ence (WMD): 0.39; 95% CI =
0.43, 1.21; p = 0.35) and the cognitive
domain of the Alzheimer's Disease Scale (WMD: 0.41; 95%
CI =
1.25, 2.07; p = 0.63). There was also no difference between
groups with respect to activities of daily living, measured by the
Bristol Activity of Daily Living instrument (WMD:
0.57; 95%
CI =
1.95, 0.81; p = 0.42).
The Cochrane review (Malouf et al., 2003) was updated in 2008
(Malouf and Grimley, 2008). The reviewers identified two
additional RCT studies examining the effect of folic acid on
progression of dementia/AD. One study failed to detect a difference
between groups (10 mg folic acid vs. placebo for 10 weeks) with
respect to measures of memory, language, speed and concentra-
tion. The second study detected no significant difference between
groups (1 mg folic acid/day vs. placebo for 24 weeks) with respect
to cognitive performance, measured by MMSE (WMD:
0.13; 95%
CI =
1.96, 1.70, P = 0.89) or Digit Symbol Substitution Test (DSST)
(WMD: 0.26; 95% CI =
4.12, 4.64, p = 0.91). Similarly, there was no
difference between groups with respect to behaviour measured by
the Social Behaviour Subscale (WMD: 1.38; 95% CI =
1.13, 3.89,
p = 0.28). Patients assigned to the intervention group, however,
performed better on the Instrumental Activities of Daily Living
Scale (WMD: 2.67; 95% CI = 0.25, 5.09, p = 0.031). When data from
the aforementioned study was combined with one of the studies
identified in the original review, there was no difference between
groups with respect to MMSE (WMD: 0.30; 95% CI =
0.45, 1.05;
Test for heterogeneity: x2 = 0.26, df = 1, p = 0.61, I2 = 0.0%). The
second study identified in the updated review failed to detect a
difference between groups (10 mg folic acid vs. placebo for 10
weeks) with respect to measures of memory, language, speed and
concentration. Overall, folic acid with or without vitamin B-12 had
no effect on cognitive decline/performance, or a number of other
outcomes, in subjects with dementia/AD (Malouf et al., 2003;
Malouf and Grimley, 2008).
The systematic review by Dangour et al. (2010) examined the
association between folic acid supplementation, with or without
vitamin B, on cognitive function among subjects with dementia
and AD. The review included 10 randomized controlled trials, six
of which examined the effect of folic acid combined with B-
vitamins and four of which examined the isolated effect of folic
acid. There was methodological heterogeneity across studies with
respect to the tools used to assess cognitive function. Three of the
four trials reporting on isolated folic acid supplementation
reported a beneficial effect. All six studies that examined folic
acid supplementation combined with vitamin B failed to detect a
beneficial effect. Overall, the reviewers concluded that the effect
of folic acid supplementation was inconclusive (Dangour et al.,
2010).
179
3.2.4. Vitamin B12
The Cochrane systematic review reported by Malouf and Areosa
Sastre (2003) examined the effect of vitamin B-12 supplementa-
tion on cognition in subjects with and without dementia. The
review included three randomized controlled trials, only one of
which specifically enrolled subjects with AD. The remaining
studies examined subjects with cognitive impairment and
unspecified dementia. According to the one RCT in AD subjects,
there was no difference between the intervention group (n = 6
receiving 10 mg over 5 months) and the placebo group (n = 5) with
respect to cognitive function measured by the cognitive domain of
the Alzheimer's Disease Scale (mean difference (MD) = 0.04; 95%
CI =
5.95, 6.03). The second RCT also reported no difference in
cognition measured by the Mini Mental State Examination (MMSE)
between experimental (10 mg/day or 50 mg/day of cyanocobalamin
for one month) and placebo among dementia subjects (50mcg vs.
placebo, MD =
0.60; 95% CI =
3.06 to 1.86, p = 0.63. 10mcg vs.
placebo, MD =
1.60 95% CI =
3.99 to 0.79, p = 0.19). The final
study failed to detect a difference in cognitive function between
intervention (1 mg/week for 4 weeks) and placebo. More
specifically, groups did not differ with respect to MMSE (MD =
0.10; 95% CI:
0.60, 0.80, P = 0.70), the Cambridge Cognitive
Examination (MD =
0.60; 95% CI =
2.20 to 0.90; P = 0.43), and the
12-words learning test immediate recall (MD =
0.20; 95% CI:

0.70, 0.30, P = 0.42). Overall, the review suggested that vitamin B-
12 did not improve cognitive function among subjects with AD,
dementia, or cognitive impairment (Malouf and Areosa, 2003). As
noted by the reviewers, the identified studies were limited by the
very small sample size of subjects.
3.2.5. Vitamin E
The Cochrane systematic review by Farina et al. (2012)
examined the effect of vitamin E supplementation on progression
of MCI and AD. The review included two randomized controlled
trials examining treatment effect in subjects with AD and one
study enrolling subjects with MCI. The RCT reporting on AD
examined the outcome of “survival time to any one of four end
points; death, institutionalisation, change in severity of dementia
to a CDR of three or loss of two basic activities of daily living”
(Farina et al., 2012). According to the adjusted analysis controlling
for baseline MMSE, time to the endpoints was delayed in AD
subjects assigned to the intervention group (2000 IU/day)
compared to the placebo group (RR 0.47; P = 0.001). The second
study examined the effect of 800 IU/day for six months on
cognitive function measured by the Clock drawing test, Blessed
Dementia Scale (BDS), and MMSE. Treatment with vitamin E had
no effect on cognitive function measured by the Clock drawing test
and the BDS. However, treated subjects who did not respond to
vitamin E supplementation, assessed by oxidative stress, experi-
enced a more pronounced decrease in cognitive function measured
by MMSE when compared to placebo. There was no difference
between placebo and treated subjects who responded. According
to the final study, compared to placebo, 2000 IU/day of vitamin E
did not prevent conversion of MCI to AD (HR = 1.02; 95%CI: 0.74,
1.41; P = 0.91). Considered cumulatively, the reviewers concluded
that vitamin E supplementation did not appear to be efficacious in
the prevention of AD progression or progression of MCI to AD.
3.2.6. Fatty acids
The systematic review by Dangour et al. (2010) examined the
effect of fatty acids on cognitive function among subjects with AD
or dementia. Four randomized controlled trials were captured in
the review, two of which detected a beneficial effect with respect
to cognitive function or quality of life. Based on the conflicting
findings across studies, the reviewers concluded that the effect of
180 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
fatty acid supplementation on cognitive function in AD was
unclear.
The systematic review by Issa et al. (2006) examined the effect
of omega-3 fatty acids on the incidence and treatment of dementia.
The review included one randomized controlled trial examining
the effect of omega-3 fatty acid supplementation on cognitive
function. Briefly, subjects assigned to receive 4.3 g/day of
docosahexaenoic acid (DHA) experienced improvements in
cognitive function measured by the Hasegawa’s Dementia Rating
Scale (HDS-R) andMini Mental State Examination (MMSE).
However, all subjects enrolled in the study had vascular dementia,
therefore, the applicability of this trial was limited as our review
was focused on AD.
3.2.7. Lecithin
The Cochrane systematic review reported by Higgins and
Flicker (2000) examined the effect of lecithin on the progression of
AD and other dementia subtypes. The review included ten relevant
trials examining progression of AD. Studies were heterogeneous
with respect to the outcome or indicator of progression examined.
Two studies reported on global impression, one reported on
functional performance, two reported on behaviour, one reported
on cognitive function, two reported on specific domains of
cognitive function, and three reported on death. The random
effects meta-analysis combining results from two studies failed to
detect a difference between groups with respect to global
impression (OR = 3.01; 95% CI = 0.92, 9.81; test for heterogeneity:
x2 = 0.94, p = 0.33, I2 = 0%). One RCT failed to detect a statistically
significant difference between the intervention and placebo group
with respect to functional performance (MD: 0.76; 95% CI =
0.91,
2.43). There was no difference in behaviour between groups
according to the pooled analysis of two studies (standard mean
difference (SMD) = 0.09; 95% CI =
0.58, 0.77; test for heterogene-
ity: x2 = 0, p = 0.97, I2 = 0%). There was no difference in cognitive
function between groups according to one trial (OR = 0.91; 95%
CI = 0.25, 3.34). One RCT failed to detect a difference between
groups with respect to memory (SMD: 0.40; 95% CI =
0.66, 1.46).
The pooled analysis of two studies failed to detect a difference
between intervention and placebo with respect to cognitive
function measured by the Paired Associated Learning Test (SMD:

0.03; 95% CI =
1.48, 1.42; test for heterogeneity: x2 = 3.83,
p = 0.05, I2 = 74%). Similarly, lecithin did not exert a beneficial effect
on orientation according to the pooled analysis of two studies
(SMD:
0.12; 95% CI =
0.84, 0.60; test for heterogeneity: x2 = 1.38,
p = 0.24, x2 = 28%). Finally, the only study to provide adequate data
on death failed to detect a statistically significant difference in
mortality between groups (OR = 0.31; 95% CI = 0.03, 3.38). Consid-
ered cumulatively, the reviewers concluded that lecithin did not
appear to delay the progression of AD.
3.2.8. Physical activity
The effect of physical activity on cognitive function, physical
function, physical fitness, and behaviour was examined by Heyn
et al. (2004). Randomized trials examining the effect of physical
activity in older ( > 65 years of age) adults with dementia, AD, or
other cognitive impairments were sought. Thirty trials satisfied the
eligibility criteria. According to the series of pooled analyses,
subjects assigned to the intervention group experienced statisti-
cally significant improvements in functional performance (ES =
0.59; 95% CI = 0.43, 0.76, P < 0.001), behaviour (ES = 0.54; 95%
CI = 0.36, 0.72, P < 0.001), and cognitive function (ES = 0.57; 95%
CI = 0.38, 0.75, P < 0.001). Exercise training also proved to be
beneficial for physical fitness (ES = 0.69; 95% CI = 0.58, 0.80), a
measure which encompassed several outcomes including cardio-
vascular fitness, strength, flexibility, and BMI. When all study
outcomes (cardiovascular, strength, flexibility, BMI, cognitive,
functional, behaviour) were combined, a moderate effect size
was derived (ES = 0.62; 95% CI = 0.55, 0.70, P < 0.001). Results of the
review suggest that physical activity was associated with a
beneficial effect on several outcomes including cognition, behav-
iour, physical function and physical fitness (Heyn et al., 2004).
The systematic review by Littbrand et al. (2011) examined the
effect of physical activity on indicators of progression. Inclusion
was restricted to randomized controlled trials. A systematic search
of the literature yielded ten relevant studies, four of which were
deemed to be of moderate methodological quality according to the
Physiotherapy Evidence Database (PEDro) instrument. The
remaining studies were of low quality. Studies were heterogeneous
with respect to the interventions under study. One (low-quality) of
four studies detected a difference between intervention and
control groups with respect to cognitive function. Of the three
moderate-quality studies reporting on the effect of physical
function, two reported beneficial effects associated with exercise.
The only study which reported on the effect of exercise on activities
of daily living (ADL) reported a beneficial effect, namely, reduced
ADL decline. The reviewers concluded that there was no evidence
of improved cognitive function or mobility associated with
physical exercise among subjects with dementia. Alternatively,
there was preliminary evidence to suggest a beneficial effect of
weight-bearing exercise with respect to walking performance and
ADLs (Littbrand et al., 2011).
3.2.9. Cognitive stimulation
The Cochrane systematic review by Woods et al. (2012)
examined the effect of cognitive stimulation on cognitive and
non-cognitive outcomes. Fifteen RCTs satisfied the inclusion
criteria of the review. Fourteen trials reported on the effect of
cognitive stimulation on cognitive function. The pooled analysis of
the fourteen trials detected a statistically significant beneficial
effect of cognitive stimulation (Standard Mean Differrence (SMD):
0.41; 95% CI = 0.25, 0.57; test for heterogeneity: x2 = 7.98, p = 0.84,
I2 = 0%). A pooled analysis (n = 4) detected a beneficial effect of
cognitive stimulation on communication and social interaction
(SMD: 0.44; 95% CI = 0.17, 0.71; test for heterogeneity: x2 = 2.24,
p = 0.52, I2 = 0%). Pooled analyses detected that cognitive stimula-
tion was effective in improving self-reported quality of life (SMD:
0.38; 95% CI = 0.11, 0.65; test for heterogeneity: x2 = 2.15, p = 0.54,
I2 = 0%) but not self- (SMD: 0.22; 95% CI =
0.09, 0.53; test for
heterogeneity: x2 = 5.30, p = 0.26, I2 = 25%) or staff-reported mood
(SMD: 0.05; 95% CI =
0.21, 0.31; test for heterogeneity: x2 = 3.16,
p = 0.37, I2 = 5%). Similarly, additional pooled analyses failed to
detect a difference between intervention and control groups with
respect to ADL (SMD = 0.21; 95% CI =
0.05, 0.47; test for
heterogeneity: x2 = 0.22, p = 0.97, I2 = 0%), behaviour (SMD = 0.13;
95% CI =
0.07, 0.32; test for heterogeneity: x2 = 2.61, p = 0.92,
I2 = 0%), or behaviour problems (SMD =
0.14; 95% CI =
0.44, 0.17;
test for heterogeneity: x2 = 1.13, p = 0.57, I2 = 0%). The review
suggested that cognitive stimulation may be associated with a
beneficial effect with respect to cognitive function, communica-
tion/social interaction, and self-reported quality of life (Woods
et al., 2012).
4. Conclusion
4.1. Risk factors associated with AD onset
According to the available evidence, head injury in males,
depression, mild cognitive impairment, age, diabetes mellitus,
conjugated equine estrogen use with medroxyprogesterone
acetate, and exposure to pesticides were all associated with an
increased risk of AD. With respect to genetic factors, APOE e4
remained the strongest predictor of Alzheimer’s disease. The T/T
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
genotype of IL-1b +3953 SNP, the T/T genotype of the IL-1a
889C/
T polymorphism,
1082G/A polymorphism of IL-10 gene, I allele of
the ACE D/I polymorphism, APOE promoter
491A/T and
219T/G
polymorphisms, TT genotype of TNF-a SNP
850, C allele of OLR1
30 -UTR (+1073) polymorphism, rs744373 polymorphism of BIN1,
rs3764650 polymorphism of ABCA7, rs670139 polymorphism of
MS4A4E, and rs9349407 polymorphism of CD2AP were all
associated with an increased risk of AD. Several identified genetic
factors appeared to confer susceptibility to AD in specific
populations. MTHFR C677T polymorphism, SORL1 SNPs
rs2070045, rs3824968, rs2282649, and VLDLR 50 -UTR genotype
2 were associated with an increased risk of AD in Asians, GA/AA
genotype of ALDH2 in East Asian men, PS-1 2/2 genotype in
Europeans, rs3818361 polymorphism of CR1 in Caucasians, SORL1
SNPs 5, 8, 9, 10, and 19 in whites, SORL1 SNPs 19 and 25 in Asians,
and CC + CG genotype of the BACE1 exon 5 (G/C) polymorphism in
APOE e4 non-carriers.
Statin use and several genetic factors, including APOE e2,
rs610932 polymorphism of MS4A6A, and rs3865444 of CD33, were
associated with a reduced risk of AD. Genetic factors that were
associated with a protective effect in specific populations include
PS-1 2/2 genotype in North Americans, rs11136000 polymorphism
of CLU and rs3851179 polymorphism of PICALM in Caucasians,
VLDLR 50 -UTR genotype 2 in non-Asians, SORL1 SNPs 4 and 12 in
whites, SORL1 SNPs 23 and 24 in Asians, and CC + CG genotype of
the BACE1 exon 5 (G/C) polymorphism in APOE e4 carriers.
There was no evidence of an association between AD risk and
metabolic syndrome, conjugated equine estrogen use (without
progestin), antihypertensive medication use, lead, and several
genetic factors including, but not limited to APOE promoter

427 T/C, CTSD polymorphism, and TNF-a SNPs
308,
863,

238,
1031.
With respect to lifestyle factors, current smoking and lower
social engagement were identified as factors associated with an
increased risk of AD. The available evidence was suggestive of a
reduced risk of AD associated with light-to-moderate alcohol
consumption, compliance with a Mediterranean diet, higher
educational attainment, and regular engagement in physically
and cognitively stimulating activities.
While a number of risk factors appeared to be associated with
risk of AD onset, the associations were weak, at best, for a majority
of factors. Conflicting findings across studies or inadequate
evidence precluded definitive conclusions with respect to the
effect of a majority of the examined factors on risk of developing
AD. Certain factors (e.g., mild cognitive impairment and depres-
sion) are suspected prodromes of AD rather than risk factors.
Finally, as we included multiple reviews reporting on the same risk
factor, there was often considerable overlap of included studies
across reviews.
Further research is required to confirm the etiological or
protective role of a number of risk factors.
4.2. Risk factors associated with AD progression
Twelve systematic reviews were identified that reported on
factors associated with cognitive and non-cognitive indicators of
AD progression. Two systematic reviews reported no association
between educational attainment and mortality (Paradise et al.,
2009; Meng and D'Arcy, 2012). However, one review reported an
association between higher educational attainment and faster
cognitive decline (Meng and D'Arcy, 2012). Apolipoprotein E4
(APoE4) did not predict progression of AD (Allan and Ebmeier,
2011). Two reviews failed to detect a beneficial effect of folic acid
supplementation, with or without vitamin B-12, on cognitive
function (Dangour et al., 2010; Malouf et al., 2003; Malouf and
Grimley, 2008). Similarly, findings of a Cochrane systematic review
181
suggest that vitamin B-12 consumed in isolation was not effective
for the improvement of cognitive function in subjects with
dementia, AD, or cognitive impairment (Malouf and Areosa,
2003). Two reviews reported that there was insufficient evidence
to support a beneficial effect of fatty acid supplementation
(Issa et al., 2006; Dangour et al., 2010) and two Cochrane reviews
reported no improvements attributed to vitamin E (Farina et al.,
2012) or lecithin (Higgins and Flicker, 2000). Although physical
activity was not associated with improvements in cognitive
function according to one review (Littbrand et al., 2011), an earlier
systematic review reported a beneficial effect with respect to
cognitive function and other outcomes (Heyn et al., 2004). Finally,
results from a Cochrane review suggested that cognitive stimula-
tion may improve cognitive function, communication/social
interaction, and self-reported quality of life (Woods et al., 2012).
Considered cumulatively, there was evidence to suggest that
physical and cognitive activities may exert a beneficial effect
on cognitive function and other indicators of dementia
progression, while higher educational attainment may contribute
to faster cognitive decline in subjects with dementia. As our
review was restricted to existing systematic reviews, this report
does not provide a comprehensive list of putative factors
associated with AD progression. The identification and synthesis
of primary studies to attain a more comprehensive list of putative
factors is warranted.
Conflict of interest statement
The authors declare that there are no conflicts of interest.
Update
This update identifies systematic reviews reporting on factors
associated with the onset or progression of AD published after the
original search date of our systematic review. As described in
methods section, a supplementary search of the Medline database
was performed to identify recently published systematic reviews
that may be of interest to reader. Study screening, data extraction,
and quality appraisal of each systematic review was conducted by a
single reviewer.
Our supplementary search for systematic reviews on risk
factors for AD onset included reports published between 2012 and
2015. We included moderate or high quality (AMSTAR score >3)
systematic reviews reporting on the association between genetic
and non-genetic risk factors and onset of AD in adults. We excluded
narrative reviews, primary research studies, low quality systematic
reviews (AMSTAR score 3), and systematic reviews which failed
to distinguish AD from all-cause dementia or dementia subtypes.
The search yielded 333 citations, 76 of which were deemed
potentially relevant. Following full-text review, two primary
research studies were excluded, 18 reviews were excluded due
to low quality, 9 reviews were excluded because they failed to
distinguish AD from all-cause dementia or dementia subtype, and
one review did not report on risk of AD onset. Two reviews
identified in the original search were also excluded from the
update. The search for systematic reviews on factors associated
with AD progression, published between 2013 and 2015, produced
241 citations, four of which were deemed potentially relevant. We
included moderate or high quality systematic reviews reporting on
the association between risk factors and the progression of AD in
adults with a previous AD diagnosis. We excluded narrative
reviews, primary research studies, low quality systematic reviews
(AMSTAR score 3). Of the four potentially relevant studies, one
was excluded due to low methodological quality. A list of excluded
systematic reviews on onset and progression, along with reasons
for exclusion, can be found in Supplementary material III.
182 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
A total of 45 systematic reviews satisfied the eligibility criteria
for the update on AD onset. Of these, 23 reviews reported on
genetic factors associated with AD onset and 22 reported on
non-genetic risk factors. Three systematic reviews were included
in the update on AD progression. A list of systematic reviews
identified in the updated search is provided in Supplementary
material III. Findings of the systematic reviews are briefly
summarized below. For more details on non genetic and genetic
risk factors see Table 123 and 124 respectively in Supplementary
material II.
Genetic factors associated with AD onset: A total of 23
systematic reviews/meta-analysis were identified that quantita-
tively assessed the association of several genes (CETP, CYP2D6*4,
CYP2D6, APOE, ESR1, HFE, BDNF, CR1, cathepsin D,Interleukin-1A,
APOC1, ACT, PLAU, UBQLN1, ACE, ABCA1, NEDD9, TF and VEGF) with
the risk of AD onset. Three systematic reviews investigated the
association between different APOE alleles and risk of AD onset. Fei
and Jianhua (2013) observed a positive and significant association
between APOE e4 allele and the risk of progression from MCI to AD.
Liu et al. (2014) reported that APOE e4 and APOE e3 alleles were
positively and negatively associated with AD in Chinese and Indian
population, respectively. However, no significant association of
APOE e2 allele with AD was observed in the Indian population
(Agarwal and Tripathi, 2014). Three systematic reviews assessed
the association between the ESR1 polymorphisms [PvuII polymor-
phism (T/C, rs223493); Xbal (A/G, rs9340799)] and the risk of AD in
different populations. In the meta-analysis by Cheng et al. (2014), a
positive association was noted between the ESR1 Pvu II polymor-
phism in the Caucasian population, whereas, no association was
found between the ESR1 XbaI polymorphisms and risk of AD in any
population. Pan et al. (2014) reported a weak protective role of the
Xbal ESR1 polymorphism in the Asian [but not Caucasian]
population. Similarly, Wang (2014) reported a weak protective
role of the PvuII ESR1 polymorphism on AD onset in the Caucasian
population, but failed to find a significant association between XbaI
ESR1 polymorphism and risk of AD. Lin et al. (2014) investigated
the association between two BDNF gene polymorphisms and risk
of AD in different subgroups and found that while the A allele of
rs6265 BDNF polymorphism was associated with increased the risk
of AD in the Caucasian females and female LOAD patients, the
rs2030324 BDNF polymorphism was not associated with the risk of
AD in any subgroup. Liu et al. (2013) investigated the association
between the two VEGF polymorphisms [-2578C/A polymorphism;

1154G/A polymorphism] and risk of AD based on their APO e4
status. The authors reported that VEGF
2578C/A and VEGF-
1154G/A polymorphisms were associated with increased and
decreased risk of AD when the APOE e4 status was negative,
respectively (Liu et al., 2013).
The following other positive associations with AD were
identified in the overall population or different subgroups: CETP
rs5882A>G; CYP2D6*4A/G; H63D HFE; IL-1A
22889C/T;
rs11568822 APOC1; rs2227564 PLAU; UBQ-8i UBQLN1[LOAD risk];
TNFa G308A [Chinese population]; rs4291 ACE [Risk of SAD in
Caucaisan and in older cases]; TF rs1049296; ACT
15A/T [AD as
well as LOAD in Caucasians]; rs6656401 and rs3818361 CR1 [LOAD
risk] (Chen et al., 2014; Lu et al., 2014; Lin et al., 2014; Luo et al.
2014; Qin et al., 2012; Zhou et al,, 2014; Guan et al., 2012; Wu et al.,
2013; Zhang and Jia, 2014; Wang et al., 2015a,b, 2013a). No
association with AD onset could be demonstrated for the following
polymorphisms in the overall population or in any subgroup: CETP
rs708272 T > C, C282Y HFE, C224T cathepsin D; rs4343 and
rs1800764 ACE; rs2230806, rs4149313 and rs2230808 ABCA 1
(Chen et al., 2014; Lin et al., 2012; Mo et al., 2014; Wang et al.,
2013a,b, 2015a). NEDD9 rs760678 polymorphism was found to be
associated with decreased risk of AD in the Caucasian subgroup but
not the Asian subgroup (Wang et al., 2012).
Other factors associated with AD onset: A total of eight
systematic reviews assessed the association between various
dietary factors and incident AD. Four reviews (Cooper et al., 2015;
Singh et al., 2014; Psaltopoulou et al., 2013; Lourida et al., 2013)
reported statistically significant protective effects of Mediterra-
nean diet on the development of AD, or conversion from mild
cognitive impairment to AD. Other studies reported significant
protective effect of high fish intake, n-3 fatty acids, higher fruit and
vegetable consumption, vitamins C and E (Loef et al., 2012; Li et al.,
2012; Beydoun et al., 2014; Wu et al., 2015). Evidence with respect
to the protective effects of antioxidants, beta-carotene and caffeine
consumption on the risk of AD onset were conflicting (Beydoun
et al., 2014).
A total of six systematic reviews reported associations between
various drugs and the risk of AD. Three reviews (Macedo et al.,
2014; Song et al., 2013; Wong et al., 2013) assessed the potential
protective effect of statin use on the development of AD. Of these,
Macedo et al. (2014) and Wong et al. (2013) reported pooled effect
sizes, both of which were statistically significant. Wang (2015)
reported that use of NSAIDs and ASA were associated with a
significantly lower risk of developing AD in observational studies,
however the only RCT identified found no significant benefit of
celecoxib. Yang et al. (2013) reviewed the effect of raloxifene on the
development of AD. They identified only one study, which reported
a statistically significant benefit, but concluded that there was
insufficient evidence to confirm benefit. One systematic review
(O’Brien et al., 2014) which included one randomized trial and
eight observational studies reported that postmenopausal hor-
monal therapy was not associated with AD.
Two reviews reported associations between co-morbid diseases
and risk of AD. Diniz et al. (2013) and Cheng et al. (2012) reported
that late-life depression and diabetes were significantly associated
with the development of AD, respectively.
Six reviews reported on associations between various serum
biomarkers or nutrient levels and risk of AD. Low Apo E levels
(Wang et al., 2014), high CRP levels (Koyama et al., 2013), high
homocysteine levels (Beydoun et al., 2014) but not elevated IL-6
levels (Koyama et al., 2013) were significantly associated with the
development of AD. Annweiler et al. (2013) reported that vitamin D
levels were lower in patients with AD, potentially implicating this
vitamin in the development of AD. No conclusions could be drawn
on associations between vitamin B12 and development of AD
(O'Leary et al., 2012).
Three additional reviews of various modifiable risk factors were
identified. Zhong et al. (2015) reported that current and ever
smokers demonstrated a significantly elevated risk of developing
AD, while former smokers did not. The risk was most pronounced
in individuals that were APOE e4 non-carriers. Beydoun et al.
(2014) also reviewed the association between smoking and AD,
and found that current or ever smokers were significantly more
likely to develop AD than never smokers. Beydoun et al. (2014) also
reported on the effects of physical activity, alcohol consumption,
and educational attainment. While the evidence for alcohol was
found to be conflicting, physical activity and higher education
[over versus under 8 years] demonstrated statistically significant
protective effect. Vergara et al. (2013) studied the association
between occupational exposure to low frequency magnetic field
exposure and AD. The authors found that such exposure was
associated with a significantly higher risk of developing AD,
although limitations such as heterogeneity across studies and
evidence of publication bias prevented firm conclusions from
being drawn.
Factors associated with AD progression: Three out of four
systematic reviews reporting on AD progression had an AMSTAR
score of >3. These systematic reviews reported that natural
medicine (Gingko), exercise and occupational therapy may have
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187 183

beneficial effects in AD patients (Yang, 2014; Farin, 2014; Mclaren, Bateman, R.J., Aisen, P.S., De Strooper, B., Fox, N.C., Lemere, C.A., Ringman, J.M.,
2014). Salloway, S., Sperling, R.A., Windisch, M., Xiong, C., 2011. Autosomal-dominant
Alzheimer's disease: a review and proposal for the prevention of Alzheimer's
disease. Alzheimer's Res.Ther. 3 (1), 1.
Acknowledgments Bates, M.N., 2006. Mercury amalgam dental fillings: an epidemiological assessment.
Int. J. Hyg. Environ. Health 209 (4), 309–316.
Beri, A., Sural, N., Mahajan, S.B., 2009. Non-atheroprotective effects of statins: a
This project was funded by the Public Health Agency of Canada systematic review. Am. J. Cardiovasc. Drugs 9 (6), 361–370.
in association with Neurological Health Charities Canada through a Bertram, L., McQueen, M.B., Mullin, K., Blacker, D., Tanzi, R.E., 2007. Systematic
contribution agreement administered through the University of meta-analyses of Alzheimer disease genetic association studies: the AlzGene
database. Nat. Genet. 39 (1), 17–23.
Ottawa. We thank Dr. Lynn Beattie for her consultation in preparing Beydoun, M.A., Beydoun, H.A., Wang, Y., 2008. Obesity and central obesity as risk
this manuscript. We also thank Lindsey Sikora for her assistance in factors for incident dementia and its subtypes: a systematic review and meta-
developing the search strategy. Daniel Krewski is the Natural analysis. Obes. Rev. 9 (3), 204–218.
Beydoun, M.A., Beydoun, H.A., Gamaldo, A.A., Teel, A., Zonderman, A.B., Wang, Y.,
Sciences and Engineering Research Council of Canada Chair in Risk
2014. Epidemiologic studies of modifiable factors associated with cognition and
Science at the University of Ottawa. dementia: systematic review and meta-analysis. BMC Public Health 14, 643.
Biessels, G.J., Staekenborg, S., Brunner, E., Brayne, C., Scheltens, P., 2006. Risk of
Appendix A. Supplementary data dementia in diabetes mellitus: a systematic review. Lancet Neurol. 5 (1), 64–74.
Blasko, I., Jellinger, K., Kemmler, G., Krampla, W., Jungwirth, S., Wichart, I., Tragl, K.
H., Fischer, P., 2008. Conversion from cognitive health to mild cognitive
Supplementary data associated with this article can be impairment and Alzheimer's disease: prediction by plasma amyloid beta 42,
found, in the online version, at http://dx.doi.org/10.1016/j. medial temporal lobe atrophy and homocysteine. Neurobiol. Aging 29 (1), 1–11.
Blasko, I., Kemmler, G., Jungwirth, S., Wichart, I., Krampla, W., Weissgram, S.,
neuro.2017.03.006. Jellinger, K., Tragl, K.H., Fischer, P., 2010. Plasma amyloid beta-42 independently
predicts both late-onset depression and Alzheimer’s disease. Am. J. Geriatr.
References Psychiatry 18 (11), 973–982.
Bottiglieri, T., Parnetti, L., Arning, E., Ortiz, T., Amici, S., Lanari, A., Gallai, V., 2001.
Plasma total homocysteine levels and the C677T mutation in the
Abellan van Kan, G., Rolland, Y., Gillette-Guyonnet, S., Gardette, V., Annweiler, C.,
methylenetetrahydrofolate reductase (MTHFR) gene: a study in an Italian
Beauchet, O., Andrieu, S., Vellas, B., 2012. Gait speed, body composition, and
population with dementia. Mech. Ageing Dev. 122 (16), 2013–2023.
dementia. The EPIDOS-Toulouse cohort. J. Gerontol. 67 (4), 425–432.
Boyle, P.A., Buchman, A.S., Barnes, L.L., Bennett, D.A., 2010. Effect of purpose in life
Agarwal, R., Tripathi, C.B., 2014. Association of apolipoprotein E genetic variation in
on risk of incident Alzheimer disease and mild cognitive impairment in
Alzheimer's disease in Indian population: a meta-analysis. Am. J. Alzheimer's
community-dwelling older persons. Arch. Gen. Psychiatry 67 (3), 304–310.
Dis. Other Dementias 29 (7), 575–582.
Breitner, J.C., Baker, L.D., Montine, T.J., Meinert, C.L., Lyketsos, C.G., Ashe, K.H.,
Aglukkaq, L., 2011. Alzheimer's Awareness Month: January 2011. Health Canada.
Brandt, J., et al., 2011. Extended results of the Alzheimer's disease anti-
http://www.hc-sc.gc.ca/ahc-asc/minist/messages/_2011/2011_01_04-eng.php.
inflammatory prevention trial. Alzheimer's Dementia 7 (4), 402–411.
Aguirre, E., Woods, R.T., Spector, A., Orrell, M., 2013. Cognitive stimulation for
Breteler, M.M., van Duijn, C.M., Chandra, V., Fratiglioni, L., Graves, A.B., Heyman, A.,
dementia: a systematic review of the evidence of effectiveness from
Jorm, A.F., et al., 1991. Medical history and the risk of Alzheimer's disease: a
randomised controlled trials. Ageing Res. Rev. 12 (1), 253–262.
collaborative re-analysis of case-control studies. EURODEM Risk Factors
Ahlbom, A., 2001. Neurodegenerative diseases and depressive symptoms in relation
Research Group. Int. J. Epidemiol. 20 (Suppl. 2), S36–S42.
to EMF. Bioelectromagnetics S132–43.
Brinker, K., 2011. Serum Lipids and Statin Therapy in Relation to Alzheimer’s
Ahtiluoto, S., Polvikoski, T., Peltonen, M., Solomon, A., Tuomilehto, J., Winblad, B.,
Disease: Exploring the Relationship Between Elevated Serum Cholesterol Levels
Sulkava, R., Kivipelto, M., 2010. Diabetes, Alzheimer disease, and vascular
and Alzheimer’s Disease Dementia. Diss. Cornell University.
dementia: a population-based neuropathologic study. Neurology 75 (13), 1195–
Bruscoli, M., Lovestone, S., 2004. Is MCI really just early dementia? A systematic
1202.
review of conversion studies. Int. Psychogeriatr. 16 (2), 129–140.
Allan, C.L., Ebmeier, K.P., 2011. The influence of ApoE4 on clinical progression of
Caamano-Isorna, F., Corral, M., Montes-Martinez, A., Takkouche, B., 2006. Education
dementia: a meta-analysis. Int. J. Geriatr. Psychiatry 26 (5), 520–526.
and dementia: a meta-analytic study. Neuroepidemiology 26 (4), 226–232.
Almeida, O.P., Hulse, G.K., Lawrence, D., Flicker, L., 2002. Smoking as a risk factor for
Cataldo, J.K., Prochaska, J.J., Glantz, S.A., 2010. Cigarette smoking is a risk factor for
Alzheimer's disease: contrasting evidence from a systematic review of case-
Alzheimer's Disease: an analysis controlling for tobacco industry affiliation. J.
control and cohort studies. Addiction 97 (1), 15–28.
Alzheimer's Disease 19 (2), 465–480.
Alzheimer Society. Rising Tide: The Impact of Dementia on Canadian Society. 2010.
Chang, C.L., 2008. Obesity and alzheimer's disease. Hong Kong J. Psychiatry 18 (1),
http://www.alzheimer.ca//media/Files/national/Advocacy/ASC_Rising%
28–35.
20Tide Executive%20Summary_Eng.ashx.
Chang-Quan, H., Hui, W., Chao-Min, W., Zheng-Rong, W., Jun-Wen, G., Yong-Hong, L.,
Amieva, H., Stoykova, R., Matharan, F., Helmer, C., Antonucci, T.C., Dartigues, J.F.,
Yan-You, L., Qing-Xiu, L., 2011. The association of antihypertensive medication
2010. What aspects of social network are protective for dementia? Not the
use with risk of cognitive decline and dementia: a meta-analysis of longitudinal
quantity but the quality of social interactions is protective up to 15 years later.
studies. Int. J. Clin. Pract. 65 (12), 1295–1305.
Psychosom. Med. 72 (9), 905–911.
Chen, Y.C., Chen, T.F., Yip, P.K., Hu, C.Y., Chu, Y.M., Chen, J.H., 2010. Body mass index
Annerbo, S., Wahlund, L.O., Lökk, J., 2006. The significance of thyroid-stimulating
(BMI) at an early age and the risk of dementia. Archiv. Gerontol. Geriatrics 50
hormone and homocysteine in the development of Alzheimer’s disease in mild
(Suppl. 1), S48–S52.
cognitive impairment: a 6-year follow-up study. Am. J. Alzheimer’s Dis. Other
Chen, J.J., Li, Y.M., Zou, W.Y., Fu, J.L., 2014. Relationships between CETP genetic
Dementias 21 (3), 182–188.
polymorphisms and Alzheimer's disease risk: a meta-analysis. DNA Cell Biol. 33,
Annweiler, C., Llewellyn, D.J., Beauchet, O., 2013. Low serum vitamin D
807–815.
concentrations in Alzheimer's disease: a systematic review and meta-analysis. J.
Cheng, D., Noble, J., Tang, M.X., Schupf, N., Mayeux, R., Luchsinger, J.A., 2011. Type 2
Alzheimer's Dis. 33, 659–674.
diabetes and late-onset Alzheimer's disease. Dementia Geriatric Cognit. Disord.
Anstey, K.J., von Sanden, C., Salim, A., O’Kearney, R., 2007. Smoking as a risk factor for
31 (6), 424–430.
dementia and cognitive decline: a meta-analysis of prospective studies. Am. J.
Cheng, G., Huang, C., Deng, H., Wang, H., 2012. Diabetes as a risk factor for dementia
Epidemiol. 166 (4), 367–378.
and mild cognitive impairment: a meta-analysis of longitudinal studies. Intern.
Anstey, K.J., Lipnicki, D.M., Low, L.F., 2008. Cholesterol as a risk factor for dementia
Med. J. 42, 484–491.
and cognitive decline: a systematic review of prospective studies with meta-
Cheng, D., Liang, B., Hao, Y., Zhou, W., 2014. Estrogen receptor alpha gene
analysis. Am. J. Geriatr. Psychiatry 16 (5), 343–354.
polymorphisms and risk of Alzheimer's disease: evidence from a meta-analysis.
Anstey, K.J., Mack, H.A., Cherbuin, N., 2009. Alcohol consumption as a risk factor for
Clin. Interv. Aging 9, 1031–1038.
dementia and cognitive decline: meta-analysis of prospective studies. Am. J.
Chu, L.W., Tam, S., Wong, R.L., Yik, P.Y., Song, Y., Cheung, B.M., Morley, J.E., Lam, K.S.,
Geriatr. Psychiatry 17 (7), 542–555.
2010. Bioavailable testosterone predicts a lower risk of Alzheimer's disease in
Anstey, K.J., Cherbuin, N., Budge, M., Young, J., 2011. Body mass index in midlife and
older men. J. Alzheimer's Dis. 21 (4), 1335–1345.
late-life as a risk factor for dementia: a meta-analysis of prospective studies.
Clarke, R., Smith, A.D., Jobst, K.A., Refsum, H., Sutton, L., Ueland, P.M., 1998. Folate,
Obes. Rev. 12 (5), e426–37.
vitamin B12, and serum total homocysteine levels in confirmed Alzheimer
Barnes, D.E., Yaffe, K., 2011. The projected effect of risk factor reduction on
disease. Arch. Neurol. 55 (11), 1449–1455.
Alzheimer's disease prevalence. Lancet Neurol. 10 (9), 819–828.
Cooper, C., Sommerlad, A., Lyketsos, C.G., Livingston, G., 2015. Modifiable predictors
Barnes, D.E., Yaffe, K., Byers, A.L., McCormick, M., Schaefer, C., Whitmer, R.A., et al.,
of dementia in mild cognitive impairment: a systematic review and meta-
2012. Midlife vs. late-life depressive symptoms and risk of dementia:
analysis. Am. J. Psychiatry 172, 323–334.
Differential effects for Alzheimer’s disease and vascular dementia. Arch. Gen.
Crichton, G.E., Bryan, J., Murphy, K.J., Buckley, J., 2010. Review of dairy consumption
Psychiatry 69 (5), 493–498.
and cognitive performance in adults: findings and methodological issues.
Barranco Quintana, J.L., Allam, M.F., Serrano Del Castillo, A., Fernandez-Crehuet
Dementia Geriatric Cognit. Disord. 30 (4), 352–361.
Navajas, R., 2007. Alzheimer's disease and coffee: a quantitative review. Neurol.
Res. 29 (1), 91–95.
184 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Cukierman, T., Gerstein, H.C., Williamson, J.D., 2005. Cognitive decline and dementia
in diabetes—systematic overview of prospective observational studies.
Diabetologia 48 (12), 2460–2469.
Dangour, A.D., Whitehouse, P.J., Rafferty, K., Mitchell, S.A., Smith, L., Hawkesworth,
S., Vellas, B., 2010. B-vitamins and fatty acids in the prevention and treatment of
Alzheimer's disease and dementia: a systematic review. J. Alzheimer's Dis. 22
(1), 205–224.
Daviglus, M.L., Plassman, B.L., Pirzada, A., Bell, C.C., Bowen, P.E., Burke, J.R., Connolly
Jr., E.S., et al., 2011. Risk factors and preventive interventions for Alzheimer
disease: state of the science. Arch. Neurol. 68 (9), 1185–1190.
Debette, S., Markus, H.S., 2010. The clinical importance of white matter
hyperintensities on brain magnetic resonance imaging: systematic review and
meta-analysis. BMJ 341, c3666.
Del Bo, R., Scarlato, M., Ghezzi, S., Martinelli-Boneschi, F., Fenoglio, C., Galimberti, G.,
Galbiati, S., et al., 2006. Is M129 V of PRNP gene associated with Alzheimer's
disease? A case-control study and a meta-analysis. Neurobiol. Aging 27 (5)
(770. e1-770.e5).
Desilets, A.R., Koslowski, D.A., Dunican, K.C., 2010. The role of statins in the
prevention and treatment of Alzheimer’s disease. J. Pharm. Technol. 26, 276–
284.
Devore, E.E., Grodstein, F., van Rooij, F.J., Hofman, A., Stampfer, M.J., Witteman, J.C.,
Breteler, M.M., 2010. Dietary antioxidants and long-term risk of dementia. Arch.
Neurol. 67 (7), 819–825.
Di Bona, D., Plaia, A., Vasto, S., Cavallone, L., Lescai, F., Franceschi, C., Licastro, F.,
Colonna-Romano, G., Lio, D., Candore, G., Caruso, C., 2008. Association between
the interleukin-1b polymorphisms and Alzheimer's disease: a systematic
review and meta-analysis. Brain Res. Rev. 59 (1), 155–163.
Di Bona, D., Candore, G., Franceschi, C., Licastro, F., Colonna-Romano, G., Cammà, C.,
Lio, D., Caruso, C., 2009. Systematic review by meta-analyses on the possible
role of TNF-b polymorphisms in association with Alzheimer’s disease. Brain
Res. Rev. 61 (2), 60–68.
Diniz, B.S., Pinto, J.A. Jr., Gonzaga, M.L., Guimaraes, F.M., Gattaz, W.F., Forlenza, O.V.,
2009. To treat or not to treat? A meta-analysis of the use of cholinesterase
inhibitors in mild cognitive impairment for delaying progression to Alzheimer's
disease. Eur. Archi. Psychiatry Clin. Neurosci. 259 (4), 248–256.
Diniz, B.S., Butters, M.A., Albert, S.M., Dew, M.A., Reynolds 3rd, C.F., 2013. Late-life
depression and risk of vascular dementia and Alzheimer's disease: systematic
review and meta-analysis of community-based cohort studies. Br. J. Psychiatry
202, 329–335.
Dotson, V.M., Beydoun, M.A., Zonderman, A.B., 2010. Recurrent depressive
symptoms and the incidence of dementia and mild cognitive impairment.
Neurology 75 (1), 27–34.
Driver, J.A., Beiser, A., Au, R., Kreger, B.E., Splansky, G.L., Kurth, T., Kiel, D.P., Lu, K.P.,
Seshadri, S., Wolf, P.A., 2012. Inverse association between cancer and
Alzheimer's disease: results from the Framingham Heart Study. BMJ 344, e1442.
Dublin, S., Anderson, M.L., Haneuse, S.J., Heckbert, S.R., Crane, P.K., Breitner, J.C.,
McCormick, W., et al., 2011. Atrial fibrillation and risk of dementia: a prospective
cohort study. J. Am. Geriatr. Soc. 59 (8), 1369–1375.
ESHRE Capri Workshop Groupi, 2011. Perimenopausal risk factors and future health.
Hum. Reprod. Update 17 (5), 706–717.
Elias-Sonnenschein, L.S., Viechtbauer, W., Ramakers, I.H., Verhey, F.R., Visser, P.J.,
2011. Predictive value of APOE-4 allele for progression from MCI to AD-type
dementia: a meta-analysis Journal of Neurology. Neurosurg. Psychiatry 82 (10),
1149–1156.
Elkins, J.S., Douglas, V.C., Johnston, S.C., 2004. Alzheimer disease risk and genetic
variation in ACE: a meta-analysis. Neurology 62 (3), 363–368.
Engelman, M., Agree, E.M., Meoni, L.A., Klag, M.J., 2010. Propositional density and
cognitive function in later life: findings from the Precursors Study. J. Gerontol.
Ser. B-Psychol.Sci. Soc. Sci. 65 (6), 706–711.
Eriksson, U.K., Bennet, A.M., Gatz, M., Dickman, P.W., Pedersen, N.L., 2010a.
Nonstroke cardiovascular disease and risk of Alzheimer disease and dementia.
Alzheimer Dis. Assoc. Disord. 24 (3), 213–219.
Eriksson, U.K., Sjoberg, B.G., Bennet, A.M., de Faire, U., Pedersen, N.L., Frostegard, J.,
2010b. Low levels of antibodies against phosphorylcholine in Alzheimer's
disease. J. Alzheimer's Dis. 21 (2), 577–584.
Etminan, M., Gill, S., Samii, A., 2003. Effect of non-steroidal anti-inflammatory drugs
on risk of Alzheimer's disease: systematic review and meta-analysis of
observational studies. BMJ 327 (7407), 128.
Farina, N., Isaac, M.G., Clark, A.R., Rusted, J., Tabet, N., 2012. Vitamin E for Alzheimer's
dementia and mild cognitive impairment. Cochrane Database Syst. Rev. 11,
CD002854.
Fei, M., Jianhua, W., 2013. Apolipoprotein epsilon4-allele as a significant risk factor
for conversion from mild cognitive impairment to Alzheimer's disease: a meta-
analysis of prospective studies. J. Mol. Neurosci. 50, 257–263.
Ferreira, P.C., Piai Kde, A., Takayanagui, A.M., Segura-Munoz, S.I., 2008. Aluminum as
a risk factor for Alzheimer's disease. Rev. Lat. Am. Enfermagem 16 (1), 151–157.
Fleminger, S., Oliver, D.L., Lovestone, S., Rabe-Hesketh, S., Giora, A., 2003. Head
injury as a risk factor for Alzheimer's disease: the evidence 10 years on; a partial
replication Journal of Neurology. Neurosurg. Psychiatry 74 (7), 857–862.
Folin, M., Baiguera, S., Gallucci, M., Conconi, M.T., Di Liddo, R., Zanardo, A.,
Parnigotto, P.P., 2005. A cross-sectional study of homocysteine-, NO-levels, and
CT-findings in Alzheimer dementia, vascular dementia and controls.
Biogerontology 6 (4), 255–260.
Forti, P., Olivelli, V., Rietti, E., Maltoni, B., Pirazzoli, G., Gatti, R., Gioia, M.G., Ravaglia,
G., 2012. Serum thyroid-stimulating hormone as a predictor of cognitive
impairment in an elderly cohort. Gerontology 58 (1), 41–49.
Fotuhi, M., Mohassel, P., Yaffe, K., 2009. Fish consumption, long-chain omega-3 fatty
acids and risk of cognitive decline or Alzheimer disease: a complex association.
Nat. Clin. Pract. Neurol. 5 (3), 140–152.
Fratiglioni, L., Paillard-Borg, S., Winblad, B., 2004. An active and socially integrated
lifestyle in late life might protect against dementia. Lancet Neurol. 3 (6), 343–
353.
Friedland, R.P., Fritsch, T., Smyth, K.A., Koss, E., Lerner, A.J., Chen, C.H., Petot, G.J.,
Debanne, S.M., 2001. Patients with Alzheimer's disease have reduced activities
in midlife compared with healthy control-group members. Proc. Natl. Acad. Sci.
U. S. A. 98 (6), 3440–3445.
Gao, S., Hendrie, H.C., Hall, K.S., Hui, S., 1998. The relationships between age, sex, and
the incidence of dementia and Alzheimer disease: a meta-analysis. Arch. Gen.
Psychiatry 55 (9), 809–815.
García, A.M., Sisternas, A., Hoyos, S.P., 2008. Occupational exposure to extremely
low frequency elective and magnetic fields and Alzheimer disease: a meta-
analysis. Int. J. Epidemiol. 37 (2), 329–340.
Gates, G.A., Anderson, M.L., McCurry, S.M., Feeney, M.P., Larson, E.B., 2011. Central
auditory dysfunction as a harbinger of Alzheimer dementia. Arch. Otolaryngol.
Head Neck Surg. 137 (4), 390–395.
Gauthier, S., Wirth, Y., Mobius, H.J., 2005. Effects of memantine on behavioural
symptoms in Alzheimer's disease patients: an analysis of the Neuropsychiatric
Inventory (NPI) data of two randomised, controlled studies. Int. J. Geriatr.
Psychiatry 20 (5), 459–464.
Gelber, R.P., Petrovitch, H., Masaki, K.H., Ross, G.W., White, L.R., 2011. Coffee intake in
midlife and risk of dementia and its neuropathologic correlates. J. Alzheimer's
Dis. 23 (4), 607–615.
Gelber, R.P., Petrovitch, H., Masaki, K.H., Abbott, R.D., Ross, G.W., Launer, L.J., White,
L.R., 2012. Lifestyle and the risk of dementia in Japanese-american men. J. Am.
Geriatr. Soc. 60 (1), 118–123.
Gorospe, E.C., Dave, J.K., 2007. The risk of dementia with increased body mass index.
Age Ageing 36 (1), 23–29.
Graves, A.B., van Duijn, C.M., Chandra, V., Fratiglioni, L., Heyman, A., Jorm, A.F.,
Kokmen, E., et al., 1991a. Alcohol and tobacco consumption as risk factors for
Alzheimer's disease: a collaborative re-analysis of case-control studies.
EURODEM Risk Factors Research Group. Int. J. Epidemiol. 20 (Suppl. 2), S48–S57.
Graves, A.B., van Duijn, C.M., Chandra, V., Fratiglioni, L., Heyman, A., Jorm, A.F.,
Kokmen, E., et al., 1991b. Occupational exposures to solvents and lead as risk
factors for Alzheimer's disease: a collaborative re-analysis of case-control
studies. EURODEM Risk Factors Research Group. Int. J. Epidemiol. 20 (Suppl. 2),
S58–S61.
Gray, S.L., Walker, R., Dublin, S., Haneuse, S., Crane, P.K., Breitner, J.C., Bowen, J.,
McCormick, W., Larson, E.B., 2011. Histamine-2 receptor agonist use and
incident dementia in an older. J. Am. Geriatric Soci. 59 (2), 251–257.
Gu, Y., Luchsinger, J.A., Stern, Y., Scarmeas, N., 2010a. Mediterranean diet,
inflammatory and metabolic biomarkers, and risk of Alzheimer's disease. J.
Alzheimer's Dis. 22 (2), 483–492.
Gu, Y., Nieves, J.W., Stern, Y., Luchsinger, J.A., Scarmeas, N., 2010b. Food combination
and Alzheimer’s disease risk: a protective diet. Arch. Neurol. 67 (6), 699–706.
Guan, F., Gu, J., Hu, F., Zhu, Y., Wang, W., 2012. Association between alpha1-
antichymotrypsin signal peptide
15A/T polymorphism and the risk of
Alzheimer's disease: a meta-analysis. Mol. Biol. Rep. 39, 6661–6669.
Guerchet, M., Aboyans, V., Nubukpo, P., Lacroix, P., Clement, J.P., Preux, P.M., 2011.
Ankle-brachial index as a marker of cognitive impairment and dementia in
general population. A systematic review. Atherosclerosis 216 (2), 251–257.
Haan, M.N., Miller, J.W., Aiello, A.E., Whitmer, R.A., Jagust, W.J., Mungas, D.M., Allen,
L.H., Green, R., 2007. Homocysteine, B vitamins, and the incidence of dementia
and cognitive impairment: results from the Sacramento Area Latino Study on
Aging. Am. J. Clin. Nutr. 85 (2), 511–517.
Hamer, M., Chida, Y., 2009. Physical activity and risk of neurodegenerative disease: a
systematic review of prospective evidence. Psychol. Med. 39 (1), 3–11.
Han, X.M., Wang, C.H., Sima, X., Liu, S.Y., 2011. Interleukin-6
174G/C polymorphism
and the risk of Alzheimer's disease in Caucasians: a meta-analysis. Neurosci.
Lett. 504 (1), 4–8.
Hao, P.P., Chen, Y.G., Wang, J.L., Wang, X.L., Zhang, Y., 2011a. Meta-analysis of
aldehyde dehydrogenase 2 gene polymorphism and Alzheimer's disease in East
Asians. Can. J. Neurol. Sci. 38 (3), 500–506.
Hao, Z., Wu, B., Wang, D., Liu, M., 2011b. Association between metabolic syndrome
and cognitive decline: a systematic review of prospective population-based
studies. Acta Neuropsychiatrica 23 (2), 69–74.
Hersi, M., Quach, P., Wang, M.D., Gomes, J., Gaskin, J., Krewski, D., 2016. Systematic
reviews of factors associated with the onset and progression of neurological
conditions in humans: a methodological overview. Neurotoxicology (in this
issue).
Hersi, M., 2015. Lifestyle-related Risk Factors for the Onset of Alzheimer’s Disease: a
Systematic Review of the Literature. Thesis. University of Ottawa.
Heyn, P., Abreu, B.C., Ottenbacher, K.J., 2004. The effects of exercise training on
elderly persons with cognitive impairment and dementia: a meta-analysis.
Arch. Phys. Med. Rehab. 85 (10), 1694–1704.
Higgins, J.P., Flicker, L., 2000. Lecithin for dementia and cognitive impairment.
Cochrane Database Syst. Rev. 4, CD001015.
Ho, R.C., Cheung, M.W., Fu, E., Win, H.H., Zaw, M.H., Ng, A., Mak, A., 2011. Is high
homocysteine level a risk factor for cognitive decline in elderly? A systematic
review, meta-analysis, and meta-regression. Am. J. Geriatric Psychiatry 19 (7),
607–617.
Hogervorst, E., Ribeiro, H.M., Molyneux, A., Budge, M., Smith, A.D., 2002. Plasma
homocysteine levels, cerebrovascular risk factors, and cerebral white matter
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
changes (Leukoaraiosis) in patients with Alzheimer disease. Arch. Neurol. 59
(5), 787–793.
Hooshmand, B., Solomon, A., Kåreholt, I., Leiviskä, J., Rusanen, M., Ahtiluoto, S.,
Winblad, B., Laatikainen, T., Soininen, H., Kivipelto, M., 2010. Homocysteine and
holotranscobalamin and the risk of Alzheimer’s disease: a longitudinal study.
Neurology 75 (16), 1408–1414.
Hua, Y., Zhao, H., Kong, Y., Ye, M., 2011. Association between the MTHFR gene and
Alzheimer's disease: a meta-analysis. Int. J. Neurosci. 121 (8), 462–471.
Hurley, B.F., Hanson, E.D., Sheaff, A.K., 2011. Strength training as a countermeasure
to aging muscle and chronic disease. Sports Med. 41 (4), 289–306.
Isaac, M.G., Quinn, R., Tabet, N., 2008. Vitamin E for Alzheimer's disease and mild
cognitive impairment. Cochrane Database Syst. Rev. 3, CD002854.
Issa, A.M., Mojica, W.A., Morton, S.C., Traina, S., Newberry, S.J., Hilton, L.G., Garland,
R.H., Maclean, C.H., 2006. The efficacy of omega-3 fatty acids on cognitive
function in aging and dementia: a systematic review. Dementia Geriatric
Cognit. Disord. 21 (2), 88–96.
James, B.D., Boyle, P.A., Buchman, A.S., Barnes, L.L., Bennett, D.A., 2011. Life space and
risk of Alzheimer’s disease, mild cognitive impairment, and cognitive decline in
old age. Am. J. Geriatr. Psychiatry 19 (11), 961–969.
Joas, E., Backman, K., Gustafson, D., Ostling, S., Waern, M., Guo, X., Skoog, I., 2012.
Blood pressure trajectories from midlife to late life in relation to dementia in
women followed for 37 years. Hypertension 59 (4), 796–801.
Jorm, A.F., 2001. History of depression as a risk factor for dementia: an updated
review. Aust. New Zealand J. Psychiatry 35 (6), 776–781.
Keage, H.A., Carare, R.O., Friedland, R.P., Ince, P.G., Love, S., Nicoll, J.A., Wharton, S.B.,
Weller, R.O., Brayne, C., 2009. Population studies of sporadic cerebral amyloid
angiopathy and dementia: a systematic review. BMC Neurol. 9, 3.
Kerola, T., Nieminen, T., Hartikainen, S., Sulkava, R., Vuolteenaho, O., Kettunen, R.,
2010. B-type natriuretic peptide as a predictor of declining cognitive function
and dementia
a cohort study of an elderly general population with a 5-year
follow-up. Ann. Med. 42 (3), 207–215.
Kheifets, L., Bowman, J.D., Checkoway, H., Feychting, M., Harrington, J.M., Kavet, R.,
Marsh, G., Mezei, G., Renew, D.C., van Wijngaarden, E., 2009. Future needs of
occupational epidemiology of extremely low frequency electric and magnetic
fields: review and recommendations. Occup. Environ. Med. 66 (2), 72–80.
Khoury, M.J., Bertram, L., Boffetta, P., Butterworth, A.S., Chanock, S.J., Dolan, S.M.,
Fortier, I., Carcia-Closas, M., Gwinn, M., et al., 2009. Genome-wide association
studies, field synopses, and the development of the knowledge base on genetic
variation and human disease. Am. J. Epidemiol. 170 (3), 269–279.
Kim, J.M., Stewart, R., Kim, S.W., Shin, I.S., Yang, S.J., Shin, H.Y., Yoon, J.S., 2008.
Changes in folate, vitamin B12 and homocysteine associated with incident
dementia. J. Neurol. Neurosurg. Psychiatry 79 (8), 864–868.
Kimm, H., Lee, P.H., Shin, Y.J., Park, K.S., Jo, J., Lee, Y., Kang, H.C., Jee, S.H., 2011. Midlife
and late-life vascular risk factors and dementia in Korean men and women.
Arch. Gerontol. Geriatr. 52 (3), e117–e122.
Kloppenborg, R.P., van den Berg, E., Kappelle, L.J., Biessels, G.J., 2008. Diabetes and
other vascular risk factors for dementia: which factor matters most? A
systematic review. Eur. J. Pharmacol. 585 (1), 97–108.
Kohler, S., van Boxtel, M., Jolles, J., Verhey, F., 2011. Depressive symptoms and risk for
dementia: a 9-year follow-up of the Maastricht Aging Study. Am. J. Geriatr.
Psychiatry 19 (10), 902–905.
Koseoglu, E., Karaman, Y., 2007. Relations between homocysteine, folate and
vitamin B12 in vascular dementia and in Alzheimer disease. Clin. Biochem. 40
(12), 859–863.
Koyama, A., O'Brien, J., Weuve, J., Blacker, D., Metti, A.L., Yaffe, K., 2013. The role of
peripheral inflammatory markers in dementia and Alzheimer's disease: a meta-
analysis. J. Gerontol. Ser. A-Biol. Sci. Med. Sci. 68, 433–440.
Kristal, A.R., Peters, U., Potter, J.D., 2005. Is it time to abandon the food frequency
questionnaire? Cancer Epidemiol. Biomarkers Prevention 14 (12), 2826–2828.
Kuo, H.K., Sorond, F.A., Chen, J.H., Hashmi, A., Milberg, W.P., Lipsitz, L.A., 2005a. The
role of homocysteine in multisystem age-related problems: a systematic
review. J. Gerontol. Ser. A-Biol. Sci. Med. Sci. 60 (9), 1190–1201.
Kuo, H.K., Yen, C.J., Chang, C.H., Kuo, C.K., Chen, J.H., Sorond, F., 2005b. Relation of C-
reactive protein to stroke, cognitive disorders, and depression in the general
population: systematic review and meta-analysis. Lancet Neurol. 4 (6), 371–
380.
LaRoia, H., Louis, E.D., 2011. Association between essential tremor and other
neurodegenerative diseases: what is the epidemiological evidence?
Neuroepidemiology 37 (1), 1–10.
LeBlanc, E.S., Janowsky, J., Chan, B.K., Nelson, H.D., 2001. Hormone replacement
therapy and cognition: systematic review and meta-analysis. JAMA 285 (11),
1489–1499.
Lee, Y., Back, J.H., Kim, J., Kim, S.H., Na, D.L., Cheong, H.K., Hong, C.H., Kim, Y.G., 2010.
Systematic review of health behavioral risks and cognitive health in older
adults. Int. Psychogeriatr. 22 (2), 174–187.
Leinonen, V., Koivisto, A.M., Savolainen, S., Rummukainen, J., Tamminen, J.N.,
Tillgren, T., Vainikka, S., et al., 2010. Amyloid and (proteins in cortical brain
biopsy and Alzheimer's disease. Ann. Neurol. 68 (4), 446–453.
Lenoir, H., Dufouil, C., Auriacombe, S., Lacombe, J.M., Dartigues, J.F., Ritchie, K.,
Tzourio, C., 2011. Depression history, depressive symptoms, and incident
dementia: the 3C Study. J. Alzheimer's Dis. 26 (1), 27–38.
Li, G., Shofer, J.B., Rhew, I.C., Kukull, W.A., Peskind, E.R., McCormick, W., Bowen, J.D.,
et al., 2010a. Age-varying association between statin use and incident
Alzheimer's disease. J. Am. Geriatr. Soc. 58 (7), 1311–1317.
185
Li, N.C., Lee, A., Whitmer, R.A., Kivipelto, M., Lawler, E., Kazis, L.E., Wolozin, B., 2010b.
Use of angiotensin receptor blockers and risk of dementia in a predominantly
male population: prospective cohort analysis. BMJ 340, b5465.
Li, G., Wang, L.Y., Shofer, J.B., Thompson, M.L., Peskind, E.R., McCormick, W., Bowen, J.
D., Crane, P.K., Larson, E.B., 2011. Temporal relationship between depression and
dementia: findings from a large community-based 15-year follow-up study.
Arch. Gen. Psychiatry 68 (9), 970–977.
Li, F.J., Shen, L., Ji, H.F., 2012. Dietary intakes of vitamin E, vitamin C, and beta-
carotene and risk of Alzheimer's disease: a meta-analysis. J. Alzheimer's Dis. 31,
253–258.
Lim, W.S., Gammack, J.K., Van Niekerk, J.K., Dangour, A., 2006. Omega 3 fatty acid for
the prevention of dementia. Cochrane Database Syst. Rev. 1, CD005379.
Lin, F.R., Metter, E.J., O'Brien, R.J., Resnick, S.M., Zonderman, A.B., Ferrucci, L., 2011.
Hearing loss and incident dementia. Arch. Neurol. 68 (2), 214–220.
Lin, M., Zhao, L., Fan, J., Lian, X.G., Ye, J.X., Wu, L., et al., 2012. Association between
HFE polymorphisms and susceptibility to Alzheimer's disease: a meta-analysis
of 22 studies including 4,365 cases and 8,652 controls. Mol. Biol. Rep. 39, 3089–
3095.
Lin, Y., Cheng, S., Xie, Z., Zhang, D., 2014. Association of rs6265 and rs2030324
polymorphisms in brain-derived neurotrophic factor gene with Alzheimer's
disease: a meta-analysis. PLoS One [Electron. Resour.] 9, e94961.
Lindsay, J., Anderson, L., 2003. Dementia and Alzheimer's Disease. Women's Health
Surveillance Report. Public Health Agency of Canada. http://www.phac-aspc.gc.
ca/publicat/whsr-rssf/chap_19-eng.php.
Lipton, R.B., Hirsch, J., Katz, M.J., Wang, C., Sanders, A.E., Verghese, J., Barzilai, N.,
Derby, C.A., 2010. Exceptional parental longevity associated with lower risk of
Alzheimer's disease and memory decline. J. Am. Geriatr. Soc. 58 (6), 1043–1049.
Littbrand, H., Stenvall, M., Rosendahl, E., 2011. Applicability and effects of physical
exercise on physical and cognitive functions and activities of daily living among
people with dementia: a systematic review. Am. J. Phys. Med. Rehab. 90 (6),
495–518.
Liu, Y., Zhang, J., Zeng, Z., 2013. Vascular endothelial growth factor polymorphisms
and risk of Alzheimer's disease: a meta-analysis. Gene 518, 296–302.
Liu, M., Bian, C., Zhang, J., Wen, F., 2014. Apolipoprotein E gene polymorphism and
Alzheimer's disease in Chinese population: a meta-analysis. Sci. Rep. 4, 4383.
Llorca, J., Rodriguez-Rodriguez, E., Dierssen-Sotos, T., Delgado-Rodriguez, M.,
Berciano, J., Combarros, O., 2008. Meta-analysis of genetic variability in the
beta-amyloid production, aggregation and degradation metabolic pathways
and the risk of Alzheimer's disease. Acta Neurol. Scand. 117 (1), 1–14.
Loef, M., Walach, H., 2012a. Copper and iron in Alzheimer's disease: a systematic
review and its dietary implications. Br. J. Nutr. 107 (1), 7–19.
Loef, M., Walach, H., 2012b. Fruit, vegetables and prevention of cognitive decline or
dementia: a systematic review of cohort studies. J. Nutr. Health Aging 16 (7),
626–630.
Loef, M., Mendoza, L.F., Walach, H., 2011a. Lead (Pb) and the risk of Alzheimer’s
disease or cognitive decline: a systematic review. Tox. Rev. 30 (4), 103–114.
Loef, M., Schrauzer, G.N., Walach, H., 2011b. Selenium and Alzheimer's disease: a
systematic review. J. Alzheimer's Dis. 26 (1), 81–104.
Loef, M., von Stillfried, N., Walach, H., 2012. Zinc diet and Alzheimer's disease: a
systematic review. Nutr. Neurosci. 15 (5), 2–12.
Lopez, L.B., Kritz-Silverstein, D., Barrett Connor, E., 2011. High dietary and plasma
levels of the omega-3 fatty acid docosahexaenoic acid are associated with
decreased dementia risk: the Rancho Bernardo study. J. Nutr. Health Aging 15
(1), 25–31.
Louis, E.D., Tang, M.X., Schupf, N., 2010. Mild parkinsonian signs are associated with
increased risk of dementia in a prospective, population-based study of elders.
Mov. Disord. 25 (2), 172–178.
Lourida, I., Soni, M., Thompson-Coon, J., Purandare, N., Lang, I.A., Ukoumunne, O.C.,
et al., 2013. Mediterranean diet, cognitive function, and dementia: a systematic
review. Epidemiology 24, 479–489.
Lu, F.P., Lin, K.P., Kuo, H.K., 2009. Diabetes and the risk of multi-system aging
phenotypes: a systematic review and meta-analysis. PLoS ONE [Electron.
Resour.] 4 (1), e4144.
Lu, Y., et al., 2014. Quantitative assessment of CYP2D6 polymorphisms and risk of
Alzheimer's disease: a meta-analysis. J. Neurol. Sci. 343, 15–22.
Luchsinger, J.A., Tang, M.X., Shea, S., Miller, J., Green, R., Mayeux, R., 2004. Plasma
homocysteine levels and risk of Alzheimer disease. Neurology 62 (11), 1972–
1976.
Luo, J., Li, S., Qin, X., Song, L., Peng, Q., Chen, S., et al., 2014. Meta-analysis of the
association between CR1 polymorphisms and risk of late-onset Alzheimer's
disease. Neurosci. Lett. 578, 165–170.
Macedo, A.F., Taylor, F.C., Casas, J.P., Adler, A., Prieto-Merino, D., Ebrahim, S., 2014.
Unintended effects of statins from observational studies in the general
population: systematic review and meta-analysis. BMC Med. 12, 51.
Malouf, R., Areosa Sastre, A., 2003. Vitamin B12 for cognition. Cochrane Database
Syst. Rev. 3, CD004326.
Malouf, R., Grimley, E.J., 2008. Folic acid with or without vitamin B12 for the
prevention and treatment of healthy elderly and demented people. Cochrane
Database Syst. Rev. 4, CD004514.
Malouf, M., Grimley, E.J., Areosa, S.A., 2003. Folic acid with or without vitamin B12
for cognition and dementia. Cochrane Database Syst. Rev. 4, CD004514.
Marengoni, A., Qiu, C., Winblad, B., Fratiglioni, L., 2011. Atrial fibrillation, stroke and
dementia in the very old: a population-based study. Neurobiol. Aging 32 (7),
1336–1337.
186 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
McGuinness, B., Todd, S., Passmore, P., Bullock, R., 2009. Blood pressure lowering in
patients without prior cerebrovascular disease for prevention of cognitive
impairment and dementia. Cochrane Database Syst. Rev. 4, CD004034.
Meng, X., D'Arcy, C., 2012. Education and dementia in the context of the cognitive
reserve hypothesis: a systematic review with meta-analyses and qualitative
analyses. PLoS ONE [Electron. Resour.] 7 (6), e38268.
Mielke, M.M., Zandi, P.P., Shao, H., Waern, M., Ostling, S., Guo, X., Bjorkelund, C.,
Lissner, L., Skoog, I., Gustafson, D.R., 2010. The 32-year relationship between
cholesterol and dementia from midlife to late life. Neurology 75 (21), 1888–
1895.
Miller, L.J., Chacko, R., 2004. The role of cholesterol and statins in Alzheimer's
disease. Ann. Pharmacother. 38 (1), 91–98.
Miller, J.W., Green, R., Mungas, D.M., Reed, B.R., Jagust, W.J., 2002. Homocysteine,
vitamin B6, and vascular disease in AD patients. Neurology 58 (10), 1471–1475.
Mitchell, A.J., Shiri-Feshki, M., 2008. Temporal trends in the long term risk of
progression of mild cognitive impairment: a pooled analysis Journal of
Neurology. Neurosurg. Psychiatry 79 (12), 1386–1391.
Mitchell, A.J., Shiri-Feshki, M., 2009. Rate of progression of mild cognitive
impairment to dementia–meta-analysis of 41 robust inception cohort studies.
Acta Psychiatr. Scand. 119 (4), 252–265.
Mo, C., Peng, Q., Sui, J., Wang, J., Deng, Y., Xie, L., et al., 2014. Lack of association
between cathepsin D C224T polymorphism and Alzheimer's disease risk: an
update meta-analysis. BMC Neurol. 14, 13.
Montlahuc, C., Soumare, A., Dufouil, C., Berr, C., Dartigues, J.F., Poncet, M., Tzourio, C.,
Alperovitch, A., 2011. Self-rated health and risk of incident dementia: a
community-based elderly cohort, the 3C study. Neurology 77 (15), 1457–1464.
Mortimer, J.A., van Duijn, C.M., Chandra, V., Fratiglioni, L., Graves, A.B., Heyman, A.,
Jorm, A.F., et al., 1991. Head trauma as a risk factor for Alzheimer's disease: a
collaborative re-analysis of case-control studies. EURODEM Risk Factors
Research Group. Int. J. Epidemiol. 20 (Suppl.2), S28–S35.
Muller, M., Schupf, N., Manly, J.J., Mayeux, R., Luchsinger, J.A., 2010. Sex hormone
binding globulin and incident Alzheimer's disease in elderly men and women.
Neurobiol. Aging 31 (10), 1758–1765.
Nägga, K., Rajani, R., Mårdh, E., Borch, K., Mårdh, S., Marcusson, J., 2003. Cobalamin,
folate, methylmalonic acid, homocysteine, and gastritis markers in dementia.
Dement. Geriatr. Cogn. Disord. 16 (4), 269–275.
National Advisory Council on Aging (Canada), 2004. The NACA Position on
Alzheimer Disease and Related Dementias. . http://books2. scholarsportal.info/
viewdoc.html?id=/ebooks/ebooks1/gibson_chrc/2010-08-06/1/10089389.
Ning, M., Yang, Y., Zhang, Z., Chen, Z., Zhao, T., Zhang, D., Zhou, D., Xu, J., Liu, Z., Wang,
Y., Liu, Y., Zhao, X., Li, W., Li, S., He, L., 2010. Amyloid-b-Related genes SORL1 and
ACE are genetically associated with risk for late-onset alzheimer disease in the
chinese population. Alzheimer Dis. Relat. Disord. 24 (4), 390–396.
Norton, M.C., Dew, J., Smith, H., Fauth, E., Piercy, K.W., Breitner, J.C., Tschanz, J.,
Wengreen, H., Welsh-Bohmer, K., Cache County Investigators, 2012. Lifestyle
behavior pattern is associated with different levels of risk for incident dementia
and Alzheimer's disease: the Cache County study. J. Am. Geriatr. Soc. 60 (3),
405–412.
Ntais, C., Polycarpou, A., Ioannidis, J.P., 2004. Meta-analysis of the association of the
cathepsin D Ala224Val gene polymorphism with the risk of Alzheimer's disease:
a HuGE gene-disease association review. Am. J. Epidemiol. 159 (6), 527–536.
O'Brien, J., Jackson, J.W., Grodstein, F., Blacker, D., Weuve, J., 2014. Postmenopausal
hormone therapy is not associated with risk of all-cause dementia and
Alzheimer's disease. Epidemiol. Rev. 36, 83–103.
O'Leary, F., Allman-Farinelli, M., Samman, S., 2012. Vitamin B12 status: cognitive
decline and dementia: a systematic review of prospective cohort studies. Br. J.
Nutr. 108, 1948–1961.
Ohara, T., Doi, Y., Ninomiya, T., Hirakawa, Y., Hata, J., Iwaki, T., Kanba, S., Kiyohara, Y.,
2011. Glucose tolerance status and risk of dementia in the community: the
Hisayama study. Neurology 77 (12), 1126–1134.
Ownby, R.L., Crocco, E., Acevedo, A., John, V., Loewenstein, D., 2006. Depression and
risk for Alzheimer disease: systematic review, meta-analysis, and
metaregression analysis. Arch. Gen. Psychiatry 63 (5), 530–538.
Pan, Y., Li, Y., Shen, H., 2014. Meta-analysis of the association between
polymorphisms of estrogen receptor alpha genes rs9340799 and rs2234693 and
Alzheimer's disease: evidence from 23 articles. Am. J. Alzheimer's Dis. Other
Dementias 29, 704–711.
Paradise, M., Cooper, C., Livingston, G., 2009. Systematic review of the effect of
education on survival in Alzheimer's disease. Int. Psychogeriatr. 21 (1), 25–32.
Parihar, M.S., Hemnani, T., 2004. Alzheimer's disease pathogenesis and therapeutic
interventions. J. Clin. Neurosci. 11 (5), 456–467.
Patatanian, E., Gales, M.A., 2005. The future of statins: alzheimer's disease? Consult.
Pharm. 20 (8), 663–673.
Patterson, C., Feightner, J., Garcia, A., MacKnight, C., 2007. General risk factors for
dementia: a systematic evidence review. Alzheimer's Dementia 3 (4), 341–347.
Penrose, F.K., 2005. Can exercise affect cognitive functioning in Alzheimer ‘s
disease? A review of the literature. Activities Adaptation Aging 29 (4), 15–40.
Peters, R., Burch, L., Warner, J., Beckett, N., Poulter, R., Bulpitt, C., 2008a.
Haemoglobin, anaemia, dementia and cognitive decline in the elderly, a
systematic review. BMC Geriatr. 8, 18.
Peters, R., Peters, J., Warner, J., Beckett, N., Bulpitt, C., 2008b. Alcohol, dementia and
cognitive decline in the elderly: a systematic review. Age Ageing 37 (5), 505–
512.
Peters, R., Poulter, R., Warner, J., Beckett, N., Burch, L., Bulpitt, C., 2008c. Smoking,
dementia, and cognitive decline in the elderly, a systematic review. BMC Geriatr.
8, 36.
Pham, D.Q., Plakogiannis, R., 2005. Vitamin E supplementation in Alzheimer's
disease, Parkinson's disease, tardive dyskinesia, and cataract: part 2. Ann.
Pharmacother. 39 (12), 2065–2072.
Pitner, J.K., 2005. Obesity in the elderly. Consult. Pharm. 20 (6), 498–513.
Podewils, L.J., 2003. Physical Activity and Dementia Risk: a Systematic Review.
(Diss., Johns Hopkins University).
Power, M.C., Weuve, J., Gagne, J.J., McQueen, M.B., Viswanathan, A., Blacker, D., 2011.
The association between blood pressure and incident alzheimer disease: a
systematic review and meta-analysis. Epidemiology 22 (5), 646–659.
Profenno, L.A., Porsteinsson, A.P., Faraone, S.V., 2010. Meta-analysis of Alzheimer's
disease risk with obesity, diabetes, and related disorders. Biol. Psychiatry 67 (6),
505–512.
Psaltopoulou, T., Sergentanis, T.N., Panagiotakos, D.B., Sergentanis, I.N., Kosti, R.,
Scarmeas, N., 2013. Mediterranean diet, stroke, cognitive impairment, and
depression: a meta-analysis. Ann. Neurol. 74, 580–591.
Public Health Agency of Canada, 2010. Chapter 3: The Health and Well-being of
Canadian Seniors. The Chief Public Health Officer's Report on The State of Public
Health in Canada 2010. . Accessed 28 Oct. 2013 http://www.phac-aspc.gc.ca/
cphorsphc-respcacsp/2010/fr-rc/cphorsphc-respcacsp-06-eng.php.
Purnell, C., Callahan, C.M., Hendrie, H.C., 2009. Cardiovascular risk factors and
incident alzheimer’s disease: a systematic review of the literature. Alzheimer
Dis. Assoc. Disord. 23 (1), 1–10.
Qin, X., Peng, Q., Zeng, Z., Chen, Z., Lin, L., Deng, Y., et al., 2012. Interleukin-1A
889C/
T polymorphism and risk of Alzheimer's disease: a meta-analysis based on 32
case-control studies. J. Neurol. 259, 1519–1529.
Qiu, C., Xu, W., Winblad, B., Fratiglioni, L., 2010. Vascular risk profiles for dementia
and Alzheimer's disease in very old people: a population-based longitudinal
study. J. Alzheimer's Dis. 20 (1), 293–300.
Quadri, P., Fragiacomo, C., Pezzati, R., Zanda, E., Forloni, G., Tettamanti, M., Lucca, U.,
2004. Homocysteine, folate, and vitamin B-12 in mild cognitive impairment,
Alzheimer disease, and vascular dementia. Am. J. Clin. Nutr. 80 (1), 114–122.
Rainero, I., Bo, M., Ferrero, M., Valfrè, W., Vaula, G., Pinessi, L., 2004. Association
between the interleukin-1a gene and Alzheimer’s disease: a meta-analysis.
Neurobiol. Aging 25 (10), 1293–1298.
Raman, G., Tatsioni, A., Chung, M., Rosenberg, I.H., Lau, J., Lichtenstein, A.H., Balk, E.
M., 2007. Heterogeneity and lack of good quality studies limit association
between folate, vitamins B-6 ad B-12, and cognitive function. J. Nutr. 137 (7),
1789–1794.
Raschetti, R., Albanese, E., Vanacore, N., Maggini, M., 2007. Cholinesterase inhibitors
in mild cognitive impairment: a systematic review of randomised trials. PLoS
Med. Public Lib. Sci. 4 (11), e338.
Ravaglia, G., Forti, P., Maioli, F., Martelli, M., Servadei, L., Brunetti, N., Porcellini, E.,
Licastro, F., 2005. Homocysteine and folate as risk factors for dementia and
Alzheimer disease. Am. J. Clin. Nutr. 82 (3), 636–643.
Reitz, C., Tang, M.X., Schupf, N., Manly, J.J., Mayeux, R., Luchsinger, J.A., 2010a.
Association of higher levels of high-density lipoprotein cholesterol in elderly
individuals and lower risk of late-onset Alzheimer disease. Arch. Neurol. 67 (12),
1491–1497.
Reitz, C., Tang, M.X., Schupf, N., Manly, J.J., Mayeux, R., Luchsinger, J.A., 2010b. b: a
summary risk score for the prediction of Alzheimer disease in elderly persons.
Arch. Neurol. 67 (7), 835–841.
Reitz, C., Cheng, R., Rogaeva, E., Lee, J.H., Tokuhiro, S., Zou, F., Bettens, K., et al., 2011.
Meta-analysis of the association between variants in SORL1 and Alzheimer
disease. Arch. Neurol. 68 (1), 99–106.
Religa, D., Styczynska, M., Peplonska, B., Gabryelewicz, T., Pfeffer, A., Chodakowska,
M., Luczywek, E., Wasiak, B., Stepien, K., Golebiowski, M., Winblad, B.,
Barcikowska, M., et al., 2003. Homocysteine, apolopoproteine E and
methylenetetrahydrofolate reductase in Alzheimer’s disease and mild cognitive
impairment. Dement. Geriatr. Cogn. Disord. 16 (2), 4–70.
Reutens, S., Sachdev, P., 2002. Homocysteine in neuropsychiatric disorders of the
elderly. Int. J. Geriatr. Psychiatry 17 (9), 859–864.
Rocca, W.A., van Duijn, C.M., Clayton, D., Chandra, V., Fratiglioni, L., Graves, A.B.,
Heyman, A., et al., 1991. Maternal age and Alzheimer's disease: a collaborative
re-analysis of case-control studies: EURODEM Risk Factors Research Group. Int.
J. Epidemiol. 20 (Suppl. 2), S21–7.
Rodriguez-Manotas, M., Amorin-Diaz, M., Canizares-Hernandez, F., Ruiz-Espejo, F.,
Martinez-Vidal, S., Gonzalez-Sarmiento, R., Martinez-Hernandez, P., Cabezas-
Herrera, J., 2007. Association study and meta-analysis of Alzheimer's disease
risk and presenilin-1 intronic polymorphism. Brain Res. 1170, 119–128.
Roe, C.M., Fitzpatrick, A.L., Xiong, C., Sieh, W., Kuller, L., Miller, J.P., Williams, M.M.,
Kopan, R., Behrens, M.I., Morris, J.C., 2010. Cancer linked to Alzheimer disease
but not vascular dementia. Neurology 74 (2), 106–112.
Rolland, Y., Abellan van Kan, G., Vellas, B., 2008. Physical activity and Alzheimer's
disease: from prevention to therapeutic perspectives. J. Am. Med. Dir. Assoc. 9
(6), 390–405.
Roman, B., Carta, L., Martinez-Gonzalez, M.A., Serra-Majem, L., 2008. Effectiveness
of the Mediterranean diet in the elderly. Clin. Interv. Aging 3 (1), 97–109.
Ronnemaa, E., Zethelius, B., Lannfelt, L., Kilander, L., 2011. Vascular risk factors and
dementia: 40-year follow-up of a population-based cohort. Dementia Geriatric
Cognit. Disord. 31 (6), 460–466.
Rusanen, M., Kivipelto, M., Quesenberry, C.P.Jr., Zhou, J., Whitmer, R.A., 2011. Heavy
smoking in midlife and long-term risk of Alzheimer disease and vascular
dementia. Arch. Intern. Med. 171 (4), 333–339.
Saczynski, J.S., Beiser, A., Seshadri, S., Auerbach, S., Wolf, P.A., Au, R., 2010. Depressive
symptoms and risk of dementia: the Framingham Heart Study. Neurology 75
(1), 35–41.
 M. Hersi et al. / NeuroToxicology 61 (2017) 143–187
Santibanez, M., Bolumar, F., Garcia, A.M., 2007. Occupational risk factors in
Alzheimer's disease: a review assessing the quality of published
epidemiological studies. Occup. Environ. Med. 64 (11), 723–732.
Santos, C., Costa, J., Santos, J., Vaz-Carneiro, A., Lunet, N., 2010. Caffeine intake and
dementia: systematic review and meta-analysis. J. Alzheimer's Dis. 20 (Suppl 1),
S187–204.
Schrijvers, E.M., Witteman, J.C., Sijbrands, E.J., Hofman, A., Koudstaal, P.J., Breteler,
M.M., 2010. Insulin metabolism and the risk of Alzheimer disease: the
Rotterdam Study. Neurology 75 (22), 1982–1987.
Schrijvers, E.M., Koudstaal, P.J., Hofman, A., Breteler, M.M., et al., 2011. Plasma
clusterin and the risk of Alzheimer disease. J. Am. Med. Assoc. 305 (13), 1322–
1326.
Seshadri, S., Beiser, A., Selhub, J., Jacques, P.F., Rosenberg, I.H., D'Agostino, R.B.,
Wilson, P.W., Wolf, P.A., 2002. Plasma homocysteine as a risk factor for dementia
and Alzheimer's disease. New Engl. J. Med. 346 (7), 476–483.
Shadlen, M.F., Larson, E.B., Yukawa, M., 2000. The epidemiology of Alzheimer's
disease and vascular dementia in Japanese and African-American populations:
the search for etiological clues. Neurobiol. Aging 21 (2), 171–181.
Shah, K., Qureshi, S.U., Johnson, M., Parikh, N., Schulz, P.E., Kunik, M.E., 2009. Does
use of antihypertensive drugs affect the incidence or progression of dementia?
A systematic review. Am. J. Geriatr. Pharmacother. 7 (5), 250–261.
Shah, R.C., Buchman, A.S., Wilson, R.S., Leurgans, S.E., Bennett, D.A., 2011.
Hemoglobin level in older persons and incident Alzheimer disease: prospective
cohort analysis. Neurology 77 (3), 219–226.
Shah, N.S., Vidal, J.S., Masaki, K., Petrovitch, H., Ross, G.W., Tilley, C., DeMattos, R.B.,
Tracy, R.P., White, L.R., Launer, L.J., 2012. Midlife blood pressure, plasma
b-Amyloid and the risk for alzheimer’s disease: the Honolulu asia aging study.
Hypertension 59 (4), 780–786.
Shea, B.J., Hamel, C., Wells, G.A., Bouter, L.M., Kristjansson, E., Grimshaw, J., Henry, D.
A., Boers, M., 2009. AMSTAR is a reliable and valid measurement tool to assess
the methodological quality of systematic reviews. J. Clin. Epidemiol. 62, 1013–
1020.
Singh, B., Parsaik, A.K., Mielke, M.M., Erwin, P.J., Knopman, D.S., Petersen, R.C., et al.,
2014. Association of mediterranean diet with mild cognitive impairment and
Alzheimer's disease: a systematic review ad meta-analysis. J. Alzheimer's Dis.
39, 271–282.
Sofi, F., Cesari, F., Abbate, R., Gensini, G.F., Casini, A., 2008. Adherence to
Mediterranean diet and health status: meta-analysis. BMJ 337, a1344.
Sofi, F., Abbate, R., Gensini, G.F., Casini, A., 2010. Accruing evidence on benefits of
adherence to the Mediterranean diet on health: an updated systematic review
and meta-analysis. Am. J. Clin. Nutr. 92 (5), 1189–1196.
Song, X., Mitnitski, A., Rockwood, K., 2011. Nontraditional risk factors predict
Alzheimer disease. Neurology 77 (3), 227–234.
Song, Y., Nie, H., Xu, Y., Zhang, L., Wu, Y., 2013. Association of statin use with risk of
dementia: a meta-analysis of prospective cohort studies. Geriatrics Gerontol.
Int. 13, 817–824.
Standridge, J.B., 2004. Pharmacotherapeutic approaches to the prevention of
Alzheimer's disease. Am. J. Geriatr. Pharmacother. 2 (2), 119–132.
Stern, C., Konno, R., 2009. Physical leisure activities and their role in preventing
dementia: a systematic review. Int. J. Evide. Based Healthcare 7 (4), 270–282.
Stern, C., Munn, Z., 2010. Cognitive leisure activities and their role in preventing
dementia: a systematic review. Inte. J. Evid. Based Healthcare 8 (1), 2–17.
Stern, Y., 2006. Cognitive reserve and Alzheimer disease. Alzheimer Dis. Assoc.
Disord. 20 (3 (Sulppl. 2)), S69–S74.
Storey, S.G., Suryadevara, V., Aronow, W.S., Ahn, C., 2003. Association of plasma
homocysteine in elderly persons with atherosclerotic vascular disease and
dementia, atherosclerotic vascular disease without dementia, dementia
without atherosclerotic vascular disease, and no dementia or atherosclerotic
vascular disease. J. Gerontol. 58 (12), 1135–1136.
Sundelöf, J., Sundström, J., Hansson, O., Eriksdotter-Jönhagen, M., Giedraitis, V.,
Larsson, A., Degerman-Gunnarsson, M., Ingelsson, M., Minthon, L., Blennow, K.,
Kilander, L., Basun, H., Lannfelt, L., 2012. Cystatin C levels are positively
correlated with Ab42 and t levels in cerebrospinal fluid in persons with
Alzheimer’s disease, mild cognitive impairment, and healthy controls. J.
Alzheimer’s Dis. 21 (2), 471–478.
Szekely, C.A., Thorne, J.E., Zandi, P.P., Ek, M., Messias, E., Breitner, J.C., Goodman, S.N.,
2004. Nonsteroidal anti-inflammatory drugs for the prevention of Alzheimer's
disease: a systematic review. Neuroepidemiology 23 (4), 159–169.
Verdelho, A., Madureira, S., Moleiro, C., Ferro, J.M., Santos, C.O., Erkinjuntti, T.,
Pantoni, L., et al., 2010. White matter changes and diabetes predict cognitive
decline in the elderly: the LADIS study. Neurology 75 (2), 160–167.
Vergara, X., Kheifets, L., Greenland, S., Oksuzyan, S., Cho, Y.S., Mezei, G., 2013.
Occupational exposure to extremely low-frequency magnetic fields and
neurodegenerative disease: a meta-analysis. J. Occup. Environ. Med. 55, 135–146.
Wang, Y., Bi, L., Wang, H., Li, Y., Di, Q., Xu, W., et al., 2012. NEDD9 rs760678
polymorphism and the risk of Alzheimer's disease: a meta-analysis. Neurosci.
Lett. 527, 121–125.
Wang, X.F., Cao, Y.W., Feng, Z.Z., Fu, D., Ma, Y.S., Zhang, F., et al., 2013a. Quantitative
assessment of the effect of ABCA1 gene polymorphism on the risk of
Alzheimer's disease. Mol. Biol. Rep. 40, 779–785.
Wang, Y., Xu, S., Liu, Z., Lai, C., Xie, Z., Zhao, C., et al., 2013b. Meta-analysis on the
association between the TF gene rs1049296 and AD. Can. J. Neurol. Sci. 40 (5),
691–697.
Wang, C., Yu, J.T., Wang, H.F., Jiang, T., Tan, C.C., Meng, X.F., et al., 2014. Meta-analysis
of peripheral blood apolipoprotein E levels in Alzheimer's disease. PLoS ONE
[Electron. Resour.] 9, e89041.
187
Wang, J., Tan, L., Wang, H.F., Tan, C.C., Meng, X.F., Wang, C., et al., 2015a. Anti-
inflammatory drugs and risk of Alzheimer's disease: an updated systematic
review and meta-analysis. J. Alzheimer's Dis. 44, 385–396.
Wang, X.B., Cui, N.H., Gao, J.J., Qiu, X.P., Yang, N., Zheng, F., 2015b. Angiotensin-
converting enzyme gene polymorphisms and risk for sporadic Alzheimer's
disease: a meta-analysis. J. Neural Transm. 122, 211–224.
Wang, T., 2014. Meta-analysis of PvuII: XbaI variants in ESR1 gene and the risk of
Alzheimer's disease: the regional European difference. Neurosci. Lett. 574, 41–
46.
Wang, T., 2015. TNF-alpha G308A polymorphism and the susceptibility to
Alzheimer's disease: an updated meta-analysis. Arch. Med. Res. 46, 24–30 (e1).
Weih, M., Wiltfang, J., Kornhuber, J., 2007. Non-pharmacologic prevention of
Alzheimer's disease: nutritional and life-style risk factors. J. Neural Transm. 114
(9), 1187–1197.
Weih, M., Degirmenci, Ü., Kreil, S., Kornhuber, J., 2010. Physical activity and
Alzheimer’s disease: a meta-analysis of cohort studies. GeroPsych 23 (1), 17–20.
Weyerer, S., Schaufele, M., Wiese, B., Maier, W., Tebarth, F., van den Bussche, H.,
Pentzek, M., et al., 2011. Current alcohol consumption and its relationship to
incident dementia: results from a 3-year follow-up study among primary care
attenders aged 75 years and older. Age Ageing 40 (4), 456–463.
Williams, J.W., Plassman, B.L., Burke, J., Benjamin, S., 2010. Preventing Alzheimer’s
Disease and Cognitive Decline. Evidence Report/Technology Assessment N0.
193. (Prepared by the Duke Evidence-based Practice Center Under Contract No.
HHSA 290–2007-10066-I.) AHRQ Publication No. 10-E005. Agency for
Healthcare Research and Quality, Rockville, MD.
Wong, W.B., Lin, V.W., Boudreau, D., Devine, E.B., 2013. Statins in the prevention of
dementia and Alzheimer's disease: a meta-analysis of observational studies and
an assessment of confounding. Pharmacoepidemiol. Drug Saf. 22, 345–358.
Woods, B., Aguirre, E., Spector, A.E., Orrell, M., 2012. Cognitive stimulation to
improve cognitive functioning in people with dementia. Cochrane Database
Syst. Rev. 2, CD005562.
Wu, W., Jiang, H., Wang, M., Zhang, D., 2013. Meta-analysis of the association
between urokinase-plasminogen activator gene rs2227564 polymorphism and
Alzheimer's disease. Am. J. Alzheimer's Dis. Other Dementias 28, 517–523.
Wu, S., Ding, Y., Wu, F., Li, R., Hou, J., Mao, P., 2015. Omega-3 fatty acids intake and
risks of dementia and Alzheimer's disease: a meta-analysis. Neurosci. Biobehav.
Rev. 48, 1–9.
Xin, X.Y., Ding, J.Q., Chen, S.D., 2010. Apolipoprotein E promoter polymorphisms and
risk of Alzheimer's disease: evidence from meta-analysis. J. Alzheimer's Dis. 19
(4), 1283–1294.
Xu, W.L., Atti, A.R., Gatz, M., Pedersen, N.L., Johansson, B., Fratiglioni, L., 2011. Midlife
overweight and obesity increase late-life dementia risk: a population-based
twin study. Neurology 76 (18), 1568–1574.
Yaffe, K., Sawaya, G., Lieberburg, I., Grady, D., 1998. Estrogen therapy in
postmenopausal women: effects on cognitive function and dementia. JAMA 279
(9), 688–695.
Yang, Y.H., Roe, C.M., Morris, J.C., 2011. Relationship between late-life hypertension,
blood pressure, and Alzheimer's disease. Am. J. Alzheimer's Dis. Other
Dementias 26 (6), 457–462.
Yang, Z.D., Yu, J., Zhang, Q., 2013. Effects of raloxifene on cognition, mental health,
sleep and sexual function in menopausal women: a systematic review of
randomized controlled trials. Maturitas 75, 341–348.
Zhang, T., Jia, Y., 2014. Meta-analysis of Ubiquilin1 gene polymorphism and
Alzheimer's disease risk. Med. Sci. Monit. 20, 2250–2255.
Zhang, M.Y., Miao, L., Li, Y.S., Hu, G.Y., 2010. Meta-analysis of the
methylenetetrahydrofolate reductase C677T polymorphism and susceptibility
to Alzheimer's disease. Neurosci. Res. 68 (2), 142–150.
Zhang, Y., Zhang, J., Tian, C., Xiao, Y., Li, X., He, C., Huang, J., Fan, H., 2011. The
1082G/
A polymorphism in IL-10 gene is associated with risk of Alzheimer's disease: a
meta-analysis. J. Neurol. Sci. 303 (1-2), 133–138.
Zhong, G., Wang, Y., Zhang, Y., Guo, J.J., Zhao, Y., 2015. Smoking is associated with an
increased risk of dementia: a meta-analysis of prospective cohort studies with
investigation of potential effect modifiers. PLoS ONE [Electron. Resour.] 10,
e0118333.
Zhou, B., Teramukai, S., Fukushima, M., 2007. Prevention and treatment of dementia
or Alzheimer's disease by statins: a meta-analysis. Dementia Geriatric Cognit.
Disord. 23 (3), 194–201.
Zhou, R., Deng, J., Zhang, M., Zhou, H.D., Wang, Y.J., 2011. Association between bone
mineral density and the risk of Alzheimer's disease. J. Alzheimer's Dis. 24 (1),
101–108.
Zhou, Q., Zhao, F., Lv, Z.P., Zheng, C.G., Zheng, W.D., Sun, L., et al., 2014. Association
between APOC1 polymorphism and Alzheimer's disease: a case-control study
and meta-analysis. PLoS ONE [Electron. Resour.] 9, e87017.
Zuo, C., Zuo, Z., 2010. Spine Surgery under general anesthesia may not increase the
risk of Alzheimer's disease. Dementia Geriatric Cognit. Disord. 29 (3), 233–239.
de Craen, A.J., Gussekloo, J., Vrijsen, B., Westendorp, R.G., 2005. Meta-analysis of
nonsteroidal antiinflammatory drug use and risk of dementia. Am. J. Epidemiol.
161 (2), 114–120.
de Jong, F.J., Schrijvers, E.M., Ikram, M.K., Koudstaal, P.J., de Jong, P.T., Hofman, A.,
Vingerling, J.R., Breteler, M.M., 2011. Retinal vascular caliber and risk of
dementia: the Rotterdam study. Neurology 76 (9), 816–821.
van Duijn, C.M., Clayton, D., Chandra, V., Fratiglioni, L., Graves, A.B., Heyman, A.,
Jorm, A.F., et al., 1991. Familial aggregation of Alzheimer's disease and related
disorders: a collaborative re-analysis of case-control studies. Int. J. Epidemiol.
20 (Suppl. 2), S13–S20.